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