1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 Address [auto]configuration 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 9 */ 10 11 /* 12 * Changes: 13 * 14 * Janos Farkas : delete timer on ifdown 15 * <chexum@bankinf.banki.hu> 16 * Andi Kleen : kill double kfree on module 17 * unload. 18 * Maciej W. Rozycki : FDDI support 19 * sekiya@USAGI : Don't send too many RS 20 * packets. 21 * yoshfuji@USAGI : Fixed interval between DAD 22 * packets. 23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of 24 * address validation timer. 25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041) 26 * support. 27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6 28 * address on a same interface. 29 * YOSHIFUJI Hideaki @USAGI : ARCnet support 30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to 31 * seq_file. 32 * YOSHIFUJI Hideaki @USAGI : improved source address 33 * selection; consider scope, 34 * status etc. 35 */ 36 37 #define pr_fmt(fmt) "IPv6: " fmt 38 39 #include <linux/errno.h> 40 #include <linux/types.h> 41 #include <linux/kernel.h> 42 #include <linux/sched/signal.h> 43 #include <linux/socket.h> 44 #include <linux/sockios.h> 45 #include <linux/net.h> 46 #include <linux/inet.h> 47 #include <linux/in6.h> 48 #include <linux/netdevice.h> 49 #include <linux/if_addr.h> 50 #include <linux/if_arp.h> 51 #include <linux/if_arcnet.h> 52 #include <linux/if_infiniband.h> 53 #include <linux/route.h> 54 #include <linux/inetdevice.h> 55 #include <linux/init.h> 56 #include <linux/slab.h> 57 #ifdef CONFIG_SYSCTL 58 #include <linux/sysctl.h> 59 #endif 60 #include <linux/capability.h> 61 #include <linux/delay.h> 62 #include <linux/notifier.h> 63 #include <linux/string.h> 64 #include <linux/hash.h> 65 66 #include <net/ip_tunnels.h> 67 #include <net/net_namespace.h> 68 #include <net/sock.h> 69 #include <net/snmp.h> 70 71 #include <net/6lowpan.h> 72 #include <net/firewire.h> 73 #include <net/ipv6.h> 74 #include <net/protocol.h> 75 #include <net/ndisc.h> 76 #include <net/ip6_route.h> 77 #include <net/addrconf.h> 78 #include <net/tcp.h> 79 #include <net/ip.h> 80 #include <net/netlink.h> 81 #include <net/pkt_sched.h> 82 #include <net/l3mdev.h> 83 #include <net/netdev_lock.h> 84 #include <linux/if_tunnel.h> 85 #include <linux/rtnetlink.h> 86 #include <linux/netconf.h> 87 #include <linux/random.h> 88 #include <linux/uaccess.h> 89 #include <linux/unaligned.h> 90 91 #include <linux/proc_fs.h> 92 #include <linux/seq_file.h> 93 #include <linux/export.h> 94 #include <linux/ioam6.h> 95 96 #define IPV6_MAX_STRLEN \ 97 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255") 98 99 static inline u32 cstamp_delta(unsigned long cstamp) 100 { 101 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ; 102 } 103 104 static inline s32 rfc3315_s14_backoff_init(s32 irt) 105 { 106 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */ 107 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt; 108 do_div(tmp, 1000000); 109 return (s32)tmp; 110 } 111 112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt) 113 { 114 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */ 115 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt; 116 do_div(tmp, 1000000); 117 if ((s32)tmp > mrt) { 118 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */ 119 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt; 120 do_div(tmp, 1000000); 121 } 122 return (s32)tmp; 123 } 124 125 #ifdef CONFIG_SYSCTL 126 static int addrconf_sysctl_register(struct inet6_dev *idev); 127 static void addrconf_sysctl_unregister(struct inet6_dev *idev); 128 #else 129 static inline int addrconf_sysctl_register(struct inet6_dev *idev) 130 { 131 return 0; 132 } 133 134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev) 135 { 136 } 137 #endif 138 139 static void ipv6_gen_rnd_iid(struct in6_addr *addr); 140 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev); 142 static int ipv6_count_addresses(const struct inet6_dev *idev); 143 static int ipv6_generate_stable_address(struct in6_addr *addr, 144 u8 dad_count, 145 const struct inet6_dev *idev); 146 147 #define IN6_ADDR_HSIZE_SHIFT 8 148 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT) 149 150 static void addrconf_verify(struct net *net); 151 static void addrconf_verify_rtnl(struct net *net); 152 153 static struct workqueue_struct *addrconf_wq; 154 155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp); 156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp); 157 158 static void addrconf_type_change(struct net_device *dev, 159 unsigned long event); 160 static int addrconf_ifdown(struct net_device *dev, bool unregister); 161 162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 163 int plen, 164 const struct net_device *dev, 165 u32 flags, u32 noflags, 166 bool no_gw); 167 168 static void addrconf_dad_start(struct inet6_ifaddr *ifp); 169 static void addrconf_dad_work(struct work_struct *w); 170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 171 bool send_na); 172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart); 173 static void addrconf_rs_timer(struct timer_list *t); 174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 176 177 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 178 struct prefix_info *pinfo); 179 180 static struct ipv6_devconf ipv6_devconf __read_mostly = { 181 .forwarding = 0, 182 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 183 .mtu6 = IPV6_MIN_MTU, 184 .accept_ra = 1, 185 .accept_redirects = 1, 186 .autoconf = 1, 187 .force_mld_version = 0, 188 .mldv1_unsolicited_report_interval = 10 * HZ, 189 .mldv2_unsolicited_report_interval = HZ, 190 .dad_transmits = 1, 191 .rtr_solicits = MAX_RTR_SOLICITATIONS, 192 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 193 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 194 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 195 .use_tempaddr = 0, 196 .temp_valid_lft = TEMP_VALID_LIFETIME, 197 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 198 .regen_min_advance = REGEN_MIN_ADVANCE, 199 .regen_max_retry = REGEN_MAX_RETRY, 200 .max_desync_factor = MAX_DESYNC_FACTOR, 201 .max_addresses = IPV6_MAX_ADDRESSES, 202 .accept_ra_defrtr = 1, 203 .ra_defrtr_metric = IP6_RT_PRIO_USER, 204 .accept_ra_from_local = 0, 205 .accept_ra_min_hop_limit= 1, 206 .accept_ra_min_lft = 0, 207 .accept_ra_pinfo = 1, 208 #ifdef CONFIG_IPV6_ROUTER_PREF 209 .accept_ra_rtr_pref = 1, 210 .rtr_probe_interval = 60 * HZ, 211 #ifdef CONFIG_IPV6_ROUTE_INFO 212 .accept_ra_rt_info_min_plen = 0, 213 .accept_ra_rt_info_max_plen = 0, 214 #endif 215 #endif 216 .proxy_ndp = 0, 217 .accept_source_route = 0, /* we do not accept RH0 by default. */ 218 .disable_ipv6 = 0, 219 .accept_dad = 0, 220 .suppress_frag_ndisc = 1, 221 .accept_ra_mtu = 1, 222 .stable_secret = { 223 .initialized = false, 224 }, 225 .use_oif_addrs_only = 0, 226 .ignore_routes_with_linkdown = 0, 227 .keep_addr_on_down = 0, 228 .seg6_enabled = 0, 229 #ifdef CONFIG_IPV6_SEG6_HMAC 230 .seg6_require_hmac = 0, 231 #endif 232 .enhanced_dad = 1, 233 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 234 .disable_policy = 0, 235 .rpl_seg_enabled = 0, 236 .ioam6_enabled = 0, 237 .ioam6_id = IOAM6_DEFAULT_IF_ID, 238 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 239 .ndisc_evict_nocarrier = 1, 240 .ra_honor_pio_life = 0, 241 .ra_honor_pio_pflag = 0, 242 }; 243 244 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = { 245 .forwarding = 0, 246 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 247 .mtu6 = IPV6_MIN_MTU, 248 .accept_ra = 1, 249 .accept_redirects = 1, 250 .autoconf = 1, 251 .force_mld_version = 0, 252 .mldv1_unsolicited_report_interval = 10 * HZ, 253 .mldv2_unsolicited_report_interval = HZ, 254 .dad_transmits = 1, 255 .rtr_solicits = MAX_RTR_SOLICITATIONS, 256 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 257 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 258 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 259 .use_tempaddr = 0, 260 .temp_valid_lft = TEMP_VALID_LIFETIME, 261 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 262 .regen_min_advance = REGEN_MIN_ADVANCE, 263 .regen_max_retry = REGEN_MAX_RETRY, 264 .max_desync_factor = MAX_DESYNC_FACTOR, 265 .max_addresses = IPV6_MAX_ADDRESSES, 266 .accept_ra_defrtr = 1, 267 .ra_defrtr_metric = IP6_RT_PRIO_USER, 268 .accept_ra_from_local = 0, 269 .accept_ra_min_hop_limit= 1, 270 .accept_ra_min_lft = 0, 271 .accept_ra_pinfo = 1, 272 #ifdef CONFIG_IPV6_ROUTER_PREF 273 .accept_ra_rtr_pref = 1, 274 .rtr_probe_interval = 60 * HZ, 275 #ifdef CONFIG_IPV6_ROUTE_INFO 276 .accept_ra_rt_info_min_plen = 0, 277 .accept_ra_rt_info_max_plen = 0, 278 #endif 279 #endif 280 .proxy_ndp = 0, 281 .accept_source_route = 0, /* we do not accept RH0 by default. */ 282 .disable_ipv6 = 0, 283 .accept_dad = 1, 284 .suppress_frag_ndisc = 1, 285 .accept_ra_mtu = 1, 286 .stable_secret = { 287 .initialized = false, 288 }, 289 .use_oif_addrs_only = 0, 290 .ignore_routes_with_linkdown = 0, 291 .keep_addr_on_down = 0, 292 .seg6_enabled = 0, 293 #ifdef CONFIG_IPV6_SEG6_HMAC 294 .seg6_require_hmac = 0, 295 #endif 296 .enhanced_dad = 1, 297 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 298 .disable_policy = 0, 299 .rpl_seg_enabled = 0, 300 .ioam6_enabled = 0, 301 .ioam6_id = IOAM6_DEFAULT_IF_ID, 302 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 303 .ndisc_evict_nocarrier = 1, 304 .ra_honor_pio_life = 0, 305 .ra_honor_pio_pflag = 0, 306 }; 307 308 /* Check if link is ready: is it up and is a valid qdisc available */ 309 static inline bool addrconf_link_ready(const struct net_device *dev) 310 { 311 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev); 312 } 313 314 static void addrconf_del_rs_timer(struct inet6_dev *idev) 315 { 316 if (timer_delete(&idev->rs_timer)) 317 __in6_dev_put(idev); 318 } 319 320 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp) 321 { 322 if (cancel_delayed_work(&ifp->dad_work)) 323 __in6_ifa_put(ifp); 324 } 325 326 static void addrconf_mod_rs_timer(struct inet6_dev *idev, 327 unsigned long when) 328 { 329 if (!mod_timer(&idev->rs_timer, jiffies + when)) 330 in6_dev_hold(idev); 331 } 332 333 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp, 334 unsigned long delay) 335 { 336 in6_ifa_hold(ifp); 337 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay)) 338 in6_ifa_put(ifp); 339 } 340 341 static int snmp6_alloc_dev(struct inet6_dev *idev) 342 { 343 int i; 344 345 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT); 346 if (!idev->stats.ipv6) 347 goto err_ip; 348 349 for_each_possible_cpu(i) { 350 struct ipstats_mib *addrconf_stats; 351 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i); 352 u64_stats_init(&addrconf_stats->syncp); 353 } 354 355 356 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device), 357 GFP_KERNEL); 358 if (!idev->stats.icmpv6dev) 359 goto err_icmp; 360 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device), 361 GFP_KERNEL_ACCOUNT); 362 if (!idev->stats.icmpv6msgdev) 363 goto err_icmpmsg; 364 365 return 0; 366 367 err_icmpmsg: 368 kfree(idev->stats.icmpv6dev); 369 err_icmp: 370 free_percpu(idev->stats.ipv6); 371 err_ip: 372 return -ENOMEM; 373 } 374 375 static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 376 { 377 struct inet6_dev *ndev; 378 int err = -ENOMEM; 379 380 ASSERT_RTNL(); 381 netdev_ops_assert_locked(dev); 382 383 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev) 384 return ERR_PTR(-EINVAL); 385 386 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT); 387 if (!ndev) 388 return ERR_PTR(err); 389 390 rwlock_init(&ndev->lock); 391 ndev->dev = dev; 392 INIT_LIST_HEAD(&ndev->addr_list); 393 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0); 394 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf)); 395 396 if (ndev->cnf.stable_secret.initialized) 397 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY; 398 399 ndev->cnf.mtu6 = dev->mtu; 400 ndev->ra_mtu = 0; 401 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl); 402 if (!ndev->nd_parms) { 403 kfree(ndev); 404 return ERR_PTR(err); 405 } 406 if (ndev->cnf.forwarding) 407 netif_disable_lro(dev); 408 /* We refer to the device */ 409 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL); 410 411 if (snmp6_alloc_dev(ndev) < 0) { 412 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n", 413 __func__); 414 neigh_parms_release(&nd_tbl, ndev->nd_parms); 415 netdev_put(dev, &ndev->dev_tracker); 416 kfree(ndev); 417 return ERR_PTR(err); 418 } 419 420 if (dev != blackhole_netdev) { 421 if (snmp6_register_dev(ndev) < 0) { 422 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n", 423 __func__, dev->name); 424 goto err_release; 425 } 426 } 427 /* One reference from device. */ 428 refcount_set(&ndev->refcnt, 1); 429 430 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK)) 431 ndev->cnf.accept_dad = -1; 432 433 #if IS_ENABLED(CONFIG_IPV6_SIT) 434 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) { 435 pr_info("%s: Disabled Multicast RS\n", dev->name); 436 ndev->cnf.rtr_solicits = 0; 437 } 438 #endif 439 440 INIT_LIST_HEAD(&ndev->tempaddr_list); 441 ndev->desync_factor = U32_MAX; 442 if ((dev->flags&IFF_LOOPBACK) || 443 dev->type == ARPHRD_TUNNEL || 444 dev->type == ARPHRD_TUNNEL6 || 445 dev->type == ARPHRD_SIT || 446 dev->type == ARPHRD_NONE) { 447 ndev->cnf.use_tempaddr = -1; 448 } 449 450 ndev->token = in6addr_any; 451 452 if (netif_running(dev) && addrconf_link_ready(dev)) 453 ndev->if_flags |= IF_READY; 454 455 ipv6_mc_init_dev(ndev); 456 ndev->tstamp = jiffies; 457 if (dev != blackhole_netdev) { 458 err = addrconf_sysctl_register(ndev); 459 if (err) { 460 ipv6_mc_destroy_dev(ndev); 461 snmp6_unregister_dev(ndev); 462 goto err_release; 463 } 464 } 465 /* protected by rtnl_lock */ 466 rcu_assign_pointer(dev->ip6_ptr, ndev); 467 468 if (dev != blackhole_netdev) { 469 /* Join interface-local all-node multicast group */ 470 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes); 471 472 /* Join all-node multicast group */ 473 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes); 474 475 /* Join all-router multicast group if forwarding is set */ 476 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST)) 477 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 478 } 479 return ndev; 480 481 err_release: 482 neigh_parms_release(&nd_tbl, ndev->nd_parms); 483 ndev->dead = 1; 484 in6_dev_finish_destroy(ndev); 485 return ERR_PTR(err); 486 } 487 488 static struct inet6_dev *ipv6_find_idev(struct net_device *dev) 489 { 490 struct inet6_dev *idev; 491 492 ASSERT_RTNL(); 493 494 idev = __in6_dev_get(dev); 495 if (!idev) { 496 idev = ipv6_add_dev(dev); 497 if (IS_ERR(idev)) 498 return idev; 499 } 500 501 if (dev->flags&IFF_UP) 502 ipv6_mc_up(idev); 503 return idev; 504 } 505 506 static int inet6_netconf_msgsize_devconf(int type) 507 { 508 int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) 509 + nla_total_size(4); /* NETCONFA_IFINDEX */ 510 bool all = false; 511 512 if (type == NETCONFA_ALL) 513 all = true; 514 515 if (all || type == NETCONFA_FORWARDING) 516 size += nla_total_size(4); 517 #ifdef CONFIG_IPV6_MROUTE 518 if (all || type == NETCONFA_MC_FORWARDING) 519 size += nla_total_size(4); 520 #endif 521 if (all || type == NETCONFA_PROXY_NEIGH) 522 size += nla_total_size(4); 523 524 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) 525 size += nla_total_size(4); 526 527 return size; 528 } 529 530 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex, 531 struct ipv6_devconf *devconf, u32 portid, 532 u32 seq, int event, unsigned int flags, 533 int type) 534 { 535 struct nlmsghdr *nlh; 536 struct netconfmsg *ncm; 537 bool all = false; 538 539 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), 540 flags); 541 if (!nlh) 542 return -EMSGSIZE; 543 544 if (type == NETCONFA_ALL) 545 all = true; 546 547 ncm = nlmsg_data(nlh); 548 ncm->ncm_family = AF_INET6; 549 550 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0) 551 goto nla_put_failure; 552 553 if (!devconf) 554 goto out; 555 556 if ((all || type == NETCONFA_FORWARDING) && 557 nla_put_s32(skb, NETCONFA_FORWARDING, 558 READ_ONCE(devconf->forwarding)) < 0) 559 goto nla_put_failure; 560 #ifdef CONFIG_IPV6_MROUTE 561 if ((all || type == NETCONFA_MC_FORWARDING) && 562 nla_put_s32(skb, NETCONFA_MC_FORWARDING, 563 atomic_read(&devconf->mc_forwarding)) < 0) 564 goto nla_put_failure; 565 #endif 566 if ((all || type == NETCONFA_PROXY_NEIGH) && 567 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, 568 READ_ONCE(devconf->proxy_ndp)) < 0) 569 goto nla_put_failure; 570 571 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) && 572 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 573 READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0) 574 goto nla_put_failure; 575 576 out: 577 nlmsg_end(skb, nlh); 578 return 0; 579 580 nla_put_failure: 581 nlmsg_cancel(skb, nlh); 582 return -EMSGSIZE; 583 } 584 585 void inet6_netconf_notify_devconf(struct net *net, int event, int type, 586 int ifindex, struct ipv6_devconf *devconf) 587 { 588 struct sk_buff *skb; 589 int err = -ENOBUFS; 590 591 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL); 592 if (!skb) 593 goto errout; 594 595 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0, 596 event, 0, type); 597 if (err < 0) { 598 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 599 WARN_ON(err == -EMSGSIZE); 600 kfree_skb(skb); 601 goto errout; 602 } 603 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL); 604 return; 605 errout: 606 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err); 607 } 608 609 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = { 610 [NETCONFA_IFINDEX] = { .len = sizeof(int) }, 611 [NETCONFA_FORWARDING] = { .len = sizeof(int) }, 612 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) }, 613 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) }, 614 }; 615 616 static int inet6_netconf_valid_get_req(struct sk_buff *skb, 617 const struct nlmsghdr *nlh, 618 struct nlattr **tb, 619 struct netlink_ext_ack *extack) 620 { 621 int i, err; 622 623 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) { 624 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request"); 625 return -EINVAL; 626 } 627 628 if (!netlink_strict_get_check(skb)) 629 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg), 630 tb, NETCONFA_MAX, 631 devconf_ipv6_policy, extack); 632 633 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg), 634 tb, NETCONFA_MAX, 635 devconf_ipv6_policy, extack); 636 if (err) 637 return err; 638 639 for (i = 0; i <= NETCONFA_MAX; i++) { 640 if (!tb[i]) 641 continue; 642 643 switch (i) { 644 case NETCONFA_IFINDEX: 645 break; 646 default: 647 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request"); 648 return -EINVAL; 649 } 650 } 651 652 return 0; 653 } 654 655 static int inet6_netconf_get_devconf(struct sk_buff *in_skb, 656 struct nlmsghdr *nlh, 657 struct netlink_ext_ack *extack) 658 { 659 struct net *net = sock_net(in_skb->sk); 660 struct nlattr *tb[NETCONFA_MAX+1]; 661 struct inet6_dev *in6_dev = NULL; 662 struct net_device *dev = NULL; 663 struct sk_buff *skb; 664 struct ipv6_devconf *devconf; 665 int ifindex; 666 int err; 667 668 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack); 669 if (err < 0) 670 return err; 671 672 if (!tb[NETCONFA_IFINDEX]) 673 return -EINVAL; 674 675 err = -EINVAL; 676 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); 677 switch (ifindex) { 678 case NETCONFA_IFINDEX_ALL: 679 devconf = net->ipv6.devconf_all; 680 break; 681 case NETCONFA_IFINDEX_DEFAULT: 682 devconf = net->ipv6.devconf_dflt; 683 break; 684 default: 685 dev = dev_get_by_index(net, ifindex); 686 if (!dev) 687 return -EINVAL; 688 in6_dev = in6_dev_get(dev); 689 if (!in6_dev) 690 goto errout; 691 devconf = &in6_dev->cnf; 692 break; 693 } 694 695 err = -ENOBUFS; 696 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); 697 if (!skb) 698 goto errout; 699 700 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 701 NETLINK_CB(in_skb).portid, 702 nlh->nlmsg_seq, RTM_NEWNETCONF, 0, 703 NETCONFA_ALL); 704 if (err < 0) { 705 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 706 WARN_ON(err == -EMSGSIZE); 707 kfree_skb(skb); 708 goto errout; 709 } 710 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 711 errout: 712 if (in6_dev) 713 in6_dev_put(in6_dev); 714 dev_put(dev); 715 return err; 716 } 717 718 /* Combine dev_addr_genid and dev_base_seq to detect changes. 719 */ 720 static u32 inet6_base_seq(const struct net *net) 721 { 722 u32 res = atomic_read(&net->ipv6.dev_addr_genid) + 723 READ_ONCE(net->dev_base_seq); 724 725 /* Must not return 0 (see nl_dump_check_consistent()). 726 * Chose a value far away from 0. 727 */ 728 if (!res) 729 res = 0x80000000; 730 return res; 731 } 732 733 static int inet6_netconf_dump_devconf(struct sk_buff *skb, 734 struct netlink_callback *cb) 735 { 736 const struct nlmsghdr *nlh = cb->nlh; 737 struct net *net = sock_net(skb->sk); 738 struct { 739 unsigned long ifindex; 740 unsigned int all_default; 741 } *ctx = (void *)cb->ctx; 742 struct net_device *dev; 743 struct inet6_dev *idev; 744 int err = 0; 745 746 if (cb->strict_check) { 747 struct netlink_ext_ack *extack = cb->extack; 748 struct netconfmsg *ncm; 749 750 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) { 751 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request"); 752 return -EINVAL; 753 } 754 755 if (nlmsg_attrlen(nlh, sizeof(*ncm))) { 756 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request"); 757 return -EINVAL; 758 } 759 } 760 761 rcu_read_lock(); 762 for_each_netdev_dump(net, dev, ctx->ifindex) { 763 idev = __in6_dev_get(dev); 764 if (!idev) 765 continue; 766 err = inet6_netconf_fill_devconf(skb, dev->ifindex, 767 &idev->cnf, 768 NETLINK_CB(cb->skb).portid, 769 nlh->nlmsg_seq, 770 RTM_NEWNETCONF, 771 NLM_F_MULTI, 772 NETCONFA_ALL); 773 if (err < 0) 774 goto done; 775 } 776 if (ctx->all_default == 0) { 777 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL, 778 net->ipv6.devconf_all, 779 NETLINK_CB(cb->skb).portid, 780 nlh->nlmsg_seq, 781 RTM_NEWNETCONF, NLM_F_MULTI, 782 NETCONFA_ALL); 783 if (err < 0) 784 goto done; 785 ctx->all_default++; 786 } 787 if (ctx->all_default == 1) { 788 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT, 789 net->ipv6.devconf_dflt, 790 NETLINK_CB(cb->skb).portid, 791 nlh->nlmsg_seq, 792 RTM_NEWNETCONF, NLM_F_MULTI, 793 NETCONFA_ALL); 794 if (err < 0) 795 goto done; 796 ctx->all_default++; 797 } 798 done: 799 rcu_read_unlock(); 800 return err; 801 } 802 803 #ifdef CONFIG_SYSCTL 804 static void dev_forward_change(struct inet6_dev *idev) 805 { 806 struct net_device *dev; 807 struct inet6_ifaddr *ifa; 808 LIST_HEAD(tmp_addr_list); 809 810 if (!idev) 811 return; 812 dev = idev->dev; 813 if (idev->cnf.forwarding) 814 dev_disable_lro(dev); 815 if (dev->flags & IFF_MULTICAST) { 816 if (idev->cnf.forwarding) { 817 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 818 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters); 819 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters); 820 } else { 821 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters); 822 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters); 823 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters); 824 } 825 } 826 827 read_lock_bh(&idev->lock); 828 list_for_each_entry(ifa, &idev->addr_list, if_list) { 829 if (ifa->flags&IFA_F_TENTATIVE) 830 continue; 831 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 832 } 833 read_unlock_bh(&idev->lock); 834 835 while (!list_empty(&tmp_addr_list)) { 836 ifa = list_first_entry(&tmp_addr_list, 837 struct inet6_ifaddr, if_list_aux); 838 list_del(&ifa->if_list_aux); 839 if (idev->cnf.forwarding) 840 addrconf_join_anycast(ifa); 841 else 842 addrconf_leave_anycast(ifa); 843 } 844 845 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 846 NETCONFA_FORWARDING, 847 dev->ifindex, &idev->cnf); 848 } 849 850 851 static void addrconf_forward_change(struct net *net, __s32 newf) 852 { 853 struct net_device *dev; 854 struct inet6_dev *idev; 855 856 for_each_netdev(net, dev) { 857 idev = __in6_dev_get_rtnl_net(dev); 858 if (idev) { 859 int changed = (!idev->cnf.forwarding) ^ (!newf); 860 861 WRITE_ONCE(idev->cnf.forwarding, newf); 862 if (changed) 863 dev_forward_change(idev); 864 } 865 } 866 } 867 868 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf) 869 { 870 struct net *net = (struct net *)table->extra2; 871 int old; 872 873 if (!rtnl_net_trylock(net)) 874 return restart_syscall(); 875 876 old = *p; 877 WRITE_ONCE(*p, newf); 878 879 if (p == &net->ipv6.devconf_dflt->forwarding) { 880 if ((!newf) ^ (!old)) 881 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 882 NETCONFA_FORWARDING, 883 NETCONFA_IFINDEX_DEFAULT, 884 net->ipv6.devconf_dflt); 885 rtnl_net_unlock(net); 886 return 0; 887 } 888 889 if (p == &net->ipv6.devconf_all->forwarding) { 890 int old_dflt = net->ipv6.devconf_dflt->forwarding; 891 892 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf); 893 if ((!newf) ^ (!old_dflt)) 894 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 895 NETCONFA_FORWARDING, 896 NETCONFA_IFINDEX_DEFAULT, 897 net->ipv6.devconf_dflt); 898 899 addrconf_forward_change(net, newf); 900 if ((!newf) ^ (!old)) 901 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 902 NETCONFA_FORWARDING, 903 NETCONFA_IFINDEX_ALL, 904 net->ipv6.devconf_all); 905 } else if ((!newf) ^ (!old)) 906 dev_forward_change((struct inet6_dev *)table->extra1); 907 rtnl_net_unlock(net); 908 909 if (newf) 910 rt6_purge_dflt_routers(net); 911 return 1; 912 } 913 914 static void addrconf_linkdown_change(struct net *net, __s32 newf) 915 { 916 struct net_device *dev; 917 struct inet6_dev *idev; 918 919 for_each_netdev(net, dev) { 920 idev = __in6_dev_get_rtnl_net(dev); 921 if (idev) { 922 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf); 923 924 WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf); 925 if (changed) 926 inet6_netconf_notify_devconf(dev_net(dev), 927 RTM_NEWNETCONF, 928 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 929 dev->ifindex, 930 &idev->cnf); 931 } 932 } 933 } 934 935 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf) 936 { 937 struct net *net = (struct net *)table->extra2; 938 int old; 939 940 if (!rtnl_net_trylock(net)) 941 return restart_syscall(); 942 943 old = *p; 944 WRITE_ONCE(*p, newf); 945 946 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) { 947 if ((!newf) ^ (!old)) 948 inet6_netconf_notify_devconf(net, 949 RTM_NEWNETCONF, 950 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 951 NETCONFA_IFINDEX_DEFAULT, 952 net->ipv6.devconf_dflt); 953 rtnl_net_unlock(net); 954 return 0; 955 } 956 957 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) { 958 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf); 959 addrconf_linkdown_change(net, newf); 960 if ((!newf) ^ (!old)) 961 inet6_netconf_notify_devconf(net, 962 RTM_NEWNETCONF, 963 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 964 NETCONFA_IFINDEX_ALL, 965 net->ipv6.devconf_all); 966 } 967 968 rtnl_net_unlock(net); 969 970 return 1; 971 } 972 973 #endif 974 975 /* Nobody refers to this ifaddr, destroy it */ 976 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp) 977 { 978 WARN_ON(!hlist_unhashed(&ifp->addr_lst)); 979 980 #ifdef NET_REFCNT_DEBUG 981 pr_debug("%s\n", __func__); 982 #endif 983 984 in6_dev_put(ifp->idev); 985 986 if (cancel_delayed_work(&ifp->dad_work)) 987 pr_notice("delayed DAD work was pending while freeing ifa=%p\n", 988 ifp); 989 990 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 991 pr_warn("Freeing alive inet6 address %p\n", ifp); 992 return; 993 } 994 995 kfree_rcu(ifp, rcu); 996 } 997 998 static void 999 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp) 1000 { 1001 struct list_head *p; 1002 int ifp_scope = ipv6_addr_src_scope(&ifp->addr); 1003 1004 /* 1005 * Each device address list is sorted in order of scope - 1006 * global before linklocal. 1007 */ 1008 list_for_each(p, &idev->addr_list) { 1009 struct inet6_ifaddr *ifa 1010 = list_entry(p, struct inet6_ifaddr, if_list); 1011 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr)) 1012 break; 1013 } 1014 1015 list_add_tail_rcu(&ifp->if_list, p); 1016 } 1017 1018 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr) 1019 { 1020 u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net)); 1021 1022 return hash_32(val, IN6_ADDR_HSIZE_SHIFT); 1023 } 1024 1025 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr, 1026 struct net_device *dev, unsigned int hash) 1027 { 1028 struct inet6_ifaddr *ifp; 1029 1030 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 1031 if (ipv6_addr_equal(&ifp->addr, addr)) { 1032 if (!dev || ifp->idev->dev == dev) 1033 return true; 1034 } 1035 } 1036 return false; 1037 } 1038 1039 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa) 1040 { 1041 struct net *net = dev_net(dev); 1042 unsigned int hash = inet6_addr_hash(net, &ifa->addr); 1043 int err = 0; 1044 1045 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1046 1047 /* Ignore adding duplicate addresses on an interface */ 1048 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) { 1049 netdev_dbg(dev, "ipv6_add_addr: already assigned\n"); 1050 err = -EEXIST; 1051 } else { 1052 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]); 1053 } 1054 1055 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1056 1057 return err; 1058 } 1059 1060 /* On success it returns ifp with increased reference count */ 1061 1062 static struct inet6_ifaddr * 1063 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg, 1064 bool can_block, struct netlink_ext_ack *extack) 1065 { 1066 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC; 1067 int addr_type = ipv6_addr_type(cfg->pfx); 1068 struct net *net = dev_net(idev->dev); 1069 struct inet6_ifaddr *ifa = NULL; 1070 struct fib6_info *f6i = NULL; 1071 int err = 0; 1072 1073 if (addr_type == IPV6_ADDR_ANY) { 1074 NL_SET_ERR_MSG_MOD(extack, "Invalid address"); 1075 return ERR_PTR(-EADDRNOTAVAIL); 1076 } else if (addr_type & IPV6_ADDR_MULTICAST && 1077 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) { 1078 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag"); 1079 return ERR_PTR(-EADDRNOTAVAIL); 1080 } else if (!(idev->dev->flags & IFF_LOOPBACK) && 1081 !netif_is_l3_master(idev->dev) && 1082 addr_type & IPV6_ADDR_LOOPBACK) { 1083 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device"); 1084 return ERR_PTR(-EADDRNOTAVAIL); 1085 } 1086 1087 if (idev->dead) { 1088 NL_SET_ERR_MSG_MOD(extack, "device is going away"); 1089 err = -ENODEV; 1090 goto out; 1091 } 1092 1093 if (idev->cnf.disable_ipv6) { 1094 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 1095 err = -EACCES; 1096 goto out; 1097 } 1098 1099 /* validator notifier needs to be blocking; 1100 * do not call in atomic context 1101 */ 1102 if (can_block) { 1103 struct in6_validator_info i6vi = { 1104 .i6vi_addr = *cfg->pfx, 1105 .i6vi_dev = idev, 1106 .extack = extack, 1107 }; 1108 1109 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi); 1110 err = notifier_to_errno(err); 1111 if (err < 0) 1112 goto out; 1113 } 1114 1115 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT); 1116 if (!ifa) { 1117 err = -ENOBUFS; 1118 goto out; 1119 } 1120 1121 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack); 1122 if (IS_ERR(f6i)) { 1123 err = PTR_ERR(f6i); 1124 f6i = NULL; 1125 goto out; 1126 } 1127 1128 neigh_parms_data_state_setall(idev->nd_parms); 1129 1130 ifa->addr = *cfg->pfx; 1131 if (cfg->peer_pfx) 1132 ifa->peer_addr = *cfg->peer_pfx; 1133 1134 spin_lock_init(&ifa->lock); 1135 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work); 1136 INIT_HLIST_NODE(&ifa->addr_lst); 1137 ifa->scope = cfg->scope; 1138 ifa->prefix_len = cfg->plen; 1139 ifa->rt_priority = cfg->rt_priority; 1140 ifa->flags = cfg->ifa_flags; 1141 ifa->ifa_proto = cfg->ifa_proto; 1142 /* No need to add the TENTATIVE flag for addresses with NODAD */ 1143 if (!(cfg->ifa_flags & IFA_F_NODAD)) 1144 ifa->flags |= IFA_F_TENTATIVE; 1145 ifa->valid_lft = cfg->valid_lft; 1146 ifa->prefered_lft = cfg->preferred_lft; 1147 ifa->cstamp = ifa->tstamp = jiffies; 1148 ifa->tokenized = false; 1149 1150 ifa->rt = f6i; 1151 1152 ifa->idev = idev; 1153 in6_dev_hold(idev); 1154 1155 /* For caller */ 1156 refcount_set(&ifa->refcnt, 1); 1157 1158 rcu_read_lock(); 1159 1160 err = ipv6_add_addr_hash(idev->dev, ifa); 1161 if (err < 0) { 1162 rcu_read_unlock(); 1163 goto out; 1164 } 1165 1166 write_lock_bh(&idev->lock); 1167 1168 /* Add to inet6_dev unicast addr list. */ 1169 ipv6_link_dev_addr(idev, ifa); 1170 1171 if (ifa->flags&IFA_F_TEMPORARY) { 1172 list_add(&ifa->tmp_list, &idev->tempaddr_list); 1173 in6_ifa_hold(ifa); 1174 } 1175 1176 in6_ifa_hold(ifa); 1177 write_unlock_bh(&idev->lock); 1178 1179 rcu_read_unlock(); 1180 1181 inet6addr_notifier_call_chain(NETDEV_UP, ifa); 1182 out: 1183 if (unlikely(err < 0)) { 1184 fib6_info_release(f6i); 1185 1186 if (ifa) { 1187 if (ifa->idev) 1188 in6_dev_put(ifa->idev); 1189 kfree(ifa); 1190 } 1191 ifa = ERR_PTR(err); 1192 } 1193 1194 return ifa; 1195 } 1196 1197 enum cleanup_prefix_rt_t { 1198 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */ 1199 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */ 1200 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */ 1201 }; 1202 1203 /* 1204 * Check, whether the prefix for ifp would still need a prefix route 1205 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_* 1206 * constants. 1207 * 1208 * 1) we don't purge prefix if address was not permanent. 1209 * prefix is managed by its own lifetime. 1210 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE. 1211 * 3) if there are no addresses, delete prefix. 1212 * 4) if there are still other permanent address(es), 1213 * corresponding prefix is still permanent. 1214 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE, 1215 * don't purge the prefix, assume user space is managing it. 1216 * 6) otherwise, update prefix lifetime to the 1217 * longest valid lifetime among the corresponding 1218 * addresses on the device. 1219 * Note: subsequent RA will update lifetime. 1220 **/ 1221 static enum cleanup_prefix_rt_t 1222 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires) 1223 { 1224 struct inet6_ifaddr *ifa; 1225 struct inet6_dev *idev = ifp->idev; 1226 unsigned long lifetime; 1227 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL; 1228 1229 *expires = jiffies; 1230 1231 list_for_each_entry(ifa, &idev->addr_list, if_list) { 1232 if (ifa == ifp) 1233 continue; 1234 if (ifa->prefix_len != ifp->prefix_len || 1235 !ipv6_prefix_equal(&ifa->addr, &ifp->addr, 1236 ifp->prefix_len)) 1237 continue; 1238 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE)) 1239 return CLEANUP_PREFIX_RT_NOP; 1240 1241 action = CLEANUP_PREFIX_RT_EXPIRE; 1242 1243 spin_lock(&ifa->lock); 1244 1245 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ); 1246 /* 1247 * Note: Because this address is 1248 * not permanent, lifetime < 1249 * LONG_MAX / HZ here. 1250 */ 1251 if (time_before(*expires, ifa->tstamp + lifetime * HZ)) 1252 *expires = ifa->tstamp + lifetime * HZ; 1253 spin_unlock(&ifa->lock); 1254 } 1255 1256 return action; 1257 } 1258 1259 static void 1260 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, 1261 bool del_rt, bool del_peer) 1262 { 1263 struct fib6_table *table; 1264 struct fib6_info *f6i; 1265 1266 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr, 1267 ifp->prefix_len, 1268 ifp->idev->dev, 0, RTF_DEFAULT, true); 1269 if (f6i) { 1270 if (del_rt) 1271 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false); 1272 else { 1273 if (!(f6i->fib6_flags & RTF_EXPIRES)) { 1274 table = f6i->fib6_table; 1275 spin_lock_bh(&table->tb6_lock); 1276 1277 fib6_set_expires(f6i, expires); 1278 fib6_add_gc_list(f6i); 1279 1280 spin_unlock_bh(&table->tb6_lock); 1281 } 1282 fib6_info_release(f6i); 1283 } 1284 } 1285 } 1286 1287 1288 /* This function wants to get referenced ifp and releases it before return */ 1289 1290 static void ipv6_del_addr(struct inet6_ifaddr *ifp) 1291 { 1292 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP; 1293 struct net *net = dev_net(ifp->idev->dev); 1294 unsigned long expires; 1295 int state; 1296 1297 ASSERT_RTNL(); 1298 1299 spin_lock_bh(&ifp->lock); 1300 state = ifp->state; 1301 ifp->state = INET6_IFADDR_STATE_DEAD; 1302 spin_unlock_bh(&ifp->lock); 1303 1304 if (state == INET6_IFADDR_STATE_DEAD) 1305 goto out; 1306 1307 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1308 hlist_del_init_rcu(&ifp->addr_lst); 1309 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1310 1311 write_lock_bh(&ifp->idev->lock); 1312 1313 if (ifp->flags&IFA_F_TEMPORARY) { 1314 list_del(&ifp->tmp_list); 1315 if (ifp->ifpub) { 1316 in6_ifa_put(ifp->ifpub); 1317 ifp->ifpub = NULL; 1318 } 1319 __in6_ifa_put(ifp); 1320 } 1321 1322 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE)) 1323 action = check_cleanup_prefix_route(ifp, &expires); 1324 1325 list_del_rcu(&ifp->if_list); 1326 __in6_ifa_put(ifp); 1327 1328 write_unlock_bh(&ifp->idev->lock); 1329 1330 addrconf_del_dad_work(ifp); 1331 1332 ipv6_ifa_notify(RTM_DELADDR, ifp); 1333 1334 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp); 1335 1336 if (action != CLEANUP_PREFIX_RT_NOP) { 1337 cleanup_prefix_route(ifp, expires, 1338 action == CLEANUP_PREFIX_RT_DEL, false); 1339 } 1340 1341 /* clean up prefsrc entries */ 1342 rt6_remove_prefsrc(ifp); 1343 out: 1344 in6_ifa_put(ifp); 1345 } 1346 1347 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev) 1348 { 1349 return READ_ONCE(idev->cnf.regen_min_advance) + 1350 READ_ONCE(idev->cnf.regen_max_retry) * 1351 READ_ONCE(idev->cnf.dad_transmits) * 1352 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ; 1353 } 1354 1355 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block) 1356 { 1357 struct inet6_dev *idev = ifp->idev; 1358 unsigned long tmp_tstamp, age; 1359 unsigned long regen_advance; 1360 unsigned long now = jiffies; 1361 u32 if_public_preferred_lft; 1362 s32 cnf_temp_preferred_lft; 1363 struct inet6_ifaddr *ift; 1364 struct ifa6_config cfg; 1365 long max_desync_factor; 1366 struct in6_addr addr; 1367 int ret = 0; 1368 1369 write_lock_bh(&idev->lock); 1370 1371 retry: 1372 in6_dev_hold(idev); 1373 if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) { 1374 write_unlock_bh(&idev->lock); 1375 pr_info("%s: use_tempaddr is disabled\n", __func__); 1376 in6_dev_put(idev); 1377 ret = -1; 1378 goto out; 1379 } 1380 spin_lock_bh(&ifp->lock); 1381 if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) { 1382 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/ 1383 spin_unlock_bh(&ifp->lock); 1384 write_unlock_bh(&idev->lock); 1385 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n", 1386 __func__); 1387 in6_dev_put(idev); 1388 ret = -1; 1389 goto out; 1390 } 1391 in6_ifa_hold(ifp); 1392 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8); 1393 ipv6_gen_rnd_iid(&addr); 1394 1395 age = (now - ifp->tstamp) / HZ; 1396 1397 regen_advance = ipv6_get_regen_advance(idev); 1398 1399 /* recalculate max_desync_factor each time and update 1400 * idev->desync_factor if it's larger 1401 */ 1402 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft); 1403 max_desync_factor = min_t(long, 1404 READ_ONCE(idev->cnf.max_desync_factor), 1405 cnf_temp_preferred_lft - regen_advance); 1406 1407 if (unlikely(idev->desync_factor > max_desync_factor)) { 1408 if (max_desync_factor > 0) { 1409 get_random_bytes(&idev->desync_factor, 1410 sizeof(idev->desync_factor)); 1411 idev->desync_factor %= max_desync_factor; 1412 } else { 1413 idev->desync_factor = 0; 1414 } 1415 } 1416 1417 if_public_preferred_lft = ifp->prefered_lft; 1418 1419 memset(&cfg, 0, sizeof(cfg)); 1420 cfg.valid_lft = min_t(__u32, ifp->valid_lft, 1421 READ_ONCE(idev->cnf.temp_valid_lft) + age); 1422 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor; 1423 cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft); 1424 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft); 1425 1426 cfg.plen = ifp->prefix_len; 1427 tmp_tstamp = ifp->tstamp; 1428 spin_unlock_bh(&ifp->lock); 1429 1430 write_unlock_bh(&idev->lock); 1431 1432 /* From RFC 4941: 1433 * 1434 * A temporary address is created only if this calculated Preferred 1435 * Lifetime is greater than REGEN_ADVANCE time units. In 1436 * particular, an implementation must not create a temporary address 1437 * with a zero Preferred Lifetime. 1438 * 1439 * ... 1440 * 1441 * When creating a temporary address, the lifetime values MUST be 1442 * derived from the corresponding prefix as follows: 1443 * 1444 * ... 1445 * 1446 * * Its Preferred Lifetime is the lower of the Preferred Lifetime 1447 * of the public address or TEMP_PREFERRED_LIFETIME - 1448 * DESYNC_FACTOR. 1449 * 1450 * To comply with the RFC's requirements, clamp the preferred lifetime 1451 * to a minimum of regen_advance, unless that would exceed valid_lft or 1452 * ifp->prefered_lft. 1453 * 1454 * Use age calculation as in addrconf_verify to avoid unnecessary 1455 * temporary addresses being generated. 1456 */ 1457 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 1458 if (cfg.preferred_lft <= regen_advance + age) { 1459 cfg.preferred_lft = regen_advance + age + 1; 1460 if (cfg.preferred_lft > cfg.valid_lft || 1461 cfg.preferred_lft > if_public_preferred_lft) { 1462 in6_ifa_put(ifp); 1463 in6_dev_put(idev); 1464 ret = -1; 1465 goto out; 1466 } 1467 } 1468 1469 cfg.ifa_flags = IFA_F_TEMPORARY; 1470 /* set in addrconf_prefix_rcv() */ 1471 if (ifp->flags & IFA_F_OPTIMISTIC) 1472 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 1473 1474 cfg.pfx = &addr; 1475 cfg.scope = ipv6_addr_scope(cfg.pfx); 1476 1477 ift = ipv6_add_addr(idev, &cfg, block, NULL); 1478 if (IS_ERR(ift)) { 1479 in6_ifa_put(ifp); 1480 in6_dev_put(idev); 1481 pr_info("%s: retry temporary address regeneration\n", __func__); 1482 write_lock_bh(&idev->lock); 1483 goto retry; 1484 } 1485 1486 spin_lock_bh(&ift->lock); 1487 ift->ifpub = ifp; 1488 ift->cstamp = now; 1489 ift->tstamp = tmp_tstamp; 1490 spin_unlock_bh(&ift->lock); 1491 1492 addrconf_dad_start(ift); 1493 in6_ifa_put(ift); 1494 in6_dev_put(idev); 1495 out: 1496 return ret; 1497 } 1498 1499 /* 1500 * Choose an appropriate source address (RFC3484) 1501 */ 1502 enum { 1503 IPV6_SADDR_RULE_INIT = 0, 1504 IPV6_SADDR_RULE_LOCAL, 1505 IPV6_SADDR_RULE_SCOPE, 1506 IPV6_SADDR_RULE_PREFERRED, 1507 #ifdef CONFIG_IPV6_MIP6 1508 IPV6_SADDR_RULE_HOA, 1509 #endif 1510 IPV6_SADDR_RULE_OIF, 1511 IPV6_SADDR_RULE_LABEL, 1512 IPV6_SADDR_RULE_PRIVACY, 1513 IPV6_SADDR_RULE_ORCHID, 1514 IPV6_SADDR_RULE_PREFIX, 1515 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1516 IPV6_SADDR_RULE_NOT_OPTIMISTIC, 1517 #endif 1518 IPV6_SADDR_RULE_MAX 1519 }; 1520 1521 struct ipv6_saddr_score { 1522 int rule; 1523 int addr_type; 1524 struct inet6_ifaddr *ifa; 1525 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX); 1526 int scopedist; 1527 int matchlen; 1528 }; 1529 1530 struct ipv6_saddr_dst { 1531 const struct in6_addr *addr; 1532 int ifindex; 1533 int scope; 1534 int label; 1535 unsigned int prefs; 1536 }; 1537 1538 static inline int ipv6_saddr_preferred(int type) 1539 { 1540 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK)) 1541 return 1; 1542 return 0; 1543 } 1544 1545 static bool ipv6_use_optimistic_addr(const struct net *net, 1546 const struct inet6_dev *idev) 1547 { 1548 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1549 if (!idev) 1550 return false; 1551 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) && 1552 !READ_ONCE(idev->cnf.optimistic_dad)) 1553 return false; 1554 if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) && 1555 !READ_ONCE(idev->cnf.use_optimistic)) 1556 return false; 1557 1558 return true; 1559 #else 1560 return false; 1561 #endif 1562 } 1563 1564 static bool ipv6_allow_optimistic_dad(const struct net *net, 1565 const struct inet6_dev *idev) 1566 { 1567 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1568 if (!idev) 1569 return false; 1570 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) && 1571 !READ_ONCE(idev->cnf.optimistic_dad)) 1572 return false; 1573 1574 return true; 1575 #else 1576 return false; 1577 #endif 1578 } 1579 1580 static int ipv6_get_saddr_eval(struct net *net, 1581 struct ipv6_saddr_score *score, 1582 struct ipv6_saddr_dst *dst, 1583 int i) 1584 { 1585 int ret; 1586 1587 if (i <= score->rule) { 1588 switch (i) { 1589 case IPV6_SADDR_RULE_SCOPE: 1590 ret = score->scopedist; 1591 break; 1592 case IPV6_SADDR_RULE_PREFIX: 1593 ret = score->matchlen; 1594 break; 1595 default: 1596 ret = !!test_bit(i, score->scorebits); 1597 } 1598 goto out; 1599 } 1600 1601 switch (i) { 1602 case IPV6_SADDR_RULE_INIT: 1603 /* Rule 0: remember if hiscore is not ready yet */ 1604 ret = !!score->ifa; 1605 break; 1606 case IPV6_SADDR_RULE_LOCAL: 1607 /* Rule 1: Prefer same address */ 1608 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr); 1609 break; 1610 case IPV6_SADDR_RULE_SCOPE: 1611 /* Rule 2: Prefer appropriate scope 1612 * 1613 * ret 1614 * ^ 1615 * -1 | d 15 1616 * ---+--+-+---> scope 1617 * | 1618 * | d is scope of the destination. 1619 * B-d | \ 1620 * | \ <- smaller scope is better if 1621 * B-15 | \ if scope is enough for destination. 1622 * | ret = B - scope (-1 <= scope >= d <= 15). 1623 * d-C-1 | / 1624 * |/ <- greater is better 1625 * -C / if scope is not enough for destination. 1626 * /| ret = scope - C (-1 <= d < scope <= 15). 1627 * 1628 * d - C - 1 < B -15 (for all -1 <= d <= 15). 1629 * C > d + 14 - B >= 15 + 14 - B = 29 - B. 1630 * Assume B = 0 and we get C > 29. 1631 */ 1632 ret = __ipv6_addr_src_scope(score->addr_type); 1633 if (ret >= dst->scope) 1634 ret = -ret; 1635 else 1636 ret -= 128; /* 30 is enough */ 1637 score->scopedist = ret; 1638 break; 1639 case IPV6_SADDR_RULE_PREFERRED: 1640 { 1641 /* Rule 3: Avoid deprecated and optimistic addresses */ 1642 u8 avoid = IFA_F_DEPRECATED; 1643 1644 if (!ipv6_use_optimistic_addr(net, score->ifa->idev)) 1645 avoid |= IFA_F_OPTIMISTIC; 1646 ret = ipv6_saddr_preferred(score->addr_type) || 1647 !(score->ifa->flags & avoid); 1648 break; 1649 } 1650 #ifdef CONFIG_IPV6_MIP6 1651 case IPV6_SADDR_RULE_HOA: 1652 { 1653 /* Rule 4: Prefer home address */ 1654 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA); 1655 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome; 1656 break; 1657 } 1658 #endif 1659 case IPV6_SADDR_RULE_OIF: 1660 /* Rule 5: Prefer outgoing interface */ 1661 ret = (!dst->ifindex || 1662 dst->ifindex == score->ifa->idev->dev->ifindex); 1663 break; 1664 case IPV6_SADDR_RULE_LABEL: 1665 /* Rule 6: Prefer matching label */ 1666 ret = ipv6_addr_label(net, 1667 &score->ifa->addr, score->addr_type, 1668 score->ifa->idev->dev->ifindex) == dst->label; 1669 break; 1670 case IPV6_SADDR_RULE_PRIVACY: 1671 { 1672 /* Rule 7: Prefer public address 1673 * Note: prefer temporary address if use_tempaddr >= 2 1674 */ 1675 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ? 1676 !!(dst->prefs & IPV6_PREFER_SRC_TMP) : 1677 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2; 1678 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp; 1679 break; 1680 } 1681 case IPV6_SADDR_RULE_ORCHID: 1682 /* Rule 8-: Prefer ORCHID vs ORCHID or 1683 * non-ORCHID vs non-ORCHID 1684 */ 1685 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^ 1686 ipv6_addr_orchid(dst->addr)); 1687 break; 1688 case IPV6_SADDR_RULE_PREFIX: 1689 /* Rule 8: Use longest matching prefix */ 1690 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr); 1691 if (ret > score->ifa->prefix_len) 1692 ret = score->ifa->prefix_len; 1693 score->matchlen = ret; 1694 break; 1695 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1696 case IPV6_SADDR_RULE_NOT_OPTIMISTIC: 1697 /* Optimistic addresses still have lower precedence than other 1698 * preferred addresses. 1699 */ 1700 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC); 1701 break; 1702 #endif 1703 default: 1704 ret = 0; 1705 } 1706 1707 if (ret) 1708 __set_bit(i, score->scorebits); 1709 score->rule = i; 1710 out: 1711 return ret; 1712 } 1713 1714 static int __ipv6_dev_get_saddr(struct net *net, 1715 struct ipv6_saddr_dst *dst, 1716 struct inet6_dev *idev, 1717 struct ipv6_saddr_score *scores, 1718 int hiscore_idx) 1719 { 1720 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx]; 1721 1722 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) { 1723 int i; 1724 1725 /* 1726 * - Tentative Address (RFC2462 section 5.4) 1727 * - A tentative address is not considered 1728 * "assigned to an interface" in the traditional 1729 * sense, unless it is also flagged as optimistic. 1730 * - Candidate Source Address (section 4) 1731 * - In any case, anycast addresses, multicast 1732 * addresses, and the unspecified address MUST 1733 * NOT be included in a candidate set. 1734 */ 1735 if ((score->ifa->flags & IFA_F_TENTATIVE) && 1736 (!(score->ifa->flags & IFA_F_OPTIMISTIC))) 1737 continue; 1738 1739 score->addr_type = __ipv6_addr_type(&score->ifa->addr); 1740 1741 if (unlikely(score->addr_type == IPV6_ADDR_ANY || 1742 score->addr_type & IPV6_ADDR_MULTICAST)) { 1743 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s", 1744 idev->dev->name); 1745 continue; 1746 } 1747 1748 score->rule = -1; 1749 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX); 1750 1751 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) { 1752 int minihiscore, miniscore; 1753 1754 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i); 1755 miniscore = ipv6_get_saddr_eval(net, score, dst, i); 1756 1757 if (minihiscore > miniscore) { 1758 if (i == IPV6_SADDR_RULE_SCOPE && 1759 score->scopedist > 0) { 1760 /* 1761 * special case: 1762 * each remaining entry 1763 * has too small (not enough) 1764 * scope, because ifa entries 1765 * are sorted by their scope 1766 * values. 1767 */ 1768 goto out; 1769 } 1770 break; 1771 } else if (minihiscore < miniscore) { 1772 swap(hiscore, score); 1773 hiscore_idx = 1 - hiscore_idx; 1774 1775 /* restore our iterator */ 1776 score->ifa = hiscore->ifa; 1777 1778 break; 1779 } 1780 } 1781 } 1782 out: 1783 return hiscore_idx; 1784 } 1785 1786 static int ipv6_get_saddr_master(struct net *net, 1787 const struct net_device *dst_dev, 1788 const struct net_device *master, 1789 struct ipv6_saddr_dst *dst, 1790 struct ipv6_saddr_score *scores, 1791 int hiscore_idx) 1792 { 1793 struct inet6_dev *idev; 1794 1795 idev = __in6_dev_get(dst_dev); 1796 if (idev) 1797 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1798 scores, hiscore_idx); 1799 1800 idev = __in6_dev_get(master); 1801 if (idev) 1802 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1803 scores, hiscore_idx); 1804 1805 return hiscore_idx; 1806 } 1807 1808 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev, 1809 const struct in6_addr *daddr, unsigned int prefs, 1810 struct in6_addr *saddr) 1811 { 1812 struct ipv6_saddr_score scores[2], *hiscore; 1813 struct ipv6_saddr_dst dst; 1814 struct inet6_dev *idev; 1815 struct net_device *dev; 1816 int dst_type; 1817 bool use_oif_addr = false; 1818 int hiscore_idx = 0; 1819 int ret = 0; 1820 1821 dst_type = __ipv6_addr_type(daddr); 1822 dst.addr = daddr; 1823 dst.ifindex = dst_dev ? dst_dev->ifindex : 0; 1824 dst.scope = __ipv6_addr_src_scope(dst_type); 1825 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex); 1826 dst.prefs = prefs; 1827 1828 scores[hiscore_idx].rule = -1; 1829 scores[hiscore_idx].ifa = NULL; 1830 1831 rcu_read_lock(); 1832 1833 /* Candidate Source Address (section 4) 1834 * - multicast and link-local destination address, 1835 * the set of candidate source address MUST only 1836 * include addresses assigned to interfaces 1837 * belonging to the same link as the outgoing 1838 * interface. 1839 * (- For site-local destination addresses, the 1840 * set of candidate source addresses MUST only 1841 * include addresses assigned to interfaces 1842 * belonging to the same site as the outgoing 1843 * interface.) 1844 * - "It is RECOMMENDED that the candidate source addresses 1845 * be the set of unicast addresses assigned to the 1846 * interface that will be used to send to the destination 1847 * (the 'outgoing' interface)." (RFC 6724) 1848 */ 1849 if (dst_dev) { 1850 idev = __in6_dev_get(dst_dev); 1851 if ((dst_type & IPV6_ADDR_MULTICAST) || 1852 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL || 1853 (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) { 1854 use_oif_addr = true; 1855 } 1856 } 1857 1858 if (use_oif_addr) { 1859 if (idev) 1860 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1861 } else { 1862 const struct net_device *master; 1863 int master_idx = 0; 1864 1865 /* if dst_dev exists and is enslaved to an L3 device, then 1866 * prefer addresses from dst_dev and then the master over 1867 * any other enslaved devices in the L3 domain. 1868 */ 1869 master = l3mdev_master_dev_rcu(dst_dev); 1870 if (master) { 1871 master_idx = master->ifindex; 1872 1873 hiscore_idx = ipv6_get_saddr_master(net, dst_dev, 1874 master, &dst, 1875 scores, hiscore_idx); 1876 1877 if (scores[hiscore_idx].ifa && 1878 scores[hiscore_idx].scopedist >= 0) 1879 goto out; 1880 } 1881 1882 for_each_netdev_rcu(net, dev) { 1883 /* only consider addresses on devices in the 1884 * same L3 domain 1885 */ 1886 if (l3mdev_master_ifindex_rcu(dev) != master_idx) 1887 continue; 1888 idev = __in6_dev_get(dev); 1889 if (!idev) 1890 continue; 1891 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1892 } 1893 } 1894 1895 out: 1896 hiscore = &scores[hiscore_idx]; 1897 if (!hiscore->ifa) 1898 ret = -EADDRNOTAVAIL; 1899 else 1900 *saddr = hiscore->ifa->addr; 1901 1902 rcu_read_unlock(); 1903 return ret; 1904 } 1905 EXPORT_SYMBOL(ipv6_dev_get_saddr); 1906 1907 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr, 1908 u32 banned_flags) 1909 { 1910 struct inet6_ifaddr *ifp; 1911 int err = -EADDRNOTAVAIL; 1912 1913 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 1914 if (ifp->scope > IFA_LINK) 1915 break; 1916 if (ifp->scope == IFA_LINK && 1917 !(ifp->flags & banned_flags)) { 1918 *addr = ifp->addr; 1919 err = 0; 1920 break; 1921 } 1922 } 1923 return err; 1924 } 1925 1926 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr, 1927 u32 banned_flags) 1928 { 1929 struct inet6_dev *idev; 1930 int err = -EADDRNOTAVAIL; 1931 1932 rcu_read_lock(); 1933 idev = __in6_dev_get(dev); 1934 if (idev) { 1935 read_lock_bh(&idev->lock); 1936 err = __ipv6_get_lladdr(idev, addr, banned_flags); 1937 read_unlock_bh(&idev->lock); 1938 } 1939 rcu_read_unlock(); 1940 return err; 1941 } 1942 1943 static int ipv6_count_addresses(const struct inet6_dev *idev) 1944 { 1945 const struct inet6_ifaddr *ifp; 1946 int cnt = 0; 1947 1948 rcu_read_lock(); 1949 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list) 1950 cnt++; 1951 rcu_read_unlock(); 1952 return cnt; 1953 } 1954 1955 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr, 1956 const struct net_device *dev, int strict) 1957 { 1958 return ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1959 strict, IFA_F_TENTATIVE); 1960 } 1961 EXPORT_SYMBOL(ipv6_chk_addr); 1962 1963 /* device argument is used to find the L3 domain of interest. If 1964 * skip_dev_check is set, then the ifp device is not checked against 1965 * the passed in dev argument. So the 2 cases for addresses checks are: 1966 * 1. does the address exist in the L3 domain that dev is part of 1967 * (skip_dev_check = true), or 1968 * 1969 * 2. does the address exist on the specific device 1970 * (skip_dev_check = false) 1971 */ 1972 static struct net_device * 1973 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 1974 const struct net_device *dev, bool skip_dev_check, 1975 int strict, u32 banned_flags) 1976 { 1977 unsigned int hash = inet6_addr_hash(net, addr); 1978 struct net_device *l3mdev, *ndev; 1979 struct inet6_ifaddr *ifp; 1980 u32 ifp_flags; 1981 1982 rcu_read_lock(); 1983 1984 l3mdev = l3mdev_master_dev_rcu(dev); 1985 if (skip_dev_check) 1986 dev = NULL; 1987 1988 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 1989 ndev = ifp->idev->dev; 1990 1991 if (l3mdev_master_dev_rcu(ndev) != l3mdev) 1992 continue; 1993 1994 /* Decouple optimistic from tentative for evaluation here. 1995 * Ban optimistic addresses explicitly, when required. 1996 */ 1997 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC) 1998 ? (ifp->flags&~IFA_F_TENTATIVE) 1999 : ifp->flags; 2000 if (ipv6_addr_equal(&ifp->addr, addr) && 2001 !(ifp_flags&banned_flags) && 2002 (!dev || ndev == dev || 2003 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) { 2004 rcu_read_unlock(); 2005 return ndev; 2006 } 2007 } 2008 2009 rcu_read_unlock(); 2010 return NULL; 2011 } 2012 2013 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 2014 const struct net_device *dev, bool skip_dev_check, 2015 int strict, u32 banned_flags) 2016 { 2017 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check, 2018 strict, banned_flags) ? 1 : 0; 2019 } 2020 EXPORT_SYMBOL(ipv6_chk_addr_and_flags); 2021 2022 2023 /* Compares an address/prefix_len with addresses on device @dev. 2024 * If one is found it returns true. 2025 */ 2026 bool ipv6_chk_custom_prefix(const struct in6_addr *addr, 2027 const unsigned int prefix_len, struct net_device *dev) 2028 { 2029 const struct inet6_ifaddr *ifa; 2030 const struct inet6_dev *idev; 2031 bool ret = false; 2032 2033 rcu_read_lock(); 2034 idev = __in6_dev_get(dev); 2035 if (idev) { 2036 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 2037 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len); 2038 if (ret) 2039 break; 2040 } 2041 } 2042 rcu_read_unlock(); 2043 2044 return ret; 2045 } 2046 EXPORT_SYMBOL(ipv6_chk_custom_prefix); 2047 2048 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev) 2049 { 2050 const struct inet6_ifaddr *ifa; 2051 const struct inet6_dev *idev; 2052 int onlink; 2053 2054 onlink = 0; 2055 rcu_read_lock(); 2056 idev = __in6_dev_get(dev); 2057 if (idev) { 2058 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 2059 onlink = ipv6_prefix_equal(addr, &ifa->addr, 2060 ifa->prefix_len); 2061 if (onlink) 2062 break; 2063 } 2064 } 2065 rcu_read_unlock(); 2066 return onlink; 2067 } 2068 EXPORT_SYMBOL(ipv6_chk_prefix); 2069 2070 /** 2071 * ipv6_dev_find - find the first device with a given source address. 2072 * @net: the net namespace 2073 * @addr: the source address 2074 * @dev: used to find the L3 domain of interest 2075 * 2076 * The caller should be protected by RCU, or RTNL. 2077 */ 2078 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr, 2079 struct net_device *dev) 2080 { 2081 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1, 2082 IFA_F_TENTATIVE); 2083 } 2084 EXPORT_SYMBOL(ipv6_dev_find); 2085 2086 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr, 2087 struct net_device *dev, int strict) 2088 { 2089 unsigned int hash = inet6_addr_hash(net, addr); 2090 struct inet6_ifaddr *ifp, *result = NULL; 2091 2092 rcu_read_lock(); 2093 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 2094 if (ipv6_addr_equal(&ifp->addr, addr)) { 2095 if (!dev || ifp->idev->dev == dev || 2096 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) { 2097 if (in6_ifa_hold_safe(ifp)) { 2098 result = ifp; 2099 break; 2100 } 2101 } 2102 } 2103 } 2104 rcu_read_unlock(); 2105 2106 return result; 2107 } 2108 2109 /* Gets referenced address, destroys ifaddr */ 2110 2111 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed) 2112 { 2113 if (dad_failed) 2114 ifp->flags |= IFA_F_DADFAILED; 2115 2116 if (ifp->flags&IFA_F_TEMPORARY) { 2117 struct inet6_ifaddr *ifpub; 2118 spin_lock_bh(&ifp->lock); 2119 ifpub = ifp->ifpub; 2120 if (ifpub) { 2121 in6_ifa_hold(ifpub); 2122 spin_unlock_bh(&ifp->lock); 2123 ipv6_create_tempaddr(ifpub, true); 2124 in6_ifa_put(ifpub); 2125 } else { 2126 spin_unlock_bh(&ifp->lock); 2127 } 2128 ipv6_del_addr(ifp); 2129 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) { 2130 spin_lock_bh(&ifp->lock); 2131 addrconf_del_dad_work(ifp); 2132 ifp->flags |= IFA_F_TENTATIVE; 2133 if (dad_failed) 2134 ifp->flags &= ~IFA_F_OPTIMISTIC; 2135 spin_unlock_bh(&ifp->lock); 2136 if (dad_failed) 2137 ipv6_ifa_notify(0, ifp); 2138 in6_ifa_put(ifp); 2139 } else { 2140 ipv6_del_addr(ifp); 2141 } 2142 } 2143 2144 static int addrconf_dad_end(struct inet6_ifaddr *ifp) 2145 { 2146 int err = -ENOENT; 2147 2148 spin_lock_bh(&ifp->lock); 2149 if (ifp->state == INET6_IFADDR_STATE_DAD) { 2150 ifp->state = INET6_IFADDR_STATE_POSTDAD; 2151 err = 0; 2152 } 2153 spin_unlock_bh(&ifp->lock); 2154 2155 return err; 2156 } 2157 2158 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp) 2159 { 2160 struct inet6_dev *idev = ifp->idev; 2161 struct net *net = dev_net(idev->dev); 2162 int max_addresses; 2163 2164 if (addrconf_dad_end(ifp)) { 2165 in6_ifa_put(ifp); 2166 return; 2167 } 2168 2169 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n", 2170 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source); 2171 2172 spin_lock_bh(&ifp->lock); 2173 2174 if (ifp->flags & IFA_F_STABLE_PRIVACY) { 2175 struct in6_addr new_addr; 2176 struct inet6_ifaddr *ifp2; 2177 int retries = ifp->stable_privacy_retry + 1; 2178 struct ifa6_config cfg = { 2179 .pfx = &new_addr, 2180 .plen = ifp->prefix_len, 2181 .ifa_flags = ifp->flags, 2182 .valid_lft = ifp->valid_lft, 2183 .preferred_lft = ifp->prefered_lft, 2184 .scope = ifp->scope, 2185 }; 2186 2187 if (retries > net->ipv6.sysctl.idgen_retries) { 2188 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n", 2189 ifp->idev->dev->name); 2190 goto errdad; 2191 } 2192 2193 new_addr = ifp->addr; 2194 if (ipv6_generate_stable_address(&new_addr, retries, 2195 idev)) 2196 goto errdad; 2197 2198 spin_unlock_bh(&ifp->lock); 2199 2200 max_addresses = READ_ONCE(idev->cnf.max_addresses); 2201 if (max_addresses && 2202 ipv6_count_addresses(idev) >= max_addresses) 2203 goto lock_errdad; 2204 2205 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n", 2206 ifp->idev->dev->name); 2207 2208 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL); 2209 if (IS_ERR(ifp2)) 2210 goto lock_errdad; 2211 2212 spin_lock_bh(&ifp2->lock); 2213 ifp2->stable_privacy_retry = retries; 2214 ifp2->state = INET6_IFADDR_STATE_PREDAD; 2215 spin_unlock_bh(&ifp2->lock); 2216 2217 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay); 2218 in6_ifa_put(ifp2); 2219 lock_errdad: 2220 spin_lock_bh(&ifp->lock); 2221 } 2222 2223 errdad: 2224 /* transition from _POSTDAD to _ERRDAD */ 2225 ifp->state = INET6_IFADDR_STATE_ERRDAD; 2226 spin_unlock_bh(&ifp->lock); 2227 2228 addrconf_mod_dad_work(ifp, 0); 2229 in6_ifa_put(ifp); 2230 } 2231 2232 /* Join to solicited addr multicast group. */ 2233 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr) 2234 { 2235 struct in6_addr maddr; 2236 2237 if (READ_ONCE(dev->flags) & (IFF_LOOPBACK | IFF_NOARP)) 2238 return; 2239 2240 addrconf_addr_solict_mult(addr, &maddr); 2241 ipv6_dev_mc_inc(dev, &maddr); 2242 } 2243 2244 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr) 2245 { 2246 struct in6_addr maddr; 2247 2248 if (READ_ONCE(idev->dev->flags) & (IFF_LOOPBACK | IFF_NOARP)) 2249 return; 2250 2251 addrconf_addr_solict_mult(addr, &maddr); 2252 __ipv6_dev_mc_dec(idev, &maddr); 2253 } 2254 2255 static void addrconf_join_anycast(struct inet6_ifaddr *ifp) 2256 { 2257 struct in6_addr addr; 2258 2259 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2260 return; 2261 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2262 if (ipv6_addr_any(&addr)) 2263 return; 2264 __ipv6_dev_ac_inc(ifp->idev, &addr); 2265 } 2266 2267 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp) 2268 { 2269 struct in6_addr addr; 2270 2271 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2272 return; 2273 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2274 if (ipv6_addr_any(&addr)) 2275 return; 2276 __ipv6_dev_ac_dec(ifp->idev, &addr); 2277 } 2278 2279 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev) 2280 { 2281 switch (dev->addr_len) { 2282 case ETH_ALEN: 2283 memcpy(eui, dev->dev_addr, 3); 2284 eui[3] = 0xFF; 2285 eui[4] = 0xFE; 2286 memcpy(eui + 5, dev->dev_addr + 3, 3); 2287 break; 2288 case EUI64_ADDR_LEN: 2289 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN); 2290 eui[0] ^= 2; 2291 break; 2292 default: 2293 return -1; 2294 } 2295 2296 return 0; 2297 } 2298 2299 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev) 2300 { 2301 const union fwnet_hwaddr *ha; 2302 2303 if (dev->addr_len != FWNET_ALEN) 2304 return -1; 2305 2306 ha = (const union fwnet_hwaddr *)dev->dev_addr; 2307 2308 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id)); 2309 eui[0] ^= 2; 2310 return 0; 2311 } 2312 2313 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev) 2314 { 2315 /* XXX: inherit EUI-64 from other interface -- yoshfuji */ 2316 if (dev->addr_len != ARCNET_ALEN) 2317 return -1; 2318 memset(eui, 0, 7); 2319 eui[7] = *(u8 *)dev->dev_addr; 2320 return 0; 2321 } 2322 2323 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev) 2324 { 2325 if (dev->addr_len != INFINIBAND_ALEN) 2326 return -1; 2327 memcpy(eui, dev->dev_addr + 12, 8); 2328 eui[0] |= 2; 2329 return 0; 2330 } 2331 2332 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr) 2333 { 2334 if (addr == 0) 2335 return -1; 2336 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) || 2337 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) || 2338 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) || 2339 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) || 2340 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) || 2341 ipv4_is_lbcast(addr)) ? 0x00 : 0x02; 2342 eui[1] = 0; 2343 eui[2] = 0x5E; 2344 eui[3] = 0xFE; 2345 memcpy(eui + 4, &addr, 4); 2346 return 0; 2347 } 2348 2349 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev) 2350 { 2351 if (dev->priv_flags & IFF_ISATAP) 2352 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2353 return -1; 2354 } 2355 2356 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev) 2357 { 2358 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2359 } 2360 2361 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev) 2362 { 2363 memcpy(eui, dev->perm_addr, 3); 2364 memcpy(eui + 5, dev->perm_addr + 3, 3); 2365 eui[3] = 0xFF; 2366 eui[4] = 0xFE; 2367 eui[0] ^= 2; 2368 return 0; 2369 } 2370 2371 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev) 2372 { 2373 switch (dev->type) { 2374 case ARPHRD_ETHER: 2375 case ARPHRD_FDDI: 2376 return addrconf_ifid_eui48(eui, dev); 2377 case ARPHRD_ARCNET: 2378 return addrconf_ifid_arcnet(eui, dev); 2379 case ARPHRD_INFINIBAND: 2380 return addrconf_ifid_infiniband(eui, dev); 2381 case ARPHRD_SIT: 2382 return addrconf_ifid_sit(eui, dev); 2383 case ARPHRD_IPGRE: 2384 case ARPHRD_TUNNEL: 2385 return addrconf_ifid_gre(eui, dev); 2386 case ARPHRD_6LOWPAN: 2387 return addrconf_ifid_6lowpan(eui, dev); 2388 case ARPHRD_IEEE1394: 2389 return addrconf_ifid_ieee1394(eui, dev); 2390 case ARPHRD_TUNNEL6: 2391 case ARPHRD_IP6GRE: 2392 case ARPHRD_RAWIP: 2393 return addrconf_ifid_ip6tnl(eui, dev); 2394 } 2395 return -1; 2396 } 2397 2398 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev) 2399 { 2400 int err = -1; 2401 struct inet6_ifaddr *ifp; 2402 2403 read_lock_bh(&idev->lock); 2404 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 2405 if (ifp->scope > IFA_LINK) 2406 break; 2407 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) { 2408 memcpy(eui, ifp->addr.s6_addr+8, 8); 2409 err = 0; 2410 break; 2411 } 2412 } 2413 read_unlock_bh(&idev->lock); 2414 return err; 2415 } 2416 2417 /* Generation of a randomized Interface Identifier 2418 * draft-ietf-6man-rfc4941bis, Section 3.3.1 2419 */ 2420 2421 static void ipv6_gen_rnd_iid(struct in6_addr *addr) 2422 { 2423 regen: 2424 get_random_bytes(&addr->s6_addr[8], 8); 2425 2426 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1: 2427 * check if generated address is not inappropriate: 2428 * 2429 * - Reserved IPv6 Interface Identifiers 2430 * - XXX: already assigned to an address on the device 2431 */ 2432 2433 /* Subnet-router anycast: 0000:0000:0000:0000 */ 2434 if (!(addr->s6_addr32[2] | addr->s6_addr32[3])) 2435 goto regen; 2436 2437 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212 2438 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213 2439 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF 2440 */ 2441 if (ntohl(addr->s6_addr32[2]) == 0x02005eff && 2442 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000) 2443 goto regen; 2444 2445 /* Reserved subnet anycast addresses */ 2446 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff && 2447 ntohl(addr->s6_addr32[3]) >= 0Xffffff80) 2448 goto regen; 2449 } 2450 2451 /* 2452 * Add prefix route. 2453 */ 2454 2455 static void 2456 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric, 2457 struct net_device *dev, unsigned long expires, 2458 u32 flags, gfp_t gfp_flags) 2459 { 2460 struct fib6_config cfg = { 2461 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX, 2462 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF, 2463 .fc_ifindex = dev->ifindex, 2464 .fc_expires = expires, 2465 .fc_dst_len = plen, 2466 .fc_flags = RTF_UP | flags, 2467 .fc_nlinfo.nl_net = dev_net(dev), 2468 .fc_protocol = RTPROT_KERNEL, 2469 .fc_type = RTN_UNICAST, 2470 }; 2471 2472 cfg.fc_dst = *pfx; 2473 2474 /* Prevent useless cloning on PtP SIT. 2475 This thing is done here expecting that the whole 2476 class of non-broadcast devices need not cloning. 2477 */ 2478 #if IS_ENABLED(CONFIG_IPV6_SIT) 2479 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT)) 2480 cfg.fc_flags |= RTF_NONEXTHOP; 2481 #endif 2482 2483 ip6_route_add(&cfg, gfp_flags, NULL); 2484 } 2485 2486 2487 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 2488 int plen, 2489 const struct net_device *dev, 2490 u32 flags, u32 noflags, 2491 bool no_gw) 2492 { 2493 struct fib6_node *fn; 2494 struct fib6_info *rt = NULL; 2495 struct fib6_table *table; 2496 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX; 2497 2498 table = fib6_get_table(dev_net(dev), tb_id); 2499 if (!table) 2500 return NULL; 2501 2502 rcu_read_lock(); 2503 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true); 2504 if (!fn) 2505 goto out; 2506 2507 for_each_fib6_node_rt_rcu(fn) { 2508 /* prefix routes only use builtin fib6_nh */ 2509 if (rt->nh) 2510 continue; 2511 2512 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex) 2513 continue; 2514 if (no_gw && rt->fib6_nh->fib_nh_gw_family) 2515 continue; 2516 if ((rt->fib6_flags & flags) != flags) 2517 continue; 2518 if ((rt->fib6_flags & noflags) != 0) 2519 continue; 2520 if (!fib6_info_hold_safe(rt)) 2521 continue; 2522 break; 2523 } 2524 out: 2525 rcu_read_unlock(); 2526 return rt; 2527 } 2528 2529 2530 /* Create "default" multicast route to the interface */ 2531 2532 static void addrconf_add_mroute(struct net_device *dev) 2533 { 2534 struct fib6_config cfg = { 2535 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL, 2536 .fc_metric = IP6_RT_PRIO_ADDRCONF, 2537 .fc_ifindex = dev->ifindex, 2538 .fc_dst_len = 8, 2539 .fc_flags = RTF_UP, 2540 .fc_type = RTN_MULTICAST, 2541 .fc_nlinfo.nl_net = dev_net(dev), 2542 .fc_protocol = RTPROT_KERNEL, 2543 }; 2544 2545 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0); 2546 2547 ip6_route_add(&cfg, GFP_KERNEL, NULL); 2548 } 2549 2550 static struct inet6_dev *addrconf_add_dev(struct net_device *dev) 2551 { 2552 struct inet6_dev *idev; 2553 2554 ASSERT_RTNL(); 2555 2556 idev = ipv6_find_idev(dev); 2557 if (IS_ERR(idev)) 2558 return idev; 2559 2560 if (idev->cnf.disable_ipv6) 2561 return ERR_PTR(-EACCES); 2562 2563 /* Add default multicast route */ 2564 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev)) 2565 addrconf_add_mroute(dev); 2566 2567 return idev; 2568 } 2569 2570 static void delete_tempaddrs(struct inet6_dev *idev, 2571 struct inet6_ifaddr *ifp) 2572 { 2573 struct inet6_ifaddr *ift, *tmp; 2574 2575 write_lock_bh(&idev->lock); 2576 list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) { 2577 if (ift->ifpub != ifp) 2578 continue; 2579 2580 in6_ifa_hold(ift); 2581 write_unlock_bh(&idev->lock); 2582 ipv6_del_addr(ift); 2583 write_lock_bh(&idev->lock); 2584 } 2585 write_unlock_bh(&idev->lock); 2586 } 2587 2588 static void manage_tempaddrs(struct inet6_dev *idev, 2589 struct inet6_ifaddr *ifp, 2590 __u32 valid_lft, __u32 prefered_lft, 2591 bool create, unsigned long now) 2592 { 2593 u32 flags; 2594 struct inet6_ifaddr *ift; 2595 2596 read_lock_bh(&idev->lock); 2597 /* update all temporary addresses in the list */ 2598 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) { 2599 int age, max_valid, max_prefered; 2600 2601 if (ifp != ift->ifpub) 2602 continue; 2603 2604 /* RFC 4941 section 3.3: 2605 * If a received option will extend the lifetime of a public 2606 * address, the lifetimes of temporary addresses should 2607 * be extended, subject to the overall constraint that no 2608 * temporary addresses should ever remain "valid" or "preferred" 2609 * for a time longer than (TEMP_VALID_LIFETIME) or 2610 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively. 2611 */ 2612 age = (now - ift->cstamp) / HZ; 2613 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age; 2614 if (max_valid < 0) 2615 max_valid = 0; 2616 2617 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) - 2618 idev->desync_factor - age; 2619 if (max_prefered < 0) 2620 max_prefered = 0; 2621 2622 if (valid_lft > max_valid) 2623 valid_lft = max_valid; 2624 2625 if (prefered_lft > max_prefered) 2626 prefered_lft = max_prefered; 2627 2628 spin_lock(&ift->lock); 2629 flags = ift->flags; 2630 ift->valid_lft = valid_lft; 2631 ift->prefered_lft = prefered_lft; 2632 ift->tstamp = now; 2633 if (prefered_lft > 0) 2634 ift->flags &= ~IFA_F_DEPRECATED; 2635 2636 spin_unlock(&ift->lock); 2637 if (!(flags&IFA_F_TENTATIVE)) 2638 ipv6_ifa_notify(0, ift); 2639 } 2640 2641 /* Also create a temporary address if it's enabled but no temporary 2642 * address currently exists. 2643 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false 2644 * as part of cleanup (ie. deleting the mngtmpaddr). 2645 * We don't want that to result in creating a new temporary ip address. 2646 */ 2647 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft)) 2648 create = true; 2649 2650 if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) { 2651 /* When a new public address is created as described 2652 * in [ADDRCONF], also create a new temporary address. 2653 */ 2654 read_unlock_bh(&idev->lock); 2655 ipv6_create_tempaddr(ifp, false); 2656 } else { 2657 read_unlock_bh(&idev->lock); 2658 } 2659 } 2660 2661 static bool is_addr_mode_generate_stable(struct inet6_dev *idev) 2662 { 2663 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY || 2664 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM; 2665 } 2666 2667 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev, 2668 const struct prefix_info *pinfo, 2669 struct inet6_dev *in6_dev, 2670 const struct in6_addr *addr, int addr_type, 2671 u32 addr_flags, bool sllao, bool tokenized, 2672 __u32 valid_lft, u32 prefered_lft) 2673 { 2674 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1); 2675 int create = 0, update_lft = 0; 2676 2677 if (!ifp && valid_lft) { 2678 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses); 2679 struct ifa6_config cfg = { 2680 .pfx = addr, 2681 .plen = pinfo->prefix_len, 2682 .ifa_flags = addr_flags, 2683 .valid_lft = valid_lft, 2684 .preferred_lft = prefered_lft, 2685 .scope = addr_type & IPV6_ADDR_SCOPE_MASK, 2686 .ifa_proto = IFAPROT_KERNEL_RA 2687 }; 2688 2689 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 2690 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) || 2691 READ_ONCE(in6_dev->cnf.optimistic_dad)) && 2692 !net->ipv6.devconf_all->forwarding && sllao) 2693 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 2694 #endif 2695 2696 /* Do not allow to create too much of autoconfigured 2697 * addresses; this would be too easy way to crash kernel. 2698 */ 2699 if (!max_addresses || 2700 ipv6_count_addresses(in6_dev) < max_addresses) 2701 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL); 2702 2703 if (IS_ERR_OR_NULL(ifp)) 2704 return -1; 2705 2706 create = 1; 2707 spin_lock_bh(&ifp->lock); 2708 ifp->flags |= IFA_F_MANAGETEMPADDR; 2709 ifp->cstamp = jiffies; 2710 ifp->tokenized = tokenized; 2711 spin_unlock_bh(&ifp->lock); 2712 addrconf_dad_start(ifp); 2713 } 2714 2715 if (ifp) { 2716 u32 flags; 2717 unsigned long now; 2718 u32 stored_lft; 2719 2720 /* update lifetime (RFC2462 5.5.3 e) */ 2721 spin_lock_bh(&ifp->lock); 2722 now = jiffies; 2723 if (ifp->valid_lft > (now - ifp->tstamp) / HZ) 2724 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ; 2725 else 2726 stored_lft = 0; 2727 2728 /* RFC4862 Section 5.5.3e: 2729 * "Note that the preferred lifetime of the 2730 * corresponding address is always reset to 2731 * the Preferred Lifetime in the received 2732 * Prefix Information option, regardless of 2733 * whether the valid lifetime is also reset or 2734 * ignored." 2735 * 2736 * So we should always update prefered_lft here. 2737 */ 2738 update_lft = !create && stored_lft; 2739 2740 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) { 2741 const u32 minimum_lft = min_t(u32, 2742 stored_lft, MIN_VALID_LIFETIME); 2743 valid_lft = max(valid_lft, minimum_lft); 2744 } 2745 2746 if (update_lft) { 2747 ifp->valid_lft = valid_lft; 2748 ifp->prefered_lft = prefered_lft; 2749 WRITE_ONCE(ifp->tstamp, now); 2750 flags = ifp->flags; 2751 ifp->flags &= ~IFA_F_DEPRECATED; 2752 spin_unlock_bh(&ifp->lock); 2753 2754 if (!(flags&IFA_F_TENTATIVE)) 2755 ipv6_ifa_notify(0, ifp); 2756 } else 2757 spin_unlock_bh(&ifp->lock); 2758 2759 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft, 2760 create, now); 2761 2762 in6_ifa_put(ifp); 2763 addrconf_verify(net); 2764 } 2765 2766 return 0; 2767 } 2768 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr); 2769 2770 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao) 2771 { 2772 struct prefix_info *pinfo; 2773 struct fib6_table *table; 2774 __u32 valid_lft; 2775 __u32 prefered_lft; 2776 int addr_type, err; 2777 u32 addr_flags = 0; 2778 struct inet6_dev *in6_dev; 2779 struct net *net = dev_net(dev); 2780 bool ignore_autoconf = false; 2781 2782 pinfo = (struct prefix_info *) opt; 2783 2784 if (len < sizeof(struct prefix_info)) { 2785 netdev_dbg(dev, "addrconf: prefix option too short\n"); 2786 return; 2787 } 2788 2789 /* 2790 * Validation checks ([ADDRCONF], page 19) 2791 */ 2792 2793 addr_type = ipv6_addr_type(&pinfo->prefix); 2794 2795 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)) 2796 return; 2797 2798 valid_lft = ntohl(pinfo->valid); 2799 prefered_lft = ntohl(pinfo->prefered); 2800 2801 if (prefered_lft > valid_lft) { 2802 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n"); 2803 return; 2804 } 2805 2806 in6_dev = in6_dev_get(dev); 2807 2808 if (!in6_dev) { 2809 net_dbg_ratelimited("addrconf: device %s not configured\n", 2810 dev->name); 2811 return; 2812 } 2813 2814 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft) 2815 goto put; 2816 2817 /* 2818 * Two things going on here: 2819 * 1) Add routes for on-link prefixes 2820 * 2) Configure prefixes with the auto flag set 2821 */ 2822 2823 if (pinfo->onlink) { 2824 struct fib6_info *rt; 2825 unsigned long rt_expires; 2826 2827 /* Avoid arithmetic overflow. Really, we could 2828 * save rt_expires in seconds, likely valid_lft, 2829 * but it would require division in fib gc, that it 2830 * not good. 2831 */ 2832 if (HZ > USER_HZ) 2833 rt_expires = addrconf_timeout_fixup(valid_lft, HZ); 2834 else 2835 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ); 2836 2837 if (addrconf_finite_timeout(rt_expires)) 2838 rt_expires *= HZ; 2839 2840 rt = addrconf_get_prefix_route(&pinfo->prefix, 2841 pinfo->prefix_len, 2842 dev, 2843 RTF_ADDRCONF | RTF_PREFIX_RT, 2844 RTF_DEFAULT, true); 2845 2846 if (rt) { 2847 /* Autoconf prefix route */ 2848 if (valid_lft == 0) { 2849 ip6_del_rt(net, rt, false); 2850 rt = NULL; 2851 } else { 2852 table = rt->fib6_table; 2853 spin_lock_bh(&table->tb6_lock); 2854 2855 if (addrconf_finite_timeout(rt_expires)) { 2856 /* not infinity */ 2857 fib6_set_expires(rt, jiffies + rt_expires); 2858 fib6_add_gc_list(rt); 2859 } else { 2860 fib6_clean_expires(rt); 2861 fib6_remove_gc_list(rt); 2862 } 2863 2864 spin_unlock_bh(&table->tb6_lock); 2865 } 2866 } else if (valid_lft) { 2867 clock_t expires = 0; 2868 int flags = RTF_ADDRCONF | RTF_PREFIX_RT; 2869 if (addrconf_finite_timeout(rt_expires)) { 2870 /* not infinity */ 2871 flags |= RTF_EXPIRES; 2872 expires = jiffies_to_clock_t(rt_expires); 2873 } 2874 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len, 2875 0, dev, expires, flags, 2876 GFP_ATOMIC); 2877 } 2878 fib6_info_release(rt); 2879 } 2880 2881 /* Try to figure out our local address for this prefix */ 2882 2883 ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd; 2884 if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) { 2885 struct in6_addr addr; 2886 bool tokenized = false, dev_addr_generated = false; 2887 2888 if (pinfo->prefix_len == 64) { 2889 memcpy(&addr, &pinfo->prefix, 8); 2890 2891 if (!ipv6_addr_any(&in6_dev->token)) { 2892 read_lock_bh(&in6_dev->lock); 2893 memcpy(addr.s6_addr + 8, 2894 in6_dev->token.s6_addr + 8, 8); 2895 read_unlock_bh(&in6_dev->lock); 2896 tokenized = true; 2897 } else if (is_addr_mode_generate_stable(in6_dev) && 2898 !ipv6_generate_stable_address(&addr, 0, 2899 in6_dev)) { 2900 addr_flags |= IFA_F_STABLE_PRIVACY; 2901 goto ok; 2902 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) && 2903 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) { 2904 goto put; 2905 } else { 2906 dev_addr_generated = true; 2907 } 2908 goto ok; 2909 } 2910 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n", 2911 pinfo->prefix_len); 2912 goto put; 2913 2914 ok: 2915 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, 2916 &addr, addr_type, 2917 addr_flags, sllao, 2918 tokenized, valid_lft, 2919 prefered_lft); 2920 if (err) 2921 goto put; 2922 2923 /* Ignore error case here because previous prefix add addr was 2924 * successful which will be notified. 2925 */ 2926 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr, 2927 addr_type, addr_flags, sllao, 2928 tokenized, valid_lft, 2929 prefered_lft, 2930 dev_addr_generated); 2931 } 2932 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo); 2933 put: 2934 in6_dev_put(in6_dev); 2935 } 2936 2937 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev, 2938 struct in6_ifreq *ireq) 2939 { 2940 struct ip_tunnel_parm_kern p = { }; 2941 int err; 2942 2943 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4)) 2944 return -EADDRNOTAVAIL; 2945 2946 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3]; 2947 p.iph.version = 4; 2948 p.iph.ihl = 5; 2949 p.iph.protocol = IPPROTO_IPV6; 2950 p.iph.ttl = 64; 2951 2952 if (!dev->netdev_ops->ndo_tunnel_ctl) 2953 return -EOPNOTSUPP; 2954 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL); 2955 if (err) 2956 return err; 2957 2958 dev = __dev_get_by_name(net, p.name); 2959 if (!dev) 2960 return -ENOBUFS; 2961 return dev_open(dev, NULL); 2962 } 2963 2964 /* 2965 * Set destination address. 2966 * Special case for SIT interfaces where we create a new "virtual" 2967 * device. 2968 */ 2969 int addrconf_set_dstaddr(struct net *net, void __user *arg) 2970 { 2971 struct net_device *dev; 2972 struct in6_ifreq ireq; 2973 int err = -ENODEV; 2974 2975 if (!IS_ENABLED(CONFIG_IPV6_SIT)) 2976 return -ENODEV; 2977 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 2978 return -EFAULT; 2979 2980 rtnl_net_lock(net); 2981 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 2982 if (dev && dev->type == ARPHRD_SIT) 2983 err = addrconf_set_sit_dstaddr(net, dev, &ireq); 2984 rtnl_net_unlock(net); 2985 return err; 2986 } 2987 2988 static int ipv6_mc_config(struct sock *sk, bool join, 2989 const struct in6_addr *addr, int ifindex) 2990 { 2991 int ret; 2992 2993 ASSERT_RTNL(); 2994 2995 lock_sock(sk); 2996 if (join) 2997 ret = ipv6_sock_mc_join(sk, ifindex, addr); 2998 else 2999 ret = ipv6_sock_mc_drop(sk, ifindex, addr); 3000 release_sock(sk); 3001 3002 return ret; 3003 } 3004 3005 /* 3006 * Manual configuration of address on an interface 3007 */ 3008 static int inet6_addr_add(struct net *net, struct net_device *dev, 3009 struct ifa6_config *cfg, clock_t expires, u32 flags, 3010 struct netlink_ext_ack *extack) 3011 { 3012 struct inet6_ifaddr *ifp; 3013 struct inet6_dev *idev; 3014 3015 ASSERT_RTNL_NET(net); 3016 3017 if (cfg->plen > 128) { 3018 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 3019 return -EINVAL; 3020 } 3021 3022 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) { 3023 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64"); 3024 return -EINVAL; 3025 } 3026 3027 idev = addrconf_add_dev(dev); 3028 if (IS_ERR(idev)) { 3029 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 3030 return PTR_ERR(idev); 3031 } 3032 3033 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 3034 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk, 3035 true, cfg->pfx, dev->ifindex); 3036 3037 if (ret < 0) { 3038 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed"); 3039 return ret; 3040 } 3041 } 3042 3043 cfg->scope = ipv6_addr_scope(cfg->pfx); 3044 3045 ifp = ipv6_add_addr(idev, cfg, true, extack); 3046 if (!IS_ERR(ifp)) { 3047 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 3048 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3049 ifp->rt_priority, dev, expires, 3050 flags, GFP_KERNEL); 3051 } 3052 3053 /* Send a netlink notification if DAD is enabled and 3054 * optimistic flag is not set 3055 */ 3056 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD))) 3057 ipv6_ifa_notify(0, ifp); 3058 /* 3059 * Note that section 3.1 of RFC 4429 indicates 3060 * that the Optimistic flag should not be set for 3061 * manually configured addresses 3062 */ 3063 addrconf_dad_start(ifp); 3064 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR) 3065 manage_tempaddrs(idev, ifp, cfg->valid_lft, 3066 cfg->preferred_lft, true, jiffies); 3067 in6_ifa_put(ifp); 3068 addrconf_verify_rtnl(net); 3069 return 0; 3070 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 3071 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false, 3072 cfg->pfx, dev->ifindex); 3073 } 3074 3075 return PTR_ERR(ifp); 3076 } 3077 3078 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags, 3079 const struct in6_addr *pfx, unsigned int plen, 3080 struct netlink_ext_ack *extack) 3081 { 3082 struct inet6_ifaddr *ifp; 3083 struct inet6_dev *idev; 3084 struct net_device *dev; 3085 3086 if (plen > 128) { 3087 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 3088 return -EINVAL; 3089 } 3090 3091 dev = __dev_get_by_index(net, ifindex); 3092 if (!dev) { 3093 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 3094 return -ENODEV; 3095 } 3096 3097 idev = __in6_dev_get_rtnl_net(dev); 3098 if (!idev) { 3099 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 3100 return -ENXIO; 3101 } 3102 3103 read_lock_bh(&idev->lock); 3104 list_for_each_entry(ifp, &idev->addr_list, if_list) { 3105 if (ifp->prefix_len == plen && 3106 ipv6_addr_equal(pfx, &ifp->addr)) { 3107 in6_ifa_hold(ifp); 3108 read_unlock_bh(&idev->lock); 3109 3110 ipv6_del_addr(ifp); 3111 3112 if (!(ifp->flags & IFA_F_TEMPORARY) && 3113 (ifp->flags & IFA_F_MANAGETEMPADDR)) 3114 delete_tempaddrs(idev, ifp); 3115 3116 addrconf_verify_rtnl(net); 3117 if (ipv6_addr_is_multicast(pfx)) { 3118 ipv6_mc_config(net->ipv6.mc_autojoin_sk, 3119 false, pfx, dev->ifindex); 3120 } 3121 return 0; 3122 } 3123 } 3124 read_unlock_bh(&idev->lock); 3125 3126 NL_SET_ERR_MSG_MOD(extack, "address not found"); 3127 return -EADDRNOTAVAIL; 3128 } 3129 3130 3131 int addrconf_add_ifaddr(struct net *net, void __user *arg) 3132 { 3133 struct ifa6_config cfg = { 3134 .ifa_flags = IFA_F_PERMANENT, 3135 .preferred_lft = INFINITY_LIFE_TIME, 3136 .valid_lft = INFINITY_LIFE_TIME, 3137 }; 3138 struct net_device *dev; 3139 struct in6_ifreq ireq; 3140 int err; 3141 3142 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 3143 return -EPERM; 3144 3145 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3146 return -EFAULT; 3147 3148 cfg.pfx = &ireq.ifr6_addr; 3149 cfg.plen = ireq.ifr6_prefixlen; 3150 3151 rtnl_net_lock(net); 3152 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 3153 if (dev) { 3154 netdev_lock_ops(dev); 3155 err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL); 3156 netdev_unlock_ops(dev); 3157 } else { 3158 err = -ENODEV; 3159 } 3160 rtnl_net_unlock(net); 3161 return err; 3162 } 3163 3164 int addrconf_del_ifaddr(struct net *net, void __user *arg) 3165 { 3166 struct in6_ifreq ireq; 3167 int err; 3168 3169 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 3170 return -EPERM; 3171 3172 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3173 return -EFAULT; 3174 3175 rtnl_net_lock(net); 3176 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr, 3177 ireq.ifr6_prefixlen, NULL); 3178 rtnl_net_unlock(net); 3179 return err; 3180 } 3181 3182 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr, 3183 int plen, int scope, u8 proto) 3184 { 3185 struct inet6_ifaddr *ifp; 3186 struct ifa6_config cfg = { 3187 .pfx = addr, 3188 .plen = plen, 3189 .ifa_flags = IFA_F_PERMANENT, 3190 .valid_lft = INFINITY_LIFE_TIME, 3191 .preferred_lft = INFINITY_LIFE_TIME, 3192 .scope = scope, 3193 .ifa_proto = proto 3194 }; 3195 3196 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3197 if (!IS_ERR(ifp)) { 3198 spin_lock_bh(&ifp->lock); 3199 ifp->flags &= ~IFA_F_TENTATIVE; 3200 spin_unlock_bh(&ifp->lock); 3201 rt_genid_bump_ipv6(dev_net(idev->dev)); 3202 ipv6_ifa_notify(RTM_NEWADDR, ifp); 3203 in6_ifa_put(ifp); 3204 } 3205 } 3206 3207 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) 3208 static void add_v4_addrs(struct inet6_dev *idev) 3209 { 3210 struct in6_addr addr; 3211 struct net_device *dev; 3212 struct net *net = dev_net(idev->dev); 3213 int scope, plen; 3214 u32 pflags = 0; 3215 3216 ASSERT_RTNL(); 3217 3218 memset(&addr, 0, sizeof(struct in6_addr)); 3219 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4); 3220 3221 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) { 3222 scope = IPV6_ADDR_COMPATv4; 3223 plen = 96; 3224 pflags |= RTF_NONEXTHOP; 3225 } else { 3226 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE) 3227 return; 3228 3229 addr.s6_addr32[0] = htonl(0xfe800000); 3230 scope = IFA_LINK; 3231 plen = 64; 3232 } 3233 3234 if (addr.s6_addr32[3]) { 3235 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC); 3236 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags, 3237 GFP_KERNEL); 3238 return; 3239 } 3240 3241 for_each_netdev(net, dev) { 3242 struct in_device *in_dev = __in_dev_get_rtnl(dev); 3243 if (in_dev && (dev->flags & IFF_UP)) { 3244 struct in_ifaddr *ifa; 3245 int flag = scope; 3246 3247 in_dev_for_each_ifa_rtnl(ifa, in_dev) { 3248 addr.s6_addr32[3] = ifa->ifa_local; 3249 3250 if (ifa->ifa_scope == RT_SCOPE_LINK) 3251 continue; 3252 if (ifa->ifa_scope >= RT_SCOPE_HOST) { 3253 if (idev->dev->flags&IFF_POINTOPOINT) 3254 continue; 3255 flag |= IFA_HOST; 3256 } 3257 3258 add_addr(idev, &addr, plen, flag, 3259 IFAPROT_UNSPEC); 3260 addrconf_prefix_route(&addr, plen, 0, idev->dev, 3261 0, pflags, GFP_KERNEL); 3262 } 3263 } 3264 } 3265 } 3266 #endif 3267 3268 static void init_loopback(struct net_device *dev) 3269 { 3270 struct inet6_dev *idev; 3271 3272 /* ::1 */ 3273 3274 ASSERT_RTNL(); 3275 3276 idev = ipv6_find_idev(dev); 3277 if (IS_ERR(idev)) { 3278 pr_debug("%s: add_dev failed\n", __func__); 3279 return; 3280 } 3281 3282 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO); 3283 } 3284 3285 void addrconf_add_linklocal(struct inet6_dev *idev, 3286 const struct in6_addr *addr, u32 flags) 3287 { 3288 struct ifa6_config cfg = { 3289 .pfx = addr, 3290 .plen = 64, 3291 .ifa_flags = flags | IFA_F_PERMANENT, 3292 .valid_lft = INFINITY_LIFE_TIME, 3293 .preferred_lft = INFINITY_LIFE_TIME, 3294 .scope = IFA_LINK, 3295 .ifa_proto = IFAPROT_KERNEL_LL 3296 }; 3297 struct inet6_ifaddr *ifp; 3298 3299 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 3300 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) || 3301 READ_ONCE(idev->cnf.optimistic_dad)) && 3302 !dev_net(idev->dev)->ipv6.devconf_all->forwarding) 3303 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 3304 #endif 3305 3306 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3307 if (!IS_ERR(ifp)) { 3308 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev, 3309 0, 0, GFP_ATOMIC); 3310 addrconf_dad_start(ifp); 3311 in6_ifa_put(ifp); 3312 } 3313 } 3314 EXPORT_SYMBOL_GPL(addrconf_add_linklocal); 3315 3316 static bool ipv6_reserved_interfaceid(struct in6_addr address) 3317 { 3318 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0) 3319 return true; 3320 3321 if (address.s6_addr32[2] == htonl(0x02005eff) && 3322 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000))) 3323 return true; 3324 3325 if (address.s6_addr32[2] == htonl(0xfdffffff) && 3326 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80))) 3327 return true; 3328 3329 return false; 3330 } 3331 3332 static int ipv6_generate_stable_address(struct in6_addr *address, 3333 u8 dad_count, 3334 const struct inet6_dev *idev) 3335 { 3336 static DEFINE_SPINLOCK(lock); 3337 static __u32 digest[SHA1_DIGEST_WORDS]; 3338 static __u32 workspace[SHA1_WORKSPACE_WORDS]; 3339 3340 static union { 3341 char __data[SHA1_BLOCK_SIZE]; 3342 struct { 3343 struct in6_addr secret; 3344 __be32 prefix[2]; 3345 unsigned char hwaddr[MAX_ADDR_LEN]; 3346 u8 dad_count; 3347 } __packed; 3348 } data; 3349 3350 struct in6_addr secret; 3351 struct in6_addr temp; 3352 struct net *net = dev_net(idev->dev); 3353 3354 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data)); 3355 3356 if (idev->cnf.stable_secret.initialized) 3357 secret = idev->cnf.stable_secret.secret; 3358 else if (net->ipv6.devconf_dflt->stable_secret.initialized) 3359 secret = net->ipv6.devconf_dflt->stable_secret.secret; 3360 else 3361 return -1; 3362 3363 retry: 3364 spin_lock_bh(&lock); 3365 3366 sha1_init(digest); 3367 memset(&data, 0, sizeof(data)); 3368 memset(workspace, 0, sizeof(workspace)); 3369 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len); 3370 data.prefix[0] = address->s6_addr32[0]; 3371 data.prefix[1] = address->s6_addr32[1]; 3372 data.secret = secret; 3373 data.dad_count = dad_count; 3374 3375 sha1_transform(digest, data.__data, workspace); 3376 3377 temp = *address; 3378 temp.s6_addr32[2] = (__force __be32)digest[0]; 3379 temp.s6_addr32[3] = (__force __be32)digest[1]; 3380 3381 spin_unlock_bh(&lock); 3382 3383 if (ipv6_reserved_interfaceid(temp)) { 3384 dad_count++; 3385 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries) 3386 return -1; 3387 goto retry; 3388 } 3389 3390 *address = temp; 3391 return 0; 3392 } 3393 3394 static void ipv6_gen_mode_random_init(struct inet6_dev *idev) 3395 { 3396 struct ipv6_stable_secret *s = &idev->cnf.stable_secret; 3397 3398 if (s->initialized) 3399 return; 3400 s = &idev->cnf.stable_secret; 3401 get_random_bytes(&s->secret, sizeof(s->secret)); 3402 s->initialized = true; 3403 } 3404 3405 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route) 3406 { 3407 struct in6_addr addr; 3408 3409 /* no link local addresses on L3 master devices */ 3410 if (netif_is_l3_master(idev->dev)) 3411 return; 3412 3413 /* no link local addresses on devices flagged as slaves */ 3414 if (idev->dev->priv_flags & IFF_NO_ADDRCONF) 3415 return; 3416 3417 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0); 3418 3419 switch (idev->cnf.addr_gen_mode) { 3420 case IN6_ADDR_GEN_MODE_RANDOM: 3421 ipv6_gen_mode_random_init(idev); 3422 fallthrough; 3423 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY: 3424 if (!ipv6_generate_stable_address(&addr, 0, idev)) 3425 addrconf_add_linklocal(idev, &addr, 3426 IFA_F_STABLE_PRIVACY); 3427 else if (prefix_route) 3428 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3429 0, 0, GFP_KERNEL); 3430 break; 3431 case IN6_ADDR_GEN_MODE_EUI64: 3432 /* addrconf_add_linklocal also adds a prefix_route and we 3433 * only need to care about prefix routes if ipv6_generate_eui64 3434 * couldn't generate one. 3435 */ 3436 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0) 3437 addrconf_add_linklocal(idev, &addr, 0); 3438 else if (prefix_route) 3439 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3440 0, 0, GFP_KERNEL); 3441 break; 3442 case IN6_ADDR_GEN_MODE_NONE: 3443 default: 3444 /* will not add any link local address */ 3445 break; 3446 } 3447 } 3448 3449 static void addrconf_dev_config(struct net_device *dev) 3450 { 3451 struct inet6_dev *idev; 3452 3453 ASSERT_RTNL(); 3454 3455 if ((dev->type != ARPHRD_ETHER) && 3456 (dev->type != ARPHRD_FDDI) && 3457 (dev->type != ARPHRD_ARCNET) && 3458 (dev->type != ARPHRD_INFINIBAND) && 3459 (dev->type != ARPHRD_IEEE1394) && 3460 (dev->type != ARPHRD_TUNNEL6) && 3461 (dev->type != ARPHRD_6LOWPAN) && 3462 (dev->type != ARPHRD_IP6GRE) && 3463 (dev->type != ARPHRD_TUNNEL) && 3464 (dev->type != ARPHRD_NONE) && 3465 (dev->type != ARPHRD_RAWIP)) { 3466 /* Alas, we support only Ethernet autoconfiguration. */ 3467 idev = __in6_dev_get(dev); 3468 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP && 3469 dev->flags & IFF_MULTICAST) 3470 ipv6_mc_up(idev); 3471 return; 3472 } 3473 3474 idev = addrconf_add_dev(dev); 3475 if (IS_ERR(idev)) 3476 return; 3477 3478 /* this device type has no EUI support */ 3479 if (dev->type == ARPHRD_NONE && 3480 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) 3481 WRITE_ONCE(idev->cnf.addr_gen_mode, 3482 IN6_ADDR_GEN_MODE_RANDOM); 3483 3484 addrconf_addr_gen(idev, false); 3485 } 3486 3487 #if IS_ENABLED(CONFIG_IPV6_SIT) 3488 static void addrconf_sit_config(struct net_device *dev) 3489 { 3490 struct inet6_dev *idev; 3491 3492 ASSERT_RTNL(); 3493 3494 /* 3495 * Configure the tunnel with one of our IPv4 3496 * addresses... we should configure all of 3497 * our v4 addrs in the tunnel 3498 */ 3499 3500 idev = ipv6_find_idev(dev); 3501 if (IS_ERR(idev)) { 3502 pr_debug("%s: add_dev failed\n", __func__); 3503 return; 3504 } 3505 3506 if (dev->priv_flags & IFF_ISATAP) { 3507 addrconf_addr_gen(idev, false); 3508 return; 3509 } 3510 3511 add_v4_addrs(idev); 3512 3513 if (dev->flags&IFF_POINTOPOINT) 3514 addrconf_add_mroute(dev); 3515 } 3516 #endif 3517 3518 #if IS_ENABLED(CONFIG_NET_IPGRE) 3519 static void addrconf_gre_config(struct net_device *dev) 3520 { 3521 struct inet6_dev *idev; 3522 3523 ASSERT_RTNL(); 3524 3525 idev = ipv6_find_idev(dev); 3526 if (IS_ERR(idev)) { 3527 pr_debug("%s: add_dev failed\n", __func__); 3528 return; 3529 } 3530 3531 /* Generate the IPv6 link-local address using addrconf_addr_gen(), 3532 * unless we have an IPv4 GRE device not bound to an IP address and 3533 * which is in EUI64 mode (as __ipv6_isatap_ifid() would fail in this 3534 * case). Such devices fall back to add_v4_addrs() instead. 3535 */ 3536 if (!(*(__be32 *)dev->dev_addr == 0 && 3537 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)) { 3538 addrconf_addr_gen(idev, true); 3539 return; 3540 } 3541 3542 add_v4_addrs(idev); 3543 3544 if (dev->flags & IFF_POINTOPOINT) 3545 addrconf_add_mroute(dev); 3546 } 3547 #endif 3548 3549 static void addrconf_init_auto_addrs(struct net_device *dev) 3550 { 3551 switch (dev->type) { 3552 #if IS_ENABLED(CONFIG_IPV6_SIT) 3553 case ARPHRD_SIT: 3554 addrconf_sit_config(dev); 3555 break; 3556 #endif 3557 #if IS_ENABLED(CONFIG_NET_IPGRE) 3558 case ARPHRD_IPGRE: 3559 addrconf_gre_config(dev); 3560 break; 3561 #endif 3562 case ARPHRD_LOOPBACK: 3563 init_loopback(dev); 3564 break; 3565 3566 default: 3567 addrconf_dev_config(dev); 3568 break; 3569 } 3570 } 3571 3572 static int fixup_permanent_addr(struct net *net, 3573 struct inet6_dev *idev, 3574 struct inet6_ifaddr *ifp) 3575 { 3576 /* !fib6_node means the host route was removed from the 3577 * FIB, for example, if 'lo' device is taken down. In that 3578 * case regenerate the host route. 3579 */ 3580 if (!ifp->rt || !ifp->rt->fib6_node) { 3581 struct fib6_info *f6i, *prev; 3582 3583 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false, 3584 GFP_ATOMIC, NULL); 3585 if (IS_ERR(f6i)) 3586 return PTR_ERR(f6i); 3587 3588 /* ifp->rt can be accessed outside of rtnl */ 3589 spin_lock(&ifp->lock); 3590 prev = ifp->rt; 3591 ifp->rt = f6i; 3592 spin_unlock(&ifp->lock); 3593 3594 fib6_info_release(prev); 3595 } 3596 3597 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) { 3598 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3599 ifp->rt_priority, idev->dev, 0, 0, 3600 GFP_ATOMIC); 3601 } 3602 3603 if (ifp->state == INET6_IFADDR_STATE_PREDAD) 3604 addrconf_dad_start(ifp); 3605 3606 return 0; 3607 } 3608 3609 static void addrconf_permanent_addr(struct net *net, struct net_device *dev) 3610 { 3611 struct inet6_ifaddr *ifp, *tmp; 3612 struct inet6_dev *idev; 3613 3614 idev = __in6_dev_get(dev); 3615 if (!idev) 3616 return; 3617 3618 write_lock_bh(&idev->lock); 3619 3620 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) { 3621 if ((ifp->flags & IFA_F_PERMANENT) && 3622 fixup_permanent_addr(net, idev, ifp) < 0) { 3623 write_unlock_bh(&idev->lock); 3624 in6_ifa_hold(ifp); 3625 ipv6_del_addr(ifp); 3626 write_lock_bh(&idev->lock); 3627 3628 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n", 3629 idev->dev->name, &ifp->addr); 3630 } 3631 } 3632 3633 write_unlock_bh(&idev->lock); 3634 } 3635 3636 static int addrconf_notify(struct notifier_block *this, unsigned long event, 3637 void *ptr) 3638 { 3639 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 3640 struct netdev_notifier_change_info *change_info; 3641 struct netdev_notifier_changeupper_info *info; 3642 struct inet6_dev *idev = __in6_dev_get(dev); 3643 struct net *net = dev_net(dev); 3644 int run_pending = 0; 3645 int err; 3646 3647 switch (event) { 3648 case NETDEV_REGISTER: 3649 if (!idev && dev->mtu >= IPV6_MIN_MTU) { 3650 idev = ipv6_add_dev(dev); 3651 if (IS_ERR(idev)) 3652 return notifier_from_errno(PTR_ERR(idev)); 3653 } 3654 break; 3655 3656 case NETDEV_CHANGEMTU: 3657 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */ 3658 if (dev->mtu < IPV6_MIN_MTU) { 3659 addrconf_ifdown(dev, dev != net->loopback_dev); 3660 break; 3661 } 3662 3663 if (idev) { 3664 rt6_mtu_change(dev, dev->mtu); 3665 WRITE_ONCE(idev->cnf.mtu6, dev->mtu); 3666 break; 3667 } 3668 3669 /* allocate new idev */ 3670 idev = ipv6_add_dev(dev); 3671 if (IS_ERR(idev)) 3672 break; 3673 3674 /* device is still not ready */ 3675 if (!(idev->if_flags & IF_READY)) 3676 break; 3677 3678 run_pending = 1; 3679 fallthrough; 3680 case NETDEV_UP: 3681 case NETDEV_CHANGE: 3682 if (idev && idev->cnf.disable_ipv6) 3683 break; 3684 3685 if (dev->priv_flags & IFF_NO_ADDRCONF) { 3686 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) && 3687 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST) 3688 ipv6_mc_up(idev); 3689 break; 3690 } 3691 3692 if (event == NETDEV_UP) { 3693 /* restore routes for permanent addresses */ 3694 addrconf_permanent_addr(net, dev); 3695 3696 if (!addrconf_link_ready(dev)) { 3697 /* device is not ready yet. */ 3698 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n", 3699 dev->name); 3700 break; 3701 } 3702 3703 if (!idev && dev->mtu >= IPV6_MIN_MTU) 3704 idev = ipv6_add_dev(dev); 3705 3706 if (!IS_ERR_OR_NULL(idev)) { 3707 idev->if_flags |= IF_READY; 3708 run_pending = 1; 3709 } 3710 } else if (event == NETDEV_CHANGE) { 3711 if (!addrconf_link_ready(dev)) { 3712 /* device is still not ready. */ 3713 rt6_sync_down_dev(dev, event); 3714 break; 3715 } 3716 3717 if (!IS_ERR_OR_NULL(idev)) { 3718 if (idev->if_flags & IF_READY) { 3719 /* device is already configured - 3720 * but resend MLD reports, we might 3721 * have roamed and need to update 3722 * multicast snooping switches 3723 */ 3724 ipv6_mc_up(idev); 3725 change_info = ptr; 3726 if (change_info->flags_changed & IFF_NOARP) 3727 addrconf_dad_run(idev, true); 3728 rt6_sync_up(dev, RTNH_F_LINKDOWN); 3729 break; 3730 } 3731 idev->if_flags |= IF_READY; 3732 } 3733 3734 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n", 3735 dev->name); 3736 3737 run_pending = 1; 3738 } 3739 3740 addrconf_init_auto_addrs(dev); 3741 3742 if (!IS_ERR_OR_NULL(idev)) { 3743 if (run_pending) 3744 addrconf_dad_run(idev, false); 3745 3746 /* Device has an address by now */ 3747 rt6_sync_up(dev, RTNH_F_DEAD); 3748 3749 /* 3750 * If the MTU changed during the interface down, 3751 * when the interface up, the changed MTU must be 3752 * reflected in the idev as well as routers. 3753 */ 3754 if (idev->cnf.mtu6 != dev->mtu && 3755 dev->mtu >= IPV6_MIN_MTU) { 3756 rt6_mtu_change(dev, dev->mtu); 3757 WRITE_ONCE(idev->cnf.mtu6, dev->mtu); 3758 } 3759 WRITE_ONCE(idev->tstamp, jiffies); 3760 inet6_ifinfo_notify(RTM_NEWLINK, idev); 3761 3762 /* 3763 * If the changed mtu during down is lower than 3764 * IPV6_MIN_MTU stop IPv6 on this interface. 3765 */ 3766 if (dev->mtu < IPV6_MIN_MTU) 3767 addrconf_ifdown(dev, dev != net->loopback_dev); 3768 } 3769 break; 3770 3771 case NETDEV_DOWN: 3772 case NETDEV_UNREGISTER: 3773 /* 3774 * Remove all addresses from this interface. 3775 */ 3776 addrconf_ifdown(dev, event != NETDEV_DOWN); 3777 break; 3778 3779 case NETDEV_CHANGENAME: 3780 if (idev) { 3781 snmp6_unregister_dev(idev); 3782 addrconf_sysctl_unregister(idev); 3783 err = addrconf_sysctl_register(idev); 3784 if (err) 3785 return notifier_from_errno(err); 3786 err = snmp6_register_dev(idev); 3787 if (err) { 3788 addrconf_sysctl_unregister(idev); 3789 return notifier_from_errno(err); 3790 } 3791 } 3792 break; 3793 3794 case NETDEV_PRE_TYPE_CHANGE: 3795 case NETDEV_POST_TYPE_CHANGE: 3796 if (idev) 3797 addrconf_type_change(dev, event); 3798 break; 3799 3800 case NETDEV_CHANGEUPPER: 3801 info = ptr; 3802 3803 /* flush all routes if dev is linked to or unlinked from 3804 * an L3 master device (e.g., VRF) 3805 */ 3806 if (info->upper_dev && netif_is_l3_master(info->upper_dev)) 3807 addrconf_ifdown(dev, false); 3808 } 3809 3810 return NOTIFY_OK; 3811 } 3812 3813 /* 3814 * addrconf module should be notified of a device going up 3815 */ 3816 static struct notifier_block ipv6_dev_notf = { 3817 .notifier_call = addrconf_notify, 3818 .priority = ADDRCONF_NOTIFY_PRIORITY, 3819 }; 3820 3821 static void addrconf_type_change(struct net_device *dev, unsigned long event) 3822 { 3823 struct inet6_dev *idev; 3824 ASSERT_RTNL(); 3825 3826 idev = __in6_dev_get(dev); 3827 3828 if (event == NETDEV_POST_TYPE_CHANGE) 3829 ipv6_mc_remap(idev); 3830 else if (event == NETDEV_PRE_TYPE_CHANGE) 3831 ipv6_mc_unmap(idev); 3832 } 3833 3834 static bool addr_is_local(const struct in6_addr *addr) 3835 { 3836 return ipv6_addr_type(addr) & 3837 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK); 3838 } 3839 3840 static int addrconf_ifdown(struct net_device *dev, bool unregister) 3841 { 3842 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN; 3843 struct net *net = dev_net(dev); 3844 struct inet6_dev *idev; 3845 struct inet6_ifaddr *ifa; 3846 LIST_HEAD(tmp_addr_list); 3847 bool keep_addr = false; 3848 bool was_ready; 3849 int state, i; 3850 3851 ASSERT_RTNL(); 3852 3853 rt6_disable_ip(dev, event); 3854 3855 idev = __in6_dev_get(dev); 3856 if (!idev) 3857 return -ENODEV; 3858 3859 /* 3860 * Step 1: remove reference to ipv6 device from parent device. 3861 * Do not dev_put! 3862 */ 3863 if (unregister) { 3864 WRITE_ONCE(idev->dead, 1); 3865 3866 /* protected by rtnl_lock */ 3867 RCU_INIT_POINTER(dev->ip6_ptr, NULL); 3868 3869 /* Step 1.5: remove snmp6 entry */ 3870 snmp6_unregister_dev(idev); 3871 3872 } 3873 3874 /* combine the user config with event to determine if permanent 3875 * addresses are to be removed from address hash table 3876 */ 3877 if (!unregister && !idev->cnf.disable_ipv6) { 3878 /* aggregate the system setting and interface setting */ 3879 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down); 3880 3881 if (!_keep_addr) 3882 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down); 3883 3884 keep_addr = (_keep_addr > 0); 3885 } 3886 3887 /* Step 2: clear hash table */ 3888 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 3889 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i]; 3890 3891 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 3892 restart: 3893 hlist_for_each_entry_rcu(ifa, h, addr_lst) { 3894 if (ifa->idev == idev) { 3895 addrconf_del_dad_work(ifa); 3896 /* combined flag + permanent flag decide if 3897 * address is retained on a down event 3898 */ 3899 if (!keep_addr || 3900 !(ifa->flags & IFA_F_PERMANENT) || 3901 addr_is_local(&ifa->addr)) { 3902 hlist_del_init_rcu(&ifa->addr_lst); 3903 goto restart; 3904 } 3905 } 3906 } 3907 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 3908 } 3909 3910 write_lock_bh(&idev->lock); 3911 3912 addrconf_del_rs_timer(idev); 3913 3914 /* Step 2: clear flags for stateless addrconf, repeated down 3915 * detection 3916 */ 3917 was_ready = idev->if_flags & IF_READY; 3918 if (!unregister) 3919 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY); 3920 3921 /* Step 3: clear tempaddr list */ 3922 while (!list_empty(&idev->tempaddr_list)) { 3923 ifa = list_first_entry(&idev->tempaddr_list, 3924 struct inet6_ifaddr, tmp_list); 3925 list_del(&ifa->tmp_list); 3926 write_unlock_bh(&idev->lock); 3927 spin_lock_bh(&ifa->lock); 3928 3929 if (ifa->ifpub) { 3930 in6_ifa_put(ifa->ifpub); 3931 ifa->ifpub = NULL; 3932 } 3933 spin_unlock_bh(&ifa->lock); 3934 in6_ifa_put(ifa); 3935 write_lock_bh(&idev->lock); 3936 } 3937 3938 list_for_each_entry(ifa, &idev->addr_list, if_list) 3939 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 3940 write_unlock_bh(&idev->lock); 3941 3942 while (!list_empty(&tmp_addr_list)) { 3943 struct fib6_info *rt = NULL; 3944 bool keep; 3945 3946 ifa = list_first_entry(&tmp_addr_list, 3947 struct inet6_ifaddr, if_list_aux); 3948 list_del(&ifa->if_list_aux); 3949 3950 addrconf_del_dad_work(ifa); 3951 3952 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) && 3953 !addr_is_local(&ifa->addr); 3954 3955 spin_lock_bh(&ifa->lock); 3956 3957 if (keep) { 3958 /* set state to skip the notifier below */ 3959 state = INET6_IFADDR_STATE_DEAD; 3960 ifa->state = INET6_IFADDR_STATE_PREDAD; 3961 if (!(ifa->flags & IFA_F_NODAD)) 3962 ifa->flags |= IFA_F_TENTATIVE; 3963 3964 rt = ifa->rt; 3965 ifa->rt = NULL; 3966 } else { 3967 state = ifa->state; 3968 ifa->state = INET6_IFADDR_STATE_DEAD; 3969 } 3970 3971 spin_unlock_bh(&ifa->lock); 3972 3973 if (rt) 3974 ip6_del_rt(net, rt, false); 3975 3976 if (state != INET6_IFADDR_STATE_DEAD) { 3977 __ipv6_ifa_notify(RTM_DELADDR, ifa); 3978 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa); 3979 } else { 3980 if (idev->cnf.forwarding) 3981 addrconf_leave_anycast(ifa); 3982 addrconf_leave_solict(ifa->idev, &ifa->addr); 3983 } 3984 3985 if (!keep) { 3986 write_lock_bh(&idev->lock); 3987 list_del_rcu(&ifa->if_list); 3988 write_unlock_bh(&idev->lock); 3989 in6_ifa_put(ifa); 3990 } 3991 } 3992 3993 /* Step 5: Discard anycast and multicast list */ 3994 if (unregister) { 3995 ipv6_ac_destroy_dev(idev); 3996 ipv6_mc_destroy_dev(idev); 3997 } else if (was_ready) { 3998 ipv6_mc_down(idev); 3999 } 4000 4001 WRITE_ONCE(idev->tstamp, jiffies); 4002 idev->ra_mtu = 0; 4003 4004 /* Last: Shot the device (if unregistered) */ 4005 if (unregister) { 4006 addrconf_sysctl_unregister(idev); 4007 neigh_parms_release(&nd_tbl, idev->nd_parms); 4008 neigh_ifdown(&nd_tbl, dev); 4009 in6_dev_put(idev); 4010 } 4011 return 0; 4012 } 4013 4014 static void addrconf_rs_timer(struct timer_list *t) 4015 { 4016 struct inet6_dev *idev = timer_container_of(idev, t, rs_timer); 4017 struct net_device *dev = idev->dev; 4018 struct in6_addr lladdr; 4019 int rtr_solicits; 4020 4021 write_lock(&idev->lock); 4022 if (idev->dead || !(idev->if_flags & IF_READY)) 4023 goto out; 4024 4025 if (!ipv6_accept_ra(idev)) 4026 goto out; 4027 4028 /* Announcement received after solicitation was sent */ 4029 if (idev->if_flags & IF_RA_RCVD) 4030 goto out; 4031 4032 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits); 4033 4034 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) { 4035 write_unlock(&idev->lock); 4036 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 4037 ndisc_send_rs(dev, &lladdr, 4038 &in6addr_linklocal_allrouters); 4039 else 4040 goto put; 4041 4042 write_lock(&idev->lock); 4043 idev->rs_interval = rfc3315_s14_backoff_update( 4044 idev->rs_interval, 4045 READ_ONCE(idev->cnf.rtr_solicit_max_interval)); 4046 /* The wait after the last probe can be shorter */ 4047 addrconf_mod_rs_timer(idev, (idev->rs_probes == 4048 READ_ONCE(idev->cnf.rtr_solicits)) ? 4049 READ_ONCE(idev->cnf.rtr_solicit_delay) : 4050 idev->rs_interval); 4051 } else { 4052 /* 4053 * Note: we do not support deprecated "all on-link" 4054 * assumption any longer. 4055 */ 4056 pr_debug("%s: no IPv6 routers present\n", idev->dev->name); 4057 } 4058 4059 out: 4060 write_unlock(&idev->lock); 4061 put: 4062 in6_dev_put(idev); 4063 } 4064 4065 /* 4066 * Duplicate Address Detection 4067 */ 4068 static void addrconf_dad_kick(struct inet6_ifaddr *ifp) 4069 { 4070 struct inet6_dev *idev = ifp->idev; 4071 unsigned long rand_num; 4072 u64 nonce; 4073 4074 if (ifp->flags & IFA_F_OPTIMISTIC) 4075 rand_num = 0; 4076 else 4077 rand_num = get_random_u32_below( 4078 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1); 4079 4080 nonce = 0; 4081 if (READ_ONCE(idev->cnf.enhanced_dad) || 4082 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) { 4083 do 4084 get_random_bytes(&nonce, 6); 4085 while (nonce == 0); 4086 } 4087 ifp->dad_nonce = nonce; 4088 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits); 4089 addrconf_mod_dad_work(ifp, rand_num); 4090 } 4091 4092 static void addrconf_dad_begin(struct inet6_ifaddr *ifp) 4093 { 4094 struct inet6_dev *idev = ifp->idev; 4095 struct net_device *dev = idev->dev; 4096 bool bump_id, notify = false; 4097 struct net *net; 4098 4099 addrconf_join_solict(dev, &ifp->addr); 4100 4101 read_lock_bh(&idev->lock); 4102 spin_lock(&ifp->lock); 4103 if (ifp->state == INET6_IFADDR_STATE_DEAD) 4104 goto out; 4105 4106 net = dev_net(dev); 4107 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) || 4108 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 && 4109 READ_ONCE(idev->cnf.accept_dad) < 1) || 4110 !(ifp->flags&IFA_F_TENTATIVE) || 4111 ifp->flags & IFA_F_NODAD) { 4112 bool send_na = false; 4113 4114 if (ifp->flags & IFA_F_TENTATIVE && 4115 !(ifp->flags & IFA_F_OPTIMISTIC)) 4116 send_na = true; 4117 bump_id = ifp->flags & IFA_F_TENTATIVE; 4118 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4119 spin_unlock(&ifp->lock); 4120 read_unlock_bh(&idev->lock); 4121 4122 addrconf_dad_completed(ifp, bump_id, send_na); 4123 return; 4124 } 4125 4126 if (!(idev->if_flags & IF_READY)) { 4127 spin_unlock(&ifp->lock); 4128 read_unlock_bh(&idev->lock); 4129 /* 4130 * If the device is not ready: 4131 * - keep it tentative if it is a permanent address. 4132 * - otherwise, kill it. 4133 */ 4134 in6_ifa_hold(ifp); 4135 addrconf_dad_stop(ifp, 0); 4136 return; 4137 } 4138 4139 /* 4140 * Optimistic nodes can start receiving 4141 * Frames right away 4142 */ 4143 if (ifp->flags & IFA_F_OPTIMISTIC) { 4144 ip6_ins_rt(net, ifp->rt); 4145 if (ipv6_use_optimistic_addr(net, idev)) { 4146 /* Because optimistic nodes can use this address, 4147 * notify listeners. If DAD fails, RTM_DELADDR is sent. 4148 */ 4149 notify = true; 4150 } 4151 } 4152 4153 addrconf_dad_kick(ifp); 4154 out: 4155 spin_unlock(&ifp->lock); 4156 read_unlock_bh(&idev->lock); 4157 if (notify) 4158 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4159 } 4160 4161 static void addrconf_dad_start(struct inet6_ifaddr *ifp) 4162 { 4163 bool begin_dad = false; 4164 4165 spin_lock_bh(&ifp->lock); 4166 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 4167 ifp->state = INET6_IFADDR_STATE_PREDAD; 4168 begin_dad = true; 4169 } 4170 spin_unlock_bh(&ifp->lock); 4171 4172 if (begin_dad) 4173 addrconf_mod_dad_work(ifp, 0); 4174 } 4175 4176 static void addrconf_dad_work(struct work_struct *w) 4177 { 4178 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w), 4179 struct inet6_ifaddr, 4180 dad_work); 4181 struct inet6_dev *idev = ifp->idev; 4182 bool bump_id, disable_ipv6 = false; 4183 struct in6_addr mcaddr; 4184 struct net *net; 4185 4186 enum { 4187 DAD_PROCESS, 4188 DAD_BEGIN, 4189 DAD_ABORT, 4190 } action = DAD_PROCESS; 4191 4192 net = dev_net(idev->dev); 4193 4194 rtnl_net_lock(net); 4195 4196 spin_lock_bh(&ifp->lock); 4197 if (ifp->state == INET6_IFADDR_STATE_PREDAD) { 4198 action = DAD_BEGIN; 4199 ifp->state = INET6_IFADDR_STATE_DAD; 4200 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) { 4201 action = DAD_ABORT; 4202 ifp->state = INET6_IFADDR_STATE_POSTDAD; 4203 4204 if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 || 4205 READ_ONCE(idev->cnf.accept_dad) > 1) && 4206 !idev->cnf.disable_ipv6 && 4207 !(ifp->flags & IFA_F_STABLE_PRIVACY)) { 4208 struct in6_addr addr; 4209 4210 addr.s6_addr32[0] = htonl(0xfe800000); 4211 addr.s6_addr32[1] = 0; 4212 4213 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) && 4214 ipv6_addr_equal(&ifp->addr, &addr)) { 4215 /* DAD failed for link-local based on MAC */ 4216 WRITE_ONCE(idev->cnf.disable_ipv6, 1); 4217 4218 pr_info("%s: IPv6 being disabled!\n", 4219 ifp->idev->dev->name); 4220 disable_ipv6 = true; 4221 } 4222 } 4223 } 4224 spin_unlock_bh(&ifp->lock); 4225 4226 if (action == DAD_BEGIN) { 4227 addrconf_dad_begin(ifp); 4228 goto out; 4229 } else if (action == DAD_ABORT) { 4230 in6_ifa_hold(ifp); 4231 addrconf_dad_stop(ifp, 1); 4232 if (disable_ipv6) 4233 addrconf_ifdown(idev->dev, false); 4234 goto out; 4235 } 4236 4237 if (!ifp->dad_probes && addrconf_dad_end(ifp)) 4238 goto out; 4239 4240 write_lock_bh(&idev->lock); 4241 if (idev->dead || !(idev->if_flags & IF_READY)) { 4242 write_unlock_bh(&idev->lock); 4243 goto out; 4244 } 4245 4246 spin_lock(&ifp->lock); 4247 if (ifp->state == INET6_IFADDR_STATE_DEAD) { 4248 spin_unlock(&ifp->lock); 4249 write_unlock_bh(&idev->lock); 4250 goto out; 4251 } 4252 4253 if (ifp->dad_probes == 0) { 4254 bool send_na = false; 4255 4256 /* 4257 * DAD was successful 4258 */ 4259 4260 if (ifp->flags & IFA_F_TENTATIVE && 4261 !(ifp->flags & IFA_F_OPTIMISTIC)) 4262 send_na = true; 4263 bump_id = ifp->flags & IFA_F_TENTATIVE; 4264 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4265 spin_unlock(&ifp->lock); 4266 write_unlock_bh(&idev->lock); 4267 4268 addrconf_dad_completed(ifp, bump_id, send_na); 4269 4270 goto out; 4271 } 4272 4273 ifp->dad_probes--; 4274 addrconf_mod_dad_work(ifp, 4275 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), 4276 HZ/100)); 4277 spin_unlock(&ifp->lock); 4278 write_unlock_bh(&idev->lock); 4279 4280 /* send a neighbour solicitation for our addr */ 4281 addrconf_addr_solict_mult(&ifp->addr, &mcaddr); 4282 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any, 4283 ifp->dad_nonce); 4284 out: 4285 in6_ifa_put(ifp); 4286 rtnl_net_unlock(net); 4287 } 4288 4289 /* ifp->idev must be at least read locked */ 4290 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp) 4291 { 4292 struct inet6_ifaddr *ifpiter; 4293 struct inet6_dev *idev = ifp->idev; 4294 4295 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) { 4296 if (ifpiter->scope > IFA_LINK) 4297 break; 4298 if (ifp != ifpiter && ifpiter->scope == IFA_LINK && 4299 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE| 4300 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) == 4301 IFA_F_PERMANENT) 4302 return false; 4303 } 4304 return true; 4305 } 4306 4307 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 4308 bool send_na) 4309 { 4310 struct net_device *dev = ifp->idev->dev; 4311 struct in6_addr lladdr; 4312 bool send_rs, send_mld; 4313 4314 addrconf_del_dad_work(ifp); 4315 4316 /* 4317 * Configure the address for reception. Now it is valid. 4318 */ 4319 4320 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4321 4322 /* If added prefix is link local and we are prepared to process 4323 router advertisements, start sending router solicitations. 4324 */ 4325 4326 read_lock_bh(&ifp->idev->lock); 4327 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp); 4328 send_rs = send_mld && 4329 ipv6_accept_ra(ifp->idev) && 4330 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 && 4331 (dev->flags & IFF_LOOPBACK) == 0 && 4332 (dev->type != ARPHRD_TUNNEL) && 4333 !netif_is_team_port(dev); 4334 read_unlock_bh(&ifp->idev->lock); 4335 4336 /* While dad is in progress mld report's source address is in6_addrany. 4337 * Resend with proper ll now. 4338 */ 4339 if (send_mld) 4340 ipv6_mc_dad_complete(ifp->idev); 4341 4342 /* send unsolicited NA if enabled */ 4343 if (send_na && 4344 (READ_ONCE(ifp->idev->cnf.ndisc_notify) || 4345 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) { 4346 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr, 4347 /*router=*/ !!ifp->idev->cnf.forwarding, 4348 /*solicited=*/ false, /*override=*/ true, 4349 /*inc_opt=*/ true); 4350 } 4351 4352 if (send_rs) { 4353 /* 4354 * If a host as already performed a random delay 4355 * [...] as part of DAD [...] there is no need 4356 * to delay again before sending the first RS 4357 */ 4358 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 4359 return; 4360 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters); 4361 4362 write_lock_bh(&ifp->idev->lock); 4363 spin_lock(&ifp->lock); 4364 ifp->idev->rs_interval = rfc3315_s14_backoff_init( 4365 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval)); 4366 ifp->idev->rs_probes = 1; 4367 ifp->idev->if_flags |= IF_RS_SENT; 4368 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval); 4369 spin_unlock(&ifp->lock); 4370 write_unlock_bh(&ifp->idev->lock); 4371 } 4372 4373 if (bump_id) 4374 rt_genid_bump_ipv6(dev_net(dev)); 4375 4376 /* Make sure that a new temporary address will be created 4377 * before this temporary address becomes deprecated. 4378 */ 4379 if (ifp->flags & IFA_F_TEMPORARY) 4380 addrconf_verify_rtnl(dev_net(dev)); 4381 } 4382 4383 static void addrconf_dad_run(struct inet6_dev *idev, bool restart) 4384 { 4385 struct inet6_ifaddr *ifp; 4386 4387 read_lock_bh(&idev->lock); 4388 list_for_each_entry(ifp, &idev->addr_list, if_list) { 4389 spin_lock(&ifp->lock); 4390 if ((ifp->flags & IFA_F_TENTATIVE && 4391 ifp->state == INET6_IFADDR_STATE_DAD) || restart) { 4392 if (restart) 4393 ifp->state = INET6_IFADDR_STATE_PREDAD; 4394 addrconf_dad_kick(ifp); 4395 } 4396 spin_unlock(&ifp->lock); 4397 } 4398 read_unlock_bh(&idev->lock); 4399 } 4400 4401 #ifdef CONFIG_PROC_FS 4402 struct if6_iter_state { 4403 struct seq_net_private p; 4404 int bucket; 4405 int offset; 4406 }; 4407 4408 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos) 4409 { 4410 struct if6_iter_state *state = seq->private; 4411 struct net *net = seq_file_net(seq); 4412 struct inet6_ifaddr *ifa = NULL; 4413 int p = 0; 4414 4415 /* initial bucket if pos is 0 */ 4416 if (pos == 0) { 4417 state->bucket = 0; 4418 state->offset = 0; 4419 } 4420 4421 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) { 4422 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket], 4423 addr_lst) { 4424 /* sync with offset */ 4425 if (p < state->offset) { 4426 p++; 4427 continue; 4428 } 4429 return ifa; 4430 } 4431 4432 /* prepare for next bucket */ 4433 state->offset = 0; 4434 p = 0; 4435 } 4436 return NULL; 4437 } 4438 4439 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, 4440 struct inet6_ifaddr *ifa) 4441 { 4442 struct if6_iter_state *state = seq->private; 4443 struct net *net = seq_file_net(seq); 4444 4445 hlist_for_each_entry_continue_rcu(ifa, addr_lst) { 4446 state->offset++; 4447 return ifa; 4448 } 4449 4450 state->offset = 0; 4451 while (++state->bucket < IN6_ADDR_HSIZE) { 4452 hlist_for_each_entry_rcu(ifa, 4453 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) { 4454 return ifa; 4455 } 4456 } 4457 4458 return NULL; 4459 } 4460 4461 static void *if6_seq_start(struct seq_file *seq, loff_t *pos) 4462 __acquires(rcu) 4463 { 4464 rcu_read_lock(); 4465 return if6_get_first(seq, *pos); 4466 } 4467 4468 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 4469 { 4470 struct inet6_ifaddr *ifa; 4471 4472 ifa = if6_get_next(seq, v); 4473 ++*pos; 4474 return ifa; 4475 } 4476 4477 static void if6_seq_stop(struct seq_file *seq, void *v) 4478 __releases(rcu) 4479 { 4480 rcu_read_unlock(); 4481 } 4482 4483 static int if6_seq_show(struct seq_file *seq, void *v) 4484 { 4485 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v; 4486 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n", 4487 &ifp->addr, 4488 ifp->idev->dev->ifindex, 4489 ifp->prefix_len, 4490 ifp->scope, 4491 (u8) ifp->flags, 4492 ifp->idev->dev->name); 4493 return 0; 4494 } 4495 4496 static const struct seq_operations if6_seq_ops = { 4497 .start = if6_seq_start, 4498 .next = if6_seq_next, 4499 .show = if6_seq_show, 4500 .stop = if6_seq_stop, 4501 }; 4502 4503 static int __net_init if6_proc_net_init(struct net *net) 4504 { 4505 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops, 4506 sizeof(struct if6_iter_state))) 4507 return -ENOMEM; 4508 return 0; 4509 } 4510 4511 static void __net_exit if6_proc_net_exit(struct net *net) 4512 { 4513 remove_proc_entry("if_inet6", net->proc_net); 4514 } 4515 4516 static struct pernet_operations if6_proc_net_ops = { 4517 .init = if6_proc_net_init, 4518 .exit = if6_proc_net_exit, 4519 }; 4520 4521 int __init if6_proc_init(void) 4522 { 4523 return register_pernet_subsys(&if6_proc_net_ops); 4524 } 4525 4526 void if6_proc_exit(void) 4527 { 4528 unregister_pernet_subsys(&if6_proc_net_ops); 4529 } 4530 #endif /* CONFIG_PROC_FS */ 4531 4532 #if IS_ENABLED(CONFIG_IPV6_MIP6) 4533 /* Check if address is a home address configured on any interface. */ 4534 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr) 4535 { 4536 unsigned int hash = inet6_addr_hash(net, addr); 4537 struct inet6_ifaddr *ifp = NULL; 4538 int ret = 0; 4539 4540 rcu_read_lock(); 4541 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4542 if (ipv6_addr_equal(&ifp->addr, addr) && 4543 (ifp->flags & IFA_F_HOMEADDRESS)) { 4544 ret = 1; 4545 break; 4546 } 4547 } 4548 rcu_read_unlock(); 4549 return ret; 4550 } 4551 #endif 4552 4553 /* RFC6554 has some algorithm to avoid loops in segment routing by 4554 * checking if the segments contains any of a local interface address. 4555 * 4556 * Quote: 4557 * 4558 * To detect loops in the SRH, a router MUST determine if the SRH 4559 * includes multiple addresses assigned to any interface on that router. 4560 * If such addresses appear more than once and are separated by at least 4561 * one address not assigned to that router. 4562 */ 4563 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs, 4564 unsigned char nsegs) 4565 { 4566 const struct in6_addr *addr; 4567 int i, ret = 0, found = 0; 4568 struct inet6_ifaddr *ifp; 4569 bool separated = false; 4570 unsigned int hash; 4571 bool hash_found; 4572 4573 rcu_read_lock(); 4574 for (i = 0; i < nsegs; i++) { 4575 addr = &segs[i]; 4576 hash = inet6_addr_hash(net, addr); 4577 4578 hash_found = false; 4579 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4580 4581 if (ipv6_addr_equal(&ifp->addr, addr)) { 4582 hash_found = true; 4583 break; 4584 } 4585 } 4586 4587 if (hash_found) { 4588 if (found > 1 && separated) { 4589 ret = 1; 4590 break; 4591 } 4592 4593 separated = false; 4594 found++; 4595 } else { 4596 separated = true; 4597 } 4598 } 4599 rcu_read_unlock(); 4600 4601 return ret; 4602 } 4603 4604 /* 4605 * Periodic address status verification 4606 */ 4607 4608 static void addrconf_verify_rtnl(struct net *net) 4609 { 4610 unsigned long now, next, next_sec, next_sched; 4611 struct inet6_ifaddr *ifp; 4612 int i; 4613 4614 ASSERT_RTNL(); 4615 4616 rcu_read_lock_bh(); 4617 now = jiffies; 4618 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY); 4619 4620 cancel_delayed_work(&net->ipv6.addr_chk_work); 4621 4622 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 4623 restart: 4624 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) { 4625 unsigned long age; 4626 4627 /* When setting preferred_lft to a value not zero or 4628 * infinity, while valid_lft is infinity 4629 * IFA_F_PERMANENT has a non-infinity life time. 4630 */ 4631 if ((ifp->flags & IFA_F_PERMANENT) && 4632 (ifp->prefered_lft == INFINITY_LIFE_TIME)) 4633 continue; 4634 4635 spin_lock(&ifp->lock); 4636 /* We try to batch several events at once. */ 4637 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 4638 4639 if ((ifp->flags&IFA_F_TEMPORARY) && 4640 !(ifp->flags&IFA_F_TENTATIVE) && 4641 ifp->prefered_lft != INFINITY_LIFE_TIME && 4642 !ifp->regen_count && ifp->ifpub) { 4643 /* This is a non-regenerated temporary addr. */ 4644 4645 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev); 4646 4647 if (age + regen_advance >= ifp->prefered_lft) { 4648 struct inet6_ifaddr *ifpub = ifp->ifpub; 4649 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4650 next = ifp->tstamp + ifp->prefered_lft * HZ; 4651 4652 ifp->regen_count++; 4653 in6_ifa_hold(ifp); 4654 in6_ifa_hold(ifpub); 4655 spin_unlock(&ifp->lock); 4656 4657 spin_lock(&ifpub->lock); 4658 ifpub->regen_count = 0; 4659 spin_unlock(&ifpub->lock); 4660 rcu_read_unlock_bh(); 4661 ipv6_create_tempaddr(ifpub, true); 4662 in6_ifa_put(ifpub); 4663 in6_ifa_put(ifp); 4664 rcu_read_lock_bh(); 4665 goto restart; 4666 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next)) 4667 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ; 4668 } 4669 4670 if (ifp->valid_lft != INFINITY_LIFE_TIME && 4671 age >= ifp->valid_lft) { 4672 spin_unlock(&ifp->lock); 4673 in6_ifa_hold(ifp); 4674 rcu_read_unlock_bh(); 4675 ipv6_del_addr(ifp); 4676 rcu_read_lock_bh(); 4677 goto restart; 4678 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) { 4679 spin_unlock(&ifp->lock); 4680 continue; 4681 } else if (age >= ifp->prefered_lft) { 4682 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */ 4683 int deprecate = 0; 4684 4685 if (!(ifp->flags&IFA_F_DEPRECATED)) { 4686 deprecate = 1; 4687 ifp->flags |= IFA_F_DEPRECATED; 4688 } 4689 4690 if ((ifp->valid_lft != INFINITY_LIFE_TIME) && 4691 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))) 4692 next = ifp->tstamp + ifp->valid_lft * HZ; 4693 4694 spin_unlock(&ifp->lock); 4695 4696 if (deprecate) { 4697 in6_ifa_hold(ifp); 4698 4699 ipv6_ifa_notify(0, ifp); 4700 in6_ifa_put(ifp); 4701 goto restart; 4702 } 4703 } else { 4704 /* ifp->prefered_lft <= ifp->valid_lft */ 4705 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4706 next = ifp->tstamp + ifp->prefered_lft * HZ; 4707 spin_unlock(&ifp->lock); 4708 } 4709 } 4710 } 4711 4712 next_sec = round_jiffies_up(next); 4713 next_sched = next; 4714 4715 /* If rounded timeout is accurate enough, accept it. */ 4716 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ)) 4717 next_sched = next_sec; 4718 4719 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */ 4720 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX)) 4721 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX; 4722 4723 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n", 4724 now, next, next_sec, next_sched); 4725 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now); 4726 rcu_read_unlock_bh(); 4727 } 4728 4729 static void addrconf_verify_work(struct work_struct *w) 4730 { 4731 struct net *net = container_of(to_delayed_work(w), struct net, 4732 ipv6.addr_chk_work); 4733 4734 rtnl_net_lock(net); 4735 addrconf_verify_rtnl(net); 4736 rtnl_net_unlock(net); 4737 } 4738 4739 static void addrconf_verify(struct net *net) 4740 { 4741 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0); 4742 } 4743 4744 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local, 4745 struct in6_addr **peer_pfx) 4746 { 4747 struct in6_addr *pfx = NULL; 4748 4749 *peer_pfx = NULL; 4750 4751 if (addr) 4752 pfx = nla_data(addr); 4753 4754 if (local) { 4755 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx))) 4756 *peer_pfx = pfx; 4757 pfx = nla_data(local); 4758 } 4759 4760 return pfx; 4761 } 4762 4763 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = { 4764 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) }, 4765 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) }, 4766 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) }, 4767 [IFA_FLAGS] = { .len = sizeof(u32) }, 4768 [IFA_RT_PRIORITY] = { .len = sizeof(u32) }, 4769 [IFA_TARGET_NETNSID] = { .type = NLA_S32 }, 4770 [IFA_PROTO] = { .type = NLA_U8 }, 4771 }; 4772 4773 static int 4774 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, 4775 struct netlink_ext_ack *extack) 4776 { 4777 struct net *net = sock_net(skb->sk); 4778 struct ifaddrmsg *ifm; 4779 struct nlattr *tb[IFA_MAX+1]; 4780 struct in6_addr *pfx, *peer_pfx; 4781 u32 ifa_flags; 4782 int err; 4783 4784 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 4785 ifa_ipv6_policy, extack); 4786 if (err < 0) 4787 return err; 4788 4789 ifm = nlmsg_data(nlh); 4790 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 4791 if (!pfx) 4792 return -EINVAL; 4793 4794 ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags); 4795 4796 /* We ignore other flags so far. */ 4797 ifa_flags &= IFA_F_MANAGETEMPADDR; 4798 4799 rtnl_net_lock(net); 4800 err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx, 4801 ifm->ifa_prefixlen, extack); 4802 rtnl_net_unlock(net); 4803 4804 return err; 4805 } 4806 4807 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp, 4808 unsigned long expires, u32 flags, 4809 bool modify_peer) 4810 { 4811 struct fib6_table *table; 4812 struct fib6_info *f6i; 4813 u32 prio; 4814 4815 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4816 ifp->prefix_len, 4817 ifp->idev->dev, 0, RTF_DEFAULT, true); 4818 if (!f6i) 4819 return -ENOENT; 4820 4821 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF; 4822 if (f6i->fib6_metric != prio) { 4823 /* delete old one */ 4824 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false); 4825 4826 /* add new one */ 4827 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4828 ifp->prefix_len, 4829 ifp->rt_priority, ifp->idev->dev, 4830 expires, flags, GFP_KERNEL); 4831 return 0; 4832 } 4833 if (f6i != net->ipv6.fib6_null_entry) { 4834 table = f6i->fib6_table; 4835 spin_lock_bh(&table->tb6_lock); 4836 4837 if (!(flags & RTF_EXPIRES)) { 4838 fib6_clean_expires(f6i); 4839 fib6_remove_gc_list(f6i); 4840 } else { 4841 fib6_set_expires(f6i, expires); 4842 fib6_add_gc_list(f6i); 4843 } 4844 4845 spin_unlock_bh(&table->tb6_lock); 4846 } 4847 fib6_info_release(f6i); 4848 4849 return 0; 4850 } 4851 4852 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp, 4853 struct ifa6_config *cfg, clock_t expires, 4854 u32 flags) 4855 { 4856 bool was_managetempaddr; 4857 bool new_peer = false; 4858 bool had_prefixroute; 4859 4860 ASSERT_RTNL_NET(net); 4861 4862 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && 4863 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64)) 4864 return -EINVAL; 4865 4866 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED) 4867 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC; 4868 4869 if (cfg->peer_pfx && 4870 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) { 4871 if (!ipv6_addr_any(&ifp->peer_addr)) 4872 cleanup_prefix_route(ifp, expires, true, true); 4873 new_peer = true; 4874 } 4875 4876 spin_lock_bh(&ifp->lock); 4877 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR; 4878 had_prefixroute = ifp->flags & IFA_F_PERMANENT && 4879 !(ifp->flags & IFA_F_NOPREFIXROUTE); 4880 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | 4881 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR | 4882 IFA_F_NOPREFIXROUTE); 4883 ifp->flags |= cfg->ifa_flags; 4884 WRITE_ONCE(ifp->tstamp, jiffies); 4885 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft); 4886 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft); 4887 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto); 4888 4889 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority) 4890 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority); 4891 4892 if (new_peer) 4893 ifp->peer_addr = *cfg->peer_pfx; 4894 4895 spin_unlock_bh(&ifp->lock); 4896 if (!(ifp->flags&IFA_F_TENTATIVE)) 4897 ipv6_ifa_notify(0, ifp); 4898 4899 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 4900 int rc = -ENOENT; 4901 4902 if (had_prefixroute) 4903 rc = modify_prefix_route(net, ifp, expires, flags, false); 4904 4905 /* prefix route could have been deleted; if so restore it */ 4906 if (rc == -ENOENT) { 4907 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 4908 ifp->rt_priority, ifp->idev->dev, 4909 expires, flags, GFP_KERNEL); 4910 } 4911 4912 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr)) 4913 rc = modify_prefix_route(net, ifp, expires, flags, true); 4914 4915 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) { 4916 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len, 4917 ifp->rt_priority, ifp->idev->dev, 4918 expires, flags, GFP_KERNEL); 4919 } 4920 } else if (had_prefixroute) { 4921 enum cleanup_prefix_rt_t action; 4922 unsigned long rt_expires; 4923 4924 write_lock_bh(&ifp->idev->lock); 4925 action = check_cleanup_prefix_route(ifp, &rt_expires); 4926 write_unlock_bh(&ifp->idev->lock); 4927 4928 if (action != CLEANUP_PREFIX_RT_NOP) { 4929 cleanup_prefix_route(ifp, rt_expires, 4930 action == CLEANUP_PREFIX_RT_DEL, false); 4931 } 4932 } 4933 4934 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) { 4935 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR)) 4936 delete_tempaddrs(ifp->idev, ifp); 4937 else 4938 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft, 4939 cfg->preferred_lft, !was_managetempaddr, 4940 jiffies); 4941 } 4942 4943 addrconf_verify_rtnl(net); 4944 4945 return 0; 4946 } 4947 4948 static int 4949 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, 4950 struct netlink_ext_ack *extack) 4951 { 4952 struct net *net = sock_net(skb->sk); 4953 struct nlattr *tb[IFA_MAX+1]; 4954 struct in6_addr *peer_pfx; 4955 struct inet6_ifaddr *ifa; 4956 struct net_device *dev; 4957 struct inet6_dev *idev; 4958 struct ifa6_config cfg; 4959 struct ifaddrmsg *ifm; 4960 unsigned long timeout; 4961 clock_t expires; 4962 u32 flags; 4963 int err; 4964 4965 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 4966 ifa_ipv6_policy, extack); 4967 if (err < 0) 4968 return err; 4969 4970 memset(&cfg, 0, sizeof(cfg)); 4971 4972 ifm = nlmsg_data(nlh); 4973 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 4974 if (!cfg.pfx) 4975 return -EINVAL; 4976 4977 cfg.peer_pfx = peer_pfx; 4978 cfg.plen = ifm->ifa_prefixlen; 4979 if (tb[IFA_RT_PRIORITY]) 4980 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]); 4981 4982 if (tb[IFA_PROTO]) 4983 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]); 4984 4985 cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags); 4986 4987 /* We ignore other flags so far. */ 4988 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | 4989 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE | 4990 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC; 4991 4992 cfg.ifa_flags |= IFA_F_PERMANENT; 4993 cfg.valid_lft = INFINITY_LIFE_TIME; 4994 cfg.preferred_lft = INFINITY_LIFE_TIME; 4995 expires = 0; 4996 flags = 0; 4997 4998 if (tb[IFA_CACHEINFO]) { 4999 struct ifa_cacheinfo *ci; 5000 5001 ci = nla_data(tb[IFA_CACHEINFO]); 5002 cfg.valid_lft = ci->ifa_valid; 5003 cfg.preferred_lft = ci->ifa_prefered; 5004 5005 if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) { 5006 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid"); 5007 return -EINVAL; 5008 } 5009 5010 timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ); 5011 if (addrconf_finite_timeout(timeout)) { 5012 cfg.ifa_flags &= ~IFA_F_PERMANENT; 5013 cfg.valid_lft = timeout; 5014 expires = jiffies_to_clock_t(timeout * HZ); 5015 flags = RTF_EXPIRES; 5016 } 5017 5018 timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ); 5019 if (addrconf_finite_timeout(timeout)) { 5020 if (timeout == 0) 5021 cfg.ifa_flags |= IFA_F_DEPRECATED; 5022 5023 cfg.preferred_lft = timeout; 5024 } 5025 } 5026 5027 rtnl_net_lock(net); 5028 5029 dev = __dev_get_by_index(net, ifm->ifa_index); 5030 if (!dev) { 5031 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 5032 err = -ENODEV; 5033 goto unlock_rtnl; 5034 } 5035 5036 netdev_lock_ops(dev); 5037 idev = ipv6_find_idev(dev); 5038 if (IS_ERR(idev)) { 5039 err = PTR_ERR(idev); 5040 goto unlock; 5041 } 5042 5043 if (!ipv6_allow_optimistic_dad(net, idev)) 5044 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC; 5045 5046 if (cfg.ifa_flags & IFA_F_NODAD && 5047 cfg.ifa_flags & IFA_F_OPTIMISTIC) { 5048 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive"); 5049 err = -EINVAL; 5050 goto unlock; 5051 } 5052 5053 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1); 5054 if (!ifa) { 5055 /* 5056 * It would be best to check for !NLM_F_CREATE here but 5057 * userspace already relies on not having to provide this. 5058 */ 5059 err = inet6_addr_add(net, dev, &cfg, expires, flags, extack); 5060 goto unlock; 5061 } 5062 5063 if (nlh->nlmsg_flags & NLM_F_EXCL || 5064 !(nlh->nlmsg_flags & NLM_F_REPLACE)) { 5065 NL_SET_ERR_MSG_MOD(extack, "address already assigned"); 5066 err = -EEXIST; 5067 } else { 5068 err = inet6_addr_modify(net, ifa, &cfg, expires, flags); 5069 } 5070 5071 in6_ifa_put(ifa); 5072 unlock: 5073 netdev_unlock_ops(dev); 5074 unlock_rtnl: 5075 rtnl_net_unlock(net); 5076 5077 return err; 5078 } 5079 5080 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags, 5081 u8 scope, int ifindex) 5082 { 5083 struct ifaddrmsg *ifm; 5084 5085 ifm = nlmsg_data(nlh); 5086 ifm->ifa_family = AF_INET6; 5087 ifm->ifa_prefixlen = prefixlen; 5088 ifm->ifa_flags = flags; 5089 ifm->ifa_scope = scope; 5090 ifm->ifa_index = ifindex; 5091 } 5092 5093 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp, 5094 unsigned long tstamp, u32 preferred, u32 valid) 5095 { 5096 struct ifa_cacheinfo ci; 5097 5098 ci.cstamp = cstamp_delta(cstamp); 5099 ci.tstamp = cstamp_delta(tstamp); 5100 ci.ifa_prefered = preferred; 5101 ci.ifa_valid = valid; 5102 5103 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci); 5104 } 5105 5106 static inline int rt_scope(int ifa_scope) 5107 { 5108 if (ifa_scope & IFA_HOST) 5109 return RT_SCOPE_HOST; 5110 else if (ifa_scope & IFA_LINK) 5111 return RT_SCOPE_LINK; 5112 else if (ifa_scope & IFA_SITE) 5113 return RT_SCOPE_SITE; 5114 else 5115 return RT_SCOPE_UNIVERSE; 5116 } 5117 5118 static inline int inet6_ifaddr_msgsize(void) 5119 { 5120 return NLMSG_ALIGN(sizeof(struct ifaddrmsg)) 5121 + nla_total_size(16) /* IFA_LOCAL */ 5122 + nla_total_size(16) /* IFA_ADDRESS */ 5123 + nla_total_size(sizeof(struct ifa_cacheinfo)) 5124 + nla_total_size(4) /* IFA_FLAGS */ 5125 + nla_total_size(1) /* IFA_PROTO */ 5126 + nla_total_size(4) /* IFA_RT_PRIORITY */; 5127 } 5128 5129 static int inet6_fill_ifaddr(struct sk_buff *skb, 5130 const struct inet6_ifaddr *ifa, 5131 struct inet6_fill_args *args) 5132 { 5133 struct nlmsghdr *nlh; 5134 u32 preferred, valid; 5135 u32 flags, priority; 5136 u8 proto; 5137 5138 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5139 sizeof(struct ifaddrmsg), args->flags); 5140 if (!nlh) 5141 return -EMSGSIZE; 5142 5143 flags = READ_ONCE(ifa->flags); 5144 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope), 5145 ifa->idev->dev->ifindex); 5146 5147 if (args->netnsid >= 0 && 5148 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) 5149 goto error; 5150 5151 preferred = READ_ONCE(ifa->prefered_lft); 5152 valid = READ_ONCE(ifa->valid_lft); 5153 5154 if (!((flags & IFA_F_PERMANENT) && 5155 (preferred == INFINITY_LIFE_TIME))) { 5156 if (preferred != INFINITY_LIFE_TIME) { 5157 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ; 5158 5159 if (preferred > tval) 5160 preferred -= tval; 5161 else 5162 preferred = 0; 5163 if (valid != INFINITY_LIFE_TIME) { 5164 if (valid > tval) 5165 valid -= tval; 5166 else 5167 valid = 0; 5168 } 5169 } 5170 } else { 5171 preferred = INFINITY_LIFE_TIME; 5172 valid = INFINITY_LIFE_TIME; 5173 } 5174 5175 if (!ipv6_addr_any(&ifa->peer_addr)) { 5176 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 || 5177 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0) 5178 goto error; 5179 } else { 5180 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0) 5181 goto error; 5182 } 5183 5184 priority = READ_ONCE(ifa->rt_priority); 5185 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority)) 5186 goto error; 5187 5188 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp), 5189 preferred, valid) < 0) 5190 goto error; 5191 5192 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0) 5193 goto error; 5194 5195 proto = READ_ONCE(ifa->ifa_proto); 5196 if (proto && nla_put_u8(skb, IFA_PROTO, proto)) 5197 goto error; 5198 5199 nlmsg_end(skb, nlh); 5200 return 0; 5201 5202 error: 5203 nlmsg_cancel(skb, nlh); 5204 return -EMSGSIZE; 5205 } 5206 5207 int inet6_fill_ifmcaddr(struct sk_buff *skb, 5208 const struct ifmcaddr6 *ifmca, 5209 struct inet6_fill_args *args) 5210 { 5211 int ifindex = ifmca->idev->dev->ifindex; 5212 u8 scope = RT_SCOPE_UNIVERSE; 5213 struct nlmsghdr *nlh; 5214 5215 if (!args->force_rt_scope_universe && 5216 ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE) 5217 scope = RT_SCOPE_SITE; 5218 5219 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5220 sizeof(struct ifaddrmsg), args->flags); 5221 if (!nlh) 5222 return -EMSGSIZE; 5223 5224 if (args->netnsid >= 0 && 5225 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) { 5226 nlmsg_cancel(skb, nlh); 5227 return -EMSGSIZE; 5228 } 5229 5230 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 5231 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 || 5232 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp), 5233 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5234 nlmsg_cancel(skb, nlh); 5235 return -EMSGSIZE; 5236 } 5237 5238 nlmsg_end(skb, nlh); 5239 return 0; 5240 } 5241 5242 int inet6_fill_ifacaddr(struct sk_buff *skb, 5243 const struct ifacaddr6 *ifaca, 5244 struct inet6_fill_args *args) 5245 { 5246 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt); 5247 int ifindex = dev ? dev->ifindex : 1; 5248 u8 scope = RT_SCOPE_UNIVERSE; 5249 struct nlmsghdr *nlh; 5250 5251 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE) 5252 scope = RT_SCOPE_SITE; 5253 5254 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5255 sizeof(struct ifaddrmsg), args->flags); 5256 if (!nlh) 5257 return -EMSGSIZE; 5258 5259 if (args->netnsid >= 0 && 5260 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) { 5261 nlmsg_cancel(skb, nlh); 5262 return -EMSGSIZE; 5263 } 5264 5265 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 5266 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 || 5267 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp), 5268 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5269 nlmsg_cancel(skb, nlh); 5270 return -EMSGSIZE; 5271 } 5272 5273 nlmsg_end(skb, nlh); 5274 return 0; 5275 } 5276 5277 /* called with rcu_read_lock() */ 5278 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb, 5279 struct netlink_callback *cb, int *s_ip_idx, 5280 struct inet6_fill_args *fillargs) 5281 { 5282 const struct ifmcaddr6 *ifmca; 5283 const struct ifacaddr6 *ifaca; 5284 int ip_idx = 0; 5285 int err = 0; 5286 5287 switch (fillargs->type) { 5288 case UNICAST_ADDR: { 5289 const struct inet6_ifaddr *ifa; 5290 fillargs->event = RTM_NEWADDR; 5291 5292 /* unicast address incl. temp addr */ 5293 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 5294 if (ip_idx < *s_ip_idx) 5295 goto next; 5296 err = inet6_fill_ifaddr(skb, ifa, fillargs); 5297 if (err < 0) 5298 break; 5299 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 5300 next: 5301 ip_idx++; 5302 } 5303 break; 5304 } 5305 case MULTICAST_ADDR: 5306 fillargs->event = RTM_GETMULTICAST; 5307 5308 /* multicast address */ 5309 for (ifmca = rcu_dereference(idev->mc_list); 5310 ifmca; 5311 ifmca = rcu_dereference(ifmca->next), ip_idx++) { 5312 if (ip_idx < *s_ip_idx) 5313 continue; 5314 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs); 5315 if (err < 0) 5316 break; 5317 } 5318 break; 5319 case ANYCAST_ADDR: 5320 fillargs->event = RTM_GETANYCAST; 5321 /* anycast address */ 5322 for (ifaca = rcu_dereference(idev->ac_list); ifaca; 5323 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) { 5324 if (ip_idx < *s_ip_idx) 5325 continue; 5326 err = inet6_fill_ifacaddr(skb, ifaca, fillargs); 5327 if (err < 0) 5328 break; 5329 } 5330 break; 5331 default: 5332 break; 5333 } 5334 *s_ip_idx = err ? ip_idx : 0; 5335 return err; 5336 } 5337 5338 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh, 5339 struct inet6_fill_args *fillargs, 5340 struct net **tgt_net, struct sock *sk, 5341 struct netlink_callback *cb) 5342 { 5343 struct netlink_ext_ack *extack = cb->extack; 5344 struct nlattr *tb[IFA_MAX+1]; 5345 struct ifaddrmsg *ifm; 5346 int err, i; 5347 5348 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 5349 if (!ifm) { 5350 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request"); 5351 return -EINVAL; 5352 } 5353 5354 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5355 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request"); 5356 return -EINVAL; 5357 } 5358 5359 fillargs->ifindex = ifm->ifa_index; 5360 if (fillargs->ifindex) { 5361 cb->answer_flags |= NLM_F_DUMP_FILTERED; 5362 fillargs->flags |= NLM_F_DUMP_FILTERED; 5363 } 5364 5365 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5366 ifa_ipv6_policy, extack); 5367 if (err < 0) 5368 return err; 5369 5370 for (i = 0; i <= IFA_MAX; ++i) { 5371 if (!tb[i]) 5372 continue; 5373 5374 if (i == IFA_TARGET_NETNSID) { 5375 struct net *net; 5376 5377 fillargs->netnsid = nla_get_s32(tb[i]); 5378 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid); 5379 if (IS_ERR(net)) { 5380 fillargs->netnsid = -1; 5381 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id"); 5382 return PTR_ERR(net); 5383 } 5384 *tgt_net = net; 5385 } else { 5386 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request"); 5387 return -EINVAL; 5388 } 5389 } 5390 5391 return 0; 5392 } 5393 5394 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb, 5395 enum addr_type_t type) 5396 { 5397 struct net *tgt_net = sock_net(skb->sk); 5398 const struct nlmsghdr *nlh = cb->nlh; 5399 struct inet6_fill_args fillargs = { 5400 .portid = NETLINK_CB(cb->skb).portid, 5401 .seq = cb->nlh->nlmsg_seq, 5402 .flags = NLM_F_MULTI, 5403 .netnsid = -1, 5404 .type = type, 5405 .force_rt_scope_universe = false, 5406 }; 5407 struct { 5408 unsigned long ifindex; 5409 int ip_idx; 5410 } *ctx = (void *)cb->ctx; 5411 struct net_device *dev; 5412 struct inet6_dev *idev; 5413 int err = 0; 5414 5415 rcu_read_lock(); 5416 if (cb->strict_check) { 5417 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net, 5418 skb->sk, cb); 5419 if (err < 0) 5420 goto done; 5421 5422 err = 0; 5423 if (fillargs.ifindex) { 5424 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex); 5425 if (!dev) { 5426 err = -ENODEV; 5427 goto done; 5428 } 5429 idev = __in6_dev_get(dev); 5430 if (idev) 5431 err = in6_dump_addrs(idev, skb, cb, 5432 &ctx->ip_idx, 5433 &fillargs); 5434 goto done; 5435 } 5436 } 5437 5438 cb->seq = inet6_base_seq(tgt_net); 5439 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) { 5440 idev = __in6_dev_get(dev); 5441 if (!idev) 5442 continue; 5443 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx, 5444 &fillargs); 5445 if (err < 0) 5446 goto done; 5447 } 5448 done: 5449 rcu_read_unlock(); 5450 if (fillargs.netnsid >= 0) 5451 put_net(tgt_net); 5452 5453 return err; 5454 } 5455 5456 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb) 5457 { 5458 enum addr_type_t type = UNICAST_ADDR; 5459 5460 return inet6_dump_addr(skb, cb, type); 5461 } 5462 5463 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb) 5464 { 5465 enum addr_type_t type = MULTICAST_ADDR; 5466 5467 return inet6_dump_addr(skb, cb, type); 5468 } 5469 5470 5471 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb) 5472 { 5473 enum addr_type_t type = ANYCAST_ADDR; 5474 5475 return inet6_dump_addr(skb, cb, type); 5476 } 5477 5478 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb, 5479 const struct nlmsghdr *nlh, 5480 struct nlattr **tb, 5481 struct netlink_ext_ack *extack) 5482 { 5483 struct ifaddrmsg *ifm; 5484 int i, err; 5485 5486 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 5487 if (!ifm) { 5488 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request"); 5489 return -EINVAL; 5490 } 5491 5492 if (!netlink_strict_get_check(skb)) 5493 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 5494 ifa_ipv6_policy, extack); 5495 5496 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5497 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request"); 5498 return -EINVAL; 5499 } 5500 5501 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5502 ifa_ipv6_policy, extack); 5503 if (err) 5504 return err; 5505 5506 for (i = 0; i <= IFA_MAX; i++) { 5507 if (!tb[i]) 5508 continue; 5509 5510 switch (i) { 5511 case IFA_TARGET_NETNSID: 5512 case IFA_ADDRESS: 5513 case IFA_LOCAL: 5514 break; 5515 default: 5516 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request"); 5517 return -EINVAL; 5518 } 5519 } 5520 5521 return 0; 5522 } 5523 5524 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh, 5525 struct netlink_ext_ack *extack) 5526 { 5527 struct net *tgt_net = sock_net(in_skb->sk); 5528 struct inet6_fill_args fillargs = { 5529 .portid = NETLINK_CB(in_skb).portid, 5530 .seq = nlh->nlmsg_seq, 5531 .event = RTM_NEWADDR, 5532 .flags = 0, 5533 .netnsid = -1, 5534 .force_rt_scope_universe = false, 5535 }; 5536 struct ifaddrmsg *ifm; 5537 struct nlattr *tb[IFA_MAX+1]; 5538 struct in6_addr *addr = NULL, *peer; 5539 struct net_device *dev = NULL; 5540 struct inet6_ifaddr *ifa; 5541 struct sk_buff *skb; 5542 int err; 5543 5544 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack); 5545 if (err < 0) 5546 return err; 5547 5548 if (tb[IFA_TARGET_NETNSID]) { 5549 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]); 5550 5551 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk, 5552 fillargs.netnsid); 5553 if (IS_ERR(tgt_net)) 5554 return PTR_ERR(tgt_net); 5555 } 5556 5557 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer); 5558 if (!addr) { 5559 err = -EINVAL; 5560 goto errout; 5561 } 5562 ifm = nlmsg_data(nlh); 5563 if (ifm->ifa_index) 5564 dev = dev_get_by_index(tgt_net, ifm->ifa_index); 5565 5566 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1); 5567 if (!ifa) { 5568 err = -EADDRNOTAVAIL; 5569 goto errout; 5570 } 5571 5572 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL); 5573 if (!skb) { 5574 err = -ENOBUFS; 5575 goto errout_ifa; 5576 } 5577 5578 err = inet6_fill_ifaddr(skb, ifa, &fillargs); 5579 if (err < 0) { 5580 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 5581 WARN_ON(err == -EMSGSIZE); 5582 kfree_skb(skb); 5583 goto errout_ifa; 5584 } 5585 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid); 5586 errout_ifa: 5587 in6_ifa_put(ifa); 5588 errout: 5589 dev_put(dev); 5590 if (fillargs.netnsid >= 0) 5591 put_net(tgt_net); 5592 5593 return err; 5594 } 5595 5596 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa) 5597 { 5598 struct sk_buff *skb; 5599 struct net *net = dev_net(ifa->idev->dev); 5600 struct inet6_fill_args fillargs = { 5601 .portid = 0, 5602 .seq = 0, 5603 .event = event, 5604 .flags = 0, 5605 .netnsid = -1, 5606 .force_rt_scope_universe = false, 5607 }; 5608 int err = -ENOBUFS; 5609 5610 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC); 5611 if (!skb) 5612 goto errout; 5613 5614 err = inet6_fill_ifaddr(skb, ifa, &fillargs); 5615 if (err < 0) { 5616 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 5617 WARN_ON(err == -EMSGSIZE); 5618 kfree_skb(skb); 5619 goto errout; 5620 } 5621 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC); 5622 return; 5623 errout: 5624 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err); 5625 } 5626 5627 static void ipv6_store_devconf(const struct ipv6_devconf *cnf, 5628 __s32 *array, int bytes) 5629 { 5630 BUG_ON(bytes < (DEVCONF_MAX * 4)); 5631 5632 memset(array, 0, bytes); 5633 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding); 5634 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit); 5635 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6); 5636 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra); 5637 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects); 5638 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf); 5639 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits); 5640 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits); 5641 array[DEVCONF_RTR_SOLICIT_INTERVAL] = 5642 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval)); 5643 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] = 5644 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval)); 5645 array[DEVCONF_RTR_SOLICIT_DELAY] = 5646 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay)); 5647 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version); 5648 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] = 5649 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval)); 5650 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] = 5651 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval)); 5652 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr); 5653 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft); 5654 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft); 5655 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry); 5656 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor); 5657 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses); 5658 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr); 5659 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric); 5660 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = 5661 READ_ONCE(cnf->accept_ra_min_hop_limit); 5662 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo); 5663 #ifdef CONFIG_IPV6_ROUTER_PREF 5664 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref); 5665 array[DEVCONF_RTR_PROBE_INTERVAL] = 5666 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval)); 5667 #ifdef CONFIG_IPV6_ROUTE_INFO 5668 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = 5669 READ_ONCE(cnf->accept_ra_rt_info_min_plen); 5670 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = 5671 READ_ONCE(cnf->accept_ra_rt_info_max_plen); 5672 #endif 5673 #endif 5674 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp); 5675 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = 5676 READ_ONCE(cnf->accept_source_route); 5677 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 5678 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad); 5679 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic); 5680 #endif 5681 #ifdef CONFIG_IPV6_MROUTE 5682 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding); 5683 #endif 5684 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6); 5685 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad); 5686 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao); 5687 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify); 5688 array[DEVCONF_SUPPRESS_FRAG_NDISC] = 5689 READ_ONCE(cnf->suppress_frag_ndisc); 5690 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = 5691 READ_ONCE(cnf->accept_ra_from_local); 5692 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu); 5693 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = 5694 READ_ONCE(cnf->ignore_routes_with_linkdown); 5695 /* we omit DEVCONF_STABLE_SECRET for now */ 5696 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only); 5697 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = 5698 READ_ONCE(cnf->drop_unicast_in_l2_multicast); 5699 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na); 5700 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down); 5701 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled); 5702 #ifdef CONFIG_IPV6_SEG6_HMAC 5703 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac); 5704 #endif 5705 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad); 5706 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode); 5707 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy); 5708 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass); 5709 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled); 5710 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled); 5711 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id); 5712 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide); 5713 array[DEVCONF_NDISC_EVICT_NOCARRIER] = 5714 READ_ONCE(cnf->ndisc_evict_nocarrier); 5715 array[DEVCONF_ACCEPT_UNTRACKED_NA] = 5716 READ_ONCE(cnf->accept_untracked_na); 5717 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft); 5718 } 5719 5720 static inline size_t inet6_ifla6_size(void) 5721 { 5722 return nla_total_size(4) /* IFLA_INET6_FLAGS */ 5723 + nla_total_size(sizeof(struct ifla_cacheinfo)) 5724 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */ 5725 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */ 5726 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */ 5727 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */ 5728 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */ 5729 + nla_total_size(4) /* IFLA_INET6_RA_MTU */ 5730 + 0; 5731 } 5732 5733 static inline size_t inet6_if_nlmsg_size(void) 5734 { 5735 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 5736 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 5737 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 5738 + nla_total_size(4) /* IFLA_MTU */ 5739 + nla_total_size(4) /* IFLA_LINK */ 5740 + nla_total_size(1) /* IFLA_OPERSTATE */ 5741 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */ 5742 } 5743 5744 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib, 5745 int bytes) 5746 { 5747 int i; 5748 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX; 5749 BUG_ON(pad < 0); 5750 5751 /* Use put_unaligned() because stats may not be aligned for u64. */ 5752 put_unaligned(ICMP6_MIB_MAX, &stats[0]); 5753 for (i = 1; i < ICMP6_MIB_MAX; i++) 5754 put_unaligned(atomic_long_read(&mib[i]), &stats[i]); 5755 5756 memset(&stats[ICMP6_MIB_MAX], 0, pad); 5757 } 5758 5759 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib, 5760 int bytes, size_t syncpoff) 5761 { 5762 int i, c; 5763 u64 buff[IPSTATS_MIB_MAX]; 5764 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX; 5765 5766 BUG_ON(pad < 0); 5767 5768 memset(buff, 0, sizeof(buff)); 5769 buff[0] = IPSTATS_MIB_MAX; 5770 5771 for_each_possible_cpu(c) { 5772 for (i = 1; i < IPSTATS_MIB_MAX; i++) 5773 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff); 5774 } 5775 5776 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64)); 5777 memset(&stats[IPSTATS_MIB_MAX], 0, pad); 5778 } 5779 5780 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype, 5781 int bytes) 5782 { 5783 switch (attrtype) { 5784 case IFLA_INET6_STATS: 5785 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes, 5786 offsetof(struct ipstats_mib, syncp)); 5787 break; 5788 case IFLA_INET6_ICMP6STATS: 5789 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes); 5790 break; 5791 } 5792 } 5793 5794 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb, 5795 struct inet6_dev *idev) 5796 { 5797 struct nlattr *nla; 5798 5799 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64)); 5800 if (!nla) 5801 goto nla_put_failure; 5802 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla)); 5803 5804 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64)); 5805 if (!nla) 5806 goto nla_put_failure; 5807 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla)); 5808 5809 return 0; 5810 5811 nla_put_failure: 5812 return -EMSGSIZE; 5813 } 5814 5815 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev, 5816 u32 ext_filter_mask) 5817 { 5818 struct ifla_cacheinfo ci; 5819 struct nlattr *nla; 5820 u32 ra_mtu; 5821 5822 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags))) 5823 goto nla_put_failure; 5824 ci.max_reasm_len = IPV6_MAXPLEN; 5825 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp)); 5826 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time); 5827 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME)); 5828 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci)) 5829 goto nla_put_failure; 5830 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32)); 5831 if (!nla) 5832 goto nla_put_failure; 5833 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla)); 5834 5835 /* XXX - MC not implemented */ 5836 5837 if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) { 5838 if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0) 5839 goto nla_put_failure; 5840 } 5841 5842 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr)); 5843 if (!nla) 5844 goto nla_put_failure; 5845 read_lock_bh(&idev->lock); 5846 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla)); 5847 read_unlock_bh(&idev->lock); 5848 5849 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, 5850 READ_ONCE(idev->cnf.addr_gen_mode))) 5851 goto nla_put_failure; 5852 5853 ra_mtu = READ_ONCE(idev->ra_mtu); 5854 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu)) 5855 goto nla_put_failure; 5856 5857 return 0; 5858 5859 nla_put_failure: 5860 return -EMSGSIZE; 5861 } 5862 5863 static size_t inet6_get_link_af_size(const struct net_device *dev, 5864 u32 ext_filter_mask) 5865 { 5866 if (!__in6_dev_get(dev)) 5867 return 0; 5868 5869 return inet6_ifla6_size(); 5870 } 5871 5872 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev, 5873 u32 ext_filter_mask) 5874 { 5875 struct inet6_dev *idev = __in6_dev_get(dev); 5876 5877 if (!idev) 5878 return -ENODATA; 5879 5880 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0) 5881 return -EMSGSIZE; 5882 5883 return 0; 5884 } 5885 5886 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token, 5887 struct netlink_ext_ack *extack) 5888 { 5889 struct inet6_ifaddr *ifp; 5890 struct net_device *dev = idev->dev; 5891 bool clear_token, update_rs = false; 5892 struct in6_addr ll_addr; 5893 5894 ASSERT_RTNL(); 5895 5896 if (!token) 5897 return -EINVAL; 5898 5899 if (dev->flags & IFF_LOOPBACK) { 5900 NL_SET_ERR_MSG_MOD(extack, "Device is loopback"); 5901 return -EINVAL; 5902 } 5903 5904 if (dev->flags & IFF_NOARP) { 5905 NL_SET_ERR_MSG_MOD(extack, 5906 "Device does not do neighbour discovery"); 5907 return -EINVAL; 5908 } 5909 5910 if (!ipv6_accept_ra(idev)) { 5911 NL_SET_ERR_MSG_MOD(extack, 5912 "Router advertisement is disabled on device"); 5913 return -EINVAL; 5914 } 5915 5916 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) { 5917 NL_SET_ERR_MSG(extack, 5918 "Router solicitation is disabled on device"); 5919 return -EINVAL; 5920 } 5921 5922 write_lock_bh(&idev->lock); 5923 5924 BUILD_BUG_ON(sizeof(token->s6_addr) != 16); 5925 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8); 5926 5927 write_unlock_bh(&idev->lock); 5928 5929 clear_token = ipv6_addr_any(token); 5930 if (clear_token) 5931 goto update_lft; 5932 5933 if (!idev->dead && (idev->if_flags & IF_READY) && 5934 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE | 5935 IFA_F_OPTIMISTIC)) { 5936 /* If we're not ready, then normal ifup will take care 5937 * of this. Otherwise, we need to request our rs here. 5938 */ 5939 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters); 5940 update_rs = true; 5941 } 5942 5943 update_lft: 5944 write_lock_bh(&idev->lock); 5945 5946 if (update_rs) { 5947 idev->if_flags |= IF_RS_SENT; 5948 idev->rs_interval = rfc3315_s14_backoff_init( 5949 READ_ONCE(idev->cnf.rtr_solicit_interval)); 5950 idev->rs_probes = 1; 5951 addrconf_mod_rs_timer(idev, idev->rs_interval); 5952 } 5953 5954 /* Well, that's kinda nasty ... */ 5955 list_for_each_entry(ifp, &idev->addr_list, if_list) { 5956 spin_lock(&ifp->lock); 5957 if (ifp->tokenized) { 5958 ifp->valid_lft = 0; 5959 ifp->prefered_lft = 0; 5960 } 5961 spin_unlock(&ifp->lock); 5962 } 5963 5964 write_unlock_bh(&idev->lock); 5965 inet6_ifinfo_notify(RTM_NEWLINK, idev); 5966 addrconf_verify_rtnl(dev_net(dev)); 5967 return 0; 5968 } 5969 5970 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = { 5971 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 }, 5972 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) }, 5973 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT, 5974 .reject_message = 5975 "IFLA_INET6_RA_MTU can not be set" }, 5976 }; 5977 5978 static int check_addr_gen_mode(int mode) 5979 { 5980 if (mode != IN6_ADDR_GEN_MODE_EUI64 && 5981 mode != IN6_ADDR_GEN_MODE_NONE && 5982 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 5983 mode != IN6_ADDR_GEN_MODE_RANDOM) 5984 return -EINVAL; 5985 return 1; 5986 } 5987 5988 static int check_stable_privacy(struct inet6_dev *idev, struct net *net, 5989 int mode) 5990 { 5991 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 5992 !idev->cnf.stable_secret.initialized && 5993 !net->ipv6.devconf_dflt->stable_secret.initialized) 5994 return -EINVAL; 5995 return 1; 5996 } 5997 5998 static int inet6_validate_link_af(const struct net_device *dev, 5999 const struct nlattr *nla, 6000 struct netlink_ext_ack *extack) 6001 { 6002 struct nlattr *tb[IFLA_INET6_MAX + 1]; 6003 struct inet6_dev *idev = NULL; 6004 int err; 6005 6006 if (dev) { 6007 idev = __in6_dev_get(dev); 6008 if (!idev) 6009 return -EAFNOSUPPORT; 6010 } 6011 6012 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, 6013 inet6_af_policy, extack); 6014 if (err) 6015 return err; 6016 6017 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE]) 6018 return -EINVAL; 6019 6020 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 6021 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 6022 6023 if (check_addr_gen_mode(mode) < 0) 6024 return -EINVAL; 6025 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0) 6026 return -EINVAL; 6027 } 6028 6029 return 0; 6030 } 6031 6032 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla, 6033 struct netlink_ext_ack *extack) 6034 { 6035 struct inet6_dev *idev = __in6_dev_get(dev); 6036 struct nlattr *tb[IFLA_INET6_MAX + 1]; 6037 int err; 6038 6039 if (!idev) 6040 return -EAFNOSUPPORT; 6041 6042 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0) 6043 return -EINVAL; 6044 6045 if (tb[IFLA_INET6_TOKEN]) { 6046 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]), 6047 extack); 6048 if (err) 6049 return err; 6050 } 6051 6052 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 6053 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 6054 6055 WRITE_ONCE(idev->cnf.addr_gen_mode, mode); 6056 } 6057 6058 return 0; 6059 } 6060 6061 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, 6062 u32 portid, u32 seq, int event, unsigned int flags) 6063 { 6064 struct net_device *dev = idev->dev; 6065 struct ifinfomsg *hdr; 6066 struct nlmsghdr *nlh; 6067 int ifindex, iflink; 6068 void *protoinfo; 6069 6070 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags); 6071 if (!nlh) 6072 return -EMSGSIZE; 6073 6074 hdr = nlmsg_data(nlh); 6075 hdr->ifi_family = AF_INET6; 6076 hdr->__ifi_pad = 0; 6077 hdr->ifi_type = dev->type; 6078 ifindex = READ_ONCE(dev->ifindex); 6079 hdr->ifi_index = ifindex; 6080 hdr->ifi_flags = dev_get_flags(dev); 6081 hdr->ifi_change = 0; 6082 6083 iflink = dev_get_iflink(dev); 6084 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 6085 (dev->addr_len && 6086 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 6087 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) || 6088 (ifindex != iflink && 6089 nla_put_u32(skb, IFLA_LINK, iflink)) || 6090 nla_put_u8(skb, IFLA_OPERSTATE, 6091 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN)) 6092 goto nla_put_failure; 6093 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO); 6094 if (!protoinfo) 6095 goto nla_put_failure; 6096 6097 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0) 6098 goto nla_put_failure; 6099 6100 nla_nest_end(skb, protoinfo); 6101 nlmsg_end(skb, nlh); 6102 return 0; 6103 6104 nla_put_failure: 6105 nlmsg_cancel(skb, nlh); 6106 return -EMSGSIZE; 6107 } 6108 6109 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh, 6110 struct netlink_ext_ack *extack) 6111 { 6112 struct ifinfomsg *ifm; 6113 6114 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 6115 if (!ifm) { 6116 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request"); 6117 return -EINVAL; 6118 } 6119 6120 if (nlmsg_attrlen(nlh, sizeof(*ifm))) { 6121 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header"); 6122 return -EINVAL; 6123 } 6124 6125 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 6126 ifm->ifi_change || ifm->ifi_index) { 6127 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request"); 6128 return -EINVAL; 6129 } 6130 6131 return 0; 6132 } 6133 6134 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 6135 { 6136 struct net *net = sock_net(skb->sk); 6137 struct { 6138 unsigned long ifindex; 6139 } *ctx = (void *)cb->ctx; 6140 struct net_device *dev; 6141 struct inet6_dev *idev; 6142 int err; 6143 6144 /* only requests using strict checking can pass data to 6145 * influence the dump 6146 */ 6147 if (cb->strict_check) { 6148 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack); 6149 6150 if (err < 0) 6151 return err; 6152 } 6153 6154 err = 0; 6155 rcu_read_lock(); 6156 for_each_netdev_dump(net, dev, ctx->ifindex) { 6157 idev = __in6_dev_get(dev); 6158 if (!idev) 6159 continue; 6160 err = inet6_fill_ifinfo(skb, idev, 6161 NETLINK_CB(cb->skb).portid, 6162 cb->nlh->nlmsg_seq, 6163 RTM_NEWLINK, NLM_F_MULTI); 6164 if (err < 0) 6165 break; 6166 } 6167 rcu_read_unlock(); 6168 6169 return err; 6170 } 6171 6172 void inet6_ifinfo_notify(int event, struct inet6_dev *idev) 6173 { 6174 struct sk_buff *skb; 6175 struct net *net = dev_net(idev->dev); 6176 int err = -ENOBUFS; 6177 6178 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC); 6179 if (!skb) 6180 goto errout; 6181 6182 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0); 6183 if (err < 0) { 6184 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */ 6185 WARN_ON(err == -EMSGSIZE); 6186 kfree_skb(skb); 6187 goto errout; 6188 } 6189 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC); 6190 return; 6191 errout: 6192 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err); 6193 } 6194 6195 static inline size_t inet6_prefix_nlmsg_size(void) 6196 { 6197 return NLMSG_ALIGN(sizeof(struct prefixmsg)) 6198 + nla_total_size(sizeof(struct in6_addr)) 6199 + nla_total_size(sizeof(struct prefix_cacheinfo)); 6200 } 6201 6202 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev, 6203 struct prefix_info *pinfo, u32 portid, u32 seq, 6204 int event, unsigned int flags) 6205 { 6206 struct prefixmsg *pmsg; 6207 struct nlmsghdr *nlh; 6208 struct prefix_cacheinfo ci; 6209 6210 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags); 6211 if (!nlh) 6212 return -EMSGSIZE; 6213 6214 pmsg = nlmsg_data(nlh); 6215 pmsg->prefix_family = AF_INET6; 6216 pmsg->prefix_pad1 = 0; 6217 pmsg->prefix_pad2 = 0; 6218 pmsg->prefix_ifindex = idev->dev->ifindex; 6219 pmsg->prefix_len = pinfo->prefix_len; 6220 pmsg->prefix_type = pinfo->type; 6221 pmsg->prefix_pad3 = 0; 6222 pmsg->prefix_flags = pinfo->flags; 6223 6224 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix)) 6225 goto nla_put_failure; 6226 ci.preferred_time = ntohl(pinfo->prefered); 6227 ci.valid_time = ntohl(pinfo->valid); 6228 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci)) 6229 goto nla_put_failure; 6230 nlmsg_end(skb, nlh); 6231 return 0; 6232 6233 nla_put_failure: 6234 nlmsg_cancel(skb, nlh); 6235 return -EMSGSIZE; 6236 } 6237 6238 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 6239 struct prefix_info *pinfo) 6240 { 6241 struct sk_buff *skb; 6242 struct net *net = dev_net(idev->dev); 6243 int err = -ENOBUFS; 6244 6245 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC); 6246 if (!skb) 6247 goto errout; 6248 6249 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0); 6250 if (err < 0) { 6251 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */ 6252 WARN_ON(err == -EMSGSIZE); 6253 kfree_skb(skb); 6254 goto errout; 6255 } 6256 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC); 6257 return; 6258 errout: 6259 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err); 6260 } 6261 6262 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6263 { 6264 struct net *net = dev_net(ifp->idev->dev); 6265 6266 if (event) 6267 ASSERT_RTNL(); 6268 6269 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp); 6270 6271 switch (event) { 6272 case RTM_NEWADDR: 6273 /* 6274 * If the address was optimistic we inserted the route at the 6275 * start of our DAD process, so we don't need to do it again. 6276 * If the device was taken down in the middle of the DAD 6277 * cycle there is a race where we could get here without a 6278 * host route, so nothing to insert. That will be fixed when 6279 * the device is brought up. 6280 */ 6281 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) { 6282 ip6_ins_rt(net, ifp->rt); 6283 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) { 6284 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n", 6285 &ifp->addr, ifp->idev->dev->name); 6286 } 6287 6288 if (ifp->idev->cnf.forwarding) 6289 addrconf_join_anycast(ifp); 6290 if (!ipv6_addr_any(&ifp->peer_addr)) 6291 addrconf_prefix_route(&ifp->peer_addr, 128, 6292 ifp->rt_priority, ifp->idev->dev, 6293 0, 0, GFP_ATOMIC); 6294 break; 6295 case RTM_DELADDR: 6296 if (ifp->idev->cnf.forwarding) 6297 addrconf_leave_anycast(ifp); 6298 addrconf_leave_solict(ifp->idev, &ifp->addr); 6299 if (!ipv6_addr_any(&ifp->peer_addr)) { 6300 struct fib6_info *rt; 6301 6302 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128, 6303 ifp->idev->dev, 0, 0, 6304 false); 6305 if (rt) 6306 ip6_del_rt(net, rt, false); 6307 } 6308 if (ifp->rt) { 6309 ip6_del_rt(net, ifp->rt, false); 6310 ifp->rt = NULL; 6311 } 6312 rt_genid_bump_ipv6(net); 6313 break; 6314 } 6315 atomic_inc(&net->ipv6.dev_addr_genid); 6316 } 6317 6318 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6319 { 6320 if (likely(ifp->idev->dead == 0)) 6321 __ipv6_ifa_notify(event, ifp); 6322 } 6323 6324 #ifdef CONFIG_SYSCTL 6325 6326 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write, 6327 void *buffer, size_t *lenp, loff_t *ppos) 6328 { 6329 int *valp = ctl->data; 6330 int val = *valp; 6331 loff_t pos = *ppos; 6332 struct ctl_table lctl; 6333 int ret; 6334 6335 /* 6336 * ctl->data points to idev->cnf.forwarding, we should 6337 * not modify it until we get the rtnl lock. 6338 */ 6339 lctl = *ctl; 6340 lctl.data = &val; 6341 6342 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6343 6344 if (write) 6345 ret = addrconf_fixup_forwarding(ctl, valp, val); 6346 if (ret) 6347 *ppos = pos; 6348 return ret; 6349 } 6350 6351 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write, 6352 void *buffer, size_t *lenp, loff_t *ppos) 6353 { 6354 struct inet6_dev *idev = ctl->extra1; 6355 int min_mtu = IPV6_MIN_MTU; 6356 struct ctl_table lctl; 6357 6358 lctl = *ctl; 6359 lctl.extra1 = &min_mtu; 6360 lctl.extra2 = idev ? &idev->dev->mtu : NULL; 6361 6362 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos); 6363 } 6364 6365 static void dev_disable_change(struct inet6_dev *idev) 6366 { 6367 struct netdev_notifier_info info; 6368 6369 if (!idev || !idev->dev) 6370 return; 6371 6372 netdev_notifier_info_init(&info, idev->dev); 6373 if (idev->cnf.disable_ipv6) 6374 addrconf_notify(NULL, NETDEV_DOWN, &info); 6375 else 6376 addrconf_notify(NULL, NETDEV_UP, &info); 6377 } 6378 6379 static void addrconf_disable_change(struct net *net, __s32 newf) 6380 { 6381 struct net_device *dev; 6382 struct inet6_dev *idev; 6383 6384 for_each_netdev(net, dev) { 6385 idev = __in6_dev_get_rtnl_net(dev); 6386 if (idev) { 6387 int changed = (!idev->cnf.disable_ipv6) ^ (!newf); 6388 6389 WRITE_ONCE(idev->cnf.disable_ipv6, newf); 6390 if (changed) 6391 dev_disable_change(idev); 6392 } 6393 } 6394 } 6395 6396 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf) 6397 { 6398 struct net *net = (struct net *)table->extra2; 6399 int old; 6400 6401 if (p == &net->ipv6.devconf_dflt->disable_ipv6) { 6402 WRITE_ONCE(*p, newf); 6403 return 0; 6404 } 6405 6406 if (!rtnl_net_trylock(net)) 6407 return restart_syscall(); 6408 6409 old = *p; 6410 WRITE_ONCE(*p, newf); 6411 6412 if (p == &net->ipv6.devconf_all->disable_ipv6) { 6413 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf); 6414 addrconf_disable_change(net, newf); 6415 } else if ((!newf) ^ (!old)) { 6416 dev_disable_change((struct inet6_dev *)table->extra1); 6417 } 6418 6419 rtnl_net_unlock(net); 6420 return 0; 6421 } 6422 6423 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write, 6424 void *buffer, size_t *lenp, loff_t *ppos) 6425 { 6426 int *valp = ctl->data; 6427 int val = *valp; 6428 loff_t pos = *ppos; 6429 struct ctl_table lctl; 6430 int ret; 6431 6432 /* 6433 * ctl->data points to idev->cnf.disable_ipv6, we should 6434 * not modify it until we get the rtnl lock. 6435 */ 6436 lctl = *ctl; 6437 lctl.data = &val; 6438 6439 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6440 6441 if (write) 6442 ret = addrconf_disable_ipv6(ctl, valp, val); 6443 if (ret) 6444 *ppos = pos; 6445 return ret; 6446 } 6447 6448 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write, 6449 void *buffer, size_t *lenp, loff_t *ppos) 6450 { 6451 int *valp = ctl->data; 6452 int ret; 6453 int old, new; 6454 6455 old = *valp; 6456 ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 6457 new = *valp; 6458 6459 if (write && old != new) { 6460 struct net *net = ctl->extra2; 6461 6462 if (!rtnl_net_trylock(net)) 6463 return restart_syscall(); 6464 6465 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) { 6466 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6467 NETCONFA_PROXY_NEIGH, 6468 NETCONFA_IFINDEX_DEFAULT, 6469 net->ipv6.devconf_dflt); 6470 } else if (valp == &net->ipv6.devconf_all->proxy_ndp) { 6471 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6472 NETCONFA_PROXY_NEIGH, 6473 NETCONFA_IFINDEX_ALL, 6474 net->ipv6.devconf_all); 6475 } else { 6476 struct inet6_dev *idev = ctl->extra1; 6477 6478 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6479 NETCONFA_PROXY_NEIGH, 6480 idev->dev->ifindex, 6481 &idev->cnf); 6482 } 6483 rtnl_net_unlock(net); 6484 } 6485 6486 return ret; 6487 } 6488 6489 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write, 6490 void *buffer, size_t *lenp, 6491 loff_t *ppos) 6492 { 6493 int ret = 0; 6494 u32 new_val; 6495 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1; 6496 struct net *net = (struct net *)ctl->extra2; 6497 struct ctl_table tmp = { 6498 .data = &new_val, 6499 .maxlen = sizeof(new_val), 6500 .mode = ctl->mode, 6501 }; 6502 6503 if (!rtnl_net_trylock(net)) 6504 return restart_syscall(); 6505 6506 new_val = *((u32 *)ctl->data); 6507 6508 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos); 6509 if (ret != 0) 6510 goto out; 6511 6512 if (write) { 6513 if (check_addr_gen_mode(new_val) < 0) { 6514 ret = -EINVAL; 6515 goto out; 6516 } 6517 6518 if (idev) { 6519 if (check_stable_privacy(idev, net, new_val) < 0) { 6520 ret = -EINVAL; 6521 goto out; 6522 } 6523 6524 if (idev->cnf.addr_gen_mode != new_val) { 6525 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val); 6526 netdev_lock_ops(idev->dev); 6527 addrconf_init_auto_addrs(idev->dev); 6528 netdev_unlock_ops(idev->dev); 6529 } 6530 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) { 6531 struct net_device *dev; 6532 6533 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val); 6534 for_each_netdev(net, dev) { 6535 idev = __in6_dev_get_rtnl_net(dev); 6536 if (idev && 6537 idev->cnf.addr_gen_mode != new_val) { 6538 WRITE_ONCE(idev->cnf.addr_gen_mode, 6539 new_val); 6540 netdev_lock_ops(idev->dev); 6541 addrconf_init_auto_addrs(idev->dev); 6542 netdev_unlock_ops(idev->dev); 6543 } 6544 } 6545 } 6546 6547 WRITE_ONCE(*((u32 *)ctl->data), new_val); 6548 } 6549 6550 out: 6551 rtnl_net_unlock(net); 6552 6553 return ret; 6554 } 6555 6556 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write, 6557 void *buffer, size_t *lenp, 6558 loff_t *ppos) 6559 { 6560 int err; 6561 struct in6_addr addr; 6562 char str[IPV6_MAX_STRLEN]; 6563 struct ctl_table lctl = *ctl; 6564 struct net *net = ctl->extra2; 6565 struct ipv6_stable_secret *secret = ctl->data; 6566 6567 if (&net->ipv6.devconf_all->stable_secret == ctl->data) 6568 return -EIO; 6569 6570 lctl.maxlen = IPV6_MAX_STRLEN; 6571 lctl.data = str; 6572 6573 if (!rtnl_net_trylock(net)) 6574 return restart_syscall(); 6575 6576 if (!write && !secret->initialized) { 6577 err = -EIO; 6578 goto out; 6579 } 6580 6581 err = snprintf(str, sizeof(str), "%pI6", &secret->secret); 6582 if (err >= sizeof(str)) { 6583 err = -EIO; 6584 goto out; 6585 } 6586 6587 err = proc_dostring(&lctl, write, buffer, lenp, ppos); 6588 if (err || !write) 6589 goto out; 6590 6591 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) { 6592 err = -EIO; 6593 goto out; 6594 } 6595 6596 secret->initialized = true; 6597 secret->secret = addr; 6598 6599 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) { 6600 struct net_device *dev; 6601 6602 for_each_netdev(net, dev) { 6603 struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev); 6604 6605 if (idev) { 6606 WRITE_ONCE(idev->cnf.addr_gen_mode, 6607 IN6_ADDR_GEN_MODE_STABLE_PRIVACY); 6608 } 6609 } 6610 } else { 6611 struct inet6_dev *idev = ctl->extra1; 6612 6613 WRITE_ONCE(idev->cnf.addr_gen_mode, 6614 IN6_ADDR_GEN_MODE_STABLE_PRIVACY); 6615 } 6616 6617 out: 6618 rtnl_net_unlock(net); 6619 6620 return err; 6621 } 6622 6623 static 6624 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl, 6625 int write, void *buffer, 6626 size_t *lenp, 6627 loff_t *ppos) 6628 { 6629 int *valp = ctl->data; 6630 int val = *valp; 6631 loff_t pos = *ppos; 6632 struct ctl_table lctl; 6633 int ret; 6634 6635 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown 6636 * we should not modify it until we get the rtnl lock. 6637 */ 6638 lctl = *ctl; 6639 lctl.data = &val; 6640 6641 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6642 6643 if (write) 6644 ret = addrconf_fixup_linkdown(ctl, valp, val); 6645 if (ret) 6646 *ppos = pos; 6647 return ret; 6648 } 6649 6650 static 6651 void addrconf_set_nopolicy(struct rt6_info *rt, int action) 6652 { 6653 if (rt) { 6654 if (action) 6655 rt->dst.flags |= DST_NOPOLICY; 6656 else 6657 rt->dst.flags &= ~DST_NOPOLICY; 6658 } 6659 } 6660 6661 static 6662 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val) 6663 { 6664 struct inet6_ifaddr *ifa; 6665 6666 read_lock_bh(&idev->lock); 6667 list_for_each_entry(ifa, &idev->addr_list, if_list) { 6668 spin_lock(&ifa->lock); 6669 if (ifa->rt) { 6670 /* host routes only use builtin fib6_nh */ 6671 struct fib6_nh *nh = ifa->rt->fib6_nh; 6672 int cpu; 6673 6674 rcu_read_lock(); 6675 ifa->rt->dst_nopolicy = val ? true : false; 6676 if (nh->rt6i_pcpu) { 6677 for_each_possible_cpu(cpu) { 6678 struct rt6_info **rtp; 6679 6680 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu); 6681 addrconf_set_nopolicy(*rtp, val); 6682 } 6683 } 6684 rcu_read_unlock(); 6685 } 6686 spin_unlock(&ifa->lock); 6687 } 6688 read_unlock_bh(&idev->lock); 6689 } 6690 6691 static 6692 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val) 6693 { 6694 struct net *net = (struct net *)ctl->extra2; 6695 struct inet6_dev *idev; 6696 6697 if (valp == &net->ipv6.devconf_dflt->disable_policy) { 6698 WRITE_ONCE(*valp, val); 6699 return 0; 6700 } 6701 6702 if (!rtnl_net_trylock(net)) 6703 return restart_syscall(); 6704 6705 WRITE_ONCE(*valp, val); 6706 6707 if (valp == &net->ipv6.devconf_all->disable_policy) { 6708 struct net_device *dev; 6709 6710 for_each_netdev(net, dev) { 6711 idev = __in6_dev_get_rtnl_net(dev); 6712 if (idev) 6713 addrconf_disable_policy_idev(idev, val); 6714 } 6715 } else { 6716 idev = (struct inet6_dev *)ctl->extra1; 6717 addrconf_disable_policy_idev(idev, val); 6718 } 6719 6720 rtnl_net_unlock(net); 6721 return 0; 6722 } 6723 6724 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write, 6725 void *buffer, size_t *lenp, loff_t *ppos) 6726 { 6727 int *valp = ctl->data; 6728 int val = *valp; 6729 loff_t pos = *ppos; 6730 struct ctl_table lctl; 6731 int ret; 6732 6733 lctl = *ctl; 6734 lctl.data = &val; 6735 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6736 6737 if (write && (*valp != val)) 6738 ret = addrconf_disable_policy(ctl, valp, val); 6739 6740 if (ret) 6741 *ppos = pos; 6742 6743 return ret; 6744 } 6745 6746 static int minus_one = -1; 6747 static const int two_five_five = 255; 6748 static u32 ioam6_if_id_max = U16_MAX; 6749 6750 static const struct ctl_table addrconf_sysctl[] = { 6751 { 6752 .procname = "forwarding", 6753 .data = &ipv6_devconf.forwarding, 6754 .maxlen = sizeof(int), 6755 .mode = 0644, 6756 .proc_handler = addrconf_sysctl_forward, 6757 }, 6758 { 6759 .procname = "hop_limit", 6760 .data = &ipv6_devconf.hop_limit, 6761 .maxlen = sizeof(int), 6762 .mode = 0644, 6763 .proc_handler = proc_dointvec_minmax, 6764 .extra1 = (void *)SYSCTL_ONE, 6765 .extra2 = (void *)&two_five_five, 6766 }, 6767 { 6768 .procname = "mtu", 6769 .data = &ipv6_devconf.mtu6, 6770 .maxlen = sizeof(int), 6771 .mode = 0644, 6772 .proc_handler = addrconf_sysctl_mtu, 6773 }, 6774 { 6775 .procname = "accept_ra", 6776 .data = &ipv6_devconf.accept_ra, 6777 .maxlen = sizeof(int), 6778 .mode = 0644, 6779 .proc_handler = proc_dointvec, 6780 }, 6781 { 6782 .procname = "accept_redirects", 6783 .data = &ipv6_devconf.accept_redirects, 6784 .maxlen = sizeof(int), 6785 .mode = 0644, 6786 .proc_handler = proc_dointvec, 6787 }, 6788 { 6789 .procname = "autoconf", 6790 .data = &ipv6_devconf.autoconf, 6791 .maxlen = sizeof(int), 6792 .mode = 0644, 6793 .proc_handler = proc_dointvec, 6794 }, 6795 { 6796 .procname = "dad_transmits", 6797 .data = &ipv6_devconf.dad_transmits, 6798 .maxlen = sizeof(int), 6799 .mode = 0644, 6800 .proc_handler = proc_dointvec, 6801 }, 6802 { 6803 .procname = "router_solicitations", 6804 .data = &ipv6_devconf.rtr_solicits, 6805 .maxlen = sizeof(int), 6806 .mode = 0644, 6807 .proc_handler = proc_dointvec_minmax, 6808 .extra1 = &minus_one, 6809 }, 6810 { 6811 .procname = "router_solicitation_interval", 6812 .data = &ipv6_devconf.rtr_solicit_interval, 6813 .maxlen = sizeof(int), 6814 .mode = 0644, 6815 .proc_handler = proc_dointvec_jiffies, 6816 }, 6817 { 6818 .procname = "router_solicitation_max_interval", 6819 .data = &ipv6_devconf.rtr_solicit_max_interval, 6820 .maxlen = sizeof(int), 6821 .mode = 0644, 6822 .proc_handler = proc_dointvec_jiffies, 6823 }, 6824 { 6825 .procname = "router_solicitation_delay", 6826 .data = &ipv6_devconf.rtr_solicit_delay, 6827 .maxlen = sizeof(int), 6828 .mode = 0644, 6829 .proc_handler = proc_dointvec_jiffies, 6830 }, 6831 { 6832 .procname = "force_mld_version", 6833 .data = &ipv6_devconf.force_mld_version, 6834 .maxlen = sizeof(int), 6835 .mode = 0644, 6836 .proc_handler = proc_dointvec, 6837 }, 6838 { 6839 .procname = "mldv1_unsolicited_report_interval", 6840 .data = 6841 &ipv6_devconf.mldv1_unsolicited_report_interval, 6842 .maxlen = sizeof(int), 6843 .mode = 0644, 6844 .proc_handler = proc_dointvec_ms_jiffies, 6845 }, 6846 { 6847 .procname = "mldv2_unsolicited_report_interval", 6848 .data = 6849 &ipv6_devconf.mldv2_unsolicited_report_interval, 6850 .maxlen = sizeof(int), 6851 .mode = 0644, 6852 .proc_handler = proc_dointvec_ms_jiffies, 6853 }, 6854 { 6855 .procname = "use_tempaddr", 6856 .data = &ipv6_devconf.use_tempaddr, 6857 .maxlen = sizeof(int), 6858 .mode = 0644, 6859 .proc_handler = proc_dointvec, 6860 }, 6861 { 6862 .procname = "temp_valid_lft", 6863 .data = &ipv6_devconf.temp_valid_lft, 6864 .maxlen = sizeof(int), 6865 .mode = 0644, 6866 .proc_handler = proc_dointvec, 6867 }, 6868 { 6869 .procname = "temp_prefered_lft", 6870 .data = &ipv6_devconf.temp_prefered_lft, 6871 .maxlen = sizeof(int), 6872 .mode = 0644, 6873 .proc_handler = proc_dointvec, 6874 }, 6875 { 6876 .procname = "regen_min_advance", 6877 .data = &ipv6_devconf.regen_min_advance, 6878 .maxlen = sizeof(int), 6879 .mode = 0644, 6880 .proc_handler = proc_dointvec, 6881 }, 6882 { 6883 .procname = "regen_max_retry", 6884 .data = &ipv6_devconf.regen_max_retry, 6885 .maxlen = sizeof(int), 6886 .mode = 0644, 6887 .proc_handler = proc_dointvec, 6888 }, 6889 { 6890 .procname = "max_desync_factor", 6891 .data = &ipv6_devconf.max_desync_factor, 6892 .maxlen = sizeof(int), 6893 .mode = 0644, 6894 .proc_handler = proc_dointvec, 6895 }, 6896 { 6897 .procname = "max_addresses", 6898 .data = &ipv6_devconf.max_addresses, 6899 .maxlen = sizeof(int), 6900 .mode = 0644, 6901 .proc_handler = proc_dointvec, 6902 }, 6903 { 6904 .procname = "accept_ra_defrtr", 6905 .data = &ipv6_devconf.accept_ra_defrtr, 6906 .maxlen = sizeof(int), 6907 .mode = 0644, 6908 .proc_handler = proc_dointvec, 6909 }, 6910 { 6911 .procname = "ra_defrtr_metric", 6912 .data = &ipv6_devconf.ra_defrtr_metric, 6913 .maxlen = sizeof(u32), 6914 .mode = 0644, 6915 .proc_handler = proc_douintvec_minmax, 6916 .extra1 = (void *)SYSCTL_ONE, 6917 }, 6918 { 6919 .procname = "accept_ra_min_hop_limit", 6920 .data = &ipv6_devconf.accept_ra_min_hop_limit, 6921 .maxlen = sizeof(int), 6922 .mode = 0644, 6923 .proc_handler = proc_dointvec, 6924 }, 6925 { 6926 .procname = "accept_ra_min_lft", 6927 .data = &ipv6_devconf.accept_ra_min_lft, 6928 .maxlen = sizeof(int), 6929 .mode = 0644, 6930 .proc_handler = proc_dointvec, 6931 }, 6932 { 6933 .procname = "accept_ra_pinfo", 6934 .data = &ipv6_devconf.accept_ra_pinfo, 6935 .maxlen = sizeof(int), 6936 .mode = 0644, 6937 .proc_handler = proc_dointvec, 6938 }, 6939 { 6940 .procname = "ra_honor_pio_life", 6941 .data = &ipv6_devconf.ra_honor_pio_life, 6942 .maxlen = sizeof(u8), 6943 .mode = 0644, 6944 .proc_handler = proc_dou8vec_minmax, 6945 .extra1 = SYSCTL_ZERO, 6946 .extra2 = SYSCTL_ONE, 6947 }, 6948 { 6949 .procname = "ra_honor_pio_pflag", 6950 .data = &ipv6_devconf.ra_honor_pio_pflag, 6951 .maxlen = sizeof(u8), 6952 .mode = 0644, 6953 .proc_handler = proc_dou8vec_minmax, 6954 .extra1 = SYSCTL_ZERO, 6955 .extra2 = SYSCTL_ONE, 6956 }, 6957 #ifdef CONFIG_IPV6_ROUTER_PREF 6958 { 6959 .procname = "accept_ra_rtr_pref", 6960 .data = &ipv6_devconf.accept_ra_rtr_pref, 6961 .maxlen = sizeof(int), 6962 .mode = 0644, 6963 .proc_handler = proc_dointvec, 6964 }, 6965 { 6966 .procname = "router_probe_interval", 6967 .data = &ipv6_devconf.rtr_probe_interval, 6968 .maxlen = sizeof(int), 6969 .mode = 0644, 6970 .proc_handler = proc_dointvec_jiffies, 6971 }, 6972 #ifdef CONFIG_IPV6_ROUTE_INFO 6973 { 6974 .procname = "accept_ra_rt_info_min_plen", 6975 .data = &ipv6_devconf.accept_ra_rt_info_min_plen, 6976 .maxlen = sizeof(int), 6977 .mode = 0644, 6978 .proc_handler = proc_dointvec, 6979 }, 6980 { 6981 .procname = "accept_ra_rt_info_max_plen", 6982 .data = &ipv6_devconf.accept_ra_rt_info_max_plen, 6983 .maxlen = sizeof(int), 6984 .mode = 0644, 6985 .proc_handler = proc_dointvec, 6986 }, 6987 #endif 6988 #endif 6989 { 6990 .procname = "proxy_ndp", 6991 .data = &ipv6_devconf.proxy_ndp, 6992 .maxlen = sizeof(int), 6993 .mode = 0644, 6994 .proc_handler = addrconf_sysctl_proxy_ndp, 6995 }, 6996 { 6997 .procname = "accept_source_route", 6998 .data = &ipv6_devconf.accept_source_route, 6999 .maxlen = sizeof(int), 7000 .mode = 0644, 7001 .proc_handler = proc_dointvec, 7002 }, 7003 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 7004 { 7005 .procname = "optimistic_dad", 7006 .data = &ipv6_devconf.optimistic_dad, 7007 .maxlen = sizeof(int), 7008 .mode = 0644, 7009 .proc_handler = proc_dointvec, 7010 }, 7011 { 7012 .procname = "use_optimistic", 7013 .data = &ipv6_devconf.use_optimistic, 7014 .maxlen = sizeof(int), 7015 .mode = 0644, 7016 .proc_handler = proc_dointvec, 7017 }, 7018 #endif 7019 #ifdef CONFIG_IPV6_MROUTE 7020 { 7021 .procname = "mc_forwarding", 7022 .data = &ipv6_devconf.mc_forwarding, 7023 .maxlen = sizeof(int), 7024 .mode = 0444, 7025 .proc_handler = proc_dointvec, 7026 }, 7027 #endif 7028 { 7029 .procname = "disable_ipv6", 7030 .data = &ipv6_devconf.disable_ipv6, 7031 .maxlen = sizeof(int), 7032 .mode = 0644, 7033 .proc_handler = addrconf_sysctl_disable, 7034 }, 7035 { 7036 .procname = "accept_dad", 7037 .data = &ipv6_devconf.accept_dad, 7038 .maxlen = sizeof(int), 7039 .mode = 0644, 7040 .proc_handler = proc_dointvec, 7041 }, 7042 { 7043 .procname = "force_tllao", 7044 .data = &ipv6_devconf.force_tllao, 7045 .maxlen = sizeof(int), 7046 .mode = 0644, 7047 .proc_handler = proc_dointvec 7048 }, 7049 { 7050 .procname = "ndisc_notify", 7051 .data = &ipv6_devconf.ndisc_notify, 7052 .maxlen = sizeof(int), 7053 .mode = 0644, 7054 .proc_handler = proc_dointvec 7055 }, 7056 { 7057 .procname = "suppress_frag_ndisc", 7058 .data = &ipv6_devconf.suppress_frag_ndisc, 7059 .maxlen = sizeof(int), 7060 .mode = 0644, 7061 .proc_handler = proc_dointvec 7062 }, 7063 { 7064 .procname = "accept_ra_from_local", 7065 .data = &ipv6_devconf.accept_ra_from_local, 7066 .maxlen = sizeof(int), 7067 .mode = 0644, 7068 .proc_handler = proc_dointvec, 7069 }, 7070 { 7071 .procname = "accept_ra_mtu", 7072 .data = &ipv6_devconf.accept_ra_mtu, 7073 .maxlen = sizeof(int), 7074 .mode = 0644, 7075 .proc_handler = proc_dointvec, 7076 }, 7077 { 7078 .procname = "stable_secret", 7079 .data = &ipv6_devconf.stable_secret, 7080 .maxlen = IPV6_MAX_STRLEN, 7081 .mode = 0600, 7082 .proc_handler = addrconf_sysctl_stable_secret, 7083 }, 7084 { 7085 .procname = "use_oif_addrs_only", 7086 .data = &ipv6_devconf.use_oif_addrs_only, 7087 .maxlen = sizeof(int), 7088 .mode = 0644, 7089 .proc_handler = proc_dointvec, 7090 }, 7091 { 7092 .procname = "ignore_routes_with_linkdown", 7093 .data = &ipv6_devconf.ignore_routes_with_linkdown, 7094 .maxlen = sizeof(int), 7095 .mode = 0644, 7096 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown, 7097 }, 7098 { 7099 .procname = "drop_unicast_in_l2_multicast", 7100 .data = &ipv6_devconf.drop_unicast_in_l2_multicast, 7101 .maxlen = sizeof(int), 7102 .mode = 0644, 7103 .proc_handler = proc_dointvec, 7104 }, 7105 { 7106 .procname = "drop_unsolicited_na", 7107 .data = &ipv6_devconf.drop_unsolicited_na, 7108 .maxlen = sizeof(int), 7109 .mode = 0644, 7110 .proc_handler = proc_dointvec, 7111 }, 7112 { 7113 .procname = "keep_addr_on_down", 7114 .data = &ipv6_devconf.keep_addr_on_down, 7115 .maxlen = sizeof(int), 7116 .mode = 0644, 7117 .proc_handler = proc_dointvec, 7118 7119 }, 7120 { 7121 .procname = "seg6_enabled", 7122 .data = &ipv6_devconf.seg6_enabled, 7123 .maxlen = sizeof(int), 7124 .mode = 0644, 7125 .proc_handler = proc_dointvec, 7126 }, 7127 #ifdef CONFIG_IPV6_SEG6_HMAC 7128 { 7129 .procname = "seg6_require_hmac", 7130 .data = &ipv6_devconf.seg6_require_hmac, 7131 .maxlen = sizeof(int), 7132 .mode = 0644, 7133 .proc_handler = proc_dointvec, 7134 }, 7135 #endif 7136 { 7137 .procname = "enhanced_dad", 7138 .data = &ipv6_devconf.enhanced_dad, 7139 .maxlen = sizeof(int), 7140 .mode = 0644, 7141 .proc_handler = proc_dointvec, 7142 }, 7143 { 7144 .procname = "addr_gen_mode", 7145 .data = &ipv6_devconf.addr_gen_mode, 7146 .maxlen = sizeof(int), 7147 .mode = 0644, 7148 .proc_handler = addrconf_sysctl_addr_gen_mode, 7149 }, 7150 { 7151 .procname = "disable_policy", 7152 .data = &ipv6_devconf.disable_policy, 7153 .maxlen = sizeof(int), 7154 .mode = 0644, 7155 .proc_handler = addrconf_sysctl_disable_policy, 7156 }, 7157 { 7158 .procname = "ndisc_tclass", 7159 .data = &ipv6_devconf.ndisc_tclass, 7160 .maxlen = sizeof(int), 7161 .mode = 0644, 7162 .proc_handler = proc_dointvec_minmax, 7163 .extra1 = (void *)SYSCTL_ZERO, 7164 .extra2 = (void *)&two_five_five, 7165 }, 7166 { 7167 .procname = "rpl_seg_enabled", 7168 .data = &ipv6_devconf.rpl_seg_enabled, 7169 .maxlen = sizeof(int), 7170 .mode = 0644, 7171 .proc_handler = proc_dointvec, 7172 }, 7173 { 7174 .procname = "ioam6_enabled", 7175 .data = &ipv6_devconf.ioam6_enabled, 7176 .maxlen = sizeof(u8), 7177 .mode = 0644, 7178 .proc_handler = proc_dou8vec_minmax, 7179 .extra1 = (void *)SYSCTL_ZERO, 7180 .extra2 = (void *)SYSCTL_ONE, 7181 }, 7182 { 7183 .procname = "ioam6_id", 7184 .data = &ipv6_devconf.ioam6_id, 7185 .maxlen = sizeof(u32), 7186 .mode = 0644, 7187 .proc_handler = proc_douintvec_minmax, 7188 .extra1 = (void *)SYSCTL_ZERO, 7189 .extra2 = (void *)&ioam6_if_id_max, 7190 }, 7191 { 7192 .procname = "ioam6_id_wide", 7193 .data = &ipv6_devconf.ioam6_id_wide, 7194 .maxlen = sizeof(u32), 7195 .mode = 0644, 7196 .proc_handler = proc_douintvec, 7197 }, 7198 { 7199 .procname = "ndisc_evict_nocarrier", 7200 .data = &ipv6_devconf.ndisc_evict_nocarrier, 7201 .maxlen = sizeof(u8), 7202 .mode = 0644, 7203 .proc_handler = proc_dou8vec_minmax, 7204 .extra1 = (void *)SYSCTL_ZERO, 7205 .extra2 = (void *)SYSCTL_ONE, 7206 }, 7207 { 7208 .procname = "accept_untracked_na", 7209 .data = &ipv6_devconf.accept_untracked_na, 7210 .maxlen = sizeof(int), 7211 .mode = 0644, 7212 .proc_handler = proc_dointvec_minmax, 7213 .extra1 = SYSCTL_ZERO, 7214 .extra2 = SYSCTL_TWO, 7215 }, 7216 }; 7217 7218 static int __addrconf_sysctl_register(struct net *net, char *dev_name, 7219 struct inet6_dev *idev, struct ipv6_devconf *p) 7220 { 7221 size_t table_size = ARRAY_SIZE(addrconf_sysctl); 7222 int i, ifindex; 7223 struct ctl_table *table; 7224 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ]; 7225 7226 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT); 7227 if (!table) 7228 goto out; 7229 7230 for (i = 0; i < table_size; i++) { 7231 table[i].data += (char *)p - (char *)&ipv6_devconf; 7232 /* If one of these is already set, then it is not safe to 7233 * overwrite either of them: this makes proc_dointvec_minmax 7234 * usable. 7235 */ 7236 if (!table[i].extra1 && !table[i].extra2) { 7237 table[i].extra1 = idev; /* embedded; no ref */ 7238 table[i].extra2 = net; 7239 } 7240 } 7241 7242 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name); 7243 7244 p->sysctl_header = register_net_sysctl_sz(net, path, table, 7245 table_size); 7246 if (!p->sysctl_header) 7247 goto free; 7248 7249 if (!strcmp(dev_name, "all")) 7250 ifindex = NETCONFA_IFINDEX_ALL; 7251 else if (!strcmp(dev_name, "default")) 7252 ifindex = NETCONFA_IFINDEX_DEFAULT; 7253 else 7254 ifindex = idev->dev->ifindex; 7255 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, 7256 ifindex, p); 7257 return 0; 7258 7259 free: 7260 kfree(table); 7261 out: 7262 return -ENOBUFS; 7263 } 7264 7265 static void __addrconf_sysctl_unregister(struct net *net, 7266 struct ipv6_devconf *p, int ifindex) 7267 { 7268 const struct ctl_table *table; 7269 7270 if (!p->sysctl_header) 7271 return; 7272 7273 table = p->sysctl_header->ctl_table_arg; 7274 unregister_net_sysctl_table(p->sysctl_header); 7275 p->sysctl_header = NULL; 7276 kfree(table); 7277 7278 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL); 7279 } 7280 7281 static int addrconf_sysctl_register(struct inet6_dev *idev) 7282 { 7283 int err; 7284 7285 if (!sysctl_dev_name_is_allowed(idev->dev->name)) 7286 return -EINVAL; 7287 7288 err = neigh_sysctl_register(idev->dev, idev->nd_parms, 7289 &ndisc_ifinfo_sysctl_change); 7290 if (err) 7291 return err; 7292 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name, 7293 idev, &idev->cnf); 7294 if (err) 7295 neigh_sysctl_unregister(idev->nd_parms); 7296 7297 return err; 7298 } 7299 7300 static void addrconf_sysctl_unregister(struct inet6_dev *idev) 7301 { 7302 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf, 7303 idev->dev->ifindex); 7304 neigh_sysctl_unregister(idev->nd_parms); 7305 } 7306 7307 7308 #endif 7309 7310 static int __net_init addrconf_init_net(struct net *net) 7311 { 7312 int err = -ENOMEM; 7313 struct ipv6_devconf *all, *dflt; 7314 7315 spin_lock_init(&net->ipv6.addrconf_hash_lock); 7316 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work); 7317 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE, 7318 sizeof(struct hlist_head), 7319 GFP_KERNEL); 7320 if (!net->ipv6.inet6_addr_lst) 7321 goto err_alloc_addr; 7322 7323 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL); 7324 if (!all) 7325 goto err_alloc_all; 7326 7327 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL); 7328 if (!dflt) 7329 goto err_alloc_dflt; 7330 7331 if (!net_eq(net, &init_net)) { 7332 switch (net_inherit_devconf()) { 7333 case 1: /* copy from init_net */ 7334 memcpy(all, init_net.ipv6.devconf_all, 7335 sizeof(ipv6_devconf)); 7336 memcpy(dflt, init_net.ipv6.devconf_dflt, 7337 sizeof(ipv6_devconf_dflt)); 7338 break; 7339 case 3: /* copy from the current netns */ 7340 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all, 7341 sizeof(ipv6_devconf)); 7342 memcpy(dflt, 7343 current->nsproxy->net_ns->ipv6.devconf_dflt, 7344 sizeof(ipv6_devconf_dflt)); 7345 break; 7346 case 0: 7347 case 2: 7348 /* use compiled values */ 7349 break; 7350 } 7351 } 7352 7353 /* these will be inherited by all namespaces */ 7354 dflt->autoconf = ipv6_defaults.autoconf; 7355 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6; 7356 7357 dflt->stable_secret.initialized = false; 7358 all->stable_secret.initialized = false; 7359 7360 net->ipv6.devconf_all = all; 7361 net->ipv6.devconf_dflt = dflt; 7362 7363 #ifdef CONFIG_SYSCTL 7364 err = __addrconf_sysctl_register(net, "all", NULL, all); 7365 if (err < 0) 7366 goto err_reg_all; 7367 7368 err = __addrconf_sysctl_register(net, "default", NULL, dflt); 7369 if (err < 0) 7370 goto err_reg_dflt; 7371 #endif 7372 return 0; 7373 7374 #ifdef CONFIG_SYSCTL 7375 err_reg_dflt: 7376 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL); 7377 err_reg_all: 7378 kfree(dflt); 7379 net->ipv6.devconf_dflt = NULL; 7380 #endif 7381 err_alloc_dflt: 7382 kfree(all); 7383 net->ipv6.devconf_all = NULL; 7384 err_alloc_all: 7385 kfree(net->ipv6.inet6_addr_lst); 7386 err_alloc_addr: 7387 return err; 7388 } 7389 7390 static void __net_exit addrconf_exit_net(struct net *net) 7391 { 7392 int i; 7393 7394 #ifdef CONFIG_SYSCTL 7395 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt, 7396 NETCONFA_IFINDEX_DEFAULT); 7397 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all, 7398 NETCONFA_IFINDEX_ALL); 7399 #endif 7400 kfree(net->ipv6.devconf_dflt); 7401 net->ipv6.devconf_dflt = NULL; 7402 kfree(net->ipv6.devconf_all); 7403 net->ipv6.devconf_all = NULL; 7404 7405 cancel_delayed_work_sync(&net->ipv6.addr_chk_work); 7406 /* 7407 * Check hash table, then free it. 7408 */ 7409 for (i = 0; i < IN6_ADDR_HSIZE; i++) 7410 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i])); 7411 7412 kfree(net->ipv6.inet6_addr_lst); 7413 net->ipv6.inet6_addr_lst = NULL; 7414 } 7415 7416 static struct pernet_operations addrconf_ops = { 7417 .init = addrconf_init_net, 7418 .exit = addrconf_exit_net, 7419 }; 7420 7421 static struct rtnl_af_ops inet6_ops __read_mostly = { 7422 .family = AF_INET6, 7423 .fill_link_af = inet6_fill_link_af, 7424 .get_link_af_size = inet6_get_link_af_size, 7425 .validate_link_af = inet6_validate_link_af, 7426 .set_link_af = inet6_set_link_af, 7427 }; 7428 7429 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = { 7430 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK, 7431 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7432 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR, 7433 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET}, 7434 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR, 7435 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET}, 7436 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR, 7437 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr, 7438 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 7439 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST, 7440 .dumpit = inet6_dump_ifmcaddr, 7441 .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7442 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST, 7443 .dumpit = inet6_dump_ifacaddr, 7444 .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7445 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF, 7446 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf, 7447 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 7448 }; 7449 7450 /* 7451 * Init / cleanup code 7452 */ 7453 7454 int __init addrconf_init(void) 7455 { 7456 struct inet6_dev *idev; 7457 int err; 7458 7459 err = ipv6_addr_label_init(); 7460 if (err < 0) { 7461 pr_crit("%s: cannot initialize default policy table: %d\n", 7462 __func__, err); 7463 goto out; 7464 } 7465 7466 err = register_pernet_subsys(&addrconf_ops); 7467 if (err < 0) 7468 goto out_addrlabel; 7469 7470 /* All works using addrconf_wq need to lock rtnl. */ 7471 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf"); 7472 if (!addrconf_wq) { 7473 err = -ENOMEM; 7474 goto out_nowq; 7475 } 7476 7477 rtnl_net_lock(&init_net); 7478 idev = ipv6_add_dev(blackhole_netdev); 7479 rtnl_net_unlock(&init_net); 7480 if (IS_ERR(idev)) { 7481 err = PTR_ERR(idev); 7482 goto errlo; 7483 } 7484 7485 ip6_route_init_special_entries(); 7486 7487 register_netdevice_notifier(&ipv6_dev_notf); 7488 7489 addrconf_verify(&init_net); 7490 7491 err = rtnl_af_register(&inet6_ops); 7492 if (err) 7493 goto erraf; 7494 7495 err = rtnl_register_many(addrconf_rtnl_msg_handlers); 7496 if (err) 7497 goto errout; 7498 7499 err = ipv6_addr_label_rtnl_register(); 7500 if (err < 0) 7501 goto errout; 7502 7503 return 0; 7504 errout: 7505 rtnl_unregister_all(PF_INET6); 7506 rtnl_af_unregister(&inet6_ops); 7507 erraf: 7508 unregister_netdevice_notifier(&ipv6_dev_notf); 7509 errlo: 7510 destroy_workqueue(addrconf_wq); 7511 out_nowq: 7512 unregister_pernet_subsys(&addrconf_ops); 7513 out_addrlabel: 7514 ipv6_addr_label_cleanup(); 7515 out: 7516 return err; 7517 } 7518 7519 void addrconf_cleanup(void) 7520 { 7521 struct net_device *dev; 7522 7523 unregister_netdevice_notifier(&ipv6_dev_notf); 7524 unregister_pernet_subsys(&addrconf_ops); 7525 ipv6_addr_label_cleanup(); 7526 7527 rtnl_af_unregister(&inet6_ops); 7528 7529 rtnl_net_lock(&init_net); 7530 7531 /* clean dev list */ 7532 for_each_netdev(&init_net, dev) { 7533 if (!__in6_dev_get_rtnl_net(dev)) 7534 continue; 7535 addrconf_ifdown(dev, true); 7536 } 7537 addrconf_ifdown(init_net.loopback_dev, true); 7538 7539 rtnl_net_unlock(&init_net); 7540 7541 destroy_workqueue(addrconf_wq); 7542 } 7543