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