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 *ip_ra_chain; 242 DEFINE_RWLOCK(ip_ra_lock); 243 244 int ip_ra_control(struct sock *sk, unsigned char on, 245 void (*destructor)(struct sock *)) 246 { 247 struct ip_ra_chain *ra, *new_ra, **rap; 248 249 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 250 return -EINVAL; 251 252 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 253 254 write_lock_bh(&ip_ra_lock); 255 for (rap = &ip_ra_chain; (ra = *rap) != NULL; rap = &ra->next) { 256 if (ra->sk == sk) { 257 if (on) { 258 write_unlock_bh(&ip_ra_lock); 259 kfree(new_ra); 260 return -EADDRINUSE; 261 } 262 *rap = ra->next; 263 write_unlock_bh(&ip_ra_lock); 264 265 if (ra->destructor) 266 ra->destructor(sk); 267 sock_put(sk); 268 kfree(ra); 269 return 0; 270 } 271 } 272 if (new_ra == NULL) { 273 write_unlock_bh(&ip_ra_lock); 274 return -ENOBUFS; 275 } 276 new_ra->sk = sk; 277 new_ra->destructor = destructor; 278 279 new_ra->next = ra; 280 *rap = new_ra; 281 sock_hold(sk); 282 write_unlock_bh(&ip_ra_lock); 283 284 return 0; 285 } 286 287 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 288 __be16 port, u32 info, u8 *payload) 289 { 290 struct inet_sock *inet = inet_sk(sk); 291 struct sock_exterr_skb *serr; 292 293 if (!inet->recverr) 294 return; 295 296 skb = skb_clone(skb, GFP_ATOMIC); 297 if (!skb) 298 return; 299 300 serr = SKB_EXT_ERR(skb); 301 serr->ee.ee_errno = err; 302 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 303 serr->ee.ee_type = icmp_hdr(skb)->type; 304 serr->ee.ee_code = icmp_hdr(skb)->code; 305 serr->ee.ee_pad = 0; 306 serr->ee.ee_info = info; 307 serr->ee.ee_data = 0; 308 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 309 skb_network_header(skb); 310 serr->port = port; 311 312 if (skb_pull(skb, payload - skb->data) != NULL) { 313 skb_reset_transport_header(skb); 314 if (sock_queue_err_skb(sk, skb) == 0) 315 return; 316 } 317 kfree_skb(skb); 318 } 319 320 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 321 { 322 struct inet_sock *inet = inet_sk(sk); 323 struct sock_exterr_skb *serr; 324 struct iphdr *iph; 325 struct sk_buff *skb; 326 327 if (!inet->recverr) 328 return; 329 330 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 331 if (!skb) 332 return; 333 334 skb_put(skb, sizeof(struct iphdr)); 335 skb_reset_network_header(skb); 336 iph = ip_hdr(skb); 337 iph->daddr = daddr; 338 339 serr = SKB_EXT_ERR(skb); 340 serr->ee.ee_errno = err; 341 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 342 serr->ee.ee_type = 0; 343 serr->ee.ee_code = 0; 344 serr->ee.ee_pad = 0; 345 serr->ee.ee_info = info; 346 serr->ee.ee_data = 0; 347 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 348 serr->port = port; 349 350 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 351 skb_reset_transport_header(skb); 352 353 if (sock_queue_err_skb(sk, skb)) 354 kfree_skb(skb); 355 } 356 357 /* 358 * Handle MSG_ERRQUEUE 359 */ 360 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len) 361 { 362 struct sock_exterr_skb *serr; 363 struct sk_buff *skb, *skb2; 364 struct sockaddr_in *sin; 365 struct { 366 struct sock_extended_err ee; 367 struct sockaddr_in offender; 368 } errhdr; 369 int err; 370 int copied; 371 372 err = -EAGAIN; 373 skb = skb_dequeue(&sk->sk_error_queue); 374 if (skb == NULL) 375 goto out; 376 377 copied = skb->len; 378 if (copied > len) { 379 msg->msg_flags |= MSG_TRUNC; 380 copied = len; 381 } 382 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 383 if (err) 384 goto out_free_skb; 385 386 sock_recv_timestamp(msg, sk, skb); 387 388 serr = SKB_EXT_ERR(skb); 389 390 sin = (struct sockaddr_in *)msg->msg_name; 391 if (sin) { 392 sin->sin_family = AF_INET; 393 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 394 serr->addr_offset); 395 sin->sin_port = serr->port; 396 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 397 } 398 399 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 400 sin = &errhdr.offender; 401 sin->sin_family = AF_UNSPEC; 402 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) { 403 struct inet_sock *inet = inet_sk(sk); 404 405 sin->sin_family = AF_INET; 406 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 407 sin->sin_port = 0; 408 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 409 if (inet->cmsg_flags) 410 ip_cmsg_recv(msg, skb); 411 } 412 413 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 414 415 /* Now we could try to dump offended packet options */ 416 417 msg->msg_flags |= MSG_ERRQUEUE; 418 err = copied; 419 420 /* Reset and regenerate socket error */ 421 spin_lock_bh(&sk->sk_error_queue.lock); 422 sk->sk_err = 0; 423 skb2 = skb_peek(&sk->sk_error_queue); 424 if (skb2 != NULL) { 425 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno; 426 spin_unlock_bh(&sk->sk_error_queue.lock); 427 sk->sk_error_report(sk); 428 } else 429 spin_unlock_bh(&sk->sk_error_queue.lock); 430 431 out_free_skb: 432 kfree_skb(skb); 433 out: 434 return err; 435 } 436 437 438 /* 439 * Socket option code for IP. This is the end of the line after any 440 * TCP,UDP etc options on an IP socket. 441 */ 442 443 static int do_ip_setsockopt(struct sock *sk, int level, 444 int optname, char __user *optval, unsigned int optlen) 445 { 446 struct inet_sock *inet = inet_sk(sk); 447 int val = 0, err; 448 449 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) | 450 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) | 451 (1<<IP_RETOPTS) | (1<<IP_TOS) | 452 (1<<IP_TTL) | (1<<IP_HDRINCL) | 453 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) | 454 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) | 455 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) | 456 (1<<IP_MINTTL))) || 457 optname == IP_MULTICAST_TTL || 458 optname == IP_MULTICAST_ALL || 459 optname == IP_MULTICAST_LOOP || 460 optname == IP_RECVORIGDSTADDR) { 461 if (optlen >= sizeof(int)) { 462 if (get_user(val, (int __user *) optval)) 463 return -EFAULT; 464 } else if (optlen >= sizeof(char)) { 465 unsigned char ucval; 466 467 if (get_user(ucval, (unsigned char __user *) optval)) 468 return -EFAULT; 469 val = (int) ucval; 470 } 471 } 472 473 /* If optlen==0, it is equivalent to val == 0 */ 474 475 if (ip_mroute_opt(optname)) 476 return ip_mroute_setsockopt(sk, optname, optval, optlen); 477 478 err = 0; 479 lock_sock(sk); 480 481 switch (optname) { 482 case IP_OPTIONS: 483 { 484 struct ip_options *opt = NULL; 485 if (optlen > 40) 486 goto e_inval; 487 err = ip_options_get_from_user(sock_net(sk), &opt, 488 optval, optlen); 489 if (err) 490 break; 491 if (inet->is_icsk) { 492 struct inet_connection_sock *icsk = inet_csk(sk); 493 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 494 if (sk->sk_family == PF_INET || 495 (!((1 << sk->sk_state) & 496 (TCPF_LISTEN | TCPF_CLOSE)) && 497 inet->inet_daddr != LOOPBACK4_IPV6)) { 498 #endif 499 if (inet->opt) 500 icsk->icsk_ext_hdr_len -= inet->opt->optlen; 501 if (opt) 502 icsk->icsk_ext_hdr_len += opt->optlen; 503 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 504 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 505 } 506 #endif 507 } 508 opt = xchg(&inet->opt, opt); 509 kfree(opt); 510 break; 511 } 512 case IP_PKTINFO: 513 if (val) 514 inet->cmsg_flags |= IP_CMSG_PKTINFO; 515 else 516 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 517 break; 518 case IP_RECVTTL: 519 if (val) 520 inet->cmsg_flags |= IP_CMSG_TTL; 521 else 522 inet->cmsg_flags &= ~IP_CMSG_TTL; 523 break; 524 case IP_RECVTOS: 525 if (val) 526 inet->cmsg_flags |= IP_CMSG_TOS; 527 else 528 inet->cmsg_flags &= ~IP_CMSG_TOS; 529 break; 530 case IP_RECVOPTS: 531 if (val) 532 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 533 else 534 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 535 break; 536 case IP_RETOPTS: 537 if (val) 538 inet->cmsg_flags |= IP_CMSG_RETOPTS; 539 else 540 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 541 break; 542 case IP_PASSSEC: 543 if (val) 544 inet->cmsg_flags |= IP_CMSG_PASSSEC; 545 else 546 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 547 break; 548 case IP_RECVORIGDSTADDR: 549 if (val) 550 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 551 else 552 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 553 break; 554 case IP_TOS: /* This sets both TOS and Precedence */ 555 if (sk->sk_type == SOCK_STREAM) { 556 val &= ~3; 557 val |= inet->tos & 3; 558 } 559 if (inet->tos != val) { 560 inet->tos = val; 561 sk->sk_priority = rt_tos2priority(val); 562 sk_dst_reset(sk); 563 } 564 break; 565 case IP_TTL: 566 if (optlen < 1) 567 goto e_inval; 568 if (val != -1 && (val < 0 || val > 255)) 569 goto e_inval; 570 inet->uc_ttl = val; 571 break; 572 case IP_HDRINCL: 573 if (sk->sk_type != SOCK_RAW) { 574 err = -ENOPROTOOPT; 575 break; 576 } 577 inet->hdrincl = val ? 1 : 0; 578 break; 579 case IP_MTU_DISCOVER: 580 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE) 581 goto e_inval; 582 inet->pmtudisc = val; 583 break; 584 case IP_RECVERR: 585 inet->recverr = !!val; 586 if (!val) 587 skb_queue_purge(&sk->sk_error_queue); 588 break; 589 case IP_MULTICAST_TTL: 590 if (sk->sk_type == SOCK_STREAM) 591 goto e_inval; 592 if (optlen < 1) 593 goto e_inval; 594 if (val == -1) 595 val = 1; 596 if (val < 0 || val > 255) 597 goto e_inval; 598 inet->mc_ttl = val; 599 break; 600 case IP_MULTICAST_LOOP: 601 if (optlen < 1) 602 goto e_inval; 603 inet->mc_loop = !!val; 604 break; 605 case IP_MULTICAST_IF: 606 { 607 struct ip_mreqn mreq; 608 struct net_device *dev = NULL; 609 610 if (sk->sk_type == SOCK_STREAM) 611 goto e_inval; 612 /* 613 * Check the arguments are allowable 614 */ 615 616 if (optlen < sizeof(struct in_addr)) 617 goto e_inval; 618 619 err = -EFAULT; 620 if (optlen >= sizeof(struct ip_mreqn)) { 621 if (copy_from_user(&mreq, optval, sizeof(mreq))) 622 break; 623 } else { 624 memset(&mreq, 0, sizeof(mreq)); 625 if (optlen >= sizeof(struct in_addr) && 626 copy_from_user(&mreq.imr_address, optval, 627 sizeof(struct in_addr))) 628 break; 629 } 630 631 if (!mreq.imr_ifindex) { 632 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 633 inet->mc_index = 0; 634 inet->mc_addr = 0; 635 err = 0; 636 break; 637 } 638 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 639 if (dev) 640 mreq.imr_ifindex = dev->ifindex; 641 } else 642 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 643 644 645 err = -EADDRNOTAVAIL; 646 if (!dev) 647 break; 648 dev_put(dev); 649 650 err = -EINVAL; 651 if (sk->sk_bound_dev_if && 652 mreq.imr_ifindex != sk->sk_bound_dev_if) 653 break; 654 655 inet->mc_index = mreq.imr_ifindex; 656 inet->mc_addr = mreq.imr_address.s_addr; 657 err = 0; 658 break; 659 } 660 661 case IP_ADD_MEMBERSHIP: 662 case IP_DROP_MEMBERSHIP: 663 { 664 struct ip_mreqn mreq; 665 666 err = -EPROTO; 667 if (inet_sk(sk)->is_icsk) 668 break; 669 670 if (optlen < sizeof(struct ip_mreq)) 671 goto e_inval; 672 err = -EFAULT; 673 if (optlen >= sizeof(struct ip_mreqn)) { 674 if (copy_from_user(&mreq, optval, sizeof(mreq))) 675 break; 676 } else { 677 memset(&mreq, 0, sizeof(mreq)); 678 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 679 break; 680 } 681 682 if (optname == IP_ADD_MEMBERSHIP) 683 err = ip_mc_join_group(sk, &mreq); 684 else 685 err = ip_mc_leave_group(sk, &mreq); 686 break; 687 } 688 case IP_MSFILTER: 689 { 690 struct ip_msfilter *msf; 691 692 if (optlen < IP_MSFILTER_SIZE(0)) 693 goto e_inval; 694 if (optlen > sysctl_optmem_max) { 695 err = -ENOBUFS; 696 break; 697 } 698 msf = kmalloc(optlen, GFP_KERNEL); 699 if (!msf) { 700 err = -ENOBUFS; 701 break; 702 } 703 err = -EFAULT; 704 if (copy_from_user(msf, optval, optlen)) { 705 kfree(msf); 706 break; 707 } 708 /* numsrc >= (1G-4) overflow in 32 bits */ 709 if (msf->imsf_numsrc >= 0x3ffffffcU || 710 msf->imsf_numsrc > sysctl_igmp_max_msf) { 711 kfree(msf); 712 err = -ENOBUFS; 713 break; 714 } 715 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 716 kfree(msf); 717 err = -EINVAL; 718 break; 719 } 720 err = ip_mc_msfilter(sk, msf, 0); 721 kfree(msf); 722 break; 723 } 724 case IP_BLOCK_SOURCE: 725 case IP_UNBLOCK_SOURCE: 726 case IP_ADD_SOURCE_MEMBERSHIP: 727 case IP_DROP_SOURCE_MEMBERSHIP: 728 { 729 struct ip_mreq_source mreqs; 730 int omode, add; 731 732 if (optlen != sizeof(struct ip_mreq_source)) 733 goto e_inval; 734 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 735 err = -EFAULT; 736 break; 737 } 738 if (optname == IP_BLOCK_SOURCE) { 739 omode = MCAST_EXCLUDE; 740 add = 1; 741 } else if (optname == IP_UNBLOCK_SOURCE) { 742 omode = MCAST_EXCLUDE; 743 add = 0; 744 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 745 struct ip_mreqn mreq; 746 747 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 748 mreq.imr_address.s_addr = mreqs.imr_interface; 749 mreq.imr_ifindex = 0; 750 err = ip_mc_join_group(sk, &mreq); 751 if (err && err != -EADDRINUSE) 752 break; 753 omode = MCAST_INCLUDE; 754 add = 1; 755 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 756 omode = MCAST_INCLUDE; 757 add = 0; 758 } 759 err = ip_mc_source(add, omode, sk, &mreqs, 0); 760 break; 761 } 762 case MCAST_JOIN_GROUP: 763 case MCAST_LEAVE_GROUP: 764 { 765 struct group_req greq; 766 struct sockaddr_in *psin; 767 struct ip_mreqn mreq; 768 769 if (optlen < sizeof(struct group_req)) 770 goto e_inval; 771 err = -EFAULT; 772 if (copy_from_user(&greq, optval, sizeof(greq))) 773 break; 774 psin = (struct sockaddr_in *)&greq.gr_group; 775 if (psin->sin_family != AF_INET) 776 goto e_inval; 777 memset(&mreq, 0, sizeof(mreq)); 778 mreq.imr_multiaddr = psin->sin_addr; 779 mreq.imr_ifindex = greq.gr_interface; 780 781 if (optname == MCAST_JOIN_GROUP) 782 err = ip_mc_join_group(sk, &mreq); 783 else 784 err = ip_mc_leave_group(sk, &mreq); 785 break; 786 } 787 case MCAST_JOIN_SOURCE_GROUP: 788 case MCAST_LEAVE_SOURCE_GROUP: 789 case MCAST_BLOCK_SOURCE: 790 case MCAST_UNBLOCK_SOURCE: 791 { 792 struct group_source_req greqs; 793 struct ip_mreq_source mreqs; 794 struct sockaddr_in *psin; 795 int omode, add; 796 797 if (optlen != sizeof(struct group_source_req)) 798 goto e_inval; 799 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 800 err = -EFAULT; 801 break; 802 } 803 if (greqs.gsr_group.ss_family != AF_INET || 804 greqs.gsr_source.ss_family != AF_INET) { 805 err = -EADDRNOTAVAIL; 806 break; 807 } 808 psin = (struct sockaddr_in *)&greqs.gsr_group; 809 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 810 psin = (struct sockaddr_in *)&greqs.gsr_source; 811 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 812 mreqs.imr_interface = 0; /* use index for mc_source */ 813 814 if (optname == MCAST_BLOCK_SOURCE) { 815 omode = MCAST_EXCLUDE; 816 add = 1; 817 } else if (optname == MCAST_UNBLOCK_SOURCE) { 818 omode = MCAST_EXCLUDE; 819 add = 0; 820 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 821 struct ip_mreqn mreq; 822 823 psin = (struct sockaddr_in *)&greqs.gsr_group; 824 mreq.imr_multiaddr = psin->sin_addr; 825 mreq.imr_address.s_addr = 0; 826 mreq.imr_ifindex = greqs.gsr_interface; 827 err = ip_mc_join_group(sk, &mreq); 828 if (err && err != -EADDRINUSE) 829 break; 830 greqs.gsr_interface = mreq.imr_ifindex; 831 omode = MCAST_INCLUDE; 832 add = 1; 833 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 834 omode = MCAST_INCLUDE; 835 add = 0; 836 } 837 err = ip_mc_source(add, omode, sk, &mreqs, 838 greqs.gsr_interface); 839 break; 840 } 841 case MCAST_MSFILTER: 842 { 843 struct sockaddr_in *psin; 844 struct ip_msfilter *msf = NULL; 845 struct group_filter *gsf = NULL; 846 int msize, i, ifindex; 847 848 if (optlen < GROUP_FILTER_SIZE(0)) 849 goto e_inval; 850 if (optlen > sysctl_optmem_max) { 851 err = -ENOBUFS; 852 break; 853 } 854 gsf = kmalloc(optlen, GFP_KERNEL); 855 if (!gsf) { 856 err = -ENOBUFS; 857 break; 858 } 859 err = -EFAULT; 860 if (copy_from_user(gsf, optval, optlen)) 861 goto mc_msf_out; 862 863 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 864 if (gsf->gf_numsrc >= 0x1ffffff || 865 gsf->gf_numsrc > sysctl_igmp_max_msf) { 866 err = -ENOBUFS; 867 goto mc_msf_out; 868 } 869 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 870 err = -EINVAL; 871 goto mc_msf_out; 872 } 873 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 874 msf = kmalloc(msize, GFP_KERNEL); 875 if (!msf) { 876 err = -ENOBUFS; 877 goto mc_msf_out; 878 } 879 ifindex = gsf->gf_interface; 880 psin = (struct sockaddr_in *)&gsf->gf_group; 881 if (psin->sin_family != AF_INET) { 882 err = -EADDRNOTAVAIL; 883 goto mc_msf_out; 884 } 885 msf->imsf_multiaddr = psin->sin_addr.s_addr; 886 msf->imsf_interface = 0; 887 msf->imsf_fmode = gsf->gf_fmode; 888 msf->imsf_numsrc = gsf->gf_numsrc; 889 err = -EADDRNOTAVAIL; 890 for (i = 0; i < gsf->gf_numsrc; ++i) { 891 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 892 893 if (psin->sin_family != AF_INET) 894 goto mc_msf_out; 895 msf->imsf_slist[i] = psin->sin_addr.s_addr; 896 } 897 kfree(gsf); 898 gsf = NULL; 899 900 err = ip_mc_msfilter(sk, msf, ifindex); 901 mc_msf_out: 902 kfree(msf); 903 kfree(gsf); 904 break; 905 } 906 case IP_MULTICAST_ALL: 907 if (optlen < 1) 908 goto e_inval; 909 if (val != 0 && val != 1) 910 goto e_inval; 911 inet->mc_all = val; 912 break; 913 case IP_ROUTER_ALERT: 914 err = ip_ra_control(sk, val ? 1 : 0, NULL); 915 break; 916 917 case IP_FREEBIND: 918 if (optlen < 1) 919 goto e_inval; 920 inet->freebind = !!val; 921 break; 922 923 case IP_IPSEC_POLICY: 924 case IP_XFRM_POLICY: 925 err = -EPERM; 926 if (!capable(CAP_NET_ADMIN)) 927 break; 928 err = xfrm_user_policy(sk, optname, optval, optlen); 929 break; 930 931 case IP_TRANSPARENT: 932 if (!capable(CAP_NET_ADMIN)) { 933 err = -EPERM; 934 break; 935 } 936 if (optlen < 1) 937 goto e_inval; 938 inet->transparent = !!val; 939 break; 940 941 case IP_MINTTL: 942 if (optlen < 1) 943 goto e_inval; 944 if (val < 0 || val > 255) 945 goto e_inval; 946 inet->min_ttl = val; 947 break; 948 949 default: 950 err = -ENOPROTOOPT; 951 break; 952 } 953 release_sock(sk); 954 return err; 955 956 e_inval: 957 release_sock(sk); 958 return -EINVAL; 959 } 960 961 int ip_setsockopt(struct sock *sk, int level, 962 int optname, char __user *optval, unsigned int optlen) 963 { 964 int err; 965 966 if (level != SOL_IP) 967 return -ENOPROTOOPT; 968 969 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 970 #ifdef CONFIG_NETFILTER 971 /* we need to exclude all possible ENOPROTOOPTs except default case */ 972 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 973 optname != IP_IPSEC_POLICY && 974 optname != IP_XFRM_POLICY && 975 !ip_mroute_opt(optname)) { 976 lock_sock(sk); 977 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 978 release_sock(sk); 979 } 980 #endif 981 return err; 982 } 983 EXPORT_SYMBOL(ip_setsockopt); 984 985 #ifdef CONFIG_COMPAT 986 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 987 char __user *optval, unsigned int optlen) 988 { 989 int err; 990 991 if (level != SOL_IP) 992 return -ENOPROTOOPT; 993 994 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 995 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 996 ip_setsockopt); 997 998 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 999 #ifdef CONFIG_NETFILTER 1000 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1001 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1002 optname != IP_IPSEC_POLICY && 1003 optname != IP_XFRM_POLICY && 1004 !ip_mroute_opt(optname)) { 1005 lock_sock(sk); 1006 err = compat_nf_setsockopt(sk, PF_INET, optname, 1007 optval, optlen); 1008 release_sock(sk); 1009 } 1010 #endif 1011 return err; 1012 } 1013 EXPORT_SYMBOL(compat_ip_setsockopt); 1014 #endif 1015 1016 /* 1017 * Get the options. Note for future reference. The GET of IP options gets 1018 * the _received_ ones. The set sets the _sent_ ones. 1019 */ 1020 1021 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1022 char __user *optval, int __user *optlen) 1023 { 1024 struct inet_sock *inet = inet_sk(sk); 1025 int val; 1026 int len; 1027 1028 if (level != SOL_IP) 1029 return -EOPNOTSUPP; 1030 1031 if (ip_mroute_opt(optname)) 1032 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1033 1034 if (get_user(len, optlen)) 1035 return -EFAULT; 1036 if (len < 0) 1037 return -EINVAL; 1038 1039 lock_sock(sk); 1040 1041 switch (optname) { 1042 case IP_OPTIONS: 1043 { 1044 unsigned char optbuf[sizeof(struct ip_options)+40]; 1045 struct ip_options * opt = (struct ip_options *)optbuf; 1046 opt->optlen = 0; 1047 if (inet->opt) 1048 memcpy(optbuf, inet->opt, 1049 sizeof(struct ip_options)+ 1050 inet->opt->optlen); 1051 release_sock(sk); 1052 1053 if (opt->optlen == 0) 1054 return put_user(0, optlen); 1055 1056 ip_options_undo(opt); 1057 1058 len = min_t(unsigned int, len, opt->optlen); 1059 if (put_user(len, optlen)) 1060 return -EFAULT; 1061 if (copy_to_user(optval, opt->__data, len)) 1062 return -EFAULT; 1063 return 0; 1064 } 1065 case IP_PKTINFO: 1066 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1067 break; 1068 case IP_RECVTTL: 1069 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1070 break; 1071 case IP_RECVTOS: 1072 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1073 break; 1074 case IP_RECVOPTS: 1075 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1076 break; 1077 case IP_RETOPTS: 1078 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1079 break; 1080 case IP_PASSSEC: 1081 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1082 break; 1083 case IP_RECVORIGDSTADDR: 1084 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1085 break; 1086 case IP_TOS: 1087 val = inet->tos; 1088 break; 1089 case IP_TTL: 1090 val = (inet->uc_ttl == -1 ? 1091 sysctl_ip_default_ttl : 1092 inet->uc_ttl); 1093 break; 1094 case IP_HDRINCL: 1095 val = inet->hdrincl; 1096 break; 1097 case IP_MTU_DISCOVER: 1098 val = inet->pmtudisc; 1099 break; 1100 case IP_MTU: 1101 { 1102 struct dst_entry *dst; 1103 val = 0; 1104 dst = sk_dst_get(sk); 1105 if (dst) { 1106 val = dst_mtu(dst); 1107 dst_release(dst); 1108 } 1109 if (!val) { 1110 release_sock(sk); 1111 return -ENOTCONN; 1112 } 1113 break; 1114 } 1115 case IP_RECVERR: 1116 val = inet->recverr; 1117 break; 1118 case IP_MULTICAST_TTL: 1119 val = inet->mc_ttl; 1120 break; 1121 case IP_MULTICAST_LOOP: 1122 val = inet->mc_loop; 1123 break; 1124 case IP_MULTICAST_IF: 1125 { 1126 struct in_addr addr; 1127 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1128 addr.s_addr = inet->mc_addr; 1129 release_sock(sk); 1130 1131 if (put_user(len, optlen)) 1132 return -EFAULT; 1133 if (copy_to_user(optval, &addr, len)) 1134 return -EFAULT; 1135 return 0; 1136 } 1137 case IP_MSFILTER: 1138 { 1139 struct ip_msfilter msf; 1140 int err; 1141 1142 if (len < IP_MSFILTER_SIZE(0)) { 1143 release_sock(sk); 1144 return -EINVAL; 1145 } 1146 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1147 release_sock(sk); 1148 return -EFAULT; 1149 } 1150 err = ip_mc_msfget(sk, &msf, 1151 (struct ip_msfilter __user *)optval, optlen); 1152 release_sock(sk); 1153 return err; 1154 } 1155 case MCAST_MSFILTER: 1156 { 1157 struct group_filter gsf; 1158 int err; 1159 1160 if (len < GROUP_FILTER_SIZE(0)) { 1161 release_sock(sk); 1162 return -EINVAL; 1163 } 1164 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1165 release_sock(sk); 1166 return -EFAULT; 1167 } 1168 err = ip_mc_gsfget(sk, &gsf, 1169 (struct group_filter __user *)optval, 1170 optlen); 1171 release_sock(sk); 1172 return err; 1173 } 1174 case IP_MULTICAST_ALL: 1175 val = inet->mc_all; 1176 break; 1177 case IP_PKTOPTIONS: 1178 { 1179 struct msghdr msg; 1180 1181 release_sock(sk); 1182 1183 if (sk->sk_type != SOCK_STREAM) 1184 return -ENOPROTOOPT; 1185 1186 msg.msg_control = optval; 1187 msg.msg_controllen = len; 1188 msg.msg_flags = 0; 1189 1190 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1191 struct in_pktinfo info; 1192 1193 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1194 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1195 info.ipi_ifindex = inet->mc_index; 1196 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1197 } 1198 if (inet->cmsg_flags & IP_CMSG_TTL) { 1199 int hlim = inet->mc_ttl; 1200 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1201 } 1202 len -= msg.msg_controllen; 1203 return put_user(len, optlen); 1204 } 1205 case IP_FREEBIND: 1206 val = inet->freebind; 1207 break; 1208 case IP_TRANSPARENT: 1209 val = inet->transparent; 1210 break; 1211 case IP_MINTTL: 1212 val = inet->min_ttl; 1213 break; 1214 default: 1215 release_sock(sk); 1216 return -ENOPROTOOPT; 1217 } 1218 release_sock(sk); 1219 1220 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1221 unsigned char ucval = (unsigned char)val; 1222 len = 1; 1223 if (put_user(len, optlen)) 1224 return -EFAULT; 1225 if (copy_to_user(optval, &ucval, 1)) 1226 return -EFAULT; 1227 } else { 1228 len = min_t(unsigned int, sizeof(int), len); 1229 if (put_user(len, optlen)) 1230 return -EFAULT; 1231 if (copy_to_user(optval, &val, len)) 1232 return -EFAULT; 1233 } 1234 return 0; 1235 } 1236 1237 int ip_getsockopt(struct sock *sk, int level, 1238 int optname, char __user *optval, int __user *optlen) 1239 { 1240 int err; 1241 1242 err = do_ip_getsockopt(sk, level, optname, optval, optlen); 1243 #ifdef CONFIG_NETFILTER 1244 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1245 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1246 !ip_mroute_opt(optname)) { 1247 int len; 1248 1249 if (get_user(len, optlen)) 1250 return -EFAULT; 1251 1252 lock_sock(sk); 1253 err = nf_getsockopt(sk, PF_INET, optname, optval, 1254 &len); 1255 release_sock(sk); 1256 if (err >= 0) 1257 err = put_user(len, optlen); 1258 return err; 1259 } 1260 #endif 1261 return err; 1262 } 1263 EXPORT_SYMBOL(ip_getsockopt); 1264 1265 #ifdef CONFIG_COMPAT 1266 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1267 char __user *optval, int __user *optlen) 1268 { 1269 int err; 1270 1271 if (optname == MCAST_MSFILTER) 1272 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1273 ip_getsockopt); 1274 1275 err = do_ip_getsockopt(sk, level, optname, optval, optlen); 1276 1277 #ifdef CONFIG_NETFILTER 1278 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1279 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1280 !ip_mroute_opt(optname)) { 1281 int len; 1282 1283 if (get_user(len, optlen)) 1284 return -EFAULT; 1285 1286 lock_sock(sk); 1287 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1288 release_sock(sk); 1289 if (err >= 0) 1290 err = put_user(len, optlen); 1291 return err; 1292 } 1293 #endif 1294 return err; 1295 } 1296 EXPORT_SYMBOL(compat_ip_getsockopt); 1297 #endif 1298