1 /* 2 * INET An implementation of the TCP/IP protocol suite for the LINUX 3 * operating system. INET is implemented using the BSD Socket 4 * interface as the means of communication with the user level. 5 * 6 * The IP to API glue. 7 * 8 * Authors: see ip.c 9 * 10 * Fixes: 11 * Many : Split from ip.c , see ip.c for history. 12 * Martin Mares : TOS setting fixed. 13 * Alan Cox : Fixed a couple of oopses in Martin's 14 * TOS tweaks. 15 * Mike McLagan : Routing by source 16 */ 17 18 #include <linux/module.h> 19 #include <linux/types.h> 20 #include <linux/mm.h> 21 #include <linux/skbuff.h> 22 #include <linux/ip.h> 23 #include <linux/icmp.h> 24 #include <linux/inetdevice.h> 25 #include <linux/netdevice.h> 26 #include <linux/slab.h> 27 #include <net/sock.h> 28 #include <net/ip.h> 29 #include <net/icmp.h> 30 #include <net/tcp_states.h> 31 #include <linux/udp.h> 32 #include <linux/igmp.h> 33 #include <linux/netfilter.h> 34 #include <linux/route.h> 35 #include <linux/mroute.h> 36 #include <net/inet_ecn.h> 37 #include <net/route.h> 38 #include <net/xfrm.h> 39 #include <net/compat.h> 40 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 41 #include <net/transp_v6.h> 42 #endif 43 44 #include <linux/errqueue.h> 45 #include <asm/uaccess.h> 46 47 #define IP_CMSG_PKTINFO 1 48 #define IP_CMSG_TTL 2 49 #define IP_CMSG_TOS 4 50 #define IP_CMSG_RECVOPTS 8 51 #define IP_CMSG_RETOPTS 16 52 #define IP_CMSG_PASSSEC 32 53 #define IP_CMSG_ORIGDSTADDR 64 54 55 /* 56 * SOL_IP control messages. 57 */ 58 59 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) 60 { 61 struct in_pktinfo info; 62 struct rtable *rt = skb_rtable(skb); 63 64 info.ipi_addr.s_addr = ip_hdr(skb)->daddr; 65 if (rt) { 66 info.ipi_ifindex = rt->rt_iif; 67 info.ipi_spec_dst.s_addr = rt->rt_spec_dst; 68 } else { 69 info.ipi_ifindex = 0; 70 info.ipi_spec_dst.s_addr = 0; 71 } 72 73 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 74 } 75 76 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb) 77 { 78 int ttl = ip_hdr(skb)->ttl; 79 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl); 80 } 81 82 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb) 83 { 84 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos); 85 } 86 87 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb) 88 { 89 if (IPCB(skb)->opt.optlen == 0) 90 return; 91 92 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen, 93 ip_hdr(skb) + 1); 94 } 95 96 97 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb) 98 { 99 unsigned char optbuf[sizeof(struct ip_options) + 40]; 100 struct ip_options * opt = (struct ip_options *)optbuf; 101 102 if (IPCB(skb)->opt.optlen == 0) 103 return; 104 105 if (ip_options_echo(opt, skb)) { 106 msg->msg_flags |= MSG_CTRUNC; 107 return; 108 } 109 ip_options_undo(opt); 110 111 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data); 112 } 113 114 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb) 115 { 116 char *secdata; 117 u32 seclen, secid; 118 int err; 119 120 err = security_socket_getpeersec_dgram(NULL, skb, &secid); 121 if (err) 122 return; 123 124 err = security_secid_to_secctx(secid, &secdata, &seclen); 125 if (err) 126 return; 127 128 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata); 129 security_release_secctx(secdata, seclen); 130 } 131 132 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb) 133 { 134 struct sockaddr_in sin; 135 const struct iphdr *iph = ip_hdr(skb); 136 __be16 *ports = (__be16 *)skb_transport_header(skb); 137 138 if (skb_transport_offset(skb) + 4 > skb->len) 139 return; 140 141 /* All current transport protocols have the port numbers in the 142 * first four bytes of the transport header and this function is 143 * written with this assumption in mind. 144 */ 145 146 sin.sin_family = AF_INET; 147 sin.sin_addr.s_addr = iph->daddr; 148 sin.sin_port = ports[1]; 149 memset(sin.sin_zero, 0, sizeof(sin.sin_zero)); 150 151 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin); 152 } 153 154 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb) 155 { 156 struct inet_sock *inet = inet_sk(skb->sk); 157 unsigned flags = inet->cmsg_flags; 158 159 /* Ordered by supposed usage frequency */ 160 if (flags & 1) 161 ip_cmsg_recv_pktinfo(msg, skb); 162 if ((flags >>= 1) == 0) 163 return; 164 165 if (flags & 1) 166 ip_cmsg_recv_ttl(msg, skb); 167 if ((flags >>= 1) == 0) 168 return; 169 170 if (flags & 1) 171 ip_cmsg_recv_tos(msg, skb); 172 if ((flags >>= 1) == 0) 173 return; 174 175 if (flags & 1) 176 ip_cmsg_recv_opts(msg, skb); 177 if ((flags >>= 1) == 0) 178 return; 179 180 if (flags & 1) 181 ip_cmsg_recv_retopts(msg, skb); 182 if ((flags >>= 1) == 0) 183 return; 184 185 if (flags & 1) 186 ip_cmsg_recv_security(msg, skb); 187 188 if ((flags >>= 1) == 0) 189 return; 190 if (flags & 1) 191 ip_cmsg_recv_dstaddr(msg, skb); 192 193 } 194 EXPORT_SYMBOL(ip_cmsg_recv); 195 196 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc) 197 { 198 int err; 199 struct cmsghdr *cmsg; 200 201 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) { 202 if (!CMSG_OK(msg, cmsg)) 203 return -EINVAL; 204 if (cmsg->cmsg_level != SOL_IP) 205 continue; 206 switch (cmsg->cmsg_type) { 207 case IP_RETOPTS: 208 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr)); 209 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), 210 err < 40 ? err : 40); 211 if (err) 212 return err; 213 break; 214 case IP_PKTINFO: 215 { 216 struct in_pktinfo *info; 217 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) 218 return -EINVAL; 219 info = (struct in_pktinfo *)CMSG_DATA(cmsg); 220 ipc->oif = info->ipi_ifindex; 221 ipc->addr = info->ipi_spec_dst.s_addr; 222 break; 223 } 224 default: 225 return -EINVAL; 226 } 227 } 228 return 0; 229 } 230 231 232 /* Special input handler for packets caught by router alert option. 233 They are selected only by protocol field, and then processed likely 234 local ones; but only if someone wants them! Otherwise, router 235 not running rsvpd will kill RSVP. 236 237 It is user level problem, what it will make with them. 238 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)), 239 but receiver should be enough clever f.e. to forward mtrace requests, 240 sent to multicast group to reach destination designated router. 241 */ 242 struct ip_ra_chain __rcu *ip_ra_chain; 243 static DEFINE_SPINLOCK(ip_ra_lock); 244 245 246 static void ip_ra_destroy_rcu(struct rcu_head *head) 247 { 248 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu); 249 250 sock_put(ra->saved_sk); 251 kfree(ra); 252 } 253 254 int ip_ra_control(struct sock *sk, unsigned char on, 255 void (*destructor)(struct sock *)) 256 { 257 struct ip_ra_chain *ra, *new_ra; 258 struct ip_ra_chain __rcu **rap; 259 260 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 261 return -EINVAL; 262 263 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 264 265 spin_lock_bh(&ip_ra_lock); 266 for (rap = &ip_ra_chain; 267 (ra = rcu_dereference_protected(*rap, 268 lockdep_is_held(&ip_ra_lock))) != NULL; 269 rap = &ra->next) { 270 if (ra->sk == sk) { 271 if (on) { 272 spin_unlock_bh(&ip_ra_lock); 273 kfree(new_ra); 274 return -EADDRINUSE; 275 } 276 /* dont let ip_call_ra_chain() use sk again */ 277 ra->sk = NULL; 278 rcu_assign_pointer(*rap, ra->next); 279 spin_unlock_bh(&ip_ra_lock); 280 281 if (ra->destructor) 282 ra->destructor(sk); 283 /* 284 * Delay sock_put(sk) and kfree(ra) after one rcu grace 285 * period. This guarantee ip_call_ra_chain() dont need 286 * to mess with socket refcounts. 287 */ 288 ra->saved_sk = sk; 289 call_rcu(&ra->rcu, ip_ra_destroy_rcu); 290 return 0; 291 } 292 } 293 if (new_ra == NULL) { 294 spin_unlock_bh(&ip_ra_lock); 295 return -ENOBUFS; 296 } 297 new_ra->sk = sk; 298 new_ra->destructor = destructor; 299 300 new_ra->next = ra; 301 rcu_assign_pointer(*rap, new_ra); 302 sock_hold(sk); 303 spin_unlock_bh(&ip_ra_lock); 304 305 return 0; 306 } 307 308 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 309 __be16 port, u32 info, u8 *payload) 310 { 311 struct sock_exterr_skb *serr; 312 313 skb = skb_clone(skb, GFP_ATOMIC); 314 if (!skb) 315 return; 316 317 serr = SKB_EXT_ERR(skb); 318 serr->ee.ee_errno = err; 319 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 320 serr->ee.ee_type = icmp_hdr(skb)->type; 321 serr->ee.ee_code = icmp_hdr(skb)->code; 322 serr->ee.ee_pad = 0; 323 serr->ee.ee_info = info; 324 serr->ee.ee_data = 0; 325 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 326 skb_network_header(skb); 327 serr->port = port; 328 329 if (skb_pull(skb, payload - skb->data) != NULL) { 330 skb_reset_transport_header(skb); 331 if (sock_queue_err_skb(sk, skb) == 0) 332 return; 333 } 334 kfree_skb(skb); 335 } 336 337 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 338 { 339 struct inet_sock *inet = inet_sk(sk); 340 struct sock_exterr_skb *serr; 341 struct iphdr *iph; 342 struct sk_buff *skb; 343 344 if (!inet->recverr) 345 return; 346 347 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 348 if (!skb) 349 return; 350 351 skb_put(skb, sizeof(struct iphdr)); 352 skb_reset_network_header(skb); 353 iph = ip_hdr(skb); 354 iph->daddr = daddr; 355 356 serr = SKB_EXT_ERR(skb); 357 serr->ee.ee_errno = err; 358 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 359 serr->ee.ee_type = 0; 360 serr->ee.ee_code = 0; 361 serr->ee.ee_pad = 0; 362 serr->ee.ee_info = info; 363 serr->ee.ee_data = 0; 364 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 365 serr->port = port; 366 367 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 368 skb_reset_transport_header(skb); 369 370 if (sock_queue_err_skb(sk, skb)) 371 kfree_skb(skb); 372 } 373 374 /* 375 * Handle MSG_ERRQUEUE 376 */ 377 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len) 378 { 379 struct sock_exterr_skb *serr; 380 struct sk_buff *skb, *skb2; 381 struct sockaddr_in *sin; 382 struct { 383 struct sock_extended_err ee; 384 struct sockaddr_in offender; 385 } errhdr; 386 int err; 387 int copied; 388 389 err = -EAGAIN; 390 skb = skb_dequeue(&sk->sk_error_queue); 391 if (skb == NULL) 392 goto out; 393 394 copied = skb->len; 395 if (copied > len) { 396 msg->msg_flags |= MSG_TRUNC; 397 copied = len; 398 } 399 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 400 if (err) 401 goto out_free_skb; 402 403 sock_recv_timestamp(msg, sk, skb); 404 405 serr = SKB_EXT_ERR(skb); 406 407 sin = (struct sockaddr_in *)msg->msg_name; 408 if (sin) { 409 sin->sin_family = AF_INET; 410 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 411 serr->addr_offset); 412 sin->sin_port = serr->port; 413 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 414 } 415 416 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 417 sin = &errhdr.offender; 418 sin->sin_family = AF_UNSPEC; 419 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) { 420 struct inet_sock *inet = inet_sk(sk); 421 422 sin->sin_family = AF_INET; 423 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 424 sin->sin_port = 0; 425 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 426 if (inet->cmsg_flags) 427 ip_cmsg_recv(msg, skb); 428 } 429 430 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 431 432 /* Now we could try to dump offended packet options */ 433 434 msg->msg_flags |= MSG_ERRQUEUE; 435 err = copied; 436 437 /* Reset and regenerate socket error */ 438 spin_lock_bh(&sk->sk_error_queue.lock); 439 sk->sk_err = 0; 440 skb2 = skb_peek(&sk->sk_error_queue); 441 if (skb2 != NULL) { 442 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno; 443 spin_unlock_bh(&sk->sk_error_queue.lock); 444 sk->sk_error_report(sk); 445 } else 446 spin_unlock_bh(&sk->sk_error_queue.lock); 447 448 out_free_skb: 449 kfree_skb(skb); 450 out: 451 return err; 452 } 453 454 455 static void opt_kfree_rcu(struct rcu_head *head) 456 { 457 kfree(container_of(head, struct ip_options_rcu, rcu)); 458 } 459 460 /* 461 * Socket option code for IP. This is the end of the line after any 462 * TCP,UDP etc options on an IP socket. 463 */ 464 465 static int do_ip_setsockopt(struct sock *sk, int level, 466 int optname, char __user *optval, unsigned int optlen) 467 { 468 struct inet_sock *inet = inet_sk(sk); 469 int val = 0, err; 470 471 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) | 472 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) | 473 (1<<IP_RETOPTS) | (1<<IP_TOS) | 474 (1<<IP_TTL) | (1<<IP_HDRINCL) | 475 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) | 476 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) | 477 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) | 478 (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) || 479 optname == IP_MULTICAST_TTL || 480 optname == IP_MULTICAST_ALL || 481 optname == IP_MULTICAST_LOOP || 482 optname == IP_RECVORIGDSTADDR) { 483 if (optlen >= sizeof(int)) { 484 if (get_user(val, (int __user *) optval)) 485 return -EFAULT; 486 } else if (optlen >= sizeof(char)) { 487 unsigned char ucval; 488 489 if (get_user(ucval, (unsigned char __user *) optval)) 490 return -EFAULT; 491 val = (int) ucval; 492 } 493 } 494 495 /* If optlen==0, it is equivalent to val == 0 */ 496 497 if (ip_mroute_opt(optname)) 498 return ip_mroute_setsockopt(sk, optname, optval, optlen); 499 500 err = 0; 501 lock_sock(sk); 502 503 switch (optname) { 504 case IP_OPTIONS: 505 { 506 struct ip_options_rcu *old, *opt = NULL; 507 508 if (optlen > 40) 509 goto e_inval; 510 err = ip_options_get_from_user(sock_net(sk), &opt, 511 optval, optlen); 512 if (err) 513 break; 514 old = rcu_dereference_protected(inet->inet_opt, 515 sock_owned_by_user(sk)); 516 if (inet->is_icsk) { 517 struct inet_connection_sock *icsk = inet_csk(sk); 518 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 519 if (sk->sk_family == PF_INET || 520 (!((1 << sk->sk_state) & 521 (TCPF_LISTEN | TCPF_CLOSE)) && 522 inet->inet_daddr != LOOPBACK4_IPV6)) { 523 #endif 524 if (old) 525 icsk->icsk_ext_hdr_len -= old->opt.optlen; 526 if (opt) 527 icsk->icsk_ext_hdr_len += opt->opt.optlen; 528 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 529 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) 530 } 531 #endif 532 } 533 rcu_assign_pointer(inet->inet_opt, opt); 534 if (old) 535 call_rcu(&old->rcu, opt_kfree_rcu); 536 break; 537 } 538 case IP_PKTINFO: 539 if (val) 540 inet->cmsg_flags |= IP_CMSG_PKTINFO; 541 else 542 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 543 break; 544 case IP_RECVTTL: 545 if (val) 546 inet->cmsg_flags |= IP_CMSG_TTL; 547 else 548 inet->cmsg_flags &= ~IP_CMSG_TTL; 549 break; 550 case IP_RECVTOS: 551 if (val) 552 inet->cmsg_flags |= IP_CMSG_TOS; 553 else 554 inet->cmsg_flags &= ~IP_CMSG_TOS; 555 break; 556 case IP_RECVOPTS: 557 if (val) 558 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 559 else 560 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 561 break; 562 case IP_RETOPTS: 563 if (val) 564 inet->cmsg_flags |= IP_CMSG_RETOPTS; 565 else 566 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 567 break; 568 case IP_PASSSEC: 569 if (val) 570 inet->cmsg_flags |= IP_CMSG_PASSSEC; 571 else 572 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 573 break; 574 case IP_RECVORIGDSTADDR: 575 if (val) 576 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 577 else 578 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 579 break; 580 case IP_TOS: /* This sets both TOS and Precedence */ 581 if (sk->sk_type == SOCK_STREAM) { 582 val &= ~INET_ECN_MASK; 583 val |= inet->tos & INET_ECN_MASK; 584 } 585 if (inet->tos != val) { 586 inet->tos = val; 587 sk->sk_priority = rt_tos2priority(val); 588 sk_dst_reset(sk); 589 } 590 break; 591 case IP_TTL: 592 if (optlen < 1) 593 goto e_inval; 594 if (val != -1 && (val < 0 || val > 255)) 595 goto e_inval; 596 inet->uc_ttl = val; 597 break; 598 case IP_HDRINCL: 599 if (sk->sk_type != SOCK_RAW) { 600 err = -ENOPROTOOPT; 601 break; 602 } 603 inet->hdrincl = val ? 1 : 0; 604 break; 605 case IP_NODEFRAG: 606 if (sk->sk_type != SOCK_RAW) { 607 err = -ENOPROTOOPT; 608 break; 609 } 610 inet->nodefrag = val ? 1 : 0; 611 break; 612 case IP_MTU_DISCOVER: 613 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE) 614 goto e_inval; 615 inet->pmtudisc = val; 616 break; 617 case IP_RECVERR: 618 inet->recverr = !!val; 619 if (!val) 620 skb_queue_purge(&sk->sk_error_queue); 621 break; 622 case IP_MULTICAST_TTL: 623 if (sk->sk_type == SOCK_STREAM) 624 goto e_inval; 625 if (optlen < 1) 626 goto e_inval; 627 if (val == -1) 628 val = 1; 629 if (val < 0 || val > 255) 630 goto e_inval; 631 inet->mc_ttl = val; 632 break; 633 case IP_MULTICAST_LOOP: 634 if (optlen < 1) 635 goto e_inval; 636 inet->mc_loop = !!val; 637 break; 638 case IP_MULTICAST_IF: 639 { 640 struct ip_mreqn mreq; 641 struct net_device *dev = NULL; 642 643 if (sk->sk_type == SOCK_STREAM) 644 goto e_inval; 645 /* 646 * Check the arguments are allowable 647 */ 648 649 if (optlen < sizeof(struct in_addr)) 650 goto e_inval; 651 652 err = -EFAULT; 653 if (optlen >= sizeof(struct ip_mreqn)) { 654 if (copy_from_user(&mreq, optval, sizeof(mreq))) 655 break; 656 } else { 657 memset(&mreq, 0, sizeof(mreq)); 658 if (optlen >= sizeof(struct in_addr) && 659 copy_from_user(&mreq.imr_address, optval, 660 sizeof(struct in_addr))) 661 break; 662 } 663 664 if (!mreq.imr_ifindex) { 665 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 666 inet->mc_index = 0; 667 inet->mc_addr = 0; 668 err = 0; 669 break; 670 } 671 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 672 if (dev) 673 mreq.imr_ifindex = dev->ifindex; 674 } else 675 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 676 677 678 err = -EADDRNOTAVAIL; 679 if (!dev) 680 break; 681 dev_put(dev); 682 683 err = -EINVAL; 684 if (sk->sk_bound_dev_if && 685 mreq.imr_ifindex != sk->sk_bound_dev_if) 686 break; 687 688 inet->mc_index = mreq.imr_ifindex; 689 inet->mc_addr = mreq.imr_address.s_addr; 690 err = 0; 691 break; 692 } 693 694 case IP_ADD_MEMBERSHIP: 695 case IP_DROP_MEMBERSHIP: 696 { 697 struct ip_mreqn mreq; 698 699 err = -EPROTO; 700 if (inet_sk(sk)->is_icsk) 701 break; 702 703 if (optlen < sizeof(struct ip_mreq)) 704 goto e_inval; 705 err = -EFAULT; 706 if (optlen >= sizeof(struct ip_mreqn)) { 707 if (copy_from_user(&mreq, optval, sizeof(mreq))) 708 break; 709 } else { 710 memset(&mreq, 0, sizeof(mreq)); 711 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 712 break; 713 } 714 715 if (optname == IP_ADD_MEMBERSHIP) 716 err = ip_mc_join_group(sk, &mreq); 717 else 718 err = ip_mc_leave_group(sk, &mreq); 719 break; 720 } 721 case IP_MSFILTER: 722 { 723 struct ip_msfilter *msf; 724 725 if (optlen < IP_MSFILTER_SIZE(0)) 726 goto e_inval; 727 if (optlen > sysctl_optmem_max) { 728 err = -ENOBUFS; 729 break; 730 } 731 msf = kmalloc(optlen, GFP_KERNEL); 732 if (!msf) { 733 err = -ENOBUFS; 734 break; 735 } 736 err = -EFAULT; 737 if (copy_from_user(msf, optval, optlen)) { 738 kfree(msf); 739 break; 740 } 741 /* numsrc >= (1G-4) overflow in 32 bits */ 742 if (msf->imsf_numsrc >= 0x3ffffffcU || 743 msf->imsf_numsrc > sysctl_igmp_max_msf) { 744 kfree(msf); 745 err = -ENOBUFS; 746 break; 747 } 748 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 749 kfree(msf); 750 err = -EINVAL; 751 break; 752 } 753 err = ip_mc_msfilter(sk, msf, 0); 754 kfree(msf); 755 break; 756 } 757 case IP_BLOCK_SOURCE: 758 case IP_UNBLOCK_SOURCE: 759 case IP_ADD_SOURCE_MEMBERSHIP: 760 case IP_DROP_SOURCE_MEMBERSHIP: 761 { 762 struct ip_mreq_source mreqs; 763 int omode, add; 764 765 if (optlen != sizeof(struct ip_mreq_source)) 766 goto e_inval; 767 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 768 err = -EFAULT; 769 break; 770 } 771 if (optname == IP_BLOCK_SOURCE) { 772 omode = MCAST_EXCLUDE; 773 add = 1; 774 } else if (optname == IP_UNBLOCK_SOURCE) { 775 omode = MCAST_EXCLUDE; 776 add = 0; 777 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 778 struct ip_mreqn mreq; 779 780 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 781 mreq.imr_address.s_addr = mreqs.imr_interface; 782 mreq.imr_ifindex = 0; 783 err = ip_mc_join_group(sk, &mreq); 784 if (err && err != -EADDRINUSE) 785 break; 786 omode = MCAST_INCLUDE; 787 add = 1; 788 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 789 omode = MCAST_INCLUDE; 790 add = 0; 791 } 792 err = ip_mc_source(add, omode, sk, &mreqs, 0); 793 break; 794 } 795 case MCAST_JOIN_GROUP: 796 case MCAST_LEAVE_GROUP: 797 { 798 struct group_req greq; 799 struct sockaddr_in *psin; 800 struct ip_mreqn mreq; 801 802 if (optlen < sizeof(struct group_req)) 803 goto e_inval; 804 err = -EFAULT; 805 if (copy_from_user(&greq, optval, sizeof(greq))) 806 break; 807 psin = (struct sockaddr_in *)&greq.gr_group; 808 if (psin->sin_family != AF_INET) 809 goto e_inval; 810 memset(&mreq, 0, sizeof(mreq)); 811 mreq.imr_multiaddr = psin->sin_addr; 812 mreq.imr_ifindex = greq.gr_interface; 813 814 if (optname == MCAST_JOIN_GROUP) 815 err = ip_mc_join_group(sk, &mreq); 816 else 817 err = ip_mc_leave_group(sk, &mreq); 818 break; 819 } 820 case MCAST_JOIN_SOURCE_GROUP: 821 case MCAST_LEAVE_SOURCE_GROUP: 822 case MCAST_BLOCK_SOURCE: 823 case MCAST_UNBLOCK_SOURCE: 824 { 825 struct group_source_req greqs; 826 struct ip_mreq_source mreqs; 827 struct sockaddr_in *psin; 828 int omode, add; 829 830 if (optlen != sizeof(struct group_source_req)) 831 goto e_inval; 832 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 833 err = -EFAULT; 834 break; 835 } 836 if (greqs.gsr_group.ss_family != AF_INET || 837 greqs.gsr_source.ss_family != AF_INET) { 838 err = -EADDRNOTAVAIL; 839 break; 840 } 841 psin = (struct sockaddr_in *)&greqs.gsr_group; 842 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 843 psin = (struct sockaddr_in *)&greqs.gsr_source; 844 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 845 mreqs.imr_interface = 0; /* use index for mc_source */ 846 847 if (optname == MCAST_BLOCK_SOURCE) { 848 omode = MCAST_EXCLUDE; 849 add = 1; 850 } else if (optname == MCAST_UNBLOCK_SOURCE) { 851 omode = MCAST_EXCLUDE; 852 add = 0; 853 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 854 struct ip_mreqn mreq; 855 856 psin = (struct sockaddr_in *)&greqs.gsr_group; 857 mreq.imr_multiaddr = psin->sin_addr; 858 mreq.imr_address.s_addr = 0; 859 mreq.imr_ifindex = greqs.gsr_interface; 860 err = ip_mc_join_group(sk, &mreq); 861 if (err && err != -EADDRINUSE) 862 break; 863 greqs.gsr_interface = mreq.imr_ifindex; 864 omode = MCAST_INCLUDE; 865 add = 1; 866 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 867 omode = MCAST_INCLUDE; 868 add = 0; 869 } 870 err = ip_mc_source(add, omode, sk, &mreqs, 871 greqs.gsr_interface); 872 break; 873 } 874 case MCAST_MSFILTER: 875 { 876 struct sockaddr_in *psin; 877 struct ip_msfilter *msf = NULL; 878 struct group_filter *gsf = NULL; 879 int msize, i, ifindex; 880 881 if (optlen < GROUP_FILTER_SIZE(0)) 882 goto e_inval; 883 if (optlen > sysctl_optmem_max) { 884 err = -ENOBUFS; 885 break; 886 } 887 gsf = kmalloc(optlen, GFP_KERNEL); 888 if (!gsf) { 889 err = -ENOBUFS; 890 break; 891 } 892 err = -EFAULT; 893 if (copy_from_user(gsf, optval, optlen)) 894 goto mc_msf_out; 895 896 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 897 if (gsf->gf_numsrc >= 0x1ffffff || 898 gsf->gf_numsrc > sysctl_igmp_max_msf) { 899 err = -ENOBUFS; 900 goto mc_msf_out; 901 } 902 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 903 err = -EINVAL; 904 goto mc_msf_out; 905 } 906 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 907 msf = kmalloc(msize, GFP_KERNEL); 908 if (!msf) { 909 err = -ENOBUFS; 910 goto mc_msf_out; 911 } 912 ifindex = gsf->gf_interface; 913 psin = (struct sockaddr_in *)&gsf->gf_group; 914 if (psin->sin_family != AF_INET) { 915 err = -EADDRNOTAVAIL; 916 goto mc_msf_out; 917 } 918 msf->imsf_multiaddr = psin->sin_addr.s_addr; 919 msf->imsf_interface = 0; 920 msf->imsf_fmode = gsf->gf_fmode; 921 msf->imsf_numsrc = gsf->gf_numsrc; 922 err = -EADDRNOTAVAIL; 923 for (i = 0; i < gsf->gf_numsrc; ++i) { 924 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 925 926 if (psin->sin_family != AF_INET) 927 goto mc_msf_out; 928 msf->imsf_slist[i] = psin->sin_addr.s_addr; 929 } 930 kfree(gsf); 931 gsf = NULL; 932 933 err = ip_mc_msfilter(sk, msf, ifindex); 934 mc_msf_out: 935 kfree(msf); 936 kfree(gsf); 937 break; 938 } 939 case IP_MULTICAST_ALL: 940 if (optlen < 1) 941 goto e_inval; 942 if (val != 0 && val != 1) 943 goto e_inval; 944 inet->mc_all = val; 945 break; 946 case IP_ROUTER_ALERT: 947 err = ip_ra_control(sk, val ? 1 : 0, NULL); 948 break; 949 950 case IP_FREEBIND: 951 if (optlen < 1) 952 goto e_inval; 953 inet->freebind = !!val; 954 break; 955 956 case IP_IPSEC_POLICY: 957 case IP_XFRM_POLICY: 958 err = -EPERM; 959 if (!capable(CAP_NET_ADMIN)) 960 break; 961 err = xfrm_user_policy(sk, optname, optval, optlen); 962 break; 963 964 case IP_TRANSPARENT: 965 if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) { 966 err = -EPERM; 967 break; 968 } 969 if (optlen < 1) 970 goto e_inval; 971 inet->transparent = !!val; 972 break; 973 974 case IP_MINTTL: 975 if (optlen < 1) 976 goto e_inval; 977 if (val < 0 || val > 255) 978 goto e_inval; 979 inet->min_ttl = val; 980 break; 981 982 default: 983 err = -ENOPROTOOPT; 984 break; 985 } 986 release_sock(sk); 987 return err; 988 989 e_inval: 990 release_sock(sk); 991 return -EINVAL; 992 } 993 994 /** 995 * ip_queue_rcv_skb - Queue an skb into sock receive queue 996 * @sk: socket 997 * @skb: buffer 998 * 999 * Queues an skb into socket receive queue. If IP_CMSG_PKTINFO option 1000 * is not set, we drop skb dst entry now, while dst cache line is hot. 1001 */ 1002 int ip_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) 1003 { 1004 if (!(inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO)) 1005 skb_dst_drop(skb); 1006 return sock_queue_rcv_skb(sk, skb); 1007 } 1008 EXPORT_SYMBOL(ip_queue_rcv_skb); 1009 1010 int ip_setsockopt(struct sock *sk, int level, 1011 int optname, char __user *optval, unsigned int optlen) 1012 { 1013 int err; 1014 1015 if (level != SOL_IP) 1016 return -ENOPROTOOPT; 1017 1018 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1019 #ifdef CONFIG_NETFILTER 1020 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1021 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1022 optname != IP_IPSEC_POLICY && 1023 optname != IP_XFRM_POLICY && 1024 !ip_mroute_opt(optname)) { 1025 lock_sock(sk); 1026 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1027 release_sock(sk); 1028 } 1029 #endif 1030 return err; 1031 } 1032 EXPORT_SYMBOL(ip_setsockopt); 1033 1034 #ifdef CONFIG_COMPAT 1035 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1036 char __user *optval, unsigned int optlen) 1037 { 1038 int err; 1039 1040 if (level != SOL_IP) 1041 return -ENOPROTOOPT; 1042 1043 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1044 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1045 ip_setsockopt); 1046 1047 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1048 #ifdef CONFIG_NETFILTER 1049 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1050 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1051 optname != IP_IPSEC_POLICY && 1052 optname != IP_XFRM_POLICY && 1053 !ip_mroute_opt(optname)) { 1054 lock_sock(sk); 1055 err = compat_nf_setsockopt(sk, PF_INET, optname, 1056 optval, optlen); 1057 release_sock(sk); 1058 } 1059 #endif 1060 return err; 1061 } 1062 EXPORT_SYMBOL(compat_ip_setsockopt); 1063 #endif 1064 1065 /* 1066 * Get the options. Note for future reference. The GET of IP options gets 1067 * the _received_ ones. The set sets the _sent_ ones. 1068 */ 1069 1070 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1071 char __user *optval, int __user *optlen, unsigned flags) 1072 { 1073 struct inet_sock *inet = inet_sk(sk); 1074 int val; 1075 int len; 1076 1077 if (level != SOL_IP) 1078 return -EOPNOTSUPP; 1079 1080 if (ip_mroute_opt(optname)) 1081 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1082 1083 if (get_user(len, optlen)) 1084 return -EFAULT; 1085 if (len < 0) 1086 return -EINVAL; 1087 1088 lock_sock(sk); 1089 1090 switch (optname) { 1091 case IP_OPTIONS: 1092 { 1093 unsigned char optbuf[sizeof(struct ip_options)+40]; 1094 struct ip_options *opt = (struct ip_options *)optbuf; 1095 struct ip_options_rcu *inet_opt; 1096 1097 inet_opt = rcu_dereference_protected(inet->inet_opt, 1098 sock_owned_by_user(sk)); 1099 opt->optlen = 0; 1100 if (inet_opt) 1101 memcpy(optbuf, &inet_opt->opt, 1102 sizeof(struct ip_options) + 1103 inet_opt->opt.optlen); 1104 release_sock(sk); 1105 1106 if (opt->optlen == 0) 1107 return put_user(0, optlen); 1108 1109 ip_options_undo(opt); 1110 1111 len = min_t(unsigned int, len, opt->optlen); 1112 if (put_user(len, optlen)) 1113 return -EFAULT; 1114 if (copy_to_user(optval, opt->__data, len)) 1115 return -EFAULT; 1116 return 0; 1117 } 1118 case IP_PKTINFO: 1119 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1120 break; 1121 case IP_RECVTTL: 1122 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1123 break; 1124 case IP_RECVTOS: 1125 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1126 break; 1127 case IP_RECVOPTS: 1128 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1129 break; 1130 case IP_RETOPTS: 1131 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1132 break; 1133 case IP_PASSSEC: 1134 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1135 break; 1136 case IP_RECVORIGDSTADDR: 1137 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1138 break; 1139 case IP_TOS: 1140 val = inet->tos; 1141 break; 1142 case IP_TTL: 1143 val = (inet->uc_ttl == -1 ? 1144 sysctl_ip_default_ttl : 1145 inet->uc_ttl); 1146 break; 1147 case IP_HDRINCL: 1148 val = inet->hdrincl; 1149 break; 1150 case IP_NODEFRAG: 1151 val = inet->nodefrag; 1152 break; 1153 case IP_MTU_DISCOVER: 1154 val = inet->pmtudisc; 1155 break; 1156 case IP_MTU: 1157 { 1158 struct dst_entry *dst; 1159 val = 0; 1160 dst = sk_dst_get(sk); 1161 if (dst) { 1162 val = dst_mtu(dst); 1163 dst_release(dst); 1164 } 1165 if (!val) { 1166 release_sock(sk); 1167 return -ENOTCONN; 1168 } 1169 break; 1170 } 1171 case IP_RECVERR: 1172 val = inet->recverr; 1173 break; 1174 case IP_MULTICAST_TTL: 1175 val = inet->mc_ttl; 1176 break; 1177 case IP_MULTICAST_LOOP: 1178 val = inet->mc_loop; 1179 break; 1180 case IP_MULTICAST_IF: 1181 { 1182 struct in_addr addr; 1183 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1184 addr.s_addr = inet->mc_addr; 1185 release_sock(sk); 1186 1187 if (put_user(len, optlen)) 1188 return -EFAULT; 1189 if (copy_to_user(optval, &addr, len)) 1190 return -EFAULT; 1191 return 0; 1192 } 1193 case IP_MSFILTER: 1194 { 1195 struct ip_msfilter msf; 1196 int err; 1197 1198 if (len < IP_MSFILTER_SIZE(0)) { 1199 release_sock(sk); 1200 return -EINVAL; 1201 } 1202 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1203 release_sock(sk); 1204 return -EFAULT; 1205 } 1206 err = ip_mc_msfget(sk, &msf, 1207 (struct ip_msfilter __user *)optval, optlen); 1208 release_sock(sk); 1209 return err; 1210 } 1211 case MCAST_MSFILTER: 1212 { 1213 struct group_filter gsf; 1214 int err; 1215 1216 if (len < GROUP_FILTER_SIZE(0)) { 1217 release_sock(sk); 1218 return -EINVAL; 1219 } 1220 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1221 release_sock(sk); 1222 return -EFAULT; 1223 } 1224 err = ip_mc_gsfget(sk, &gsf, 1225 (struct group_filter __user *)optval, 1226 optlen); 1227 release_sock(sk); 1228 return err; 1229 } 1230 case IP_MULTICAST_ALL: 1231 val = inet->mc_all; 1232 break; 1233 case IP_PKTOPTIONS: 1234 { 1235 struct msghdr msg; 1236 1237 release_sock(sk); 1238 1239 if (sk->sk_type != SOCK_STREAM) 1240 return -ENOPROTOOPT; 1241 1242 msg.msg_control = optval; 1243 msg.msg_controllen = len; 1244 msg.msg_flags = flags; 1245 1246 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1247 struct in_pktinfo info; 1248 1249 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1250 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1251 info.ipi_ifindex = inet->mc_index; 1252 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1253 } 1254 if (inet->cmsg_flags & IP_CMSG_TTL) { 1255 int hlim = inet->mc_ttl; 1256 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1257 } 1258 len -= msg.msg_controllen; 1259 return put_user(len, optlen); 1260 } 1261 case IP_FREEBIND: 1262 val = inet->freebind; 1263 break; 1264 case IP_TRANSPARENT: 1265 val = inet->transparent; 1266 break; 1267 case IP_MINTTL: 1268 val = inet->min_ttl; 1269 break; 1270 default: 1271 release_sock(sk); 1272 return -ENOPROTOOPT; 1273 } 1274 release_sock(sk); 1275 1276 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1277 unsigned char ucval = (unsigned char)val; 1278 len = 1; 1279 if (put_user(len, optlen)) 1280 return -EFAULT; 1281 if (copy_to_user(optval, &ucval, 1)) 1282 return -EFAULT; 1283 } else { 1284 len = min_t(unsigned int, sizeof(int), len); 1285 if (put_user(len, optlen)) 1286 return -EFAULT; 1287 if (copy_to_user(optval, &val, len)) 1288 return -EFAULT; 1289 } 1290 return 0; 1291 } 1292 1293 int ip_getsockopt(struct sock *sk, int level, 1294 int optname, char __user *optval, int __user *optlen) 1295 { 1296 int err; 1297 1298 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1299 #ifdef CONFIG_NETFILTER 1300 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1301 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1302 !ip_mroute_opt(optname)) { 1303 int len; 1304 1305 if (get_user(len, optlen)) 1306 return -EFAULT; 1307 1308 lock_sock(sk); 1309 err = nf_getsockopt(sk, PF_INET, optname, optval, 1310 &len); 1311 release_sock(sk); 1312 if (err >= 0) 1313 err = put_user(len, optlen); 1314 return err; 1315 } 1316 #endif 1317 return err; 1318 } 1319 EXPORT_SYMBOL(ip_getsockopt); 1320 1321 #ifdef CONFIG_COMPAT 1322 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1323 char __user *optval, int __user *optlen) 1324 { 1325 int err; 1326 1327 if (optname == MCAST_MSFILTER) 1328 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1329 ip_getsockopt); 1330 1331 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1332 MSG_CMSG_COMPAT); 1333 1334 #ifdef CONFIG_NETFILTER 1335 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1336 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1337 !ip_mroute_opt(optname)) { 1338 int len; 1339 1340 if (get_user(len, optlen)) 1341 return -EFAULT; 1342 1343 lock_sock(sk); 1344 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1345 release_sock(sk); 1346 if (err >= 0) 1347 err = put_user(len, optlen); 1348 return err; 1349 } 1350 #endif 1351 return err; 1352 } 1353 EXPORT_SYMBOL(compat_ip_getsockopt); 1354 #endif 1355