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 /* 449 * Socket option code for IP. This is the end of the line after any 450 * TCP,UDP etc options on an IP socket. 451 */ 452 453 static int do_ip_setsockopt(struct sock *sk, int level, 454 int optname, char __user *optval, unsigned int optlen) 455 { 456 struct inet_sock *inet = inet_sk(sk); 457 int val = 0, err; 458 459 if (((1<<optname) & ((1<<IP_PKTINFO) | (1<<IP_RECVTTL) | 460 (1<<IP_RECVOPTS) | (1<<IP_RECVTOS) | 461 (1<<IP_RETOPTS) | (1<<IP_TOS) | 462 (1<<IP_TTL) | (1<<IP_HDRINCL) | 463 (1<<IP_MTU_DISCOVER) | (1<<IP_RECVERR) | 464 (1<<IP_ROUTER_ALERT) | (1<<IP_FREEBIND) | 465 (1<<IP_PASSSEC) | (1<<IP_TRANSPARENT) | 466 (1<<IP_MINTTL) | (1<<IP_NODEFRAG))) || 467 optname == IP_UNICAST_IF || 468 optname == IP_MULTICAST_TTL || 469 optname == IP_MULTICAST_ALL || 470 optname == IP_MULTICAST_LOOP || 471 optname == IP_RECVORIGDSTADDR) { 472 if (optlen >= sizeof(int)) { 473 if (get_user(val, (int __user *) optval)) 474 return -EFAULT; 475 } else if (optlen >= sizeof(char)) { 476 unsigned char ucval; 477 478 if (get_user(ucval, (unsigned char __user *) optval)) 479 return -EFAULT; 480 val = (int) ucval; 481 } 482 } 483 484 /* If optlen==0, it is equivalent to val == 0 */ 485 486 if (ip_mroute_opt(optname)) 487 return ip_mroute_setsockopt(sk, optname, optval, optlen); 488 489 err = 0; 490 lock_sock(sk); 491 492 switch (optname) { 493 case IP_OPTIONS: 494 { 495 struct ip_options_rcu *old, *opt = NULL; 496 497 if (optlen > 40) 498 goto e_inval; 499 err = ip_options_get_from_user(sock_net(sk), &opt, 500 optval, optlen); 501 if (err) 502 break; 503 old = rcu_dereference_protected(inet->inet_opt, 504 sock_owned_by_user(sk)); 505 if (inet->is_icsk) { 506 struct inet_connection_sock *icsk = inet_csk(sk); 507 #if IS_ENABLED(CONFIG_IPV6) 508 if (sk->sk_family == PF_INET || 509 (!((1 << sk->sk_state) & 510 (TCPF_LISTEN | TCPF_CLOSE)) && 511 inet->inet_daddr != LOOPBACK4_IPV6)) { 512 #endif 513 if (old) 514 icsk->icsk_ext_hdr_len -= old->opt.optlen; 515 if (opt) 516 icsk->icsk_ext_hdr_len += opt->opt.optlen; 517 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 518 #if IS_ENABLED(CONFIG_IPV6) 519 } 520 #endif 521 } 522 rcu_assign_pointer(inet->inet_opt, opt); 523 if (old) 524 kfree_rcu(old, rcu); 525 break; 526 } 527 case IP_PKTINFO: 528 if (val) 529 inet->cmsg_flags |= IP_CMSG_PKTINFO; 530 else 531 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 532 break; 533 case IP_RECVTTL: 534 if (val) 535 inet->cmsg_flags |= IP_CMSG_TTL; 536 else 537 inet->cmsg_flags &= ~IP_CMSG_TTL; 538 break; 539 case IP_RECVTOS: 540 if (val) 541 inet->cmsg_flags |= IP_CMSG_TOS; 542 else 543 inet->cmsg_flags &= ~IP_CMSG_TOS; 544 break; 545 case IP_RECVOPTS: 546 if (val) 547 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 548 else 549 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 550 break; 551 case IP_RETOPTS: 552 if (val) 553 inet->cmsg_flags |= IP_CMSG_RETOPTS; 554 else 555 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 556 break; 557 case IP_PASSSEC: 558 if (val) 559 inet->cmsg_flags |= IP_CMSG_PASSSEC; 560 else 561 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 562 break; 563 case IP_RECVORIGDSTADDR: 564 if (val) 565 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 566 else 567 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 568 break; 569 case IP_TOS: /* This sets both TOS and Precedence */ 570 if (sk->sk_type == SOCK_STREAM) { 571 val &= ~INET_ECN_MASK; 572 val |= inet->tos & INET_ECN_MASK; 573 } 574 if (inet->tos != val) { 575 inet->tos = val; 576 sk->sk_priority = rt_tos2priority(val); 577 sk_dst_reset(sk); 578 } 579 break; 580 case IP_TTL: 581 if (optlen < 1) 582 goto e_inval; 583 if (val != -1 && (val < 0 || val > 255)) 584 goto e_inval; 585 inet->uc_ttl = val; 586 break; 587 case IP_HDRINCL: 588 if (sk->sk_type != SOCK_RAW) { 589 err = -ENOPROTOOPT; 590 break; 591 } 592 inet->hdrincl = val ? 1 : 0; 593 break; 594 case IP_NODEFRAG: 595 if (sk->sk_type != SOCK_RAW) { 596 err = -ENOPROTOOPT; 597 break; 598 } 599 inet->nodefrag = val ? 1 : 0; 600 break; 601 case IP_MTU_DISCOVER: 602 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_PROBE) 603 goto e_inval; 604 inet->pmtudisc = val; 605 break; 606 case IP_RECVERR: 607 inet->recverr = !!val; 608 if (!val) 609 skb_queue_purge(&sk->sk_error_queue); 610 break; 611 case IP_MULTICAST_TTL: 612 if (sk->sk_type == SOCK_STREAM) 613 goto e_inval; 614 if (optlen < 1) 615 goto e_inval; 616 if (val == -1) 617 val = 1; 618 if (val < 0 || val > 255) 619 goto e_inval; 620 inet->mc_ttl = val; 621 break; 622 case IP_MULTICAST_LOOP: 623 if (optlen < 1) 624 goto e_inval; 625 inet->mc_loop = !!val; 626 break; 627 case IP_UNICAST_IF: 628 { 629 struct net_device *dev = NULL; 630 int ifindex; 631 632 if (optlen != sizeof(int)) 633 goto e_inval; 634 635 ifindex = (__force int)ntohl((__force __be32)val); 636 if (ifindex == 0) { 637 inet->uc_index = 0; 638 err = 0; 639 break; 640 } 641 642 dev = dev_get_by_index(sock_net(sk), ifindex); 643 err = -EADDRNOTAVAIL; 644 if (!dev) 645 break; 646 dev_put(dev); 647 648 err = -EINVAL; 649 if (sk->sk_bound_dev_if) 650 break; 651 652 inet->uc_index = ifindex; 653 err = 0; 654 break; 655 } 656 case IP_MULTICAST_IF: 657 { 658 struct ip_mreqn mreq; 659 struct net_device *dev = NULL; 660 661 if (sk->sk_type == SOCK_STREAM) 662 goto e_inval; 663 /* 664 * Check the arguments are allowable 665 */ 666 667 if (optlen < sizeof(struct in_addr)) 668 goto e_inval; 669 670 err = -EFAULT; 671 if (optlen >= sizeof(struct ip_mreqn)) { 672 if (copy_from_user(&mreq, optval, sizeof(mreq))) 673 break; 674 } else { 675 memset(&mreq, 0, sizeof(mreq)); 676 if (optlen >= sizeof(struct in_addr) && 677 copy_from_user(&mreq.imr_address, optval, 678 sizeof(struct in_addr))) 679 break; 680 } 681 682 if (!mreq.imr_ifindex) { 683 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 684 inet->mc_index = 0; 685 inet->mc_addr = 0; 686 err = 0; 687 break; 688 } 689 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 690 if (dev) 691 mreq.imr_ifindex = dev->ifindex; 692 } else 693 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 694 695 696 err = -EADDRNOTAVAIL; 697 if (!dev) 698 break; 699 dev_put(dev); 700 701 err = -EINVAL; 702 if (sk->sk_bound_dev_if && 703 mreq.imr_ifindex != sk->sk_bound_dev_if) 704 break; 705 706 inet->mc_index = mreq.imr_ifindex; 707 inet->mc_addr = mreq.imr_address.s_addr; 708 err = 0; 709 break; 710 } 711 712 case IP_ADD_MEMBERSHIP: 713 case IP_DROP_MEMBERSHIP: 714 { 715 struct ip_mreqn mreq; 716 717 err = -EPROTO; 718 if (inet_sk(sk)->is_icsk) 719 break; 720 721 if (optlen < sizeof(struct ip_mreq)) 722 goto e_inval; 723 err = -EFAULT; 724 if (optlen >= sizeof(struct ip_mreqn)) { 725 if (copy_from_user(&mreq, optval, sizeof(mreq))) 726 break; 727 } else { 728 memset(&mreq, 0, sizeof(mreq)); 729 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 730 break; 731 } 732 733 if (optname == IP_ADD_MEMBERSHIP) 734 err = ip_mc_join_group(sk, &mreq); 735 else 736 err = ip_mc_leave_group(sk, &mreq); 737 break; 738 } 739 case IP_MSFILTER: 740 { 741 struct ip_msfilter *msf; 742 743 if (optlen < IP_MSFILTER_SIZE(0)) 744 goto e_inval; 745 if (optlen > sysctl_optmem_max) { 746 err = -ENOBUFS; 747 break; 748 } 749 msf = kmalloc(optlen, GFP_KERNEL); 750 if (!msf) { 751 err = -ENOBUFS; 752 break; 753 } 754 err = -EFAULT; 755 if (copy_from_user(msf, optval, optlen)) { 756 kfree(msf); 757 break; 758 } 759 /* numsrc >= (1G-4) overflow in 32 bits */ 760 if (msf->imsf_numsrc >= 0x3ffffffcU || 761 msf->imsf_numsrc > sysctl_igmp_max_msf) { 762 kfree(msf); 763 err = -ENOBUFS; 764 break; 765 } 766 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 767 kfree(msf); 768 err = -EINVAL; 769 break; 770 } 771 err = ip_mc_msfilter(sk, msf, 0); 772 kfree(msf); 773 break; 774 } 775 case IP_BLOCK_SOURCE: 776 case IP_UNBLOCK_SOURCE: 777 case IP_ADD_SOURCE_MEMBERSHIP: 778 case IP_DROP_SOURCE_MEMBERSHIP: 779 { 780 struct ip_mreq_source mreqs; 781 int omode, add; 782 783 if (optlen != sizeof(struct ip_mreq_source)) 784 goto e_inval; 785 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 786 err = -EFAULT; 787 break; 788 } 789 if (optname == IP_BLOCK_SOURCE) { 790 omode = MCAST_EXCLUDE; 791 add = 1; 792 } else if (optname == IP_UNBLOCK_SOURCE) { 793 omode = MCAST_EXCLUDE; 794 add = 0; 795 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 796 struct ip_mreqn mreq; 797 798 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 799 mreq.imr_address.s_addr = mreqs.imr_interface; 800 mreq.imr_ifindex = 0; 801 err = ip_mc_join_group(sk, &mreq); 802 if (err && err != -EADDRINUSE) 803 break; 804 omode = MCAST_INCLUDE; 805 add = 1; 806 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 807 omode = MCAST_INCLUDE; 808 add = 0; 809 } 810 err = ip_mc_source(add, omode, sk, &mreqs, 0); 811 break; 812 } 813 case MCAST_JOIN_GROUP: 814 case MCAST_LEAVE_GROUP: 815 { 816 struct group_req greq; 817 struct sockaddr_in *psin; 818 struct ip_mreqn mreq; 819 820 if (optlen < sizeof(struct group_req)) 821 goto e_inval; 822 err = -EFAULT; 823 if (copy_from_user(&greq, optval, sizeof(greq))) 824 break; 825 psin = (struct sockaddr_in *)&greq.gr_group; 826 if (psin->sin_family != AF_INET) 827 goto e_inval; 828 memset(&mreq, 0, sizeof(mreq)); 829 mreq.imr_multiaddr = psin->sin_addr; 830 mreq.imr_ifindex = greq.gr_interface; 831 832 if (optname == MCAST_JOIN_GROUP) 833 err = ip_mc_join_group(sk, &mreq); 834 else 835 err = ip_mc_leave_group(sk, &mreq); 836 break; 837 } 838 case MCAST_JOIN_SOURCE_GROUP: 839 case MCAST_LEAVE_SOURCE_GROUP: 840 case MCAST_BLOCK_SOURCE: 841 case MCAST_UNBLOCK_SOURCE: 842 { 843 struct group_source_req greqs; 844 struct ip_mreq_source mreqs; 845 struct sockaddr_in *psin; 846 int omode, add; 847 848 if (optlen != sizeof(struct group_source_req)) 849 goto e_inval; 850 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 851 err = -EFAULT; 852 break; 853 } 854 if (greqs.gsr_group.ss_family != AF_INET || 855 greqs.gsr_source.ss_family != AF_INET) { 856 err = -EADDRNOTAVAIL; 857 break; 858 } 859 psin = (struct sockaddr_in *)&greqs.gsr_group; 860 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 861 psin = (struct sockaddr_in *)&greqs.gsr_source; 862 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 863 mreqs.imr_interface = 0; /* use index for mc_source */ 864 865 if (optname == MCAST_BLOCK_SOURCE) { 866 omode = MCAST_EXCLUDE; 867 add = 1; 868 } else if (optname == MCAST_UNBLOCK_SOURCE) { 869 omode = MCAST_EXCLUDE; 870 add = 0; 871 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 872 struct ip_mreqn mreq; 873 874 psin = (struct sockaddr_in *)&greqs.gsr_group; 875 mreq.imr_multiaddr = psin->sin_addr; 876 mreq.imr_address.s_addr = 0; 877 mreq.imr_ifindex = greqs.gsr_interface; 878 err = ip_mc_join_group(sk, &mreq); 879 if (err && err != -EADDRINUSE) 880 break; 881 greqs.gsr_interface = mreq.imr_ifindex; 882 omode = MCAST_INCLUDE; 883 add = 1; 884 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 885 omode = MCAST_INCLUDE; 886 add = 0; 887 } 888 err = ip_mc_source(add, omode, sk, &mreqs, 889 greqs.gsr_interface); 890 break; 891 } 892 case MCAST_MSFILTER: 893 { 894 struct sockaddr_in *psin; 895 struct ip_msfilter *msf = NULL; 896 struct group_filter *gsf = NULL; 897 int msize, i, ifindex; 898 899 if (optlen < GROUP_FILTER_SIZE(0)) 900 goto e_inval; 901 if (optlen > sysctl_optmem_max) { 902 err = -ENOBUFS; 903 break; 904 } 905 gsf = kmalloc(optlen, GFP_KERNEL); 906 if (!gsf) { 907 err = -ENOBUFS; 908 break; 909 } 910 err = -EFAULT; 911 if (copy_from_user(gsf, optval, optlen)) 912 goto mc_msf_out; 913 914 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 915 if (gsf->gf_numsrc >= 0x1ffffff || 916 gsf->gf_numsrc > sysctl_igmp_max_msf) { 917 err = -ENOBUFS; 918 goto mc_msf_out; 919 } 920 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 921 err = -EINVAL; 922 goto mc_msf_out; 923 } 924 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 925 msf = kmalloc(msize, GFP_KERNEL); 926 if (!msf) { 927 err = -ENOBUFS; 928 goto mc_msf_out; 929 } 930 ifindex = gsf->gf_interface; 931 psin = (struct sockaddr_in *)&gsf->gf_group; 932 if (psin->sin_family != AF_INET) { 933 err = -EADDRNOTAVAIL; 934 goto mc_msf_out; 935 } 936 msf->imsf_multiaddr = psin->sin_addr.s_addr; 937 msf->imsf_interface = 0; 938 msf->imsf_fmode = gsf->gf_fmode; 939 msf->imsf_numsrc = gsf->gf_numsrc; 940 err = -EADDRNOTAVAIL; 941 for (i = 0; i < gsf->gf_numsrc; ++i) { 942 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 943 944 if (psin->sin_family != AF_INET) 945 goto mc_msf_out; 946 msf->imsf_slist[i] = psin->sin_addr.s_addr; 947 } 948 kfree(gsf); 949 gsf = NULL; 950 951 err = ip_mc_msfilter(sk, msf, ifindex); 952 mc_msf_out: 953 kfree(msf); 954 kfree(gsf); 955 break; 956 } 957 case IP_MULTICAST_ALL: 958 if (optlen < 1) 959 goto e_inval; 960 if (val != 0 && val != 1) 961 goto e_inval; 962 inet->mc_all = val; 963 break; 964 case IP_ROUTER_ALERT: 965 err = ip_ra_control(sk, val ? 1 : 0, NULL); 966 break; 967 968 case IP_FREEBIND: 969 if (optlen < 1) 970 goto e_inval; 971 inet->freebind = !!val; 972 break; 973 974 case IP_IPSEC_POLICY: 975 case IP_XFRM_POLICY: 976 err = -EPERM; 977 if (!capable(CAP_NET_ADMIN)) 978 break; 979 err = xfrm_user_policy(sk, optname, optval, optlen); 980 break; 981 982 case IP_TRANSPARENT: 983 if (!!val && !capable(CAP_NET_RAW) && !capable(CAP_NET_ADMIN)) { 984 err = -EPERM; 985 break; 986 } 987 if (optlen < 1) 988 goto e_inval; 989 inet->transparent = !!val; 990 break; 991 992 case IP_MINTTL: 993 if (optlen < 1) 994 goto e_inval; 995 if (val < 0 || val > 255) 996 goto e_inval; 997 inet->min_ttl = val; 998 break; 999 1000 default: 1001 err = -ENOPROTOOPT; 1002 break; 1003 } 1004 release_sock(sk); 1005 return err; 1006 1007 e_inval: 1008 release_sock(sk); 1009 return -EINVAL; 1010 } 1011 1012 /** 1013 * ipv4_pktinfo_prepare - transfert some info from rtable to skb 1014 * @sk: socket 1015 * @skb: buffer 1016 * 1017 * To support IP_CMSG_PKTINFO option, we store rt_iif and rt_spec_dst 1018 * in skb->cb[] before dst drop. 1019 * This way, receiver doesnt make cache line misses to read rtable. 1020 */ 1021 void ipv4_pktinfo_prepare(struct sk_buff *skb) 1022 { 1023 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb); 1024 const struct rtable *rt = skb_rtable(skb); 1025 1026 if (rt) { 1027 pktinfo->ipi_ifindex = rt->rt_iif; 1028 pktinfo->ipi_spec_dst.s_addr = rt->rt_spec_dst; 1029 } else { 1030 pktinfo->ipi_ifindex = 0; 1031 pktinfo->ipi_spec_dst.s_addr = 0; 1032 } 1033 skb_dst_drop(skb); 1034 } 1035 1036 int ip_setsockopt(struct sock *sk, int level, 1037 int optname, char __user *optval, unsigned int optlen) 1038 { 1039 int err; 1040 1041 if (level != SOL_IP) 1042 return -ENOPROTOOPT; 1043 1044 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1045 #ifdef CONFIG_NETFILTER 1046 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1047 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1048 optname != IP_IPSEC_POLICY && 1049 optname != IP_XFRM_POLICY && 1050 !ip_mroute_opt(optname)) { 1051 lock_sock(sk); 1052 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1053 release_sock(sk); 1054 } 1055 #endif 1056 return err; 1057 } 1058 EXPORT_SYMBOL(ip_setsockopt); 1059 1060 #ifdef CONFIG_COMPAT 1061 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1062 char __user *optval, unsigned int optlen) 1063 { 1064 int err; 1065 1066 if (level != SOL_IP) 1067 return -ENOPROTOOPT; 1068 1069 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1070 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1071 ip_setsockopt); 1072 1073 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1074 #ifdef CONFIG_NETFILTER 1075 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1076 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1077 optname != IP_IPSEC_POLICY && 1078 optname != IP_XFRM_POLICY && 1079 !ip_mroute_opt(optname)) { 1080 lock_sock(sk); 1081 err = compat_nf_setsockopt(sk, PF_INET, optname, 1082 optval, optlen); 1083 release_sock(sk); 1084 } 1085 #endif 1086 return err; 1087 } 1088 EXPORT_SYMBOL(compat_ip_setsockopt); 1089 #endif 1090 1091 /* 1092 * Get the options. Note for future reference. The GET of IP options gets 1093 * the _received_ ones. The set sets the _sent_ ones. 1094 */ 1095 1096 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1097 char __user *optval, int __user *optlen, unsigned flags) 1098 { 1099 struct inet_sock *inet = inet_sk(sk); 1100 int val; 1101 int len; 1102 1103 if (level != SOL_IP) 1104 return -EOPNOTSUPP; 1105 1106 if (ip_mroute_opt(optname)) 1107 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1108 1109 if (get_user(len, optlen)) 1110 return -EFAULT; 1111 if (len < 0) 1112 return -EINVAL; 1113 1114 lock_sock(sk); 1115 1116 switch (optname) { 1117 case IP_OPTIONS: 1118 { 1119 unsigned char optbuf[sizeof(struct ip_options)+40]; 1120 struct ip_options *opt = (struct ip_options *)optbuf; 1121 struct ip_options_rcu *inet_opt; 1122 1123 inet_opt = rcu_dereference_protected(inet->inet_opt, 1124 sock_owned_by_user(sk)); 1125 opt->optlen = 0; 1126 if (inet_opt) 1127 memcpy(optbuf, &inet_opt->opt, 1128 sizeof(struct ip_options) + 1129 inet_opt->opt.optlen); 1130 release_sock(sk); 1131 1132 if (opt->optlen == 0) 1133 return put_user(0, optlen); 1134 1135 ip_options_undo(opt); 1136 1137 len = min_t(unsigned int, len, opt->optlen); 1138 if (put_user(len, optlen)) 1139 return -EFAULT; 1140 if (copy_to_user(optval, opt->__data, len)) 1141 return -EFAULT; 1142 return 0; 1143 } 1144 case IP_PKTINFO: 1145 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1146 break; 1147 case IP_RECVTTL: 1148 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1149 break; 1150 case IP_RECVTOS: 1151 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1152 break; 1153 case IP_RECVOPTS: 1154 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1155 break; 1156 case IP_RETOPTS: 1157 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1158 break; 1159 case IP_PASSSEC: 1160 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1161 break; 1162 case IP_RECVORIGDSTADDR: 1163 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1164 break; 1165 case IP_TOS: 1166 val = inet->tos; 1167 break; 1168 case IP_TTL: 1169 val = (inet->uc_ttl == -1 ? 1170 sysctl_ip_default_ttl : 1171 inet->uc_ttl); 1172 break; 1173 case IP_HDRINCL: 1174 val = inet->hdrincl; 1175 break; 1176 case IP_NODEFRAG: 1177 val = inet->nodefrag; 1178 break; 1179 case IP_MTU_DISCOVER: 1180 val = inet->pmtudisc; 1181 break; 1182 case IP_MTU: 1183 { 1184 struct dst_entry *dst; 1185 val = 0; 1186 dst = sk_dst_get(sk); 1187 if (dst) { 1188 val = dst_mtu(dst); 1189 dst_release(dst); 1190 } 1191 if (!val) { 1192 release_sock(sk); 1193 return -ENOTCONN; 1194 } 1195 break; 1196 } 1197 case IP_RECVERR: 1198 val = inet->recverr; 1199 break; 1200 case IP_MULTICAST_TTL: 1201 val = inet->mc_ttl; 1202 break; 1203 case IP_MULTICAST_LOOP: 1204 val = inet->mc_loop; 1205 break; 1206 case IP_UNICAST_IF: 1207 val = (__force int)htonl((__u32) inet->uc_index); 1208 break; 1209 case IP_MULTICAST_IF: 1210 { 1211 struct in_addr addr; 1212 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1213 addr.s_addr = inet->mc_addr; 1214 release_sock(sk); 1215 1216 if (put_user(len, optlen)) 1217 return -EFAULT; 1218 if (copy_to_user(optval, &addr, len)) 1219 return -EFAULT; 1220 return 0; 1221 } 1222 case IP_MSFILTER: 1223 { 1224 struct ip_msfilter msf; 1225 int err; 1226 1227 if (len < IP_MSFILTER_SIZE(0)) { 1228 release_sock(sk); 1229 return -EINVAL; 1230 } 1231 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1232 release_sock(sk); 1233 return -EFAULT; 1234 } 1235 err = ip_mc_msfget(sk, &msf, 1236 (struct ip_msfilter __user *)optval, optlen); 1237 release_sock(sk); 1238 return err; 1239 } 1240 case MCAST_MSFILTER: 1241 { 1242 struct group_filter gsf; 1243 int err; 1244 1245 if (len < GROUP_FILTER_SIZE(0)) { 1246 release_sock(sk); 1247 return -EINVAL; 1248 } 1249 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1250 release_sock(sk); 1251 return -EFAULT; 1252 } 1253 err = ip_mc_gsfget(sk, &gsf, 1254 (struct group_filter __user *)optval, 1255 optlen); 1256 release_sock(sk); 1257 return err; 1258 } 1259 case IP_MULTICAST_ALL: 1260 val = inet->mc_all; 1261 break; 1262 case IP_PKTOPTIONS: 1263 { 1264 struct msghdr msg; 1265 1266 release_sock(sk); 1267 1268 if (sk->sk_type != SOCK_STREAM) 1269 return -ENOPROTOOPT; 1270 1271 msg.msg_control = optval; 1272 msg.msg_controllen = len; 1273 msg.msg_flags = flags; 1274 1275 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1276 struct in_pktinfo info; 1277 1278 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1279 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1280 info.ipi_ifindex = inet->mc_index; 1281 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1282 } 1283 if (inet->cmsg_flags & IP_CMSG_TTL) { 1284 int hlim = inet->mc_ttl; 1285 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1286 } 1287 if (inet->cmsg_flags & IP_CMSG_TOS) { 1288 int tos = inet->rcv_tos; 1289 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos); 1290 } 1291 len -= msg.msg_controllen; 1292 return put_user(len, optlen); 1293 } 1294 case IP_FREEBIND: 1295 val = inet->freebind; 1296 break; 1297 case IP_TRANSPARENT: 1298 val = inet->transparent; 1299 break; 1300 case IP_MINTTL: 1301 val = inet->min_ttl; 1302 break; 1303 default: 1304 release_sock(sk); 1305 return -ENOPROTOOPT; 1306 } 1307 release_sock(sk); 1308 1309 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1310 unsigned char ucval = (unsigned char)val; 1311 len = 1; 1312 if (put_user(len, optlen)) 1313 return -EFAULT; 1314 if (copy_to_user(optval, &ucval, 1)) 1315 return -EFAULT; 1316 } else { 1317 len = min_t(unsigned int, sizeof(int), len); 1318 if (put_user(len, optlen)) 1319 return -EFAULT; 1320 if (copy_to_user(optval, &val, len)) 1321 return -EFAULT; 1322 } 1323 return 0; 1324 } 1325 1326 int ip_getsockopt(struct sock *sk, int level, 1327 int optname, char __user *optval, int __user *optlen) 1328 { 1329 int err; 1330 1331 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1332 #ifdef CONFIG_NETFILTER 1333 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1334 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1335 !ip_mroute_opt(optname)) { 1336 int len; 1337 1338 if (get_user(len, optlen)) 1339 return -EFAULT; 1340 1341 lock_sock(sk); 1342 err = nf_getsockopt(sk, PF_INET, optname, optval, 1343 &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(ip_getsockopt); 1353 1354 #ifdef CONFIG_COMPAT 1355 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1356 char __user *optval, int __user *optlen) 1357 { 1358 int err; 1359 1360 if (optname == MCAST_MSFILTER) 1361 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1362 ip_getsockopt); 1363 1364 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1365 MSG_CMSG_COMPAT); 1366 1367 #ifdef CONFIG_NETFILTER 1368 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1369 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1370 !ip_mroute_opt(optname)) { 1371 int len; 1372 1373 if (get_user(len, optlen)) 1374 return -EFAULT; 1375 1376 lock_sock(sk); 1377 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1378 release_sock(sk); 1379 if (err >= 0) 1380 err = put_user(len, optlen); 1381 return err; 1382 } 1383 #endif 1384 return err; 1385 } 1386 EXPORT_SYMBOL(compat_ip_getsockopt); 1387 #endif 1388