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/route.h> 37 #include <net/xfrm.h> 38 #include <net/compat.h> 39 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 40 #include <net/transp_v6.h> 41 #endif 42 43 #include <linux/errqueue.h> 44 #include <asm/uaccess.h> 45 46 #define IP_CMSG_PKTINFO 1 47 #define IP_CMSG_TTL 2 48 #define IP_CMSG_TOS 4 49 #define IP_CMSG_RECVOPTS 8 50 #define IP_CMSG_RETOPTS 16 51 #define IP_CMSG_PASSSEC 32 52 #define IP_CMSG_ORIGDSTADDR 64 53 54 /* 55 * SOL_IP control messages. 56 */ 57 58 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) 59 { 60 struct in_pktinfo info; 61 struct rtable *rt = skb_rtable(skb); 62 63 info.ipi_addr.s_addr = ip_hdr(skb)->daddr; 64 if (rt) { 65 info.ipi_ifindex = rt->rt_iif; 66 info.ipi_spec_dst.s_addr = rt->rt_spec_dst; 67 } else { 68 info.ipi_ifindex = 0; 69 info.ipi_spec_dst.s_addr = 0; 70 } 71 72 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 73 } 74 75 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb) 76 { 77 int ttl = ip_hdr(skb)->ttl; 78 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl); 79 } 80 81 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb) 82 { 83 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos); 84 } 85 86 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb) 87 { 88 if (IPCB(skb)->opt.optlen == 0) 89 return; 90 91 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen, 92 ip_hdr(skb) + 1); 93 } 94 95 96 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb) 97 { 98 unsigned char optbuf[sizeof(struct ip_options) + 40]; 99 struct ip_options * opt = (struct ip_options *)optbuf; 100 101 if (IPCB(skb)->opt.optlen == 0) 102 return; 103 104 if (ip_options_echo(opt, skb)) { 105 msg->msg_flags |= MSG_CTRUNC; 106 return; 107 } 108 ip_options_undo(opt); 109 110 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data); 111 } 112 113 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb) 114 { 115 char *secdata; 116 u32 seclen, secid; 117 int err; 118 119 err = security_socket_getpeersec_dgram(NULL, skb, &secid); 120 if (err) 121 return; 122 123 err = security_secid_to_secctx(secid, &secdata, &seclen); 124 if (err) 125 return; 126 127 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata); 128 security_release_secctx(secdata, seclen); 129 } 130 131 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb) 132 { 133 struct sockaddr_in sin; 134 struct iphdr *iph = ip_hdr(skb); 135 __be16 *ports = (__be16 *)skb_transport_header(skb); 136 137 if (skb_transport_offset(skb) + 4 > skb->len) 138 return; 139 140 /* All current transport protocols have the port numbers in the 141 * first four bytes of the transport header and this function is 142 * written with this assumption in mind. 143 */ 144 145 sin.sin_family = AF_INET; 146 sin.sin_addr.s_addr = iph->daddr; 147 sin.sin_port = ports[1]; 148 memset(sin.sin_zero, 0, sizeof(sin.sin_zero)); 149 150 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin); 151 } 152 153 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb) 154 { 155 struct inet_sock *inet = inet_sk(skb->sk); 156 unsigned flags = inet->cmsg_flags; 157 158 /* Ordered by supposed usage frequency */ 159 if (flags & 1) 160 ip_cmsg_recv_pktinfo(msg, skb); 161 if ((flags >>= 1) == 0) 162 return; 163 164 if (flags & 1) 165 ip_cmsg_recv_ttl(msg, skb); 166 if ((flags >>= 1) == 0) 167 return; 168 169 if (flags & 1) 170 ip_cmsg_recv_tos(msg, skb); 171 if ((flags >>= 1) == 0) 172 return; 173 174 if (flags & 1) 175 ip_cmsg_recv_opts(msg, skb); 176 if ((flags >>= 1) == 0) 177 return; 178 179 if (flags & 1) 180 ip_cmsg_recv_retopts(msg, skb); 181 if ((flags >>= 1) == 0) 182 return; 183 184 if (flags & 1) 185 ip_cmsg_recv_security(msg, skb); 186 187 if ((flags >>= 1) == 0) 188 return; 189 if (flags & 1) 190 ip_cmsg_recv_dstaddr(msg, skb); 191 192 } 193 EXPORT_SYMBOL(ip_cmsg_recv); 194 195 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc) 196 { 197 int err; 198 struct cmsghdr *cmsg; 199 200 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) { 201 if (!CMSG_OK(msg, cmsg)) 202 return -EINVAL; 203 if (cmsg->cmsg_level != SOL_IP) 204 continue; 205 switch (cmsg->cmsg_type) { 206 case IP_RETOPTS: 207 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr)); 208 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), 209 err < 40 ? err : 40); 210 if (err) 211 return err; 212 break; 213 case IP_PKTINFO: 214 { 215 struct in_pktinfo *info; 216 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) 217 return -EINVAL; 218 info = (struct in_pktinfo *)CMSG_DATA(cmsg); 219 ipc->oif = info->ipi_ifindex; 220 ipc->addr = info->ipi_spec_dst.s_addr; 221 break; 222 } 223 default: 224 return -EINVAL; 225 } 226 } 227 return 0; 228 } 229 230 231 /* Special input handler for packets caught by router alert option. 232 They are selected only by protocol field, and then processed likely 233 local ones; but only if someone wants them! Otherwise, router 234 not running rsvpd will kill RSVP. 235 236 It is user level problem, what it will make with them. 237 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)), 238 but receiver should be enough clever f.e. to forward mtrace requests, 239 sent to multicast group to reach destination designated router. 240 */ 241 struct ip_ra_chain __rcu *ip_ra_chain; 242 static DEFINE_SPINLOCK(ip_ra_lock); 243 244 245 static void ip_ra_destroy_rcu(struct rcu_head *head) 246 { 247 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu); 248 249 sock_put(ra->saved_sk); 250 kfree(ra); 251 } 252 253 int ip_ra_control(struct sock *sk, unsigned char on, 254 void (*destructor)(struct sock *)) 255 { 256 struct ip_ra_chain *ra, *new_ra; 257 struct ip_ra_chain __rcu **rap; 258 259 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 260 return -EINVAL; 261 262 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 263 264 spin_lock_bh(&ip_ra_lock); 265 for (rap = &ip_ra_chain; 266 (ra = rcu_dereference_protected(*rap, 267 lockdep_is_held(&ip_ra_lock))) != NULL; 268 rap = &ra->next) { 269 if (ra->sk == sk) { 270 if (on) { 271 spin_unlock_bh(&ip_ra_lock); 272 kfree(new_ra); 273 return -EADDRINUSE; 274 } 275 /* dont let ip_call_ra_chain() use sk again */ 276 ra->sk = NULL; 277 rcu_assign_pointer(*rap, ra->next); 278 spin_unlock_bh(&ip_ra_lock); 279 280 if (ra->destructor) 281 ra->destructor(sk); 282 /* 283 * Delay sock_put(sk) and kfree(ra) after one rcu grace 284 * period. This guarantee ip_call_ra_chain() dont need 285 * to mess with socket refcounts. 286 */ 287 ra->saved_sk = sk; 288 call_rcu(&ra->rcu, ip_ra_destroy_rcu); 289 return 0; 290 } 291 } 292 if (new_ra == NULL) { 293 spin_unlock_bh(&ip_ra_lock); 294 return -ENOBUFS; 295 } 296 new_ra->sk = sk; 297 new_ra->destructor = destructor; 298 299 new_ra->next = ra; 300 rcu_assign_pointer(*rap, new_ra); 301 sock_hold(sk); 302 spin_unlock_bh(&ip_ra_lock); 303 304 return 0; 305 } 306 307 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 308 __be16 port, u32 info, u8 *payload) 309 { 310 struct sock_exterr_skb *serr; 311 312 skb = skb_clone(skb, GFP_ATOMIC); 313 if (!skb) 314 return; 315 316 serr = SKB_EXT_ERR(skb); 317 serr->ee.ee_errno = err; 318 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 319 serr->ee.ee_type = icmp_hdr(skb)->type; 320 serr->ee.ee_code = icmp_hdr(skb)->code; 321 serr->ee.ee_pad = 0; 322 serr->ee.ee_info = info; 323 serr->ee.ee_data = 0; 324 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 325 skb_network_header(skb); 326 serr->port = port; 327 328 if (skb_pull(skb, payload - skb->data) != NULL) { 329 skb_reset_transport_header(skb); 330 if (sock_queue_err_skb(sk, skb) == 0) 331 return; 332 } 333 kfree_skb(skb); 334 } 335 336 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 337 { 338 struct inet_sock *inet = inet_sk(sk); 339 struct sock_exterr_skb *serr; 340 struct iphdr *iph; 341 struct sk_buff *skb; 342 343 if (!inet->recverr) 344 return; 345 346 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 347 if (!skb) 348 return; 349 350 skb_put(skb, sizeof(struct iphdr)); 351 skb_reset_network_header(skb); 352 iph = ip_hdr(skb); 353 iph->daddr = daddr; 354 355 serr = SKB_EXT_ERR(skb); 356 serr->ee.ee_errno = err; 357 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 358 serr->ee.ee_type = 0; 359 serr->ee.ee_code = 0; 360 serr->ee.ee_pad = 0; 361 serr->ee.ee_info = info; 362 serr->ee.ee_data = 0; 363 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 364 serr->port = port; 365 366 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 367 skb_reset_transport_header(skb); 368 369 if (sock_queue_err_skb(sk, skb)) 370 kfree_skb(skb); 371 } 372 373 /* 374 * Handle MSG_ERRQUEUE 375 */ 376 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len) 377 { 378 struct sock_exterr_skb *serr; 379 struct sk_buff *skb, *skb2; 380 struct sockaddr_in *sin; 381 struct { 382 struct sock_extended_err ee; 383 struct sockaddr_in offender; 384 } errhdr; 385 int err; 386 int copied; 387 388 err = -EAGAIN; 389 skb = skb_dequeue(&sk->sk_error_queue); 390 if (skb == NULL) 391 goto out; 392 393 copied = skb->len; 394 if (copied > len) { 395 msg->msg_flags |= MSG_TRUNC; 396 copied = len; 397 } 398 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 399 if (err) 400 goto out_free_skb; 401 402 sock_recv_timestamp(msg, sk, skb); 403 404 serr = SKB_EXT_ERR(skb); 405 406 sin = (struct sockaddr_in *)msg->msg_name; 407 if (sin) { 408 sin->sin_family = AF_INET; 409 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 410 serr->addr_offset); 411 sin->sin_port = serr->port; 412 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 413 } 414 415 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 416 sin = &errhdr.offender; 417 sin->sin_family = AF_UNSPEC; 418 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) { 419 struct inet_sock *inet = inet_sk(sk); 420 421 sin->sin_family = AF_INET; 422 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 423 sin->sin_port = 0; 424 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 425 if (inet->cmsg_flags) 426 ip_cmsg_recv(msg, skb); 427 } 428 429 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 430 431 /* Now we could try to dump offended packet options */ 432 433 msg->msg_flags |= MSG_ERRQUEUE; 434 err = copied; 435 436 /* Reset and regenerate socket error */ 437 spin_lock_bh(&sk->sk_error_queue.lock); 438 sk->sk_err = 0; 439 skb2 = skb_peek(&sk->sk_error_queue); 440 if (skb2 != NULL) { 441 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno; 442 spin_unlock_bh(&sk->sk_error_queue.lock); 443 sk->sk_error_report(sk); 444 } else 445 spin_unlock_bh(&sk->sk_error_queue.lock); 446 447 out_free_skb: 448 kfree_skb(skb); 449 out: 450 return err; 451 } 452 453 454 /* 455 * Socket option code for IP. This is the end of the line after any 456 * TCP,UDP etc options on an IP socket. 457 */ 458 459 static int do_ip_setsockopt(struct sock *sk, int level, 460 int optname, char __user *optval, unsigned int optlen) 461 { 462 struct inet_sock *inet = inet_sk(sk); 463 int val = 0, err; 464 465 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) | 466 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) | 467 (1<<IP_RETOPTS) | (1<<IP_TOS) | 468 (1<<IP_TTL) | (1<<IP_HDRINCL) | 469 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) | 470 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) | 471 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) | 472 (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) || 473 optname == IP_MULTICAST_TTL || 474 optname == IP_MULTICAST_ALL || 475 optname == IP_MULTICAST_LOOP || 476 optname == IP_RECVORIGDSTADDR) { 477 if (optlen >= sizeof(int)) { 478 if (get_user(val, (int __user *) optval)) 479 return -EFAULT; 480 } else if (optlen >= sizeof(char)) { 481 unsigned char ucval; 482 483 if (get_user(ucval, (unsigned char __user *) optval)) 484 return -EFAULT; 485 val = (int) ucval; 486 } 487 } 488 489 /* If optlen==0, it is equivalent to val == 0 */ 490 491 if (ip_mroute_opt(optname)) 492 return ip_mroute_setsockopt(sk, optname, optval, optlen); 493 494 err = 0; 495 lock_sock(sk); 496 497 switch (optname) { 498 case IP_OPTIONS: 499 { 500 struct ip_options *opt = NULL; 501 if (optlen > 40) 502 goto e_inval; 503 err = ip_options_get_from_user(sock_net(sk), &opt, 504 optval, optlen); 505 if (err) 506 break; 507 if (inet->is_icsk) { 508 struct inet_connection_sock *icsk = inet_csk(sk); 509 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 510 if (sk->sk_family == PF_INET || 511 (!((1 << sk->sk_state) & 512 (TCPF_LISTEN | TCPF_CLOSE)) && 513 inet->inet_daddr != LOOPBACK4_IPV6)) { 514 #endif 515 if (inet->opt) 516 icsk->icsk_ext_hdr_len -= inet->opt->optlen; 517 if (opt) 518 icsk->icsk_ext_hdr_len += opt->optlen; 519 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 520 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 521 } 522 #endif 523 } 524 opt = xchg(&inet->opt, opt); 525 kfree(opt); 526 break; 527 } 528 case IP_PKTINFO: 529 if (val) 530 inet->cmsg_flags |= IP_CMSG_PKTINFO; 531 else 532 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 533 break; 534 case IP_RECVTTL: 535 if (val) 536 inet->cmsg_flags |= IP_CMSG_TTL; 537 else 538 inet->cmsg_flags &= ~IP_CMSG_TTL; 539 break; 540 case IP_RECVTOS: 541 if (val) 542 inet->cmsg_flags |= IP_CMSG_TOS; 543 else 544 inet->cmsg_flags &= ~IP_CMSG_TOS; 545 break; 546 case IP_RECVOPTS: 547 if (val) 548 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 549 else 550 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 551 break; 552 case IP_RETOPTS: 553 if (val) 554 inet->cmsg_flags |= IP_CMSG_RETOPTS; 555 else 556 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 557 break; 558 case IP_PASSSEC: 559 if (val) 560 inet->cmsg_flags |= IP_CMSG_PASSSEC; 561 else 562 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 563 break; 564 case IP_RECVORIGDSTADDR: 565 if (val) 566 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 567 else 568 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 569 break; 570 case IP_TOS: /* This sets both TOS and Precedence */ 571 if (sk->sk_type == SOCK_STREAM) { 572 val &= ~3; 573 val |= inet->tos & 3; 574 } 575 if (inet->tos != val) { 576 inet->tos = val; 577 sk->sk_priority = rt_tos2priority(val); 578 sk_dst_reset(sk); 579 } 580 break; 581 case IP_TTL: 582 if (optlen < 1) 583 goto e_inval; 584 if (val != -1 && (val < 0 || val > 255)) 585 goto e_inval; 586 inet->uc_ttl = val; 587 break; 588 case IP_HDRINCL: 589 if (sk->sk_type != SOCK_RAW) { 590 err = -ENOPROTOOPT; 591 break; 592 } 593 inet->hdrincl = val ? 1 : 0; 594 break; 595 case IP_NODEFRAG: 596 if (sk->sk_type != SOCK_RAW) { 597 err = -ENOPROTOOPT; 598 break; 599 } 600 inet->nodefrag = val ? 1 : 0; 601 break; 602 case IP_MTU_DISCOVER: 603 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE) 604 goto e_inval; 605 inet->pmtudisc = val; 606 break; 607 case IP_RECVERR: 608 inet->recverr = !!val; 609 if (!val) 610 skb_queue_purge(&sk->sk_error_queue); 611 break; 612 case IP_MULTICAST_TTL: 613 if (sk->sk_type == SOCK_STREAM) 614 goto e_inval; 615 if (optlen < 1) 616 goto e_inval; 617 if (val == -1) 618 val = 1; 619 if (val < 0 || val > 255) 620 goto e_inval; 621 inet->mc_ttl = val; 622 break; 623 case IP_MULTICAST_LOOP: 624 if (optlen < 1) 625 goto e_inval; 626 inet->mc_loop = !!val; 627 break; 628 case IP_MULTICAST_IF: 629 { 630 struct ip_mreqn mreq; 631 struct net_device *dev = NULL; 632 633 if (sk->sk_type == SOCK_STREAM) 634 goto e_inval; 635 /* 636 * Check the arguments are allowable 637 */ 638 639 if (optlen < sizeof(struct in_addr)) 640 goto e_inval; 641 642 err = -EFAULT; 643 if (optlen >= sizeof(struct ip_mreqn)) { 644 if (copy_from_user(&mreq, optval, sizeof(mreq))) 645 break; 646 } else { 647 memset(&mreq, 0, sizeof(mreq)); 648 if (optlen >= sizeof(struct in_addr) && 649 copy_from_user(&mreq.imr_address, optval, 650 sizeof(struct in_addr))) 651 break; 652 } 653 654 if (!mreq.imr_ifindex) { 655 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 656 inet->mc_index = 0; 657 inet->mc_addr = 0; 658 err = 0; 659 break; 660 } 661 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 662 if (dev) 663 mreq.imr_ifindex = dev->ifindex; 664 } else 665 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 666 667 668 err = -EADDRNOTAVAIL; 669 if (!dev) 670 break; 671 dev_put(dev); 672 673 err = -EINVAL; 674 if (sk->sk_bound_dev_if && 675 mreq.imr_ifindex != sk->sk_bound_dev_if) 676 break; 677 678 inet->mc_index = mreq.imr_ifindex; 679 inet->mc_addr = mreq.imr_address.s_addr; 680 err = 0; 681 break; 682 } 683 684 case IP_ADD_MEMBERSHIP: 685 case IP_DROP_MEMBERSHIP: 686 { 687 struct ip_mreqn mreq; 688 689 err = -EPROTO; 690 if (inet_sk(sk)->is_icsk) 691 break; 692 693 if (optlen < sizeof(struct ip_mreq)) 694 goto e_inval; 695 err = -EFAULT; 696 if (optlen >= sizeof(struct ip_mreqn)) { 697 if (copy_from_user(&mreq, optval, sizeof(mreq))) 698 break; 699 } else { 700 memset(&mreq, 0, sizeof(mreq)); 701 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 702 break; 703 } 704 705 if (optname == IP_ADD_MEMBERSHIP) 706 err = ip_mc_join_group(sk, &mreq); 707 else 708 err = ip_mc_leave_group(sk, &mreq); 709 break; 710 } 711 case IP_MSFILTER: 712 { 713 struct ip_msfilter *msf; 714 715 if (optlen < IP_MSFILTER_SIZE(0)) 716 goto e_inval; 717 if (optlen > sysctl_optmem_max) { 718 err = -ENOBUFS; 719 break; 720 } 721 msf = kmalloc(optlen, GFP_KERNEL); 722 if (!msf) { 723 err = -ENOBUFS; 724 break; 725 } 726 err = -EFAULT; 727 if (copy_from_user(msf, optval, optlen)) { 728 kfree(msf); 729 break; 730 } 731 /* numsrc >= (1G-4) overflow in 32 bits */ 732 if (msf->imsf_numsrc >= 0x3ffffffcU || 733 msf->imsf_numsrc > sysctl_igmp_max_msf) { 734 kfree(msf); 735 err = -ENOBUFS; 736 break; 737 } 738 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 739 kfree(msf); 740 err = -EINVAL; 741 break; 742 } 743 err = ip_mc_msfilter(sk, msf, 0); 744 kfree(msf); 745 break; 746 } 747 case IP_BLOCK_SOURCE: 748 case IP_UNBLOCK_SOURCE: 749 case IP_ADD_SOURCE_MEMBERSHIP: 750 case IP_DROP_SOURCE_MEMBERSHIP: 751 { 752 struct ip_mreq_source mreqs; 753 int omode, add; 754 755 if (optlen != sizeof(struct ip_mreq_source)) 756 goto e_inval; 757 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 758 err = -EFAULT; 759 break; 760 } 761 if (optname == IP_BLOCK_SOURCE) { 762 omode = MCAST_EXCLUDE; 763 add = 1; 764 } else if (optname == IP_UNBLOCK_SOURCE) { 765 omode = MCAST_EXCLUDE; 766 add = 0; 767 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 768 struct ip_mreqn mreq; 769 770 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 771 mreq.imr_address.s_addr = mreqs.imr_interface; 772 mreq.imr_ifindex = 0; 773 err = ip_mc_join_group(sk, &mreq); 774 if (err && err != -EADDRINUSE) 775 break; 776 omode = MCAST_INCLUDE; 777 add = 1; 778 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 779 omode = MCAST_INCLUDE; 780 add = 0; 781 } 782 err = ip_mc_source(add, omode, sk, &mreqs, 0); 783 break; 784 } 785 case MCAST_JOIN_GROUP: 786 case MCAST_LEAVE_GROUP: 787 { 788 struct group_req greq; 789 struct sockaddr_in *psin; 790 struct ip_mreqn mreq; 791 792 if (optlen < sizeof(struct group_req)) 793 goto e_inval; 794 err = -EFAULT; 795 if (copy_from_user(&greq, optval, sizeof(greq))) 796 break; 797 psin = (struct sockaddr_in *)&greq.gr_group; 798 if (psin->sin_family != AF_INET) 799 goto e_inval; 800 memset(&mreq, 0, sizeof(mreq)); 801 mreq.imr_multiaddr = psin->sin_addr; 802 mreq.imr_ifindex = greq.gr_interface; 803 804 if (optname == MCAST_JOIN_GROUP) 805 err = ip_mc_join_group(sk, &mreq); 806 else 807 err = ip_mc_leave_group(sk, &mreq); 808 break; 809 } 810 case MCAST_JOIN_SOURCE_GROUP: 811 case MCAST_LEAVE_SOURCE_GROUP: 812 case MCAST_BLOCK_SOURCE: 813 case MCAST_UNBLOCK_SOURCE: 814 { 815 struct group_source_req greqs; 816 struct ip_mreq_source mreqs; 817 struct sockaddr_in *psin; 818 int omode, add; 819 820 if (optlen != sizeof(struct group_source_req)) 821 goto e_inval; 822 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 823 err = -EFAULT; 824 break; 825 } 826 if (greqs.gsr_group.ss_family != AF_INET || 827 greqs.gsr_source.ss_family != AF_INET) { 828 err = -EADDRNOTAVAIL; 829 break; 830 } 831 psin = (struct sockaddr_in *)&greqs.gsr_group; 832 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 833 psin = (struct sockaddr_in *)&greqs.gsr_source; 834 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 835 mreqs.imr_interface = 0; /* use index for mc_source */ 836 837 if (optname == MCAST_BLOCK_SOURCE) { 838 omode = MCAST_EXCLUDE; 839 add = 1; 840 } else if (optname == MCAST_UNBLOCK_SOURCE) { 841 omode = MCAST_EXCLUDE; 842 add = 0; 843 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 844 struct ip_mreqn mreq; 845 846 psin = (struct sockaddr_in *)&greqs.gsr_group; 847 mreq.imr_multiaddr = psin->sin_addr; 848 mreq.imr_address.s_addr = 0; 849 mreq.imr_ifindex = greqs.gsr_interface; 850 err = ip_mc_join_group(sk, &mreq); 851 if (err && err != -EADDRINUSE) 852 break; 853 greqs.gsr_interface = mreq.imr_ifindex; 854 omode = MCAST_INCLUDE; 855 add = 1; 856 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 857 omode = MCAST_INCLUDE; 858 add = 0; 859 } 860 err = ip_mc_source(add, omode, sk, &mreqs, 861 greqs.gsr_interface); 862 break; 863 } 864 case MCAST_MSFILTER: 865 { 866 struct sockaddr_in *psin; 867 struct ip_msfilter *msf = NULL; 868 struct group_filter *gsf = NULL; 869 int msize, i, ifindex; 870 871 if (optlen < GROUP_FILTER_SIZE(0)) 872 goto e_inval; 873 if (optlen > sysctl_optmem_max) { 874 err = -ENOBUFS; 875 break; 876 } 877 gsf = kmalloc(optlen, GFP_KERNEL); 878 if (!gsf) { 879 err = -ENOBUFS; 880 break; 881 } 882 err = -EFAULT; 883 if (copy_from_user(gsf, optval, optlen)) 884 goto mc_msf_out; 885 886 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 887 if (gsf->gf_numsrc >= 0x1ffffff || 888 gsf->gf_numsrc > sysctl_igmp_max_msf) { 889 err = -ENOBUFS; 890 goto mc_msf_out; 891 } 892 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 893 err = -EINVAL; 894 goto mc_msf_out; 895 } 896 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 897 msf = kmalloc(msize, GFP_KERNEL); 898 if (!msf) { 899 err = -ENOBUFS; 900 goto mc_msf_out; 901 } 902 ifindex = gsf->gf_interface; 903 psin = (struct sockaddr_in *)&gsf->gf_group; 904 if (psin->sin_family != AF_INET) { 905 err = -EADDRNOTAVAIL; 906 goto mc_msf_out; 907 } 908 msf->imsf_multiaddr = psin->sin_addr.s_addr; 909 msf->imsf_interface = 0; 910 msf->imsf_fmode = gsf->gf_fmode; 911 msf->imsf_numsrc = gsf->gf_numsrc; 912 err = -EADDRNOTAVAIL; 913 for (i = 0; i < gsf->gf_numsrc; ++i) { 914 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 915 916 if (psin->sin_family != AF_INET) 917 goto mc_msf_out; 918 msf->imsf_slist[i] = psin->sin_addr.s_addr; 919 } 920 kfree(gsf); 921 gsf = NULL; 922 923 err = ip_mc_msfilter(sk, msf, ifindex); 924 mc_msf_out: 925 kfree(msf); 926 kfree(gsf); 927 break; 928 } 929 case IP_MULTICAST_ALL: 930 if (optlen < 1) 931 goto e_inval; 932 if (val != 0 && val != 1) 933 goto e_inval; 934 inet->mc_all = val; 935 break; 936 case IP_ROUTER_ALERT: 937 err = ip_ra_control(sk, val ? 1 : 0, NULL); 938 break; 939 940 case IP_FREEBIND: 941 if (optlen < 1) 942 goto e_inval; 943 inet->freebind = !!val; 944 break; 945 946 case IP_IPSEC_POLICY: 947 case IP_XFRM_POLICY: 948 err = -EPERM; 949 if (!capable(CAP_NET_ADMIN)) 950 break; 951 err = xfrm_user_policy(sk, optname, optval, optlen); 952 break; 953 954 case IP_TRANSPARENT: 955 if (!capable(CAP_NET_ADMIN)) { 956 err = -EPERM; 957 break; 958 } 959 if (optlen < 1) 960 goto e_inval; 961 inet->transparent = !!val; 962 break; 963 964 case IP_MINTTL: 965 if (optlen < 1) 966 goto e_inval; 967 if (val < 0 || val > 255) 968 goto e_inval; 969 inet->min_ttl = val; 970 break; 971 972 default: 973 err = -ENOPROTOOPT; 974 break; 975 } 976 release_sock(sk); 977 return err; 978 979 e_inval: 980 release_sock(sk); 981 return -EINVAL; 982 } 983 984 /** 985 * ip_queue_rcv_skb - Queue an skb into sock receive queue 986 * @sk: socket 987 * @skb: buffer 988 * 989 * Queues an skb into socket receive queue. If IP_CMSG_PKTINFO option 990 * is not set, we drop skb dst entry now, while dst cache line is hot. 991 */ 992 int ip_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 993 { 994 if (!(inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO)) 995 skb_dst_drop(skb); 996 return sock_queue_rcv_skb(sk, skb); 997 } 998 EXPORT_SYMBOL(ip_queue_rcv_skb); 999 1000 int ip_setsockopt(struct sock *sk, int level, 1001 int optname, char __user *optval, unsigned int optlen) 1002 { 1003 int err; 1004 1005 if (level != SOL_IP) 1006 return -ENOPROTOOPT; 1007 1008 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1009 #ifdef CONFIG_NETFILTER 1010 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1011 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1012 optname != IP_IPSEC_POLICY && 1013 optname != IP_XFRM_POLICY && 1014 !ip_mroute_opt(optname)) { 1015 lock_sock(sk); 1016 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1017 release_sock(sk); 1018 } 1019 #endif 1020 return err; 1021 } 1022 EXPORT_SYMBOL(ip_setsockopt); 1023 1024 #ifdef CONFIG_COMPAT 1025 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1026 char __user *optval, unsigned int optlen) 1027 { 1028 int err; 1029 1030 if (level != SOL_IP) 1031 return -ENOPROTOOPT; 1032 1033 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1034 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1035 ip_setsockopt); 1036 1037 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1038 #ifdef CONFIG_NETFILTER 1039 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1040 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1041 optname != IP_IPSEC_POLICY && 1042 optname != IP_XFRM_POLICY && 1043 !ip_mroute_opt(optname)) { 1044 lock_sock(sk); 1045 err = compat_nf_setsockopt(sk, PF_INET, optname, 1046 optval, optlen); 1047 release_sock(sk); 1048 } 1049 #endif 1050 return err; 1051 } 1052 EXPORT_SYMBOL(compat_ip_setsockopt); 1053 #endif 1054 1055 /* 1056 * Get the options. Note for future reference. The GET of IP options gets 1057 * the _received_ ones. The set sets the _sent_ ones. 1058 */ 1059 1060 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1061 char __user *optval, int __user *optlen) 1062 { 1063 struct inet_sock *inet = inet_sk(sk); 1064 int val; 1065 int len; 1066 1067 if (level != SOL_IP) 1068 return -EOPNOTSUPP; 1069 1070 if (ip_mroute_opt(optname)) 1071 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1072 1073 if (get_user(len, optlen)) 1074 return -EFAULT; 1075 if (len < 0) 1076 return -EINVAL; 1077 1078 lock_sock(sk); 1079 1080 switch (optname) { 1081 case IP_OPTIONS: 1082 { 1083 unsigned char optbuf[sizeof(struct ip_options)+40]; 1084 struct ip_options * opt = (struct ip_options *)optbuf; 1085 opt->optlen = 0; 1086 if (inet->opt) 1087 memcpy(optbuf, inet->opt, 1088 sizeof(struct ip_options)+ 1089 inet->opt->optlen); 1090 release_sock(sk); 1091 1092 if (opt->optlen == 0) 1093 return put_user(0, optlen); 1094 1095 ip_options_undo(opt); 1096 1097 len = min_t(unsigned int, len, opt->optlen); 1098 if (put_user(len, optlen)) 1099 return -EFAULT; 1100 if (copy_to_user(optval, opt->__data, len)) 1101 return -EFAULT; 1102 return 0; 1103 } 1104 case IP_PKTINFO: 1105 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1106 break; 1107 case IP_RECVTTL: 1108 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1109 break; 1110 case IP_RECVTOS: 1111 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1112 break; 1113 case IP_RECVOPTS: 1114 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1115 break; 1116 case IP_RETOPTS: 1117 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1118 break; 1119 case IP_PASSSEC: 1120 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1121 break; 1122 case IP_RECVORIGDSTADDR: 1123 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1124 break; 1125 case IP_TOS: 1126 val = inet->tos; 1127 break; 1128 case IP_TTL: 1129 val = (inet->uc_ttl == -1 ? 1130 sysctl_ip_default_ttl : 1131 inet->uc_ttl); 1132 break; 1133 case IP_HDRINCL: 1134 val = inet->hdrincl; 1135 break; 1136 case IP_NODEFRAG: 1137 val = inet->nodefrag; 1138 break; 1139 case IP_MTU_DISCOVER: 1140 val = inet->pmtudisc; 1141 break; 1142 case IP_MTU: 1143 { 1144 struct dst_entry *dst; 1145 val = 0; 1146 dst = sk_dst_get(sk); 1147 if (dst) { 1148 val = dst_mtu(dst); 1149 dst_release(dst); 1150 } 1151 if (!val) { 1152 release_sock(sk); 1153 return -ENOTCONN; 1154 } 1155 break; 1156 } 1157 case IP_RECVERR: 1158 val = inet->recverr; 1159 break; 1160 case IP_MULTICAST_TTL: 1161 val = inet->mc_ttl; 1162 break; 1163 case IP_MULTICAST_LOOP: 1164 val = inet->mc_loop; 1165 break; 1166 case IP_MULTICAST_IF: 1167 { 1168 struct in_addr addr; 1169 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1170 addr.s_addr = inet->mc_addr; 1171 release_sock(sk); 1172 1173 if (put_user(len, optlen)) 1174 return -EFAULT; 1175 if (copy_to_user(optval, &addr, len)) 1176 return -EFAULT; 1177 return 0; 1178 } 1179 case IP_MSFILTER: 1180 { 1181 struct ip_msfilter msf; 1182 int err; 1183 1184 if (len < IP_MSFILTER_SIZE(0)) { 1185 release_sock(sk); 1186 return -EINVAL; 1187 } 1188 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1189 release_sock(sk); 1190 return -EFAULT; 1191 } 1192 err = ip_mc_msfget(sk, &msf, 1193 (struct ip_msfilter __user *)optval, optlen); 1194 release_sock(sk); 1195 return err; 1196 } 1197 case MCAST_MSFILTER: 1198 { 1199 struct group_filter gsf; 1200 int err; 1201 1202 if (len < GROUP_FILTER_SIZE(0)) { 1203 release_sock(sk); 1204 return -EINVAL; 1205 } 1206 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1207 release_sock(sk); 1208 return -EFAULT; 1209 } 1210 err = ip_mc_gsfget(sk, &gsf, 1211 (struct group_filter __user *)optval, 1212 optlen); 1213 release_sock(sk); 1214 return err; 1215 } 1216 case IP_MULTICAST_ALL: 1217 val = inet->mc_all; 1218 break; 1219 case IP_PKTOPTIONS: 1220 { 1221 struct msghdr msg; 1222 1223 release_sock(sk); 1224 1225 if (sk->sk_type != SOCK_STREAM) 1226 return -ENOPROTOOPT; 1227 1228 msg.msg_control = optval; 1229 msg.msg_controllen = len; 1230 msg.msg_flags = 0; 1231 1232 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1233 struct in_pktinfo info; 1234 1235 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1236 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1237 info.ipi_ifindex = inet->mc_index; 1238 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1239 } 1240 if (inet->cmsg_flags & IP_CMSG_TTL) { 1241 int hlim = inet->mc_ttl; 1242 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1243 } 1244 len -= msg.msg_controllen; 1245 return put_user(len, optlen); 1246 } 1247 case IP_FREEBIND: 1248 val = inet->freebind; 1249 break; 1250 case IP_TRANSPARENT: 1251 val = inet->transparent; 1252 break; 1253 case IP_MINTTL: 1254 val = inet->min_ttl; 1255 break; 1256 default: 1257 release_sock(sk); 1258 return -ENOPROTOOPT; 1259 } 1260 release_sock(sk); 1261 1262 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1263 unsigned char ucval = (unsigned char)val; 1264 len = 1; 1265 if (put_user(len, optlen)) 1266 return -EFAULT; 1267 if (copy_to_user(optval, &ucval, 1)) 1268 return -EFAULT; 1269 } else { 1270 len = min_t(unsigned int, sizeof(int), len); 1271 if (put_user(len, optlen)) 1272 return -EFAULT; 1273 if (copy_to_user(optval, &val, len)) 1274 return -EFAULT; 1275 } 1276 return 0; 1277 } 1278 1279 int ip_getsockopt(struct sock *sk, int level, 1280 int optname, char __user *optval, int __user *optlen) 1281 { 1282 int err; 1283 1284 err = do_ip_getsockopt(sk, level, optname, optval, optlen); 1285 #ifdef CONFIG_NETFILTER 1286 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1287 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1288 !ip_mroute_opt(optname)) { 1289 int len; 1290 1291 if (get_user(len, optlen)) 1292 return -EFAULT; 1293 1294 lock_sock(sk); 1295 err = nf_getsockopt(sk, PF_INET, optname, optval, 1296 &len); 1297 release_sock(sk); 1298 if (err >= 0) 1299 err = put_user(len, optlen); 1300 return err; 1301 } 1302 #endif 1303 return err; 1304 } 1305 EXPORT_SYMBOL(ip_getsockopt); 1306 1307 #ifdef CONFIG_COMPAT 1308 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1309 char __user *optval, int __user *optlen) 1310 { 1311 int err; 1312 1313 if (optname == MCAST_MSFILTER) 1314 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1315 ip_getsockopt); 1316 1317 err = do_ip_getsockopt(sk, level, optname, optval, optlen); 1318 1319 #ifdef CONFIG_NETFILTER 1320 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1321 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1322 !ip_mroute_opt(optname)) { 1323 int len; 1324 1325 if (get_user(len, optlen)) 1326 return -EFAULT; 1327 1328 lock_sock(sk); 1329 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1330 release_sock(sk); 1331 if (err >= 0) 1332 err = put_user(len, optlen); 1333 return err; 1334 } 1335 #endif 1336 return err; 1337 } 1338 EXPORT_SYMBOL(compat_ip_getsockopt); 1339 #endif 1340