1 /* 2 * IPv6 Address [auto]configuration 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 8 * 9 * This program is free software; you can redistribute it and/or 10 * modify it under the terms of the GNU General Public License 11 * as published by the Free Software Foundation; either version 12 * 2 of the License, or (at your option) any later version. 13 */ 14 15 /* 16 * Changes: 17 * 18 * Janos Farkas : delete timer on ifdown 19 * <chexum@bankinf.banki.hu> 20 * Andi Kleen : kill double kfree on module 21 * unload. 22 * Maciej W. Rozycki : FDDI support 23 * sekiya@USAGI : Don't send too many RS 24 * packets. 25 * yoshfuji@USAGI : Fixed interval between DAD 26 * packets. 27 * YOSHIFUJI Hideaki @USAGI : improved accuracy of 28 * address validation timer. 29 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041) 30 * support. 31 * Yuji SEKIYA @USAGI : Don't assign a same IPv6 32 * address on a same interface. 33 * YOSHIFUJI Hideaki @USAGI : ARCnet support 34 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to 35 * seq_file. 36 * YOSHIFUJI Hideaki @USAGI : improved source address 37 * selection; consider scope, 38 * status etc. 39 */ 40 41 #define pr_fmt(fmt) "IPv6: " fmt 42 43 #include <linux/errno.h> 44 #include <linux/types.h> 45 #include <linux/kernel.h> 46 #include <linux/socket.h> 47 #include <linux/sockios.h> 48 #include <linux/net.h> 49 #include <linux/in6.h> 50 #include <linux/netdevice.h> 51 #include <linux/if_addr.h> 52 #include <linux/if_arp.h> 53 #include <linux/if_arcnet.h> 54 #include <linux/if_infiniband.h> 55 #include <linux/route.h> 56 #include <linux/inetdevice.h> 57 #include <linux/init.h> 58 #include <linux/slab.h> 59 #ifdef CONFIG_SYSCTL 60 #include <linux/sysctl.h> 61 #endif 62 #include <linux/capability.h> 63 #include <linux/delay.h> 64 #include <linux/notifier.h> 65 #include <linux/string.h> 66 #include <linux/hash.h> 67 68 #include <net/net_namespace.h> 69 #include <net/sock.h> 70 #include <net/snmp.h> 71 72 #include <net/af_ieee802154.h> 73 #include <net/firewire.h> 74 #include <net/ipv6.h> 75 #include <net/protocol.h> 76 #include <net/ndisc.h> 77 #include <net/ip6_route.h> 78 #include <net/addrconf.h> 79 #include <net/tcp.h> 80 #include <net/ip.h> 81 #include <net/netlink.h> 82 #include <net/pkt_sched.h> 83 #include <linux/if_tunnel.h> 84 #include <linux/rtnetlink.h> 85 #include <linux/netconf.h> 86 #include <linux/random.h> 87 #include <linux/uaccess.h> 88 #include <asm/unaligned.h> 89 90 #include <linux/proc_fs.h> 91 #include <linux/seq_file.h> 92 #include <linux/export.h> 93 94 /* Set to 3 to get tracing... */ 95 #define ACONF_DEBUG 2 96 97 #if ACONF_DEBUG >= 3 98 #define ADBG(fmt, ...) printk(fmt, ##__VA_ARGS__) 99 #else 100 #define ADBG(fmt, ...) do { if (0) printk(fmt, ##__VA_ARGS__); } while (0) 101 #endif 102 103 #define INFINITY_LIFE_TIME 0xFFFFFFFF 104 105 static inline u32 cstamp_delta(unsigned long cstamp) 106 { 107 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ; 108 } 109 110 #ifdef CONFIG_SYSCTL 111 static void addrconf_sysctl_register(struct inet6_dev *idev); 112 static void addrconf_sysctl_unregister(struct inet6_dev *idev); 113 #else 114 static inline void addrconf_sysctl_register(struct inet6_dev *idev) 115 { 116 } 117 118 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev) 119 { 120 } 121 #endif 122 123 static void __ipv6_regen_rndid(struct inet6_dev *idev); 124 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr); 125 static void ipv6_regen_rndid(unsigned long data); 126 127 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev); 128 static int ipv6_count_addresses(struct inet6_dev *idev); 129 130 /* 131 * Configured unicast address hash table 132 */ 133 static struct hlist_head inet6_addr_lst[IN6_ADDR_HSIZE]; 134 static DEFINE_SPINLOCK(addrconf_hash_lock); 135 136 static void addrconf_verify(unsigned long); 137 138 static DEFINE_TIMER(addr_chk_timer, addrconf_verify, 0, 0); 139 static DEFINE_SPINLOCK(addrconf_verify_lock); 140 141 static void addrconf_join_anycast(struct inet6_ifaddr *ifp); 142 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp); 143 144 static void addrconf_type_change(struct net_device *dev, 145 unsigned long event); 146 static int addrconf_ifdown(struct net_device *dev, int how); 147 148 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 149 int plen, 150 const struct net_device *dev, 151 u32 flags, u32 noflags); 152 153 static void addrconf_dad_start(struct inet6_ifaddr *ifp); 154 static void addrconf_dad_timer(unsigned long data); 155 static void addrconf_dad_completed(struct inet6_ifaddr *ifp); 156 static void addrconf_dad_run(struct inet6_dev *idev); 157 static void addrconf_rs_timer(unsigned long data); 158 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 159 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 160 161 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 162 struct prefix_info *pinfo); 163 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr, 164 struct net_device *dev); 165 166 static struct ipv6_devconf ipv6_devconf __read_mostly = { 167 .forwarding = 0, 168 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 169 .mtu6 = IPV6_MIN_MTU, 170 .accept_ra = 1, 171 .accept_redirects = 1, 172 .autoconf = 1, 173 .force_mld_version = 0, 174 .mldv1_unsolicited_report_interval = 10 * HZ, 175 .mldv2_unsolicited_report_interval = HZ, 176 .dad_transmits = 1, 177 .rtr_solicits = MAX_RTR_SOLICITATIONS, 178 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 179 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 180 .use_tempaddr = 0, 181 .temp_valid_lft = TEMP_VALID_LIFETIME, 182 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 183 .regen_max_retry = REGEN_MAX_RETRY, 184 .max_desync_factor = MAX_DESYNC_FACTOR, 185 .max_addresses = IPV6_MAX_ADDRESSES, 186 .accept_ra_defrtr = 1, 187 .accept_ra_pinfo = 1, 188 #ifdef CONFIG_IPV6_ROUTER_PREF 189 .accept_ra_rtr_pref = 1, 190 .rtr_probe_interval = 60 * HZ, 191 #ifdef CONFIG_IPV6_ROUTE_INFO 192 .accept_ra_rt_info_max_plen = 0, 193 #endif 194 #endif 195 .proxy_ndp = 0, 196 .accept_source_route = 0, /* we do not accept RH0 by default. */ 197 .disable_ipv6 = 0, 198 .accept_dad = 1, 199 .suppress_frag_ndisc = 1, 200 }; 201 202 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = { 203 .forwarding = 0, 204 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 205 .mtu6 = IPV6_MIN_MTU, 206 .accept_ra = 1, 207 .accept_redirects = 1, 208 .autoconf = 1, 209 .force_mld_version = 0, 210 .mldv1_unsolicited_report_interval = 10 * HZ, 211 .mldv2_unsolicited_report_interval = HZ, 212 .dad_transmits = 1, 213 .rtr_solicits = MAX_RTR_SOLICITATIONS, 214 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 215 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 216 .use_tempaddr = 0, 217 .temp_valid_lft = TEMP_VALID_LIFETIME, 218 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 219 .regen_max_retry = REGEN_MAX_RETRY, 220 .max_desync_factor = MAX_DESYNC_FACTOR, 221 .max_addresses = IPV6_MAX_ADDRESSES, 222 .accept_ra_defrtr = 1, 223 .accept_ra_pinfo = 1, 224 #ifdef CONFIG_IPV6_ROUTER_PREF 225 .accept_ra_rtr_pref = 1, 226 .rtr_probe_interval = 60 * HZ, 227 #ifdef CONFIG_IPV6_ROUTE_INFO 228 .accept_ra_rt_info_max_plen = 0, 229 #endif 230 #endif 231 .proxy_ndp = 0, 232 .accept_source_route = 0, /* we do not accept RH0 by default. */ 233 .disable_ipv6 = 0, 234 .accept_dad = 1, 235 .suppress_frag_ndisc = 1, 236 }; 237 238 /* Check if a valid qdisc is available */ 239 static inline bool addrconf_qdisc_ok(const struct net_device *dev) 240 { 241 return !qdisc_tx_is_noop(dev); 242 } 243 244 static void addrconf_del_rs_timer(struct inet6_dev *idev) 245 { 246 if (del_timer(&idev->rs_timer)) 247 __in6_dev_put(idev); 248 } 249 250 static void addrconf_del_dad_timer(struct inet6_ifaddr *ifp) 251 { 252 if (del_timer(&ifp->dad_timer)) 253 __in6_ifa_put(ifp); 254 } 255 256 static void addrconf_mod_rs_timer(struct inet6_dev *idev, 257 unsigned long when) 258 { 259 if (!timer_pending(&idev->rs_timer)) 260 in6_dev_hold(idev); 261 mod_timer(&idev->rs_timer, jiffies + when); 262 } 263 264 static void addrconf_mod_dad_timer(struct inet6_ifaddr *ifp, 265 unsigned long when) 266 { 267 if (!timer_pending(&ifp->dad_timer)) 268 in6_ifa_hold(ifp); 269 mod_timer(&ifp->dad_timer, jiffies + when); 270 } 271 272 static int snmp6_alloc_dev(struct inet6_dev *idev) 273 { 274 int i; 275 276 if (snmp_mib_init((void __percpu **)idev->stats.ipv6, 277 sizeof(struct ipstats_mib), 278 __alignof__(struct ipstats_mib)) < 0) 279 goto err_ip; 280 281 for_each_possible_cpu(i) { 282 struct ipstats_mib *addrconf_stats; 283 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[0], i); 284 u64_stats_init(&addrconf_stats->syncp); 285 #if SNMP_ARRAY_SZ == 2 286 addrconf_stats = per_cpu_ptr(idev->stats.ipv6[1], i); 287 u64_stats_init(&addrconf_stats->syncp); 288 #endif 289 } 290 291 292 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device), 293 GFP_KERNEL); 294 if (!idev->stats.icmpv6dev) 295 goto err_icmp; 296 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device), 297 GFP_KERNEL); 298 if (!idev->stats.icmpv6msgdev) 299 goto err_icmpmsg; 300 301 return 0; 302 303 err_icmpmsg: 304 kfree(idev->stats.icmpv6dev); 305 err_icmp: 306 snmp_mib_free((void __percpu **)idev->stats.ipv6); 307 err_ip: 308 return -ENOMEM; 309 } 310 311 static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 312 { 313 struct inet6_dev *ndev; 314 315 ASSERT_RTNL(); 316 317 if (dev->mtu < IPV6_MIN_MTU) 318 return NULL; 319 320 ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL); 321 322 if (ndev == NULL) 323 return NULL; 324 325 rwlock_init(&ndev->lock); 326 ndev->dev = dev; 327 INIT_LIST_HEAD(&ndev->addr_list); 328 setup_timer(&ndev->rs_timer, addrconf_rs_timer, 329 (unsigned long)ndev); 330 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf)); 331 ndev->cnf.mtu6 = dev->mtu; 332 ndev->cnf.sysctl = NULL; 333 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl); 334 if (ndev->nd_parms == NULL) { 335 kfree(ndev); 336 return NULL; 337 } 338 if (ndev->cnf.forwarding) 339 dev_disable_lro(dev); 340 /* We refer to the device */ 341 dev_hold(dev); 342 343 if (snmp6_alloc_dev(ndev) < 0) { 344 ADBG(KERN_WARNING 345 "%s: cannot allocate memory for statistics; dev=%s.\n", 346 __func__, dev->name); 347 neigh_parms_release(&nd_tbl, ndev->nd_parms); 348 dev_put(dev); 349 kfree(ndev); 350 return NULL; 351 } 352 353 if (snmp6_register_dev(ndev) < 0) { 354 ADBG(KERN_WARNING 355 "%s: cannot create /proc/net/dev_snmp6/%s\n", 356 __func__, dev->name); 357 neigh_parms_release(&nd_tbl, ndev->nd_parms); 358 ndev->dead = 1; 359 in6_dev_finish_destroy(ndev); 360 return NULL; 361 } 362 363 /* One reference from device. We must do this before 364 * we invoke __ipv6_regen_rndid(). 365 */ 366 in6_dev_hold(ndev); 367 368 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK)) 369 ndev->cnf.accept_dad = -1; 370 371 #if IS_ENABLED(CONFIG_IPV6_SIT) 372 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) { 373 pr_info("%s: Disabled Multicast RS\n", dev->name); 374 ndev->cnf.rtr_solicits = 0; 375 } 376 #endif 377 378 INIT_LIST_HEAD(&ndev->tempaddr_list); 379 setup_timer(&ndev->regen_timer, ipv6_regen_rndid, (unsigned long)ndev); 380 if ((dev->flags&IFF_LOOPBACK) || 381 dev->type == ARPHRD_TUNNEL || 382 dev->type == ARPHRD_TUNNEL6 || 383 dev->type == ARPHRD_SIT || 384 dev->type == ARPHRD_NONE) { 385 ndev->cnf.use_tempaddr = -1; 386 } else { 387 in6_dev_hold(ndev); 388 ipv6_regen_rndid((unsigned long) ndev); 389 } 390 391 ndev->token = in6addr_any; 392 393 if (netif_running(dev) && addrconf_qdisc_ok(dev)) 394 ndev->if_flags |= IF_READY; 395 396 ipv6_mc_init_dev(ndev); 397 ndev->tstamp = jiffies; 398 addrconf_sysctl_register(ndev); 399 /* protected by rtnl_lock */ 400 rcu_assign_pointer(dev->ip6_ptr, ndev); 401 402 /* Join interface-local all-node multicast group */ 403 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes); 404 405 /* Join all-node multicast group */ 406 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes); 407 408 /* Join all-router multicast group if forwarding is set */ 409 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST)) 410 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 411 412 return ndev; 413 } 414 415 static struct inet6_dev *ipv6_find_idev(struct net_device *dev) 416 { 417 struct inet6_dev *idev; 418 419 ASSERT_RTNL(); 420 421 idev = __in6_dev_get(dev); 422 if (!idev) { 423 idev = ipv6_add_dev(dev); 424 if (!idev) 425 return NULL; 426 } 427 428 if (dev->flags&IFF_UP) 429 ipv6_mc_up(idev); 430 return idev; 431 } 432 433 static int inet6_netconf_msgsize_devconf(int type) 434 { 435 int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) 436 + nla_total_size(4); /* NETCONFA_IFINDEX */ 437 438 /* type -1 is used for ALL */ 439 if (type == -1 || type == NETCONFA_FORWARDING) 440 size += nla_total_size(4); 441 #ifdef CONFIG_IPV6_MROUTE 442 if (type == -1 || type == NETCONFA_MC_FORWARDING) 443 size += nla_total_size(4); 444 #endif 445 if (type == -1 || type == NETCONFA_PROXY_NEIGH) 446 size += nla_total_size(4); 447 448 return size; 449 } 450 451 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex, 452 struct ipv6_devconf *devconf, u32 portid, 453 u32 seq, int event, unsigned int flags, 454 int type) 455 { 456 struct nlmsghdr *nlh; 457 struct netconfmsg *ncm; 458 459 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), 460 flags); 461 if (nlh == NULL) 462 return -EMSGSIZE; 463 464 ncm = nlmsg_data(nlh); 465 ncm->ncm_family = AF_INET6; 466 467 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0) 468 goto nla_put_failure; 469 470 /* type -1 is used for ALL */ 471 if ((type == -1 || type == NETCONFA_FORWARDING) && 472 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0) 473 goto nla_put_failure; 474 #ifdef CONFIG_IPV6_MROUTE 475 if ((type == -1 || type == NETCONFA_MC_FORWARDING) && 476 nla_put_s32(skb, NETCONFA_MC_FORWARDING, 477 devconf->mc_forwarding) < 0) 478 goto nla_put_failure; 479 #endif 480 if ((type == -1 || type == NETCONFA_PROXY_NEIGH) && 481 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0) 482 goto nla_put_failure; 483 484 return nlmsg_end(skb, nlh); 485 486 nla_put_failure: 487 nlmsg_cancel(skb, nlh); 488 return -EMSGSIZE; 489 } 490 491 void inet6_netconf_notify_devconf(struct net *net, int type, int ifindex, 492 struct ipv6_devconf *devconf) 493 { 494 struct sk_buff *skb; 495 int err = -ENOBUFS; 496 497 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_ATOMIC); 498 if (skb == NULL) 499 goto errout; 500 501 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0, 502 RTM_NEWNETCONF, 0, type); 503 if (err < 0) { 504 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 505 WARN_ON(err == -EMSGSIZE); 506 kfree_skb(skb); 507 goto errout; 508 } 509 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_ATOMIC); 510 return; 511 errout: 512 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err); 513 } 514 515 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = { 516 [NETCONFA_IFINDEX] = { .len = sizeof(int) }, 517 [NETCONFA_FORWARDING] = { .len = sizeof(int) }, 518 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) }, 519 }; 520 521 static int inet6_netconf_get_devconf(struct sk_buff *in_skb, 522 struct nlmsghdr *nlh) 523 { 524 struct net *net = sock_net(in_skb->sk); 525 struct nlattr *tb[NETCONFA_MAX+1]; 526 struct netconfmsg *ncm; 527 struct sk_buff *skb; 528 struct ipv6_devconf *devconf; 529 struct inet6_dev *in6_dev; 530 struct net_device *dev; 531 int ifindex; 532 int err; 533 534 err = nlmsg_parse(nlh, sizeof(*ncm), tb, NETCONFA_MAX, 535 devconf_ipv6_policy); 536 if (err < 0) 537 goto errout; 538 539 err = EINVAL; 540 if (!tb[NETCONFA_IFINDEX]) 541 goto errout; 542 543 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); 544 switch (ifindex) { 545 case NETCONFA_IFINDEX_ALL: 546 devconf = net->ipv6.devconf_all; 547 break; 548 case NETCONFA_IFINDEX_DEFAULT: 549 devconf = net->ipv6.devconf_dflt; 550 break; 551 default: 552 dev = __dev_get_by_index(net, ifindex); 553 if (dev == NULL) 554 goto errout; 555 in6_dev = __in6_dev_get(dev); 556 if (in6_dev == NULL) 557 goto errout; 558 devconf = &in6_dev->cnf; 559 break; 560 } 561 562 err = -ENOBUFS; 563 skb = nlmsg_new(inet6_netconf_msgsize_devconf(-1), GFP_ATOMIC); 564 if (skb == NULL) 565 goto errout; 566 567 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 568 NETLINK_CB(in_skb).portid, 569 nlh->nlmsg_seq, RTM_NEWNETCONF, 0, 570 -1); 571 if (err < 0) { 572 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 573 WARN_ON(err == -EMSGSIZE); 574 kfree_skb(skb); 575 goto errout; 576 } 577 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 578 errout: 579 return err; 580 } 581 582 static int inet6_netconf_dump_devconf(struct sk_buff *skb, 583 struct netlink_callback *cb) 584 { 585 struct net *net = sock_net(skb->sk); 586 int h, s_h; 587 int idx, s_idx; 588 struct net_device *dev; 589 struct inet6_dev *idev; 590 struct hlist_head *head; 591 592 s_h = cb->args[0]; 593 s_idx = idx = cb->args[1]; 594 595 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 596 idx = 0; 597 head = &net->dev_index_head[h]; 598 rcu_read_lock(); 599 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ 600 net->dev_base_seq; 601 hlist_for_each_entry_rcu(dev, head, index_hlist) { 602 if (idx < s_idx) 603 goto cont; 604 idev = __in6_dev_get(dev); 605 if (!idev) 606 goto cont; 607 608 if (inet6_netconf_fill_devconf(skb, dev->ifindex, 609 &idev->cnf, 610 NETLINK_CB(cb->skb).portid, 611 cb->nlh->nlmsg_seq, 612 RTM_NEWNETCONF, 613 NLM_F_MULTI, 614 -1) <= 0) { 615 rcu_read_unlock(); 616 goto done; 617 } 618 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 619 cont: 620 idx++; 621 } 622 rcu_read_unlock(); 623 } 624 if (h == NETDEV_HASHENTRIES) { 625 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL, 626 net->ipv6.devconf_all, 627 NETLINK_CB(cb->skb).portid, 628 cb->nlh->nlmsg_seq, 629 RTM_NEWNETCONF, NLM_F_MULTI, 630 -1) <= 0) 631 goto done; 632 else 633 h++; 634 } 635 if (h == NETDEV_HASHENTRIES + 1) { 636 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT, 637 net->ipv6.devconf_dflt, 638 NETLINK_CB(cb->skb).portid, 639 cb->nlh->nlmsg_seq, 640 RTM_NEWNETCONF, NLM_F_MULTI, 641 -1) <= 0) 642 goto done; 643 else 644 h++; 645 } 646 done: 647 cb->args[0] = h; 648 cb->args[1] = idx; 649 650 return skb->len; 651 } 652 653 #ifdef CONFIG_SYSCTL 654 static void dev_forward_change(struct inet6_dev *idev) 655 { 656 struct net_device *dev; 657 struct inet6_ifaddr *ifa; 658 659 if (!idev) 660 return; 661 dev = idev->dev; 662 if (idev->cnf.forwarding) 663 dev_disable_lro(dev); 664 if (dev->flags & IFF_MULTICAST) { 665 if (idev->cnf.forwarding) { 666 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 667 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters); 668 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters); 669 } else { 670 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters); 671 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters); 672 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters); 673 } 674 } 675 676 list_for_each_entry(ifa, &idev->addr_list, if_list) { 677 if (ifa->flags&IFA_F_TENTATIVE) 678 continue; 679 if (idev->cnf.forwarding) 680 addrconf_join_anycast(ifa); 681 else 682 addrconf_leave_anycast(ifa); 683 } 684 inet6_netconf_notify_devconf(dev_net(dev), NETCONFA_FORWARDING, 685 dev->ifindex, &idev->cnf); 686 } 687 688 689 static void addrconf_forward_change(struct net *net, __s32 newf) 690 { 691 struct net_device *dev; 692 struct inet6_dev *idev; 693 694 for_each_netdev(net, dev) { 695 idev = __in6_dev_get(dev); 696 if (idev) { 697 int changed = (!idev->cnf.forwarding) ^ (!newf); 698 idev->cnf.forwarding = newf; 699 if (changed) 700 dev_forward_change(idev); 701 } 702 } 703 } 704 705 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf) 706 { 707 struct net *net; 708 int old; 709 710 if (!rtnl_trylock()) 711 return restart_syscall(); 712 713 net = (struct net *)table->extra2; 714 old = *p; 715 *p = newf; 716 717 if (p == &net->ipv6.devconf_dflt->forwarding) { 718 if ((!newf) ^ (!old)) 719 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING, 720 NETCONFA_IFINDEX_DEFAULT, 721 net->ipv6.devconf_dflt); 722 rtnl_unlock(); 723 return 0; 724 } 725 726 if (p == &net->ipv6.devconf_all->forwarding) { 727 net->ipv6.devconf_dflt->forwarding = newf; 728 addrconf_forward_change(net, newf); 729 if ((!newf) ^ (!old)) 730 inet6_netconf_notify_devconf(net, NETCONFA_FORWARDING, 731 NETCONFA_IFINDEX_ALL, 732 net->ipv6.devconf_all); 733 } else if ((!newf) ^ (!old)) 734 dev_forward_change((struct inet6_dev *)table->extra1); 735 rtnl_unlock(); 736 737 if (newf) 738 rt6_purge_dflt_routers(net); 739 return 1; 740 } 741 #endif 742 743 /* Nobody refers to this ifaddr, destroy it */ 744 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp) 745 { 746 WARN_ON(!hlist_unhashed(&ifp->addr_lst)); 747 748 #ifdef NET_REFCNT_DEBUG 749 pr_debug("%s\n", __func__); 750 #endif 751 752 in6_dev_put(ifp->idev); 753 754 if (del_timer(&ifp->dad_timer)) 755 pr_notice("Timer is still running, when freeing ifa=%p\n", ifp); 756 757 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 758 pr_warn("Freeing alive inet6 address %p\n", ifp); 759 return; 760 } 761 ip6_rt_put(ifp->rt); 762 763 kfree_rcu(ifp, rcu); 764 } 765 766 static void 767 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp) 768 { 769 struct list_head *p; 770 int ifp_scope = ipv6_addr_src_scope(&ifp->addr); 771 772 /* 773 * Each device address list is sorted in order of scope - 774 * global before linklocal. 775 */ 776 list_for_each(p, &idev->addr_list) { 777 struct inet6_ifaddr *ifa 778 = list_entry(p, struct inet6_ifaddr, if_list); 779 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr)) 780 break; 781 } 782 783 list_add_tail(&ifp->if_list, p); 784 } 785 786 static u32 inet6_addr_hash(const struct in6_addr *addr) 787 { 788 return hash_32(ipv6_addr_hash(addr), IN6_ADDR_HSIZE_SHIFT); 789 } 790 791 /* On success it returns ifp with increased reference count */ 792 793 static struct inet6_ifaddr * 794 ipv6_add_addr(struct inet6_dev *idev, const struct in6_addr *addr, 795 const struct in6_addr *peer_addr, int pfxlen, 796 int scope, u32 flags, u32 valid_lft, u32 prefered_lft) 797 { 798 struct inet6_ifaddr *ifa = NULL; 799 struct rt6_info *rt; 800 unsigned int hash; 801 int err = 0; 802 int addr_type = ipv6_addr_type(addr); 803 804 if (addr_type == IPV6_ADDR_ANY || 805 addr_type & IPV6_ADDR_MULTICAST || 806 (!(idev->dev->flags & IFF_LOOPBACK) && 807 addr_type & IPV6_ADDR_LOOPBACK)) 808 return ERR_PTR(-EADDRNOTAVAIL); 809 810 rcu_read_lock_bh(); 811 if (idev->dead) { 812 err = -ENODEV; /*XXX*/ 813 goto out2; 814 } 815 816 if (idev->cnf.disable_ipv6) { 817 err = -EACCES; 818 goto out2; 819 } 820 821 spin_lock(&addrconf_hash_lock); 822 823 /* Ignore adding duplicate addresses on an interface */ 824 if (ipv6_chk_same_addr(dev_net(idev->dev), addr, idev->dev)) { 825 ADBG("ipv6_add_addr: already assigned\n"); 826 err = -EEXIST; 827 goto out; 828 } 829 830 ifa = kzalloc(sizeof(struct inet6_ifaddr), GFP_ATOMIC); 831 832 if (ifa == NULL) { 833 ADBG("ipv6_add_addr: malloc failed\n"); 834 err = -ENOBUFS; 835 goto out; 836 } 837 838 rt = addrconf_dst_alloc(idev, addr, false); 839 if (IS_ERR(rt)) { 840 err = PTR_ERR(rt); 841 goto out; 842 } 843 844 neigh_parms_data_state_setall(idev->nd_parms); 845 846 ifa->addr = *addr; 847 if (peer_addr) 848 ifa->peer_addr = *peer_addr; 849 850 spin_lock_init(&ifa->lock); 851 spin_lock_init(&ifa->state_lock); 852 setup_timer(&ifa->dad_timer, addrconf_dad_timer, 853 (unsigned long)ifa); 854 INIT_HLIST_NODE(&ifa->addr_lst); 855 ifa->scope = scope; 856 ifa->prefix_len = pfxlen; 857 ifa->flags = flags | IFA_F_TENTATIVE; 858 ifa->valid_lft = valid_lft; 859 ifa->prefered_lft = prefered_lft; 860 ifa->cstamp = ifa->tstamp = jiffies; 861 ifa->tokenized = false; 862 863 ifa->rt = rt; 864 865 ifa->idev = idev; 866 in6_dev_hold(idev); 867 /* For caller */ 868 in6_ifa_hold(ifa); 869 870 /* Add to big hash table */ 871 hash = inet6_addr_hash(addr); 872 873 hlist_add_head_rcu(&ifa->addr_lst, &inet6_addr_lst[hash]); 874 spin_unlock(&addrconf_hash_lock); 875 876 write_lock(&idev->lock); 877 /* Add to inet6_dev unicast addr list. */ 878 ipv6_link_dev_addr(idev, ifa); 879 880 if (ifa->flags&IFA_F_TEMPORARY) { 881 list_add(&ifa->tmp_list, &idev->tempaddr_list); 882 in6_ifa_hold(ifa); 883 } 884 885 in6_ifa_hold(ifa); 886 write_unlock(&idev->lock); 887 out2: 888 rcu_read_unlock_bh(); 889 890 if (likely(err == 0)) 891 inet6addr_notifier_call_chain(NETDEV_UP, ifa); 892 else { 893 kfree(ifa); 894 ifa = ERR_PTR(err); 895 } 896 897 return ifa; 898 out: 899 spin_unlock(&addrconf_hash_lock); 900 goto out2; 901 } 902 903 /* This function wants to get referenced ifp and releases it before return */ 904 905 static void ipv6_del_addr(struct inet6_ifaddr *ifp) 906 { 907 struct inet6_ifaddr *ifa, *ifn; 908 struct inet6_dev *idev = ifp->idev; 909 int state; 910 int deleted = 0, onlink = 0; 911 unsigned long expires = jiffies; 912 913 spin_lock_bh(&ifp->state_lock); 914 state = ifp->state; 915 ifp->state = INET6_IFADDR_STATE_DEAD; 916 spin_unlock_bh(&ifp->state_lock); 917 918 if (state == INET6_IFADDR_STATE_DEAD) 919 goto out; 920 921 spin_lock_bh(&addrconf_hash_lock); 922 hlist_del_init_rcu(&ifp->addr_lst); 923 spin_unlock_bh(&addrconf_hash_lock); 924 925 write_lock_bh(&idev->lock); 926 927 if (ifp->flags&IFA_F_TEMPORARY) { 928 list_del(&ifp->tmp_list); 929 if (ifp->ifpub) { 930 in6_ifa_put(ifp->ifpub); 931 ifp->ifpub = NULL; 932 } 933 __in6_ifa_put(ifp); 934 } 935 936 list_for_each_entry_safe(ifa, ifn, &idev->addr_list, if_list) { 937 if (ifa == ifp) { 938 list_del_init(&ifp->if_list); 939 __in6_ifa_put(ifp); 940 941 if (!(ifp->flags & IFA_F_PERMANENT) || onlink > 0) 942 break; 943 deleted = 1; 944 continue; 945 } else if (ifp->flags & IFA_F_PERMANENT) { 946 if (ipv6_prefix_equal(&ifa->addr, &ifp->addr, 947 ifp->prefix_len)) { 948 if (ifa->flags & IFA_F_PERMANENT) { 949 onlink = 1; 950 if (deleted) 951 break; 952 } else { 953 unsigned long lifetime; 954 955 if (!onlink) 956 onlink = -1; 957 958 spin_lock(&ifa->lock); 959 960 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ); 961 /* 962 * Note: Because this address is 963 * not permanent, lifetime < 964 * LONG_MAX / HZ here. 965 */ 966 if (time_before(expires, 967 ifa->tstamp + lifetime * HZ)) 968 expires = ifa->tstamp + lifetime * HZ; 969 spin_unlock(&ifa->lock); 970 } 971 } 972 } 973 } 974 write_unlock_bh(&idev->lock); 975 976 addrconf_del_dad_timer(ifp); 977 978 ipv6_ifa_notify(RTM_DELADDR, ifp); 979 980 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp); 981 982 /* 983 * Purge or update corresponding prefix 984 * 985 * 1) we don't purge prefix here if address was not permanent. 986 * prefix is managed by its own lifetime. 987 * 2) if there're no addresses, delete prefix. 988 * 3) if there're still other permanent address(es), 989 * corresponding prefix is still permanent. 990 * 4) otherwise, update prefix lifetime to the 991 * longest valid lifetime among the corresponding 992 * addresses on the device. 993 * Note: subsequent RA will update lifetime. 994 * 995 * --yoshfuji 996 */ 997 if ((ifp->flags & IFA_F_PERMANENT) && onlink < 1) { 998 struct rt6_info *rt; 999 1000 rt = addrconf_get_prefix_route(&ifp->addr, 1001 ifp->prefix_len, 1002 ifp->idev->dev, 1003 0, RTF_GATEWAY | RTF_DEFAULT); 1004 1005 if (rt) { 1006 if (onlink == 0) { 1007 ip6_del_rt(rt); 1008 rt = NULL; 1009 } else if (!(rt->rt6i_flags & RTF_EXPIRES)) { 1010 rt6_set_expires(rt, expires); 1011 } 1012 } 1013 ip6_rt_put(rt); 1014 } 1015 1016 /* clean up prefsrc entries */ 1017 rt6_remove_prefsrc(ifp); 1018 out: 1019 in6_ifa_put(ifp); 1020 } 1021 1022 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, struct inet6_ifaddr *ift) 1023 { 1024 struct inet6_dev *idev = ifp->idev; 1025 struct in6_addr addr, *tmpaddr; 1026 unsigned long tmp_prefered_lft, tmp_valid_lft, tmp_tstamp, age; 1027 unsigned long regen_advance; 1028 int tmp_plen; 1029 int ret = 0; 1030 u32 addr_flags; 1031 unsigned long now = jiffies; 1032 1033 write_lock_bh(&idev->lock); 1034 if (ift) { 1035 spin_lock_bh(&ift->lock); 1036 memcpy(&addr.s6_addr[8], &ift->addr.s6_addr[8], 8); 1037 spin_unlock_bh(&ift->lock); 1038 tmpaddr = &addr; 1039 } else { 1040 tmpaddr = NULL; 1041 } 1042 retry: 1043 in6_dev_hold(idev); 1044 if (idev->cnf.use_tempaddr <= 0) { 1045 write_unlock_bh(&idev->lock); 1046 pr_info("%s: use_tempaddr is disabled\n", __func__); 1047 in6_dev_put(idev); 1048 ret = -1; 1049 goto out; 1050 } 1051 spin_lock_bh(&ifp->lock); 1052 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) { 1053 idev->cnf.use_tempaddr = -1; /*XXX*/ 1054 spin_unlock_bh(&ifp->lock); 1055 write_unlock_bh(&idev->lock); 1056 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n", 1057 __func__); 1058 in6_dev_put(idev); 1059 ret = -1; 1060 goto out; 1061 } 1062 in6_ifa_hold(ifp); 1063 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8); 1064 __ipv6_try_regen_rndid(idev, tmpaddr); 1065 memcpy(&addr.s6_addr[8], idev->rndid, 8); 1066 age = (now - ifp->tstamp) / HZ; 1067 tmp_valid_lft = min_t(__u32, 1068 ifp->valid_lft, 1069 idev->cnf.temp_valid_lft + age); 1070 tmp_prefered_lft = min_t(__u32, 1071 ifp->prefered_lft, 1072 idev->cnf.temp_prefered_lft + age - 1073 idev->cnf.max_desync_factor); 1074 tmp_plen = ifp->prefix_len; 1075 tmp_tstamp = ifp->tstamp; 1076 spin_unlock_bh(&ifp->lock); 1077 1078 regen_advance = idev->cnf.regen_max_retry * 1079 idev->cnf.dad_transmits * 1080 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) / HZ; 1081 write_unlock_bh(&idev->lock); 1082 1083 /* A temporary address is created only if this calculated Preferred 1084 * Lifetime is greater than REGEN_ADVANCE time units. In particular, 1085 * an implementation must not create a temporary address with a zero 1086 * Preferred Lifetime. 1087 */ 1088 if (tmp_prefered_lft <= regen_advance) { 1089 in6_ifa_put(ifp); 1090 in6_dev_put(idev); 1091 ret = -1; 1092 goto out; 1093 } 1094 1095 addr_flags = IFA_F_TEMPORARY; 1096 /* set in addrconf_prefix_rcv() */ 1097 if (ifp->flags & IFA_F_OPTIMISTIC) 1098 addr_flags |= IFA_F_OPTIMISTIC; 1099 1100 ift = ipv6_add_addr(idev, &addr, NULL, tmp_plen, 1101 ipv6_addr_scope(&addr), addr_flags, 1102 tmp_valid_lft, tmp_prefered_lft); 1103 if (IS_ERR(ift)) { 1104 in6_ifa_put(ifp); 1105 in6_dev_put(idev); 1106 pr_info("%s: retry temporary address regeneration\n", __func__); 1107 tmpaddr = &addr; 1108 write_lock_bh(&idev->lock); 1109 goto retry; 1110 } 1111 1112 spin_lock_bh(&ift->lock); 1113 ift->ifpub = ifp; 1114 ift->cstamp = now; 1115 ift->tstamp = tmp_tstamp; 1116 spin_unlock_bh(&ift->lock); 1117 1118 addrconf_dad_start(ift); 1119 in6_ifa_put(ift); 1120 in6_dev_put(idev); 1121 out: 1122 return ret; 1123 } 1124 1125 /* 1126 * Choose an appropriate source address (RFC3484) 1127 */ 1128 enum { 1129 IPV6_SADDR_RULE_INIT = 0, 1130 IPV6_SADDR_RULE_LOCAL, 1131 IPV6_SADDR_RULE_SCOPE, 1132 IPV6_SADDR_RULE_PREFERRED, 1133 #ifdef CONFIG_IPV6_MIP6 1134 IPV6_SADDR_RULE_HOA, 1135 #endif 1136 IPV6_SADDR_RULE_OIF, 1137 IPV6_SADDR_RULE_LABEL, 1138 IPV6_SADDR_RULE_PRIVACY, 1139 IPV6_SADDR_RULE_ORCHID, 1140 IPV6_SADDR_RULE_PREFIX, 1141 IPV6_SADDR_RULE_MAX 1142 }; 1143 1144 struct ipv6_saddr_score { 1145 int rule; 1146 int addr_type; 1147 struct inet6_ifaddr *ifa; 1148 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX); 1149 int scopedist; 1150 int matchlen; 1151 }; 1152 1153 struct ipv6_saddr_dst { 1154 const struct in6_addr *addr; 1155 int ifindex; 1156 int scope; 1157 int label; 1158 unsigned int prefs; 1159 }; 1160 1161 static inline int ipv6_saddr_preferred(int type) 1162 { 1163 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK)) 1164 return 1; 1165 return 0; 1166 } 1167 1168 static int ipv6_get_saddr_eval(struct net *net, 1169 struct ipv6_saddr_score *score, 1170 struct ipv6_saddr_dst *dst, 1171 int i) 1172 { 1173 int ret; 1174 1175 if (i <= score->rule) { 1176 switch (i) { 1177 case IPV6_SADDR_RULE_SCOPE: 1178 ret = score->scopedist; 1179 break; 1180 case IPV6_SADDR_RULE_PREFIX: 1181 ret = score->matchlen; 1182 break; 1183 default: 1184 ret = !!test_bit(i, score->scorebits); 1185 } 1186 goto out; 1187 } 1188 1189 switch (i) { 1190 case IPV6_SADDR_RULE_INIT: 1191 /* Rule 0: remember if hiscore is not ready yet */ 1192 ret = !!score->ifa; 1193 break; 1194 case IPV6_SADDR_RULE_LOCAL: 1195 /* Rule 1: Prefer same address */ 1196 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr); 1197 break; 1198 case IPV6_SADDR_RULE_SCOPE: 1199 /* Rule 2: Prefer appropriate scope 1200 * 1201 * ret 1202 * ^ 1203 * -1 | d 15 1204 * ---+--+-+---> scope 1205 * | 1206 * | d is scope of the destination. 1207 * B-d | \ 1208 * | \ <- smaller scope is better if 1209 * B-15 | \ if scope is enough for destinaion. 1210 * | ret = B - scope (-1 <= scope >= d <= 15). 1211 * d-C-1 | / 1212 * |/ <- greater is better 1213 * -C / if scope is not enough for destination. 1214 * /| ret = scope - C (-1 <= d < scope <= 15). 1215 * 1216 * d - C - 1 < B -15 (for all -1 <= d <= 15). 1217 * C > d + 14 - B >= 15 + 14 - B = 29 - B. 1218 * Assume B = 0 and we get C > 29. 1219 */ 1220 ret = __ipv6_addr_src_scope(score->addr_type); 1221 if (ret >= dst->scope) 1222 ret = -ret; 1223 else 1224 ret -= 128; /* 30 is enough */ 1225 score->scopedist = ret; 1226 break; 1227 case IPV6_SADDR_RULE_PREFERRED: 1228 /* Rule 3: Avoid deprecated and optimistic addresses */ 1229 ret = ipv6_saddr_preferred(score->addr_type) || 1230 !(score->ifa->flags & (IFA_F_DEPRECATED|IFA_F_OPTIMISTIC)); 1231 break; 1232 #ifdef CONFIG_IPV6_MIP6 1233 case IPV6_SADDR_RULE_HOA: 1234 { 1235 /* Rule 4: Prefer home address */ 1236 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA); 1237 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome; 1238 break; 1239 } 1240 #endif 1241 case IPV6_SADDR_RULE_OIF: 1242 /* Rule 5: Prefer outgoing interface */ 1243 ret = (!dst->ifindex || 1244 dst->ifindex == score->ifa->idev->dev->ifindex); 1245 break; 1246 case IPV6_SADDR_RULE_LABEL: 1247 /* Rule 6: Prefer matching label */ 1248 ret = ipv6_addr_label(net, 1249 &score->ifa->addr, score->addr_type, 1250 score->ifa->idev->dev->ifindex) == dst->label; 1251 break; 1252 case IPV6_SADDR_RULE_PRIVACY: 1253 { 1254 /* Rule 7: Prefer public address 1255 * Note: prefer temporary address if use_tempaddr >= 2 1256 */ 1257 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ? 1258 !!(dst->prefs & IPV6_PREFER_SRC_TMP) : 1259 score->ifa->idev->cnf.use_tempaddr >= 2; 1260 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp; 1261 break; 1262 } 1263 case IPV6_SADDR_RULE_ORCHID: 1264 /* Rule 8-: Prefer ORCHID vs ORCHID or 1265 * non-ORCHID vs non-ORCHID 1266 */ 1267 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^ 1268 ipv6_addr_orchid(dst->addr)); 1269 break; 1270 case IPV6_SADDR_RULE_PREFIX: 1271 /* Rule 8: Use longest matching prefix */ 1272 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr); 1273 if (ret > score->ifa->prefix_len) 1274 ret = score->ifa->prefix_len; 1275 score->matchlen = ret; 1276 break; 1277 default: 1278 ret = 0; 1279 } 1280 1281 if (ret) 1282 __set_bit(i, score->scorebits); 1283 score->rule = i; 1284 out: 1285 return ret; 1286 } 1287 1288 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev, 1289 const struct in6_addr *daddr, unsigned int prefs, 1290 struct in6_addr *saddr) 1291 { 1292 struct ipv6_saddr_score scores[2], 1293 *score = &scores[0], *hiscore = &scores[1]; 1294 struct ipv6_saddr_dst dst; 1295 struct net_device *dev; 1296 int dst_type; 1297 1298 dst_type = __ipv6_addr_type(daddr); 1299 dst.addr = daddr; 1300 dst.ifindex = dst_dev ? dst_dev->ifindex : 0; 1301 dst.scope = __ipv6_addr_src_scope(dst_type); 1302 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex); 1303 dst.prefs = prefs; 1304 1305 hiscore->rule = -1; 1306 hiscore->ifa = NULL; 1307 1308 rcu_read_lock(); 1309 1310 for_each_netdev_rcu(net, dev) { 1311 struct inet6_dev *idev; 1312 1313 /* Candidate Source Address (section 4) 1314 * - multicast and link-local destination address, 1315 * the set of candidate source address MUST only 1316 * include addresses assigned to interfaces 1317 * belonging to the same link as the outgoing 1318 * interface. 1319 * (- For site-local destination addresses, the 1320 * set of candidate source addresses MUST only 1321 * include addresses assigned to interfaces 1322 * belonging to the same site as the outgoing 1323 * interface.) 1324 */ 1325 if (((dst_type & IPV6_ADDR_MULTICAST) || 1326 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL) && 1327 dst.ifindex && dev->ifindex != dst.ifindex) 1328 continue; 1329 1330 idev = __in6_dev_get(dev); 1331 if (!idev) 1332 continue; 1333 1334 read_lock_bh(&idev->lock); 1335 list_for_each_entry(score->ifa, &idev->addr_list, if_list) { 1336 int i; 1337 1338 /* 1339 * - Tentative Address (RFC2462 section 5.4) 1340 * - A tentative address is not considered 1341 * "assigned to an interface" in the traditional 1342 * sense, unless it is also flagged as optimistic. 1343 * - Candidate Source Address (section 4) 1344 * - In any case, anycast addresses, multicast 1345 * addresses, and the unspecified address MUST 1346 * NOT be included in a candidate set. 1347 */ 1348 if ((score->ifa->flags & IFA_F_TENTATIVE) && 1349 (!(score->ifa->flags & IFA_F_OPTIMISTIC))) 1350 continue; 1351 1352 score->addr_type = __ipv6_addr_type(&score->ifa->addr); 1353 1354 if (unlikely(score->addr_type == IPV6_ADDR_ANY || 1355 score->addr_type & IPV6_ADDR_MULTICAST)) { 1356 LIMIT_NETDEBUG(KERN_DEBUG 1357 "ADDRCONF: unspecified / multicast address " 1358 "assigned as unicast address on %s", 1359 dev->name); 1360 continue; 1361 } 1362 1363 score->rule = -1; 1364 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX); 1365 1366 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) { 1367 int minihiscore, miniscore; 1368 1369 minihiscore = ipv6_get_saddr_eval(net, hiscore, &dst, i); 1370 miniscore = ipv6_get_saddr_eval(net, score, &dst, i); 1371 1372 if (minihiscore > miniscore) { 1373 if (i == IPV6_SADDR_RULE_SCOPE && 1374 score->scopedist > 0) { 1375 /* 1376 * special case: 1377 * each remaining entry 1378 * has too small (not enough) 1379 * scope, because ifa entries 1380 * are sorted by their scope 1381 * values. 1382 */ 1383 goto try_nextdev; 1384 } 1385 break; 1386 } else if (minihiscore < miniscore) { 1387 if (hiscore->ifa) 1388 in6_ifa_put(hiscore->ifa); 1389 1390 in6_ifa_hold(score->ifa); 1391 1392 swap(hiscore, score); 1393 1394 /* restore our iterator */ 1395 score->ifa = hiscore->ifa; 1396 1397 break; 1398 } 1399 } 1400 } 1401 try_nextdev: 1402 read_unlock_bh(&idev->lock); 1403 } 1404 rcu_read_unlock(); 1405 1406 if (!hiscore->ifa) 1407 return -EADDRNOTAVAIL; 1408 1409 *saddr = hiscore->ifa->addr; 1410 in6_ifa_put(hiscore->ifa); 1411 return 0; 1412 } 1413 EXPORT_SYMBOL(ipv6_dev_get_saddr); 1414 1415 int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr, 1416 u32 banned_flags) 1417 { 1418 struct inet6_ifaddr *ifp; 1419 int err = -EADDRNOTAVAIL; 1420 1421 list_for_each_entry(ifp, &idev->addr_list, if_list) { 1422 if (ifp->scope == IFA_LINK && 1423 !(ifp->flags & banned_flags)) { 1424 *addr = ifp->addr; 1425 err = 0; 1426 break; 1427 } 1428 } 1429 return err; 1430 } 1431 1432 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr, 1433 u32 banned_flags) 1434 { 1435 struct inet6_dev *idev; 1436 int err = -EADDRNOTAVAIL; 1437 1438 rcu_read_lock(); 1439 idev = __in6_dev_get(dev); 1440 if (idev) { 1441 read_lock_bh(&idev->lock); 1442 err = __ipv6_get_lladdr(idev, addr, banned_flags); 1443 read_unlock_bh(&idev->lock); 1444 } 1445 rcu_read_unlock(); 1446 return err; 1447 } 1448 1449 static int ipv6_count_addresses(struct inet6_dev *idev) 1450 { 1451 int cnt = 0; 1452 struct inet6_ifaddr *ifp; 1453 1454 read_lock_bh(&idev->lock); 1455 list_for_each_entry(ifp, &idev->addr_list, if_list) 1456 cnt++; 1457 read_unlock_bh(&idev->lock); 1458 return cnt; 1459 } 1460 1461 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr, 1462 const struct net_device *dev, int strict) 1463 { 1464 struct inet6_ifaddr *ifp; 1465 unsigned int hash = inet6_addr_hash(addr); 1466 1467 rcu_read_lock_bh(); 1468 hlist_for_each_entry_rcu(ifp, &inet6_addr_lst[hash], addr_lst) { 1469 if (!net_eq(dev_net(ifp->idev->dev), net)) 1470 continue; 1471 if (ipv6_addr_equal(&ifp->addr, addr) && 1472 !(ifp->flags&IFA_F_TENTATIVE) && 1473 (dev == NULL || ifp->idev->dev == dev || 1474 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) { 1475 rcu_read_unlock_bh(); 1476 return 1; 1477 } 1478 } 1479 1480 rcu_read_unlock_bh(); 1481 return 0; 1482 } 1483 EXPORT_SYMBOL(ipv6_chk_addr); 1484 1485 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr, 1486 struct net_device *dev) 1487 { 1488 unsigned int hash = inet6_addr_hash(addr); 1489 struct inet6_ifaddr *ifp; 1490 1491 hlist_for_each_entry(ifp, &inet6_addr_lst[hash], addr_lst) { 1492 if (!net_eq(dev_net(ifp->idev->dev), net)) 1493 continue; 1494 if (ipv6_addr_equal(&ifp->addr, addr)) { 1495 if (dev == NULL || ifp->idev->dev == dev) 1496 return true; 1497 } 1498 } 1499 return false; 1500 } 1501 1502 /* Compares an address/prefix_len with addresses on device @dev. 1503 * If one is found it returns true. 1504 */ 1505 bool ipv6_chk_custom_prefix(const struct in6_addr *addr, 1506 const unsigned int prefix_len, struct net_device *dev) 1507 { 1508 struct inet6_dev *idev; 1509 struct inet6_ifaddr *ifa; 1510 bool ret = false; 1511 1512 rcu_read_lock(); 1513 idev = __in6_dev_get(dev); 1514 if (idev) { 1515 read_lock_bh(&idev->lock); 1516 list_for_each_entry(ifa, &idev->addr_list, if_list) { 1517 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len); 1518 if (ret) 1519 break; 1520 } 1521 read_unlock_bh(&idev->lock); 1522 } 1523 rcu_read_unlock(); 1524 1525 return ret; 1526 } 1527 EXPORT_SYMBOL(ipv6_chk_custom_prefix); 1528 1529 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev) 1530 { 1531 struct inet6_dev *idev; 1532 struct inet6_ifaddr *ifa; 1533 int onlink; 1534 1535 onlink = 0; 1536 rcu_read_lock(); 1537 idev = __in6_dev_get(dev); 1538 if (idev) { 1539 read_lock_bh(&idev->lock); 1540 list_for_each_entry(ifa, &idev->addr_list, if_list) { 1541 onlink = ipv6_prefix_equal(addr, &ifa->addr, 1542 ifa->prefix_len); 1543 if (onlink) 1544 break; 1545 } 1546 read_unlock_bh(&idev->lock); 1547 } 1548 rcu_read_unlock(); 1549 return onlink; 1550 } 1551 EXPORT_SYMBOL(ipv6_chk_prefix); 1552 1553 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr, 1554 struct net_device *dev, int strict) 1555 { 1556 struct inet6_ifaddr *ifp, *result = NULL; 1557 unsigned int hash = inet6_addr_hash(addr); 1558 1559 rcu_read_lock_bh(); 1560 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) { 1561 if (!net_eq(dev_net(ifp->idev->dev), net)) 1562 continue; 1563 if (ipv6_addr_equal(&ifp->addr, addr)) { 1564 if (dev == NULL || ifp->idev->dev == dev || 1565 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) { 1566 result = ifp; 1567 in6_ifa_hold(ifp); 1568 break; 1569 } 1570 } 1571 } 1572 rcu_read_unlock_bh(); 1573 1574 return result; 1575 } 1576 1577 /* Gets referenced address, destroys ifaddr */ 1578 1579 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed) 1580 { 1581 if (ifp->flags&IFA_F_PERMANENT) { 1582 spin_lock_bh(&ifp->lock); 1583 addrconf_del_dad_timer(ifp); 1584 ifp->flags |= IFA_F_TENTATIVE; 1585 if (dad_failed) 1586 ifp->flags |= IFA_F_DADFAILED; 1587 spin_unlock_bh(&ifp->lock); 1588 if (dad_failed) 1589 ipv6_ifa_notify(0, ifp); 1590 in6_ifa_put(ifp); 1591 } else if (ifp->flags&IFA_F_TEMPORARY) { 1592 struct inet6_ifaddr *ifpub; 1593 spin_lock_bh(&ifp->lock); 1594 ifpub = ifp->ifpub; 1595 if (ifpub) { 1596 in6_ifa_hold(ifpub); 1597 spin_unlock_bh(&ifp->lock); 1598 ipv6_create_tempaddr(ifpub, ifp); 1599 in6_ifa_put(ifpub); 1600 } else { 1601 spin_unlock_bh(&ifp->lock); 1602 } 1603 ipv6_del_addr(ifp); 1604 } else 1605 ipv6_del_addr(ifp); 1606 } 1607 1608 static int addrconf_dad_end(struct inet6_ifaddr *ifp) 1609 { 1610 int err = -ENOENT; 1611 1612 spin_lock(&ifp->state_lock); 1613 if (ifp->state == INET6_IFADDR_STATE_DAD) { 1614 ifp->state = INET6_IFADDR_STATE_POSTDAD; 1615 err = 0; 1616 } 1617 spin_unlock(&ifp->state_lock); 1618 1619 return err; 1620 } 1621 1622 void addrconf_dad_failure(struct inet6_ifaddr *ifp) 1623 { 1624 struct inet6_dev *idev = ifp->idev; 1625 1626 if (addrconf_dad_end(ifp)) { 1627 in6_ifa_put(ifp); 1628 return; 1629 } 1630 1631 net_info_ratelimited("%s: IPv6 duplicate address %pI6c detected!\n", 1632 ifp->idev->dev->name, &ifp->addr); 1633 1634 if (idev->cnf.accept_dad > 1 && !idev->cnf.disable_ipv6) { 1635 struct in6_addr addr; 1636 1637 addr.s6_addr32[0] = htonl(0xfe800000); 1638 addr.s6_addr32[1] = 0; 1639 1640 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) && 1641 ipv6_addr_equal(&ifp->addr, &addr)) { 1642 /* DAD failed for link-local based on MAC address */ 1643 idev->cnf.disable_ipv6 = 1; 1644 1645 pr_info("%s: IPv6 being disabled!\n", 1646 ifp->idev->dev->name); 1647 } 1648 } 1649 1650 addrconf_dad_stop(ifp, 1); 1651 } 1652 1653 /* Join to solicited addr multicast group. */ 1654 1655 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr) 1656 { 1657 struct in6_addr maddr; 1658 1659 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP)) 1660 return; 1661 1662 addrconf_addr_solict_mult(addr, &maddr); 1663 ipv6_dev_mc_inc(dev, &maddr); 1664 } 1665 1666 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr) 1667 { 1668 struct in6_addr maddr; 1669 1670 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP)) 1671 return; 1672 1673 addrconf_addr_solict_mult(addr, &maddr); 1674 __ipv6_dev_mc_dec(idev, &maddr); 1675 } 1676 1677 static void addrconf_join_anycast(struct inet6_ifaddr *ifp) 1678 { 1679 struct in6_addr addr; 1680 if (ifp->prefix_len == 127) /* RFC 6164 */ 1681 return; 1682 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 1683 if (ipv6_addr_any(&addr)) 1684 return; 1685 ipv6_dev_ac_inc(ifp->idev->dev, &addr); 1686 } 1687 1688 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp) 1689 { 1690 struct in6_addr addr; 1691 if (ifp->prefix_len == 127) /* RFC 6164 */ 1692 return; 1693 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 1694 if (ipv6_addr_any(&addr)) 1695 return; 1696 __ipv6_dev_ac_dec(ifp->idev, &addr); 1697 } 1698 1699 static int addrconf_ifid_eui48(u8 *eui, struct net_device *dev) 1700 { 1701 if (dev->addr_len != ETH_ALEN) 1702 return -1; 1703 memcpy(eui, dev->dev_addr, 3); 1704 memcpy(eui + 5, dev->dev_addr + 3, 3); 1705 1706 /* 1707 * The zSeries OSA network cards can be shared among various 1708 * OS instances, but the OSA cards have only one MAC address. 1709 * This leads to duplicate address conflicts in conjunction 1710 * with IPv6 if more than one instance uses the same card. 1711 * 1712 * The driver for these cards can deliver a unique 16-bit 1713 * identifier for each instance sharing the same card. It is 1714 * placed instead of 0xFFFE in the interface identifier. The 1715 * "u" bit of the interface identifier is not inverted in this 1716 * case. Hence the resulting interface identifier has local 1717 * scope according to RFC2373. 1718 */ 1719 if (dev->dev_id) { 1720 eui[3] = (dev->dev_id >> 8) & 0xFF; 1721 eui[4] = dev->dev_id & 0xFF; 1722 } else { 1723 eui[3] = 0xFF; 1724 eui[4] = 0xFE; 1725 eui[0] ^= 2; 1726 } 1727 return 0; 1728 } 1729 1730 static int addrconf_ifid_eui64(u8 *eui, struct net_device *dev) 1731 { 1732 if (dev->addr_len != IEEE802154_ADDR_LEN) 1733 return -1; 1734 memcpy(eui, dev->dev_addr, 8); 1735 eui[0] ^= 2; 1736 return 0; 1737 } 1738 1739 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev) 1740 { 1741 union fwnet_hwaddr *ha; 1742 1743 if (dev->addr_len != FWNET_ALEN) 1744 return -1; 1745 1746 ha = (union fwnet_hwaddr *)dev->dev_addr; 1747 1748 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id)); 1749 eui[0] ^= 2; 1750 return 0; 1751 } 1752 1753 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev) 1754 { 1755 /* XXX: inherit EUI-64 from other interface -- yoshfuji */ 1756 if (dev->addr_len != ARCNET_ALEN) 1757 return -1; 1758 memset(eui, 0, 7); 1759 eui[7] = *(u8 *)dev->dev_addr; 1760 return 0; 1761 } 1762 1763 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev) 1764 { 1765 if (dev->addr_len != INFINIBAND_ALEN) 1766 return -1; 1767 memcpy(eui, dev->dev_addr + 12, 8); 1768 eui[0] |= 2; 1769 return 0; 1770 } 1771 1772 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr) 1773 { 1774 if (addr == 0) 1775 return -1; 1776 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) || 1777 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) || 1778 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) || 1779 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) || 1780 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) || 1781 ipv4_is_lbcast(addr)) ? 0x00 : 0x02; 1782 eui[1] = 0; 1783 eui[2] = 0x5E; 1784 eui[3] = 0xFE; 1785 memcpy(eui + 4, &addr, 4); 1786 return 0; 1787 } 1788 1789 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev) 1790 { 1791 if (dev->priv_flags & IFF_ISATAP) 1792 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 1793 return -1; 1794 } 1795 1796 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev) 1797 { 1798 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 1799 } 1800 1801 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev) 1802 { 1803 memcpy(eui, dev->perm_addr, 3); 1804 memcpy(eui + 5, dev->perm_addr + 3, 3); 1805 eui[3] = 0xFF; 1806 eui[4] = 0xFE; 1807 eui[0] ^= 2; 1808 return 0; 1809 } 1810 1811 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev) 1812 { 1813 switch (dev->type) { 1814 case ARPHRD_ETHER: 1815 case ARPHRD_FDDI: 1816 return addrconf_ifid_eui48(eui, dev); 1817 case ARPHRD_ARCNET: 1818 return addrconf_ifid_arcnet(eui, dev); 1819 case ARPHRD_INFINIBAND: 1820 return addrconf_ifid_infiniband(eui, dev); 1821 case ARPHRD_SIT: 1822 return addrconf_ifid_sit(eui, dev); 1823 case ARPHRD_IPGRE: 1824 return addrconf_ifid_gre(eui, dev); 1825 case ARPHRD_IEEE802154: 1826 return addrconf_ifid_eui64(eui, dev); 1827 case ARPHRD_IEEE1394: 1828 return addrconf_ifid_ieee1394(eui, dev); 1829 case ARPHRD_TUNNEL6: 1830 return addrconf_ifid_ip6tnl(eui, dev); 1831 } 1832 return -1; 1833 } 1834 1835 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev) 1836 { 1837 int err = -1; 1838 struct inet6_ifaddr *ifp; 1839 1840 read_lock_bh(&idev->lock); 1841 list_for_each_entry(ifp, &idev->addr_list, if_list) { 1842 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) { 1843 memcpy(eui, ifp->addr.s6_addr+8, 8); 1844 err = 0; 1845 break; 1846 } 1847 } 1848 read_unlock_bh(&idev->lock); 1849 return err; 1850 } 1851 1852 /* (re)generation of randomized interface identifier (RFC 3041 3.2, 3.5) */ 1853 static void __ipv6_regen_rndid(struct inet6_dev *idev) 1854 { 1855 regen: 1856 get_random_bytes(idev->rndid, sizeof(idev->rndid)); 1857 idev->rndid[0] &= ~0x02; 1858 1859 /* 1860 * <draft-ietf-ipngwg-temp-addresses-v2-00.txt>: 1861 * check if generated address is not inappropriate 1862 * 1863 * - Reserved subnet anycast (RFC 2526) 1864 * 11111101 11....11 1xxxxxxx 1865 * - ISATAP (RFC4214) 6.1 1866 * 00-00-5E-FE-xx-xx-xx-xx 1867 * - value 0 1868 * - XXX: already assigned to an address on the device 1869 */ 1870 if (idev->rndid[0] == 0xfd && 1871 (idev->rndid[1]&idev->rndid[2]&idev->rndid[3]&idev->rndid[4]&idev->rndid[5]&idev->rndid[6]) == 0xff && 1872 (idev->rndid[7]&0x80)) 1873 goto regen; 1874 if ((idev->rndid[0]|idev->rndid[1]) == 0) { 1875 if (idev->rndid[2] == 0x5e && idev->rndid[3] == 0xfe) 1876 goto regen; 1877 if ((idev->rndid[2]|idev->rndid[3]|idev->rndid[4]|idev->rndid[5]|idev->rndid[6]|idev->rndid[7]) == 0x00) 1878 goto regen; 1879 } 1880 } 1881 1882 static void ipv6_regen_rndid(unsigned long data) 1883 { 1884 struct inet6_dev *idev = (struct inet6_dev *) data; 1885 unsigned long expires; 1886 1887 rcu_read_lock_bh(); 1888 write_lock_bh(&idev->lock); 1889 1890 if (idev->dead) 1891 goto out; 1892 1893 __ipv6_regen_rndid(idev); 1894 1895 expires = jiffies + 1896 idev->cnf.temp_prefered_lft * HZ - 1897 idev->cnf.regen_max_retry * idev->cnf.dad_transmits * 1898 NEIGH_VAR(idev->nd_parms, RETRANS_TIME) - 1899 idev->cnf.max_desync_factor * HZ; 1900 if (time_before(expires, jiffies)) { 1901 pr_warn("%s: too short regeneration interval; timer disabled for %s\n", 1902 __func__, idev->dev->name); 1903 goto out; 1904 } 1905 1906 if (!mod_timer(&idev->regen_timer, expires)) 1907 in6_dev_hold(idev); 1908 1909 out: 1910 write_unlock_bh(&idev->lock); 1911 rcu_read_unlock_bh(); 1912 in6_dev_put(idev); 1913 } 1914 1915 static void __ipv6_try_regen_rndid(struct inet6_dev *idev, struct in6_addr *tmpaddr) 1916 { 1917 if (tmpaddr && memcmp(idev->rndid, &tmpaddr->s6_addr[8], 8) == 0) 1918 __ipv6_regen_rndid(idev); 1919 } 1920 1921 /* 1922 * Add prefix route. 1923 */ 1924 1925 static void 1926 addrconf_prefix_route(struct in6_addr *pfx, int plen, struct net_device *dev, 1927 unsigned long expires, u32 flags) 1928 { 1929 struct fib6_config cfg = { 1930 .fc_table = RT6_TABLE_PREFIX, 1931 .fc_metric = IP6_RT_PRIO_ADDRCONF, 1932 .fc_ifindex = dev->ifindex, 1933 .fc_expires = expires, 1934 .fc_dst_len = plen, 1935 .fc_flags = RTF_UP | flags, 1936 .fc_nlinfo.nl_net = dev_net(dev), 1937 .fc_protocol = RTPROT_KERNEL, 1938 }; 1939 1940 cfg.fc_dst = *pfx; 1941 1942 /* Prevent useless cloning on PtP SIT. 1943 This thing is done here expecting that the whole 1944 class of non-broadcast devices need not cloning. 1945 */ 1946 #if IS_ENABLED(CONFIG_IPV6_SIT) 1947 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT)) 1948 cfg.fc_flags |= RTF_NONEXTHOP; 1949 #endif 1950 1951 ip6_route_add(&cfg); 1952 } 1953 1954 1955 static struct rt6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 1956 int plen, 1957 const struct net_device *dev, 1958 u32 flags, u32 noflags) 1959 { 1960 struct fib6_node *fn; 1961 struct rt6_info *rt = NULL; 1962 struct fib6_table *table; 1963 1964 table = fib6_get_table(dev_net(dev), RT6_TABLE_PREFIX); 1965 if (table == NULL) 1966 return NULL; 1967 1968 read_lock_bh(&table->tb6_lock); 1969 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0); 1970 if (!fn) 1971 goto out; 1972 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) { 1973 if (rt->dst.dev->ifindex != dev->ifindex) 1974 continue; 1975 if ((rt->rt6i_flags & flags) != flags) 1976 continue; 1977 if ((rt->rt6i_flags & noflags) != 0) 1978 continue; 1979 dst_hold(&rt->dst); 1980 break; 1981 } 1982 out: 1983 read_unlock_bh(&table->tb6_lock); 1984 return rt; 1985 } 1986 1987 1988 /* Create "default" multicast route to the interface */ 1989 1990 static void addrconf_add_mroute(struct net_device *dev) 1991 { 1992 struct fib6_config cfg = { 1993 .fc_table = RT6_TABLE_LOCAL, 1994 .fc_metric = IP6_RT_PRIO_ADDRCONF, 1995 .fc_ifindex = dev->ifindex, 1996 .fc_dst_len = 8, 1997 .fc_flags = RTF_UP, 1998 .fc_nlinfo.nl_net = dev_net(dev), 1999 }; 2000 2001 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0); 2002 2003 ip6_route_add(&cfg); 2004 } 2005 2006 static struct inet6_dev *addrconf_add_dev(struct net_device *dev) 2007 { 2008 struct inet6_dev *idev; 2009 2010 ASSERT_RTNL(); 2011 2012 idev = ipv6_find_idev(dev); 2013 if (!idev) 2014 return ERR_PTR(-ENOBUFS); 2015 2016 if (idev->cnf.disable_ipv6) 2017 return ERR_PTR(-EACCES); 2018 2019 /* Add default multicast route */ 2020 if (!(dev->flags & IFF_LOOPBACK)) 2021 addrconf_add_mroute(dev); 2022 2023 return idev; 2024 } 2025 2026 static void manage_tempaddrs(struct inet6_dev *idev, 2027 struct inet6_ifaddr *ifp, 2028 __u32 valid_lft, __u32 prefered_lft, 2029 bool create, unsigned long now) 2030 { 2031 u32 flags; 2032 struct inet6_ifaddr *ift; 2033 2034 read_lock_bh(&idev->lock); 2035 /* update all temporary addresses in the list */ 2036 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) { 2037 int age, max_valid, max_prefered; 2038 2039 if (ifp != ift->ifpub) 2040 continue; 2041 2042 /* RFC 4941 section 3.3: 2043 * If a received option will extend the lifetime of a public 2044 * address, the lifetimes of temporary addresses should 2045 * be extended, subject to the overall constraint that no 2046 * temporary addresses should ever remain "valid" or "preferred" 2047 * for a time longer than (TEMP_VALID_LIFETIME) or 2048 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively. 2049 */ 2050 age = (now - ift->cstamp) / HZ; 2051 max_valid = idev->cnf.temp_valid_lft - age; 2052 if (max_valid < 0) 2053 max_valid = 0; 2054 2055 max_prefered = idev->cnf.temp_prefered_lft - 2056 idev->cnf.max_desync_factor - age; 2057 if (max_prefered < 0) 2058 max_prefered = 0; 2059 2060 if (valid_lft > max_valid) 2061 valid_lft = max_valid; 2062 2063 if (prefered_lft > max_prefered) 2064 prefered_lft = max_prefered; 2065 2066 spin_lock(&ift->lock); 2067 flags = ift->flags; 2068 ift->valid_lft = valid_lft; 2069 ift->prefered_lft = prefered_lft; 2070 ift->tstamp = now; 2071 if (prefered_lft > 0) 2072 ift->flags &= ~IFA_F_DEPRECATED; 2073 2074 spin_unlock(&ift->lock); 2075 if (!(flags&IFA_F_TENTATIVE)) 2076 ipv6_ifa_notify(0, ift); 2077 } 2078 2079 if ((create || list_empty(&idev->tempaddr_list)) && 2080 idev->cnf.use_tempaddr > 0) { 2081 /* When a new public address is created as described 2082 * in [ADDRCONF], also create a new temporary address. 2083 * Also create a temporary address if it's enabled but 2084 * no temporary address currently exists. 2085 */ 2086 read_unlock_bh(&idev->lock); 2087 ipv6_create_tempaddr(ifp, NULL); 2088 } else { 2089 read_unlock_bh(&idev->lock); 2090 } 2091 } 2092 2093 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao) 2094 { 2095 struct prefix_info *pinfo; 2096 __u32 valid_lft; 2097 __u32 prefered_lft; 2098 int addr_type; 2099 struct inet6_dev *in6_dev; 2100 struct net *net = dev_net(dev); 2101 2102 pinfo = (struct prefix_info *) opt; 2103 2104 if (len < sizeof(struct prefix_info)) { 2105 ADBG("addrconf: prefix option too short\n"); 2106 return; 2107 } 2108 2109 /* 2110 * Validation checks ([ADDRCONF], page 19) 2111 */ 2112 2113 addr_type = ipv6_addr_type(&pinfo->prefix); 2114 2115 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)) 2116 return; 2117 2118 valid_lft = ntohl(pinfo->valid); 2119 prefered_lft = ntohl(pinfo->prefered); 2120 2121 if (prefered_lft > valid_lft) { 2122 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n"); 2123 return; 2124 } 2125 2126 in6_dev = in6_dev_get(dev); 2127 2128 if (in6_dev == NULL) { 2129 net_dbg_ratelimited("addrconf: device %s not configured\n", 2130 dev->name); 2131 return; 2132 } 2133 2134 /* 2135 * Two things going on here: 2136 * 1) Add routes for on-link prefixes 2137 * 2) Configure prefixes with the auto flag set 2138 */ 2139 2140 if (pinfo->onlink) { 2141 struct rt6_info *rt; 2142 unsigned long rt_expires; 2143 2144 /* Avoid arithmetic overflow. Really, we could 2145 * save rt_expires in seconds, likely valid_lft, 2146 * but it would require division in fib gc, that it 2147 * not good. 2148 */ 2149 if (HZ > USER_HZ) 2150 rt_expires = addrconf_timeout_fixup(valid_lft, HZ); 2151 else 2152 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ); 2153 2154 if (addrconf_finite_timeout(rt_expires)) 2155 rt_expires *= HZ; 2156 2157 rt = addrconf_get_prefix_route(&pinfo->prefix, 2158 pinfo->prefix_len, 2159 dev, 2160 RTF_ADDRCONF | RTF_PREFIX_RT, 2161 RTF_GATEWAY | RTF_DEFAULT); 2162 2163 if (rt) { 2164 /* Autoconf prefix route */ 2165 if (valid_lft == 0) { 2166 ip6_del_rt(rt); 2167 rt = NULL; 2168 } else if (addrconf_finite_timeout(rt_expires)) { 2169 /* not infinity */ 2170 rt6_set_expires(rt, jiffies + rt_expires); 2171 } else { 2172 rt6_clean_expires(rt); 2173 } 2174 } else if (valid_lft) { 2175 clock_t expires = 0; 2176 int flags = RTF_ADDRCONF | RTF_PREFIX_RT; 2177 if (addrconf_finite_timeout(rt_expires)) { 2178 /* not infinity */ 2179 flags |= RTF_EXPIRES; 2180 expires = jiffies_to_clock_t(rt_expires); 2181 } 2182 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len, 2183 dev, expires, flags); 2184 } 2185 ip6_rt_put(rt); 2186 } 2187 2188 /* Try to figure out our local address for this prefix */ 2189 2190 if (pinfo->autoconf && in6_dev->cnf.autoconf) { 2191 struct inet6_ifaddr *ifp; 2192 struct in6_addr addr; 2193 int create = 0, update_lft = 0; 2194 bool tokenized = false; 2195 2196 if (pinfo->prefix_len == 64) { 2197 memcpy(&addr, &pinfo->prefix, 8); 2198 2199 if (!ipv6_addr_any(&in6_dev->token)) { 2200 read_lock_bh(&in6_dev->lock); 2201 memcpy(addr.s6_addr + 8, 2202 in6_dev->token.s6_addr + 8, 8); 2203 read_unlock_bh(&in6_dev->lock); 2204 tokenized = true; 2205 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) && 2206 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) { 2207 in6_dev_put(in6_dev); 2208 return; 2209 } 2210 goto ok; 2211 } 2212 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n", 2213 pinfo->prefix_len); 2214 in6_dev_put(in6_dev); 2215 return; 2216 2217 ok: 2218 2219 ifp = ipv6_get_ifaddr(net, &addr, dev, 1); 2220 2221 if (ifp == NULL && valid_lft) { 2222 int max_addresses = in6_dev->cnf.max_addresses; 2223 u32 addr_flags = 0; 2224 2225 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 2226 if (in6_dev->cnf.optimistic_dad && 2227 !net->ipv6.devconf_all->forwarding && sllao) 2228 addr_flags = IFA_F_OPTIMISTIC; 2229 #endif 2230 2231 /* Do not allow to create too much of autoconfigured 2232 * addresses; this would be too easy way to crash kernel. 2233 */ 2234 if (!max_addresses || 2235 ipv6_count_addresses(in6_dev) < max_addresses) 2236 ifp = ipv6_add_addr(in6_dev, &addr, NULL, 2237 pinfo->prefix_len, 2238 addr_type&IPV6_ADDR_SCOPE_MASK, 2239 addr_flags, valid_lft, 2240 prefered_lft); 2241 2242 if (IS_ERR_OR_NULL(ifp)) { 2243 in6_dev_put(in6_dev); 2244 return; 2245 } 2246 2247 ifp->flags |= IFA_F_MANAGETEMPADDR; 2248 update_lft = 0; 2249 create = 1; 2250 ifp->cstamp = jiffies; 2251 ifp->tokenized = tokenized; 2252 addrconf_dad_start(ifp); 2253 } 2254 2255 if (ifp) { 2256 u32 flags; 2257 unsigned long now; 2258 u32 stored_lft; 2259 2260 /* update lifetime (RFC2462 5.5.3 e) */ 2261 spin_lock(&ifp->lock); 2262 now = jiffies; 2263 if (ifp->valid_lft > (now - ifp->tstamp) / HZ) 2264 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ; 2265 else 2266 stored_lft = 0; 2267 if (!update_lft && !create && stored_lft) { 2268 const u32 minimum_lft = min( 2269 stored_lft, (u32)MIN_VALID_LIFETIME); 2270 valid_lft = max(valid_lft, minimum_lft); 2271 2272 /* RFC4862 Section 5.5.3e: 2273 * "Note that the preferred lifetime of the 2274 * corresponding address is always reset to 2275 * the Preferred Lifetime in the received 2276 * Prefix Information option, regardless of 2277 * whether the valid lifetime is also reset or 2278 * ignored." 2279 * 2280 * So we should always update prefered_lft here. 2281 */ 2282 update_lft = 1; 2283 } 2284 2285 if (update_lft) { 2286 ifp->valid_lft = valid_lft; 2287 ifp->prefered_lft = prefered_lft; 2288 ifp->tstamp = now; 2289 flags = ifp->flags; 2290 ifp->flags &= ~IFA_F_DEPRECATED; 2291 spin_unlock(&ifp->lock); 2292 2293 if (!(flags&IFA_F_TENTATIVE)) 2294 ipv6_ifa_notify(0, ifp); 2295 } else 2296 spin_unlock(&ifp->lock); 2297 2298 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft, 2299 create, now); 2300 2301 in6_ifa_put(ifp); 2302 addrconf_verify(0); 2303 } 2304 } 2305 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo); 2306 in6_dev_put(in6_dev); 2307 } 2308 2309 /* 2310 * Set destination address. 2311 * Special case for SIT interfaces where we create a new "virtual" 2312 * device. 2313 */ 2314 int addrconf_set_dstaddr(struct net *net, void __user *arg) 2315 { 2316 struct in6_ifreq ireq; 2317 struct net_device *dev; 2318 int err = -EINVAL; 2319 2320 rtnl_lock(); 2321 2322 err = -EFAULT; 2323 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 2324 goto err_exit; 2325 2326 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 2327 2328 err = -ENODEV; 2329 if (dev == NULL) 2330 goto err_exit; 2331 2332 #if IS_ENABLED(CONFIG_IPV6_SIT) 2333 if (dev->type == ARPHRD_SIT) { 2334 const struct net_device_ops *ops = dev->netdev_ops; 2335 struct ifreq ifr; 2336 struct ip_tunnel_parm p; 2337 2338 err = -EADDRNOTAVAIL; 2339 if (!(ipv6_addr_type(&ireq.ifr6_addr) & IPV6_ADDR_COMPATv4)) 2340 goto err_exit; 2341 2342 memset(&p, 0, sizeof(p)); 2343 p.iph.daddr = ireq.ifr6_addr.s6_addr32[3]; 2344 p.iph.saddr = 0; 2345 p.iph.version = 4; 2346 p.iph.ihl = 5; 2347 p.iph.protocol = IPPROTO_IPV6; 2348 p.iph.ttl = 64; 2349 ifr.ifr_ifru.ifru_data = (__force void __user *)&p; 2350 2351 if (ops->ndo_do_ioctl) { 2352 mm_segment_t oldfs = get_fs(); 2353 2354 set_fs(KERNEL_DS); 2355 err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL); 2356 set_fs(oldfs); 2357 } else 2358 err = -EOPNOTSUPP; 2359 2360 if (err == 0) { 2361 err = -ENOBUFS; 2362 dev = __dev_get_by_name(net, p.name); 2363 if (!dev) 2364 goto err_exit; 2365 err = dev_open(dev); 2366 } 2367 } 2368 #endif 2369 2370 err_exit: 2371 rtnl_unlock(); 2372 return err; 2373 } 2374 2375 /* 2376 * Manual configuration of address on an interface 2377 */ 2378 static int inet6_addr_add(struct net *net, int ifindex, 2379 const struct in6_addr *pfx, 2380 const struct in6_addr *peer_pfx, 2381 unsigned int plen, __u32 ifa_flags, 2382 __u32 prefered_lft, __u32 valid_lft) 2383 { 2384 struct inet6_ifaddr *ifp; 2385 struct inet6_dev *idev; 2386 struct net_device *dev; 2387 int scope; 2388 u32 flags; 2389 clock_t expires; 2390 unsigned long timeout; 2391 2392 ASSERT_RTNL(); 2393 2394 if (plen > 128) 2395 return -EINVAL; 2396 2397 /* check the lifetime */ 2398 if (!valid_lft || prefered_lft > valid_lft) 2399 return -EINVAL; 2400 2401 if (ifa_flags & IFA_F_MANAGETEMPADDR && plen != 64) 2402 return -EINVAL; 2403 2404 dev = __dev_get_by_index(net, ifindex); 2405 if (!dev) 2406 return -ENODEV; 2407 2408 idev = addrconf_add_dev(dev); 2409 if (IS_ERR(idev)) 2410 return PTR_ERR(idev); 2411 2412 scope = ipv6_addr_scope(pfx); 2413 2414 timeout = addrconf_timeout_fixup(valid_lft, HZ); 2415 if (addrconf_finite_timeout(timeout)) { 2416 expires = jiffies_to_clock_t(timeout * HZ); 2417 valid_lft = timeout; 2418 flags = RTF_EXPIRES; 2419 } else { 2420 expires = 0; 2421 flags = 0; 2422 ifa_flags |= IFA_F_PERMANENT; 2423 } 2424 2425 timeout = addrconf_timeout_fixup(prefered_lft, HZ); 2426 if (addrconf_finite_timeout(timeout)) { 2427 if (timeout == 0) 2428 ifa_flags |= IFA_F_DEPRECATED; 2429 prefered_lft = timeout; 2430 } 2431 2432 ifp = ipv6_add_addr(idev, pfx, peer_pfx, plen, scope, ifa_flags, 2433 valid_lft, prefered_lft); 2434 2435 if (!IS_ERR(ifp)) { 2436 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, dev, 2437 expires, flags); 2438 /* 2439 * Note that section 3.1 of RFC 4429 indicates 2440 * that the Optimistic flag should not be set for 2441 * manually configured addresses 2442 */ 2443 addrconf_dad_start(ifp); 2444 if (ifa_flags & IFA_F_MANAGETEMPADDR) 2445 manage_tempaddrs(idev, ifp, valid_lft, prefered_lft, 2446 true, jiffies); 2447 in6_ifa_put(ifp); 2448 addrconf_verify(0); 2449 return 0; 2450 } 2451 2452 return PTR_ERR(ifp); 2453 } 2454 2455 static int inet6_addr_del(struct net *net, int ifindex, const struct in6_addr *pfx, 2456 unsigned int plen) 2457 { 2458 struct inet6_ifaddr *ifp; 2459 struct inet6_dev *idev; 2460 struct net_device *dev; 2461 2462 if (plen > 128) 2463 return -EINVAL; 2464 2465 dev = __dev_get_by_index(net, ifindex); 2466 if (!dev) 2467 return -ENODEV; 2468 2469 if ((idev = __in6_dev_get(dev)) == NULL) 2470 return -ENXIO; 2471 2472 read_lock_bh(&idev->lock); 2473 list_for_each_entry(ifp, &idev->addr_list, if_list) { 2474 if (ifp->prefix_len == plen && 2475 ipv6_addr_equal(pfx, &ifp->addr)) { 2476 in6_ifa_hold(ifp); 2477 read_unlock_bh(&idev->lock); 2478 2479 ipv6_del_addr(ifp); 2480 return 0; 2481 } 2482 } 2483 read_unlock_bh(&idev->lock); 2484 return -EADDRNOTAVAIL; 2485 } 2486 2487 2488 int addrconf_add_ifaddr(struct net *net, void __user *arg) 2489 { 2490 struct in6_ifreq ireq; 2491 int err; 2492 2493 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 2494 return -EPERM; 2495 2496 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 2497 return -EFAULT; 2498 2499 rtnl_lock(); 2500 err = inet6_addr_add(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, NULL, 2501 ireq.ifr6_prefixlen, IFA_F_PERMANENT, 2502 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME); 2503 rtnl_unlock(); 2504 return err; 2505 } 2506 2507 int addrconf_del_ifaddr(struct net *net, void __user *arg) 2508 { 2509 struct in6_ifreq ireq; 2510 int err; 2511 2512 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 2513 return -EPERM; 2514 2515 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 2516 return -EFAULT; 2517 2518 rtnl_lock(); 2519 err = inet6_addr_del(net, ireq.ifr6_ifindex, &ireq.ifr6_addr, 2520 ireq.ifr6_prefixlen); 2521 rtnl_unlock(); 2522 return err; 2523 } 2524 2525 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr, 2526 int plen, int scope) 2527 { 2528 struct inet6_ifaddr *ifp; 2529 2530 ifp = ipv6_add_addr(idev, addr, NULL, plen, 2531 scope, IFA_F_PERMANENT, 0, 0); 2532 if (!IS_ERR(ifp)) { 2533 spin_lock_bh(&ifp->lock); 2534 ifp->flags &= ~IFA_F_TENTATIVE; 2535 spin_unlock_bh(&ifp->lock); 2536 ipv6_ifa_notify(RTM_NEWADDR, ifp); 2537 in6_ifa_put(ifp); 2538 } 2539 } 2540 2541 #if IS_ENABLED(CONFIG_IPV6_SIT) 2542 static void sit_add_v4_addrs(struct inet6_dev *idev) 2543 { 2544 struct in6_addr addr; 2545 struct net_device *dev; 2546 struct net *net = dev_net(idev->dev); 2547 int scope, plen; 2548 u32 pflags = 0; 2549 2550 ASSERT_RTNL(); 2551 2552 memset(&addr, 0, sizeof(struct in6_addr)); 2553 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4); 2554 2555 if (idev->dev->flags&IFF_POINTOPOINT) { 2556 addr.s6_addr32[0] = htonl(0xfe800000); 2557 scope = IFA_LINK; 2558 plen = 64; 2559 } else { 2560 scope = IPV6_ADDR_COMPATv4; 2561 plen = 96; 2562 pflags |= RTF_NONEXTHOP; 2563 } 2564 2565 if (addr.s6_addr32[3]) { 2566 add_addr(idev, &addr, plen, scope); 2567 addrconf_prefix_route(&addr, plen, idev->dev, 0, pflags); 2568 return; 2569 } 2570 2571 for_each_netdev(net, dev) { 2572 struct in_device *in_dev = __in_dev_get_rtnl(dev); 2573 if (in_dev && (dev->flags & IFF_UP)) { 2574 struct in_ifaddr *ifa; 2575 2576 int flag = scope; 2577 2578 for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) { 2579 2580 addr.s6_addr32[3] = ifa->ifa_local; 2581 2582 if (ifa->ifa_scope == RT_SCOPE_LINK) 2583 continue; 2584 if (ifa->ifa_scope >= RT_SCOPE_HOST) { 2585 if (idev->dev->flags&IFF_POINTOPOINT) 2586 continue; 2587 flag |= IFA_HOST; 2588 } 2589 2590 add_addr(idev, &addr, plen, flag); 2591 addrconf_prefix_route(&addr, plen, idev->dev, 0, 2592 pflags); 2593 } 2594 } 2595 } 2596 } 2597 #endif 2598 2599 static void init_loopback(struct net_device *dev) 2600 { 2601 struct inet6_dev *idev; 2602 struct net_device *sp_dev; 2603 struct inet6_ifaddr *sp_ifa; 2604 struct rt6_info *sp_rt; 2605 2606 /* ::1 */ 2607 2608 ASSERT_RTNL(); 2609 2610 if ((idev = ipv6_find_idev(dev)) == NULL) { 2611 pr_debug("%s: add_dev failed\n", __func__); 2612 return; 2613 } 2614 2615 add_addr(idev, &in6addr_loopback, 128, IFA_HOST); 2616 2617 /* Add routes to other interface's IPv6 addresses */ 2618 for_each_netdev(dev_net(dev), sp_dev) { 2619 if (!strcmp(sp_dev->name, dev->name)) 2620 continue; 2621 2622 idev = __in6_dev_get(sp_dev); 2623 if (!idev) 2624 continue; 2625 2626 read_lock_bh(&idev->lock); 2627 list_for_each_entry(sp_ifa, &idev->addr_list, if_list) { 2628 2629 if (sp_ifa->flags & (IFA_F_DADFAILED | IFA_F_TENTATIVE)) 2630 continue; 2631 2632 if (sp_ifa->rt) 2633 continue; 2634 2635 sp_rt = addrconf_dst_alloc(idev, &sp_ifa->addr, false); 2636 2637 /* Failure cases are ignored */ 2638 if (!IS_ERR(sp_rt)) { 2639 sp_ifa->rt = sp_rt; 2640 ip6_ins_rt(sp_rt); 2641 } 2642 } 2643 read_unlock_bh(&idev->lock); 2644 } 2645 } 2646 2647 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr) 2648 { 2649 struct inet6_ifaddr *ifp; 2650 u32 addr_flags = IFA_F_PERMANENT; 2651 2652 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 2653 if (idev->cnf.optimistic_dad && 2654 !dev_net(idev->dev)->ipv6.devconf_all->forwarding) 2655 addr_flags |= IFA_F_OPTIMISTIC; 2656 #endif 2657 2658 2659 ifp = ipv6_add_addr(idev, addr, NULL, 64, IFA_LINK, addr_flags, 0, 0); 2660 if (!IS_ERR(ifp)) { 2661 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, idev->dev, 0, 0); 2662 addrconf_dad_start(ifp); 2663 in6_ifa_put(ifp); 2664 } 2665 } 2666 2667 static void addrconf_dev_config(struct net_device *dev) 2668 { 2669 struct in6_addr addr; 2670 struct inet6_dev *idev; 2671 2672 ASSERT_RTNL(); 2673 2674 if ((dev->type != ARPHRD_ETHER) && 2675 (dev->type != ARPHRD_FDDI) && 2676 (dev->type != ARPHRD_ARCNET) && 2677 (dev->type != ARPHRD_INFINIBAND) && 2678 (dev->type != ARPHRD_IEEE802154) && 2679 (dev->type != ARPHRD_IEEE1394) && 2680 (dev->type != ARPHRD_TUNNEL6)) { 2681 /* Alas, we support only Ethernet autoconfiguration. */ 2682 return; 2683 } 2684 2685 idev = addrconf_add_dev(dev); 2686 if (IS_ERR(idev)) 2687 return; 2688 2689 memset(&addr, 0, sizeof(struct in6_addr)); 2690 addr.s6_addr32[0] = htonl(0xFE800000); 2691 2692 if (ipv6_generate_eui64(addr.s6_addr + 8, dev) == 0) 2693 addrconf_add_linklocal(idev, &addr); 2694 } 2695 2696 #if IS_ENABLED(CONFIG_IPV6_SIT) 2697 static void addrconf_sit_config(struct net_device *dev) 2698 { 2699 struct inet6_dev *idev; 2700 2701 ASSERT_RTNL(); 2702 2703 /* 2704 * Configure the tunnel with one of our IPv4 2705 * addresses... we should configure all of 2706 * our v4 addrs in the tunnel 2707 */ 2708 2709 if ((idev = ipv6_find_idev(dev)) == NULL) { 2710 pr_debug("%s: add_dev failed\n", __func__); 2711 return; 2712 } 2713 2714 if (dev->priv_flags & IFF_ISATAP) { 2715 struct in6_addr addr; 2716 2717 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0); 2718 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev)) 2719 addrconf_add_linklocal(idev, &addr); 2720 return; 2721 } 2722 2723 sit_add_v4_addrs(idev); 2724 2725 if (dev->flags&IFF_POINTOPOINT) 2726 addrconf_add_mroute(dev); 2727 } 2728 #endif 2729 2730 #if IS_ENABLED(CONFIG_NET_IPGRE) 2731 static void addrconf_gre_config(struct net_device *dev) 2732 { 2733 struct inet6_dev *idev; 2734 struct in6_addr addr; 2735 2736 ASSERT_RTNL(); 2737 2738 if ((idev = ipv6_find_idev(dev)) == NULL) { 2739 pr_debug("%s: add_dev failed\n", __func__); 2740 return; 2741 } 2742 2743 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0); 2744 if (!ipv6_generate_eui64(addr.s6_addr + 8, dev)) 2745 addrconf_add_linklocal(idev, &addr); 2746 } 2747 #endif 2748 2749 static inline int 2750 ipv6_inherit_linklocal(struct inet6_dev *idev, struct net_device *link_dev) 2751 { 2752 struct in6_addr lladdr; 2753 2754 if (!ipv6_get_lladdr(link_dev, &lladdr, IFA_F_TENTATIVE)) { 2755 addrconf_add_linklocal(idev, &lladdr); 2756 return 0; 2757 } 2758 return -1; 2759 } 2760 2761 static int addrconf_notify(struct notifier_block *this, unsigned long event, 2762 void *ptr) 2763 { 2764 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 2765 struct inet6_dev *idev = __in6_dev_get(dev); 2766 int run_pending = 0; 2767 int err; 2768 2769 switch (event) { 2770 case NETDEV_REGISTER: 2771 if (!idev && dev->mtu >= IPV6_MIN_MTU) { 2772 idev = ipv6_add_dev(dev); 2773 if (!idev) 2774 return notifier_from_errno(-ENOMEM); 2775 } 2776 break; 2777 2778 case NETDEV_UP: 2779 case NETDEV_CHANGE: 2780 if (dev->flags & IFF_SLAVE) 2781 break; 2782 2783 if (event == NETDEV_UP) { 2784 if (!addrconf_qdisc_ok(dev)) { 2785 /* device is not ready yet. */ 2786 pr_info("ADDRCONF(NETDEV_UP): %s: link is not ready\n", 2787 dev->name); 2788 break; 2789 } 2790 2791 if (!idev && dev->mtu >= IPV6_MIN_MTU) 2792 idev = ipv6_add_dev(dev); 2793 2794 if (idev) { 2795 idev->if_flags |= IF_READY; 2796 run_pending = 1; 2797 } 2798 } else { 2799 if (!addrconf_qdisc_ok(dev)) { 2800 /* device is still not ready. */ 2801 break; 2802 } 2803 2804 if (idev) { 2805 if (idev->if_flags & IF_READY) 2806 /* device is already configured. */ 2807 break; 2808 idev->if_flags |= IF_READY; 2809 } 2810 2811 pr_info("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n", 2812 dev->name); 2813 2814 run_pending = 1; 2815 } 2816 2817 switch (dev->type) { 2818 #if IS_ENABLED(CONFIG_IPV6_SIT) 2819 case ARPHRD_SIT: 2820 addrconf_sit_config(dev); 2821 break; 2822 #endif 2823 #if IS_ENABLED(CONFIG_NET_IPGRE) 2824 case ARPHRD_IPGRE: 2825 addrconf_gre_config(dev); 2826 break; 2827 #endif 2828 case ARPHRD_LOOPBACK: 2829 init_loopback(dev); 2830 break; 2831 2832 default: 2833 addrconf_dev_config(dev); 2834 break; 2835 } 2836 2837 if (idev) { 2838 if (run_pending) 2839 addrconf_dad_run(idev); 2840 2841 /* 2842 * If the MTU changed during the interface down, 2843 * when the interface up, the changed MTU must be 2844 * reflected in the idev as well as routers. 2845 */ 2846 if (idev->cnf.mtu6 != dev->mtu && 2847 dev->mtu >= IPV6_MIN_MTU) { 2848 rt6_mtu_change(dev, dev->mtu); 2849 idev->cnf.mtu6 = dev->mtu; 2850 } 2851 idev->tstamp = jiffies; 2852 inet6_ifinfo_notify(RTM_NEWLINK, idev); 2853 2854 /* 2855 * If the changed mtu during down is lower than 2856 * IPV6_MIN_MTU stop IPv6 on this interface. 2857 */ 2858 if (dev->mtu < IPV6_MIN_MTU) 2859 addrconf_ifdown(dev, 1); 2860 } 2861 break; 2862 2863 case NETDEV_CHANGEMTU: 2864 if (idev && dev->mtu >= IPV6_MIN_MTU) { 2865 rt6_mtu_change(dev, dev->mtu); 2866 idev->cnf.mtu6 = dev->mtu; 2867 break; 2868 } 2869 2870 if (!idev && dev->mtu >= IPV6_MIN_MTU) { 2871 idev = ipv6_add_dev(dev); 2872 if (idev) 2873 break; 2874 } 2875 2876 /* 2877 * MTU falled under IPV6_MIN_MTU. 2878 * Stop IPv6 on this interface. 2879 */ 2880 2881 case NETDEV_DOWN: 2882 case NETDEV_UNREGISTER: 2883 /* 2884 * Remove all addresses from this interface. 2885 */ 2886 addrconf_ifdown(dev, event != NETDEV_DOWN); 2887 break; 2888 2889 case NETDEV_CHANGENAME: 2890 if (idev) { 2891 snmp6_unregister_dev(idev); 2892 addrconf_sysctl_unregister(idev); 2893 addrconf_sysctl_register(idev); 2894 err = snmp6_register_dev(idev); 2895 if (err) 2896 return notifier_from_errno(err); 2897 } 2898 break; 2899 2900 case NETDEV_PRE_TYPE_CHANGE: 2901 case NETDEV_POST_TYPE_CHANGE: 2902 addrconf_type_change(dev, event); 2903 break; 2904 } 2905 2906 return NOTIFY_OK; 2907 } 2908 2909 /* 2910 * addrconf module should be notified of a device going up 2911 */ 2912 static struct notifier_block ipv6_dev_notf = { 2913 .notifier_call = addrconf_notify, 2914 }; 2915 2916 static void addrconf_type_change(struct net_device *dev, unsigned long event) 2917 { 2918 struct inet6_dev *idev; 2919 ASSERT_RTNL(); 2920 2921 idev = __in6_dev_get(dev); 2922 2923 if (event == NETDEV_POST_TYPE_CHANGE) 2924 ipv6_mc_remap(idev); 2925 else if (event == NETDEV_PRE_TYPE_CHANGE) 2926 ipv6_mc_unmap(idev); 2927 } 2928 2929 static int addrconf_ifdown(struct net_device *dev, int how) 2930 { 2931 struct net *net = dev_net(dev); 2932 struct inet6_dev *idev; 2933 struct inet6_ifaddr *ifa; 2934 int state, i; 2935 2936 ASSERT_RTNL(); 2937 2938 rt6_ifdown(net, dev); 2939 neigh_ifdown(&nd_tbl, dev); 2940 2941 idev = __in6_dev_get(dev); 2942 if (idev == NULL) 2943 return -ENODEV; 2944 2945 /* 2946 * Step 1: remove reference to ipv6 device from parent device. 2947 * Do not dev_put! 2948 */ 2949 if (how) { 2950 idev->dead = 1; 2951 2952 /* protected by rtnl_lock */ 2953 RCU_INIT_POINTER(dev->ip6_ptr, NULL); 2954 2955 /* Step 1.5: remove snmp6 entry */ 2956 snmp6_unregister_dev(idev); 2957 2958 } 2959 2960 /* Step 2: clear hash table */ 2961 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 2962 struct hlist_head *h = &inet6_addr_lst[i]; 2963 2964 spin_lock_bh(&addrconf_hash_lock); 2965 restart: 2966 hlist_for_each_entry_rcu(ifa, h, addr_lst) { 2967 if (ifa->idev == idev) { 2968 hlist_del_init_rcu(&ifa->addr_lst); 2969 addrconf_del_dad_timer(ifa); 2970 goto restart; 2971 } 2972 } 2973 spin_unlock_bh(&addrconf_hash_lock); 2974 } 2975 2976 write_lock_bh(&idev->lock); 2977 2978 addrconf_del_rs_timer(idev); 2979 2980 /* Step 2: clear flags for stateless addrconf */ 2981 if (!how) 2982 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY); 2983 2984 if (how && del_timer(&idev->regen_timer)) 2985 in6_dev_put(idev); 2986 2987 /* Step 3: clear tempaddr list */ 2988 while (!list_empty(&idev->tempaddr_list)) { 2989 ifa = list_first_entry(&idev->tempaddr_list, 2990 struct inet6_ifaddr, tmp_list); 2991 list_del(&ifa->tmp_list); 2992 write_unlock_bh(&idev->lock); 2993 spin_lock_bh(&ifa->lock); 2994 2995 if (ifa->ifpub) { 2996 in6_ifa_put(ifa->ifpub); 2997 ifa->ifpub = NULL; 2998 } 2999 spin_unlock_bh(&ifa->lock); 3000 in6_ifa_put(ifa); 3001 write_lock_bh(&idev->lock); 3002 } 3003 3004 while (!list_empty(&idev->addr_list)) { 3005 ifa = list_first_entry(&idev->addr_list, 3006 struct inet6_ifaddr, if_list); 3007 addrconf_del_dad_timer(ifa); 3008 3009 list_del(&ifa->if_list); 3010 3011 write_unlock_bh(&idev->lock); 3012 3013 spin_lock_bh(&ifa->state_lock); 3014 state = ifa->state; 3015 ifa->state = INET6_IFADDR_STATE_DEAD; 3016 spin_unlock_bh(&ifa->state_lock); 3017 3018 if (state != INET6_IFADDR_STATE_DEAD) { 3019 __ipv6_ifa_notify(RTM_DELADDR, ifa); 3020 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa); 3021 } 3022 in6_ifa_put(ifa); 3023 3024 write_lock_bh(&idev->lock); 3025 } 3026 3027 write_unlock_bh(&idev->lock); 3028 3029 /* Step 5: Discard multicast list */ 3030 if (how) 3031 ipv6_mc_destroy_dev(idev); 3032 else 3033 ipv6_mc_down(idev); 3034 3035 idev->tstamp = jiffies; 3036 3037 /* Last: Shot the device (if unregistered) */ 3038 if (how) { 3039 addrconf_sysctl_unregister(idev); 3040 neigh_parms_release(&nd_tbl, idev->nd_parms); 3041 neigh_ifdown(&nd_tbl, dev); 3042 in6_dev_put(idev); 3043 } 3044 return 0; 3045 } 3046 3047 static void addrconf_rs_timer(unsigned long data) 3048 { 3049 struct inet6_dev *idev = (struct inet6_dev *)data; 3050 struct net_device *dev = idev->dev; 3051 struct in6_addr lladdr; 3052 3053 write_lock(&idev->lock); 3054 if (idev->dead || !(idev->if_flags & IF_READY)) 3055 goto out; 3056 3057 if (!ipv6_accept_ra(idev)) 3058 goto out; 3059 3060 /* Announcement received after solicitation was sent */ 3061 if (idev->if_flags & IF_RA_RCVD) 3062 goto out; 3063 3064 if (idev->rs_probes++ < idev->cnf.rtr_solicits) { 3065 write_unlock(&idev->lock); 3066 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 3067 ndisc_send_rs(dev, &lladdr, 3068 &in6addr_linklocal_allrouters); 3069 else 3070 goto put; 3071 3072 write_lock(&idev->lock); 3073 /* The wait after the last probe can be shorter */ 3074 addrconf_mod_rs_timer(idev, (idev->rs_probes == 3075 idev->cnf.rtr_solicits) ? 3076 idev->cnf.rtr_solicit_delay : 3077 idev->cnf.rtr_solicit_interval); 3078 } else { 3079 /* 3080 * Note: we do not support deprecated "all on-link" 3081 * assumption any longer. 3082 */ 3083 pr_debug("%s: no IPv6 routers present\n", idev->dev->name); 3084 } 3085 3086 out: 3087 write_unlock(&idev->lock); 3088 put: 3089 in6_dev_put(idev); 3090 } 3091 3092 /* 3093 * Duplicate Address Detection 3094 */ 3095 static void addrconf_dad_kick(struct inet6_ifaddr *ifp) 3096 { 3097 unsigned long rand_num; 3098 struct inet6_dev *idev = ifp->idev; 3099 3100 if (ifp->flags & IFA_F_OPTIMISTIC) 3101 rand_num = 0; 3102 else 3103 rand_num = net_random() % (idev->cnf.rtr_solicit_delay ? : 1); 3104 3105 ifp->dad_probes = idev->cnf.dad_transmits; 3106 addrconf_mod_dad_timer(ifp, rand_num); 3107 } 3108 3109 static void addrconf_dad_start(struct inet6_ifaddr *ifp) 3110 { 3111 struct inet6_dev *idev = ifp->idev; 3112 struct net_device *dev = idev->dev; 3113 3114 addrconf_join_solict(dev, &ifp->addr); 3115 3116 net_srandom(ifp->addr.s6_addr32[3]); 3117 3118 read_lock_bh(&idev->lock); 3119 spin_lock(&ifp->lock); 3120 if (ifp->state == INET6_IFADDR_STATE_DEAD) 3121 goto out; 3122 3123 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) || 3124 idev->cnf.accept_dad < 1 || 3125 !(ifp->flags&IFA_F_TENTATIVE) || 3126 ifp->flags & IFA_F_NODAD) { 3127 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 3128 spin_unlock(&ifp->lock); 3129 read_unlock_bh(&idev->lock); 3130 3131 addrconf_dad_completed(ifp); 3132 return; 3133 } 3134 3135 if (!(idev->if_flags & IF_READY)) { 3136 spin_unlock(&ifp->lock); 3137 read_unlock_bh(&idev->lock); 3138 /* 3139 * If the device is not ready: 3140 * - keep it tentative if it is a permanent address. 3141 * - otherwise, kill it. 3142 */ 3143 in6_ifa_hold(ifp); 3144 addrconf_dad_stop(ifp, 0); 3145 return; 3146 } 3147 3148 /* 3149 * Optimistic nodes can start receiving 3150 * Frames right away 3151 */ 3152 if (ifp->flags & IFA_F_OPTIMISTIC) 3153 ip6_ins_rt(ifp->rt); 3154 3155 addrconf_dad_kick(ifp); 3156 out: 3157 spin_unlock(&ifp->lock); 3158 read_unlock_bh(&idev->lock); 3159 } 3160 3161 static void addrconf_dad_timer(unsigned long data) 3162 { 3163 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *) data; 3164 struct inet6_dev *idev = ifp->idev; 3165 struct in6_addr mcaddr; 3166 3167 if (!ifp->dad_probes && addrconf_dad_end(ifp)) 3168 goto out; 3169 3170 write_lock(&idev->lock); 3171 if (idev->dead || !(idev->if_flags & IF_READY)) { 3172 write_unlock(&idev->lock); 3173 goto out; 3174 } 3175 3176 spin_lock(&ifp->lock); 3177 if (ifp->state == INET6_IFADDR_STATE_DEAD) { 3178 spin_unlock(&ifp->lock); 3179 write_unlock(&idev->lock); 3180 goto out; 3181 } 3182 3183 if (ifp->dad_probes == 0) { 3184 /* 3185 * DAD was successful 3186 */ 3187 3188 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 3189 spin_unlock(&ifp->lock); 3190 write_unlock(&idev->lock); 3191 3192 addrconf_dad_completed(ifp); 3193 3194 goto out; 3195 } 3196 3197 ifp->dad_probes--; 3198 addrconf_mod_dad_timer(ifp, 3199 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME)); 3200 spin_unlock(&ifp->lock); 3201 write_unlock(&idev->lock); 3202 3203 /* send a neighbour solicitation for our addr */ 3204 addrconf_addr_solict_mult(&ifp->addr, &mcaddr); 3205 ndisc_send_ns(ifp->idev->dev, NULL, &ifp->addr, &mcaddr, &in6addr_any); 3206 out: 3207 in6_ifa_put(ifp); 3208 } 3209 3210 static void addrconf_dad_completed(struct inet6_ifaddr *ifp) 3211 { 3212 struct net_device *dev = ifp->idev->dev; 3213 struct in6_addr lladdr; 3214 bool send_rs, send_mld; 3215 3216 addrconf_del_dad_timer(ifp); 3217 3218 /* 3219 * Configure the address for reception. Now it is valid. 3220 */ 3221 3222 ipv6_ifa_notify(RTM_NEWADDR, ifp); 3223 3224 /* If added prefix is link local and we are prepared to process 3225 router advertisements, start sending router solicitations. 3226 */ 3227 3228 read_lock_bh(&ifp->idev->lock); 3229 spin_lock(&ifp->lock); 3230 send_mld = ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL && 3231 ifp->idev->valid_ll_addr_cnt == 1; 3232 send_rs = send_mld && 3233 ipv6_accept_ra(ifp->idev) && 3234 ifp->idev->cnf.rtr_solicits > 0 && 3235 (dev->flags&IFF_LOOPBACK) == 0; 3236 spin_unlock(&ifp->lock); 3237 read_unlock_bh(&ifp->idev->lock); 3238 3239 /* While dad is in progress mld report's source address is in6_addrany. 3240 * Resend with proper ll now. 3241 */ 3242 if (send_mld) 3243 ipv6_mc_dad_complete(ifp->idev); 3244 3245 if (send_rs) { 3246 /* 3247 * If a host as already performed a random delay 3248 * [...] as part of DAD [...] there is no need 3249 * to delay again before sending the first RS 3250 */ 3251 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 3252 return; 3253 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters); 3254 3255 write_lock_bh(&ifp->idev->lock); 3256 spin_lock(&ifp->lock); 3257 ifp->idev->rs_probes = 1; 3258 ifp->idev->if_flags |= IF_RS_SENT; 3259 addrconf_mod_rs_timer(ifp->idev, 3260 ifp->idev->cnf.rtr_solicit_interval); 3261 spin_unlock(&ifp->lock); 3262 write_unlock_bh(&ifp->idev->lock); 3263 } 3264 } 3265 3266 static void addrconf_dad_run(struct inet6_dev *idev) 3267 { 3268 struct inet6_ifaddr *ifp; 3269 3270 read_lock_bh(&idev->lock); 3271 list_for_each_entry(ifp, &idev->addr_list, if_list) { 3272 spin_lock(&ifp->lock); 3273 if (ifp->flags & IFA_F_TENTATIVE && 3274 ifp->state == INET6_IFADDR_STATE_DAD) 3275 addrconf_dad_kick(ifp); 3276 spin_unlock(&ifp->lock); 3277 } 3278 read_unlock_bh(&idev->lock); 3279 } 3280 3281 #ifdef CONFIG_PROC_FS 3282 struct if6_iter_state { 3283 struct seq_net_private p; 3284 int bucket; 3285 int offset; 3286 }; 3287 3288 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos) 3289 { 3290 struct inet6_ifaddr *ifa = NULL; 3291 struct if6_iter_state *state = seq->private; 3292 struct net *net = seq_file_net(seq); 3293 int p = 0; 3294 3295 /* initial bucket if pos is 0 */ 3296 if (pos == 0) { 3297 state->bucket = 0; 3298 state->offset = 0; 3299 } 3300 3301 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) { 3302 hlist_for_each_entry_rcu_bh(ifa, &inet6_addr_lst[state->bucket], 3303 addr_lst) { 3304 if (!net_eq(dev_net(ifa->idev->dev), net)) 3305 continue; 3306 /* sync with offset */ 3307 if (p < state->offset) { 3308 p++; 3309 continue; 3310 } 3311 state->offset++; 3312 return ifa; 3313 } 3314 3315 /* prepare for next bucket */ 3316 state->offset = 0; 3317 p = 0; 3318 } 3319 return NULL; 3320 } 3321 3322 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, 3323 struct inet6_ifaddr *ifa) 3324 { 3325 struct if6_iter_state *state = seq->private; 3326 struct net *net = seq_file_net(seq); 3327 3328 hlist_for_each_entry_continue_rcu_bh(ifa, addr_lst) { 3329 if (!net_eq(dev_net(ifa->idev->dev), net)) 3330 continue; 3331 state->offset++; 3332 return ifa; 3333 } 3334 3335 while (++state->bucket < IN6_ADDR_HSIZE) { 3336 state->offset = 0; 3337 hlist_for_each_entry_rcu_bh(ifa, 3338 &inet6_addr_lst[state->bucket], addr_lst) { 3339 if (!net_eq(dev_net(ifa->idev->dev), net)) 3340 continue; 3341 state->offset++; 3342 return ifa; 3343 } 3344 } 3345 3346 return NULL; 3347 } 3348 3349 static void *if6_seq_start(struct seq_file *seq, loff_t *pos) 3350 __acquires(rcu_bh) 3351 { 3352 rcu_read_lock_bh(); 3353 return if6_get_first(seq, *pos); 3354 } 3355 3356 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 3357 { 3358 struct inet6_ifaddr *ifa; 3359 3360 ifa = if6_get_next(seq, v); 3361 ++*pos; 3362 return ifa; 3363 } 3364 3365 static void if6_seq_stop(struct seq_file *seq, void *v) 3366 __releases(rcu_bh) 3367 { 3368 rcu_read_unlock_bh(); 3369 } 3370 3371 static int if6_seq_show(struct seq_file *seq, void *v) 3372 { 3373 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v; 3374 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n", 3375 &ifp->addr, 3376 ifp->idev->dev->ifindex, 3377 ifp->prefix_len, 3378 ifp->scope, 3379 (u8) ifp->flags, 3380 ifp->idev->dev->name); 3381 return 0; 3382 } 3383 3384 static const struct seq_operations if6_seq_ops = { 3385 .start = if6_seq_start, 3386 .next = if6_seq_next, 3387 .show = if6_seq_show, 3388 .stop = if6_seq_stop, 3389 }; 3390 3391 static int if6_seq_open(struct inode *inode, struct file *file) 3392 { 3393 return seq_open_net(inode, file, &if6_seq_ops, 3394 sizeof(struct if6_iter_state)); 3395 } 3396 3397 static const struct file_operations if6_fops = { 3398 .owner = THIS_MODULE, 3399 .open = if6_seq_open, 3400 .read = seq_read, 3401 .llseek = seq_lseek, 3402 .release = seq_release_net, 3403 }; 3404 3405 static int __net_init if6_proc_net_init(struct net *net) 3406 { 3407 if (!proc_create("if_inet6", S_IRUGO, net->proc_net, &if6_fops)) 3408 return -ENOMEM; 3409 return 0; 3410 } 3411 3412 static void __net_exit if6_proc_net_exit(struct net *net) 3413 { 3414 remove_proc_entry("if_inet6", net->proc_net); 3415 } 3416 3417 static struct pernet_operations if6_proc_net_ops = { 3418 .init = if6_proc_net_init, 3419 .exit = if6_proc_net_exit, 3420 }; 3421 3422 int __init if6_proc_init(void) 3423 { 3424 return register_pernet_subsys(&if6_proc_net_ops); 3425 } 3426 3427 void if6_proc_exit(void) 3428 { 3429 unregister_pernet_subsys(&if6_proc_net_ops); 3430 } 3431 #endif /* CONFIG_PROC_FS */ 3432 3433 #if IS_ENABLED(CONFIG_IPV6_MIP6) 3434 /* Check if address is a home address configured on any interface. */ 3435 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr) 3436 { 3437 int ret = 0; 3438 struct inet6_ifaddr *ifp = NULL; 3439 unsigned int hash = inet6_addr_hash(addr); 3440 3441 rcu_read_lock_bh(); 3442 hlist_for_each_entry_rcu_bh(ifp, &inet6_addr_lst[hash], addr_lst) { 3443 if (!net_eq(dev_net(ifp->idev->dev), net)) 3444 continue; 3445 if (ipv6_addr_equal(&ifp->addr, addr) && 3446 (ifp->flags & IFA_F_HOMEADDRESS)) { 3447 ret = 1; 3448 break; 3449 } 3450 } 3451 rcu_read_unlock_bh(); 3452 return ret; 3453 } 3454 #endif 3455 3456 /* 3457 * Periodic address status verification 3458 */ 3459 3460 static void addrconf_verify(unsigned long foo) 3461 { 3462 unsigned long now, next, next_sec, next_sched; 3463 struct inet6_ifaddr *ifp; 3464 int i; 3465 3466 rcu_read_lock_bh(); 3467 spin_lock(&addrconf_verify_lock); 3468 now = jiffies; 3469 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY); 3470 3471 del_timer(&addr_chk_timer); 3472 3473 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 3474 restart: 3475 hlist_for_each_entry_rcu_bh(ifp, 3476 &inet6_addr_lst[i], addr_lst) { 3477 unsigned long age; 3478 3479 if (ifp->flags & IFA_F_PERMANENT) 3480 continue; 3481 3482 spin_lock(&ifp->lock); 3483 /* We try to batch several events at once. */ 3484 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 3485 3486 if (ifp->valid_lft != INFINITY_LIFE_TIME && 3487 age >= ifp->valid_lft) { 3488 spin_unlock(&ifp->lock); 3489 in6_ifa_hold(ifp); 3490 ipv6_del_addr(ifp); 3491 goto restart; 3492 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) { 3493 spin_unlock(&ifp->lock); 3494 continue; 3495 } else if (age >= ifp->prefered_lft) { 3496 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */ 3497 int deprecate = 0; 3498 3499 if (!(ifp->flags&IFA_F_DEPRECATED)) { 3500 deprecate = 1; 3501 ifp->flags |= IFA_F_DEPRECATED; 3502 } 3503 3504 if (time_before(ifp->tstamp + ifp->valid_lft * HZ, next)) 3505 next = ifp->tstamp + ifp->valid_lft * HZ; 3506 3507 spin_unlock(&ifp->lock); 3508 3509 if (deprecate) { 3510 in6_ifa_hold(ifp); 3511 3512 ipv6_ifa_notify(0, ifp); 3513 in6_ifa_put(ifp); 3514 goto restart; 3515 } 3516 } else if ((ifp->flags&IFA_F_TEMPORARY) && 3517 !(ifp->flags&IFA_F_TENTATIVE)) { 3518 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry * 3519 ifp->idev->cnf.dad_transmits * 3520 NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME) / HZ; 3521 3522 if (age >= ifp->prefered_lft - regen_advance) { 3523 struct inet6_ifaddr *ifpub = ifp->ifpub; 3524 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 3525 next = ifp->tstamp + ifp->prefered_lft * HZ; 3526 if (!ifp->regen_count && ifpub) { 3527 ifp->regen_count++; 3528 in6_ifa_hold(ifp); 3529 in6_ifa_hold(ifpub); 3530 spin_unlock(&ifp->lock); 3531 3532 spin_lock(&ifpub->lock); 3533 ifpub->regen_count = 0; 3534 spin_unlock(&ifpub->lock); 3535 ipv6_create_tempaddr(ifpub, ifp); 3536 in6_ifa_put(ifpub); 3537 in6_ifa_put(ifp); 3538 goto restart; 3539 } 3540 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next)) 3541 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ; 3542 spin_unlock(&ifp->lock); 3543 } else { 3544 /* ifp->prefered_lft <= ifp->valid_lft */ 3545 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 3546 next = ifp->tstamp + ifp->prefered_lft * HZ; 3547 spin_unlock(&ifp->lock); 3548 } 3549 } 3550 } 3551 3552 next_sec = round_jiffies_up(next); 3553 next_sched = next; 3554 3555 /* If rounded timeout is accurate enough, accept it. */ 3556 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ)) 3557 next_sched = next_sec; 3558 3559 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */ 3560 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX)) 3561 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX; 3562 3563 ADBG(KERN_DEBUG "now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n", 3564 now, next, next_sec, next_sched); 3565 3566 addr_chk_timer.expires = next_sched; 3567 add_timer(&addr_chk_timer); 3568 spin_unlock(&addrconf_verify_lock); 3569 rcu_read_unlock_bh(); 3570 } 3571 3572 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local, 3573 struct in6_addr **peer_pfx) 3574 { 3575 struct in6_addr *pfx = NULL; 3576 3577 *peer_pfx = NULL; 3578 3579 if (addr) 3580 pfx = nla_data(addr); 3581 3582 if (local) { 3583 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx))) 3584 *peer_pfx = pfx; 3585 pfx = nla_data(local); 3586 } 3587 3588 return pfx; 3589 } 3590 3591 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = { 3592 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) }, 3593 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) }, 3594 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) }, 3595 [IFA_FLAGS] = { .len = sizeof(u32) }, 3596 }; 3597 3598 static int 3599 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh) 3600 { 3601 struct net *net = sock_net(skb->sk); 3602 struct ifaddrmsg *ifm; 3603 struct nlattr *tb[IFA_MAX+1]; 3604 struct in6_addr *pfx, *peer_pfx; 3605 int err; 3606 3607 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy); 3608 if (err < 0) 3609 return err; 3610 3611 ifm = nlmsg_data(nlh); 3612 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 3613 if (pfx == NULL) 3614 return -EINVAL; 3615 3616 return inet6_addr_del(net, ifm->ifa_index, pfx, ifm->ifa_prefixlen); 3617 } 3618 3619 static int inet6_addr_modify(struct inet6_ifaddr *ifp, u32 ifa_flags, 3620 u32 prefered_lft, u32 valid_lft) 3621 { 3622 u32 flags; 3623 clock_t expires; 3624 unsigned long timeout; 3625 bool was_managetempaddr; 3626 3627 if (!valid_lft || (prefered_lft > valid_lft)) 3628 return -EINVAL; 3629 3630 if (ifa_flags & IFA_F_MANAGETEMPADDR && 3631 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64)) 3632 return -EINVAL; 3633 3634 timeout = addrconf_timeout_fixup(valid_lft, HZ); 3635 if (addrconf_finite_timeout(timeout)) { 3636 expires = jiffies_to_clock_t(timeout * HZ); 3637 valid_lft = timeout; 3638 flags = RTF_EXPIRES; 3639 } else { 3640 expires = 0; 3641 flags = 0; 3642 ifa_flags |= IFA_F_PERMANENT; 3643 } 3644 3645 timeout = addrconf_timeout_fixup(prefered_lft, HZ); 3646 if (addrconf_finite_timeout(timeout)) { 3647 if (timeout == 0) 3648 ifa_flags |= IFA_F_DEPRECATED; 3649 prefered_lft = timeout; 3650 } 3651 3652 spin_lock_bh(&ifp->lock); 3653 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR; 3654 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | 3655 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR); 3656 ifp->flags |= ifa_flags; 3657 ifp->tstamp = jiffies; 3658 ifp->valid_lft = valid_lft; 3659 ifp->prefered_lft = prefered_lft; 3660 3661 spin_unlock_bh(&ifp->lock); 3662 if (!(ifp->flags&IFA_F_TENTATIVE)) 3663 ipv6_ifa_notify(0, ifp); 3664 3665 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, ifp->idev->dev, 3666 expires, flags); 3667 3668 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) { 3669 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR)) 3670 valid_lft = prefered_lft = 0; 3671 manage_tempaddrs(ifp->idev, ifp, valid_lft, prefered_lft, 3672 !was_managetempaddr, jiffies); 3673 } 3674 3675 addrconf_verify(0); 3676 3677 return 0; 3678 } 3679 3680 static int 3681 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh) 3682 { 3683 struct net *net = sock_net(skb->sk); 3684 struct ifaddrmsg *ifm; 3685 struct nlattr *tb[IFA_MAX+1]; 3686 struct in6_addr *pfx, *peer_pfx; 3687 struct inet6_ifaddr *ifa; 3688 struct net_device *dev; 3689 u32 valid_lft = INFINITY_LIFE_TIME, preferred_lft = INFINITY_LIFE_TIME; 3690 u32 ifa_flags; 3691 int err; 3692 3693 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy); 3694 if (err < 0) 3695 return err; 3696 3697 ifm = nlmsg_data(nlh); 3698 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 3699 if (pfx == NULL) 3700 return -EINVAL; 3701 3702 if (tb[IFA_CACHEINFO]) { 3703 struct ifa_cacheinfo *ci; 3704 3705 ci = nla_data(tb[IFA_CACHEINFO]); 3706 valid_lft = ci->ifa_valid; 3707 preferred_lft = ci->ifa_prefered; 3708 } else { 3709 preferred_lft = INFINITY_LIFE_TIME; 3710 valid_lft = INFINITY_LIFE_TIME; 3711 } 3712 3713 dev = __dev_get_by_index(net, ifm->ifa_index); 3714 if (dev == NULL) 3715 return -ENODEV; 3716 3717 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags; 3718 3719 /* We ignore other flags so far. */ 3720 ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR; 3721 3722 ifa = ipv6_get_ifaddr(net, pfx, dev, 1); 3723 if (ifa == NULL) { 3724 /* 3725 * It would be best to check for !NLM_F_CREATE here but 3726 * userspace alreay relies on not having to provide this. 3727 */ 3728 return inet6_addr_add(net, ifm->ifa_index, pfx, peer_pfx, 3729 ifm->ifa_prefixlen, ifa_flags, 3730 preferred_lft, valid_lft); 3731 } 3732 3733 if (nlh->nlmsg_flags & NLM_F_EXCL || 3734 !(nlh->nlmsg_flags & NLM_F_REPLACE)) 3735 err = -EEXIST; 3736 else 3737 err = inet6_addr_modify(ifa, ifa_flags, preferred_lft, valid_lft); 3738 3739 in6_ifa_put(ifa); 3740 3741 return err; 3742 } 3743 3744 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags, 3745 u8 scope, int ifindex) 3746 { 3747 struct ifaddrmsg *ifm; 3748 3749 ifm = nlmsg_data(nlh); 3750 ifm->ifa_family = AF_INET6; 3751 ifm->ifa_prefixlen = prefixlen; 3752 ifm->ifa_flags = flags; 3753 ifm->ifa_scope = scope; 3754 ifm->ifa_index = ifindex; 3755 } 3756 3757 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp, 3758 unsigned long tstamp, u32 preferred, u32 valid) 3759 { 3760 struct ifa_cacheinfo ci; 3761 3762 ci.cstamp = cstamp_delta(cstamp); 3763 ci.tstamp = cstamp_delta(tstamp); 3764 ci.ifa_prefered = preferred; 3765 ci.ifa_valid = valid; 3766 3767 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci); 3768 } 3769 3770 static inline int rt_scope(int ifa_scope) 3771 { 3772 if (ifa_scope & IFA_HOST) 3773 return RT_SCOPE_HOST; 3774 else if (ifa_scope & IFA_LINK) 3775 return RT_SCOPE_LINK; 3776 else if (ifa_scope & IFA_SITE) 3777 return RT_SCOPE_SITE; 3778 else 3779 return RT_SCOPE_UNIVERSE; 3780 } 3781 3782 static inline int inet6_ifaddr_msgsize(void) 3783 { 3784 return NLMSG_ALIGN(sizeof(struct ifaddrmsg)) 3785 + nla_total_size(16) /* IFA_LOCAL */ 3786 + nla_total_size(16) /* IFA_ADDRESS */ 3787 + nla_total_size(sizeof(struct ifa_cacheinfo)) 3788 + nla_total_size(4) /* IFA_FLAGS */; 3789 } 3790 3791 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa, 3792 u32 portid, u32 seq, int event, unsigned int flags) 3793 { 3794 struct nlmsghdr *nlh; 3795 u32 preferred, valid; 3796 3797 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags); 3798 if (nlh == NULL) 3799 return -EMSGSIZE; 3800 3801 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope), 3802 ifa->idev->dev->ifindex); 3803 3804 if (!(ifa->flags&IFA_F_PERMANENT)) { 3805 preferred = ifa->prefered_lft; 3806 valid = ifa->valid_lft; 3807 if (preferred != INFINITY_LIFE_TIME) { 3808 long tval = (jiffies - ifa->tstamp)/HZ; 3809 if (preferred > tval) 3810 preferred -= tval; 3811 else 3812 preferred = 0; 3813 if (valid != INFINITY_LIFE_TIME) { 3814 if (valid > tval) 3815 valid -= tval; 3816 else 3817 valid = 0; 3818 } 3819 } 3820 } else { 3821 preferred = INFINITY_LIFE_TIME; 3822 valid = INFINITY_LIFE_TIME; 3823 } 3824 3825 if (!ipv6_addr_any(&ifa->peer_addr)) { 3826 if (nla_put(skb, IFA_LOCAL, 16, &ifa->addr) < 0 || 3827 nla_put(skb, IFA_ADDRESS, 16, &ifa->peer_addr) < 0) 3828 goto error; 3829 } else 3830 if (nla_put(skb, IFA_ADDRESS, 16, &ifa->addr) < 0) 3831 goto error; 3832 3833 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) 3834 goto error; 3835 3836 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0) 3837 goto error; 3838 3839 return nlmsg_end(skb, nlh); 3840 3841 error: 3842 nlmsg_cancel(skb, nlh); 3843 return -EMSGSIZE; 3844 } 3845 3846 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca, 3847 u32 portid, u32 seq, int event, u16 flags) 3848 { 3849 struct nlmsghdr *nlh; 3850 u8 scope = RT_SCOPE_UNIVERSE; 3851 int ifindex = ifmca->idev->dev->ifindex; 3852 3853 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE) 3854 scope = RT_SCOPE_SITE; 3855 3856 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags); 3857 if (nlh == NULL) 3858 return -EMSGSIZE; 3859 3860 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 3861 if (nla_put(skb, IFA_MULTICAST, 16, &ifmca->mca_addr) < 0 || 3862 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp, 3863 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 3864 nlmsg_cancel(skb, nlh); 3865 return -EMSGSIZE; 3866 } 3867 3868 return nlmsg_end(skb, nlh); 3869 } 3870 3871 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca, 3872 u32 portid, u32 seq, int event, unsigned int flags) 3873 { 3874 struct nlmsghdr *nlh; 3875 u8 scope = RT_SCOPE_UNIVERSE; 3876 int ifindex = ifaca->aca_idev->dev->ifindex; 3877 3878 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE) 3879 scope = RT_SCOPE_SITE; 3880 3881 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct ifaddrmsg), flags); 3882 if (nlh == NULL) 3883 return -EMSGSIZE; 3884 3885 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 3886 if (nla_put(skb, IFA_ANYCAST, 16, &ifaca->aca_addr) < 0 || 3887 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp, 3888 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 3889 nlmsg_cancel(skb, nlh); 3890 return -EMSGSIZE; 3891 } 3892 3893 return nlmsg_end(skb, nlh); 3894 } 3895 3896 enum addr_type_t { 3897 UNICAST_ADDR, 3898 MULTICAST_ADDR, 3899 ANYCAST_ADDR, 3900 }; 3901 3902 /* called with rcu_read_lock() */ 3903 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb, 3904 struct netlink_callback *cb, enum addr_type_t type, 3905 int s_ip_idx, int *p_ip_idx) 3906 { 3907 struct ifmcaddr6 *ifmca; 3908 struct ifacaddr6 *ifaca; 3909 int err = 1; 3910 int ip_idx = *p_ip_idx; 3911 3912 read_lock_bh(&idev->lock); 3913 switch (type) { 3914 case UNICAST_ADDR: { 3915 struct inet6_ifaddr *ifa; 3916 3917 /* unicast address incl. temp addr */ 3918 list_for_each_entry(ifa, &idev->addr_list, if_list) { 3919 if (++ip_idx < s_ip_idx) 3920 continue; 3921 err = inet6_fill_ifaddr(skb, ifa, 3922 NETLINK_CB(cb->skb).portid, 3923 cb->nlh->nlmsg_seq, 3924 RTM_NEWADDR, 3925 NLM_F_MULTI); 3926 if (err <= 0) 3927 break; 3928 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 3929 } 3930 break; 3931 } 3932 case MULTICAST_ADDR: 3933 /* multicast address */ 3934 for (ifmca = idev->mc_list; ifmca; 3935 ifmca = ifmca->next, ip_idx++) { 3936 if (ip_idx < s_ip_idx) 3937 continue; 3938 err = inet6_fill_ifmcaddr(skb, ifmca, 3939 NETLINK_CB(cb->skb).portid, 3940 cb->nlh->nlmsg_seq, 3941 RTM_GETMULTICAST, 3942 NLM_F_MULTI); 3943 if (err <= 0) 3944 break; 3945 } 3946 break; 3947 case ANYCAST_ADDR: 3948 /* anycast address */ 3949 for (ifaca = idev->ac_list; ifaca; 3950 ifaca = ifaca->aca_next, ip_idx++) { 3951 if (ip_idx < s_ip_idx) 3952 continue; 3953 err = inet6_fill_ifacaddr(skb, ifaca, 3954 NETLINK_CB(cb->skb).portid, 3955 cb->nlh->nlmsg_seq, 3956 RTM_GETANYCAST, 3957 NLM_F_MULTI); 3958 if (err <= 0) 3959 break; 3960 } 3961 break; 3962 default: 3963 break; 3964 } 3965 read_unlock_bh(&idev->lock); 3966 *p_ip_idx = ip_idx; 3967 return err; 3968 } 3969 3970 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb, 3971 enum addr_type_t type) 3972 { 3973 struct net *net = sock_net(skb->sk); 3974 int h, s_h; 3975 int idx, ip_idx; 3976 int s_idx, s_ip_idx; 3977 struct net_device *dev; 3978 struct inet6_dev *idev; 3979 struct hlist_head *head; 3980 3981 s_h = cb->args[0]; 3982 s_idx = idx = cb->args[1]; 3983 s_ip_idx = ip_idx = cb->args[2]; 3984 3985 rcu_read_lock(); 3986 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ net->dev_base_seq; 3987 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 3988 idx = 0; 3989 head = &net->dev_index_head[h]; 3990 hlist_for_each_entry_rcu(dev, head, index_hlist) { 3991 if (idx < s_idx) 3992 goto cont; 3993 if (h > s_h || idx > s_idx) 3994 s_ip_idx = 0; 3995 ip_idx = 0; 3996 idev = __in6_dev_get(dev); 3997 if (!idev) 3998 goto cont; 3999 4000 if (in6_dump_addrs(idev, skb, cb, type, 4001 s_ip_idx, &ip_idx) <= 0) 4002 goto done; 4003 cont: 4004 idx++; 4005 } 4006 } 4007 done: 4008 rcu_read_unlock(); 4009 cb->args[0] = h; 4010 cb->args[1] = idx; 4011 cb->args[2] = ip_idx; 4012 4013 return skb->len; 4014 } 4015 4016 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb) 4017 { 4018 enum addr_type_t type = UNICAST_ADDR; 4019 4020 return inet6_dump_addr(skb, cb, type); 4021 } 4022 4023 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb) 4024 { 4025 enum addr_type_t type = MULTICAST_ADDR; 4026 4027 return inet6_dump_addr(skb, cb, type); 4028 } 4029 4030 4031 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb) 4032 { 4033 enum addr_type_t type = ANYCAST_ADDR; 4034 4035 return inet6_dump_addr(skb, cb, type); 4036 } 4037 4038 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh) 4039 { 4040 struct net *net = sock_net(in_skb->sk); 4041 struct ifaddrmsg *ifm; 4042 struct nlattr *tb[IFA_MAX+1]; 4043 struct in6_addr *addr = NULL, *peer; 4044 struct net_device *dev = NULL; 4045 struct inet6_ifaddr *ifa; 4046 struct sk_buff *skb; 4047 int err; 4048 4049 err = nlmsg_parse(nlh, sizeof(*ifm), tb, IFA_MAX, ifa_ipv6_policy); 4050 if (err < 0) 4051 goto errout; 4052 4053 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer); 4054 if (addr == NULL) { 4055 err = -EINVAL; 4056 goto errout; 4057 } 4058 4059 ifm = nlmsg_data(nlh); 4060 if (ifm->ifa_index) 4061 dev = __dev_get_by_index(net, ifm->ifa_index); 4062 4063 ifa = ipv6_get_ifaddr(net, addr, dev, 1); 4064 if (!ifa) { 4065 err = -EADDRNOTAVAIL; 4066 goto errout; 4067 } 4068 4069 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL); 4070 if (!skb) { 4071 err = -ENOBUFS; 4072 goto errout_ifa; 4073 } 4074 4075 err = inet6_fill_ifaddr(skb, ifa, NETLINK_CB(in_skb).portid, 4076 nlh->nlmsg_seq, RTM_NEWADDR, 0); 4077 if (err < 0) { 4078 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 4079 WARN_ON(err == -EMSGSIZE); 4080 kfree_skb(skb); 4081 goto errout_ifa; 4082 } 4083 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 4084 errout_ifa: 4085 in6_ifa_put(ifa); 4086 errout: 4087 return err; 4088 } 4089 4090 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa) 4091 { 4092 struct sk_buff *skb; 4093 struct net *net = dev_net(ifa->idev->dev); 4094 int err = -ENOBUFS; 4095 4096 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC); 4097 if (skb == NULL) 4098 goto errout; 4099 4100 err = inet6_fill_ifaddr(skb, ifa, 0, 0, event, 0); 4101 if (err < 0) { 4102 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 4103 WARN_ON(err == -EMSGSIZE); 4104 kfree_skb(skb); 4105 goto errout; 4106 } 4107 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC); 4108 return; 4109 errout: 4110 if (err < 0) 4111 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err); 4112 } 4113 4114 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf, 4115 __s32 *array, int bytes) 4116 { 4117 BUG_ON(bytes < (DEVCONF_MAX * 4)); 4118 4119 memset(array, 0, bytes); 4120 array[DEVCONF_FORWARDING] = cnf->forwarding; 4121 array[DEVCONF_HOPLIMIT] = cnf->hop_limit; 4122 array[DEVCONF_MTU6] = cnf->mtu6; 4123 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra; 4124 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects; 4125 array[DEVCONF_AUTOCONF] = cnf->autoconf; 4126 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits; 4127 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits; 4128 array[DEVCONF_RTR_SOLICIT_INTERVAL] = 4129 jiffies_to_msecs(cnf->rtr_solicit_interval); 4130 array[DEVCONF_RTR_SOLICIT_DELAY] = 4131 jiffies_to_msecs(cnf->rtr_solicit_delay); 4132 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version; 4133 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] = 4134 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval); 4135 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] = 4136 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval); 4137 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr; 4138 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft; 4139 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft; 4140 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry; 4141 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor; 4142 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses; 4143 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr; 4144 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo; 4145 #ifdef CONFIG_IPV6_ROUTER_PREF 4146 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref; 4147 array[DEVCONF_RTR_PROBE_INTERVAL] = 4148 jiffies_to_msecs(cnf->rtr_probe_interval); 4149 #ifdef CONFIG_IPV6_ROUTE_INFO 4150 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen; 4151 #endif 4152 #endif 4153 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp; 4154 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route; 4155 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 4156 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad; 4157 #endif 4158 #ifdef CONFIG_IPV6_MROUTE 4159 array[DEVCONF_MC_FORWARDING] = cnf->mc_forwarding; 4160 #endif 4161 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6; 4162 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad; 4163 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao; 4164 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify; 4165 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc; 4166 } 4167 4168 static inline size_t inet6_ifla6_size(void) 4169 { 4170 return nla_total_size(4) /* IFLA_INET6_FLAGS */ 4171 + nla_total_size(sizeof(struct ifla_cacheinfo)) 4172 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */ 4173 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */ 4174 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */ 4175 + nla_total_size(sizeof(struct in6_addr)); /* IFLA_INET6_TOKEN */ 4176 } 4177 4178 static inline size_t inet6_if_nlmsg_size(void) 4179 { 4180 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 4181 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 4182 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 4183 + nla_total_size(4) /* IFLA_MTU */ 4184 + nla_total_size(4) /* IFLA_LINK */ 4185 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */ 4186 } 4187 4188 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib, 4189 int items, int bytes) 4190 { 4191 int i; 4192 int pad = bytes - sizeof(u64) * items; 4193 BUG_ON(pad < 0); 4194 4195 /* Use put_unaligned() because stats may not be aligned for u64. */ 4196 put_unaligned(items, &stats[0]); 4197 for (i = 1; i < items; i++) 4198 put_unaligned(atomic_long_read(&mib[i]), &stats[i]); 4199 4200 memset(&stats[items], 0, pad); 4201 } 4202 4203 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu **mib, 4204 int items, int bytes, size_t syncpoff) 4205 { 4206 int i; 4207 int pad = bytes - sizeof(u64) * items; 4208 BUG_ON(pad < 0); 4209 4210 /* Use put_unaligned() because stats may not be aligned for u64. */ 4211 put_unaligned(items, &stats[0]); 4212 for (i = 1; i < items; i++) 4213 put_unaligned(snmp_fold_field64(mib, i, syncpoff), &stats[i]); 4214 4215 memset(&stats[items], 0, pad); 4216 } 4217 4218 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype, 4219 int bytes) 4220 { 4221 switch (attrtype) { 4222 case IFLA_INET6_STATS: 4223 __snmp6_fill_stats64(stats, (void __percpu **)idev->stats.ipv6, 4224 IPSTATS_MIB_MAX, bytes, offsetof(struct ipstats_mib, syncp)); 4225 break; 4226 case IFLA_INET6_ICMP6STATS: 4227 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, ICMP6_MIB_MAX, bytes); 4228 break; 4229 } 4230 } 4231 4232 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev) 4233 { 4234 struct nlattr *nla; 4235 struct ifla_cacheinfo ci; 4236 4237 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags)) 4238 goto nla_put_failure; 4239 ci.max_reasm_len = IPV6_MAXPLEN; 4240 ci.tstamp = cstamp_delta(idev->tstamp); 4241 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time); 4242 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME)); 4243 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci)) 4244 goto nla_put_failure; 4245 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32)); 4246 if (nla == NULL) 4247 goto nla_put_failure; 4248 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla)); 4249 4250 /* XXX - MC not implemented */ 4251 4252 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64)); 4253 if (nla == NULL) 4254 goto nla_put_failure; 4255 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla)); 4256 4257 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64)); 4258 if (nla == NULL) 4259 goto nla_put_failure; 4260 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla)); 4261 4262 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr)); 4263 if (nla == NULL) 4264 goto nla_put_failure; 4265 read_lock_bh(&idev->lock); 4266 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla)); 4267 read_unlock_bh(&idev->lock); 4268 4269 return 0; 4270 4271 nla_put_failure: 4272 return -EMSGSIZE; 4273 } 4274 4275 static size_t inet6_get_link_af_size(const struct net_device *dev) 4276 { 4277 if (!__in6_dev_get(dev)) 4278 return 0; 4279 4280 return inet6_ifla6_size(); 4281 } 4282 4283 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev) 4284 { 4285 struct inet6_dev *idev = __in6_dev_get(dev); 4286 4287 if (!idev) 4288 return -ENODATA; 4289 4290 if (inet6_fill_ifla6_attrs(skb, idev) < 0) 4291 return -EMSGSIZE; 4292 4293 return 0; 4294 } 4295 4296 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token) 4297 { 4298 struct inet6_ifaddr *ifp; 4299 struct net_device *dev = idev->dev; 4300 bool update_rs = false; 4301 struct in6_addr ll_addr; 4302 4303 if (token == NULL) 4304 return -EINVAL; 4305 if (ipv6_addr_any(token)) 4306 return -EINVAL; 4307 if (dev->flags & (IFF_LOOPBACK | IFF_NOARP)) 4308 return -EINVAL; 4309 if (!ipv6_accept_ra(idev)) 4310 return -EINVAL; 4311 if (idev->cnf.rtr_solicits <= 0) 4312 return -EINVAL; 4313 4314 write_lock_bh(&idev->lock); 4315 4316 BUILD_BUG_ON(sizeof(token->s6_addr) != 16); 4317 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8); 4318 4319 write_unlock_bh(&idev->lock); 4320 4321 if (!idev->dead && (idev->if_flags & IF_READY) && 4322 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE | 4323 IFA_F_OPTIMISTIC)) { 4324 4325 /* If we're not ready, then normal ifup will take care 4326 * of this. Otherwise, we need to request our rs here. 4327 */ 4328 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters); 4329 update_rs = true; 4330 } 4331 4332 write_lock_bh(&idev->lock); 4333 4334 if (update_rs) { 4335 idev->if_flags |= IF_RS_SENT; 4336 idev->rs_probes = 1; 4337 addrconf_mod_rs_timer(idev, idev->cnf.rtr_solicit_interval); 4338 } 4339 4340 /* Well, that's kinda nasty ... */ 4341 list_for_each_entry(ifp, &idev->addr_list, if_list) { 4342 spin_lock(&ifp->lock); 4343 if (ifp->tokenized) { 4344 ifp->valid_lft = 0; 4345 ifp->prefered_lft = 0; 4346 } 4347 spin_unlock(&ifp->lock); 4348 } 4349 4350 write_unlock_bh(&idev->lock); 4351 addrconf_verify(0); 4352 return 0; 4353 } 4354 4355 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla) 4356 { 4357 int err = -EINVAL; 4358 struct inet6_dev *idev = __in6_dev_get(dev); 4359 struct nlattr *tb[IFLA_INET6_MAX + 1]; 4360 4361 if (!idev) 4362 return -EAFNOSUPPORT; 4363 4364 if (nla_parse_nested(tb, IFLA_INET6_MAX, nla, NULL) < 0) 4365 BUG(); 4366 4367 if (tb[IFLA_INET6_TOKEN]) 4368 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN])); 4369 4370 return err; 4371 } 4372 4373 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, 4374 u32 portid, u32 seq, int event, unsigned int flags) 4375 { 4376 struct net_device *dev = idev->dev; 4377 struct ifinfomsg *hdr; 4378 struct nlmsghdr *nlh; 4379 void *protoinfo; 4380 4381 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags); 4382 if (nlh == NULL) 4383 return -EMSGSIZE; 4384 4385 hdr = nlmsg_data(nlh); 4386 hdr->ifi_family = AF_INET6; 4387 hdr->__ifi_pad = 0; 4388 hdr->ifi_type = dev->type; 4389 hdr->ifi_index = dev->ifindex; 4390 hdr->ifi_flags = dev_get_flags(dev); 4391 hdr->ifi_change = 0; 4392 4393 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 4394 (dev->addr_len && 4395 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 4396 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 4397 (dev->ifindex != dev->iflink && 4398 nla_put_u32(skb, IFLA_LINK, dev->iflink))) 4399 goto nla_put_failure; 4400 protoinfo = nla_nest_start(skb, IFLA_PROTINFO); 4401 if (protoinfo == NULL) 4402 goto nla_put_failure; 4403 4404 if (inet6_fill_ifla6_attrs(skb, idev) < 0) 4405 goto nla_put_failure; 4406 4407 nla_nest_end(skb, protoinfo); 4408 return nlmsg_end(skb, nlh); 4409 4410 nla_put_failure: 4411 nlmsg_cancel(skb, nlh); 4412 return -EMSGSIZE; 4413 } 4414 4415 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 4416 { 4417 struct net *net = sock_net(skb->sk); 4418 int h, s_h; 4419 int idx = 0, s_idx; 4420 struct net_device *dev; 4421 struct inet6_dev *idev; 4422 struct hlist_head *head; 4423 4424 s_h = cb->args[0]; 4425 s_idx = cb->args[1]; 4426 4427 rcu_read_lock(); 4428 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 4429 idx = 0; 4430 head = &net->dev_index_head[h]; 4431 hlist_for_each_entry_rcu(dev, head, index_hlist) { 4432 if (idx < s_idx) 4433 goto cont; 4434 idev = __in6_dev_get(dev); 4435 if (!idev) 4436 goto cont; 4437 if (inet6_fill_ifinfo(skb, idev, 4438 NETLINK_CB(cb->skb).portid, 4439 cb->nlh->nlmsg_seq, 4440 RTM_NEWLINK, NLM_F_MULTI) <= 0) 4441 goto out; 4442 cont: 4443 idx++; 4444 } 4445 } 4446 out: 4447 rcu_read_unlock(); 4448 cb->args[1] = idx; 4449 cb->args[0] = h; 4450 4451 return skb->len; 4452 } 4453 4454 void inet6_ifinfo_notify(int event, struct inet6_dev *idev) 4455 { 4456 struct sk_buff *skb; 4457 struct net *net = dev_net(idev->dev); 4458 int err = -ENOBUFS; 4459 4460 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC); 4461 if (skb == NULL) 4462 goto errout; 4463 4464 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0); 4465 if (err < 0) { 4466 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */ 4467 WARN_ON(err == -EMSGSIZE); 4468 kfree_skb(skb); 4469 goto errout; 4470 } 4471 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC); 4472 return; 4473 errout: 4474 if (err < 0) 4475 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err); 4476 } 4477 4478 static inline size_t inet6_prefix_nlmsg_size(void) 4479 { 4480 return NLMSG_ALIGN(sizeof(struct prefixmsg)) 4481 + nla_total_size(sizeof(struct in6_addr)) 4482 + nla_total_size(sizeof(struct prefix_cacheinfo)); 4483 } 4484 4485 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev, 4486 struct prefix_info *pinfo, u32 portid, u32 seq, 4487 int event, unsigned int flags) 4488 { 4489 struct prefixmsg *pmsg; 4490 struct nlmsghdr *nlh; 4491 struct prefix_cacheinfo ci; 4492 4493 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags); 4494 if (nlh == NULL) 4495 return -EMSGSIZE; 4496 4497 pmsg = nlmsg_data(nlh); 4498 pmsg->prefix_family = AF_INET6; 4499 pmsg->prefix_pad1 = 0; 4500 pmsg->prefix_pad2 = 0; 4501 pmsg->prefix_ifindex = idev->dev->ifindex; 4502 pmsg->prefix_len = pinfo->prefix_len; 4503 pmsg->prefix_type = pinfo->type; 4504 pmsg->prefix_pad3 = 0; 4505 pmsg->prefix_flags = 0; 4506 if (pinfo->onlink) 4507 pmsg->prefix_flags |= IF_PREFIX_ONLINK; 4508 if (pinfo->autoconf) 4509 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF; 4510 4511 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix)) 4512 goto nla_put_failure; 4513 ci.preferred_time = ntohl(pinfo->prefered); 4514 ci.valid_time = ntohl(pinfo->valid); 4515 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci)) 4516 goto nla_put_failure; 4517 return nlmsg_end(skb, nlh); 4518 4519 nla_put_failure: 4520 nlmsg_cancel(skb, nlh); 4521 return -EMSGSIZE; 4522 } 4523 4524 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 4525 struct prefix_info *pinfo) 4526 { 4527 struct sk_buff *skb; 4528 struct net *net = dev_net(idev->dev); 4529 int err = -ENOBUFS; 4530 4531 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC); 4532 if (skb == NULL) 4533 goto errout; 4534 4535 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0); 4536 if (err < 0) { 4537 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */ 4538 WARN_ON(err == -EMSGSIZE); 4539 kfree_skb(skb); 4540 goto errout; 4541 } 4542 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC); 4543 return; 4544 errout: 4545 if (err < 0) 4546 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err); 4547 } 4548 4549 static void update_valid_ll_addr_cnt(struct inet6_ifaddr *ifp, int count) 4550 { 4551 write_lock_bh(&ifp->idev->lock); 4552 spin_lock(&ifp->lock); 4553 if (((ifp->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE|IFA_F_OPTIMISTIC| 4554 IFA_F_DADFAILED)) == IFA_F_PERMANENT) && 4555 (ipv6_addr_type(&ifp->addr) & IPV6_ADDR_LINKLOCAL)) 4556 ifp->idev->valid_ll_addr_cnt += count; 4557 WARN_ON(ifp->idev->valid_ll_addr_cnt < 0); 4558 spin_unlock(&ifp->lock); 4559 write_unlock_bh(&ifp->idev->lock); 4560 } 4561 4562 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 4563 { 4564 struct net *net = dev_net(ifp->idev->dev); 4565 4566 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp); 4567 4568 switch (event) { 4569 case RTM_NEWADDR: 4570 update_valid_ll_addr_cnt(ifp, 1); 4571 4572 /* 4573 * If the address was optimistic 4574 * we inserted the route at the start of 4575 * our DAD process, so we don't need 4576 * to do it again 4577 */ 4578 if (!(ifp->rt->rt6i_node)) 4579 ip6_ins_rt(ifp->rt); 4580 if (ifp->idev->cnf.forwarding) 4581 addrconf_join_anycast(ifp); 4582 if (!ipv6_addr_any(&ifp->peer_addr)) 4583 addrconf_prefix_route(&ifp->peer_addr, 128, 4584 ifp->idev->dev, 0, 0); 4585 break; 4586 case RTM_DELADDR: 4587 update_valid_ll_addr_cnt(ifp, -1); 4588 4589 if (ifp->idev->cnf.forwarding) 4590 addrconf_leave_anycast(ifp); 4591 addrconf_leave_solict(ifp->idev, &ifp->addr); 4592 if (!ipv6_addr_any(&ifp->peer_addr)) { 4593 struct rt6_info *rt; 4594 struct net_device *dev = ifp->idev->dev; 4595 4596 rt = rt6_lookup(dev_net(dev), &ifp->peer_addr, NULL, 4597 dev->ifindex, 1); 4598 if (rt) { 4599 dst_hold(&rt->dst); 4600 if (ip6_del_rt(rt)) 4601 dst_free(&rt->dst); 4602 } 4603 } 4604 dst_hold(&ifp->rt->dst); 4605 4606 if (ip6_del_rt(ifp->rt)) 4607 dst_free(&ifp->rt->dst); 4608 break; 4609 } 4610 atomic_inc(&net->ipv6.dev_addr_genid); 4611 rt_genid_bump_ipv6(net); 4612 } 4613 4614 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 4615 { 4616 rcu_read_lock_bh(); 4617 if (likely(ifp->idev->dead == 0)) 4618 __ipv6_ifa_notify(event, ifp); 4619 rcu_read_unlock_bh(); 4620 } 4621 4622 #ifdef CONFIG_SYSCTL 4623 4624 static 4625 int addrconf_sysctl_forward(struct ctl_table *ctl, int write, 4626 void __user *buffer, size_t *lenp, loff_t *ppos) 4627 { 4628 int *valp = ctl->data; 4629 int val = *valp; 4630 loff_t pos = *ppos; 4631 struct ctl_table lctl; 4632 int ret; 4633 4634 /* 4635 * ctl->data points to idev->cnf.forwarding, we should 4636 * not modify it until we get the rtnl lock. 4637 */ 4638 lctl = *ctl; 4639 lctl.data = &val; 4640 4641 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 4642 4643 if (write) 4644 ret = addrconf_fixup_forwarding(ctl, valp, val); 4645 if (ret) 4646 *ppos = pos; 4647 return ret; 4648 } 4649 4650 static void dev_disable_change(struct inet6_dev *idev) 4651 { 4652 struct netdev_notifier_info info; 4653 4654 if (!idev || !idev->dev) 4655 return; 4656 4657 netdev_notifier_info_init(&info, idev->dev); 4658 if (idev->cnf.disable_ipv6) 4659 addrconf_notify(NULL, NETDEV_DOWN, &info); 4660 else 4661 addrconf_notify(NULL, NETDEV_UP, &info); 4662 } 4663 4664 static void addrconf_disable_change(struct net *net, __s32 newf) 4665 { 4666 struct net_device *dev; 4667 struct inet6_dev *idev; 4668 4669 rcu_read_lock(); 4670 for_each_netdev_rcu(net, dev) { 4671 idev = __in6_dev_get(dev); 4672 if (idev) { 4673 int changed = (!idev->cnf.disable_ipv6) ^ (!newf); 4674 idev->cnf.disable_ipv6 = newf; 4675 if (changed) 4676 dev_disable_change(idev); 4677 } 4678 } 4679 rcu_read_unlock(); 4680 } 4681 4682 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf) 4683 { 4684 struct net *net; 4685 int old; 4686 4687 if (!rtnl_trylock()) 4688 return restart_syscall(); 4689 4690 net = (struct net *)table->extra2; 4691 old = *p; 4692 *p = newf; 4693 4694 if (p == &net->ipv6.devconf_dflt->disable_ipv6) { 4695 rtnl_unlock(); 4696 return 0; 4697 } 4698 4699 if (p == &net->ipv6.devconf_all->disable_ipv6) { 4700 net->ipv6.devconf_dflt->disable_ipv6 = newf; 4701 addrconf_disable_change(net, newf); 4702 } else if ((!newf) ^ (!old)) 4703 dev_disable_change((struct inet6_dev *)table->extra1); 4704 4705 rtnl_unlock(); 4706 return 0; 4707 } 4708 4709 static 4710 int addrconf_sysctl_disable(struct ctl_table *ctl, int write, 4711 void __user *buffer, size_t *lenp, loff_t *ppos) 4712 { 4713 int *valp = ctl->data; 4714 int val = *valp; 4715 loff_t pos = *ppos; 4716 struct ctl_table lctl; 4717 int ret; 4718 4719 /* 4720 * ctl->data points to idev->cnf.disable_ipv6, we should 4721 * not modify it until we get the rtnl lock. 4722 */ 4723 lctl = *ctl; 4724 lctl.data = &val; 4725 4726 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 4727 4728 if (write) 4729 ret = addrconf_disable_ipv6(ctl, valp, val); 4730 if (ret) 4731 *ppos = pos; 4732 return ret; 4733 } 4734 4735 static 4736 int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write, 4737 void __user *buffer, size_t *lenp, loff_t *ppos) 4738 { 4739 int *valp = ctl->data; 4740 int ret; 4741 int old, new; 4742 4743 old = *valp; 4744 ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 4745 new = *valp; 4746 4747 if (write && old != new) { 4748 struct net *net = ctl->extra2; 4749 4750 if (!rtnl_trylock()) 4751 return restart_syscall(); 4752 4753 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) 4754 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH, 4755 NETCONFA_IFINDEX_DEFAULT, 4756 net->ipv6.devconf_dflt); 4757 else if (valp == &net->ipv6.devconf_all->proxy_ndp) 4758 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH, 4759 NETCONFA_IFINDEX_ALL, 4760 net->ipv6.devconf_all); 4761 else { 4762 struct inet6_dev *idev = ctl->extra1; 4763 4764 inet6_netconf_notify_devconf(net, NETCONFA_PROXY_NEIGH, 4765 idev->dev->ifindex, 4766 &idev->cnf); 4767 } 4768 rtnl_unlock(); 4769 } 4770 4771 return ret; 4772 } 4773 4774 4775 static struct addrconf_sysctl_table 4776 { 4777 struct ctl_table_header *sysctl_header; 4778 struct ctl_table addrconf_vars[DEVCONF_MAX+1]; 4779 } addrconf_sysctl __read_mostly = { 4780 .sysctl_header = NULL, 4781 .addrconf_vars = { 4782 { 4783 .procname = "forwarding", 4784 .data = &ipv6_devconf.forwarding, 4785 .maxlen = sizeof(int), 4786 .mode = 0644, 4787 .proc_handler = addrconf_sysctl_forward, 4788 }, 4789 { 4790 .procname = "hop_limit", 4791 .data = &ipv6_devconf.hop_limit, 4792 .maxlen = sizeof(int), 4793 .mode = 0644, 4794 .proc_handler = proc_dointvec, 4795 }, 4796 { 4797 .procname = "mtu", 4798 .data = &ipv6_devconf.mtu6, 4799 .maxlen = sizeof(int), 4800 .mode = 0644, 4801 .proc_handler = proc_dointvec, 4802 }, 4803 { 4804 .procname = "accept_ra", 4805 .data = &ipv6_devconf.accept_ra, 4806 .maxlen = sizeof(int), 4807 .mode = 0644, 4808 .proc_handler = proc_dointvec, 4809 }, 4810 { 4811 .procname = "accept_redirects", 4812 .data = &ipv6_devconf.accept_redirects, 4813 .maxlen = sizeof(int), 4814 .mode = 0644, 4815 .proc_handler = proc_dointvec, 4816 }, 4817 { 4818 .procname = "autoconf", 4819 .data = &ipv6_devconf.autoconf, 4820 .maxlen = sizeof(int), 4821 .mode = 0644, 4822 .proc_handler = proc_dointvec, 4823 }, 4824 { 4825 .procname = "dad_transmits", 4826 .data = &ipv6_devconf.dad_transmits, 4827 .maxlen = sizeof(int), 4828 .mode = 0644, 4829 .proc_handler = proc_dointvec, 4830 }, 4831 { 4832 .procname = "router_solicitations", 4833 .data = &ipv6_devconf.rtr_solicits, 4834 .maxlen = sizeof(int), 4835 .mode = 0644, 4836 .proc_handler = proc_dointvec, 4837 }, 4838 { 4839 .procname = "router_solicitation_interval", 4840 .data = &ipv6_devconf.rtr_solicit_interval, 4841 .maxlen = sizeof(int), 4842 .mode = 0644, 4843 .proc_handler = proc_dointvec_jiffies, 4844 }, 4845 { 4846 .procname = "router_solicitation_delay", 4847 .data = &ipv6_devconf.rtr_solicit_delay, 4848 .maxlen = sizeof(int), 4849 .mode = 0644, 4850 .proc_handler = proc_dointvec_jiffies, 4851 }, 4852 { 4853 .procname = "force_mld_version", 4854 .data = &ipv6_devconf.force_mld_version, 4855 .maxlen = sizeof(int), 4856 .mode = 0644, 4857 .proc_handler = proc_dointvec, 4858 }, 4859 { 4860 .procname = "mldv1_unsolicited_report_interval", 4861 .data = 4862 &ipv6_devconf.mldv1_unsolicited_report_interval, 4863 .maxlen = sizeof(int), 4864 .mode = 0644, 4865 .proc_handler = proc_dointvec_ms_jiffies, 4866 }, 4867 { 4868 .procname = "mldv2_unsolicited_report_interval", 4869 .data = 4870 &ipv6_devconf.mldv2_unsolicited_report_interval, 4871 .maxlen = sizeof(int), 4872 .mode = 0644, 4873 .proc_handler = proc_dointvec_ms_jiffies, 4874 }, 4875 { 4876 .procname = "use_tempaddr", 4877 .data = &ipv6_devconf.use_tempaddr, 4878 .maxlen = sizeof(int), 4879 .mode = 0644, 4880 .proc_handler = proc_dointvec, 4881 }, 4882 { 4883 .procname = "temp_valid_lft", 4884 .data = &ipv6_devconf.temp_valid_lft, 4885 .maxlen = sizeof(int), 4886 .mode = 0644, 4887 .proc_handler = proc_dointvec, 4888 }, 4889 { 4890 .procname = "temp_prefered_lft", 4891 .data = &ipv6_devconf.temp_prefered_lft, 4892 .maxlen = sizeof(int), 4893 .mode = 0644, 4894 .proc_handler = proc_dointvec, 4895 }, 4896 { 4897 .procname = "regen_max_retry", 4898 .data = &ipv6_devconf.regen_max_retry, 4899 .maxlen = sizeof(int), 4900 .mode = 0644, 4901 .proc_handler = proc_dointvec, 4902 }, 4903 { 4904 .procname = "max_desync_factor", 4905 .data = &ipv6_devconf.max_desync_factor, 4906 .maxlen = sizeof(int), 4907 .mode = 0644, 4908 .proc_handler = proc_dointvec, 4909 }, 4910 { 4911 .procname = "max_addresses", 4912 .data = &ipv6_devconf.max_addresses, 4913 .maxlen = sizeof(int), 4914 .mode = 0644, 4915 .proc_handler = proc_dointvec, 4916 }, 4917 { 4918 .procname = "accept_ra_defrtr", 4919 .data = &ipv6_devconf.accept_ra_defrtr, 4920 .maxlen = sizeof(int), 4921 .mode = 0644, 4922 .proc_handler = proc_dointvec, 4923 }, 4924 { 4925 .procname = "accept_ra_pinfo", 4926 .data = &ipv6_devconf.accept_ra_pinfo, 4927 .maxlen = sizeof(int), 4928 .mode = 0644, 4929 .proc_handler = proc_dointvec, 4930 }, 4931 #ifdef CONFIG_IPV6_ROUTER_PREF 4932 { 4933 .procname = "accept_ra_rtr_pref", 4934 .data = &ipv6_devconf.accept_ra_rtr_pref, 4935 .maxlen = sizeof(int), 4936 .mode = 0644, 4937 .proc_handler = proc_dointvec, 4938 }, 4939 { 4940 .procname = "router_probe_interval", 4941 .data = &ipv6_devconf.rtr_probe_interval, 4942 .maxlen = sizeof(int), 4943 .mode = 0644, 4944 .proc_handler = proc_dointvec_jiffies, 4945 }, 4946 #ifdef CONFIG_IPV6_ROUTE_INFO 4947 { 4948 .procname = "accept_ra_rt_info_max_plen", 4949 .data = &ipv6_devconf.accept_ra_rt_info_max_plen, 4950 .maxlen = sizeof(int), 4951 .mode = 0644, 4952 .proc_handler = proc_dointvec, 4953 }, 4954 #endif 4955 #endif 4956 { 4957 .procname = "proxy_ndp", 4958 .data = &ipv6_devconf.proxy_ndp, 4959 .maxlen = sizeof(int), 4960 .mode = 0644, 4961 .proc_handler = addrconf_sysctl_proxy_ndp, 4962 }, 4963 { 4964 .procname = "accept_source_route", 4965 .data = &ipv6_devconf.accept_source_route, 4966 .maxlen = sizeof(int), 4967 .mode = 0644, 4968 .proc_handler = proc_dointvec, 4969 }, 4970 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 4971 { 4972 .procname = "optimistic_dad", 4973 .data = &ipv6_devconf.optimistic_dad, 4974 .maxlen = sizeof(int), 4975 .mode = 0644, 4976 .proc_handler = proc_dointvec, 4977 4978 }, 4979 #endif 4980 #ifdef CONFIG_IPV6_MROUTE 4981 { 4982 .procname = "mc_forwarding", 4983 .data = &ipv6_devconf.mc_forwarding, 4984 .maxlen = sizeof(int), 4985 .mode = 0444, 4986 .proc_handler = proc_dointvec, 4987 }, 4988 #endif 4989 { 4990 .procname = "disable_ipv6", 4991 .data = &ipv6_devconf.disable_ipv6, 4992 .maxlen = sizeof(int), 4993 .mode = 0644, 4994 .proc_handler = addrconf_sysctl_disable, 4995 }, 4996 { 4997 .procname = "accept_dad", 4998 .data = &ipv6_devconf.accept_dad, 4999 .maxlen = sizeof(int), 5000 .mode = 0644, 5001 .proc_handler = proc_dointvec, 5002 }, 5003 { 5004 .procname = "force_tllao", 5005 .data = &ipv6_devconf.force_tllao, 5006 .maxlen = sizeof(int), 5007 .mode = 0644, 5008 .proc_handler = proc_dointvec 5009 }, 5010 { 5011 .procname = "ndisc_notify", 5012 .data = &ipv6_devconf.ndisc_notify, 5013 .maxlen = sizeof(int), 5014 .mode = 0644, 5015 .proc_handler = proc_dointvec 5016 }, 5017 { 5018 .procname = "suppress_frag_ndisc", 5019 .data = &ipv6_devconf.suppress_frag_ndisc, 5020 .maxlen = sizeof(int), 5021 .mode = 0644, 5022 .proc_handler = proc_dointvec 5023 }, 5024 { 5025 /* sentinel */ 5026 } 5027 }, 5028 }; 5029 5030 static int __addrconf_sysctl_register(struct net *net, char *dev_name, 5031 struct inet6_dev *idev, struct ipv6_devconf *p) 5032 { 5033 int i; 5034 struct addrconf_sysctl_table *t; 5035 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ]; 5036 5037 t = kmemdup(&addrconf_sysctl, sizeof(*t), GFP_KERNEL); 5038 if (t == NULL) 5039 goto out; 5040 5041 for (i = 0; t->addrconf_vars[i].data; i++) { 5042 t->addrconf_vars[i].data += (char *)p - (char *)&ipv6_devconf; 5043 t->addrconf_vars[i].extra1 = idev; /* embedded; no ref */ 5044 t->addrconf_vars[i].extra2 = net; 5045 } 5046 5047 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name); 5048 5049 t->sysctl_header = register_net_sysctl(net, path, t->addrconf_vars); 5050 if (t->sysctl_header == NULL) 5051 goto free; 5052 5053 p->sysctl = t; 5054 return 0; 5055 5056 free: 5057 kfree(t); 5058 out: 5059 return -ENOBUFS; 5060 } 5061 5062 static void __addrconf_sysctl_unregister(struct ipv6_devconf *p) 5063 { 5064 struct addrconf_sysctl_table *t; 5065 5066 if (p->sysctl == NULL) 5067 return; 5068 5069 t = p->sysctl; 5070 p->sysctl = NULL; 5071 unregister_net_sysctl_table(t->sysctl_header); 5072 kfree(t); 5073 } 5074 5075 static void addrconf_sysctl_register(struct inet6_dev *idev) 5076 { 5077 neigh_sysctl_register(idev->dev, idev->nd_parms, 5078 &ndisc_ifinfo_sysctl_change); 5079 __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name, 5080 idev, &idev->cnf); 5081 } 5082 5083 static void addrconf_sysctl_unregister(struct inet6_dev *idev) 5084 { 5085 __addrconf_sysctl_unregister(&idev->cnf); 5086 neigh_sysctl_unregister(idev->nd_parms); 5087 } 5088 5089 5090 #endif 5091 5092 static int __net_init addrconf_init_net(struct net *net) 5093 { 5094 int err = -ENOMEM; 5095 struct ipv6_devconf *all, *dflt; 5096 5097 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL); 5098 if (all == NULL) 5099 goto err_alloc_all; 5100 5101 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL); 5102 if (dflt == NULL) 5103 goto err_alloc_dflt; 5104 5105 /* these will be inherited by all namespaces */ 5106 dflt->autoconf = ipv6_defaults.autoconf; 5107 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6; 5108 5109 net->ipv6.devconf_all = all; 5110 net->ipv6.devconf_dflt = dflt; 5111 5112 #ifdef CONFIG_SYSCTL 5113 err = __addrconf_sysctl_register(net, "all", NULL, all); 5114 if (err < 0) 5115 goto err_reg_all; 5116 5117 err = __addrconf_sysctl_register(net, "default", NULL, dflt); 5118 if (err < 0) 5119 goto err_reg_dflt; 5120 #endif 5121 return 0; 5122 5123 #ifdef CONFIG_SYSCTL 5124 err_reg_dflt: 5125 __addrconf_sysctl_unregister(all); 5126 err_reg_all: 5127 kfree(dflt); 5128 #endif 5129 err_alloc_dflt: 5130 kfree(all); 5131 err_alloc_all: 5132 return err; 5133 } 5134 5135 static void __net_exit addrconf_exit_net(struct net *net) 5136 { 5137 #ifdef CONFIG_SYSCTL 5138 __addrconf_sysctl_unregister(net->ipv6.devconf_dflt); 5139 __addrconf_sysctl_unregister(net->ipv6.devconf_all); 5140 #endif 5141 if (!net_eq(net, &init_net)) { 5142 kfree(net->ipv6.devconf_dflt); 5143 kfree(net->ipv6.devconf_all); 5144 } 5145 } 5146 5147 static struct pernet_operations addrconf_ops = { 5148 .init = addrconf_init_net, 5149 .exit = addrconf_exit_net, 5150 }; 5151 5152 static struct rtnl_af_ops inet6_ops = { 5153 .family = AF_INET6, 5154 .fill_link_af = inet6_fill_link_af, 5155 .get_link_af_size = inet6_get_link_af_size, 5156 .set_link_af = inet6_set_link_af, 5157 }; 5158 5159 /* 5160 * Init / cleanup code 5161 */ 5162 5163 int __init addrconf_init(void) 5164 { 5165 int i, err; 5166 5167 err = ipv6_addr_label_init(); 5168 if (err < 0) { 5169 pr_crit("%s: cannot initialize default policy table: %d\n", 5170 __func__, err); 5171 goto out; 5172 } 5173 5174 err = register_pernet_subsys(&addrconf_ops); 5175 if (err < 0) 5176 goto out_addrlabel; 5177 5178 /* The addrconf netdev notifier requires that loopback_dev 5179 * has it's ipv6 private information allocated and setup 5180 * before it can bring up and give link-local addresses 5181 * to other devices which are up. 5182 * 5183 * Unfortunately, loopback_dev is not necessarily the first 5184 * entry in the global dev_base list of net devices. In fact, 5185 * it is likely to be the very last entry on that list. 5186 * So this causes the notifier registry below to try and 5187 * give link-local addresses to all devices besides loopback_dev 5188 * first, then loopback_dev, which cases all the non-loopback_dev 5189 * devices to fail to get a link-local address. 5190 * 5191 * So, as a temporary fix, allocate the ipv6 structure for 5192 * loopback_dev first by hand. 5193 * Longer term, all of the dependencies ipv6 has upon the loopback 5194 * device and it being up should be removed. 5195 */ 5196 rtnl_lock(); 5197 if (!ipv6_add_dev(init_net.loopback_dev)) 5198 err = -ENOMEM; 5199 rtnl_unlock(); 5200 if (err) 5201 goto errlo; 5202 5203 for (i = 0; i < IN6_ADDR_HSIZE; i++) 5204 INIT_HLIST_HEAD(&inet6_addr_lst[i]); 5205 5206 register_netdevice_notifier(&ipv6_dev_notf); 5207 5208 addrconf_verify(0); 5209 5210 err = rtnl_af_register(&inet6_ops); 5211 if (err < 0) 5212 goto errout_af; 5213 5214 err = __rtnl_register(PF_INET6, RTM_GETLINK, NULL, inet6_dump_ifinfo, 5215 NULL); 5216 if (err < 0) 5217 goto errout; 5218 5219 /* Only the first call to __rtnl_register can fail */ 5220 __rtnl_register(PF_INET6, RTM_NEWADDR, inet6_rtm_newaddr, NULL, NULL); 5221 __rtnl_register(PF_INET6, RTM_DELADDR, inet6_rtm_deladdr, NULL, NULL); 5222 __rtnl_register(PF_INET6, RTM_GETADDR, inet6_rtm_getaddr, 5223 inet6_dump_ifaddr, NULL); 5224 __rtnl_register(PF_INET6, RTM_GETMULTICAST, NULL, 5225 inet6_dump_ifmcaddr, NULL); 5226 __rtnl_register(PF_INET6, RTM_GETANYCAST, NULL, 5227 inet6_dump_ifacaddr, NULL); 5228 __rtnl_register(PF_INET6, RTM_GETNETCONF, inet6_netconf_get_devconf, 5229 inet6_netconf_dump_devconf, NULL); 5230 5231 ipv6_addr_label_rtnl_register(); 5232 5233 return 0; 5234 errout: 5235 rtnl_af_unregister(&inet6_ops); 5236 errout_af: 5237 unregister_netdevice_notifier(&ipv6_dev_notf); 5238 errlo: 5239 unregister_pernet_subsys(&addrconf_ops); 5240 out_addrlabel: 5241 ipv6_addr_label_cleanup(); 5242 out: 5243 return err; 5244 } 5245 5246 void addrconf_cleanup(void) 5247 { 5248 struct net_device *dev; 5249 int i; 5250 5251 unregister_netdevice_notifier(&ipv6_dev_notf); 5252 unregister_pernet_subsys(&addrconf_ops); 5253 ipv6_addr_label_cleanup(); 5254 5255 rtnl_lock(); 5256 5257 __rtnl_af_unregister(&inet6_ops); 5258 5259 /* clean dev list */ 5260 for_each_netdev(&init_net, dev) { 5261 if (__in6_dev_get(dev) == NULL) 5262 continue; 5263 addrconf_ifdown(dev, 1); 5264 } 5265 addrconf_ifdown(init_net.loopback_dev, 2); 5266 5267 /* 5268 * Check hash table. 5269 */ 5270 spin_lock_bh(&addrconf_hash_lock); 5271 for (i = 0; i < IN6_ADDR_HSIZE; i++) 5272 WARN_ON(!hlist_empty(&inet6_addr_lst[i])); 5273 spin_unlock_bh(&addrconf_hash_lock); 5274 5275 del_timer(&addr_chk_timer); 5276 rtnl_unlock(); 5277 } 5278