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 #include <net/checksum.h> 41 #if IS_ENABLED(CONFIG_IPV6) 42 #include <net/transp_v6.h> 43 #endif 44 #include <net/ip_fib.h> 45 46 #include <linux/errqueue.h> 47 #include <asm/uaccess.h> 48 49 /* 50 * SOL_IP control messages. 51 */ 52 53 static void ip_cmsg_recv_pktinfo(struct msghdr *msg, struct sk_buff *skb) 54 { 55 struct in_pktinfo info = *PKTINFO_SKB_CB(skb); 56 57 info.ipi_addr.s_addr = ip_hdr(skb)->daddr; 58 59 put_cmsg(msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 60 } 61 62 static void ip_cmsg_recv_ttl(struct msghdr *msg, struct sk_buff *skb) 63 { 64 int ttl = ip_hdr(skb)->ttl; 65 put_cmsg(msg, SOL_IP, IP_TTL, sizeof(int), &ttl); 66 } 67 68 static void ip_cmsg_recv_tos(struct msghdr *msg, struct sk_buff *skb) 69 { 70 put_cmsg(msg, SOL_IP, IP_TOS, 1, &ip_hdr(skb)->tos); 71 } 72 73 static void ip_cmsg_recv_opts(struct msghdr *msg, struct sk_buff *skb) 74 { 75 if (IPCB(skb)->opt.optlen == 0) 76 return; 77 78 put_cmsg(msg, SOL_IP, IP_RECVOPTS, IPCB(skb)->opt.optlen, 79 ip_hdr(skb) + 1); 80 } 81 82 83 static void ip_cmsg_recv_retopts(struct msghdr *msg, struct sk_buff *skb) 84 { 85 unsigned char optbuf[sizeof(struct ip_options) + 40]; 86 struct ip_options *opt = (struct ip_options *)optbuf; 87 88 if (IPCB(skb)->opt.optlen == 0) 89 return; 90 91 if (ip_options_echo(opt, skb)) { 92 msg->msg_flags |= MSG_CTRUNC; 93 return; 94 } 95 ip_options_undo(opt); 96 97 put_cmsg(msg, SOL_IP, IP_RETOPTS, opt->optlen, opt->__data); 98 } 99 100 static void ip_cmsg_recv_checksum(struct msghdr *msg, struct sk_buff *skb, 101 int offset) 102 { 103 __wsum csum = skb->csum; 104 105 if (skb->ip_summed != CHECKSUM_COMPLETE) 106 return; 107 108 if (offset != 0) 109 csum = csum_sub(csum, csum_partial(skb->data, offset, 0)); 110 111 put_cmsg(msg, SOL_IP, IP_CHECKSUM, sizeof(__wsum), &csum); 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_offset(struct msghdr *msg, struct sk_buff *skb, 155 int offset) 156 { 157 struct inet_sock *inet = inet_sk(skb->sk); 158 unsigned int flags = inet->cmsg_flags; 159 160 /* Ordered by supposed usage frequency */ 161 if (flags & IP_CMSG_PKTINFO) { 162 ip_cmsg_recv_pktinfo(msg, skb); 163 164 flags &= ~IP_CMSG_PKTINFO; 165 if (!flags) 166 return; 167 } 168 169 if (flags & IP_CMSG_TTL) { 170 ip_cmsg_recv_ttl(msg, skb); 171 172 flags &= ~IP_CMSG_TTL; 173 if (!flags) 174 return; 175 } 176 177 if (flags & IP_CMSG_TOS) { 178 ip_cmsg_recv_tos(msg, skb); 179 180 flags &= ~IP_CMSG_TOS; 181 if (!flags) 182 return; 183 } 184 185 if (flags & IP_CMSG_RECVOPTS) { 186 ip_cmsg_recv_opts(msg, skb); 187 188 flags &= ~IP_CMSG_RECVOPTS; 189 if (!flags) 190 return; 191 } 192 193 if (flags & IP_CMSG_RETOPTS) { 194 ip_cmsg_recv_retopts(msg, skb); 195 196 flags &= ~IP_CMSG_RETOPTS; 197 if (!flags) 198 return; 199 } 200 201 if (flags & IP_CMSG_PASSSEC) { 202 ip_cmsg_recv_security(msg, skb); 203 204 flags &= ~IP_CMSG_PASSSEC; 205 if (!flags) 206 return; 207 } 208 209 if (flags & IP_CMSG_ORIGDSTADDR) { 210 ip_cmsg_recv_dstaddr(msg, skb); 211 212 flags &= ~IP_CMSG_ORIGDSTADDR; 213 if (!flags) 214 return; 215 } 216 217 if (flags & IP_CMSG_CHECKSUM) 218 ip_cmsg_recv_checksum(msg, skb, offset); 219 } 220 EXPORT_SYMBOL(ip_cmsg_recv_offset); 221 222 int ip_cmsg_send(struct net *net, struct msghdr *msg, struct ipcm_cookie *ipc, 223 bool allow_ipv6) 224 { 225 int err, val; 226 struct cmsghdr *cmsg; 227 228 for_each_cmsghdr(cmsg, msg) { 229 if (!CMSG_OK(msg, cmsg)) 230 return -EINVAL; 231 #if IS_ENABLED(CONFIG_IPV6) 232 if (allow_ipv6 && 233 cmsg->cmsg_level == SOL_IPV6 && 234 cmsg->cmsg_type == IPV6_PKTINFO) { 235 struct in6_pktinfo *src_info; 236 237 if (cmsg->cmsg_len < CMSG_LEN(sizeof(*src_info))) 238 return -EINVAL; 239 src_info = (struct in6_pktinfo *)CMSG_DATA(cmsg); 240 if (!ipv6_addr_v4mapped(&src_info->ipi6_addr)) 241 return -EINVAL; 242 ipc->oif = src_info->ipi6_ifindex; 243 ipc->addr = src_info->ipi6_addr.s6_addr32[3]; 244 continue; 245 } 246 #endif 247 if (cmsg->cmsg_level != SOL_IP) 248 continue; 249 switch (cmsg->cmsg_type) { 250 case IP_RETOPTS: 251 err = cmsg->cmsg_len - CMSG_ALIGN(sizeof(struct cmsghdr)); 252 err = ip_options_get(net, &ipc->opt, CMSG_DATA(cmsg), 253 err < 40 ? err : 40); 254 if (err) 255 return err; 256 break; 257 case IP_PKTINFO: 258 { 259 struct in_pktinfo *info; 260 if (cmsg->cmsg_len != CMSG_LEN(sizeof(struct in_pktinfo))) 261 return -EINVAL; 262 info = (struct in_pktinfo *)CMSG_DATA(cmsg); 263 ipc->oif = info->ipi_ifindex; 264 ipc->addr = info->ipi_spec_dst.s_addr; 265 break; 266 } 267 case IP_TTL: 268 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 269 return -EINVAL; 270 val = *(int *)CMSG_DATA(cmsg); 271 if (val < 1 || val > 255) 272 return -EINVAL; 273 ipc->ttl = val; 274 break; 275 case IP_TOS: 276 if (cmsg->cmsg_len != CMSG_LEN(sizeof(int))) 277 return -EINVAL; 278 val = *(int *)CMSG_DATA(cmsg); 279 if (val < 0 || val > 255) 280 return -EINVAL; 281 ipc->tos = val; 282 ipc->priority = rt_tos2priority(ipc->tos); 283 break; 284 285 default: 286 return -EINVAL; 287 } 288 } 289 return 0; 290 } 291 292 293 /* Special input handler for packets caught by router alert option. 294 They are selected only by protocol field, and then processed likely 295 local ones; but only if someone wants them! Otherwise, router 296 not running rsvpd will kill RSVP. 297 298 It is user level problem, what it will make with them. 299 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)), 300 but receiver should be enough clever f.e. to forward mtrace requests, 301 sent to multicast group to reach destination designated router. 302 */ 303 struct ip_ra_chain __rcu *ip_ra_chain; 304 static DEFINE_SPINLOCK(ip_ra_lock); 305 306 307 static void ip_ra_destroy_rcu(struct rcu_head *head) 308 { 309 struct ip_ra_chain *ra = container_of(head, struct ip_ra_chain, rcu); 310 311 sock_put(ra->saved_sk); 312 kfree(ra); 313 } 314 315 int ip_ra_control(struct sock *sk, unsigned char on, 316 void (*destructor)(struct sock *)) 317 { 318 struct ip_ra_chain *ra, *new_ra; 319 struct ip_ra_chain __rcu **rap; 320 321 if (sk->sk_type != SOCK_RAW || inet_sk(sk)->inet_num == IPPROTO_RAW) 322 return -EINVAL; 323 324 new_ra = on ? kmalloc(sizeof(*new_ra), GFP_KERNEL) : NULL; 325 326 spin_lock_bh(&ip_ra_lock); 327 for (rap = &ip_ra_chain; 328 (ra = rcu_dereference_protected(*rap, 329 lockdep_is_held(&ip_ra_lock))) != NULL; 330 rap = &ra->next) { 331 if (ra->sk == sk) { 332 if (on) { 333 spin_unlock_bh(&ip_ra_lock); 334 kfree(new_ra); 335 return -EADDRINUSE; 336 } 337 /* dont let ip_call_ra_chain() use sk again */ 338 ra->sk = NULL; 339 RCU_INIT_POINTER(*rap, ra->next); 340 spin_unlock_bh(&ip_ra_lock); 341 342 if (ra->destructor) 343 ra->destructor(sk); 344 /* 345 * Delay sock_put(sk) and kfree(ra) after one rcu grace 346 * period. This guarantee ip_call_ra_chain() dont need 347 * to mess with socket refcounts. 348 */ 349 ra->saved_sk = sk; 350 call_rcu(&ra->rcu, ip_ra_destroy_rcu); 351 return 0; 352 } 353 } 354 if (new_ra == NULL) { 355 spin_unlock_bh(&ip_ra_lock); 356 return -ENOBUFS; 357 } 358 new_ra->sk = sk; 359 new_ra->destructor = destructor; 360 361 RCU_INIT_POINTER(new_ra->next, ra); 362 rcu_assign_pointer(*rap, new_ra); 363 sock_hold(sk); 364 spin_unlock_bh(&ip_ra_lock); 365 366 return 0; 367 } 368 369 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, 370 __be16 port, u32 info, u8 *payload) 371 { 372 struct sock_exterr_skb *serr; 373 374 skb = skb_clone(skb, GFP_ATOMIC); 375 if (!skb) 376 return; 377 378 serr = SKB_EXT_ERR(skb); 379 serr->ee.ee_errno = err; 380 serr->ee.ee_origin = SO_EE_ORIGIN_ICMP; 381 serr->ee.ee_type = icmp_hdr(skb)->type; 382 serr->ee.ee_code = icmp_hdr(skb)->code; 383 serr->ee.ee_pad = 0; 384 serr->ee.ee_info = info; 385 serr->ee.ee_data = 0; 386 serr->addr_offset = (u8 *)&(((struct iphdr *)(icmp_hdr(skb) + 1))->daddr) - 387 skb_network_header(skb); 388 serr->port = port; 389 390 if (skb_pull(skb, payload - skb->data) != NULL) { 391 skb_reset_transport_header(skb); 392 if (sock_queue_err_skb(sk, skb) == 0) 393 return; 394 } 395 kfree_skb(skb); 396 } 397 398 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 port, u32 info) 399 { 400 struct inet_sock *inet = inet_sk(sk); 401 struct sock_exterr_skb *serr; 402 struct iphdr *iph; 403 struct sk_buff *skb; 404 405 if (!inet->recverr) 406 return; 407 408 skb = alloc_skb(sizeof(struct iphdr), GFP_ATOMIC); 409 if (!skb) 410 return; 411 412 skb_put(skb, sizeof(struct iphdr)); 413 skb_reset_network_header(skb); 414 iph = ip_hdr(skb); 415 iph->daddr = daddr; 416 417 serr = SKB_EXT_ERR(skb); 418 serr->ee.ee_errno = err; 419 serr->ee.ee_origin = SO_EE_ORIGIN_LOCAL; 420 serr->ee.ee_type = 0; 421 serr->ee.ee_code = 0; 422 serr->ee.ee_pad = 0; 423 serr->ee.ee_info = info; 424 serr->ee.ee_data = 0; 425 serr->addr_offset = (u8 *)&iph->daddr - skb_network_header(skb); 426 serr->port = port; 427 428 __skb_pull(skb, skb_tail_pointer(skb) - skb->data); 429 skb_reset_transport_header(skb); 430 431 if (sock_queue_err_skb(sk, skb)) 432 kfree_skb(skb); 433 } 434 435 /* IPv4 supports cmsg on all imcp errors and some timestamps 436 * 437 * Timestamp code paths do not initialize the fields expected by cmsg: 438 * the PKTINFO fields in skb->cb[]. Fill those in here. 439 */ 440 static bool ipv4_datagram_support_cmsg(const struct sock *sk, 441 struct sk_buff *skb, 442 int ee_origin) 443 { 444 struct in_pktinfo *info; 445 446 if (ee_origin == SO_EE_ORIGIN_ICMP) 447 return true; 448 449 if (ee_origin == SO_EE_ORIGIN_LOCAL) 450 return false; 451 452 /* Support IP_PKTINFO on tstamp packets if requested, to correlate 453 * timestamp with egress dev. Not possible for packets without dev 454 * or without payload (SOF_TIMESTAMPING_OPT_TSONLY). 455 */ 456 if ((!(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_CMSG)) || 457 (!skb->dev)) 458 return false; 459 460 info = PKTINFO_SKB_CB(skb); 461 info->ipi_spec_dst.s_addr = ip_hdr(skb)->saddr; 462 info->ipi_ifindex = skb->dev->ifindex; 463 return true; 464 } 465 466 /* 467 * Handle MSG_ERRQUEUE 468 */ 469 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len) 470 { 471 struct sock_exterr_skb *serr; 472 struct sk_buff *skb; 473 DECLARE_SOCKADDR(struct sockaddr_in *, sin, msg->msg_name); 474 struct { 475 struct sock_extended_err ee; 476 struct sockaddr_in offender; 477 } errhdr; 478 int err; 479 int copied; 480 481 WARN_ON_ONCE(sk->sk_family == AF_INET6); 482 483 err = -EAGAIN; 484 skb = sock_dequeue_err_skb(sk); 485 if (skb == NULL) 486 goto out; 487 488 copied = skb->len; 489 if (copied > len) { 490 msg->msg_flags |= MSG_TRUNC; 491 copied = len; 492 } 493 err = skb_copy_datagram_msg(skb, 0, msg, copied); 494 if (err) 495 goto out_free_skb; 496 497 sock_recv_timestamp(msg, sk, skb); 498 499 serr = SKB_EXT_ERR(skb); 500 501 if (sin && serr->port) { 502 sin->sin_family = AF_INET; 503 sin->sin_addr.s_addr = *(__be32 *)(skb_network_header(skb) + 504 serr->addr_offset); 505 sin->sin_port = serr->port; 506 memset(&sin->sin_zero, 0, sizeof(sin->sin_zero)); 507 *addr_len = sizeof(*sin); 508 } 509 510 memcpy(&errhdr.ee, &serr->ee, sizeof(struct sock_extended_err)); 511 sin = &errhdr.offender; 512 memset(sin, 0, sizeof(*sin)); 513 514 if (ipv4_datagram_support_cmsg(sk, skb, serr->ee.ee_origin)) { 515 sin->sin_family = AF_INET; 516 sin->sin_addr.s_addr = ip_hdr(skb)->saddr; 517 if (inet_sk(sk)->cmsg_flags) 518 ip_cmsg_recv(msg, skb); 519 } 520 521 put_cmsg(msg, SOL_IP, IP_RECVERR, sizeof(errhdr), &errhdr); 522 523 /* Now we could try to dump offended packet options */ 524 525 msg->msg_flags |= MSG_ERRQUEUE; 526 err = copied; 527 528 out_free_skb: 529 kfree_skb(skb); 530 out: 531 return err; 532 } 533 534 535 /* 536 * Socket option code for IP. This is the end of the line after any 537 * TCP,UDP etc options on an IP socket. 538 */ 539 540 static int do_ip_setsockopt(struct sock *sk, int level, 541 int optname, char __user *optval, unsigned int optlen) 542 { 543 struct inet_sock *inet = inet_sk(sk); 544 int val = 0, err; 545 546 switch (optname) { 547 case IP_PKTINFO: 548 case IP_RECVTTL: 549 case IP_RECVOPTS: 550 case IP_RECVTOS: 551 case IP_RETOPTS: 552 case IP_TOS: 553 case IP_TTL: 554 case IP_HDRINCL: 555 case IP_MTU_DISCOVER: 556 case IP_RECVERR: 557 case IP_ROUTER_ALERT: 558 case IP_FREEBIND: 559 case IP_PASSSEC: 560 case IP_TRANSPARENT: 561 case IP_MINTTL: 562 case IP_NODEFRAG: 563 case IP_UNICAST_IF: 564 case IP_MULTICAST_TTL: 565 case IP_MULTICAST_ALL: 566 case IP_MULTICAST_LOOP: 567 case IP_RECVORIGDSTADDR: 568 case IP_CHECKSUM: 569 if (optlen >= sizeof(int)) { 570 if (get_user(val, (int __user *) optval)) 571 return -EFAULT; 572 } else if (optlen >= sizeof(char)) { 573 unsigned char ucval; 574 575 if (get_user(ucval, (unsigned char __user *) optval)) 576 return -EFAULT; 577 val = (int) ucval; 578 } 579 } 580 581 /* If optlen==0, it is equivalent to val == 0 */ 582 583 if (ip_mroute_opt(optname)) 584 return ip_mroute_setsockopt(sk, optname, optval, optlen); 585 586 err = 0; 587 lock_sock(sk); 588 589 switch (optname) { 590 case IP_OPTIONS: 591 { 592 struct ip_options_rcu *old, *opt = NULL; 593 594 if (optlen > 40) 595 goto e_inval; 596 err = ip_options_get_from_user(sock_net(sk), &opt, 597 optval, optlen); 598 if (err) 599 break; 600 old = rcu_dereference_protected(inet->inet_opt, 601 sock_owned_by_user(sk)); 602 if (inet->is_icsk) { 603 struct inet_connection_sock *icsk = inet_csk(sk); 604 #if IS_ENABLED(CONFIG_IPV6) 605 if (sk->sk_family == PF_INET || 606 (!((1 << sk->sk_state) & 607 (TCPF_LISTEN | TCPF_CLOSE)) && 608 inet->inet_daddr != LOOPBACK4_IPV6)) { 609 #endif 610 if (old) 611 icsk->icsk_ext_hdr_len -= old->opt.optlen; 612 if (opt) 613 icsk->icsk_ext_hdr_len += opt->opt.optlen; 614 icsk->icsk_sync_mss(sk, icsk->icsk_pmtu_cookie); 615 #if IS_ENABLED(CONFIG_IPV6) 616 } 617 #endif 618 } 619 rcu_assign_pointer(inet->inet_opt, opt); 620 if (old) 621 kfree_rcu(old, rcu); 622 break; 623 } 624 case IP_PKTINFO: 625 if (val) 626 inet->cmsg_flags |= IP_CMSG_PKTINFO; 627 else 628 inet->cmsg_flags &= ~IP_CMSG_PKTINFO; 629 break; 630 case IP_RECVTTL: 631 if (val) 632 inet->cmsg_flags |= IP_CMSG_TTL; 633 else 634 inet->cmsg_flags &= ~IP_CMSG_TTL; 635 break; 636 case IP_RECVTOS: 637 if (val) 638 inet->cmsg_flags |= IP_CMSG_TOS; 639 else 640 inet->cmsg_flags &= ~IP_CMSG_TOS; 641 break; 642 case IP_RECVOPTS: 643 if (val) 644 inet->cmsg_flags |= IP_CMSG_RECVOPTS; 645 else 646 inet->cmsg_flags &= ~IP_CMSG_RECVOPTS; 647 break; 648 case IP_RETOPTS: 649 if (val) 650 inet->cmsg_flags |= IP_CMSG_RETOPTS; 651 else 652 inet->cmsg_flags &= ~IP_CMSG_RETOPTS; 653 break; 654 case IP_PASSSEC: 655 if (val) 656 inet->cmsg_flags |= IP_CMSG_PASSSEC; 657 else 658 inet->cmsg_flags &= ~IP_CMSG_PASSSEC; 659 break; 660 case IP_RECVORIGDSTADDR: 661 if (val) 662 inet->cmsg_flags |= IP_CMSG_ORIGDSTADDR; 663 else 664 inet->cmsg_flags &= ~IP_CMSG_ORIGDSTADDR; 665 break; 666 case IP_CHECKSUM: 667 if (val) { 668 if (!(inet->cmsg_flags & IP_CMSG_CHECKSUM)) { 669 inet_inc_convert_csum(sk); 670 inet->cmsg_flags |= IP_CMSG_CHECKSUM; 671 } 672 } else { 673 if (inet->cmsg_flags & IP_CMSG_CHECKSUM) { 674 inet_dec_convert_csum(sk); 675 inet->cmsg_flags &= ~IP_CMSG_CHECKSUM; 676 } 677 } 678 break; 679 case IP_TOS: /* This sets both TOS and Precedence */ 680 if (sk->sk_type == SOCK_STREAM) { 681 val &= ~INET_ECN_MASK; 682 val |= inet->tos & INET_ECN_MASK; 683 } 684 if (inet->tos != val) { 685 inet->tos = val; 686 sk->sk_priority = rt_tos2priority(val); 687 sk_dst_reset(sk); 688 } 689 break; 690 case IP_TTL: 691 if (optlen < 1) 692 goto e_inval; 693 if (val != -1 && (val < 1 || val > 255)) 694 goto e_inval; 695 inet->uc_ttl = val; 696 break; 697 case IP_HDRINCL: 698 if (sk->sk_type != SOCK_RAW) { 699 err = -ENOPROTOOPT; 700 break; 701 } 702 inet->hdrincl = val ? 1 : 0; 703 break; 704 case IP_NODEFRAG: 705 if (sk->sk_type != SOCK_RAW) { 706 err = -ENOPROTOOPT; 707 break; 708 } 709 inet->nodefrag = val ? 1 : 0; 710 break; 711 case IP_MTU_DISCOVER: 712 if (val < IP_PMTUDISC_DONT || val > IP_PMTUDISC_OMIT) 713 goto e_inval; 714 inet->pmtudisc = val; 715 break; 716 case IP_RECVERR: 717 inet->recverr = !!val; 718 if (!val) 719 skb_queue_purge(&sk->sk_error_queue); 720 break; 721 case IP_MULTICAST_TTL: 722 if (sk->sk_type == SOCK_STREAM) 723 goto e_inval; 724 if (optlen < 1) 725 goto e_inval; 726 if (val == -1) 727 val = 1; 728 if (val < 0 || val > 255) 729 goto e_inval; 730 inet->mc_ttl = val; 731 break; 732 case IP_MULTICAST_LOOP: 733 if (optlen < 1) 734 goto e_inval; 735 inet->mc_loop = !!val; 736 break; 737 case IP_UNICAST_IF: 738 { 739 struct net_device *dev = NULL; 740 int ifindex; 741 742 if (optlen != sizeof(int)) 743 goto e_inval; 744 745 ifindex = (__force int)ntohl((__force __be32)val); 746 if (ifindex == 0) { 747 inet->uc_index = 0; 748 err = 0; 749 break; 750 } 751 752 dev = dev_get_by_index(sock_net(sk), ifindex); 753 err = -EADDRNOTAVAIL; 754 if (!dev) 755 break; 756 dev_put(dev); 757 758 err = -EINVAL; 759 if (sk->sk_bound_dev_if) 760 break; 761 762 inet->uc_index = ifindex; 763 err = 0; 764 break; 765 } 766 case IP_MULTICAST_IF: 767 { 768 struct ip_mreqn mreq; 769 struct net_device *dev = NULL; 770 771 if (sk->sk_type == SOCK_STREAM) 772 goto e_inval; 773 /* 774 * Check the arguments are allowable 775 */ 776 777 if (optlen < sizeof(struct in_addr)) 778 goto e_inval; 779 780 err = -EFAULT; 781 if (optlen >= sizeof(struct ip_mreqn)) { 782 if (copy_from_user(&mreq, optval, sizeof(mreq))) 783 break; 784 } else { 785 memset(&mreq, 0, sizeof(mreq)); 786 if (optlen >= sizeof(struct ip_mreq)) { 787 if (copy_from_user(&mreq, optval, 788 sizeof(struct ip_mreq))) 789 break; 790 } else if (optlen >= sizeof(struct in_addr)) { 791 if (copy_from_user(&mreq.imr_address, optval, 792 sizeof(struct in_addr))) 793 break; 794 } 795 } 796 797 if (!mreq.imr_ifindex) { 798 if (mreq.imr_address.s_addr == htonl(INADDR_ANY)) { 799 inet->mc_index = 0; 800 inet->mc_addr = 0; 801 err = 0; 802 break; 803 } 804 dev = ip_dev_find(sock_net(sk), mreq.imr_address.s_addr); 805 if (dev) 806 mreq.imr_ifindex = dev->ifindex; 807 } else 808 dev = dev_get_by_index(sock_net(sk), mreq.imr_ifindex); 809 810 811 err = -EADDRNOTAVAIL; 812 if (!dev) 813 break; 814 dev_put(dev); 815 816 err = -EINVAL; 817 if (sk->sk_bound_dev_if && 818 mreq.imr_ifindex != sk->sk_bound_dev_if) 819 break; 820 821 inet->mc_index = mreq.imr_ifindex; 822 inet->mc_addr = mreq.imr_address.s_addr; 823 err = 0; 824 break; 825 } 826 827 case IP_ADD_MEMBERSHIP: 828 case IP_DROP_MEMBERSHIP: 829 { 830 struct ip_mreqn mreq; 831 832 err = -EPROTO; 833 if (inet_sk(sk)->is_icsk) 834 break; 835 836 if (optlen < sizeof(struct ip_mreq)) 837 goto e_inval; 838 err = -EFAULT; 839 if (optlen >= sizeof(struct ip_mreqn)) { 840 if (copy_from_user(&mreq, optval, sizeof(mreq))) 841 break; 842 } else { 843 memset(&mreq, 0, sizeof(mreq)); 844 if (copy_from_user(&mreq, optval, sizeof(struct ip_mreq))) 845 break; 846 } 847 848 if (optname == IP_ADD_MEMBERSHIP) 849 err = ip_mc_join_group(sk, &mreq); 850 else 851 err = ip_mc_leave_group(sk, &mreq); 852 break; 853 } 854 case IP_MSFILTER: 855 { 856 struct ip_msfilter *msf; 857 858 if (optlen < IP_MSFILTER_SIZE(0)) 859 goto e_inval; 860 if (optlen > sysctl_optmem_max) { 861 err = -ENOBUFS; 862 break; 863 } 864 msf = kmalloc(optlen, GFP_KERNEL); 865 if (!msf) { 866 err = -ENOBUFS; 867 break; 868 } 869 err = -EFAULT; 870 if (copy_from_user(msf, optval, optlen)) { 871 kfree(msf); 872 break; 873 } 874 /* numsrc >= (1G-4) overflow in 32 bits */ 875 if (msf->imsf_numsrc >= 0x3ffffffcU || 876 msf->imsf_numsrc > sysctl_igmp_max_msf) { 877 kfree(msf); 878 err = -ENOBUFS; 879 break; 880 } 881 if (IP_MSFILTER_SIZE(msf->imsf_numsrc) > optlen) { 882 kfree(msf); 883 err = -EINVAL; 884 break; 885 } 886 err = ip_mc_msfilter(sk, msf, 0); 887 kfree(msf); 888 break; 889 } 890 case IP_BLOCK_SOURCE: 891 case IP_UNBLOCK_SOURCE: 892 case IP_ADD_SOURCE_MEMBERSHIP: 893 case IP_DROP_SOURCE_MEMBERSHIP: 894 { 895 struct ip_mreq_source mreqs; 896 int omode, add; 897 898 if (optlen != sizeof(struct ip_mreq_source)) 899 goto e_inval; 900 if (copy_from_user(&mreqs, optval, sizeof(mreqs))) { 901 err = -EFAULT; 902 break; 903 } 904 if (optname == IP_BLOCK_SOURCE) { 905 omode = MCAST_EXCLUDE; 906 add = 1; 907 } else if (optname == IP_UNBLOCK_SOURCE) { 908 omode = MCAST_EXCLUDE; 909 add = 0; 910 } else if (optname == IP_ADD_SOURCE_MEMBERSHIP) { 911 struct ip_mreqn mreq; 912 913 mreq.imr_multiaddr.s_addr = mreqs.imr_multiaddr; 914 mreq.imr_address.s_addr = mreqs.imr_interface; 915 mreq.imr_ifindex = 0; 916 err = ip_mc_join_group(sk, &mreq); 917 if (err && err != -EADDRINUSE) 918 break; 919 omode = MCAST_INCLUDE; 920 add = 1; 921 } else /* IP_DROP_SOURCE_MEMBERSHIP */ { 922 omode = MCAST_INCLUDE; 923 add = 0; 924 } 925 err = ip_mc_source(add, omode, sk, &mreqs, 0); 926 break; 927 } 928 case MCAST_JOIN_GROUP: 929 case MCAST_LEAVE_GROUP: 930 { 931 struct group_req greq; 932 struct sockaddr_in *psin; 933 struct ip_mreqn mreq; 934 935 if (optlen < sizeof(struct group_req)) 936 goto e_inval; 937 err = -EFAULT; 938 if (copy_from_user(&greq, optval, sizeof(greq))) 939 break; 940 psin = (struct sockaddr_in *)&greq.gr_group; 941 if (psin->sin_family != AF_INET) 942 goto e_inval; 943 memset(&mreq, 0, sizeof(mreq)); 944 mreq.imr_multiaddr = psin->sin_addr; 945 mreq.imr_ifindex = greq.gr_interface; 946 947 if (optname == MCAST_JOIN_GROUP) 948 err = ip_mc_join_group(sk, &mreq); 949 else 950 err = ip_mc_leave_group(sk, &mreq); 951 break; 952 } 953 case MCAST_JOIN_SOURCE_GROUP: 954 case MCAST_LEAVE_SOURCE_GROUP: 955 case MCAST_BLOCK_SOURCE: 956 case MCAST_UNBLOCK_SOURCE: 957 { 958 struct group_source_req greqs; 959 struct ip_mreq_source mreqs; 960 struct sockaddr_in *psin; 961 int omode, add; 962 963 if (optlen != sizeof(struct group_source_req)) 964 goto e_inval; 965 if (copy_from_user(&greqs, optval, sizeof(greqs))) { 966 err = -EFAULT; 967 break; 968 } 969 if (greqs.gsr_group.ss_family != AF_INET || 970 greqs.gsr_source.ss_family != AF_INET) { 971 err = -EADDRNOTAVAIL; 972 break; 973 } 974 psin = (struct sockaddr_in *)&greqs.gsr_group; 975 mreqs.imr_multiaddr = psin->sin_addr.s_addr; 976 psin = (struct sockaddr_in *)&greqs.gsr_source; 977 mreqs.imr_sourceaddr = psin->sin_addr.s_addr; 978 mreqs.imr_interface = 0; /* use index for mc_source */ 979 980 if (optname == MCAST_BLOCK_SOURCE) { 981 omode = MCAST_EXCLUDE; 982 add = 1; 983 } else if (optname == MCAST_UNBLOCK_SOURCE) { 984 omode = MCAST_EXCLUDE; 985 add = 0; 986 } else if (optname == MCAST_JOIN_SOURCE_GROUP) { 987 struct ip_mreqn mreq; 988 989 psin = (struct sockaddr_in *)&greqs.gsr_group; 990 mreq.imr_multiaddr = psin->sin_addr; 991 mreq.imr_address.s_addr = 0; 992 mreq.imr_ifindex = greqs.gsr_interface; 993 err = ip_mc_join_group(sk, &mreq); 994 if (err && err != -EADDRINUSE) 995 break; 996 greqs.gsr_interface = mreq.imr_ifindex; 997 omode = MCAST_INCLUDE; 998 add = 1; 999 } else /* MCAST_LEAVE_SOURCE_GROUP */ { 1000 omode = MCAST_INCLUDE; 1001 add = 0; 1002 } 1003 err = ip_mc_source(add, omode, sk, &mreqs, 1004 greqs.gsr_interface); 1005 break; 1006 } 1007 case MCAST_MSFILTER: 1008 { 1009 struct sockaddr_in *psin; 1010 struct ip_msfilter *msf = NULL; 1011 struct group_filter *gsf = NULL; 1012 int msize, i, ifindex; 1013 1014 if (optlen < GROUP_FILTER_SIZE(0)) 1015 goto e_inval; 1016 if (optlen > sysctl_optmem_max) { 1017 err = -ENOBUFS; 1018 break; 1019 } 1020 gsf = kmalloc(optlen, GFP_KERNEL); 1021 if (!gsf) { 1022 err = -ENOBUFS; 1023 break; 1024 } 1025 err = -EFAULT; 1026 if (copy_from_user(gsf, optval, optlen)) 1027 goto mc_msf_out; 1028 1029 /* numsrc >= (4G-140)/128 overflow in 32 bits */ 1030 if (gsf->gf_numsrc >= 0x1ffffff || 1031 gsf->gf_numsrc > sysctl_igmp_max_msf) { 1032 err = -ENOBUFS; 1033 goto mc_msf_out; 1034 } 1035 if (GROUP_FILTER_SIZE(gsf->gf_numsrc) > optlen) { 1036 err = -EINVAL; 1037 goto mc_msf_out; 1038 } 1039 msize = IP_MSFILTER_SIZE(gsf->gf_numsrc); 1040 msf = kmalloc(msize, GFP_KERNEL); 1041 if (!msf) { 1042 err = -ENOBUFS; 1043 goto mc_msf_out; 1044 } 1045 ifindex = gsf->gf_interface; 1046 psin = (struct sockaddr_in *)&gsf->gf_group; 1047 if (psin->sin_family != AF_INET) { 1048 err = -EADDRNOTAVAIL; 1049 goto mc_msf_out; 1050 } 1051 msf->imsf_multiaddr = psin->sin_addr.s_addr; 1052 msf->imsf_interface = 0; 1053 msf->imsf_fmode = gsf->gf_fmode; 1054 msf->imsf_numsrc = gsf->gf_numsrc; 1055 err = -EADDRNOTAVAIL; 1056 for (i = 0; i < gsf->gf_numsrc; ++i) { 1057 psin = (struct sockaddr_in *)&gsf->gf_slist[i]; 1058 1059 if (psin->sin_family != AF_INET) 1060 goto mc_msf_out; 1061 msf->imsf_slist[i] = psin->sin_addr.s_addr; 1062 } 1063 kfree(gsf); 1064 gsf = NULL; 1065 1066 err = ip_mc_msfilter(sk, msf, ifindex); 1067 mc_msf_out: 1068 kfree(msf); 1069 kfree(gsf); 1070 break; 1071 } 1072 case IP_MULTICAST_ALL: 1073 if (optlen < 1) 1074 goto e_inval; 1075 if (val != 0 && val != 1) 1076 goto e_inval; 1077 inet->mc_all = val; 1078 break; 1079 case IP_ROUTER_ALERT: 1080 err = ip_ra_control(sk, val ? 1 : 0, NULL); 1081 break; 1082 1083 case IP_FREEBIND: 1084 if (optlen < 1) 1085 goto e_inval; 1086 inet->freebind = !!val; 1087 break; 1088 1089 case IP_IPSEC_POLICY: 1090 case IP_XFRM_POLICY: 1091 err = -EPERM; 1092 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) 1093 break; 1094 err = xfrm_user_policy(sk, optname, optval, optlen); 1095 break; 1096 1097 case IP_TRANSPARENT: 1098 if (!!val && !ns_capable(sock_net(sk)->user_ns, CAP_NET_RAW) && 1099 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) { 1100 err = -EPERM; 1101 break; 1102 } 1103 if (optlen < 1) 1104 goto e_inval; 1105 inet->transparent = !!val; 1106 break; 1107 1108 case IP_MINTTL: 1109 if (optlen < 1) 1110 goto e_inval; 1111 if (val < 0 || val > 255) 1112 goto e_inval; 1113 inet->min_ttl = val; 1114 break; 1115 1116 default: 1117 err = -ENOPROTOOPT; 1118 break; 1119 } 1120 release_sock(sk); 1121 return err; 1122 1123 e_inval: 1124 release_sock(sk); 1125 return -EINVAL; 1126 } 1127 1128 /** 1129 * ipv4_pktinfo_prepare - transfer some info from rtable to skb 1130 * @sk: socket 1131 * @skb: buffer 1132 * 1133 * To support IP_CMSG_PKTINFO option, we store rt_iif and specific 1134 * destination in skb->cb[] before dst drop. 1135 * This way, receiver doesn't make cache line misses to read rtable. 1136 */ 1137 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb) 1138 { 1139 struct in_pktinfo *pktinfo = PKTINFO_SKB_CB(skb); 1140 bool prepare = (inet_sk(sk)->cmsg_flags & IP_CMSG_PKTINFO) || 1141 ipv6_sk_rxinfo(sk); 1142 1143 if (prepare && skb_rtable(skb)) { 1144 pktinfo->ipi_ifindex = inet_iif(skb); 1145 pktinfo->ipi_spec_dst.s_addr = fib_compute_spec_dst(skb); 1146 } else { 1147 pktinfo->ipi_ifindex = 0; 1148 pktinfo->ipi_spec_dst.s_addr = 0; 1149 } 1150 skb_dst_drop(skb); 1151 } 1152 1153 int ip_setsockopt(struct sock *sk, int level, 1154 int optname, char __user *optval, unsigned int optlen) 1155 { 1156 int err; 1157 1158 if (level != SOL_IP) 1159 return -ENOPROTOOPT; 1160 1161 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1162 #ifdef CONFIG_NETFILTER 1163 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1164 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1165 optname != IP_IPSEC_POLICY && 1166 optname != IP_XFRM_POLICY && 1167 !ip_mroute_opt(optname)) { 1168 lock_sock(sk); 1169 err = nf_setsockopt(sk, PF_INET, optname, optval, optlen); 1170 release_sock(sk); 1171 } 1172 #endif 1173 return err; 1174 } 1175 EXPORT_SYMBOL(ip_setsockopt); 1176 1177 #ifdef CONFIG_COMPAT 1178 int compat_ip_setsockopt(struct sock *sk, int level, int optname, 1179 char __user *optval, unsigned int optlen) 1180 { 1181 int err; 1182 1183 if (level != SOL_IP) 1184 return -ENOPROTOOPT; 1185 1186 if (optname >= MCAST_JOIN_GROUP && optname <= MCAST_MSFILTER) 1187 return compat_mc_setsockopt(sk, level, optname, optval, optlen, 1188 ip_setsockopt); 1189 1190 err = do_ip_setsockopt(sk, level, optname, optval, optlen); 1191 #ifdef CONFIG_NETFILTER 1192 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1193 if (err == -ENOPROTOOPT && optname != IP_HDRINCL && 1194 optname != IP_IPSEC_POLICY && 1195 optname != IP_XFRM_POLICY && 1196 !ip_mroute_opt(optname)) { 1197 lock_sock(sk); 1198 err = compat_nf_setsockopt(sk, PF_INET, optname, 1199 optval, optlen); 1200 release_sock(sk); 1201 } 1202 #endif 1203 return err; 1204 } 1205 EXPORT_SYMBOL(compat_ip_setsockopt); 1206 #endif 1207 1208 /* 1209 * Get the options. Note for future reference. The GET of IP options gets 1210 * the _received_ ones. The set sets the _sent_ ones. 1211 */ 1212 1213 static int do_ip_getsockopt(struct sock *sk, int level, int optname, 1214 char __user *optval, int __user *optlen, unsigned int flags) 1215 { 1216 struct inet_sock *inet = inet_sk(sk); 1217 int val; 1218 int len; 1219 1220 if (level != SOL_IP) 1221 return -EOPNOTSUPP; 1222 1223 if (ip_mroute_opt(optname)) 1224 return ip_mroute_getsockopt(sk, optname, optval, optlen); 1225 1226 if (get_user(len, optlen)) 1227 return -EFAULT; 1228 if (len < 0) 1229 return -EINVAL; 1230 1231 lock_sock(sk); 1232 1233 switch (optname) { 1234 case IP_OPTIONS: 1235 { 1236 unsigned char optbuf[sizeof(struct ip_options)+40]; 1237 struct ip_options *opt = (struct ip_options *)optbuf; 1238 struct ip_options_rcu *inet_opt; 1239 1240 inet_opt = rcu_dereference_protected(inet->inet_opt, 1241 sock_owned_by_user(sk)); 1242 opt->optlen = 0; 1243 if (inet_opt) 1244 memcpy(optbuf, &inet_opt->opt, 1245 sizeof(struct ip_options) + 1246 inet_opt->opt.optlen); 1247 release_sock(sk); 1248 1249 if (opt->optlen == 0) 1250 return put_user(0, optlen); 1251 1252 ip_options_undo(opt); 1253 1254 len = min_t(unsigned int, len, opt->optlen); 1255 if (put_user(len, optlen)) 1256 return -EFAULT; 1257 if (copy_to_user(optval, opt->__data, len)) 1258 return -EFAULT; 1259 return 0; 1260 } 1261 case IP_PKTINFO: 1262 val = (inet->cmsg_flags & IP_CMSG_PKTINFO) != 0; 1263 break; 1264 case IP_RECVTTL: 1265 val = (inet->cmsg_flags & IP_CMSG_TTL) != 0; 1266 break; 1267 case IP_RECVTOS: 1268 val = (inet->cmsg_flags & IP_CMSG_TOS) != 0; 1269 break; 1270 case IP_RECVOPTS: 1271 val = (inet->cmsg_flags & IP_CMSG_RECVOPTS) != 0; 1272 break; 1273 case IP_RETOPTS: 1274 val = (inet->cmsg_flags & IP_CMSG_RETOPTS) != 0; 1275 break; 1276 case IP_PASSSEC: 1277 val = (inet->cmsg_flags & IP_CMSG_PASSSEC) != 0; 1278 break; 1279 case IP_RECVORIGDSTADDR: 1280 val = (inet->cmsg_flags & IP_CMSG_ORIGDSTADDR) != 0; 1281 break; 1282 case IP_CHECKSUM: 1283 val = (inet->cmsg_flags & IP_CMSG_CHECKSUM) != 0; 1284 break; 1285 case IP_TOS: 1286 val = inet->tos; 1287 break; 1288 case IP_TTL: 1289 val = (inet->uc_ttl == -1 ? 1290 sysctl_ip_default_ttl : 1291 inet->uc_ttl); 1292 break; 1293 case IP_HDRINCL: 1294 val = inet->hdrincl; 1295 break; 1296 case IP_NODEFRAG: 1297 val = inet->nodefrag; 1298 break; 1299 case IP_MTU_DISCOVER: 1300 val = inet->pmtudisc; 1301 break; 1302 case IP_MTU: 1303 { 1304 struct dst_entry *dst; 1305 val = 0; 1306 dst = sk_dst_get(sk); 1307 if (dst) { 1308 val = dst_mtu(dst); 1309 dst_release(dst); 1310 } 1311 if (!val) { 1312 release_sock(sk); 1313 return -ENOTCONN; 1314 } 1315 break; 1316 } 1317 case IP_RECVERR: 1318 val = inet->recverr; 1319 break; 1320 case IP_MULTICAST_TTL: 1321 val = inet->mc_ttl; 1322 break; 1323 case IP_MULTICAST_LOOP: 1324 val = inet->mc_loop; 1325 break; 1326 case IP_UNICAST_IF: 1327 val = (__force int)htonl((__u32) inet->uc_index); 1328 break; 1329 case IP_MULTICAST_IF: 1330 { 1331 struct in_addr addr; 1332 len = min_t(unsigned int, len, sizeof(struct in_addr)); 1333 addr.s_addr = inet->mc_addr; 1334 release_sock(sk); 1335 1336 if (put_user(len, optlen)) 1337 return -EFAULT; 1338 if (copy_to_user(optval, &addr, len)) 1339 return -EFAULT; 1340 return 0; 1341 } 1342 case IP_MSFILTER: 1343 { 1344 struct ip_msfilter msf; 1345 int err; 1346 1347 if (len < IP_MSFILTER_SIZE(0)) { 1348 release_sock(sk); 1349 return -EINVAL; 1350 } 1351 if (copy_from_user(&msf, optval, IP_MSFILTER_SIZE(0))) { 1352 release_sock(sk); 1353 return -EFAULT; 1354 } 1355 err = ip_mc_msfget(sk, &msf, 1356 (struct ip_msfilter __user *)optval, optlen); 1357 release_sock(sk); 1358 return err; 1359 } 1360 case MCAST_MSFILTER: 1361 { 1362 struct group_filter gsf; 1363 int err; 1364 1365 if (len < GROUP_FILTER_SIZE(0)) { 1366 release_sock(sk); 1367 return -EINVAL; 1368 } 1369 if (copy_from_user(&gsf, optval, GROUP_FILTER_SIZE(0))) { 1370 release_sock(sk); 1371 return -EFAULT; 1372 } 1373 err = ip_mc_gsfget(sk, &gsf, 1374 (struct group_filter __user *)optval, 1375 optlen); 1376 release_sock(sk); 1377 return err; 1378 } 1379 case IP_MULTICAST_ALL: 1380 val = inet->mc_all; 1381 break; 1382 case IP_PKTOPTIONS: 1383 { 1384 struct msghdr msg; 1385 1386 release_sock(sk); 1387 1388 if (sk->sk_type != SOCK_STREAM) 1389 return -ENOPROTOOPT; 1390 1391 msg.msg_control = (__force void *) optval; 1392 msg.msg_controllen = len; 1393 msg.msg_flags = flags; 1394 1395 if (inet->cmsg_flags & IP_CMSG_PKTINFO) { 1396 struct in_pktinfo info; 1397 1398 info.ipi_addr.s_addr = inet->inet_rcv_saddr; 1399 info.ipi_spec_dst.s_addr = inet->inet_rcv_saddr; 1400 info.ipi_ifindex = inet->mc_index; 1401 put_cmsg(&msg, SOL_IP, IP_PKTINFO, sizeof(info), &info); 1402 } 1403 if (inet->cmsg_flags & IP_CMSG_TTL) { 1404 int hlim = inet->mc_ttl; 1405 put_cmsg(&msg, SOL_IP, IP_TTL, sizeof(hlim), &hlim); 1406 } 1407 if (inet->cmsg_flags & IP_CMSG_TOS) { 1408 int tos = inet->rcv_tos; 1409 put_cmsg(&msg, SOL_IP, IP_TOS, sizeof(tos), &tos); 1410 } 1411 len -= msg.msg_controllen; 1412 return put_user(len, optlen); 1413 } 1414 case IP_FREEBIND: 1415 val = inet->freebind; 1416 break; 1417 case IP_TRANSPARENT: 1418 val = inet->transparent; 1419 break; 1420 case IP_MINTTL: 1421 val = inet->min_ttl; 1422 break; 1423 default: 1424 release_sock(sk); 1425 return -ENOPROTOOPT; 1426 } 1427 release_sock(sk); 1428 1429 if (len < sizeof(int) && len > 0 && val >= 0 && val <= 255) { 1430 unsigned char ucval = (unsigned char)val; 1431 len = 1; 1432 if (put_user(len, optlen)) 1433 return -EFAULT; 1434 if (copy_to_user(optval, &ucval, 1)) 1435 return -EFAULT; 1436 } else { 1437 len = min_t(unsigned int, sizeof(int), len); 1438 if (put_user(len, optlen)) 1439 return -EFAULT; 1440 if (copy_to_user(optval, &val, len)) 1441 return -EFAULT; 1442 } 1443 return 0; 1444 } 1445 1446 int ip_getsockopt(struct sock *sk, int level, 1447 int optname, char __user *optval, int __user *optlen) 1448 { 1449 int err; 1450 1451 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 0); 1452 #ifdef CONFIG_NETFILTER 1453 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1454 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1455 !ip_mroute_opt(optname)) { 1456 int len; 1457 1458 if (get_user(len, optlen)) 1459 return -EFAULT; 1460 1461 lock_sock(sk); 1462 err = nf_getsockopt(sk, PF_INET, optname, optval, 1463 &len); 1464 release_sock(sk); 1465 if (err >= 0) 1466 err = put_user(len, optlen); 1467 return err; 1468 } 1469 #endif 1470 return err; 1471 } 1472 EXPORT_SYMBOL(ip_getsockopt); 1473 1474 #ifdef CONFIG_COMPAT 1475 int compat_ip_getsockopt(struct sock *sk, int level, int optname, 1476 char __user *optval, int __user *optlen) 1477 { 1478 int err; 1479 1480 if (optname == MCAST_MSFILTER) 1481 return compat_mc_getsockopt(sk, level, optname, optval, optlen, 1482 ip_getsockopt); 1483 1484 err = do_ip_getsockopt(sk, level, optname, optval, optlen, 1485 MSG_CMSG_COMPAT); 1486 1487 #ifdef CONFIG_NETFILTER 1488 /* we need to exclude all possible ENOPROTOOPTs except default case */ 1489 if (err == -ENOPROTOOPT && optname != IP_PKTOPTIONS && 1490 !ip_mroute_opt(optname)) { 1491 int len; 1492 1493 if (get_user(len, optlen)) 1494 return -EFAULT; 1495 1496 lock_sock(sk); 1497 err = compat_nf_getsockopt(sk, PF_INET, optname, optval, &len); 1498 release_sock(sk); 1499 if (err >= 0) 1500 err = put_user(len, optlen); 1501 return err; 1502 } 1503 #endif 1504 return err; 1505 } 1506 EXPORT_SYMBOL(compat_ip_getsockopt); 1507 #endif 1508