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 const 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 static void opt_kfree_rcu(struct rcu_head *head) 455 { 456 kfree(container_of(head, struct ip_options_rcu, rcu)); 457 } 458 459 /* 460 * Socket option code for IP. This is the end of the line after any 461 * TCP,UDP etc options on an IP socket. 462 */ 463 464 static int do_ip_setsockopt(struct sock *sk, int level, 465 int optname, char __user *optval, unsigned int optlen) 466 { 467 struct inet_sock *inet = inet_sk(sk); 468 int val = 0, err; 469 470 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) | 471 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) | 472 (1<<IP_RETOPTS) | (1<<IP_TOS) | 473 (1<<IP_TTL) | (1<<IP_HDRINCL) | 474 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) | 475 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) | 476 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) | 477 (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) || 478 optname == IP_MULTICAST_TTL || 479 optname == IP_MULTICAST_ALL || 480 optname == IP_MULTICAST_LOOP || 481 optname == IP_RECVORIGDSTADDR) { 482 if (optlen >= sizeof(int)) { 483 if (get_user(val, (int __user *) optval)) 484 return -EFAULT; 485 } else if (optlen >= sizeof(char)) { 486 unsigned char ucval; 487 488 if (get_user(ucval, (unsigned char __user *) optval)) 489 return -EFAULT; 490 val = (int) ucval; 491 } 492 } 493 494 /* If optlen==0, it is equivalent to val == 0 */ 495 496 if (ip_mroute_opt(optname)) 497 return ip_mroute_setsockopt(sk, optname, optval, optlen); 498 499 err = 0; 500 lock_sock(sk); 501 502 switch (optname) { 503 case IP_OPTIONS: 504 { 505 struct ip_options_rcu *old, *opt = NULL; 506 507 if (optlen > 40) 508 goto e_inval; 509 err = ip_options_get_from_user(sock_net(sk), &opt, 510 optval, optlen); 511 if (err) 512 break; 513 old = rcu_dereference_protected(inet->inet_opt, 514 sock_owned_by_user(sk)); 515 if (inet->is_icsk) { 516 struct inet_connection_sock *icsk = inet_csk(sk); 517 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 518 if (sk->sk_family == PF_INET || 519 (!((1 << sk->sk_state) & 520 (TCPF_LISTEN | TCPF_CLOSE)) && 521 inet->inet_daddr != LOOPBACK4_IPV6)) { 522 #endif 523 if (old) 524 icsk->icsk_ext_hdr_len -= old->opt.optlen; 525 if (opt) 526 icsk->icsk_ext_hdr_len += opt->opt.optlen; 527 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 528 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 529 } 530 #endif 531 } 532 rcu_assign_pointer(inet->inet_opt, opt); 533 if (old) 534 call_rcu(&old->rcu, opt_kfree_rcu); 535 break; 536 } 537 case IP_PKTINFO: 538 if (val) 539 inet->cmsg_flags |= IP_CMSG_PKTINFO; 540 else 541 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 542 break; 543 case IP_RECVTTL: 544 if (val) 545 inet->cmsg_flags |= IP_CMSG_TTL; 546 else 547 inet->cmsg_flags &= ~IP_CMSG_TTL; 548 break; 549 case IP_RECVTOS: 550 if (val) 551 inet->cmsg_flags |= IP_CMSG_TOS; 552 else 553 inet->cmsg_flags &= ~IP_CMSG_TOS; 554 break; 555 case IP_RECVOPTS: 556 if (val) 557 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 558 else 559 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 560 break; 561 case IP_RETOPTS: 562 if (val) 563 inet->cmsg_flags |= IP_CMSG_RETOPTS; 564 else 565 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 566 break; 567 case IP_PASSSEC: 568 if (val) 569 inet->cmsg_flags |= IP_CMSG_PASSSEC; 570 else 571 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 572 break; 573 case IP_RECVORIGDSTADDR: 574 if (val) 575 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 576 else 577 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 578 break; 579 case IP_TOS: /* This sets both TOS and Precedence */ 580 if (sk->sk_type == SOCK_STREAM) { 581 val &= ~3; 582 val |= inet->tos & 3; 583 } 584 if (inet->tos != val) { 585 inet->tos = val; 586 sk->sk_priority = rt_tos2priority(val); 587 sk_dst_reset(sk); 588 } 589 break; 590 case IP_TTL: 591 if (optlen < 1) 592 goto e_inval; 593 if (val != -1 && (val < 0 || val > 255)) 594 goto e_inval; 595 inet->uc_ttl = val; 596 break; 597 case IP_HDRINCL: 598 if (sk->sk_type != SOCK_RAW) { 599 err = -ENOPROTOOPT; 600 break; 601 } 602 inet->hdrincl = val ? 1 : 0; 603 break; 604 case IP_NODEFRAG: 605 if (sk->sk_type != SOCK_RAW) { 606 err = -ENOPROTOOPT; 607 break; 608 } 609 inet->nodefrag = val ? 1 : 0; 610 break; 611 case IP_MTU_DISCOVER: 612 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE) 613 goto e_inval; 614 inet->pmtudisc = val; 615 break; 616 case IP_RECVERR: 617 inet->recverr = !!val; 618 if (!val) 619 skb_queue_purge(&sk->sk_error_queue); 620 break; 621 case IP_MULTICAST_TTL: 622 if (sk->sk_type == SOCK_STREAM) 623 goto e_inval; 624 if (optlen < 1) 625 goto e_inval; 626 if (val == -1) 627 val = 1; 628 if (val < 0 || val > 255) 629 goto e_inval; 630 inet->mc_ttl = val; 631 break; 632 case IP_MULTICAST_LOOP: 633 if (optlen < 1) 634 goto e_inval; 635 inet->mc_loop = !!val; 636 break; 637 case IP_MULTICAST_IF: 638 { 639 struct ip_mreqn mreq; 640 struct net_device *dev = NULL; 641 642 if (sk->sk_type == SOCK_STREAM) 643 goto e_inval; 644 /* 645 * Check the arguments are allowable 646 */ 647 648 if (optlen < sizeof(struct in_addr)) 649 goto e_inval; 650 651 err = -EFAULT; 652 if (optlen >= sizeof(struct ip_mreqn)) { 653 if (copy_from_user(&mreq, optval, sizeof(mreq))) 654 break; 655 } else { 656 memset(&mreq, 0, sizeof(mreq)); 657 if (optlen >= sizeof(struct in_addr) && 658 copy_from_user(&mreq.imr_address, optval, 659 sizeof(struct in_addr))) 660 break; 661 } 662 663 if (!mreq.imr_ifindex) { 664 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 665 inet->mc_index = 0; 666 inet->mc_addr = 0; 667 err = 0; 668 break; 669 } 670 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 671 if (dev) 672 mreq.imr_ifindex = dev->ifindex; 673 } else 674 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 675 676 677 err = -EADDRNOTAVAIL; 678 if (!dev) 679 break; 680 dev_put(dev); 681 682 err = -EINVAL; 683 if (sk->sk_bound_dev_if && 684 mreq.imr_ifindex != sk->sk_bound_dev_if) 685 break; 686 687 inet->mc_index = mreq.imr_ifindex; 688 inet->mc_addr = mreq.imr_address.s_addr; 689 err = 0; 690 break; 691 } 692 693 case IP_ADD_MEMBERSHIP: 694 case IP_DROP_MEMBERSHIP: 695 { 696 struct ip_mreqn mreq; 697 698 err = -EPROTO; 699 if (inet_sk(sk)->is_icsk) 700 break; 701 702 if (optlen < sizeof(struct ip_mreq)) 703 goto e_inval; 704 err = -EFAULT; 705 if (optlen >= sizeof(struct ip_mreqn)) { 706 if (copy_from_user(&mreq, optval, sizeof(mreq))) 707 break; 708 } else { 709 memset(&mreq, 0, sizeof(mreq)); 710 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 711 break; 712 } 713 714 if (optname == IP_ADD_MEMBERSHIP) 715 err = ip_mc_join_group(sk, &mreq); 716 else 717 err = ip_mc_leave_group(sk, &mreq); 718 break; 719 } 720 case IP_MSFILTER: 721 { 722 struct ip_msfilter *msf; 723 724 if (optlen < IP_MSFILTER_SIZE(0)) 725 goto e_inval; 726 if (optlen > sysctl_optmem_max) { 727 err = -ENOBUFS; 728 break; 729 } 730 msf = kmalloc(optlen, GFP_KERNEL); 731 if (!msf) { 732 err = -ENOBUFS; 733 break; 734 } 735 err = -EFAULT; 736 if (copy_from_user(msf, optval, optlen)) { 737 kfree(msf); 738 break; 739 } 740 /* numsrc >= (1G-4) overflow in 32 bits */ 741 if (msf->imsf_numsrc >= 0x3ffffffcU || 742 msf->imsf_numsrc > sysctl_igmp_max_msf) { 743 kfree(msf); 744 err = -ENOBUFS; 745 break; 746 } 747 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 748 kfree(msf); 749 err = -EINVAL; 750 break; 751 } 752 err = ip_mc_msfilter(sk, msf, 0); 753 kfree(msf); 754 break; 755 } 756 case IP_BLOCK_SOURCE: 757 case IP_UNBLOCK_SOURCE: 758 case IP_ADD_SOURCE_MEMBERSHIP: 759 case IP_DROP_SOURCE_MEMBERSHIP: 760 { 761 struct ip_mreq_source mreqs; 762 int omode, add; 763 764 if (optlen != sizeof(struct ip_mreq_source)) 765 goto e_inval; 766 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 767 err = -EFAULT; 768 break; 769 } 770 if (optname == IP_BLOCK_SOURCE) { 771 omode = MCAST_EXCLUDE; 772 add = 1; 773 } else if (optname == IP_UNBLOCK_SOURCE) { 774 omode = MCAST_EXCLUDE; 775 add = 0; 776 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 777 struct ip_mreqn mreq; 778 779 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 780 mreq.imr_address.s_addr = mreqs.imr_interface; 781 mreq.imr_ifindex = 0; 782 err = ip_mc_join_group(sk, &mreq); 783 if (err && err != -EADDRINUSE) 784 break; 785 omode = MCAST_INCLUDE; 786 add = 1; 787 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 788 omode = MCAST_INCLUDE; 789 add = 0; 790 } 791 err = ip_mc_source(add, omode, sk, &mreqs, 0); 792 break; 793 } 794 case MCAST_JOIN_GROUP: 795 case MCAST_LEAVE_GROUP: 796 { 797 struct group_req greq; 798 struct sockaddr_in *psin; 799 struct ip_mreqn mreq; 800 801 if (optlen < sizeof(struct group_req)) 802 goto e_inval; 803 err = -EFAULT; 804 if (copy_from_user(&greq, optval, sizeof(greq))) 805 break; 806 psin = (struct sockaddr_in *)&greq.gr_group; 807 if (psin->sin_family != AF_INET) 808 goto e_inval; 809 memset(&mreq, 0, sizeof(mreq)); 810 mreq.imr_multiaddr = psin->sin_addr; 811 mreq.imr_ifindex = greq.gr_interface; 812 813 if (optname == MCAST_JOIN_GROUP) 814 err = ip_mc_join_group(sk, &mreq); 815 else 816 err = ip_mc_leave_group(sk, &mreq); 817 break; 818 } 819 case MCAST_JOIN_SOURCE_GROUP: 820 case MCAST_LEAVE_SOURCE_GROUP: 821 case MCAST_BLOCK_SOURCE: 822 case MCAST_UNBLOCK_SOURCE: 823 { 824 struct group_source_req greqs; 825 struct ip_mreq_source mreqs; 826 struct sockaddr_in *psin; 827 int omode, add; 828 829 if (optlen != sizeof(struct group_source_req)) 830 goto e_inval; 831 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 832 err = -EFAULT; 833 break; 834 } 835 if (greqs.gsr_group.ss_family != AF_INET || 836 greqs.gsr_source.ss_family != AF_INET) { 837 err = -EADDRNOTAVAIL; 838 break; 839 } 840 psin = (struct sockaddr_in *)&greqs.gsr_group; 841 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 842 psin = (struct sockaddr_in *)&greqs.gsr_source; 843 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 844 mreqs.imr_interface = 0; /* use index for mc_source */ 845 846 if (optname == MCAST_BLOCK_SOURCE) { 847 omode = MCAST_EXCLUDE; 848 add = 1; 849 } else if (optname == MCAST_UNBLOCK_SOURCE) { 850 omode = MCAST_EXCLUDE; 851 add = 0; 852 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 853 struct ip_mreqn mreq; 854 855 psin = (struct sockaddr_in *)&greqs.gsr_group; 856 mreq.imr_multiaddr = psin->sin_addr; 857 mreq.imr_address.s_addr = 0; 858 mreq.imr_ifindex = greqs.gsr_interface; 859 err = ip_mc_join_group(sk, &mreq); 860 if (err && err != -EADDRINUSE) 861 break; 862 greqs.gsr_interface = mreq.imr_ifindex; 863 omode = MCAST_INCLUDE; 864 add = 1; 865 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 866 omode = MCAST_INCLUDE; 867 add = 0; 868 } 869 err = ip_mc_source(add, omode, sk, &mreqs, 870 greqs.gsr_interface); 871 break; 872 } 873 case MCAST_MSFILTER: 874 { 875 struct sockaddr_in *psin; 876 struct ip_msfilter *msf = NULL; 877 struct group_filter *gsf = NULL; 878 int msize, i, ifindex; 879 880 if (optlen < GROUP_FILTER_SIZE(0)) 881 goto e_inval; 882 if (optlen > sysctl_optmem_max) { 883 err = -ENOBUFS; 884 break; 885 } 886 gsf = kmalloc(optlen, GFP_KERNEL); 887 if (!gsf) { 888 err = -ENOBUFS; 889 break; 890 } 891 err = -EFAULT; 892 if (copy_from_user(gsf, optval, optlen)) 893 goto mc_msf_out; 894 895 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 896 if (gsf->gf_numsrc >= 0x1ffffff || 897 gsf->gf_numsrc > sysctl_igmp_max_msf) { 898 err = -ENOBUFS; 899 goto mc_msf_out; 900 } 901 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 902 err = -EINVAL; 903 goto mc_msf_out; 904 } 905 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 906 msf = kmalloc(msize, GFP_KERNEL); 907 if (!msf) { 908 err = -ENOBUFS; 909 goto mc_msf_out; 910 } 911 ifindex = gsf->gf_interface; 912 psin = (struct sockaddr_in *)&gsf->gf_group; 913 if (psin->sin_family != AF_INET) { 914 err = -EADDRNOTAVAIL; 915 goto mc_msf_out; 916 } 917 msf->imsf_multiaddr = psin->sin_addr.s_addr; 918 msf->imsf_interface = 0; 919 msf->imsf_fmode = gsf->gf_fmode; 920 msf->imsf_numsrc = gsf->gf_numsrc; 921 err = -EADDRNOTAVAIL; 922 for (i = 0; i < gsf->gf_numsrc; ++i) { 923 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 924 925 if (psin->sin_family != AF_INET) 926 goto mc_msf_out; 927 msf->imsf_slist[i] = psin->sin_addr.s_addr; 928 } 929 kfree(gsf); 930 gsf = NULL; 931 932 err = ip_mc_msfilter(sk, msf, ifindex); 933 mc_msf_out: 934 kfree(msf); 935 kfree(gsf); 936 break; 937 } 938 case IP_MULTICAST_ALL: 939 if (optlen < 1) 940 goto e_inval; 941 if (val != 0 && val != 1) 942 goto e_inval; 943 inet->mc_all = val; 944 break; 945 case IP_ROUTER_ALERT: 946 err = ip_ra_control(sk, val ? 1 : 0, NULL); 947 break; 948 949 case IP_FREEBIND: 950 if (optlen < 1) 951 goto e_inval; 952 inet->freebind = !!val; 953 break; 954 955 case IP_IPSEC_POLICY: 956 case IP_XFRM_POLICY: 957 err = -EPERM; 958 if (!capable(CAP_NET_ADMIN)) 959 break; 960 err = xfrm_user_policy(sk, optname, optval, optlen); 961 break; 962 963 case IP_TRANSPARENT: 964 if (!capable(CAP_NET_ADMIN)) { 965 err = -EPERM; 966 break; 967 } 968 if (optlen < 1) 969 goto e_inval; 970 inet->transparent = !!val; 971 break; 972 973 case IP_MINTTL: 974 if (optlen < 1) 975 goto e_inval; 976 if (val < 0 || val > 255) 977 goto e_inval; 978 inet->min_ttl = val; 979 break; 980 981 default: 982 err = -ENOPROTOOPT; 983 break; 984 } 985 release_sock(sk); 986 return err; 987 988 e_inval: 989 release_sock(sk); 990 return -EINVAL; 991 } 992 993 /** 994 * ip_queue_rcv_skb - Queue an skb into sock receive queue 995 * @sk: socket 996 * @skb: buffer 997 * 998 * Queues an skb into socket receive queue. If IP_CMSG_PKTINFO option 999 * is not set, we drop skb dst entry now, while dst cache line is hot. 1000 */ 1001 int ip_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 1002 { 1003 if (!(inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO)) 1004 skb_dst_drop(skb); 1005 return sock_queue_rcv_skb(sk, skb); 1006 } 1007 EXPORT_SYMBOL(ip_queue_rcv_skb); 1008 1009 int ip_setsockopt(struct sock *sk, int level, 1010 int optname, char __user *optval, unsigned int optlen) 1011 { 1012 int err; 1013 1014 if (level != SOL_IP) 1015 return -ENOPROTOOPT; 1016 1017 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1018 #ifdef CONFIG_NETFILTER 1019 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1020 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1021 optname != IP_IPSEC_POLICY && 1022 optname != IP_XFRM_POLICY && 1023 !ip_mroute_opt(optname)) { 1024 lock_sock(sk); 1025 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1026 release_sock(sk); 1027 } 1028 #endif 1029 return err; 1030 } 1031 EXPORT_SYMBOL(ip_setsockopt); 1032 1033 #ifdef CONFIG_COMPAT 1034 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1035 char __user *optval, unsigned int optlen) 1036 { 1037 int err; 1038 1039 if (level != SOL_IP) 1040 return -ENOPROTOOPT; 1041 1042 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1043 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1044 ip_setsockopt); 1045 1046 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1047 #ifdef CONFIG_NETFILTER 1048 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1049 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1050 optname != IP_IPSEC_POLICY && 1051 optname != IP_XFRM_POLICY && 1052 !ip_mroute_opt(optname)) { 1053 lock_sock(sk); 1054 err = compat_nf_setsockopt(sk, PF_INET, optname, 1055 optval, optlen); 1056 release_sock(sk); 1057 } 1058 #endif 1059 return err; 1060 } 1061 EXPORT_SYMBOL(compat_ip_setsockopt); 1062 #endif 1063 1064 /* 1065 * Get the options. Note for future reference. The GET of IP options gets 1066 * the _received_ ones. The set sets the _sent_ ones. 1067 */ 1068 1069 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1070 char __user *optval, int __user *optlen, unsigned flags) 1071 { 1072 struct inet_sock *inet = inet_sk(sk); 1073 int val; 1074 int len; 1075 1076 if (level != SOL_IP) 1077 return -EOPNOTSUPP; 1078 1079 if (ip_mroute_opt(optname)) 1080 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1081 1082 if (get_user(len, optlen)) 1083 return -EFAULT; 1084 if (len < 0) 1085 return -EINVAL; 1086 1087 lock_sock(sk); 1088 1089 switch (optname) { 1090 case IP_OPTIONS: 1091 { 1092 unsigned char optbuf[sizeof(struct ip_options)+40]; 1093 struct ip_options *opt = (struct ip_options *)optbuf; 1094 struct ip_options_rcu *inet_opt; 1095 1096 inet_opt = rcu_dereference_protected(inet->inet_opt, 1097 sock_owned_by_user(sk)); 1098 opt->optlen = 0; 1099 if (inet_opt) 1100 memcpy(optbuf, &inet_opt->opt, 1101 sizeof(struct ip_options) + 1102 inet_opt->opt.optlen); 1103 release_sock(sk); 1104 1105 if (opt->optlen == 0) 1106 return put_user(0, optlen); 1107 1108 ip_options_undo(opt); 1109 1110 len = min_t(unsigned int, len, opt->optlen); 1111 if (put_user(len, optlen)) 1112 return -EFAULT; 1113 if (copy_to_user(optval, opt->__data, len)) 1114 return -EFAULT; 1115 return 0; 1116 } 1117 case IP_PKTINFO: 1118 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1119 break; 1120 case IP_RECVTTL: 1121 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1122 break; 1123 case IP_RECVTOS: 1124 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1125 break; 1126 case IP_RECVOPTS: 1127 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1128 break; 1129 case IP_RETOPTS: 1130 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1131 break; 1132 case IP_PASSSEC: 1133 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1134 break; 1135 case IP_RECVORIGDSTADDR: 1136 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1137 break; 1138 case IP_TOS: 1139 val = inet->tos; 1140 break; 1141 case IP_TTL: 1142 val = (inet->uc_ttl == -1 ? 1143 sysctl_ip_default_ttl : 1144 inet->uc_ttl); 1145 break; 1146 case IP_HDRINCL: 1147 val = inet->hdrincl; 1148 break; 1149 case IP_NODEFRAG: 1150 val = inet->nodefrag; 1151 break; 1152 case IP_MTU_DISCOVER: 1153 val = inet->pmtudisc; 1154 break; 1155 case IP_MTU: 1156 { 1157 struct dst_entry *dst; 1158 val = 0; 1159 dst = sk_dst_get(sk); 1160 if (dst) { 1161 val = dst_mtu(dst); 1162 dst_release(dst); 1163 } 1164 if (!val) { 1165 release_sock(sk); 1166 return -ENOTCONN; 1167 } 1168 break; 1169 } 1170 case IP_RECVERR: 1171 val = inet->recverr; 1172 break; 1173 case IP_MULTICAST_TTL: 1174 val = inet->mc_ttl; 1175 break; 1176 case IP_MULTICAST_LOOP: 1177 val = inet->mc_loop; 1178 break; 1179 case IP_MULTICAST_IF: 1180 { 1181 struct in_addr addr; 1182 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1183 addr.s_addr = inet->mc_addr; 1184 release_sock(sk); 1185 1186 if (put_user(len, optlen)) 1187 return -EFAULT; 1188 if (copy_to_user(optval, &addr, len)) 1189 return -EFAULT; 1190 return 0; 1191 } 1192 case IP_MSFILTER: 1193 { 1194 struct ip_msfilter msf; 1195 int err; 1196 1197 if (len < IP_MSFILTER_SIZE(0)) { 1198 release_sock(sk); 1199 return -EINVAL; 1200 } 1201 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1202 release_sock(sk); 1203 return -EFAULT; 1204 } 1205 err = ip_mc_msfget(sk, &msf, 1206 (struct ip_msfilter __user *)optval, optlen); 1207 release_sock(sk); 1208 return err; 1209 } 1210 case MCAST_MSFILTER: 1211 { 1212 struct group_filter gsf; 1213 int err; 1214 1215 if (len < GROUP_FILTER_SIZE(0)) { 1216 release_sock(sk); 1217 return -EINVAL; 1218 } 1219 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1220 release_sock(sk); 1221 return -EFAULT; 1222 } 1223 err = ip_mc_gsfget(sk, &gsf, 1224 (struct group_filter __user *)optval, 1225 optlen); 1226 release_sock(sk); 1227 return err; 1228 } 1229 case IP_MULTICAST_ALL: 1230 val = inet->mc_all; 1231 break; 1232 case IP_PKTOPTIONS: 1233 { 1234 struct msghdr msg; 1235 1236 release_sock(sk); 1237 1238 if (sk->sk_type != SOCK_STREAM) 1239 return -ENOPROTOOPT; 1240 1241 msg.msg_control = optval; 1242 msg.msg_controllen = len; 1243 msg.msg_flags = flags; 1244 1245 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1246 struct in_pktinfo info; 1247 1248 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1249 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1250 info.ipi_ifindex = inet->mc_index; 1251 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1252 } 1253 if (inet->cmsg_flags & IP_CMSG_TTL) { 1254 int hlim = inet->mc_ttl; 1255 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1256 } 1257 len -= msg.msg_controllen; 1258 return put_user(len, optlen); 1259 } 1260 case IP_FREEBIND: 1261 val = inet->freebind; 1262 break; 1263 case IP_TRANSPARENT: 1264 val = inet->transparent; 1265 break; 1266 case IP_MINTTL: 1267 val = inet->min_ttl; 1268 break; 1269 default: 1270 release_sock(sk); 1271 return -ENOPROTOOPT; 1272 } 1273 release_sock(sk); 1274 1275 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1276 unsigned char ucval = (unsigned char)val; 1277 len = 1; 1278 if (put_user(len, optlen)) 1279 return -EFAULT; 1280 if (copy_to_user(optval, &ucval, 1)) 1281 return -EFAULT; 1282 } else { 1283 len = min_t(unsigned int, sizeof(int), len); 1284 if (put_user(len, optlen)) 1285 return -EFAULT; 1286 if (copy_to_user(optval, &val, len)) 1287 return -EFAULT; 1288 } 1289 return 0; 1290 } 1291 1292 int ip_getsockopt(struct sock *sk, int level, 1293 int optname, char __user *optval, int __user *optlen) 1294 { 1295 int err; 1296 1297 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1298 #ifdef CONFIG_NETFILTER 1299 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1300 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1301 !ip_mroute_opt(optname)) { 1302 int len; 1303 1304 if (get_user(len, optlen)) 1305 return -EFAULT; 1306 1307 lock_sock(sk); 1308 err = nf_getsockopt(sk, PF_INET, optname, optval, 1309 &len); 1310 release_sock(sk); 1311 if (err >= 0) 1312 err = put_user(len, optlen); 1313 return err; 1314 } 1315 #endif 1316 return err; 1317 } 1318 EXPORT_SYMBOL(ip_getsockopt); 1319 1320 #ifdef CONFIG_COMPAT 1321 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1322 char __user *optval, int __user *optlen) 1323 { 1324 int err; 1325 1326 if (optname == MCAST_MSFILTER) 1327 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1328 ip_getsockopt); 1329 1330 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1331 MSG_CMSG_COMPAT); 1332 1333 #ifdef CONFIG_NETFILTER 1334 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1335 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1336 !ip_mroute_opt(optname)) { 1337 int len; 1338 1339 if (get_user(len, optlen)) 1340 return -EFAULT; 1341 1342 lock_sock(sk); 1343 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1344 release_sock(sk); 1345 if (err >= 0) 1346 err = put_user(len, optlen); 1347 return err; 1348 } 1349 #endif 1350 return err; 1351 } 1352 EXPORT_SYMBOL(compat_ip_getsockopt); 1353 #endif 1354