1 // SPDX-License-Identifier: GPL-2.0 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 * The IP to API glue. 8 * 9 * Authors: see ip.c 10 * 11 * Fixes: 12 * Many : Split from ip.c , see ip.c for history. 13 * Martin Mares : TOS setting fixed. 14 * Alan Cox : Fixed a couple of oopses in Martin's 15 * TOS tweaks. 16 * Mike McLagan : Routing by source 17 */ 18 19 #include <linux/module.h> 20 #include <linux/types.h> 21 #include <linux/mm.h> 22 #include <linux/skbuff.h> 23 #include <linux/ip.h> 24 #include <linux/icmp.h> 25 #include <linux/inetdevice.h> 26 #include <linux/netdevice.h> 27 #include <linux/slab.h> 28 #include <net/sock.h> 29 #include <net/ip.h> 30 #include <net/icmp.h> 31 #include <net/tcp_states.h> 32 #include <linux/udp.h> 33 #include <linux/igmp.h> 34 #include <linux/netfilter.h> 35 #include <linux/route.h> 36 #include <linux/mroute.h> 37 #include <net/inet_ecn.h> 38 #include <net/route.h> 39 #include <net/xfrm.h> 40 #include <net/compat.h> 41 #include <net/checksum.h> 42 #if IS_ENABLED(CONFIG_IPV6) 43 #include <net/transp_v6.h> 44 #endif 45 #include <net/ip_fib.h> 46 47 #include <linux/errqueue.h> 48 #include <linux/uaccess.h> 49 50 /* 51 * SOL_IP control messages. 52 */ 53 54 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) 55 { 56 struct in_pktinfo info = *PKTINFO_SKB_CB(skb); 57 58 info.ipi_addr.s_addr = ip_hdr(skb)->daddr; 59 60 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 61 } 62 63 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb) 64 { 65 int ttl = ip_hdr(skb)->ttl; 66 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl); 67 } 68 69 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb) 70 { 71 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos); 72 } 73 74 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb) 75 { 76 if (IPCB(skb)->opt.optlen == 0) 77 return; 78 79 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen, 80 ip_hdr(skb) + 1); 81 } 82 83 84 static void ip_cmsg_recv_retopts(struct net *net, struct msghdr *msg, 85 struct sk_buff *skb) 86 { 87 unsigned char optbuf[sizeof(struct ip_options) + 40]; 88 struct ip_options *opt = (struct ip_options *)optbuf; 89 90 if (IPCB(skb)->opt.optlen == 0) 91 return; 92 93 if (ip_options_echo(net, opt, skb)) { 94 msg->msg_flags |= MSG_CTRUNC; 95 return; 96 } 97 ip_options_undo(opt); 98 99 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data); 100 } 101 102 static void ip_cmsg_recv_fragsize(struct msghdr *msg, struct sk_buff *skb) 103 { 104 int val; 105 106 if (IPCB(skb)->frag_max_size == 0) 107 return; 108 109 val = IPCB(skb)->frag_max_size; 110 put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val); 111 } 112 113 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb, 114 int tlen, int offset) 115 { 116 __wsum csum = skb->csum; 117 118 if (skb->ip_summed != CHECKSUM_COMPLETE) 119 return; 120 121 if (offset != 0) { 122 int tend_off = skb_transport_offset(skb) + tlen; 123 csum = csum_sub(csum, skb_checksum(skb, tend_off, offset, 0)); 124 } 125 126 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum); 127 } 128 129 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb) 130 { 131 char *secdata; 132 u32 seclen, secid; 133 int err; 134 135 err = security_socket_getpeersec_dgram(NULL, skb, &secid); 136 if (err) 137 return; 138 139 err = security_secid_to_secctx(secid, &secdata, &seclen); 140 if (err) 141 return; 142 143 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata); 144 security_release_secctx(secdata, seclen); 145 } 146 147 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb) 148 { 149 struct sockaddr_in sin; 150 const struct iphdr *iph = ip_hdr(skb); 151 __be16 *ports = (__be16 *)skb_transport_header(skb); 152 153 if (skb_transport_offset(skb) + 4 > (int)skb->len) 154 return; 155 156 /* All current transport protocols have the port numbers in the 157 * first four bytes of the transport header and this function is 158 * written with this assumption in mind. 159 */ 160 161 sin.sin_family = AF_INET; 162 sin.sin_addr.s_addr = iph->daddr; 163 sin.sin_port = ports[1]; 164 memset(sin.sin_zero, 0, sizeof(sin.sin_zero)); 165 166 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin); 167 } 168 169 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk, 170 struct sk_buff *skb, int tlen, int offset) 171 { 172 struct inet_sock *inet = inet_sk(sk); 173 unsigned int flags = inet->cmsg_flags; 174 175 /* Ordered by supposed usage frequency */ 176 if (flags & IP_CMSG_PKTINFO) { 177 ip_cmsg_recv_pktinfo(msg, skb); 178 179 flags &= ~IP_CMSG_PKTINFO; 180 if (!flags) 181 return; 182 } 183 184 if (flags & IP_CMSG_TTL) { 185 ip_cmsg_recv_ttl(msg, skb); 186 187 flags &= ~IP_CMSG_TTL; 188 if (!flags) 189 return; 190 } 191 192 if (flags & IP_CMSG_TOS) { 193 ip_cmsg_recv_tos(msg, skb); 194 195 flags &= ~IP_CMSG_TOS; 196 if (!flags) 197 return; 198 } 199 200 if (flags & IP_CMSG_RECVOPTS) { 201 ip_cmsg_recv_opts(msg, skb); 202 203 flags &= ~IP_CMSG_RECVOPTS; 204 if (!flags) 205 return; 206 } 207 208 if (flags & IP_CMSG_RETOPTS) { 209 ip_cmsg_recv_retopts(sock_net(sk), msg, skb); 210 211 flags &= ~IP_CMSG_RETOPTS; 212 if (!flags) 213 return; 214 } 215 216 if (flags & IP_CMSG_PASSSEC) { 217 ip_cmsg_recv_security(msg, skb); 218 219 flags &= ~IP_CMSG_PASSSEC; 220 if (!flags) 221 return; 222 } 223 224 if (flags & IP_CMSG_ORIGDSTADDR) { 225 ip_cmsg_recv_dstaddr(msg, skb); 226 227 flags &= ~IP_CMSG_ORIGDSTADDR; 228 if (!flags) 229 return; 230 } 231 232 if (flags & IP_CMSG_CHECKSUM) 233 ip_cmsg_recv_checksum(msg, skb, tlen, offset); 234 235 if (flags & IP_CMSG_RECVFRAGSIZE) 236 ip_cmsg_recv_fragsize(msg, skb); 237 } 238 EXPORT_SYMBOL(ip_cmsg_recv_offset); 239 240 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc, 241 bool allow_ipv6) 242 { 243 int err, val; 244 struct cmsghdr *cmsg; 245 struct net *net = sock_net(sk); 246 247 for_each_cmsghdr(cmsg, msg) { 248 if (!CMSG_OK(msg, cmsg)) 249 return -EINVAL; 250 #if IS_ENABLED(CONFIG_IPV6) 251 if (allow_ipv6 && 252 cmsg->cmsg_level == SOL_IPV6 && 253 cmsg->cmsg_type == IPV6_PKTINFO) { 254 struct in6_pktinfo *src_info; 255 256 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info))) 257 return -EINVAL; 258 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg); 259 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr)) 260 return -EINVAL; 261 ipc->oif = src_info->ipi6_ifindex; 262 ipc->addr = src_info->ipi6_addr.s6_addr32[3]; 263 continue; 264 } 265 #endif 266 if (cmsg->cmsg_level == SOL_SOCKET) { 267 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc); 268 if (err) 269 return err; 270 continue; 271 } 272 273 if (cmsg->cmsg_level != SOL_IP) 274 continue; 275 switch (cmsg->cmsg_type) { 276 case IP_RETOPTS: 277 err = cmsg->cmsg_len - sizeof(struct cmsghdr); 278 279 /* Our caller is responsible for freeing ipc->opt */ 280 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), 281 err < 40 ? err : 40); 282 if (err) 283 return err; 284 break; 285 case IP_PKTINFO: 286 { 287 struct in_pktinfo *info; 288 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) 289 return -EINVAL; 290 info = (struct in_pktinfo *)CMSG_DATA(cmsg); 291 ipc->oif = info->ipi_ifindex; 292 ipc->addr = info->ipi_spec_dst.s_addr; 293 break; 294 } 295 case IP_TTL: 296 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 297 return -EINVAL; 298 val = *(int *)CMSG_DATA(cmsg); 299 if (val < 1 || val > 255) 300 return -EINVAL; 301 ipc->ttl = val; 302 break; 303 case IP_TOS: 304 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int))) 305 val = *(int *)CMSG_DATA(cmsg); 306 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8))) 307 val = *(u8 *)CMSG_DATA(cmsg); 308 else 309 return -EINVAL; 310 if (val < 0 || val > 255) 311 return -EINVAL; 312 ipc->tos = val; 313 ipc->priority = rt_tos2priority(ipc->tos); 314 break; 315 316 default: 317 return -EINVAL; 318 } 319 } 320 return 0; 321 } 322 323 324 /* Special input handler for packets caught by router alert option. 325 They are selected only by protocol field, and then processed likely 326 local ones; but only if someone wants them! Otherwise, router 327 not running rsvpd will kill RSVP. 328 329 It is user level problem, what it will make with them. 330 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)), 331 but receiver should be enough clever f.e. to forward mtrace requests, 332 sent to multicast group to reach destination designated router. 333 */ 334 struct ip_ra_chain __rcu *ip_ra_chain; 335 336 337 static void ip_ra_destroy_rcu(struct rcu_head *head) 338 { 339 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu); 340 341 sock_put(ra->saved_sk); 342 kfree(ra); 343 } 344 345 int ip_ra_control(struct sock *sk, unsigned char on, 346 void (*destructor)(struct sock *)) 347 { 348 struct ip_ra_chain *ra, *new_ra; 349 struct ip_ra_chain __rcu **rap; 350 351 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 352 return -EINVAL; 353 354 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 355 356 for (rap = &ip_ra_chain; 357 (ra = rtnl_dereference(*rap)) != NULL; 358 rap = &ra->next) { 359 if (ra->sk == sk) { 360 if (on) { 361 kfree(new_ra); 362 return -EADDRINUSE; 363 } 364 /* dont let ip_call_ra_chain() use sk again */ 365 ra->sk = NULL; 366 RCU_INIT_POINTER(*rap, ra->next); 367 368 if (ra->destructor) 369 ra->destructor(sk); 370 /* 371 * Delay sock_put(sk) and kfree(ra) after one rcu grace 372 * period. This guarantee ip_call_ra_chain() dont need 373 * to mess with socket refcounts. 374 */ 375 ra->saved_sk = sk; 376 call_rcu(&ra->rcu, ip_ra_destroy_rcu); 377 return 0; 378 } 379 } 380 if (!new_ra) 381 return -ENOBUFS; 382 new_ra->sk = sk; 383 new_ra->destructor = destructor; 384 385 RCU_INIT_POINTER(new_ra->next, ra); 386 rcu_assign_pointer(*rap, new_ra); 387 sock_hold(sk); 388 389 return 0; 390 } 391 392 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 393 __be16 port, u32 info, u8 *payload) 394 { 395 struct sock_exterr_skb *serr; 396 397 skb = skb_clone(skb, GFP_ATOMIC); 398 if (!skb) 399 return; 400 401 serr = SKB_EXT_ERR(skb); 402 serr->ee.ee_errno = err; 403 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 404 serr->ee.ee_type = icmp_hdr(skb)->type; 405 serr->ee.ee_code = icmp_hdr(skb)->code; 406 serr->ee.ee_pad = 0; 407 serr->ee.ee_info = info; 408 serr->ee.ee_data = 0; 409 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 410 skb_network_header(skb); 411 serr->port = port; 412 413 if (skb_pull(skb, payload - skb->data)) { 414 skb_reset_transport_header(skb); 415 if (sock_queue_err_skb(sk, skb) == 0) 416 return; 417 } 418 kfree_skb(skb); 419 } 420 421 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 422 { 423 struct inet_sock *inet = inet_sk(sk); 424 struct sock_exterr_skb *serr; 425 struct iphdr *iph; 426 struct sk_buff *skb; 427 428 if (!inet->recverr) 429 return; 430 431 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 432 if (!skb) 433 return; 434 435 skb_put(skb, sizeof(struct iphdr)); 436 skb_reset_network_header(skb); 437 iph = ip_hdr(skb); 438 iph->daddr = daddr; 439 440 serr = SKB_EXT_ERR(skb); 441 serr->ee.ee_errno = err; 442 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 443 serr->ee.ee_type = 0; 444 serr->ee.ee_code = 0; 445 serr->ee.ee_pad = 0; 446 serr->ee.ee_info = info; 447 serr->ee.ee_data = 0; 448 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 449 serr->port = port; 450 451 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 452 skb_reset_transport_header(skb); 453 454 if (sock_queue_err_skb(sk, skb)) 455 kfree_skb(skb); 456 } 457 458 /* For some errors we have valid addr_offset even with zero payload and 459 * zero port. Also, addr_offset should be supported if port is set. 460 */ 461 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr) 462 { 463 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP || 464 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port; 465 } 466 467 /* IPv4 supports cmsg on all imcp errors and some timestamps 468 * 469 * Timestamp code paths do not initialize the fields expected by cmsg: 470 * the PKTINFO fields in skb->cb[]. Fill those in here. 471 */ 472 static bool ipv4_datagram_support_cmsg(const struct sock *sk, 473 struct sk_buff *skb, 474 int ee_origin) 475 { 476 struct in_pktinfo *info; 477 478 if (ee_origin == SO_EE_ORIGIN_ICMP) 479 return true; 480 481 if (ee_origin == SO_EE_ORIGIN_LOCAL) 482 return false; 483 484 /* Support IP_PKTINFO on tstamp packets if requested, to correlate 485 * timestamp with egress dev. Not possible for packets without iif 486 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY). 487 */ 488 info = PKTINFO_SKB_CB(skb); 489 if (!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG) || 490 !info->ipi_ifindex) 491 return false; 492 493 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr; 494 return true; 495 } 496 497 /* 498 * Handle MSG_ERRQUEUE 499 */ 500 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 501 { 502 struct sock_exterr_skb *serr; 503 struct sk_buff *skb; 504 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 505 struct { 506 struct sock_extended_err ee; 507 struct sockaddr_in offender; 508 } errhdr; 509 int err; 510 int copied; 511 512 WARN_ON_ONCE(sk->sk_family == AF_INET6); 513 514 err = -EAGAIN; 515 skb = sock_dequeue_err_skb(sk); 516 if (!skb) 517 goto out; 518 519 copied = skb->len; 520 if (copied > len) { 521 msg->msg_flags |= MSG_TRUNC; 522 copied = len; 523 } 524 err = skb_copy_datagram_msg(skb, 0, msg, copied); 525 if (unlikely(err)) { 526 kfree_skb(skb); 527 return err; 528 } 529 sock_recv_timestamp(msg, sk, skb); 530 531 serr = SKB_EXT_ERR(skb); 532 533 if (sin && ipv4_datagram_support_addr(serr)) { 534 sin->sin_family = AF_INET; 535 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 536 serr->addr_offset); 537 sin->sin_port = serr->port; 538 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 539 *addr_len = sizeof(*sin); 540 } 541 542 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 543 sin = &errhdr.offender; 544 memset(sin, 0, sizeof(*sin)); 545 546 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) { 547 sin->sin_family = AF_INET; 548 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 549 if (inet_sk(sk)->cmsg_flags) 550 ip_cmsg_recv(msg, skb); 551 } 552 553 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 554 555 /* Now we could try to dump offended packet options */ 556 557 msg->msg_flags |= MSG_ERRQUEUE; 558 err = copied; 559 560 consume_skb(skb); 561 out: 562 return err; 563 } 564 565 566 /* 567 * Socket option code for IP. This is the end of the line after any 568 * TCP,UDP etc options on an IP socket. 569 */ 570 static bool setsockopt_needs_rtnl(int optname) 571 { 572 switch (optname) { 573 case IP_ADD_MEMBERSHIP: 574 case IP_ADD_SOURCE_MEMBERSHIP: 575 case IP_BLOCK_SOURCE: 576 case IP_DROP_MEMBERSHIP: 577 case IP_DROP_SOURCE_MEMBERSHIP: 578 case IP_MSFILTER: 579 case IP_UNBLOCK_SOURCE: 580 case MCAST_BLOCK_SOURCE: 581 case MCAST_MSFILTER: 582 case MCAST_JOIN_GROUP: 583 case MCAST_JOIN_SOURCE_GROUP: 584 case MCAST_LEAVE_GROUP: 585 case MCAST_LEAVE_SOURCE_GROUP: 586 case MCAST_UNBLOCK_SOURCE: 587 case IP_ROUTER_ALERT: 588 return true; 589 } 590 return false; 591 } 592 593 static int do_ip_setsockopt(struct sock *sk, int level, 594 int optname, char __user *optval, unsigned int optlen) 595 { 596 struct inet_sock *inet = inet_sk(sk); 597 struct net *net = sock_net(sk); 598 int val = 0, err; 599 bool needs_rtnl = setsockopt_needs_rtnl(optname); 600 601 switch (optname) { 602 case IP_PKTINFO: 603 case IP_RECVTTL: 604 case IP_RECVOPTS: 605 case IP_RECVTOS: 606 case IP_RETOPTS: 607 case IP_TOS: 608 case IP_TTL: 609 case IP_HDRINCL: 610 case IP_MTU_DISCOVER: 611 case IP_RECVERR: 612 case IP_ROUTER_ALERT: 613 case IP_FREEBIND: 614 case IP_PASSSEC: 615 case IP_TRANSPARENT: 616 case IP_MINTTL: 617 case IP_NODEFRAG: 618 case IP_BIND_ADDRESS_NO_PORT: 619 case IP_UNICAST_IF: 620 case IP_MULTICAST_TTL: 621 case IP_MULTICAST_ALL: 622 case IP_MULTICAST_LOOP: 623 case IP_RECVORIGDSTADDR: 624 case IP_CHECKSUM: 625 case IP_RECVFRAGSIZE: 626 if (optlen >= sizeof(int)) { 627 if (get_user(val, (int __user *) optval)) 628 return -EFAULT; 629 } else if (optlen >= sizeof(char)) { 630 unsigned char ucval; 631 632 if (get_user(ucval, (unsigned char __user *) optval)) 633 return -EFAULT; 634 val = (int) ucval; 635 } 636 } 637 638 /* If optlen==0, it is equivalent to val == 0 */ 639 640 if (ip_mroute_opt(optname)) 641 return ip_mroute_setsockopt(sk, optname, optval, optlen); 642 643 err = 0; 644 if (needs_rtnl) 645 rtnl_lock(); 646 lock_sock(sk); 647 648 switch (optname) { 649 case IP_OPTIONS: 650 { 651 struct ip_options_rcu *old, *opt = NULL; 652 653 if (optlen > 40) 654 goto e_inval; 655 err = ip_options_get_from_user(sock_net(sk), &opt, 656 optval, optlen); 657 if (err) 658 break; 659 old = rcu_dereference_protected(inet->inet_opt, 660 lockdep_sock_is_held(sk)); 661 if (inet->is_icsk) { 662 struct inet_connection_sock *icsk = inet_csk(sk); 663 #if IS_ENABLED(CONFIG_IPV6) 664 if (sk->sk_family == PF_INET || 665 (!((1 << sk->sk_state) & 666 (TCPF_LISTEN | TCPF_CLOSE)) && 667 inet->inet_daddr != LOOPBACK4_IPV6)) { 668 #endif 669 if (old) 670 icsk->icsk_ext_hdr_len -= old->opt.optlen; 671 if (opt) 672 icsk->icsk_ext_hdr_len += opt->opt.optlen; 673 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 674 #if IS_ENABLED(CONFIG_IPV6) 675 } 676 #endif 677 } 678 rcu_assign_pointer(inet->inet_opt, opt); 679 if (old) 680 kfree_rcu(old, rcu); 681 break; 682 } 683 case IP_PKTINFO: 684 if (val) 685 inet->cmsg_flags |= IP_CMSG_PKTINFO; 686 else 687 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 688 break; 689 case IP_RECVTTL: 690 if (val) 691 inet->cmsg_flags |= IP_CMSG_TTL; 692 else 693 inet->cmsg_flags &= ~IP_CMSG_TTL; 694 break; 695 case IP_RECVTOS: 696 if (val) 697 inet->cmsg_flags |= IP_CMSG_TOS; 698 else 699 inet->cmsg_flags &= ~IP_CMSG_TOS; 700 break; 701 case IP_RECVOPTS: 702 if (val) 703 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 704 else 705 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 706 break; 707 case IP_RETOPTS: 708 if (val) 709 inet->cmsg_flags |= IP_CMSG_RETOPTS; 710 else 711 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 712 break; 713 case IP_PASSSEC: 714 if (val) 715 inet->cmsg_flags |= IP_CMSG_PASSSEC; 716 else 717 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 718 break; 719 case IP_RECVORIGDSTADDR: 720 if (val) 721 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 722 else 723 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 724 break; 725 case IP_CHECKSUM: 726 if (val) { 727 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) { 728 inet_inc_convert_csum(sk); 729 inet->cmsg_flags |= IP_CMSG_CHECKSUM; 730 } 731 } else { 732 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) { 733 inet_dec_convert_csum(sk); 734 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM; 735 } 736 } 737 break; 738 case IP_RECVFRAGSIZE: 739 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM) 740 goto e_inval; 741 if (val) 742 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE; 743 else 744 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE; 745 break; 746 case IP_TOS: /* This sets both TOS and Precedence */ 747 if (sk->sk_type == SOCK_STREAM) { 748 val &= ~INET_ECN_MASK; 749 val |= inet->tos & INET_ECN_MASK; 750 } 751 if (inet->tos != val) { 752 inet->tos = val; 753 sk->sk_priority = rt_tos2priority(val); 754 sk_dst_reset(sk); 755 } 756 break; 757 case IP_TTL: 758 if (optlen < 1) 759 goto e_inval; 760 if (val != -1 && (val < 1 || val > 255)) 761 goto e_inval; 762 inet->uc_ttl = val; 763 break; 764 case IP_HDRINCL: 765 if (sk->sk_type != SOCK_RAW) { 766 err = -ENOPROTOOPT; 767 break; 768 } 769 inet->hdrincl = val ? 1 : 0; 770 break; 771 case IP_NODEFRAG: 772 if (sk->sk_type != SOCK_RAW) { 773 err = -ENOPROTOOPT; 774 break; 775 } 776 inet->nodefrag = val ? 1 : 0; 777 break; 778 case IP_BIND_ADDRESS_NO_PORT: 779 inet->bind_address_no_port = val ? 1 : 0; 780 break; 781 case IP_MTU_DISCOVER: 782 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT) 783 goto e_inval; 784 inet->pmtudisc = val; 785 break; 786 case IP_RECVERR: 787 inet->recverr = !!val; 788 if (!val) 789 skb_queue_purge(&sk->sk_error_queue); 790 break; 791 case IP_MULTICAST_TTL: 792 if (sk->sk_type == SOCK_STREAM) 793 goto e_inval; 794 if (optlen < 1) 795 goto e_inval; 796 if (val == -1) 797 val = 1; 798 if (val < 0 || val > 255) 799 goto e_inval; 800 inet->mc_ttl = val; 801 break; 802 case IP_MULTICAST_LOOP: 803 if (optlen < 1) 804 goto e_inval; 805 inet->mc_loop = !!val; 806 break; 807 case IP_UNICAST_IF: 808 { 809 struct net_device *dev = NULL; 810 int ifindex; 811 int midx; 812 813 if (optlen != sizeof(int)) 814 goto e_inval; 815 816 ifindex = (__force int)ntohl((__force __be32)val); 817 if (ifindex == 0) { 818 inet->uc_index = 0; 819 err = 0; 820 break; 821 } 822 823 dev = dev_get_by_index(sock_net(sk), ifindex); 824 err = -EADDRNOTAVAIL; 825 if (!dev) 826 break; 827 828 midx = l3mdev_master_ifindex(dev); 829 dev_put(dev); 830 831 err = -EINVAL; 832 if (sk->sk_bound_dev_if && 833 (!midx || midx != sk->sk_bound_dev_if)) 834 break; 835 836 inet->uc_index = ifindex; 837 err = 0; 838 break; 839 } 840 case IP_MULTICAST_IF: 841 { 842 struct ip_mreqn mreq; 843 struct net_device *dev = NULL; 844 int midx; 845 846 if (sk->sk_type == SOCK_STREAM) 847 goto e_inval; 848 /* 849 * Check the arguments are allowable 850 */ 851 852 if (optlen < sizeof(struct in_addr)) 853 goto e_inval; 854 855 err = -EFAULT; 856 if (optlen >= sizeof(struct ip_mreqn)) { 857 if (copy_from_user(&mreq, optval, sizeof(mreq))) 858 break; 859 } else { 860 memset(&mreq, 0, sizeof(mreq)); 861 if (optlen >= sizeof(struct ip_mreq)) { 862 if (copy_from_user(&mreq, optval, 863 sizeof(struct ip_mreq))) 864 break; 865 } else if (optlen >= sizeof(struct in_addr)) { 866 if (copy_from_user(&mreq.imr_address, optval, 867 sizeof(struct in_addr))) 868 break; 869 } 870 } 871 872 if (!mreq.imr_ifindex) { 873 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 874 inet->mc_index = 0; 875 inet->mc_addr = 0; 876 err = 0; 877 break; 878 } 879 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 880 if (dev) 881 mreq.imr_ifindex = dev->ifindex; 882 } else 883 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 884 885 886 err = -EADDRNOTAVAIL; 887 if (!dev) 888 break; 889 890 midx = l3mdev_master_ifindex(dev); 891 892 dev_put(dev); 893 894 err = -EINVAL; 895 if (sk->sk_bound_dev_if && 896 mreq.imr_ifindex != sk->sk_bound_dev_if && 897 (!midx || midx != sk->sk_bound_dev_if)) 898 break; 899 900 inet->mc_index = mreq.imr_ifindex; 901 inet->mc_addr = mreq.imr_address.s_addr; 902 err = 0; 903 break; 904 } 905 906 case IP_ADD_MEMBERSHIP: 907 case IP_DROP_MEMBERSHIP: 908 { 909 struct ip_mreqn mreq; 910 911 err = -EPROTO; 912 if (inet_sk(sk)->is_icsk) 913 break; 914 915 if (optlen < sizeof(struct ip_mreq)) 916 goto e_inval; 917 err = -EFAULT; 918 if (optlen >= sizeof(struct ip_mreqn)) { 919 if (copy_from_user(&mreq, optval, sizeof(mreq))) 920 break; 921 } else { 922 memset(&mreq, 0, sizeof(mreq)); 923 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 924 break; 925 } 926 927 if (optname == IP_ADD_MEMBERSHIP) 928 err = ip_mc_join_group(sk, &mreq); 929 else 930 err = ip_mc_leave_group(sk, &mreq); 931 break; 932 } 933 case IP_MSFILTER: 934 { 935 struct ip_msfilter *msf; 936 937 if (optlen < IP_MSFILTER_SIZE(0)) 938 goto e_inval; 939 if (optlen > sysctl_optmem_max) { 940 err = -ENOBUFS; 941 break; 942 } 943 msf = memdup_user(optval, optlen); 944 if (IS_ERR(msf)) { 945 err = PTR_ERR(msf); 946 break; 947 } 948 /* numsrc >= (1G-4) overflow in 32 bits */ 949 if (msf->imsf_numsrc >= 0x3ffffffcU || 950 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) { 951 kfree(msf); 952 err = -ENOBUFS; 953 break; 954 } 955 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 956 kfree(msf); 957 err = -EINVAL; 958 break; 959 } 960 err = ip_mc_msfilter(sk, msf, 0); 961 kfree(msf); 962 break; 963 } 964 case IP_BLOCK_SOURCE: 965 case IP_UNBLOCK_SOURCE: 966 case IP_ADD_SOURCE_MEMBERSHIP: 967 case IP_DROP_SOURCE_MEMBERSHIP: 968 { 969 struct ip_mreq_source mreqs; 970 int omode, add; 971 972 if (optlen != sizeof(struct ip_mreq_source)) 973 goto e_inval; 974 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 975 err = -EFAULT; 976 break; 977 } 978 if (optname == IP_BLOCK_SOURCE) { 979 omode = MCAST_EXCLUDE; 980 add = 1; 981 } else if (optname == IP_UNBLOCK_SOURCE) { 982 omode = MCAST_EXCLUDE; 983 add = 0; 984 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 985 struct ip_mreqn mreq; 986 987 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 988 mreq.imr_address.s_addr = mreqs.imr_interface; 989 mreq.imr_ifindex = 0; 990 err = ip_mc_join_group(sk, &mreq); 991 if (err && err != -EADDRINUSE) 992 break; 993 omode = MCAST_INCLUDE; 994 add = 1; 995 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 996 omode = MCAST_INCLUDE; 997 add = 0; 998 } 999 err = ip_mc_source(add, omode, sk, &mreqs, 0); 1000 break; 1001 } 1002 case MCAST_JOIN_GROUP: 1003 case MCAST_LEAVE_GROUP: 1004 { 1005 struct group_req greq; 1006 struct sockaddr_in *psin; 1007 struct ip_mreqn mreq; 1008 1009 if (optlen < sizeof(struct group_req)) 1010 goto e_inval; 1011 err = -EFAULT; 1012 if (copy_from_user(&greq, optval, sizeof(greq))) 1013 break; 1014 psin = (struct sockaddr_in *)&greq.gr_group; 1015 if (psin->sin_family != AF_INET) 1016 goto e_inval; 1017 memset(&mreq, 0, sizeof(mreq)); 1018 mreq.imr_multiaddr = psin->sin_addr; 1019 mreq.imr_ifindex = greq.gr_interface; 1020 1021 if (optname == MCAST_JOIN_GROUP) 1022 err = ip_mc_join_group(sk, &mreq); 1023 else 1024 err = ip_mc_leave_group(sk, &mreq); 1025 break; 1026 } 1027 case MCAST_JOIN_SOURCE_GROUP: 1028 case MCAST_LEAVE_SOURCE_GROUP: 1029 case MCAST_BLOCK_SOURCE: 1030 case MCAST_UNBLOCK_SOURCE: 1031 { 1032 struct group_source_req greqs; 1033 struct ip_mreq_source mreqs; 1034 struct sockaddr_in *psin; 1035 int omode, add; 1036 1037 if (optlen != sizeof(struct group_source_req)) 1038 goto e_inval; 1039 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 1040 err = -EFAULT; 1041 break; 1042 } 1043 if (greqs.gsr_group.ss_family != AF_INET || 1044 greqs.gsr_source.ss_family != AF_INET) { 1045 err = -EADDRNOTAVAIL; 1046 break; 1047 } 1048 psin = (struct sockaddr_in *)&greqs.gsr_group; 1049 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 1050 psin = (struct sockaddr_in *)&greqs.gsr_source; 1051 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 1052 mreqs.imr_interface = 0; /* use index for mc_source */ 1053 1054 if (optname == MCAST_BLOCK_SOURCE) { 1055 omode = MCAST_EXCLUDE; 1056 add = 1; 1057 } else if (optname == MCAST_UNBLOCK_SOURCE) { 1058 omode = MCAST_EXCLUDE; 1059 add = 0; 1060 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 1061 struct ip_mreqn mreq; 1062 1063 psin = (struct sockaddr_in *)&greqs.gsr_group; 1064 mreq.imr_multiaddr = psin->sin_addr; 1065 mreq.imr_address.s_addr = 0; 1066 mreq.imr_ifindex = greqs.gsr_interface; 1067 err = ip_mc_join_group(sk, &mreq); 1068 if (err && err != -EADDRINUSE) 1069 break; 1070 greqs.gsr_interface = mreq.imr_ifindex; 1071 omode = MCAST_INCLUDE; 1072 add = 1; 1073 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 1074 omode = MCAST_INCLUDE; 1075 add = 0; 1076 } 1077 err = ip_mc_source(add, omode, sk, &mreqs, 1078 greqs.gsr_interface); 1079 break; 1080 } 1081 case MCAST_MSFILTER: 1082 { 1083 struct sockaddr_in *psin; 1084 struct ip_msfilter *msf = NULL; 1085 struct group_filter *gsf = NULL; 1086 int msize, i, ifindex; 1087 1088 if (optlen < GROUP_FILTER_SIZE(0)) 1089 goto e_inval; 1090 if (optlen > sysctl_optmem_max) { 1091 err = -ENOBUFS; 1092 break; 1093 } 1094 gsf = memdup_user(optval, optlen); 1095 if (IS_ERR(gsf)) { 1096 err = PTR_ERR(gsf); 1097 break; 1098 } 1099 1100 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 1101 if (gsf->gf_numsrc >= 0x1ffffff || 1102 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) { 1103 err = -ENOBUFS; 1104 goto mc_msf_out; 1105 } 1106 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 1107 err = -EINVAL; 1108 goto mc_msf_out; 1109 } 1110 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 1111 msf = kmalloc(msize, GFP_KERNEL); 1112 if (!msf) { 1113 err = -ENOBUFS; 1114 goto mc_msf_out; 1115 } 1116 ifindex = gsf->gf_interface; 1117 psin = (struct sockaddr_in *)&gsf->gf_group; 1118 if (psin->sin_family != AF_INET) { 1119 err = -EADDRNOTAVAIL; 1120 goto mc_msf_out; 1121 } 1122 msf->imsf_multiaddr = psin->sin_addr.s_addr; 1123 msf->imsf_interface = 0; 1124 msf->imsf_fmode = gsf->gf_fmode; 1125 msf->imsf_numsrc = gsf->gf_numsrc; 1126 err = -EADDRNOTAVAIL; 1127 for (i = 0; i < gsf->gf_numsrc; ++i) { 1128 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 1129 1130 if (psin->sin_family != AF_INET) 1131 goto mc_msf_out; 1132 msf->imsf_slist[i] = psin->sin_addr.s_addr; 1133 } 1134 kfree(gsf); 1135 gsf = NULL; 1136 1137 err = ip_mc_msfilter(sk, msf, ifindex); 1138 mc_msf_out: 1139 kfree(msf); 1140 kfree(gsf); 1141 break; 1142 } 1143 case IP_MULTICAST_ALL: 1144 if (optlen < 1) 1145 goto e_inval; 1146 if (val != 0 && val != 1) 1147 goto e_inval; 1148 inet->mc_all = val; 1149 break; 1150 case IP_ROUTER_ALERT: 1151 err = ip_ra_control(sk, val ? 1 : 0, NULL); 1152 break; 1153 1154 case IP_FREEBIND: 1155 if (optlen < 1) 1156 goto e_inval; 1157 inet->freebind = !!val; 1158 break; 1159 1160 case IP_IPSEC_POLICY: 1161 case IP_XFRM_POLICY: 1162 err = -EPERM; 1163 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 1164 break; 1165 err = xfrm_user_policy(sk, optname, optval, optlen); 1166 break; 1167 1168 case IP_TRANSPARENT: 1169 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) && 1170 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) { 1171 err = -EPERM; 1172 break; 1173 } 1174 if (optlen < 1) 1175 goto e_inval; 1176 inet->transparent = !!val; 1177 break; 1178 1179 case IP_MINTTL: 1180 if (optlen < 1) 1181 goto e_inval; 1182 if (val < 0 || val > 255) 1183 goto e_inval; 1184 inet->min_ttl = val; 1185 break; 1186 1187 default: 1188 err = -ENOPROTOOPT; 1189 break; 1190 } 1191 release_sock(sk); 1192 if (needs_rtnl) 1193 rtnl_unlock(); 1194 return err; 1195 1196 e_inval: 1197 release_sock(sk); 1198 if (needs_rtnl) 1199 rtnl_unlock(); 1200 return -EINVAL; 1201 } 1202 1203 /** 1204 * ipv4_pktinfo_prepare - transfer some info from rtable to skb 1205 * @sk: socket 1206 * @skb: buffer 1207 * 1208 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific 1209 * destination in skb->cb[] before dst drop. 1210 * This way, receiver doesn't make cache line misses to read rtable. 1211 */ 1212 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb) 1213 { 1214 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb); 1215 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) || 1216 ipv6_sk_rxinfo(sk); 1217 1218 if (prepare && skb_rtable(skb)) { 1219 /* skb->cb is overloaded: prior to this point it is IP{6}CB 1220 * which has interface index (iif) as the first member of the 1221 * underlying inet{6}_skb_parm struct. This code then overlays 1222 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first 1223 * element so the iif is picked up from the prior IPCB. If iif 1224 * is the loopback interface, then return the sending interface 1225 * (e.g., process binds socket to eth0 for Tx which is 1226 * redirected to loopback in the rtable/dst). 1227 */ 1228 struct rtable *rt = skb_rtable(skb); 1229 bool l3slave = ipv4_l3mdev_skb(IPCB(skb)->flags); 1230 1231 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX) 1232 pktinfo->ipi_ifindex = inet_iif(skb); 1233 else if (l3slave && rt && rt->rt_iif) 1234 pktinfo->ipi_ifindex = rt->rt_iif; 1235 1236 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb); 1237 } else { 1238 pktinfo->ipi_ifindex = 0; 1239 pktinfo->ipi_spec_dst.s_addr = 0; 1240 } 1241 skb_dst_drop(skb); 1242 } 1243 1244 int ip_setsockopt(struct sock *sk, int level, 1245 int optname, char __user *optval, unsigned int optlen) 1246 { 1247 int err; 1248 1249 if (level != SOL_IP) 1250 return -ENOPROTOOPT; 1251 1252 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1253 #ifdef CONFIG_NETFILTER 1254 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1255 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1256 optname != IP_IPSEC_POLICY && 1257 optname != IP_XFRM_POLICY && 1258 !ip_mroute_opt(optname)) 1259 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1260 #endif 1261 return err; 1262 } 1263 EXPORT_SYMBOL(ip_setsockopt); 1264 1265 #ifdef CONFIG_COMPAT 1266 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1267 char __user *optval, unsigned int optlen) 1268 { 1269 int err; 1270 1271 if (level != SOL_IP) 1272 return -ENOPROTOOPT; 1273 1274 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1275 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1276 ip_setsockopt); 1277 1278 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1279 #ifdef CONFIG_NETFILTER 1280 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1281 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1282 optname != IP_IPSEC_POLICY && 1283 optname != IP_XFRM_POLICY && 1284 !ip_mroute_opt(optname)) 1285 err = compat_nf_setsockopt(sk, PF_INET, optname, optval, 1286 optlen); 1287 #endif 1288 return err; 1289 } 1290 EXPORT_SYMBOL(compat_ip_setsockopt); 1291 #endif 1292 1293 /* 1294 * Get the options. Note for future reference. The GET of IP options gets 1295 * the _received_ ones. The set sets the _sent_ ones. 1296 */ 1297 1298 static bool getsockopt_needs_rtnl(int optname) 1299 { 1300 switch (optname) { 1301 case IP_MSFILTER: 1302 case MCAST_MSFILTER: 1303 return true; 1304 } 1305 return false; 1306 } 1307 1308 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1309 char __user *optval, int __user *optlen, unsigned int flags) 1310 { 1311 struct inet_sock *inet = inet_sk(sk); 1312 bool needs_rtnl = getsockopt_needs_rtnl(optname); 1313 int val, err = 0; 1314 int len; 1315 1316 if (level != SOL_IP) 1317 return -EOPNOTSUPP; 1318 1319 if (ip_mroute_opt(optname)) 1320 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1321 1322 if (get_user(len, optlen)) 1323 return -EFAULT; 1324 if (len < 0) 1325 return -EINVAL; 1326 1327 if (needs_rtnl) 1328 rtnl_lock(); 1329 lock_sock(sk); 1330 1331 switch (optname) { 1332 case IP_OPTIONS: 1333 { 1334 unsigned char optbuf[sizeof(struct ip_options)+40]; 1335 struct ip_options *opt = (struct ip_options *)optbuf; 1336 struct ip_options_rcu *inet_opt; 1337 1338 inet_opt = rcu_dereference_protected(inet->inet_opt, 1339 lockdep_sock_is_held(sk)); 1340 opt->optlen = 0; 1341 if (inet_opt) 1342 memcpy(optbuf, &inet_opt->opt, 1343 sizeof(struct ip_options) + 1344 inet_opt->opt.optlen); 1345 release_sock(sk); 1346 1347 if (opt->optlen == 0) 1348 return put_user(0, optlen); 1349 1350 ip_options_undo(opt); 1351 1352 len = min_t(unsigned int, len, opt->optlen); 1353 if (put_user(len, optlen)) 1354 return -EFAULT; 1355 if (copy_to_user(optval, opt->__data, len)) 1356 return -EFAULT; 1357 return 0; 1358 } 1359 case IP_PKTINFO: 1360 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1361 break; 1362 case IP_RECVTTL: 1363 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1364 break; 1365 case IP_RECVTOS: 1366 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1367 break; 1368 case IP_RECVOPTS: 1369 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1370 break; 1371 case IP_RETOPTS: 1372 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1373 break; 1374 case IP_PASSSEC: 1375 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1376 break; 1377 case IP_RECVORIGDSTADDR: 1378 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1379 break; 1380 case IP_CHECKSUM: 1381 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0; 1382 break; 1383 case IP_RECVFRAGSIZE: 1384 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0; 1385 break; 1386 case IP_TOS: 1387 val = inet->tos; 1388 break; 1389 case IP_TTL: 1390 { 1391 struct net *net = sock_net(sk); 1392 val = (inet->uc_ttl == -1 ? 1393 net->ipv4.sysctl_ip_default_ttl : 1394 inet->uc_ttl); 1395 break; 1396 } 1397 case IP_HDRINCL: 1398 val = inet->hdrincl; 1399 break; 1400 case IP_NODEFRAG: 1401 val = inet->nodefrag; 1402 break; 1403 case IP_BIND_ADDRESS_NO_PORT: 1404 val = inet->bind_address_no_port; 1405 break; 1406 case IP_MTU_DISCOVER: 1407 val = inet->pmtudisc; 1408 break; 1409 case IP_MTU: 1410 { 1411 struct dst_entry *dst; 1412 val = 0; 1413 dst = sk_dst_get(sk); 1414 if (dst) { 1415 val = dst_mtu(dst); 1416 dst_release(dst); 1417 } 1418 if (!val) { 1419 release_sock(sk); 1420 return -ENOTCONN; 1421 } 1422 break; 1423 } 1424 case IP_RECVERR: 1425 val = inet->recverr; 1426 break; 1427 case IP_MULTICAST_TTL: 1428 val = inet->mc_ttl; 1429 break; 1430 case IP_MULTICAST_LOOP: 1431 val = inet->mc_loop; 1432 break; 1433 case IP_UNICAST_IF: 1434 val = (__force int)htonl((__u32) inet->uc_index); 1435 break; 1436 case IP_MULTICAST_IF: 1437 { 1438 struct in_addr addr; 1439 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1440 addr.s_addr = inet->mc_addr; 1441 release_sock(sk); 1442 1443 if (put_user(len, optlen)) 1444 return -EFAULT; 1445 if (copy_to_user(optval, &addr, len)) 1446 return -EFAULT; 1447 return 0; 1448 } 1449 case IP_MSFILTER: 1450 { 1451 struct ip_msfilter msf; 1452 1453 if (len < IP_MSFILTER_SIZE(0)) { 1454 err = -EINVAL; 1455 goto out; 1456 } 1457 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1458 err = -EFAULT; 1459 goto out; 1460 } 1461 err = ip_mc_msfget(sk, &msf, 1462 (struct ip_msfilter __user *)optval, optlen); 1463 goto out; 1464 } 1465 case MCAST_MSFILTER: 1466 { 1467 struct group_filter gsf; 1468 1469 if (len < GROUP_FILTER_SIZE(0)) { 1470 err = -EINVAL; 1471 goto out; 1472 } 1473 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1474 err = -EFAULT; 1475 goto out; 1476 } 1477 err = ip_mc_gsfget(sk, &gsf, 1478 (struct group_filter __user *)optval, 1479 optlen); 1480 goto out; 1481 } 1482 case IP_MULTICAST_ALL: 1483 val = inet->mc_all; 1484 break; 1485 case IP_PKTOPTIONS: 1486 { 1487 struct msghdr msg; 1488 1489 release_sock(sk); 1490 1491 if (sk->sk_type != SOCK_STREAM) 1492 return -ENOPROTOOPT; 1493 1494 msg.msg_control = (__force void *) optval; 1495 msg.msg_controllen = len; 1496 msg.msg_flags = flags; 1497 1498 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1499 struct in_pktinfo info; 1500 1501 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1502 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1503 info.ipi_ifindex = inet->mc_index; 1504 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1505 } 1506 if (inet->cmsg_flags & IP_CMSG_TTL) { 1507 int hlim = inet->mc_ttl; 1508 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1509 } 1510 if (inet->cmsg_flags & IP_CMSG_TOS) { 1511 int tos = inet->rcv_tos; 1512 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos); 1513 } 1514 len -= msg.msg_controllen; 1515 return put_user(len, optlen); 1516 } 1517 case IP_FREEBIND: 1518 val = inet->freebind; 1519 break; 1520 case IP_TRANSPARENT: 1521 val = inet->transparent; 1522 break; 1523 case IP_MINTTL: 1524 val = inet->min_ttl; 1525 break; 1526 default: 1527 release_sock(sk); 1528 return -ENOPROTOOPT; 1529 } 1530 release_sock(sk); 1531 1532 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1533 unsigned char ucval = (unsigned char)val; 1534 len = 1; 1535 if (put_user(len, optlen)) 1536 return -EFAULT; 1537 if (copy_to_user(optval, &ucval, 1)) 1538 return -EFAULT; 1539 } else { 1540 len = min_t(unsigned int, sizeof(int), len); 1541 if (put_user(len, optlen)) 1542 return -EFAULT; 1543 if (copy_to_user(optval, &val, len)) 1544 return -EFAULT; 1545 } 1546 return 0; 1547 1548 out: 1549 release_sock(sk); 1550 if (needs_rtnl) 1551 rtnl_unlock(); 1552 return err; 1553 } 1554 1555 int ip_getsockopt(struct sock *sk, int level, 1556 int optname, char __user *optval, int __user *optlen) 1557 { 1558 int err; 1559 1560 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1561 #ifdef CONFIG_NETFILTER 1562 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1563 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1564 !ip_mroute_opt(optname)) { 1565 int len; 1566 1567 if (get_user(len, optlen)) 1568 return -EFAULT; 1569 1570 lock_sock(sk); 1571 err = nf_getsockopt(sk, PF_INET, optname, optval, 1572 &len); 1573 release_sock(sk); 1574 if (err >= 0) 1575 err = put_user(len, optlen); 1576 return err; 1577 } 1578 #endif 1579 return err; 1580 } 1581 EXPORT_SYMBOL(ip_getsockopt); 1582 1583 #ifdef CONFIG_COMPAT 1584 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1585 char __user *optval, int __user *optlen) 1586 { 1587 int err; 1588 1589 if (optname == MCAST_MSFILTER) 1590 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1591 ip_getsockopt); 1592 1593 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1594 MSG_CMSG_COMPAT); 1595 1596 #ifdef CONFIG_NETFILTER 1597 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1598 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1599 !ip_mroute_opt(optname)) { 1600 int len; 1601 1602 if (get_user(len, optlen)) 1603 return -EFAULT; 1604 1605 lock_sock(sk); 1606 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1607 release_sock(sk); 1608 if (err >= 0) 1609 err = put_user(len, optlen); 1610 return err; 1611 } 1612 #endif 1613 return err; 1614 } 1615 EXPORT_SYMBOL(compat_ip_getsockopt); 1616 #endif 1617