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 * The IP to API glue. 7 * 8 * Authors: see ip.c 9 * 10 * Fixes: 11 * Many : Split from ip.c , see ip.c for history. 12 * Martin Mares : TOS setting fixed. 13 * Alan Cox : Fixed a couple of oopses in Martin's 14 * TOS tweaks. 15 * Mike McLagan : Routing by source 16 */ 17 18 #include <linux/module.h> 19 #include <linux/types.h> 20 #include <linux/mm.h> 21 #include <linux/skbuff.h> 22 #include <linux/ip.h> 23 #include <linux/icmp.h> 24 #include <linux/inetdevice.h> 25 #include <linux/netdevice.h> 26 #include <linux/slab.h> 27 #include <net/sock.h> 28 #include <net/ip.h> 29 #include <net/icmp.h> 30 #include <net/tcp_states.h> 31 #include <linux/udp.h> 32 #include <linux/igmp.h> 33 #include <linux/netfilter.h> 34 #include <linux/route.h> 35 #include <linux/mroute.h> 36 #include <net/inet_ecn.h> 37 #include <net/route.h> 38 #include <net/xfrm.h> 39 #include <net/compat.h> 40 #include <net/checksum.h> 41 #if IS_ENABLED(CONFIG_IPV6) 42 #include <net/transp_v6.h> 43 #endif 44 #include <net/ip_fib.h> 45 46 #include <linux/errqueue.h> 47 #include <asm/uaccess.h> 48 49 /* 50 * SOL_IP control messages. 51 */ 52 53 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) 54 { 55 struct in_pktinfo info = *PKTINFO_SKB_CB(skb); 56 57 info.ipi_addr.s_addr = ip_hdr(skb)->daddr; 58 59 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 60 } 61 62 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb) 63 { 64 int ttl = ip_hdr(skb)->ttl; 65 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl); 66 } 67 68 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb) 69 { 70 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos); 71 } 72 73 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb) 74 { 75 if (IPCB(skb)->opt.optlen == 0) 76 return; 77 78 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen, 79 ip_hdr(skb) + 1); 80 } 81 82 83 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb) 84 { 85 unsigned char optbuf[sizeof(struct ip_options) + 40]; 86 struct ip_options *opt = (struct ip_options *)optbuf; 87 88 if (IPCB(skb)->opt.optlen == 0) 89 return; 90 91 if (ip_options_echo(opt, skb)) { 92 msg->msg_flags |= MSG_CTRUNC; 93 return; 94 } 95 ip_options_undo(opt); 96 97 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data); 98 } 99 100 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb, 101 int offset) 102 { 103 __wsum csum = skb->csum; 104 105 if (skb->ip_summed != CHECKSUM_COMPLETE) 106 return; 107 108 if (offset != 0) 109 csum = csum_sub(csum, csum_partial(skb_transport_header(skb), 110 offset, 0)); 111 112 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum); 113 } 114 115 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb) 116 { 117 char *secdata; 118 u32 seclen, secid; 119 int err; 120 121 err = security_socket_getpeersec_dgram(NULL, skb, &secid); 122 if (err) 123 return; 124 125 err = security_secid_to_secctx(secid, &secdata, &seclen); 126 if (err) 127 return; 128 129 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata); 130 security_release_secctx(secdata, seclen); 131 } 132 133 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb) 134 { 135 struct sockaddr_in sin; 136 const struct iphdr *iph = ip_hdr(skb); 137 __be16 *ports = (__be16 *)skb_transport_header(skb); 138 139 if (skb_transport_offset(skb) + 4 > skb->len) 140 return; 141 142 /* All current transport protocols have the port numbers in the 143 * first four bytes of the transport header and this function is 144 * written with this assumption in mind. 145 */ 146 147 sin.sin_family = AF_INET; 148 sin.sin_addr.s_addr = iph->daddr; 149 sin.sin_port = ports[1]; 150 memset(sin.sin_zero, 0, sizeof(sin.sin_zero)); 151 152 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin); 153 } 154 155 void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb, 156 int offset) 157 { 158 struct inet_sock *inet = inet_sk(skb->sk); 159 unsigned int flags = inet->cmsg_flags; 160 161 /* Ordered by supposed usage frequency */ 162 if (flags & IP_CMSG_PKTINFO) { 163 ip_cmsg_recv_pktinfo(msg, skb); 164 165 flags &= ~IP_CMSG_PKTINFO; 166 if (!flags) 167 return; 168 } 169 170 if (flags & IP_CMSG_TTL) { 171 ip_cmsg_recv_ttl(msg, skb); 172 173 flags &= ~IP_CMSG_TTL; 174 if (!flags) 175 return; 176 } 177 178 if (flags & IP_CMSG_TOS) { 179 ip_cmsg_recv_tos(msg, skb); 180 181 flags &= ~IP_CMSG_TOS; 182 if (!flags) 183 return; 184 } 185 186 if (flags & IP_CMSG_RECVOPTS) { 187 ip_cmsg_recv_opts(msg, skb); 188 189 flags &= ~IP_CMSG_RECVOPTS; 190 if (!flags) 191 return; 192 } 193 194 if (flags & IP_CMSG_RETOPTS) { 195 ip_cmsg_recv_retopts(msg, skb); 196 197 flags &= ~IP_CMSG_RETOPTS; 198 if (!flags) 199 return; 200 } 201 202 if (flags & IP_CMSG_PASSSEC) { 203 ip_cmsg_recv_security(msg, skb); 204 205 flags &= ~IP_CMSG_PASSSEC; 206 if (!flags) 207 return; 208 } 209 210 if (flags & IP_CMSG_ORIGDSTADDR) { 211 ip_cmsg_recv_dstaddr(msg, skb); 212 213 flags &= ~IP_CMSG_ORIGDSTADDR; 214 if (!flags) 215 return; 216 } 217 218 if (flags & IP_CMSG_CHECKSUM) 219 ip_cmsg_recv_checksum(msg, skb, offset); 220 } 221 EXPORT_SYMBOL(ip_cmsg_recv_offset); 222 223 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc, 224 bool allow_ipv6) 225 { 226 int err, val; 227 struct cmsghdr *cmsg; 228 struct net *net = sock_net(sk); 229 230 for_each_cmsghdr(cmsg, msg) { 231 if (!CMSG_OK(msg, cmsg)) 232 return -EINVAL; 233 #if IS_ENABLED(CONFIG_IPV6) 234 if (allow_ipv6 && 235 cmsg->cmsg_level == SOL_IPV6 && 236 cmsg->cmsg_type == IPV6_PKTINFO) { 237 struct in6_pktinfo *src_info; 238 239 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info))) 240 return -EINVAL; 241 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg); 242 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr)) 243 return -EINVAL; 244 ipc->oif = src_info->ipi6_ifindex; 245 ipc->addr = src_info->ipi6_addr.s6_addr32[3]; 246 continue; 247 } 248 #endif 249 if (cmsg->cmsg_level == SOL_SOCKET) { 250 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc); 251 if (err) 252 return err; 253 continue; 254 } 255 256 if (cmsg->cmsg_level != SOL_IP) 257 continue; 258 switch (cmsg->cmsg_type) { 259 case IP_RETOPTS: 260 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr)); 261 262 /* Our caller is responsible for freeing ipc->opt */ 263 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), 264 err < 40 ? err : 40); 265 if (err) 266 return err; 267 break; 268 case IP_PKTINFO: 269 { 270 struct in_pktinfo *info; 271 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) 272 return -EINVAL; 273 info = (struct in_pktinfo *)CMSG_DATA(cmsg); 274 ipc->oif = info->ipi_ifindex; 275 ipc->addr = info->ipi_spec_dst.s_addr; 276 break; 277 } 278 case IP_TTL: 279 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 280 return -EINVAL; 281 val = *(int *)CMSG_DATA(cmsg); 282 if (val < 1 || val > 255) 283 return -EINVAL; 284 ipc->ttl = val; 285 break; 286 case IP_TOS: 287 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 288 return -EINVAL; 289 val = *(int *)CMSG_DATA(cmsg); 290 if (val < 0 || val > 255) 291 return -EINVAL; 292 ipc->tos = val; 293 ipc->priority = rt_tos2priority(ipc->tos); 294 break; 295 296 default: 297 return -EINVAL; 298 } 299 } 300 return 0; 301 } 302 303 304 /* Special input handler for packets caught by router alert option. 305 They are selected only by protocol field, and then processed likely 306 local ones; but only if someone wants them! Otherwise, router 307 not running rsvpd will kill RSVP. 308 309 It is user level problem, what it will make with them. 310 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)), 311 but receiver should be enough clever f.e. to forward mtrace requests, 312 sent to multicast group to reach destination designated router. 313 */ 314 struct ip_ra_chain __rcu *ip_ra_chain; 315 static DEFINE_SPINLOCK(ip_ra_lock); 316 317 318 static void ip_ra_destroy_rcu(struct rcu_head *head) 319 { 320 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu); 321 322 sock_put(ra->saved_sk); 323 kfree(ra); 324 } 325 326 int ip_ra_control(struct sock *sk, unsigned char on, 327 void (*destructor)(struct sock *)) 328 { 329 struct ip_ra_chain *ra, *new_ra; 330 struct ip_ra_chain __rcu **rap; 331 332 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 333 return -EINVAL; 334 335 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 336 337 spin_lock_bh(&ip_ra_lock); 338 for (rap = &ip_ra_chain; 339 (ra = rcu_dereference_protected(*rap, 340 lockdep_is_held(&ip_ra_lock))) != NULL; 341 rap = &ra->next) { 342 if (ra->sk == sk) { 343 if (on) { 344 spin_unlock_bh(&ip_ra_lock); 345 kfree(new_ra); 346 return -EADDRINUSE; 347 } 348 /* dont let ip_call_ra_chain() use sk again */ 349 ra->sk = NULL; 350 RCU_INIT_POINTER(*rap, ra->next); 351 spin_unlock_bh(&ip_ra_lock); 352 353 if (ra->destructor) 354 ra->destructor(sk); 355 /* 356 * Delay sock_put(sk) and kfree(ra) after one rcu grace 357 * period. This guarantee ip_call_ra_chain() dont need 358 * to mess with socket refcounts. 359 */ 360 ra->saved_sk = sk; 361 call_rcu(&ra->rcu, ip_ra_destroy_rcu); 362 return 0; 363 } 364 } 365 if (!new_ra) { 366 spin_unlock_bh(&ip_ra_lock); 367 return -ENOBUFS; 368 } 369 new_ra->sk = sk; 370 new_ra->destructor = destructor; 371 372 RCU_INIT_POINTER(new_ra->next, ra); 373 rcu_assign_pointer(*rap, new_ra); 374 sock_hold(sk); 375 spin_unlock_bh(&ip_ra_lock); 376 377 return 0; 378 } 379 380 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 381 __be16 port, u32 info, u8 *payload) 382 { 383 struct sock_exterr_skb *serr; 384 385 skb = skb_clone(skb, GFP_ATOMIC); 386 if (!skb) 387 return; 388 389 serr = SKB_EXT_ERR(skb); 390 serr->ee.ee_errno = err; 391 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 392 serr->ee.ee_type = icmp_hdr(skb)->type; 393 serr->ee.ee_code = icmp_hdr(skb)->code; 394 serr->ee.ee_pad = 0; 395 serr->ee.ee_info = info; 396 serr->ee.ee_data = 0; 397 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 398 skb_network_header(skb); 399 serr->port = port; 400 401 if (skb_pull(skb, payload - skb->data)) { 402 skb_reset_transport_header(skb); 403 if (sock_queue_err_skb(sk, skb) == 0) 404 return; 405 } 406 kfree_skb(skb); 407 } 408 409 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 410 { 411 struct inet_sock *inet = inet_sk(sk); 412 struct sock_exterr_skb *serr; 413 struct iphdr *iph; 414 struct sk_buff *skb; 415 416 if (!inet->recverr) 417 return; 418 419 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 420 if (!skb) 421 return; 422 423 skb_put(skb, sizeof(struct iphdr)); 424 skb_reset_network_header(skb); 425 iph = ip_hdr(skb); 426 iph->daddr = daddr; 427 428 serr = SKB_EXT_ERR(skb); 429 serr->ee.ee_errno = err; 430 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 431 serr->ee.ee_type = 0; 432 serr->ee.ee_code = 0; 433 serr->ee.ee_pad = 0; 434 serr->ee.ee_info = info; 435 serr->ee.ee_data = 0; 436 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 437 serr->port = port; 438 439 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 440 skb_reset_transport_header(skb); 441 442 if (sock_queue_err_skb(sk, skb)) 443 kfree_skb(skb); 444 } 445 446 /* For some errors we have valid addr_offset even with zero payload and 447 * zero port. Also, addr_offset should be supported if port is set. 448 */ 449 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr) 450 { 451 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP || 452 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port; 453 } 454 455 /* IPv4 supports cmsg on all imcp errors and some timestamps 456 * 457 * Timestamp code paths do not initialize the fields expected by cmsg: 458 * the PKTINFO fields in skb->cb[]. Fill those in here. 459 */ 460 static bool ipv4_datagram_support_cmsg(const struct sock *sk, 461 struct sk_buff *skb, 462 int ee_origin) 463 { 464 struct in_pktinfo *info; 465 466 if (ee_origin == SO_EE_ORIGIN_ICMP) 467 return true; 468 469 if (ee_origin == SO_EE_ORIGIN_LOCAL) 470 return false; 471 472 /* Support IP_PKTINFO on tstamp packets if requested, to correlate 473 * timestamp with egress dev. Not possible for packets without dev 474 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY). 475 */ 476 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) || 477 (!skb->dev)) 478 return false; 479 480 info = PKTINFO_SKB_CB(skb); 481 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr; 482 info->ipi_ifindex = skb->dev->ifindex; 483 return true; 484 } 485 486 /* 487 * Handle MSG_ERRQUEUE 488 */ 489 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 490 { 491 struct sock_exterr_skb *serr; 492 struct sk_buff *skb; 493 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 494 struct { 495 struct sock_extended_err ee; 496 struct sockaddr_in offender; 497 } errhdr; 498 int err; 499 int copied; 500 501 WARN_ON_ONCE(sk->sk_family == AF_INET6); 502 503 err = -EAGAIN; 504 skb = sock_dequeue_err_skb(sk); 505 if (!skb) 506 goto out; 507 508 copied = skb->len; 509 if (copied > len) { 510 msg->msg_flags |= MSG_TRUNC; 511 copied = len; 512 } 513 err = skb_copy_datagram_msg(skb, 0, msg, copied); 514 if (unlikely(err)) { 515 kfree_skb(skb); 516 return err; 517 } 518 sock_recv_timestamp(msg, sk, skb); 519 520 serr = SKB_EXT_ERR(skb); 521 522 if (sin && ipv4_datagram_support_addr(serr)) { 523 sin->sin_family = AF_INET; 524 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 525 serr->addr_offset); 526 sin->sin_port = serr->port; 527 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 528 *addr_len = sizeof(*sin); 529 } 530 531 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 532 sin = &errhdr.offender; 533 memset(sin, 0, sizeof(*sin)); 534 535 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) { 536 sin->sin_family = AF_INET; 537 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 538 if (inet_sk(sk)->cmsg_flags) 539 ip_cmsg_recv(msg, skb); 540 } 541 542 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 543 544 /* Now we could try to dump offended packet options */ 545 546 msg->msg_flags |= MSG_ERRQUEUE; 547 err = copied; 548 549 consume_skb(skb); 550 out: 551 return err; 552 } 553 554 555 /* 556 * Socket option code for IP. This is the end of the line after any 557 * TCP,UDP etc options on an IP socket. 558 */ 559 static bool setsockopt_needs_rtnl(int optname) 560 { 561 switch (optname) { 562 case IP_ADD_MEMBERSHIP: 563 case IP_ADD_SOURCE_MEMBERSHIP: 564 case IP_BLOCK_SOURCE: 565 case IP_DROP_MEMBERSHIP: 566 case IP_DROP_SOURCE_MEMBERSHIP: 567 case IP_MSFILTER: 568 case IP_UNBLOCK_SOURCE: 569 case MCAST_BLOCK_SOURCE: 570 case MCAST_MSFILTER: 571 case MCAST_JOIN_GROUP: 572 case MCAST_JOIN_SOURCE_GROUP: 573 case MCAST_LEAVE_GROUP: 574 case MCAST_LEAVE_SOURCE_GROUP: 575 case MCAST_UNBLOCK_SOURCE: 576 return true; 577 } 578 return false; 579 } 580 581 static int do_ip_setsockopt(struct sock *sk, int level, 582 int optname, char __user *optval, unsigned int optlen) 583 { 584 struct inet_sock *inet = inet_sk(sk); 585 struct net *net = sock_net(sk); 586 int val = 0, err; 587 bool needs_rtnl = setsockopt_needs_rtnl(optname); 588 589 switch (optname) { 590 case IP_PKTINFO: 591 case IP_RECVTTL: 592 case IP_RECVOPTS: 593 case IP_RECVTOS: 594 case IP_RETOPTS: 595 case IP_TOS: 596 case IP_TTL: 597 case IP_HDRINCL: 598 case IP_MTU_DISCOVER: 599 case IP_RECVERR: 600 case IP_ROUTER_ALERT: 601 case IP_FREEBIND: 602 case IP_PASSSEC: 603 case IP_TRANSPARENT: 604 case IP_MINTTL: 605 case IP_NODEFRAG: 606 case IP_BIND_ADDRESS_NO_PORT: 607 case IP_UNICAST_IF: 608 case IP_MULTICAST_TTL: 609 case IP_MULTICAST_ALL: 610 case IP_MULTICAST_LOOP: 611 case IP_RECVORIGDSTADDR: 612 case IP_CHECKSUM: 613 if (optlen >= sizeof(int)) { 614 if (get_user(val, (int __user *) optval)) 615 return -EFAULT; 616 } else if (optlen >= sizeof(char)) { 617 unsigned char ucval; 618 619 if (get_user(ucval, (unsigned char __user *) optval)) 620 return -EFAULT; 621 val = (int) ucval; 622 } 623 } 624 625 /* If optlen==0, it is equivalent to val == 0 */ 626 627 if (ip_mroute_opt(optname)) 628 return ip_mroute_setsockopt(sk, optname, optval, optlen); 629 630 err = 0; 631 if (needs_rtnl) 632 rtnl_lock(); 633 lock_sock(sk); 634 635 switch (optname) { 636 case IP_OPTIONS: 637 { 638 struct ip_options_rcu *old, *opt = NULL; 639 640 if (optlen > 40) 641 goto e_inval; 642 err = ip_options_get_from_user(sock_net(sk), &opt, 643 optval, optlen); 644 if (err) 645 break; 646 old = rcu_dereference_protected(inet->inet_opt, 647 lockdep_sock_is_held(sk)); 648 if (inet->is_icsk) { 649 struct inet_connection_sock *icsk = inet_csk(sk); 650 #if IS_ENABLED(CONFIG_IPV6) 651 if (sk->sk_family == PF_INET || 652 (!((1 << sk->sk_state) & 653 (TCPF_LISTEN | TCPF_CLOSE)) && 654 inet->inet_daddr != LOOPBACK4_IPV6)) { 655 #endif 656 if (old) 657 icsk->icsk_ext_hdr_len -= old->opt.optlen; 658 if (opt) 659 icsk->icsk_ext_hdr_len += opt->opt.optlen; 660 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 661 #if IS_ENABLED(CONFIG_IPV6) 662 } 663 #endif 664 } 665 rcu_assign_pointer(inet->inet_opt, opt); 666 if (old) 667 kfree_rcu(old, rcu); 668 break; 669 } 670 case IP_PKTINFO: 671 if (val) 672 inet->cmsg_flags |= IP_CMSG_PKTINFO; 673 else 674 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 675 break; 676 case IP_RECVTTL: 677 if (val) 678 inet->cmsg_flags |= IP_CMSG_TTL; 679 else 680 inet->cmsg_flags &= ~IP_CMSG_TTL; 681 break; 682 case IP_RECVTOS: 683 if (val) 684 inet->cmsg_flags |= IP_CMSG_TOS; 685 else 686 inet->cmsg_flags &= ~IP_CMSG_TOS; 687 break; 688 case IP_RECVOPTS: 689 if (val) 690 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 691 else 692 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 693 break; 694 case IP_RETOPTS: 695 if (val) 696 inet->cmsg_flags |= IP_CMSG_RETOPTS; 697 else 698 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 699 break; 700 case IP_PASSSEC: 701 if (val) 702 inet->cmsg_flags |= IP_CMSG_PASSSEC; 703 else 704 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 705 break; 706 case IP_RECVORIGDSTADDR: 707 if (val) 708 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 709 else 710 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 711 break; 712 case IP_CHECKSUM: 713 if (val) { 714 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) { 715 inet_inc_convert_csum(sk); 716 inet->cmsg_flags |= IP_CMSG_CHECKSUM; 717 } 718 } else { 719 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) { 720 inet_dec_convert_csum(sk); 721 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM; 722 } 723 } 724 break; 725 case IP_TOS: /* This sets both TOS and Precedence */ 726 if (sk->sk_type == SOCK_STREAM) { 727 val &= ~INET_ECN_MASK; 728 val |= inet->tos & INET_ECN_MASK; 729 } 730 if (inet->tos != val) { 731 inet->tos = val; 732 sk->sk_priority = rt_tos2priority(val); 733 sk_dst_reset(sk); 734 } 735 break; 736 case IP_TTL: 737 if (optlen < 1) 738 goto e_inval; 739 if (val != -1 && (val < 1 || val > 255)) 740 goto e_inval; 741 inet->uc_ttl = val; 742 break; 743 case IP_HDRINCL: 744 if (sk->sk_type != SOCK_RAW) { 745 err = -ENOPROTOOPT; 746 break; 747 } 748 inet->hdrincl = val ? 1 : 0; 749 break; 750 case IP_NODEFRAG: 751 if (sk->sk_type != SOCK_RAW) { 752 err = -ENOPROTOOPT; 753 break; 754 } 755 inet->nodefrag = val ? 1 : 0; 756 break; 757 case IP_BIND_ADDRESS_NO_PORT: 758 inet->bind_address_no_port = val ? 1 : 0; 759 break; 760 case IP_MTU_DISCOVER: 761 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT) 762 goto e_inval; 763 inet->pmtudisc = val; 764 break; 765 case IP_RECVERR: 766 inet->recverr = !!val; 767 if (!val) 768 skb_queue_purge(&sk->sk_error_queue); 769 break; 770 case IP_MULTICAST_TTL: 771 if (sk->sk_type == SOCK_STREAM) 772 goto e_inval; 773 if (optlen < 1) 774 goto e_inval; 775 if (val == -1) 776 val = 1; 777 if (val < 0 || val > 255) 778 goto e_inval; 779 inet->mc_ttl = val; 780 break; 781 case IP_MULTICAST_LOOP: 782 if (optlen < 1) 783 goto e_inval; 784 inet->mc_loop = !!val; 785 break; 786 case IP_UNICAST_IF: 787 { 788 struct net_device *dev = NULL; 789 int ifindex; 790 791 if (optlen != sizeof(int)) 792 goto e_inval; 793 794 ifindex = (__force int)ntohl((__force __be32)val); 795 if (ifindex == 0) { 796 inet->uc_index = 0; 797 err = 0; 798 break; 799 } 800 801 dev = dev_get_by_index(sock_net(sk), ifindex); 802 err = -EADDRNOTAVAIL; 803 if (!dev) 804 break; 805 dev_put(dev); 806 807 err = -EINVAL; 808 if (sk->sk_bound_dev_if) 809 break; 810 811 inet->uc_index = ifindex; 812 err = 0; 813 break; 814 } 815 case IP_MULTICAST_IF: 816 { 817 struct ip_mreqn mreq; 818 struct net_device *dev = NULL; 819 820 if (sk->sk_type == SOCK_STREAM) 821 goto e_inval; 822 /* 823 * Check the arguments are allowable 824 */ 825 826 if (optlen < sizeof(struct in_addr)) 827 goto e_inval; 828 829 err = -EFAULT; 830 if (optlen >= sizeof(struct ip_mreqn)) { 831 if (copy_from_user(&mreq, optval, sizeof(mreq))) 832 break; 833 } else { 834 memset(&mreq, 0, sizeof(mreq)); 835 if (optlen >= sizeof(struct ip_mreq)) { 836 if (copy_from_user(&mreq, optval, 837 sizeof(struct ip_mreq))) 838 break; 839 } else if (optlen >= sizeof(struct in_addr)) { 840 if (copy_from_user(&mreq.imr_address, optval, 841 sizeof(struct in_addr))) 842 break; 843 } 844 } 845 846 if (!mreq.imr_ifindex) { 847 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 848 inet->mc_index = 0; 849 inet->mc_addr = 0; 850 err = 0; 851 break; 852 } 853 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 854 if (dev) 855 mreq.imr_ifindex = dev->ifindex; 856 } else 857 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 858 859 860 err = -EADDRNOTAVAIL; 861 if (!dev) 862 break; 863 dev_put(dev); 864 865 err = -EINVAL; 866 if (sk->sk_bound_dev_if && 867 mreq.imr_ifindex != sk->sk_bound_dev_if) 868 break; 869 870 inet->mc_index = mreq.imr_ifindex; 871 inet->mc_addr = mreq.imr_address.s_addr; 872 err = 0; 873 break; 874 } 875 876 case IP_ADD_MEMBERSHIP: 877 case IP_DROP_MEMBERSHIP: 878 { 879 struct ip_mreqn mreq; 880 881 err = -EPROTO; 882 if (inet_sk(sk)->is_icsk) 883 break; 884 885 if (optlen < sizeof(struct ip_mreq)) 886 goto e_inval; 887 err = -EFAULT; 888 if (optlen >= sizeof(struct ip_mreqn)) { 889 if (copy_from_user(&mreq, optval, sizeof(mreq))) 890 break; 891 } else { 892 memset(&mreq, 0, sizeof(mreq)); 893 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 894 break; 895 } 896 897 if (optname == IP_ADD_MEMBERSHIP) 898 err = ip_mc_join_group(sk, &mreq); 899 else 900 err = ip_mc_leave_group(sk, &mreq); 901 break; 902 } 903 case IP_MSFILTER: 904 { 905 struct ip_msfilter *msf; 906 907 if (optlen < IP_MSFILTER_SIZE(0)) 908 goto e_inval; 909 if (optlen > sysctl_optmem_max) { 910 err = -ENOBUFS; 911 break; 912 } 913 msf = kmalloc(optlen, GFP_KERNEL); 914 if (!msf) { 915 err = -ENOBUFS; 916 break; 917 } 918 err = -EFAULT; 919 if (copy_from_user(msf, optval, optlen)) { 920 kfree(msf); 921 break; 922 } 923 /* numsrc >= (1G-4) overflow in 32 bits */ 924 if (msf->imsf_numsrc >= 0x3ffffffcU || 925 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) { 926 kfree(msf); 927 err = -ENOBUFS; 928 break; 929 } 930 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 931 kfree(msf); 932 err = -EINVAL; 933 break; 934 } 935 err = ip_mc_msfilter(sk, msf, 0); 936 kfree(msf); 937 break; 938 } 939 case IP_BLOCK_SOURCE: 940 case IP_UNBLOCK_SOURCE: 941 case IP_ADD_SOURCE_MEMBERSHIP: 942 case IP_DROP_SOURCE_MEMBERSHIP: 943 { 944 struct ip_mreq_source mreqs; 945 int omode, add; 946 947 if (optlen != sizeof(struct ip_mreq_source)) 948 goto e_inval; 949 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 950 err = -EFAULT; 951 break; 952 } 953 if (optname == IP_BLOCK_SOURCE) { 954 omode = MCAST_EXCLUDE; 955 add = 1; 956 } else if (optname == IP_UNBLOCK_SOURCE) { 957 omode = MCAST_EXCLUDE; 958 add = 0; 959 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 960 struct ip_mreqn mreq; 961 962 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 963 mreq.imr_address.s_addr = mreqs.imr_interface; 964 mreq.imr_ifindex = 0; 965 err = ip_mc_join_group(sk, &mreq); 966 if (err && err != -EADDRINUSE) 967 break; 968 omode = MCAST_INCLUDE; 969 add = 1; 970 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 971 omode = MCAST_INCLUDE; 972 add = 0; 973 } 974 err = ip_mc_source(add, omode, sk, &mreqs, 0); 975 break; 976 } 977 case MCAST_JOIN_GROUP: 978 case MCAST_LEAVE_GROUP: 979 { 980 struct group_req greq; 981 struct sockaddr_in *psin; 982 struct ip_mreqn mreq; 983 984 if (optlen < sizeof(struct group_req)) 985 goto e_inval; 986 err = -EFAULT; 987 if (copy_from_user(&greq, optval, sizeof(greq))) 988 break; 989 psin = (struct sockaddr_in *)&greq.gr_group; 990 if (psin->sin_family != AF_INET) 991 goto e_inval; 992 memset(&mreq, 0, sizeof(mreq)); 993 mreq.imr_multiaddr = psin->sin_addr; 994 mreq.imr_ifindex = greq.gr_interface; 995 996 if (optname == MCAST_JOIN_GROUP) 997 err = ip_mc_join_group(sk, &mreq); 998 else 999 err = ip_mc_leave_group(sk, &mreq); 1000 break; 1001 } 1002 case MCAST_JOIN_SOURCE_GROUP: 1003 case MCAST_LEAVE_SOURCE_GROUP: 1004 case MCAST_BLOCK_SOURCE: 1005 case MCAST_UNBLOCK_SOURCE: 1006 { 1007 struct group_source_req greqs; 1008 struct ip_mreq_source mreqs; 1009 struct sockaddr_in *psin; 1010 int omode, add; 1011 1012 if (optlen != sizeof(struct group_source_req)) 1013 goto e_inval; 1014 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 1015 err = -EFAULT; 1016 break; 1017 } 1018 if (greqs.gsr_group.ss_family != AF_INET || 1019 greqs.gsr_source.ss_family != AF_INET) { 1020 err = -EADDRNOTAVAIL; 1021 break; 1022 } 1023 psin = (struct sockaddr_in *)&greqs.gsr_group; 1024 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 1025 psin = (struct sockaddr_in *)&greqs.gsr_source; 1026 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 1027 mreqs.imr_interface = 0; /* use index for mc_source */ 1028 1029 if (optname == MCAST_BLOCK_SOURCE) { 1030 omode = MCAST_EXCLUDE; 1031 add = 1; 1032 } else if (optname == MCAST_UNBLOCK_SOURCE) { 1033 omode = MCAST_EXCLUDE; 1034 add = 0; 1035 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 1036 struct ip_mreqn mreq; 1037 1038 psin = (struct sockaddr_in *)&greqs.gsr_group; 1039 mreq.imr_multiaddr = psin->sin_addr; 1040 mreq.imr_address.s_addr = 0; 1041 mreq.imr_ifindex = greqs.gsr_interface; 1042 err = ip_mc_join_group(sk, &mreq); 1043 if (err && err != -EADDRINUSE) 1044 break; 1045 greqs.gsr_interface = mreq.imr_ifindex; 1046 omode = MCAST_INCLUDE; 1047 add = 1; 1048 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 1049 omode = MCAST_INCLUDE; 1050 add = 0; 1051 } 1052 err = ip_mc_source(add, omode, sk, &mreqs, 1053 greqs.gsr_interface); 1054 break; 1055 } 1056 case MCAST_MSFILTER: 1057 { 1058 struct sockaddr_in *psin; 1059 struct ip_msfilter *msf = NULL; 1060 struct group_filter *gsf = NULL; 1061 int msize, i, ifindex; 1062 1063 if (optlen < GROUP_FILTER_SIZE(0)) 1064 goto e_inval; 1065 if (optlen > sysctl_optmem_max) { 1066 err = -ENOBUFS; 1067 break; 1068 } 1069 gsf = kmalloc(optlen, GFP_KERNEL); 1070 if (!gsf) { 1071 err = -ENOBUFS; 1072 break; 1073 } 1074 err = -EFAULT; 1075 if (copy_from_user(gsf, optval, optlen)) 1076 goto mc_msf_out; 1077 1078 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 1079 if (gsf->gf_numsrc >= 0x1ffffff || 1080 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) { 1081 err = -ENOBUFS; 1082 goto mc_msf_out; 1083 } 1084 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 1085 err = -EINVAL; 1086 goto mc_msf_out; 1087 } 1088 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 1089 msf = kmalloc(msize, GFP_KERNEL); 1090 if (!msf) { 1091 err = -ENOBUFS; 1092 goto mc_msf_out; 1093 } 1094 ifindex = gsf->gf_interface; 1095 psin = (struct sockaddr_in *)&gsf->gf_group; 1096 if (psin->sin_family != AF_INET) { 1097 err = -EADDRNOTAVAIL; 1098 goto mc_msf_out; 1099 } 1100 msf->imsf_multiaddr = psin->sin_addr.s_addr; 1101 msf->imsf_interface = 0; 1102 msf->imsf_fmode = gsf->gf_fmode; 1103 msf->imsf_numsrc = gsf->gf_numsrc; 1104 err = -EADDRNOTAVAIL; 1105 for (i = 0; i < gsf->gf_numsrc; ++i) { 1106 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 1107 1108 if (psin->sin_family != AF_INET) 1109 goto mc_msf_out; 1110 msf->imsf_slist[i] = psin->sin_addr.s_addr; 1111 } 1112 kfree(gsf); 1113 gsf = NULL; 1114 1115 err = ip_mc_msfilter(sk, msf, ifindex); 1116 mc_msf_out: 1117 kfree(msf); 1118 kfree(gsf); 1119 break; 1120 } 1121 case IP_MULTICAST_ALL: 1122 if (optlen < 1) 1123 goto e_inval; 1124 if (val != 0 && val != 1) 1125 goto e_inval; 1126 inet->mc_all = val; 1127 break; 1128 case IP_ROUTER_ALERT: 1129 err = ip_ra_control(sk, val ? 1 : 0, NULL); 1130 break; 1131 1132 case IP_FREEBIND: 1133 if (optlen < 1) 1134 goto e_inval; 1135 inet->freebind = !!val; 1136 break; 1137 1138 case IP_IPSEC_POLICY: 1139 case IP_XFRM_POLICY: 1140 err = -EPERM; 1141 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 1142 break; 1143 err = xfrm_user_policy(sk, optname, optval, optlen); 1144 break; 1145 1146 case IP_TRANSPARENT: 1147 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) && 1148 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) { 1149 err = -EPERM; 1150 break; 1151 } 1152 if (optlen < 1) 1153 goto e_inval; 1154 inet->transparent = !!val; 1155 break; 1156 1157 case IP_MINTTL: 1158 if (optlen < 1) 1159 goto e_inval; 1160 if (val < 0 || val > 255) 1161 goto e_inval; 1162 inet->min_ttl = val; 1163 break; 1164 1165 default: 1166 err = -ENOPROTOOPT; 1167 break; 1168 } 1169 release_sock(sk); 1170 if (needs_rtnl) 1171 rtnl_unlock(); 1172 return err; 1173 1174 e_inval: 1175 release_sock(sk); 1176 if (needs_rtnl) 1177 rtnl_unlock(); 1178 return -EINVAL; 1179 } 1180 1181 /** 1182 * ipv4_pktinfo_prepare - transfer some info from rtable to skb 1183 * @sk: socket 1184 * @skb: buffer 1185 * 1186 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific 1187 * destination in skb->cb[] before dst drop. 1188 * This way, receiver doesn't make cache line misses to read rtable. 1189 */ 1190 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb) 1191 { 1192 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb); 1193 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) || 1194 ipv6_sk_rxinfo(sk); 1195 1196 if (prepare && skb_rtable(skb)) { 1197 /* skb->cb is overloaded: prior to this point it is IP{6}CB 1198 * which has interface index (iif) as the first member of the 1199 * underlying inet{6}_skb_parm struct. This code then overlays 1200 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first 1201 * element so the iif is picked up from the prior IPCB 1202 */ 1203 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb); 1204 } else { 1205 pktinfo->ipi_ifindex = 0; 1206 pktinfo->ipi_spec_dst.s_addr = 0; 1207 } 1208 skb_dst_drop(skb); 1209 } 1210 1211 int ip_setsockopt(struct sock *sk, int level, 1212 int optname, char __user *optval, unsigned int optlen) 1213 { 1214 int err; 1215 1216 if (level != SOL_IP) 1217 return -ENOPROTOOPT; 1218 1219 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1220 #ifdef CONFIG_NETFILTER 1221 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1222 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1223 optname != IP_IPSEC_POLICY && 1224 optname != IP_XFRM_POLICY && 1225 !ip_mroute_opt(optname)) { 1226 lock_sock(sk); 1227 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1228 release_sock(sk); 1229 } 1230 #endif 1231 return err; 1232 } 1233 EXPORT_SYMBOL(ip_setsockopt); 1234 1235 #ifdef CONFIG_COMPAT 1236 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1237 char __user *optval, unsigned int optlen) 1238 { 1239 int err; 1240 1241 if (level != SOL_IP) 1242 return -ENOPROTOOPT; 1243 1244 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1245 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1246 ip_setsockopt); 1247 1248 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1249 #ifdef CONFIG_NETFILTER 1250 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1251 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1252 optname != IP_IPSEC_POLICY && 1253 optname != IP_XFRM_POLICY && 1254 !ip_mroute_opt(optname)) { 1255 lock_sock(sk); 1256 err = compat_nf_setsockopt(sk, PF_INET, optname, 1257 optval, optlen); 1258 release_sock(sk); 1259 } 1260 #endif 1261 return err; 1262 } 1263 EXPORT_SYMBOL(compat_ip_setsockopt); 1264 #endif 1265 1266 /* 1267 * Get the options. Note for future reference. The GET of IP options gets 1268 * the _received_ ones. The set sets the _sent_ ones. 1269 */ 1270 1271 static bool getsockopt_needs_rtnl(int optname) 1272 { 1273 switch (optname) { 1274 case IP_MSFILTER: 1275 case MCAST_MSFILTER: 1276 return true; 1277 } 1278 return false; 1279 } 1280 1281 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1282 char __user *optval, int __user *optlen, unsigned int flags) 1283 { 1284 struct inet_sock *inet = inet_sk(sk); 1285 bool needs_rtnl = getsockopt_needs_rtnl(optname); 1286 int val, err = 0; 1287 int len; 1288 1289 if (level != SOL_IP) 1290 return -EOPNOTSUPP; 1291 1292 if (ip_mroute_opt(optname)) 1293 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1294 1295 if (get_user(len, optlen)) 1296 return -EFAULT; 1297 if (len < 0) 1298 return -EINVAL; 1299 1300 if (needs_rtnl) 1301 rtnl_lock(); 1302 lock_sock(sk); 1303 1304 switch (optname) { 1305 case IP_OPTIONS: 1306 { 1307 unsigned char optbuf[sizeof(struct ip_options)+40]; 1308 struct ip_options *opt = (struct ip_options *)optbuf; 1309 struct ip_options_rcu *inet_opt; 1310 1311 inet_opt = rcu_dereference_protected(inet->inet_opt, 1312 lockdep_sock_is_held(sk)); 1313 opt->optlen = 0; 1314 if (inet_opt) 1315 memcpy(optbuf, &inet_opt->opt, 1316 sizeof(struct ip_options) + 1317 inet_opt->opt.optlen); 1318 release_sock(sk); 1319 1320 if (opt->optlen == 0) 1321 return put_user(0, optlen); 1322 1323 ip_options_undo(opt); 1324 1325 len = min_t(unsigned int, len, opt->optlen); 1326 if (put_user(len, optlen)) 1327 return -EFAULT; 1328 if (copy_to_user(optval, opt->__data, len)) 1329 return -EFAULT; 1330 return 0; 1331 } 1332 case IP_PKTINFO: 1333 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1334 break; 1335 case IP_RECVTTL: 1336 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1337 break; 1338 case IP_RECVTOS: 1339 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1340 break; 1341 case IP_RECVOPTS: 1342 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1343 break; 1344 case IP_RETOPTS: 1345 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1346 break; 1347 case IP_PASSSEC: 1348 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1349 break; 1350 case IP_RECVORIGDSTADDR: 1351 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1352 break; 1353 case IP_CHECKSUM: 1354 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0; 1355 break; 1356 case IP_TOS: 1357 val = inet->tos; 1358 break; 1359 case IP_TTL: 1360 { 1361 struct net *net = sock_net(sk); 1362 val = (inet->uc_ttl == -1 ? 1363 net->ipv4.sysctl_ip_default_ttl : 1364 inet->uc_ttl); 1365 break; 1366 } 1367 case IP_HDRINCL: 1368 val = inet->hdrincl; 1369 break; 1370 case IP_NODEFRAG: 1371 val = inet->nodefrag; 1372 break; 1373 case IP_BIND_ADDRESS_NO_PORT: 1374 val = inet->bind_address_no_port; 1375 break; 1376 case IP_MTU_DISCOVER: 1377 val = inet->pmtudisc; 1378 break; 1379 case IP_MTU: 1380 { 1381 struct dst_entry *dst; 1382 val = 0; 1383 dst = sk_dst_get(sk); 1384 if (dst) { 1385 val = dst_mtu(dst); 1386 dst_release(dst); 1387 } 1388 if (!val) { 1389 release_sock(sk); 1390 return -ENOTCONN; 1391 } 1392 break; 1393 } 1394 case IP_RECVERR: 1395 val = inet->recverr; 1396 break; 1397 case IP_MULTICAST_TTL: 1398 val = inet->mc_ttl; 1399 break; 1400 case IP_MULTICAST_LOOP: 1401 val = inet->mc_loop; 1402 break; 1403 case IP_UNICAST_IF: 1404 val = (__force int)htonl((__u32) inet->uc_index); 1405 break; 1406 case IP_MULTICAST_IF: 1407 { 1408 struct in_addr addr; 1409 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1410 addr.s_addr = inet->mc_addr; 1411 release_sock(sk); 1412 1413 if (put_user(len, optlen)) 1414 return -EFAULT; 1415 if (copy_to_user(optval, &addr, len)) 1416 return -EFAULT; 1417 return 0; 1418 } 1419 case IP_MSFILTER: 1420 { 1421 struct ip_msfilter msf; 1422 1423 if (len < IP_MSFILTER_SIZE(0)) { 1424 err = -EINVAL; 1425 goto out; 1426 } 1427 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1428 err = -EFAULT; 1429 goto out; 1430 } 1431 err = ip_mc_msfget(sk, &msf, 1432 (struct ip_msfilter __user *)optval, optlen); 1433 goto out; 1434 } 1435 case MCAST_MSFILTER: 1436 { 1437 struct group_filter gsf; 1438 1439 if (len < GROUP_FILTER_SIZE(0)) { 1440 err = -EINVAL; 1441 goto out; 1442 } 1443 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1444 err = -EFAULT; 1445 goto out; 1446 } 1447 err = ip_mc_gsfget(sk, &gsf, 1448 (struct group_filter __user *)optval, 1449 optlen); 1450 goto out; 1451 } 1452 case IP_MULTICAST_ALL: 1453 val = inet->mc_all; 1454 break; 1455 case IP_PKTOPTIONS: 1456 { 1457 struct msghdr msg; 1458 1459 release_sock(sk); 1460 1461 if (sk->sk_type != SOCK_STREAM) 1462 return -ENOPROTOOPT; 1463 1464 msg.msg_control = (__force void *) optval; 1465 msg.msg_controllen = len; 1466 msg.msg_flags = flags; 1467 1468 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1469 struct in_pktinfo info; 1470 1471 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1472 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1473 info.ipi_ifindex = inet->mc_index; 1474 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1475 } 1476 if (inet->cmsg_flags & IP_CMSG_TTL) { 1477 int hlim = inet->mc_ttl; 1478 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1479 } 1480 if (inet->cmsg_flags & IP_CMSG_TOS) { 1481 int tos = inet->rcv_tos; 1482 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos); 1483 } 1484 len -= msg.msg_controllen; 1485 return put_user(len, optlen); 1486 } 1487 case IP_FREEBIND: 1488 val = inet->freebind; 1489 break; 1490 case IP_TRANSPARENT: 1491 val = inet->transparent; 1492 break; 1493 case IP_MINTTL: 1494 val = inet->min_ttl; 1495 break; 1496 default: 1497 release_sock(sk); 1498 return -ENOPROTOOPT; 1499 } 1500 release_sock(sk); 1501 1502 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1503 unsigned char ucval = (unsigned char)val; 1504 len = 1; 1505 if (put_user(len, optlen)) 1506 return -EFAULT; 1507 if (copy_to_user(optval, &ucval, 1)) 1508 return -EFAULT; 1509 } else { 1510 len = min_t(unsigned int, sizeof(int), len); 1511 if (put_user(len, optlen)) 1512 return -EFAULT; 1513 if (copy_to_user(optval, &val, len)) 1514 return -EFAULT; 1515 } 1516 return 0; 1517 1518 out: 1519 release_sock(sk); 1520 if (needs_rtnl) 1521 rtnl_unlock(); 1522 return err; 1523 } 1524 1525 int ip_getsockopt(struct sock *sk, int level, 1526 int optname, char __user *optval, int __user *optlen) 1527 { 1528 int err; 1529 1530 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1531 #ifdef CONFIG_NETFILTER 1532 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1533 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1534 !ip_mroute_opt(optname)) { 1535 int len; 1536 1537 if (get_user(len, optlen)) 1538 return -EFAULT; 1539 1540 lock_sock(sk); 1541 err = nf_getsockopt(sk, PF_INET, optname, optval, 1542 &len); 1543 release_sock(sk); 1544 if (err >= 0) 1545 err = put_user(len, optlen); 1546 return err; 1547 } 1548 #endif 1549 return err; 1550 } 1551 EXPORT_SYMBOL(ip_getsockopt); 1552 1553 #ifdef CONFIG_COMPAT 1554 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1555 char __user *optval, int __user *optlen) 1556 { 1557 int err; 1558 1559 if (optname == MCAST_MSFILTER) 1560 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1561 ip_getsockopt); 1562 1563 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1564 MSG_CMSG_COMPAT); 1565 1566 #ifdef CONFIG_NETFILTER 1567 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1568 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1569 !ip_mroute_opt(optname)) { 1570 int len; 1571 1572 if (get_user(len, optlen)) 1573 return -EFAULT; 1574 1575 lock_sock(sk); 1576 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1577 release_sock(sk); 1578 if (err >= 0) 1579 err = put_user(len, optlen); 1580 return err; 1581 } 1582 #endif 1583 return err; 1584 } 1585 EXPORT_SYMBOL(compat_ip_getsockopt); 1586 #endif 1587