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 IS_ENABLED(CONFIG_IPV6) 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 #define PKTINFO_SKB_CB(__skb) ((struct in_pktinfo *)((__skb)->cb)) 59 60 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) 61 { 62 struct in_pktinfo info = *PKTINFO_SKB_CB(skb); 63 64 info.ipi_addr.s_addr = ip_hdr(skb)->daddr; 65 66 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 67 } 68 69 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb) 70 { 71 int ttl = ip_hdr(skb)->ttl; 72 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl); 73 } 74 75 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb) 76 { 77 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos); 78 } 79 80 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb) 81 { 82 if (IPCB(skb)->opt.optlen == 0) 83 return; 84 85 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen, 86 ip_hdr(skb) + 1); 87 } 88 89 90 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb) 91 { 92 unsigned char optbuf[sizeof(struct ip_options) + 40]; 93 struct ip_options * opt = (struct ip_options *)optbuf; 94 95 if (IPCB(skb)->opt.optlen == 0) 96 return; 97 98 if (ip_options_echo(opt, skb)) { 99 msg->msg_flags |= MSG_CTRUNC; 100 return; 101 } 102 ip_options_undo(opt); 103 104 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data); 105 } 106 107 static void ip_cmsg_recv_security(struct msghdr *msg, struct sk_buff *skb) 108 { 109 char *secdata; 110 u32 seclen, secid; 111 int err; 112 113 err = security_socket_getpeersec_dgram(NULL, skb, &secid); 114 if (err) 115 return; 116 117 err = security_secid_to_secctx(secid, &secdata, &seclen); 118 if (err) 119 return; 120 121 put_cmsg(msg, SOL_IP, SCM_SECURITY, seclen, secdata); 122 security_release_secctx(secdata, seclen); 123 } 124 125 static void ip_cmsg_recv_dstaddr(struct msghdr *msg, struct sk_buff *skb) 126 { 127 struct sockaddr_in sin; 128 const struct iphdr *iph = ip_hdr(skb); 129 __be16 *ports = (__be16 *)skb_transport_header(skb); 130 131 if (skb_transport_offset(skb) + 4 > skb->len) 132 return; 133 134 /* All current transport protocols have the port numbers in the 135 * first four bytes of the transport header and this function is 136 * written with this assumption in mind. 137 */ 138 139 sin.sin_family = AF_INET; 140 sin.sin_addr.s_addr = iph->daddr; 141 sin.sin_port = ports[1]; 142 memset(sin.sin_zero, 0, sizeof(sin.sin_zero)); 143 144 put_cmsg(msg, SOL_IP, IP_ORIGDSTADDR, sizeof(sin), &sin); 145 } 146 147 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb) 148 { 149 struct inet_sock *inet = inet_sk(skb->sk); 150 unsigned flags = inet->cmsg_flags; 151 152 /* Ordered by supposed usage frequency */ 153 if (flags & 1) 154 ip_cmsg_recv_pktinfo(msg, skb); 155 if ((flags >>= 1) == 0) 156 return; 157 158 if (flags & 1) 159 ip_cmsg_recv_ttl(msg, skb); 160 if ((flags >>= 1) == 0) 161 return; 162 163 if (flags & 1) 164 ip_cmsg_recv_tos(msg, skb); 165 if ((flags >>= 1) == 0) 166 return; 167 168 if (flags & 1) 169 ip_cmsg_recv_opts(msg, skb); 170 if ((flags >>= 1) == 0) 171 return; 172 173 if (flags & 1) 174 ip_cmsg_recv_retopts(msg, skb); 175 if ((flags >>= 1) == 0) 176 return; 177 178 if (flags & 1) 179 ip_cmsg_recv_security(msg, skb); 180 181 if ((flags >>= 1) == 0) 182 return; 183 if (flags & 1) 184 ip_cmsg_recv_dstaddr(msg, skb); 185 186 } 187 EXPORT_SYMBOL(ip_cmsg_recv); 188 189 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc) 190 { 191 int err; 192 struct cmsghdr *cmsg; 193 194 for (cmsg = CMSG_FIRSTHDR(msg); cmsg; cmsg = CMSG_NXTHDR(msg, cmsg)) { 195 if (!CMSG_OK(msg, cmsg)) 196 return -EINVAL; 197 if (cmsg->cmsg_level != SOL_IP) 198 continue; 199 switch (cmsg->cmsg_type) { 200 case IP_RETOPTS: 201 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr)); 202 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), 203 err < 40 ? err : 40); 204 if (err) 205 return err; 206 break; 207 case IP_PKTINFO: 208 { 209 struct in_pktinfo *info; 210 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) 211 return -EINVAL; 212 info = (struct in_pktinfo *)CMSG_DATA(cmsg); 213 ipc->oif = info->ipi_ifindex; 214 ipc->addr = info->ipi_spec_dst.s_addr; 215 break; 216 } 217 default: 218 return -EINVAL; 219 } 220 } 221 return 0; 222 } 223 224 225 /* Special input handler for packets caught by router alert option. 226 They are selected only by protocol field, and then processed likely 227 local ones; but only if someone wants them! Otherwise, router 228 not running rsvpd will kill RSVP. 229 230 It is user level problem, what it will make with them. 231 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)), 232 but receiver should be enough clever f.e. to forward mtrace requests, 233 sent to multicast group to reach destination designated router. 234 */ 235 struct ip_ra_chain __rcu *ip_ra_chain; 236 static DEFINE_SPINLOCK(ip_ra_lock); 237 238 239 static void ip_ra_destroy_rcu(struct rcu_head *head) 240 { 241 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu); 242 243 sock_put(ra->saved_sk); 244 kfree(ra); 245 } 246 247 int ip_ra_control(struct sock *sk, unsigned char on, 248 void (*destructor)(struct sock *)) 249 { 250 struct ip_ra_chain *ra, *new_ra; 251 struct ip_ra_chain __rcu **rap; 252 253 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 254 return -EINVAL; 255 256 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 257 258 spin_lock_bh(&ip_ra_lock); 259 for (rap = &ip_ra_chain; 260 (ra = rcu_dereference_protected(*rap, 261 lockdep_is_held(&ip_ra_lock))) != NULL; 262 rap = &ra->next) { 263 if (ra->sk == sk) { 264 if (on) { 265 spin_unlock_bh(&ip_ra_lock); 266 kfree(new_ra); 267 return -EADDRINUSE; 268 } 269 /* dont let ip_call_ra_chain() use sk again */ 270 ra->sk = NULL; 271 rcu_assign_pointer(*rap, ra->next); 272 spin_unlock_bh(&ip_ra_lock); 273 274 if (ra->destructor) 275 ra->destructor(sk); 276 /* 277 * Delay sock_put(sk) and kfree(ra) after one rcu grace 278 * period. This guarantee ip_call_ra_chain() dont need 279 * to mess with socket refcounts. 280 */ 281 ra->saved_sk = sk; 282 call_rcu(&ra->rcu, ip_ra_destroy_rcu); 283 return 0; 284 } 285 } 286 if (new_ra == NULL) { 287 spin_unlock_bh(&ip_ra_lock); 288 return -ENOBUFS; 289 } 290 new_ra->sk = sk; 291 new_ra->destructor = destructor; 292 293 new_ra->next = ra; 294 rcu_assign_pointer(*rap, new_ra); 295 sock_hold(sk); 296 spin_unlock_bh(&ip_ra_lock); 297 298 return 0; 299 } 300 301 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 302 __be16 port, u32 info, u8 *payload) 303 { 304 struct sock_exterr_skb *serr; 305 306 skb = skb_clone(skb, GFP_ATOMIC); 307 if (!skb) 308 return; 309 310 serr = SKB_EXT_ERR(skb); 311 serr->ee.ee_errno = err; 312 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 313 serr->ee.ee_type = icmp_hdr(skb)->type; 314 serr->ee.ee_code = icmp_hdr(skb)->code; 315 serr->ee.ee_pad = 0; 316 serr->ee.ee_info = info; 317 serr->ee.ee_data = 0; 318 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 319 skb_network_header(skb); 320 serr->port = port; 321 322 if (skb_pull(skb, payload - skb->data) != NULL) { 323 skb_reset_transport_header(skb); 324 if (sock_queue_err_skb(sk, skb) == 0) 325 return; 326 } 327 kfree_skb(skb); 328 } 329 330 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 331 { 332 struct inet_sock *inet = inet_sk(sk); 333 struct sock_exterr_skb *serr; 334 struct iphdr *iph; 335 struct sk_buff *skb; 336 337 if (!inet->recverr) 338 return; 339 340 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 341 if (!skb) 342 return; 343 344 skb_put(skb, sizeof(struct iphdr)); 345 skb_reset_network_header(skb); 346 iph = ip_hdr(skb); 347 iph->daddr = daddr; 348 349 serr = SKB_EXT_ERR(skb); 350 serr->ee.ee_errno = err; 351 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 352 serr->ee.ee_type = 0; 353 serr->ee.ee_code = 0; 354 serr->ee.ee_pad = 0; 355 serr->ee.ee_info = info; 356 serr->ee.ee_data = 0; 357 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 358 serr->port = port; 359 360 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 361 skb_reset_transport_header(skb); 362 363 if (sock_queue_err_skb(sk, skb)) 364 kfree_skb(skb); 365 } 366 367 /* 368 * Handle MSG_ERRQUEUE 369 */ 370 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len) 371 { 372 struct sock_exterr_skb *serr; 373 struct sk_buff *skb, *skb2; 374 struct sockaddr_in *sin; 375 struct { 376 struct sock_extended_err ee; 377 struct sockaddr_in offender; 378 } errhdr; 379 int err; 380 int copied; 381 382 err = -EAGAIN; 383 skb = skb_dequeue(&sk->sk_error_queue); 384 if (skb == NULL) 385 goto out; 386 387 copied = skb->len; 388 if (copied > len) { 389 msg->msg_flags |= MSG_TRUNC; 390 copied = len; 391 } 392 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied); 393 if (err) 394 goto out_free_skb; 395 396 sock_recv_timestamp(msg, sk, skb); 397 398 serr = SKB_EXT_ERR(skb); 399 400 sin = (struct sockaddr_in *)msg->msg_name; 401 if (sin) { 402 sin->sin_family = AF_INET; 403 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 404 serr->addr_offset); 405 sin->sin_port = serr->port; 406 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 407 } 408 409 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 410 sin = &errhdr.offender; 411 sin->sin_family = AF_UNSPEC; 412 if (serr->ee.ee_origin == SO_EE_ORIGIN_ICMP) { 413 struct inet_sock *inet = inet_sk(sk); 414 415 sin->sin_family = AF_INET; 416 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 417 sin->sin_port = 0; 418 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 419 if (inet->cmsg_flags) 420 ip_cmsg_recv(msg, skb); 421 } 422 423 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 424 425 /* Now we could try to dump offended packet options */ 426 427 msg->msg_flags |= MSG_ERRQUEUE; 428 err = copied; 429 430 /* Reset and regenerate socket error */ 431 spin_lock_bh(&sk->sk_error_queue.lock); 432 sk->sk_err = 0; 433 skb2 = skb_peek(&sk->sk_error_queue); 434 if (skb2 != NULL) { 435 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno; 436 spin_unlock_bh(&sk->sk_error_queue.lock); 437 sk->sk_error_report(sk); 438 } else 439 spin_unlock_bh(&sk->sk_error_queue.lock); 440 441 out_free_skb: 442 kfree_skb(skb); 443 out: 444 return err; 445 } 446 447 448 static void opt_kfree_rcu(struct rcu_head *head) 449 { 450 kfree(container_of(head, struct ip_options_rcu, rcu)); 451 } 452 453 /* 454 * Socket option code for IP. This is the end of the line after any 455 * TCP,UDP etc options on an IP socket. 456 */ 457 458 static int do_ip_setsockopt(struct sock *sk, int level, 459 int optname, char __user *optval, unsigned int optlen) 460 { 461 struct inet_sock *inet = inet_sk(sk); 462 int val = 0, err; 463 464 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) | 465 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) | 466 (1<<IP_RETOPTS) | (1<<IP_TOS) | 467 (1<<IP_TTL) | (1<<IP_HDRINCL) | 468 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) | 469 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) | 470 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) | 471 (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) || 472 optname == IP_MULTICAST_TTL || 473 optname == IP_MULTICAST_ALL || 474 optname == IP_MULTICAST_LOOP || 475 optname == IP_RECVORIGDSTADDR) { 476 if (optlen >= sizeof(int)) { 477 if (get_user(val, (int __user *) optval)) 478 return -EFAULT; 479 } else if (optlen >= sizeof(char)) { 480 unsigned char ucval; 481 482 if (get_user(ucval, (unsigned char __user *) optval)) 483 return -EFAULT; 484 val = (int) ucval; 485 } 486 } 487 488 /* If optlen==0, it is equivalent to val == 0 */ 489 490 if (ip_mroute_opt(optname)) 491 return ip_mroute_setsockopt(sk, optname, optval, optlen); 492 493 err = 0; 494 lock_sock(sk); 495 496 switch (optname) { 497 case IP_OPTIONS: 498 { 499 struct ip_options_rcu *old, *opt = NULL; 500 501 if (optlen > 40) 502 goto e_inval; 503 err = ip_options_get_from_user(sock_net(sk), &opt, 504 optval, optlen); 505 if (err) 506 break; 507 old = rcu_dereference_protected(inet->inet_opt, 508 sock_owned_by_user(sk)); 509 if (inet->is_icsk) { 510 struct inet_connection_sock *icsk = inet_csk(sk); 511 #if IS_ENABLED(CONFIG_IPV6) 512 if (sk->sk_family == PF_INET || 513 (!((1 << sk->sk_state) & 514 (TCPF_LISTEN | TCPF_CLOSE)) && 515 inet->inet_daddr != LOOPBACK4_IPV6)) { 516 #endif 517 if (old) 518 icsk->icsk_ext_hdr_len -= old->opt.optlen; 519 if (opt) 520 icsk->icsk_ext_hdr_len += opt->opt.optlen; 521 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 522 #if IS_ENABLED(CONFIG_IPV6) 523 } 524 #endif 525 } 526 rcu_assign_pointer(inet->inet_opt, opt); 527 if (old) 528 call_rcu(&old->rcu, opt_kfree_rcu); 529 break; 530 } 531 case IP_PKTINFO: 532 if (val) 533 inet->cmsg_flags |= IP_CMSG_PKTINFO; 534 else 535 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 536 break; 537 case IP_RECVTTL: 538 if (val) 539 inet->cmsg_flags |= IP_CMSG_TTL; 540 else 541 inet->cmsg_flags &= ~IP_CMSG_TTL; 542 break; 543 case IP_RECVTOS: 544 if (val) 545 inet->cmsg_flags |= IP_CMSG_TOS; 546 else 547 inet->cmsg_flags &= ~IP_CMSG_TOS; 548 break; 549 case IP_RECVOPTS: 550 if (val) 551 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 552 else 553 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 554 break; 555 case IP_RETOPTS: 556 if (val) 557 inet->cmsg_flags |= IP_CMSG_RETOPTS; 558 else 559 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 560 break; 561 case IP_PASSSEC: 562 if (val) 563 inet->cmsg_flags |= IP_CMSG_PASSSEC; 564 else 565 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 566 break; 567 case IP_RECVORIGDSTADDR: 568 if (val) 569 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 570 else 571 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 572 break; 573 case IP_TOS: /* This sets both TOS and Precedence */ 574 if (sk->sk_type == SOCK_STREAM) { 575 val &= ~INET_ECN_MASK; 576 val |= inet->tos & INET_ECN_MASK; 577 } 578 if (inet->tos != val) { 579 inet->tos = val; 580 sk->sk_priority = rt_tos2priority(val); 581 sk_dst_reset(sk); 582 } 583 break; 584 case IP_TTL: 585 if (optlen < 1) 586 goto e_inval; 587 if (val != -1 && (val < 0 || val > 255)) 588 goto e_inval; 589 inet->uc_ttl = val; 590 break; 591 case IP_HDRINCL: 592 if (sk->sk_type != SOCK_RAW) { 593 err = -ENOPROTOOPT; 594 break; 595 } 596 inet->hdrincl = val ? 1 : 0; 597 break; 598 case IP_NODEFRAG: 599 if (sk->sk_type != SOCK_RAW) { 600 err = -ENOPROTOOPT; 601 break; 602 } 603 inet->nodefrag = val ? 1 : 0; 604 break; 605 case IP_MTU_DISCOVER: 606 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE) 607 goto e_inval; 608 inet->pmtudisc = val; 609 break; 610 case IP_RECVERR: 611 inet->recverr = !!val; 612 if (!val) 613 skb_queue_purge(&sk->sk_error_queue); 614 break; 615 case IP_MULTICAST_TTL: 616 if (sk->sk_type == SOCK_STREAM) 617 goto e_inval; 618 if (optlen < 1) 619 goto e_inval; 620 if (val == -1) 621 val = 1; 622 if (val < 0 || val > 255) 623 goto e_inval; 624 inet->mc_ttl = val; 625 break; 626 case IP_MULTICAST_LOOP: 627 if (optlen < 1) 628 goto e_inval; 629 inet->mc_loop = !!val; 630 break; 631 case IP_MULTICAST_IF: 632 { 633 struct ip_mreqn mreq; 634 struct net_device *dev = NULL; 635 636 if (sk->sk_type == SOCK_STREAM) 637 goto e_inval; 638 /* 639 * Check the arguments are allowable 640 */ 641 642 if (optlen < sizeof(struct in_addr)) 643 goto e_inval; 644 645 err = -EFAULT; 646 if (optlen >= sizeof(struct ip_mreqn)) { 647 if (copy_from_user(&mreq, optval, sizeof(mreq))) 648 break; 649 } else { 650 memset(&mreq, 0, sizeof(mreq)); 651 if (optlen >= sizeof(struct in_addr) && 652 copy_from_user(&mreq.imr_address, optval, 653 sizeof(struct in_addr))) 654 break; 655 } 656 657 if (!mreq.imr_ifindex) { 658 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 659 inet->mc_index = 0; 660 inet->mc_addr = 0; 661 err = 0; 662 break; 663 } 664 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 665 if (dev) 666 mreq.imr_ifindex = dev->ifindex; 667 } else 668 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 669 670 671 err = -EADDRNOTAVAIL; 672 if (!dev) 673 break; 674 dev_put(dev); 675 676 err = -EINVAL; 677 if (sk->sk_bound_dev_if && 678 mreq.imr_ifindex != sk->sk_bound_dev_if) 679 break; 680 681 inet->mc_index = mreq.imr_ifindex; 682 inet->mc_addr = mreq.imr_address.s_addr; 683 err = 0; 684 break; 685 } 686 687 case IP_ADD_MEMBERSHIP: 688 case IP_DROP_MEMBERSHIP: 689 { 690 struct ip_mreqn mreq; 691 692 err = -EPROTO; 693 if (inet_sk(sk)->is_icsk) 694 break; 695 696 if (optlen < sizeof(struct ip_mreq)) 697 goto e_inval; 698 err = -EFAULT; 699 if (optlen >= sizeof(struct ip_mreqn)) { 700 if (copy_from_user(&mreq, optval, sizeof(mreq))) 701 break; 702 } else { 703 memset(&mreq, 0, sizeof(mreq)); 704 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 705 break; 706 } 707 708 if (optname == IP_ADD_MEMBERSHIP) 709 err = ip_mc_join_group(sk, &mreq); 710 else 711 err = ip_mc_leave_group(sk, &mreq); 712 break; 713 } 714 case IP_MSFILTER: 715 { 716 struct ip_msfilter *msf; 717 718 if (optlen < IP_MSFILTER_SIZE(0)) 719 goto e_inval; 720 if (optlen > sysctl_optmem_max) { 721 err = -ENOBUFS; 722 break; 723 } 724 msf = kmalloc(optlen, GFP_KERNEL); 725 if (!msf) { 726 err = -ENOBUFS; 727 break; 728 } 729 err = -EFAULT; 730 if (copy_from_user(msf, optval, optlen)) { 731 kfree(msf); 732 break; 733 } 734 /* numsrc >= (1G-4) overflow in 32 bits */ 735 if (msf->imsf_numsrc >= 0x3ffffffcU || 736 msf->imsf_numsrc > sysctl_igmp_max_msf) { 737 kfree(msf); 738 err = -ENOBUFS; 739 break; 740 } 741 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 742 kfree(msf); 743 err = -EINVAL; 744 break; 745 } 746 err = ip_mc_msfilter(sk, msf, 0); 747 kfree(msf); 748 break; 749 } 750 case IP_BLOCK_SOURCE: 751 case IP_UNBLOCK_SOURCE: 752 case IP_ADD_SOURCE_MEMBERSHIP: 753 case IP_DROP_SOURCE_MEMBERSHIP: 754 { 755 struct ip_mreq_source mreqs; 756 int omode, add; 757 758 if (optlen != sizeof(struct ip_mreq_source)) 759 goto e_inval; 760 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 761 err = -EFAULT; 762 break; 763 } 764 if (optname == IP_BLOCK_SOURCE) { 765 omode = MCAST_EXCLUDE; 766 add = 1; 767 } else if (optname == IP_UNBLOCK_SOURCE) { 768 omode = MCAST_EXCLUDE; 769 add = 0; 770 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 771 struct ip_mreqn mreq; 772 773 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 774 mreq.imr_address.s_addr = mreqs.imr_interface; 775 mreq.imr_ifindex = 0; 776 err = ip_mc_join_group(sk, &mreq); 777 if (err && err != -EADDRINUSE) 778 break; 779 omode = MCAST_INCLUDE; 780 add = 1; 781 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 782 omode = MCAST_INCLUDE; 783 add = 0; 784 } 785 err = ip_mc_source(add, omode, sk, &mreqs, 0); 786 break; 787 } 788 case MCAST_JOIN_GROUP: 789 case MCAST_LEAVE_GROUP: 790 { 791 struct group_req greq; 792 struct sockaddr_in *psin; 793 struct ip_mreqn mreq; 794 795 if (optlen < sizeof(struct group_req)) 796 goto e_inval; 797 err = -EFAULT; 798 if (copy_from_user(&greq, optval, sizeof(greq))) 799 break; 800 psin = (struct sockaddr_in *)&greq.gr_group; 801 if (psin->sin_family != AF_INET) 802 goto e_inval; 803 memset(&mreq, 0, sizeof(mreq)); 804 mreq.imr_multiaddr = psin->sin_addr; 805 mreq.imr_ifindex = greq.gr_interface; 806 807 if (optname == MCAST_JOIN_GROUP) 808 err = ip_mc_join_group(sk, &mreq); 809 else 810 err = ip_mc_leave_group(sk, &mreq); 811 break; 812 } 813 case MCAST_JOIN_SOURCE_GROUP: 814 case MCAST_LEAVE_SOURCE_GROUP: 815 case MCAST_BLOCK_SOURCE: 816 case MCAST_UNBLOCK_SOURCE: 817 { 818 struct group_source_req greqs; 819 struct ip_mreq_source mreqs; 820 struct sockaddr_in *psin; 821 int omode, add; 822 823 if (optlen != sizeof(struct group_source_req)) 824 goto e_inval; 825 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 826 err = -EFAULT; 827 break; 828 } 829 if (greqs.gsr_group.ss_family != AF_INET || 830 greqs.gsr_source.ss_family != AF_INET) { 831 err = -EADDRNOTAVAIL; 832 break; 833 } 834 psin = (struct sockaddr_in *)&greqs.gsr_group; 835 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 836 psin = (struct sockaddr_in *)&greqs.gsr_source; 837 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 838 mreqs.imr_interface = 0; /* use index for mc_source */ 839 840 if (optname == MCAST_BLOCK_SOURCE) { 841 omode = MCAST_EXCLUDE; 842 add = 1; 843 } else if (optname == MCAST_UNBLOCK_SOURCE) { 844 omode = MCAST_EXCLUDE; 845 add = 0; 846 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 847 struct ip_mreqn mreq; 848 849 psin = (struct sockaddr_in *)&greqs.gsr_group; 850 mreq.imr_multiaddr = psin->sin_addr; 851 mreq.imr_address.s_addr = 0; 852 mreq.imr_ifindex = greqs.gsr_interface; 853 err = ip_mc_join_group(sk, &mreq); 854 if (err && err != -EADDRINUSE) 855 break; 856 greqs.gsr_interface = mreq.imr_ifindex; 857 omode = MCAST_INCLUDE; 858 add = 1; 859 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 860 omode = MCAST_INCLUDE; 861 add = 0; 862 } 863 err = ip_mc_source(add, omode, sk, &mreqs, 864 greqs.gsr_interface); 865 break; 866 } 867 case MCAST_MSFILTER: 868 { 869 struct sockaddr_in *psin; 870 struct ip_msfilter *msf = NULL; 871 struct group_filter *gsf = NULL; 872 int msize, i, ifindex; 873 874 if (optlen < GROUP_FILTER_SIZE(0)) 875 goto e_inval; 876 if (optlen > sysctl_optmem_max) { 877 err = -ENOBUFS; 878 break; 879 } 880 gsf = kmalloc(optlen, GFP_KERNEL); 881 if (!gsf) { 882 err = -ENOBUFS; 883 break; 884 } 885 err = -EFAULT; 886 if (copy_from_user(gsf, optval, optlen)) 887 goto mc_msf_out; 888 889 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 890 if (gsf->gf_numsrc >= 0x1ffffff || 891 gsf->gf_numsrc > sysctl_igmp_max_msf) { 892 err = -ENOBUFS; 893 goto mc_msf_out; 894 } 895 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 896 err = -EINVAL; 897 goto mc_msf_out; 898 } 899 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 900 msf = kmalloc(msize, GFP_KERNEL); 901 if (!msf) { 902 err = -ENOBUFS; 903 goto mc_msf_out; 904 } 905 ifindex = gsf->gf_interface; 906 psin = (struct sockaddr_in *)&gsf->gf_group; 907 if (psin->sin_family != AF_INET) { 908 err = -EADDRNOTAVAIL; 909 goto mc_msf_out; 910 } 911 msf->imsf_multiaddr = psin->sin_addr.s_addr; 912 msf->imsf_interface = 0; 913 msf->imsf_fmode = gsf->gf_fmode; 914 msf->imsf_numsrc = gsf->gf_numsrc; 915 err = -EADDRNOTAVAIL; 916 for (i = 0; i < gsf->gf_numsrc; ++i) { 917 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 918 919 if (psin->sin_family != AF_INET) 920 goto mc_msf_out; 921 msf->imsf_slist[i] = psin->sin_addr.s_addr; 922 } 923 kfree(gsf); 924 gsf = NULL; 925 926 err = ip_mc_msfilter(sk, msf, ifindex); 927 mc_msf_out: 928 kfree(msf); 929 kfree(gsf); 930 break; 931 } 932 case IP_MULTICAST_ALL: 933 if (optlen < 1) 934 goto e_inval; 935 if (val != 0 && val != 1) 936 goto e_inval; 937 inet->mc_all = val; 938 break; 939 case IP_ROUTER_ALERT: 940 err = ip_ra_control(sk, val ? 1 : 0, NULL); 941 break; 942 943 case IP_FREEBIND: 944 if (optlen < 1) 945 goto e_inval; 946 inet->freebind = !!val; 947 break; 948 949 case IP_IPSEC_POLICY: 950 case IP_XFRM_POLICY: 951 err = -EPERM; 952 if (!capable(CAP_NET_ADMIN)) 953 break; 954 err = xfrm_user_policy(sk, optname, optval, optlen); 955 break; 956 957 case IP_TRANSPARENT: 958 if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) { 959 err = -EPERM; 960 break; 961 } 962 if (optlen < 1) 963 goto e_inval; 964 inet->transparent = !!val; 965 break; 966 967 case IP_MINTTL: 968 if (optlen < 1) 969 goto e_inval; 970 if (val < 0 || val > 255) 971 goto e_inval; 972 inet->min_ttl = val; 973 break; 974 975 default: 976 err = -ENOPROTOOPT; 977 break; 978 } 979 release_sock(sk); 980 return err; 981 982 e_inval: 983 release_sock(sk); 984 return -EINVAL; 985 } 986 987 /** 988 * ipv4_pktinfo_prepare - transfert some info from rtable to skb 989 * @sk: socket 990 * @skb: buffer 991 * 992 * To support IP_CMSG_PKTINFO option, we store rt_iif and rt_spec_dst 993 * in skb->cb[] before dst drop. 994 * This way, receiver doesnt make cache line misses to read rtable. 995 */ 996 void ipv4_pktinfo_prepare(struct sk_buff *skb) 997 { 998 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb); 999 const struct rtable *rt = skb_rtable(skb); 1000 1001 if (rt) { 1002 pktinfo->ipi_ifindex = rt->rt_iif; 1003 pktinfo->ipi_spec_dst.s_addr = rt->rt_spec_dst; 1004 } else { 1005 pktinfo->ipi_ifindex = 0; 1006 pktinfo->ipi_spec_dst.s_addr = 0; 1007 } 1008 skb_dst_drop(skb); 1009 } 1010 1011 int ip_setsockopt(struct sock *sk, int level, 1012 int optname, char __user *optval, unsigned int optlen) 1013 { 1014 int err; 1015 1016 if (level != SOL_IP) 1017 return -ENOPROTOOPT; 1018 1019 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1020 #ifdef CONFIG_NETFILTER 1021 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1022 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1023 optname != IP_IPSEC_POLICY && 1024 optname != IP_XFRM_POLICY && 1025 !ip_mroute_opt(optname)) { 1026 lock_sock(sk); 1027 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1028 release_sock(sk); 1029 } 1030 #endif 1031 return err; 1032 } 1033 EXPORT_SYMBOL(ip_setsockopt); 1034 1035 #ifdef CONFIG_COMPAT 1036 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1037 char __user *optval, unsigned int optlen) 1038 { 1039 int err; 1040 1041 if (level != SOL_IP) 1042 return -ENOPROTOOPT; 1043 1044 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1045 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1046 ip_setsockopt); 1047 1048 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1049 #ifdef CONFIG_NETFILTER 1050 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1051 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1052 optname != IP_IPSEC_POLICY && 1053 optname != IP_XFRM_POLICY && 1054 !ip_mroute_opt(optname)) { 1055 lock_sock(sk); 1056 err = compat_nf_setsockopt(sk, PF_INET, optname, 1057 optval, optlen); 1058 release_sock(sk); 1059 } 1060 #endif 1061 return err; 1062 } 1063 EXPORT_SYMBOL(compat_ip_setsockopt); 1064 #endif 1065 1066 /* 1067 * Get the options. Note for future reference. The GET of IP options gets 1068 * the _received_ ones. The set sets the _sent_ ones. 1069 */ 1070 1071 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1072 char __user *optval, int __user *optlen, unsigned flags) 1073 { 1074 struct inet_sock *inet = inet_sk(sk); 1075 int val; 1076 int len; 1077 1078 if (level != SOL_IP) 1079 return -EOPNOTSUPP; 1080 1081 if (ip_mroute_opt(optname)) 1082 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1083 1084 if (get_user(len, optlen)) 1085 return -EFAULT; 1086 if (len < 0) 1087 return -EINVAL; 1088 1089 lock_sock(sk); 1090 1091 switch (optname) { 1092 case IP_OPTIONS: 1093 { 1094 unsigned char optbuf[sizeof(struct ip_options)+40]; 1095 struct ip_options *opt = (struct ip_options *)optbuf; 1096 struct ip_options_rcu *inet_opt; 1097 1098 inet_opt = rcu_dereference_protected(inet->inet_opt, 1099 sock_owned_by_user(sk)); 1100 opt->optlen = 0; 1101 if (inet_opt) 1102 memcpy(optbuf, &inet_opt->opt, 1103 sizeof(struct ip_options) + 1104 inet_opt->opt.optlen); 1105 release_sock(sk); 1106 1107 if (opt->optlen == 0) 1108 return put_user(0, optlen); 1109 1110 ip_options_undo(opt); 1111 1112 len = min_t(unsigned int, len, opt->optlen); 1113 if (put_user(len, optlen)) 1114 return -EFAULT; 1115 if (copy_to_user(optval, opt->__data, len)) 1116 return -EFAULT; 1117 return 0; 1118 } 1119 case IP_PKTINFO: 1120 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1121 break; 1122 case IP_RECVTTL: 1123 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1124 break; 1125 case IP_RECVTOS: 1126 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1127 break; 1128 case IP_RECVOPTS: 1129 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1130 break; 1131 case IP_RETOPTS: 1132 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1133 break; 1134 case IP_PASSSEC: 1135 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1136 break; 1137 case IP_RECVORIGDSTADDR: 1138 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1139 break; 1140 case IP_TOS: 1141 val = inet->tos; 1142 break; 1143 case IP_TTL: 1144 val = (inet->uc_ttl == -1 ? 1145 sysctl_ip_default_ttl : 1146 inet->uc_ttl); 1147 break; 1148 case IP_HDRINCL: 1149 val = inet->hdrincl; 1150 break; 1151 case IP_NODEFRAG: 1152 val = inet->nodefrag; 1153 break; 1154 case IP_MTU_DISCOVER: 1155 val = inet->pmtudisc; 1156 break; 1157 case IP_MTU: 1158 { 1159 struct dst_entry *dst; 1160 val = 0; 1161 dst = sk_dst_get(sk); 1162 if (dst) { 1163 val = dst_mtu(dst); 1164 dst_release(dst); 1165 } 1166 if (!val) { 1167 release_sock(sk); 1168 return -ENOTCONN; 1169 } 1170 break; 1171 } 1172 case IP_RECVERR: 1173 val = inet->recverr; 1174 break; 1175 case IP_MULTICAST_TTL: 1176 val = inet->mc_ttl; 1177 break; 1178 case IP_MULTICAST_LOOP: 1179 val = inet->mc_loop; 1180 break; 1181 case IP_MULTICAST_IF: 1182 { 1183 struct in_addr addr; 1184 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1185 addr.s_addr = inet->mc_addr; 1186 release_sock(sk); 1187 1188 if (put_user(len, optlen)) 1189 return -EFAULT; 1190 if (copy_to_user(optval, &addr, len)) 1191 return -EFAULT; 1192 return 0; 1193 } 1194 case IP_MSFILTER: 1195 { 1196 struct ip_msfilter msf; 1197 int err; 1198 1199 if (len < IP_MSFILTER_SIZE(0)) { 1200 release_sock(sk); 1201 return -EINVAL; 1202 } 1203 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1204 release_sock(sk); 1205 return -EFAULT; 1206 } 1207 err = ip_mc_msfget(sk, &msf, 1208 (struct ip_msfilter __user *)optval, optlen); 1209 release_sock(sk); 1210 return err; 1211 } 1212 case MCAST_MSFILTER: 1213 { 1214 struct group_filter gsf; 1215 int err; 1216 1217 if (len < GROUP_FILTER_SIZE(0)) { 1218 release_sock(sk); 1219 return -EINVAL; 1220 } 1221 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1222 release_sock(sk); 1223 return -EFAULT; 1224 } 1225 err = ip_mc_gsfget(sk, &gsf, 1226 (struct group_filter __user *)optval, 1227 optlen); 1228 release_sock(sk); 1229 return err; 1230 } 1231 case IP_MULTICAST_ALL: 1232 val = inet->mc_all; 1233 break; 1234 case IP_PKTOPTIONS: 1235 { 1236 struct msghdr msg; 1237 1238 release_sock(sk); 1239 1240 if (sk->sk_type != SOCK_STREAM) 1241 return -ENOPROTOOPT; 1242 1243 msg.msg_control = optval; 1244 msg.msg_controllen = len; 1245 msg.msg_flags = flags; 1246 1247 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1248 struct in_pktinfo info; 1249 1250 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1251 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1252 info.ipi_ifindex = inet->mc_index; 1253 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1254 } 1255 if (inet->cmsg_flags & IP_CMSG_TTL) { 1256 int hlim = inet->mc_ttl; 1257 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1258 } 1259 len -= msg.msg_controllen; 1260 return put_user(len, optlen); 1261 } 1262 case IP_FREEBIND: 1263 val = inet->freebind; 1264 break; 1265 case IP_TRANSPARENT: 1266 val = inet->transparent; 1267 break; 1268 case IP_MINTTL: 1269 val = inet->min_ttl; 1270 break; 1271 default: 1272 release_sock(sk); 1273 return -ENOPROTOOPT; 1274 } 1275 release_sock(sk); 1276 1277 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1278 unsigned char ucval = (unsigned char)val; 1279 len = 1; 1280 if (put_user(len, optlen)) 1281 return -EFAULT; 1282 if (copy_to_user(optval, &ucval, 1)) 1283 return -EFAULT; 1284 } else { 1285 len = min_t(unsigned int, sizeof(int), len); 1286 if (put_user(len, optlen)) 1287 return -EFAULT; 1288 if (copy_to_user(optval, &val, len)) 1289 return -EFAULT; 1290 } 1291 return 0; 1292 } 1293 1294 int ip_getsockopt(struct sock *sk, int level, 1295 int optname, char __user *optval, int __user *optlen) 1296 { 1297 int err; 1298 1299 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1300 #ifdef CONFIG_NETFILTER 1301 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1302 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1303 !ip_mroute_opt(optname)) { 1304 int len; 1305 1306 if (get_user(len, optlen)) 1307 return -EFAULT; 1308 1309 lock_sock(sk); 1310 err = nf_getsockopt(sk, PF_INET, optname, optval, 1311 &len); 1312 release_sock(sk); 1313 if (err >= 0) 1314 err = put_user(len, optlen); 1315 return err; 1316 } 1317 #endif 1318 return err; 1319 } 1320 EXPORT_SYMBOL(ip_getsockopt); 1321 1322 #ifdef CONFIG_COMPAT 1323 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1324 char __user *optval, int __user *optlen) 1325 { 1326 int err; 1327 1328 if (optname == MCAST_MSFILTER) 1329 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1330 ip_getsockopt); 1331 1332 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1333 MSG_CMSG_COMPAT); 1334 1335 #ifdef CONFIG_NETFILTER 1336 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1337 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1338 !ip_mroute_opt(optname)) { 1339 int len; 1340 1341 if (get_user(len, optlen)) 1342 return -EFAULT; 1343 1344 lock_sock(sk); 1345 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1346 release_sock(sk); 1347 if (err >= 0) 1348 err = put_user(len, optlen); 1349 return err; 1350 } 1351 #endif 1352 return err; 1353 } 1354 EXPORT_SYMBOL(compat_ip_getsockopt); 1355 #endif 1356