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 <linux/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_fragsize(struct msghdr *msg, struct sk_buff *skb) 101 { 102 int val; 103 104 if (IPCB(skb)->frag_max_size == 0) 105 return; 106 107 val = IPCB(skb)->frag_max_size; 108 put_cmsg(msg, SOL_IP, IP_RECVFRAGSIZE, sizeof(val), &val); 109 } 110 111 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb, 112 int tlen, int offset) 113 { 114 __wsum csum = skb->csum; 115 116 if (skb->ip_summed != CHECKSUM_COMPLETE) 117 return; 118 119 if (offset != 0) 120 csum = csum_sub(csum, 121 csum_partial(skb_transport_header(skb) + tlen, 122 offset, 0)); 123 124 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum); 125 } 126 127 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb) 128 { 129 char *secdata; 130 u32 seclen, secid; 131 int err; 132 133 err = security_socket_getpeersec_dgram(NULL, skb, &secid); 134 if (err) 135 return; 136 137 err = security_secid_to_secctx(secid, &secdata, &seclen); 138 if (err) 139 return; 140 141 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata); 142 security_release_secctx(secdata, seclen); 143 } 144 145 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb) 146 { 147 struct sockaddr_in sin; 148 const struct iphdr *iph = ip_hdr(skb); 149 __be16 *ports = (__be16 *)skb_transport_header(skb); 150 151 if (skb_transport_offset(skb) + 4 > (int)skb->len) 152 return; 153 154 /* All current transport protocols have the port numbers in the 155 * first four bytes of the transport header and this function is 156 * written with this assumption in mind. 157 */ 158 159 sin.sin_family = AF_INET; 160 sin.sin_addr.s_addr = iph->daddr; 161 sin.sin_port = ports[1]; 162 memset(sin.sin_zero, 0, sizeof(sin.sin_zero)); 163 164 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin); 165 } 166 167 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk, 168 struct sk_buff *skb, int tlen, int offset) 169 { 170 struct inet_sock *inet = inet_sk(sk); 171 unsigned int flags = inet->cmsg_flags; 172 173 /* Ordered by supposed usage frequency */ 174 if (flags & IP_CMSG_PKTINFO) { 175 ip_cmsg_recv_pktinfo(msg, skb); 176 177 flags &= ~IP_CMSG_PKTINFO; 178 if (!flags) 179 return; 180 } 181 182 if (flags & IP_CMSG_TTL) { 183 ip_cmsg_recv_ttl(msg, skb); 184 185 flags &= ~IP_CMSG_TTL; 186 if (!flags) 187 return; 188 } 189 190 if (flags & IP_CMSG_TOS) { 191 ip_cmsg_recv_tos(msg, skb); 192 193 flags &= ~IP_CMSG_TOS; 194 if (!flags) 195 return; 196 } 197 198 if (flags & IP_CMSG_RECVOPTS) { 199 ip_cmsg_recv_opts(msg, skb); 200 201 flags &= ~IP_CMSG_RECVOPTS; 202 if (!flags) 203 return; 204 } 205 206 if (flags & IP_CMSG_RETOPTS) { 207 ip_cmsg_recv_retopts(msg, skb); 208 209 flags &= ~IP_CMSG_RETOPTS; 210 if (!flags) 211 return; 212 } 213 214 if (flags & IP_CMSG_PASSSEC) { 215 ip_cmsg_recv_security(msg, skb); 216 217 flags &= ~IP_CMSG_PASSSEC; 218 if (!flags) 219 return; 220 } 221 222 if (flags & IP_CMSG_ORIGDSTADDR) { 223 ip_cmsg_recv_dstaddr(msg, skb); 224 225 flags &= ~IP_CMSG_ORIGDSTADDR; 226 if (!flags) 227 return; 228 } 229 230 if (flags & IP_CMSG_CHECKSUM) 231 ip_cmsg_recv_checksum(msg, skb, tlen, offset); 232 233 if (flags & IP_CMSG_RECVFRAGSIZE) 234 ip_cmsg_recv_fragsize(msg, skb); 235 } 236 EXPORT_SYMBOL(ip_cmsg_recv_offset); 237 238 int ip_cmsg_send(struct sock *sk, struct msghdr *msg, struct ipcm_cookie *ipc, 239 bool allow_ipv6) 240 { 241 int err, val; 242 struct cmsghdr *cmsg; 243 struct net *net = sock_net(sk); 244 245 for_each_cmsghdr(cmsg, msg) { 246 if (!CMSG_OK(msg, cmsg)) 247 return -EINVAL; 248 #if IS_ENABLED(CONFIG_IPV6) 249 if (allow_ipv6 && 250 cmsg->cmsg_level == SOL_IPV6 && 251 cmsg->cmsg_type == IPV6_PKTINFO) { 252 struct in6_pktinfo *src_info; 253 254 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info))) 255 return -EINVAL; 256 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg); 257 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr)) 258 return -EINVAL; 259 ipc->oif = src_info->ipi6_ifindex; 260 ipc->addr = src_info->ipi6_addr.s6_addr32[3]; 261 continue; 262 } 263 #endif 264 if (cmsg->cmsg_level == SOL_SOCKET) { 265 err = __sock_cmsg_send(sk, msg, cmsg, &ipc->sockc); 266 if (err) 267 return err; 268 continue; 269 } 270 271 if (cmsg->cmsg_level != SOL_IP) 272 continue; 273 switch (cmsg->cmsg_type) { 274 case IP_RETOPTS: 275 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr)); 276 277 /* Our caller is responsible for freeing ipc->opt */ 278 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), 279 err < 40 ? err : 40); 280 if (err) 281 return err; 282 break; 283 case IP_PKTINFO: 284 { 285 struct in_pktinfo *info; 286 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) 287 return -EINVAL; 288 info = (struct in_pktinfo *)CMSG_DATA(cmsg); 289 ipc->oif = info->ipi_ifindex; 290 ipc->addr = info->ipi_spec_dst.s_addr; 291 break; 292 } 293 case IP_TTL: 294 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 295 return -EINVAL; 296 val = *(int *)CMSG_DATA(cmsg); 297 if (val < 1 || val > 255) 298 return -EINVAL; 299 ipc->ttl = val; 300 break; 301 case IP_TOS: 302 if (cmsg->cmsg_len == CMSG_LEN(sizeof(int))) 303 val = *(int *)CMSG_DATA(cmsg); 304 else if (cmsg->cmsg_len == CMSG_LEN(sizeof(u8))) 305 val = *(u8 *)CMSG_DATA(cmsg); 306 else 307 return -EINVAL; 308 if (val < 0 || val > 255) 309 return -EINVAL; 310 ipc->tos = val; 311 ipc->priority = rt_tos2priority(ipc->tos); 312 break; 313 314 default: 315 return -EINVAL; 316 } 317 } 318 return 0; 319 } 320 321 322 /* Special input handler for packets caught by router alert option. 323 They are selected only by protocol field, and then processed likely 324 local ones; but only if someone wants them! Otherwise, router 325 not running rsvpd will kill RSVP. 326 327 It is user level problem, what it will make with them. 328 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)), 329 but receiver should be enough clever f.e. to forward mtrace requests, 330 sent to multicast group to reach destination designated router. 331 */ 332 struct ip_ra_chain __rcu *ip_ra_chain; 333 static DEFINE_SPINLOCK(ip_ra_lock); 334 335 336 static void ip_ra_destroy_rcu(struct rcu_head *head) 337 { 338 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu); 339 340 sock_put(ra->saved_sk); 341 kfree(ra); 342 } 343 344 int ip_ra_control(struct sock *sk, unsigned char on, 345 void (*destructor)(struct sock *)) 346 { 347 struct ip_ra_chain *ra, *new_ra; 348 struct ip_ra_chain __rcu **rap; 349 350 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 351 return -EINVAL; 352 353 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 354 355 spin_lock_bh(&ip_ra_lock); 356 for (rap = &ip_ra_chain; 357 (ra = rcu_dereference_protected(*rap, 358 lockdep_is_held(&ip_ra_lock))) != NULL; 359 rap = &ra->next) { 360 if (ra->sk == sk) { 361 if (on) { 362 spin_unlock_bh(&ip_ra_lock); 363 kfree(new_ra); 364 return -EADDRINUSE; 365 } 366 /* dont let ip_call_ra_chain() use sk again */ 367 ra->sk = NULL; 368 RCU_INIT_POINTER(*rap, ra->next); 369 spin_unlock_bh(&ip_ra_lock); 370 371 if (ra->destructor) 372 ra->destructor(sk); 373 /* 374 * Delay sock_put(sk) and kfree(ra) after one rcu grace 375 * period. This guarantee ip_call_ra_chain() dont need 376 * to mess with socket refcounts. 377 */ 378 ra->saved_sk = sk; 379 call_rcu(&ra->rcu, ip_ra_destroy_rcu); 380 return 0; 381 } 382 } 383 if (!new_ra) { 384 spin_unlock_bh(&ip_ra_lock); 385 return -ENOBUFS; 386 } 387 new_ra->sk = sk; 388 new_ra->destructor = destructor; 389 390 RCU_INIT_POINTER(new_ra->next, ra); 391 rcu_assign_pointer(*rap, new_ra); 392 sock_hold(sk); 393 spin_unlock_bh(&ip_ra_lock); 394 395 return 0; 396 } 397 398 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 399 __be16 port, u32 info, u8 *payload) 400 { 401 struct sock_exterr_skb *serr; 402 403 skb = skb_clone(skb, GFP_ATOMIC); 404 if (!skb) 405 return; 406 407 serr = SKB_EXT_ERR(skb); 408 serr->ee.ee_errno = err; 409 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 410 serr->ee.ee_type = icmp_hdr(skb)->type; 411 serr->ee.ee_code = icmp_hdr(skb)->code; 412 serr->ee.ee_pad = 0; 413 serr->ee.ee_info = info; 414 serr->ee.ee_data = 0; 415 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 416 skb_network_header(skb); 417 serr->port = port; 418 419 if (skb_pull(skb, payload - skb->data)) { 420 skb_reset_transport_header(skb); 421 if (sock_queue_err_skb(sk, skb) == 0) 422 return; 423 } 424 kfree_skb(skb); 425 } 426 427 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 428 { 429 struct inet_sock *inet = inet_sk(sk); 430 struct sock_exterr_skb *serr; 431 struct iphdr *iph; 432 struct sk_buff *skb; 433 434 if (!inet->recverr) 435 return; 436 437 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 438 if (!skb) 439 return; 440 441 skb_put(skb, sizeof(struct iphdr)); 442 skb_reset_network_header(skb); 443 iph = ip_hdr(skb); 444 iph->daddr = daddr; 445 446 serr = SKB_EXT_ERR(skb); 447 serr->ee.ee_errno = err; 448 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 449 serr->ee.ee_type = 0; 450 serr->ee.ee_code = 0; 451 serr->ee.ee_pad = 0; 452 serr->ee.ee_info = info; 453 serr->ee.ee_data = 0; 454 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 455 serr->port = port; 456 457 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 458 skb_reset_transport_header(skb); 459 460 if (sock_queue_err_skb(sk, skb)) 461 kfree_skb(skb); 462 } 463 464 /* For some errors we have valid addr_offset even with zero payload and 465 * zero port. Also, addr_offset should be supported if port is set. 466 */ 467 static inline bool ipv4_datagram_support_addr(struct sock_exterr_skb *serr) 468 { 469 return serr->ee.ee_origin == SO_EE_ORIGIN_ICMP || 470 serr->ee.ee_origin == SO_EE_ORIGIN_LOCAL || serr->port; 471 } 472 473 /* IPv4 supports cmsg on all imcp errors and some timestamps 474 * 475 * Timestamp code paths do not initialize the fields expected by cmsg: 476 * the PKTINFO fields in skb->cb[]. Fill those in here. 477 */ 478 static bool ipv4_datagram_support_cmsg(const struct sock *sk, 479 struct sk_buff *skb, 480 int ee_origin) 481 { 482 struct in_pktinfo *info; 483 484 if (ee_origin == SO_EE_ORIGIN_ICMP) 485 return true; 486 487 if (ee_origin == SO_EE_ORIGIN_LOCAL) 488 return false; 489 490 /* Support IP_PKTINFO on tstamp packets if requested, to correlate 491 * timestamp with egress dev. Not possible for packets without dev 492 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY). 493 */ 494 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) || 495 (!skb->dev)) 496 return false; 497 498 info = PKTINFO_SKB_CB(skb); 499 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr; 500 info->ipi_ifindex = skb->dev->ifindex; 501 return true; 502 } 503 504 /* 505 * Handle MSG_ERRQUEUE 506 */ 507 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 508 { 509 struct sock_exterr_skb *serr; 510 struct sk_buff *skb; 511 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 512 struct { 513 struct sock_extended_err ee; 514 struct sockaddr_in offender; 515 } errhdr; 516 int err; 517 int copied; 518 519 WARN_ON_ONCE(sk->sk_family == AF_INET6); 520 521 err = -EAGAIN; 522 skb = sock_dequeue_err_skb(sk); 523 if (!skb) 524 goto out; 525 526 copied = skb->len; 527 if (copied > len) { 528 msg->msg_flags |= MSG_TRUNC; 529 copied = len; 530 } 531 err = skb_copy_datagram_msg(skb, 0, msg, copied); 532 if (unlikely(err)) { 533 kfree_skb(skb); 534 return err; 535 } 536 sock_recv_timestamp(msg, sk, skb); 537 538 serr = SKB_EXT_ERR(skb); 539 540 if (sin && ipv4_datagram_support_addr(serr)) { 541 sin->sin_family = AF_INET; 542 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 543 serr->addr_offset); 544 sin->sin_port = serr->port; 545 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 546 *addr_len = sizeof(*sin); 547 } 548 549 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 550 sin = &errhdr.offender; 551 memset(sin, 0, sizeof(*sin)); 552 553 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) { 554 sin->sin_family = AF_INET; 555 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 556 if (inet_sk(sk)->cmsg_flags) 557 ip_cmsg_recv(msg, skb); 558 } 559 560 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 561 562 /* Now we could try to dump offended packet options */ 563 564 msg->msg_flags |= MSG_ERRQUEUE; 565 err = copied; 566 567 consume_skb(skb); 568 out: 569 return err; 570 } 571 572 573 /* 574 * Socket option code for IP. This is the end of the line after any 575 * TCP,UDP etc options on an IP socket. 576 */ 577 static bool setsockopt_needs_rtnl(int optname) 578 { 579 switch (optname) { 580 case IP_ADD_MEMBERSHIP: 581 case IP_ADD_SOURCE_MEMBERSHIP: 582 case IP_BLOCK_SOURCE: 583 case IP_DROP_MEMBERSHIP: 584 case IP_DROP_SOURCE_MEMBERSHIP: 585 case IP_MSFILTER: 586 case IP_UNBLOCK_SOURCE: 587 case MCAST_BLOCK_SOURCE: 588 case MCAST_MSFILTER: 589 case MCAST_JOIN_GROUP: 590 case MCAST_JOIN_SOURCE_GROUP: 591 case MCAST_LEAVE_GROUP: 592 case MCAST_LEAVE_SOURCE_GROUP: 593 case MCAST_UNBLOCK_SOURCE: 594 return true; 595 } 596 return false; 597 } 598 599 static int do_ip_setsockopt(struct sock *sk, int level, 600 int optname, char __user *optval, unsigned int optlen) 601 { 602 struct inet_sock *inet = inet_sk(sk); 603 struct net *net = sock_net(sk); 604 int val = 0, err; 605 bool needs_rtnl = setsockopt_needs_rtnl(optname); 606 607 switch (optname) { 608 case IP_PKTINFO: 609 case IP_RECVTTL: 610 case IP_RECVOPTS: 611 case IP_RECVTOS: 612 case IP_RETOPTS: 613 case IP_TOS: 614 case IP_TTL: 615 case IP_HDRINCL: 616 case IP_MTU_DISCOVER: 617 case IP_RECVERR: 618 case IP_ROUTER_ALERT: 619 case IP_FREEBIND: 620 case IP_PASSSEC: 621 case IP_TRANSPARENT: 622 case IP_MINTTL: 623 case IP_NODEFRAG: 624 case IP_BIND_ADDRESS_NO_PORT: 625 case IP_UNICAST_IF: 626 case IP_MULTICAST_TTL: 627 case IP_MULTICAST_ALL: 628 case IP_MULTICAST_LOOP: 629 case IP_RECVORIGDSTADDR: 630 case IP_CHECKSUM: 631 case IP_RECVFRAGSIZE: 632 if (optlen >= sizeof(int)) { 633 if (get_user(val, (int __user *) optval)) 634 return -EFAULT; 635 } else if (optlen >= sizeof(char)) { 636 unsigned char ucval; 637 638 if (get_user(ucval, (unsigned char __user *) optval)) 639 return -EFAULT; 640 val = (int) ucval; 641 } 642 } 643 644 /* If optlen==0, it is equivalent to val == 0 */ 645 646 if (ip_mroute_opt(optname)) 647 return ip_mroute_setsockopt(sk, optname, optval, optlen); 648 649 err = 0; 650 if (needs_rtnl) 651 rtnl_lock(); 652 lock_sock(sk); 653 654 switch (optname) { 655 case IP_OPTIONS: 656 { 657 struct ip_options_rcu *old, *opt = NULL; 658 659 if (optlen > 40) 660 goto e_inval; 661 err = ip_options_get_from_user(sock_net(sk), &opt, 662 optval, optlen); 663 if (err) 664 break; 665 old = rcu_dereference_protected(inet->inet_opt, 666 lockdep_sock_is_held(sk)); 667 if (inet->is_icsk) { 668 struct inet_connection_sock *icsk = inet_csk(sk); 669 #if IS_ENABLED(CONFIG_IPV6) 670 if (sk->sk_family == PF_INET || 671 (!((1 << sk->sk_state) & 672 (TCPF_LISTEN | TCPF_CLOSE)) && 673 inet->inet_daddr != LOOPBACK4_IPV6)) { 674 #endif 675 if (old) 676 icsk->icsk_ext_hdr_len -= old->opt.optlen; 677 if (opt) 678 icsk->icsk_ext_hdr_len += opt->opt.optlen; 679 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 680 #if IS_ENABLED(CONFIG_IPV6) 681 } 682 #endif 683 } 684 rcu_assign_pointer(inet->inet_opt, opt); 685 if (old) 686 kfree_rcu(old, rcu); 687 break; 688 } 689 case IP_PKTINFO: 690 if (val) 691 inet->cmsg_flags |= IP_CMSG_PKTINFO; 692 else 693 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 694 break; 695 case IP_RECVTTL: 696 if (val) 697 inet->cmsg_flags |= IP_CMSG_TTL; 698 else 699 inet->cmsg_flags &= ~IP_CMSG_TTL; 700 break; 701 case IP_RECVTOS: 702 if (val) 703 inet->cmsg_flags |= IP_CMSG_TOS; 704 else 705 inet->cmsg_flags &= ~IP_CMSG_TOS; 706 break; 707 case IP_RECVOPTS: 708 if (val) 709 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 710 else 711 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 712 break; 713 case IP_RETOPTS: 714 if (val) 715 inet->cmsg_flags |= IP_CMSG_RETOPTS; 716 else 717 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 718 break; 719 case IP_PASSSEC: 720 if (val) 721 inet->cmsg_flags |= IP_CMSG_PASSSEC; 722 else 723 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 724 break; 725 case IP_RECVORIGDSTADDR: 726 if (val) 727 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 728 else 729 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 730 break; 731 case IP_CHECKSUM: 732 if (val) { 733 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) { 734 inet_inc_convert_csum(sk); 735 inet->cmsg_flags |= IP_CMSG_CHECKSUM; 736 } 737 } else { 738 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) { 739 inet_dec_convert_csum(sk); 740 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM; 741 } 742 } 743 break; 744 case IP_RECVFRAGSIZE: 745 if (sk->sk_type != SOCK_RAW && sk->sk_type != SOCK_DGRAM) 746 goto e_inval; 747 if (val) 748 inet->cmsg_flags |= IP_CMSG_RECVFRAGSIZE; 749 else 750 inet->cmsg_flags &= ~IP_CMSG_RECVFRAGSIZE; 751 break; 752 case IP_TOS: /* This sets both TOS and Precedence */ 753 if (sk->sk_type == SOCK_STREAM) { 754 val &= ~INET_ECN_MASK; 755 val |= inet->tos & INET_ECN_MASK; 756 } 757 if (inet->tos != val) { 758 inet->tos = val; 759 sk->sk_priority = rt_tos2priority(val); 760 sk_dst_reset(sk); 761 } 762 break; 763 case IP_TTL: 764 if (optlen < 1) 765 goto e_inval; 766 if (val != -1 && (val < 1 || val > 255)) 767 goto e_inval; 768 inet->uc_ttl = val; 769 break; 770 case IP_HDRINCL: 771 if (sk->sk_type != SOCK_RAW) { 772 err = -ENOPROTOOPT; 773 break; 774 } 775 inet->hdrincl = val ? 1 : 0; 776 break; 777 case IP_NODEFRAG: 778 if (sk->sk_type != SOCK_RAW) { 779 err = -ENOPROTOOPT; 780 break; 781 } 782 inet->nodefrag = val ? 1 : 0; 783 break; 784 case IP_BIND_ADDRESS_NO_PORT: 785 inet->bind_address_no_port = val ? 1 : 0; 786 break; 787 case IP_MTU_DISCOVER: 788 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT) 789 goto e_inval; 790 inet->pmtudisc = val; 791 break; 792 case IP_RECVERR: 793 inet->recverr = !!val; 794 if (!val) 795 skb_queue_purge(&sk->sk_error_queue); 796 break; 797 case IP_MULTICAST_TTL: 798 if (sk->sk_type == SOCK_STREAM) 799 goto e_inval; 800 if (optlen < 1) 801 goto e_inval; 802 if (val == -1) 803 val = 1; 804 if (val < 0 || val > 255) 805 goto e_inval; 806 inet->mc_ttl = val; 807 break; 808 case IP_MULTICAST_LOOP: 809 if (optlen < 1) 810 goto e_inval; 811 inet->mc_loop = !!val; 812 break; 813 case IP_UNICAST_IF: 814 { 815 struct net_device *dev = NULL; 816 int ifindex; 817 818 if (optlen != sizeof(int)) 819 goto e_inval; 820 821 ifindex = (__force int)ntohl((__force __be32)val); 822 if (ifindex == 0) { 823 inet->uc_index = 0; 824 err = 0; 825 break; 826 } 827 828 dev = dev_get_by_index(sock_net(sk), ifindex); 829 err = -EADDRNOTAVAIL; 830 if (!dev) 831 break; 832 dev_put(dev); 833 834 err = -EINVAL; 835 if (sk->sk_bound_dev_if) 836 break; 837 838 inet->uc_index = ifindex; 839 err = 0; 840 break; 841 } 842 case IP_MULTICAST_IF: 843 { 844 struct ip_mreqn mreq; 845 struct net_device *dev = NULL; 846 847 if (sk->sk_type == SOCK_STREAM) 848 goto e_inval; 849 /* 850 * Check the arguments are allowable 851 */ 852 853 if (optlen < sizeof(struct in_addr)) 854 goto e_inval; 855 856 err = -EFAULT; 857 if (optlen >= sizeof(struct ip_mreqn)) { 858 if (copy_from_user(&mreq, optval, sizeof(mreq))) 859 break; 860 } else { 861 memset(&mreq, 0, sizeof(mreq)); 862 if (optlen >= sizeof(struct ip_mreq)) { 863 if (copy_from_user(&mreq, optval, 864 sizeof(struct ip_mreq))) 865 break; 866 } else if (optlen >= sizeof(struct in_addr)) { 867 if (copy_from_user(&mreq.imr_address, optval, 868 sizeof(struct in_addr))) 869 break; 870 } 871 } 872 873 if (!mreq.imr_ifindex) { 874 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 875 inet->mc_index = 0; 876 inet->mc_addr = 0; 877 err = 0; 878 break; 879 } 880 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 881 if (dev) 882 mreq.imr_ifindex = dev->ifindex; 883 } else 884 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 885 886 887 err = -EADDRNOTAVAIL; 888 if (!dev) 889 break; 890 dev_put(dev); 891 892 err = -EINVAL; 893 if (sk->sk_bound_dev_if && 894 mreq.imr_ifindex != sk->sk_bound_dev_if) 895 break; 896 897 inet->mc_index = mreq.imr_ifindex; 898 inet->mc_addr = mreq.imr_address.s_addr; 899 err = 0; 900 break; 901 } 902 903 case IP_ADD_MEMBERSHIP: 904 case IP_DROP_MEMBERSHIP: 905 { 906 struct ip_mreqn mreq; 907 908 err = -EPROTO; 909 if (inet_sk(sk)->is_icsk) 910 break; 911 912 if (optlen < sizeof(struct ip_mreq)) 913 goto e_inval; 914 err = -EFAULT; 915 if (optlen >= sizeof(struct ip_mreqn)) { 916 if (copy_from_user(&mreq, optval, sizeof(mreq))) 917 break; 918 } else { 919 memset(&mreq, 0, sizeof(mreq)); 920 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 921 break; 922 } 923 924 if (optname == IP_ADD_MEMBERSHIP) 925 err = ip_mc_join_group(sk, &mreq); 926 else 927 err = ip_mc_leave_group(sk, &mreq); 928 break; 929 } 930 case IP_MSFILTER: 931 { 932 struct ip_msfilter *msf; 933 934 if (optlen < IP_MSFILTER_SIZE(0)) 935 goto e_inval; 936 if (optlen > sysctl_optmem_max) { 937 err = -ENOBUFS; 938 break; 939 } 940 msf = kmalloc(optlen, GFP_KERNEL); 941 if (!msf) { 942 err = -ENOBUFS; 943 break; 944 } 945 err = -EFAULT; 946 if (copy_from_user(msf, optval, optlen)) { 947 kfree(msf); 948 break; 949 } 950 /* numsrc >= (1G-4) overflow in 32 bits */ 951 if (msf->imsf_numsrc >= 0x3ffffffcU || 952 msf->imsf_numsrc > net->ipv4.sysctl_igmp_max_msf) { 953 kfree(msf); 954 err = -ENOBUFS; 955 break; 956 } 957 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 958 kfree(msf); 959 err = -EINVAL; 960 break; 961 } 962 err = ip_mc_msfilter(sk, msf, 0); 963 kfree(msf); 964 break; 965 } 966 case IP_BLOCK_SOURCE: 967 case IP_UNBLOCK_SOURCE: 968 case IP_ADD_SOURCE_MEMBERSHIP: 969 case IP_DROP_SOURCE_MEMBERSHIP: 970 { 971 struct ip_mreq_source mreqs; 972 int omode, add; 973 974 if (optlen != sizeof(struct ip_mreq_source)) 975 goto e_inval; 976 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 977 err = -EFAULT; 978 break; 979 } 980 if (optname == IP_BLOCK_SOURCE) { 981 omode = MCAST_EXCLUDE; 982 add = 1; 983 } else if (optname == IP_UNBLOCK_SOURCE) { 984 omode = MCAST_EXCLUDE; 985 add = 0; 986 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 987 struct ip_mreqn mreq; 988 989 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 990 mreq.imr_address.s_addr = mreqs.imr_interface; 991 mreq.imr_ifindex = 0; 992 err = ip_mc_join_group(sk, &mreq); 993 if (err && err != -EADDRINUSE) 994 break; 995 omode = MCAST_INCLUDE; 996 add = 1; 997 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 998 omode = MCAST_INCLUDE; 999 add = 0; 1000 } 1001 err = ip_mc_source(add, omode, sk, &mreqs, 0); 1002 break; 1003 } 1004 case MCAST_JOIN_GROUP: 1005 case MCAST_LEAVE_GROUP: 1006 { 1007 struct group_req greq; 1008 struct sockaddr_in *psin; 1009 struct ip_mreqn mreq; 1010 1011 if (optlen < sizeof(struct group_req)) 1012 goto e_inval; 1013 err = -EFAULT; 1014 if (copy_from_user(&greq, optval, sizeof(greq))) 1015 break; 1016 psin = (struct sockaddr_in *)&greq.gr_group; 1017 if (psin->sin_family != AF_INET) 1018 goto e_inval; 1019 memset(&mreq, 0, sizeof(mreq)); 1020 mreq.imr_multiaddr = psin->sin_addr; 1021 mreq.imr_ifindex = greq.gr_interface; 1022 1023 if (optname == MCAST_JOIN_GROUP) 1024 err = ip_mc_join_group(sk, &mreq); 1025 else 1026 err = ip_mc_leave_group(sk, &mreq); 1027 break; 1028 } 1029 case MCAST_JOIN_SOURCE_GROUP: 1030 case MCAST_LEAVE_SOURCE_GROUP: 1031 case MCAST_BLOCK_SOURCE: 1032 case MCAST_UNBLOCK_SOURCE: 1033 { 1034 struct group_source_req greqs; 1035 struct ip_mreq_source mreqs; 1036 struct sockaddr_in *psin; 1037 int omode, add; 1038 1039 if (optlen != sizeof(struct group_source_req)) 1040 goto e_inval; 1041 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 1042 err = -EFAULT; 1043 break; 1044 } 1045 if (greqs.gsr_group.ss_family != AF_INET || 1046 greqs.gsr_source.ss_family != AF_INET) { 1047 err = -EADDRNOTAVAIL; 1048 break; 1049 } 1050 psin = (struct sockaddr_in *)&greqs.gsr_group; 1051 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 1052 psin = (struct sockaddr_in *)&greqs.gsr_source; 1053 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 1054 mreqs.imr_interface = 0; /* use index for mc_source */ 1055 1056 if (optname == MCAST_BLOCK_SOURCE) { 1057 omode = MCAST_EXCLUDE; 1058 add = 1; 1059 } else if (optname == MCAST_UNBLOCK_SOURCE) { 1060 omode = MCAST_EXCLUDE; 1061 add = 0; 1062 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 1063 struct ip_mreqn mreq; 1064 1065 psin = (struct sockaddr_in *)&greqs.gsr_group; 1066 mreq.imr_multiaddr = psin->sin_addr; 1067 mreq.imr_address.s_addr = 0; 1068 mreq.imr_ifindex = greqs.gsr_interface; 1069 err = ip_mc_join_group(sk, &mreq); 1070 if (err && err != -EADDRINUSE) 1071 break; 1072 greqs.gsr_interface = mreq.imr_ifindex; 1073 omode = MCAST_INCLUDE; 1074 add = 1; 1075 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 1076 omode = MCAST_INCLUDE; 1077 add = 0; 1078 } 1079 err = ip_mc_source(add, omode, sk, &mreqs, 1080 greqs.gsr_interface); 1081 break; 1082 } 1083 case MCAST_MSFILTER: 1084 { 1085 struct sockaddr_in *psin; 1086 struct ip_msfilter *msf = NULL; 1087 struct group_filter *gsf = NULL; 1088 int msize, i, ifindex; 1089 1090 if (optlen < GROUP_FILTER_SIZE(0)) 1091 goto e_inval; 1092 if (optlen > sysctl_optmem_max) { 1093 err = -ENOBUFS; 1094 break; 1095 } 1096 gsf = kmalloc(optlen, GFP_KERNEL); 1097 if (!gsf) { 1098 err = -ENOBUFS; 1099 break; 1100 } 1101 err = -EFAULT; 1102 if (copy_from_user(gsf, optval, optlen)) 1103 goto mc_msf_out; 1104 1105 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 1106 if (gsf->gf_numsrc >= 0x1ffffff || 1107 gsf->gf_numsrc > net->ipv4.sysctl_igmp_max_msf) { 1108 err = -ENOBUFS; 1109 goto mc_msf_out; 1110 } 1111 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 1112 err = -EINVAL; 1113 goto mc_msf_out; 1114 } 1115 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 1116 msf = kmalloc(msize, GFP_KERNEL); 1117 if (!msf) { 1118 err = -ENOBUFS; 1119 goto mc_msf_out; 1120 } 1121 ifindex = gsf->gf_interface; 1122 psin = (struct sockaddr_in *)&gsf->gf_group; 1123 if (psin->sin_family != AF_INET) { 1124 err = -EADDRNOTAVAIL; 1125 goto mc_msf_out; 1126 } 1127 msf->imsf_multiaddr = psin->sin_addr.s_addr; 1128 msf->imsf_interface = 0; 1129 msf->imsf_fmode = gsf->gf_fmode; 1130 msf->imsf_numsrc = gsf->gf_numsrc; 1131 err = -EADDRNOTAVAIL; 1132 for (i = 0; i < gsf->gf_numsrc; ++i) { 1133 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 1134 1135 if (psin->sin_family != AF_INET) 1136 goto mc_msf_out; 1137 msf->imsf_slist[i] = psin->sin_addr.s_addr; 1138 } 1139 kfree(gsf); 1140 gsf = NULL; 1141 1142 err = ip_mc_msfilter(sk, msf, ifindex); 1143 mc_msf_out: 1144 kfree(msf); 1145 kfree(gsf); 1146 break; 1147 } 1148 case IP_MULTICAST_ALL: 1149 if (optlen < 1) 1150 goto e_inval; 1151 if (val != 0 && val != 1) 1152 goto e_inval; 1153 inet->mc_all = val; 1154 break; 1155 case IP_ROUTER_ALERT: 1156 err = ip_ra_control(sk, val ? 1 : 0, NULL); 1157 break; 1158 1159 case IP_FREEBIND: 1160 if (optlen < 1) 1161 goto e_inval; 1162 inet->freebind = !!val; 1163 break; 1164 1165 case IP_IPSEC_POLICY: 1166 case IP_XFRM_POLICY: 1167 err = -EPERM; 1168 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 1169 break; 1170 err = xfrm_user_policy(sk, optname, optval, optlen); 1171 break; 1172 1173 case IP_TRANSPARENT: 1174 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) && 1175 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) { 1176 err = -EPERM; 1177 break; 1178 } 1179 if (optlen < 1) 1180 goto e_inval; 1181 inet->transparent = !!val; 1182 break; 1183 1184 case IP_MINTTL: 1185 if (optlen < 1) 1186 goto e_inval; 1187 if (val < 0 || val > 255) 1188 goto e_inval; 1189 inet->min_ttl = val; 1190 break; 1191 1192 default: 1193 err = -ENOPROTOOPT; 1194 break; 1195 } 1196 release_sock(sk); 1197 if (needs_rtnl) 1198 rtnl_unlock(); 1199 return err; 1200 1201 e_inval: 1202 release_sock(sk); 1203 if (needs_rtnl) 1204 rtnl_unlock(); 1205 return -EINVAL; 1206 } 1207 1208 /** 1209 * ipv4_pktinfo_prepare - transfer some info from rtable to skb 1210 * @sk: socket 1211 * @skb: buffer 1212 * 1213 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific 1214 * destination in skb->cb[] before dst drop. 1215 * This way, receiver doesn't make cache line misses to read rtable. 1216 */ 1217 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb) 1218 { 1219 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb); 1220 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) || 1221 ipv6_sk_rxinfo(sk); 1222 1223 if (prepare && skb_rtable(skb)) { 1224 /* skb->cb is overloaded: prior to this point it is IP{6}CB 1225 * which has interface index (iif) as the first member of the 1226 * underlying inet{6}_skb_parm struct. This code then overlays 1227 * PKTINFO_SKB_CB and in_pktinfo also has iif as the first 1228 * element so the iif is picked up from the prior IPCB. If iif 1229 * is the loopback interface, then return the sending interface 1230 * (e.g., process binds socket to eth0 for Tx which is 1231 * redirected to loopback in the rtable/dst). 1232 */ 1233 if (pktinfo->ipi_ifindex == LOOPBACK_IFINDEX) 1234 pktinfo->ipi_ifindex = inet_iif(skb); 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 lock_sock(sk); 1260 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1261 release_sock(sk); 1262 } 1263 #endif 1264 return err; 1265 } 1266 EXPORT_SYMBOL(ip_setsockopt); 1267 1268 #ifdef CONFIG_COMPAT 1269 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1270 char __user *optval, unsigned int optlen) 1271 { 1272 int err; 1273 1274 if (level != SOL_IP) 1275 return -ENOPROTOOPT; 1276 1277 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1278 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1279 ip_setsockopt); 1280 1281 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1282 #ifdef CONFIG_NETFILTER 1283 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1284 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1285 optname != IP_IPSEC_POLICY && 1286 optname != IP_XFRM_POLICY && 1287 !ip_mroute_opt(optname)) { 1288 lock_sock(sk); 1289 err = compat_nf_setsockopt(sk, PF_INET, optname, 1290 optval, optlen); 1291 release_sock(sk); 1292 } 1293 #endif 1294 return err; 1295 } 1296 EXPORT_SYMBOL(compat_ip_setsockopt); 1297 #endif 1298 1299 /* 1300 * Get the options. Note for future reference. The GET of IP options gets 1301 * the _received_ ones. The set sets the _sent_ ones. 1302 */ 1303 1304 static bool getsockopt_needs_rtnl(int optname) 1305 { 1306 switch (optname) { 1307 case IP_MSFILTER: 1308 case MCAST_MSFILTER: 1309 return true; 1310 } 1311 return false; 1312 } 1313 1314 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1315 char __user *optval, int __user *optlen, unsigned int flags) 1316 { 1317 struct inet_sock *inet = inet_sk(sk); 1318 bool needs_rtnl = getsockopt_needs_rtnl(optname); 1319 int val, err = 0; 1320 int len; 1321 1322 if (level != SOL_IP) 1323 return -EOPNOTSUPP; 1324 1325 if (ip_mroute_opt(optname)) 1326 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1327 1328 if (get_user(len, optlen)) 1329 return -EFAULT; 1330 if (len < 0) 1331 return -EINVAL; 1332 1333 if (needs_rtnl) 1334 rtnl_lock(); 1335 lock_sock(sk); 1336 1337 switch (optname) { 1338 case IP_OPTIONS: 1339 { 1340 unsigned char optbuf[sizeof(struct ip_options)+40]; 1341 struct ip_options *opt = (struct ip_options *)optbuf; 1342 struct ip_options_rcu *inet_opt; 1343 1344 inet_opt = rcu_dereference_protected(inet->inet_opt, 1345 lockdep_sock_is_held(sk)); 1346 opt->optlen = 0; 1347 if (inet_opt) 1348 memcpy(optbuf, &inet_opt->opt, 1349 sizeof(struct ip_options) + 1350 inet_opt->opt.optlen); 1351 release_sock(sk); 1352 1353 if (opt->optlen == 0) 1354 return put_user(0, optlen); 1355 1356 ip_options_undo(opt); 1357 1358 len = min_t(unsigned int, len, opt->optlen); 1359 if (put_user(len, optlen)) 1360 return -EFAULT; 1361 if (copy_to_user(optval, opt->__data, len)) 1362 return -EFAULT; 1363 return 0; 1364 } 1365 case IP_PKTINFO: 1366 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1367 break; 1368 case IP_RECVTTL: 1369 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1370 break; 1371 case IP_RECVTOS: 1372 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1373 break; 1374 case IP_RECVOPTS: 1375 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1376 break; 1377 case IP_RETOPTS: 1378 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1379 break; 1380 case IP_PASSSEC: 1381 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1382 break; 1383 case IP_RECVORIGDSTADDR: 1384 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1385 break; 1386 case IP_CHECKSUM: 1387 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0; 1388 break; 1389 case IP_RECVFRAGSIZE: 1390 val = (inet->cmsg_flags & IP_CMSG_RECVFRAGSIZE) != 0; 1391 break; 1392 case IP_TOS: 1393 val = inet->tos; 1394 break; 1395 case IP_TTL: 1396 { 1397 struct net *net = sock_net(sk); 1398 val = (inet->uc_ttl == -1 ? 1399 net->ipv4.sysctl_ip_default_ttl : 1400 inet->uc_ttl); 1401 break; 1402 } 1403 case IP_HDRINCL: 1404 val = inet->hdrincl; 1405 break; 1406 case IP_NODEFRAG: 1407 val = inet->nodefrag; 1408 break; 1409 case IP_BIND_ADDRESS_NO_PORT: 1410 val = inet->bind_address_no_port; 1411 break; 1412 case IP_MTU_DISCOVER: 1413 val = inet->pmtudisc; 1414 break; 1415 case IP_MTU: 1416 { 1417 struct dst_entry *dst; 1418 val = 0; 1419 dst = sk_dst_get(sk); 1420 if (dst) { 1421 val = dst_mtu(dst); 1422 dst_release(dst); 1423 } 1424 if (!val) { 1425 release_sock(sk); 1426 return -ENOTCONN; 1427 } 1428 break; 1429 } 1430 case IP_RECVERR: 1431 val = inet->recverr; 1432 break; 1433 case IP_MULTICAST_TTL: 1434 val = inet->mc_ttl; 1435 break; 1436 case IP_MULTICAST_LOOP: 1437 val = inet->mc_loop; 1438 break; 1439 case IP_UNICAST_IF: 1440 val = (__force int)htonl((__u32) inet->uc_index); 1441 break; 1442 case IP_MULTICAST_IF: 1443 { 1444 struct in_addr addr; 1445 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1446 addr.s_addr = inet->mc_addr; 1447 release_sock(sk); 1448 1449 if (put_user(len, optlen)) 1450 return -EFAULT; 1451 if (copy_to_user(optval, &addr, len)) 1452 return -EFAULT; 1453 return 0; 1454 } 1455 case IP_MSFILTER: 1456 { 1457 struct ip_msfilter msf; 1458 1459 if (len < IP_MSFILTER_SIZE(0)) { 1460 err = -EINVAL; 1461 goto out; 1462 } 1463 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1464 err = -EFAULT; 1465 goto out; 1466 } 1467 err = ip_mc_msfget(sk, &msf, 1468 (struct ip_msfilter __user *)optval, optlen); 1469 goto out; 1470 } 1471 case MCAST_MSFILTER: 1472 { 1473 struct group_filter gsf; 1474 1475 if (len < GROUP_FILTER_SIZE(0)) { 1476 err = -EINVAL; 1477 goto out; 1478 } 1479 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1480 err = -EFAULT; 1481 goto out; 1482 } 1483 err = ip_mc_gsfget(sk, &gsf, 1484 (struct group_filter __user *)optval, 1485 optlen); 1486 goto out; 1487 } 1488 case IP_MULTICAST_ALL: 1489 val = inet->mc_all; 1490 break; 1491 case IP_PKTOPTIONS: 1492 { 1493 struct msghdr msg; 1494 1495 release_sock(sk); 1496 1497 if (sk->sk_type != SOCK_STREAM) 1498 return -ENOPROTOOPT; 1499 1500 msg.msg_control = (__force void *) optval; 1501 msg.msg_controllen = len; 1502 msg.msg_flags = flags; 1503 1504 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1505 struct in_pktinfo info; 1506 1507 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1508 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1509 info.ipi_ifindex = inet->mc_index; 1510 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1511 } 1512 if (inet->cmsg_flags & IP_CMSG_TTL) { 1513 int hlim = inet->mc_ttl; 1514 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1515 } 1516 if (inet->cmsg_flags & IP_CMSG_TOS) { 1517 int tos = inet->rcv_tos; 1518 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos); 1519 } 1520 len -= msg.msg_controllen; 1521 return put_user(len, optlen); 1522 } 1523 case IP_FREEBIND: 1524 val = inet->freebind; 1525 break; 1526 case IP_TRANSPARENT: 1527 val = inet->transparent; 1528 break; 1529 case IP_MINTTL: 1530 val = inet->min_ttl; 1531 break; 1532 default: 1533 release_sock(sk); 1534 return -ENOPROTOOPT; 1535 } 1536 release_sock(sk); 1537 1538 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1539 unsigned char ucval = (unsigned char)val; 1540 len = 1; 1541 if (put_user(len, optlen)) 1542 return -EFAULT; 1543 if (copy_to_user(optval, &ucval, 1)) 1544 return -EFAULT; 1545 } else { 1546 len = min_t(unsigned int, sizeof(int), len); 1547 if (put_user(len, optlen)) 1548 return -EFAULT; 1549 if (copy_to_user(optval, &val, len)) 1550 return -EFAULT; 1551 } 1552 return 0; 1553 1554 out: 1555 release_sock(sk); 1556 if (needs_rtnl) 1557 rtnl_unlock(); 1558 return err; 1559 } 1560 1561 int ip_getsockopt(struct sock *sk, int level, 1562 int optname, char __user *optval, int __user *optlen) 1563 { 1564 int err; 1565 1566 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1567 #ifdef CONFIG_NETFILTER 1568 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1569 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1570 !ip_mroute_opt(optname)) { 1571 int len; 1572 1573 if (get_user(len, optlen)) 1574 return -EFAULT; 1575 1576 lock_sock(sk); 1577 err = nf_getsockopt(sk, PF_INET, optname, optval, 1578 &len); 1579 release_sock(sk); 1580 if (err >= 0) 1581 err = put_user(len, optlen); 1582 return err; 1583 } 1584 #endif 1585 return err; 1586 } 1587 EXPORT_SYMBOL(ip_getsockopt); 1588 1589 #ifdef CONFIG_COMPAT 1590 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1591 char __user *optval, int __user *optlen) 1592 { 1593 int err; 1594 1595 if (optname == MCAST_MSFILTER) 1596 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1597 ip_getsockopt); 1598 1599 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1600 MSG_CMSG_COMPAT); 1601 1602 #ifdef CONFIG_NETFILTER 1603 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1604 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1605 !ip_mroute_opt(optname)) { 1606 int len; 1607 1608 if (get_user(len, optlen)) 1609 return -EFAULT; 1610 1611 lock_sock(sk); 1612 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1613 release_sock(sk); 1614 if (err >= 0) 1615 err = put_user(len, optlen); 1616 return err; 1617 } 1618 #endif 1619 return err; 1620 } 1621 EXPORT_SYMBOL(compat_ip_getsockopt); 1622 #endif 1623