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 <crypto/sha1.h> 40 #include <linux/errno.h> 41 #include <linux/types.h> 42 #include <linux/kernel.h> 43 #include <linux/sched/signal.h> 44 #include <linux/socket.h> 45 #include <linux/sockios.h> 46 #include <linux/net.h> 47 #include <linux/inet.h> 48 #include <linux/in6.h> 49 #include <linux/netdevice.h> 50 #include <linux/if_addr.h> 51 #include <linux/if_arp.h> 52 #include <linux/if_arcnet.h> 53 #include <linux/if_infiniband.h> 54 #include <linux/route.h> 55 #include <linux/inetdevice.h> 56 #include <linux/init.h> 57 #include <linux/slab.h> 58 #ifdef CONFIG_SYSCTL 59 #include <linux/sysctl.h> 60 #endif 61 #include <linux/capability.h> 62 #include <linux/delay.h> 63 #include <linux/notifier.h> 64 #include <linux/string.h> 65 #include <linux/hash.h> 66 67 #include <net/ip_tunnels.h> 68 #include <net/net_namespace.h> 69 #include <net/sock.h> 70 #include <net/snmp.h> 71 72 #include <net/6lowpan.h> 73 #include <net/firewire.h> 74 #include <net/ipv6.h> 75 #include <net/protocol.h> 76 #include <net/ndisc.h> 77 #include <net/ip6_route.h> 78 #include <net/addrconf.h> 79 #include <net/tcp.h> 80 #include <net/ip.h> 81 #include <net/netlink.h> 82 #include <net/pkt_sched.h> 83 #include <net/l3mdev.h> 84 #include <net/netdev_lock.h> 85 #include <linux/if_tunnel.h> 86 #include <linux/rtnetlink.h> 87 #include <linux/netconf.h> 88 #include <linux/random.h> 89 #include <linux/uaccess.h> 90 #include <linux/unaligned.h> 91 92 #include <linux/proc_fs.h> 93 #include <linux/seq_file.h> 94 #include <linux/export.h> 95 #include <linux/ioam6.h> 96 97 #define IPV6_MAX_STRLEN \ 98 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255") 99 100 static inline u32 cstamp_delta(unsigned long cstamp) 101 { 102 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ; 103 } 104 105 static inline s32 rfc3315_s14_backoff_init(s32 irt) 106 { 107 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */ 108 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt; 109 do_div(tmp, 1000000); 110 return (s32)tmp; 111 } 112 113 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt) 114 { 115 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */ 116 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt; 117 do_div(tmp, 1000000); 118 if ((s32)tmp > mrt) { 119 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */ 120 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt; 121 do_div(tmp, 1000000); 122 } 123 return (s32)tmp; 124 } 125 126 #ifdef CONFIG_SYSCTL 127 static int addrconf_sysctl_register(struct inet6_dev *idev); 128 static void addrconf_sysctl_unregister(struct inet6_dev *idev); 129 #else 130 static inline int addrconf_sysctl_register(struct inet6_dev *idev) 131 { 132 return 0; 133 } 134 135 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev) 136 { 137 } 138 #endif 139 140 static void ipv6_gen_rnd_iid(struct in6_addr *addr); 141 142 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev); 143 static int ipv6_count_addresses(const struct inet6_dev *idev); 144 static int ipv6_generate_stable_address(struct in6_addr *addr, 145 u8 dad_count, 146 const struct inet6_dev *idev); 147 148 #define IN6_ADDR_HSIZE_SHIFT 8 149 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT) 150 151 static void addrconf_verify(struct net *net); 152 static void addrconf_verify_rtnl(struct net *net); 153 154 static struct workqueue_struct *addrconf_wq; 155 156 static void addrconf_join_anycast(struct inet6_ifaddr *ifp); 157 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp); 158 159 static void addrconf_type_change(struct net_device *dev, 160 unsigned long event); 161 static int addrconf_ifdown(struct net_device *dev, bool unregister); 162 163 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 164 int plen, 165 const struct net_device *dev, 166 u32 flags, u32 noflags, 167 bool no_gw); 168 169 static void addrconf_dad_start(struct inet6_ifaddr *ifp); 170 static void addrconf_dad_work(struct work_struct *w); 171 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 172 bool send_na); 173 static void addrconf_dad_run(struct inet6_dev *idev, bool restart); 174 static void addrconf_rs_timer(struct timer_list *t); 175 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 176 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 177 178 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 179 struct prefix_info *pinfo); 180 181 static struct ipv6_devconf ipv6_devconf __read_mostly = { 182 .forwarding = 0, 183 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 184 .mtu6 = IPV6_MIN_MTU, 185 .accept_ra = 1, 186 .accept_redirects = 1, 187 .autoconf = 1, 188 .force_mld_version = 0, 189 .mldv1_unsolicited_report_interval = 10 * HZ, 190 .mldv2_unsolicited_report_interval = HZ, 191 .dad_transmits = 1, 192 .rtr_solicits = MAX_RTR_SOLICITATIONS, 193 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 194 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 195 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 196 .use_tempaddr = 0, 197 .temp_valid_lft = TEMP_VALID_LIFETIME, 198 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 199 .regen_min_advance = REGEN_MIN_ADVANCE, 200 .regen_max_retry = REGEN_MAX_RETRY, 201 .max_desync_factor = MAX_DESYNC_FACTOR, 202 .max_addresses = IPV6_MAX_ADDRESSES, 203 .accept_ra_defrtr = 1, 204 .ra_defrtr_metric = IP6_RT_PRIO_USER, 205 .accept_ra_from_local = 0, 206 .accept_ra_min_hop_limit= 1, 207 .accept_ra_min_lft = 0, 208 .accept_ra_pinfo = 1, 209 #ifdef CONFIG_IPV6_ROUTER_PREF 210 .accept_ra_rtr_pref = 1, 211 .rtr_probe_interval = 60 * HZ, 212 #ifdef CONFIG_IPV6_ROUTE_INFO 213 .accept_ra_rt_info_min_plen = 0, 214 .accept_ra_rt_info_max_plen = 0, 215 #endif 216 #endif 217 .proxy_ndp = 0, 218 .accept_source_route = 0, /* we do not accept RH0 by default. */ 219 .disable_ipv6 = 0, 220 .accept_dad = 0, 221 .suppress_frag_ndisc = 1, 222 .accept_ra_mtu = 1, 223 .stable_secret = { 224 .initialized = false, 225 }, 226 .use_oif_addrs_only = 0, 227 .ignore_routes_with_linkdown = 0, 228 .keep_addr_on_down = 0, 229 .seg6_enabled = 0, 230 #ifdef CONFIG_IPV6_SEG6_HMAC 231 .seg6_require_hmac = 0, 232 #endif 233 .enhanced_dad = 1, 234 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 235 .disable_policy = 0, 236 .rpl_seg_enabled = 0, 237 .ioam6_enabled = 0, 238 .ioam6_id = IOAM6_DEFAULT_IF_ID, 239 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 240 .ndisc_evict_nocarrier = 1, 241 .ra_honor_pio_life = 0, 242 .ra_honor_pio_pflag = 0, 243 .force_forwarding = 0, 244 }; 245 246 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = { 247 .forwarding = 0, 248 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 249 .mtu6 = IPV6_MIN_MTU, 250 .accept_ra = 1, 251 .accept_redirects = 1, 252 .autoconf = 1, 253 .force_mld_version = 0, 254 .mldv1_unsolicited_report_interval = 10 * HZ, 255 .mldv2_unsolicited_report_interval = HZ, 256 .dad_transmits = 1, 257 .rtr_solicits = MAX_RTR_SOLICITATIONS, 258 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 259 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 260 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 261 .use_tempaddr = 0, 262 .temp_valid_lft = TEMP_VALID_LIFETIME, 263 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 264 .regen_min_advance = REGEN_MIN_ADVANCE, 265 .regen_max_retry = REGEN_MAX_RETRY, 266 .max_desync_factor = MAX_DESYNC_FACTOR, 267 .max_addresses = IPV6_MAX_ADDRESSES, 268 .accept_ra_defrtr = 1, 269 .ra_defrtr_metric = IP6_RT_PRIO_USER, 270 .accept_ra_from_local = 0, 271 .accept_ra_min_hop_limit= 1, 272 .accept_ra_min_lft = 0, 273 .accept_ra_pinfo = 1, 274 #ifdef CONFIG_IPV6_ROUTER_PREF 275 .accept_ra_rtr_pref = 1, 276 .rtr_probe_interval = 60 * HZ, 277 #ifdef CONFIG_IPV6_ROUTE_INFO 278 .accept_ra_rt_info_min_plen = 0, 279 .accept_ra_rt_info_max_plen = 0, 280 #endif 281 #endif 282 .proxy_ndp = 0, 283 .accept_source_route = 0, /* we do not accept RH0 by default. */ 284 .disable_ipv6 = 0, 285 .accept_dad = 1, 286 .suppress_frag_ndisc = 1, 287 .accept_ra_mtu = 1, 288 .stable_secret = { 289 .initialized = false, 290 }, 291 .use_oif_addrs_only = 0, 292 .ignore_routes_with_linkdown = 0, 293 .keep_addr_on_down = 0, 294 .seg6_enabled = 0, 295 #ifdef CONFIG_IPV6_SEG6_HMAC 296 .seg6_require_hmac = 0, 297 #endif 298 .enhanced_dad = 1, 299 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 300 .disable_policy = 0, 301 .rpl_seg_enabled = 0, 302 .ioam6_enabled = 0, 303 .ioam6_id = IOAM6_DEFAULT_IF_ID, 304 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 305 .ndisc_evict_nocarrier = 1, 306 .ra_honor_pio_life = 0, 307 .ra_honor_pio_pflag = 0, 308 .force_forwarding = 0, 309 }; 310 311 /* Check if link is ready: is it up and is a valid qdisc available */ 312 static inline bool addrconf_link_ready(const struct net_device *dev) 313 { 314 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev); 315 } 316 317 static void addrconf_del_rs_timer(struct inet6_dev *idev) 318 { 319 if (timer_delete(&idev->rs_timer)) 320 __in6_dev_put(idev); 321 } 322 323 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp) 324 { 325 if (cancel_delayed_work(&ifp->dad_work)) 326 __in6_ifa_put(ifp); 327 } 328 329 static void addrconf_mod_rs_timer(struct inet6_dev *idev, 330 unsigned long when) 331 { 332 if (!mod_timer(&idev->rs_timer, jiffies + when)) 333 in6_dev_hold(idev); 334 } 335 336 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp, 337 unsigned long delay) 338 { 339 in6_ifa_hold(ifp); 340 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay)) 341 in6_ifa_put(ifp); 342 } 343 344 static int snmp6_alloc_dev(struct inet6_dev *idev) 345 { 346 int i; 347 348 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT); 349 if (!idev->stats.ipv6) 350 goto err_ip; 351 352 for_each_possible_cpu(i) { 353 struct ipstats_mib *addrconf_stats; 354 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i); 355 u64_stats_init(&addrconf_stats->syncp); 356 } 357 358 359 idev->stats.icmpv6dev = kzalloc_obj(struct icmpv6_mib_device); 360 if (!idev->stats.icmpv6dev) 361 goto err_icmp; 362 idev->stats.icmpv6msgdev = kzalloc_obj(struct icmpv6msg_mib_device, 363 GFP_KERNEL_ACCOUNT); 364 if (!idev->stats.icmpv6msgdev) 365 goto err_icmpmsg; 366 367 return 0; 368 369 err_icmpmsg: 370 kfree(idev->stats.icmpv6dev); 371 err_icmp: 372 free_percpu(idev->stats.ipv6); 373 err_ip: 374 return -ENOMEM; 375 } 376 377 static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 378 { 379 struct net *net = dev_net(dev); 380 struct neigh_table *tbl; 381 struct inet6_dev *ndev; 382 int err = -ENOMEM; 383 384 ASSERT_RTNL(); 385 netdev_assert_locked_ops_compat(dev); 386 387 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev) 388 return ERR_PTR(-EINVAL); 389 390 ndev = kzalloc_obj(*ndev, GFP_KERNEL_ACCOUNT); 391 if (!ndev) 392 return ERR_PTR(err); 393 394 rwlock_init(&ndev->lock); 395 ndev->dev = dev; 396 INIT_LIST_HEAD(&ndev->addr_list); 397 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0); 398 memcpy(&ndev->cnf, net->ipv6.devconf_dflt, sizeof(ndev->cnf)); 399 400 if (ndev->cnf.stable_secret.initialized) 401 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY; 402 403 tbl = nd_table(net); 404 ndev->cnf.mtu6 = dev->mtu; 405 ndev->ra_mtu = 0; 406 ndev->nd_parms = neigh_parms_alloc(dev, tbl); 407 if (!ndev->nd_parms) { 408 kfree(ndev); 409 return ERR_PTR(err); 410 } 411 if (ndev->cnf.forwarding) 412 netif_disable_lro(dev); 413 /* We refer to the device */ 414 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL); 415 416 if (snmp6_alloc_dev(ndev) < 0) { 417 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n", 418 __func__); 419 neigh_parms_release(tbl, ndev->nd_parms); 420 netdev_put(dev, &ndev->dev_tracker); 421 kfree(ndev); 422 return ERR_PTR(err); 423 } 424 425 if (dev != blackhole_netdev) { 426 if (snmp6_register_dev(ndev) < 0) { 427 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n", 428 __func__, dev->name); 429 goto err_release; 430 } 431 } 432 /* One reference from device. */ 433 refcount_set(&ndev->refcnt, 1); 434 435 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK)) 436 ndev->cnf.accept_dad = -1; 437 438 #if IS_ENABLED(CONFIG_IPV6_SIT) 439 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) { 440 pr_info("%s: Disabled Multicast RS\n", dev->name); 441 ndev->cnf.rtr_solicits = 0; 442 } 443 #endif 444 445 INIT_LIST_HEAD(&ndev->tempaddr_list); 446 ndev->desync_factor = U32_MAX; 447 if ((dev->flags&IFF_LOOPBACK) || 448 dev->type == ARPHRD_TUNNEL || 449 dev->type == ARPHRD_TUNNEL6 || 450 dev->type == ARPHRD_SIT || 451 dev->type == ARPHRD_NONE) { 452 ndev->cnf.use_tempaddr = -1; 453 } 454 455 ndev->token = in6addr_any; 456 457 if (netif_running(dev) && addrconf_link_ready(dev)) 458 ndev->if_flags |= IF_READY; 459 460 ipv6_mc_init_dev(ndev); 461 ndev->tstamp = jiffies; 462 if (dev != blackhole_netdev) { 463 err = addrconf_sysctl_register(ndev); 464 if (err) { 465 ipv6_mc_destroy_dev(ndev); 466 snmp6_unregister_dev(ndev); 467 goto err_release; 468 } 469 } 470 /* protected by rtnl_lock */ 471 rcu_assign_pointer(dev->ip6_ptr, ndev); 472 473 if (dev != blackhole_netdev) { 474 /* Join interface-local all-node multicast group */ 475 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes); 476 477 /* Join all-node multicast group */ 478 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes); 479 480 /* Join all-router multicast group if forwarding is set */ 481 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST)) 482 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 483 } 484 return ndev; 485 486 err_release: 487 neigh_parms_release(tbl, ndev->nd_parms); 488 ndev->dead = 1; 489 in6_dev_finish_destroy(ndev); 490 return ERR_PTR(err); 491 } 492 493 static struct inet6_dev *ipv6_find_idev(struct net_device *dev) 494 { 495 struct inet6_dev *idev; 496 497 ASSERT_RTNL(); 498 499 idev = __in6_dev_get(dev); 500 if (!idev) { 501 idev = ipv6_add_dev(dev); 502 if (IS_ERR(idev)) 503 return idev; 504 } 505 506 if (dev->flags&IFF_UP) 507 ipv6_mc_up(idev); 508 return idev; 509 } 510 511 static int inet6_netconf_msgsize_devconf(int type) 512 { 513 int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) 514 + nla_total_size(4); /* NETCONFA_IFINDEX */ 515 bool all = false; 516 517 if (type == NETCONFA_ALL) 518 all = true; 519 520 if (all || type == NETCONFA_FORWARDING) 521 size += nla_total_size(4); 522 #ifdef CONFIG_IPV6_MROUTE 523 if (all || type == NETCONFA_MC_FORWARDING) 524 size += nla_total_size(4); 525 #endif 526 if (all || type == NETCONFA_PROXY_NEIGH) 527 size += nla_total_size(4); 528 529 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) 530 size += nla_total_size(4); 531 532 return size; 533 } 534 535 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex, 536 struct ipv6_devconf *devconf, u32 portid, 537 u32 seq, int event, unsigned int flags, 538 int type) 539 { 540 struct nlmsghdr *nlh; 541 struct netconfmsg *ncm; 542 bool all = false; 543 544 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), 545 flags); 546 if (!nlh) 547 return -EMSGSIZE; 548 549 if (type == NETCONFA_ALL) 550 all = true; 551 552 ncm = nlmsg_data(nlh); 553 ncm->ncm_family = AF_INET6; 554 555 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0) 556 goto nla_put_failure; 557 558 if (!devconf) 559 goto out; 560 561 if ((all || type == NETCONFA_FORWARDING) && 562 nla_put_s32(skb, NETCONFA_FORWARDING, 563 READ_ONCE(devconf->forwarding)) < 0) 564 goto nla_put_failure; 565 #ifdef CONFIG_IPV6_MROUTE 566 if ((all || type == NETCONFA_MC_FORWARDING) && 567 nla_put_s32(skb, NETCONFA_MC_FORWARDING, 568 atomic_read(&devconf->mc_forwarding)) < 0) 569 goto nla_put_failure; 570 #endif 571 if ((all || type == NETCONFA_PROXY_NEIGH) && 572 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, 573 READ_ONCE(devconf->proxy_ndp)) < 0) 574 goto nla_put_failure; 575 576 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) && 577 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 578 READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0) 579 goto nla_put_failure; 580 581 out: 582 nlmsg_end(skb, nlh); 583 return 0; 584 585 nla_put_failure: 586 nlmsg_cancel(skb, nlh); 587 return -EMSGSIZE; 588 } 589 590 void inet6_netconf_notify_devconf(struct net *net, int event, int type, 591 int ifindex, struct ipv6_devconf *devconf) 592 { 593 struct sk_buff *skb; 594 int err = -ENOBUFS; 595 596 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL); 597 if (!skb) 598 goto errout; 599 600 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0, 601 event, 0, type); 602 if (err < 0) { 603 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 604 WARN_ON(err == -EMSGSIZE); 605 kfree_skb(skb); 606 goto errout; 607 } 608 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL); 609 return; 610 errout: 611 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err); 612 } 613 614 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = { 615 [NETCONFA_IFINDEX] = { .len = sizeof(int) }, 616 [NETCONFA_FORWARDING] = { .len = sizeof(int) }, 617 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) }, 618 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) }, 619 }; 620 621 static int inet6_netconf_valid_get_req(struct sk_buff *skb, 622 const struct nlmsghdr *nlh, 623 struct nlattr **tb, 624 struct netlink_ext_ack *extack) 625 { 626 int i, err; 627 628 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) { 629 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request"); 630 return -EINVAL; 631 } 632 633 if (!netlink_strict_get_check(skb)) 634 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg), 635 tb, NETCONFA_MAX, 636 devconf_ipv6_policy, extack); 637 638 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg), 639 tb, NETCONFA_MAX, 640 devconf_ipv6_policy, extack); 641 if (err) 642 return err; 643 644 for (i = 0; i <= NETCONFA_MAX; i++) { 645 if (!tb[i]) 646 continue; 647 648 switch (i) { 649 case NETCONFA_IFINDEX: 650 break; 651 default: 652 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request"); 653 return -EINVAL; 654 } 655 } 656 657 return 0; 658 } 659 660 static int inet6_netconf_get_devconf(struct sk_buff *in_skb, 661 struct nlmsghdr *nlh, 662 struct netlink_ext_ack *extack) 663 { 664 struct net *net = sock_net(in_skb->sk); 665 struct nlattr *tb[NETCONFA_MAX+1]; 666 struct inet6_dev *in6_dev = NULL; 667 struct net_device *dev = NULL; 668 struct sk_buff *skb; 669 struct ipv6_devconf *devconf; 670 int ifindex; 671 int err; 672 673 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack); 674 if (err < 0) 675 return err; 676 677 if (!tb[NETCONFA_IFINDEX]) 678 return -EINVAL; 679 680 err = -EINVAL; 681 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); 682 switch (ifindex) { 683 case NETCONFA_IFINDEX_ALL: 684 devconf = net->ipv6.devconf_all; 685 break; 686 case NETCONFA_IFINDEX_DEFAULT: 687 devconf = net->ipv6.devconf_dflt; 688 break; 689 default: 690 dev = dev_get_by_index(net, ifindex); 691 if (!dev) 692 return -EINVAL; 693 in6_dev = in6_dev_get(dev); 694 if (!in6_dev) 695 goto errout; 696 devconf = &in6_dev->cnf; 697 break; 698 } 699 700 err = -ENOBUFS; 701 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); 702 if (!skb) 703 goto errout; 704 705 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 706 NETLINK_CB(in_skb).portid, 707 nlh->nlmsg_seq, RTM_NEWNETCONF, 0, 708 NETCONFA_ALL); 709 if (err < 0) { 710 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 711 WARN_ON(err == -EMSGSIZE); 712 kfree_skb(skb); 713 goto errout; 714 } 715 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 716 errout: 717 if (in6_dev) 718 in6_dev_put(in6_dev); 719 dev_put(dev); 720 return err; 721 } 722 723 /* Combine dev_addr_genid and dev_base_seq to detect changes. 724 */ 725 static u32 inet6_base_seq(const struct net *net) 726 { 727 u32 res = atomic_read(&net->ipv6.dev_addr_genid) + 728 READ_ONCE(net->dev_base_seq); 729 730 /* Must not return 0 (see nl_dump_check_consistent()). 731 * Chose a value far away from 0. 732 */ 733 if (!res) 734 res = 0x80000000; 735 return res; 736 } 737 738 static int inet6_netconf_dump_devconf(struct sk_buff *skb, 739 struct netlink_callback *cb) 740 { 741 const struct nlmsghdr *nlh = cb->nlh; 742 struct net *net = sock_net(skb->sk); 743 struct { 744 unsigned long ifindex; 745 unsigned int all_default; 746 } *ctx = (void *)cb->ctx; 747 struct net_device *dev; 748 struct inet6_dev *idev; 749 int err = 0; 750 751 if (cb->strict_check) { 752 struct netlink_ext_ack *extack = cb->extack; 753 struct netconfmsg *ncm; 754 755 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) { 756 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request"); 757 return -EINVAL; 758 } 759 760 if (nlmsg_attrlen(nlh, sizeof(*ncm))) { 761 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request"); 762 return -EINVAL; 763 } 764 } 765 766 rcu_read_lock(); 767 for_each_netdev_dump(net, dev, ctx->ifindex) { 768 idev = __in6_dev_get(dev); 769 if (!idev) 770 continue; 771 err = inet6_netconf_fill_devconf(skb, dev->ifindex, 772 &idev->cnf, 773 NETLINK_CB(cb->skb).portid, 774 nlh->nlmsg_seq, 775 RTM_NEWNETCONF, 776 NLM_F_MULTI, 777 NETCONFA_ALL); 778 if (err < 0) 779 goto done; 780 } 781 if (ctx->all_default == 0) { 782 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL, 783 net->ipv6.devconf_all, 784 NETLINK_CB(cb->skb).portid, 785 nlh->nlmsg_seq, 786 RTM_NEWNETCONF, NLM_F_MULTI, 787 NETCONFA_ALL); 788 if (err < 0) 789 goto done; 790 ctx->all_default++; 791 } 792 if (ctx->all_default == 1) { 793 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT, 794 net->ipv6.devconf_dflt, 795 NETLINK_CB(cb->skb).portid, 796 nlh->nlmsg_seq, 797 RTM_NEWNETCONF, NLM_F_MULTI, 798 NETCONFA_ALL); 799 if (err < 0) 800 goto done; 801 ctx->all_default++; 802 } 803 done: 804 rcu_read_unlock(); 805 return err; 806 } 807 808 #ifdef CONFIG_SYSCTL 809 static void dev_forward_change(struct inet6_dev *idev) 810 { 811 struct net_device *dev; 812 struct inet6_ifaddr *ifa; 813 LIST_HEAD(tmp_addr_list); 814 815 if (!idev) 816 return; 817 dev = idev->dev; 818 if (idev->cnf.forwarding) 819 dev_disable_lro(dev); 820 if (dev->flags & IFF_MULTICAST) { 821 if (idev->cnf.forwarding) { 822 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 823 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters); 824 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters); 825 } else { 826 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters); 827 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters); 828 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters); 829 } 830 } 831 832 read_lock_bh(&idev->lock); 833 list_for_each_entry(ifa, &idev->addr_list, if_list) { 834 if (ifa->flags&IFA_F_TENTATIVE) 835 continue; 836 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 837 } 838 read_unlock_bh(&idev->lock); 839 840 while (!list_empty(&tmp_addr_list)) { 841 ifa = list_first_entry(&tmp_addr_list, 842 struct inet6_ifaddr, if_list_aux); 843 list_del(&ifa->if_list_aux); 844 if (idev->cnf.forwarding) 845 addrconf_join_anycast(ifa); 846 else 847 addrconf_leave_anycast(ifa); 848 } 849 850 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 851 NETCONFA_FORWARDING, 852 dev->ifindex, &idev->cnf); 853 } 854 855 856 static void addrconf_forward_change(struct net *net, __s32 newf) 857 { 858 struct net_device *dev; 859 struct inet6_dev *idev; 860 861 for_each_netdev(net, dev) { 862 idev = __in6_dev_get_rtnl_net(dev); 863 if (idev) { 864 int changed = (!idev->cnf.forwarding) ^ (!newf); 865 /* Disabling all.forwarding sets 0 to force_forwarding for all interfaces */ 866 if (newf == 0) 867 WRITE_ONCE(idev->cnf.force_forwarding, 0); 868 869 WRITE_ONCE(idev->cnf.forwarding, newf); 870 if (changed) 871 dev_forward_change(idev); 872 } 873 } 874 } 875 876 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf) 877 { 878 struct net *net = (struct net *)table->extra2; 879 int old; 880 881 if (!rtnl_net_trylock(net)) 882 return restart_syscall(); 883 884 old = *p; 885 WRITE_ONCE(*p, newf); 886 887 if (p == &net->ipv6.devconf_dflt->forwarding) { 888 if ((!newf) ^ (!old)) 889 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 890 NETCONFA_FORWARDING, 891 NETCONFA_IFINDEX_DEFAULT, 892 net->ipv6.devconf_dflt); 893 rtnl_net_unlock(net); 894 return 0; 895 } 896 897 if (p == &net->ipv6.devconf_all->forwarding) { 898 int old_dflt = net->ipv6.devconf_dflt->forwarding; 899 900 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf); 901 if ((!newf) ^ (!old_dflt)) 902 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 903 NETCONFA_FORWARDING, 904 NETCONFA_IFINDEX_DEFAULT, 905 net->ipv6.devconf_dflt); 906 907 addrconf_forward_change(net, newf); 908 if ((!newf) ^ (!old)) 909 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 910 NETCONFA_FORWARDING, 911 NETCONFA_IFINDEX_ALL, 912 net->ipv6.devconf_all); 913 } else if ((!newf) ^ (!old)) 914 dev_forward_change((struct inet6_dev *)table->extra1); 915 rtnl_net_unlock(net); 916 917 if (newf) 918 rt6_purge_dflt_routers(net); 919 return 0; 920 } 921 922 static void addrconf_linkdown_change(struct net *net, __s32 newf) 923 { 924 struct net_device *dev; 925 struct inet6_dev *idev; 926 927 for_each_netdev(net, dev) { 928 idev = __in6_dev_get_rtnl_net(dev); 929 if (idev) { 930 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf); 931 932 WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf); 933 if (changed) 934 inet6_netconf_notify_devconf(dev_net(dev), 935 RTM_NEWNETCONF, 936 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 937 dev->ifindex, 938 &idev->cnf); 939 } 940 } 941 } 942 943 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf) 944 { 945 struct net *net = (struct net *)table->extra2; 946 int old; 947 948 if (!rtnl_net_trylock(net)) 949 return restart_syscall(); 950 951 old = *p; 952 WRITE_ONCE(*p, newf); 953 954 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) { 955 if ((!newf) ^ (!old)) 956 inet6_netconf_notify_devconf(net, 957 RTM_NEWNETCONF, 958 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 959 NETCONFA_IFINDEX_DEFAULT, 960 net->ipv6.devconf_dflt); 961 } else if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) { 962 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf); 963 addrconf_linkdown_change(net, newf); 964 if ((!newf) ^ (!old)) 965 inet6_netconf_notify_devconf(net, 966 RTM_NEWNETCONF, 967 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 968 NETCONFA_IFINDEX_ALL, 969 net->ipv6.devconf_all); 970 } else { 971 if (!newf ^ !old) { 972 struct inet6_dev *idev = table->extra1; 973 974 inet6_netconf_notify_devconf(net, 975 RTM_NEWNETCONF, 976 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 977 idev->dev->ifindex, 978 &idev->cnf); 979 } 980 } 981 982 rtnl_net_unlock(net); 983 984 return 0; 985 } 986 987 #endif 988 989 /* Nobody refers to this ifaddr, destroy it */ 990 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp) 991 { 992 WARN_ON(!hlist_unhashed(&ifp->addr_lst)); 993 994 #ifdef NET_REFCNT_DEBUG 995 pr_debug("%s\n", __func__); 996 #endif 997 998 in6_dev_put(ifp->idev); 999 1000 if (cancel_delayed_work(&ifp->dad_work)) 1001 pr_notice("delayed DAD work was pending while freeing ifa=%p\n", 1002 ifp); 1003 1004 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 1005 pr_warn("Freeing alive inet6 address %p\n", ifp); 1006 return; 1007 } 1008 1009 kfree_rcu(ifp, rcu); 1010 } 1011 1012 static void 1013 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp) 1014 { 1015 struct list_head *p; 1016 int ifp_scope = ipv6_addr_src_scope(&ifp->addr); 1017 1018 /* 1019 * Each device address list is sorted in order of scope - 1020 * global before linklocal. 1021 */ 1022 list_for_each(p, &idev->addr_list) { 1023 struct inet6_ifaddr *ifa 1024 = list_entry(p, struct inet6_ifaddr, if_list); 1025 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr)) 1026 break; 1027 } 1028 1029 list_add_tail_rcu(&ifp->if_list, p); 1030 } 1031 1032 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr) 1033 { 1034 u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net)); 1035 1036 return hash_32(val, IN6_ADDR_HSIZE_SHIFT); 1037 } 1038 1039 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr, 1040 struct net_device *dev, unsigned int hash) 1041 { 1042 struct inet6_ifaddr *ifp; 1043 1044 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 1045 if (ipv6_addr_equal(&ifp->addr, addr)) { 1046 if (!dev || ifp->idev->dev == dev) 1047 return true; 1048 } 1049 } 1050 return false; 1051 } 1052 1053 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa) 1054 { 1055 struct net *net = dev_net(dev); 1056 unsigned int hash = inet6_addr_hash(net, &ifa->addr); 1057 int err = 0; 1058 1059 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1060 1061 /* Ignore adding duplicate addresses on an interface */ 1062 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) { 1063 netdev_dbg(dev, "ipv6_add_addr: already assigned\n"); 1064 err = -EEXIST; 1065 } else { 1066 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]); 1067 } 1068 1069 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1070 1071 return err; 1072 } 1073 1074 /* On success it returns ifp with increased reference count */ 1075 1076 static struct inet6_ifaddr * 1077 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg, 1078 bool can_block, struct netlink_ext_ack *extack) 1079 { 1080 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC; 1081 int addr_type = ipv6_addr_type(cfg->pfx); 1082 struct net *net = dev_net(idev->dev); 1083 struct inet6_ifaddr *ifa = NULL; 1084 struct fib6_info *f6i = NULL; 1085 int err = 0; 1086 1087 if (addr_type == IPV6_ADDR_ANY) { 1088 NL_SET_ERR_MSG_MOD(extack, "Invalid address"); 1089 return ERR_PTR(-EADDRNOTAVAIL); 1090 } else if (addr_type & IPV6_ADDR_MULTICAST && 1091 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) { 1092 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag"); 1093 return ERR_PTR(-EADDRNOTAVAIL); 1094 } else if (!(idev->dev->flags & IFF_LOOPBACK) && 1095 !netif_is_l3_master(idev->dev) && 1096 addr_type & IPV6_ADDR_LOOPBACK) { 1097 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device"); 1098 return ERR_PTR(-EADDRNOTAVAIL); 1099 } 1100 1101 if (idev->dead) { 1102 NL_SET_ERR_MSG_MOD(extack, "device is going away"); 1103 err = -ENODEV; 1104 goto out; 1105 } 1106 1107 if (idev->cnf.disable_ipv6) { 1108 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 1109 err = -EACCES; 1110 goto out; 1111 } 1112 1113 /* validator notifier needs to be blocking; 1114 * do not call in atomic context 1115 */ 1116 if (can_block) { 1117 struct in6_validator_info i6vi = { 1118 .i6vi_addr = *cfg->pfx, 1119 .i6vi_dev = idev, 1120 .extack = extack, 1121 }; 1122 1123 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi); 1124 err = notifier_to_errno(err); 1125 if (err < 0) 1126 goto out; 1127 } 1128 1129 ifa = kzalloc_obj(*ifa, gfp_flags | __GFP_ACCOUNT); 1130 if (!ifa) { 1131 err = -ENOBUFS; 1132 goto out; 1133 } 1134 1135 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack); 1136 if (IS_ERR(f6i)) { 1137 err = PTR_ERR(f6i); 1138 f6i = NULL; 1139 goto out; 1140 } 1141 1142 neigh_parms_data_state_setall(idev->nd_parms); 1143 1144 ifa->addr = *cfg->pfx; 1145 if (cfg->peer_pfx) 1146 ifa->peer_addr = *cfg->peer_pfx; 1147 1148 spin_lock_init(&ifa->lock); 1149 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work); 1150 INIT_HLIST_NODE(&ifa->addr_lst); 1151 ifa->scope = cfg->scope; 1152 ifa->prefix_len = cfg->plen; 1153 ifa->rt_priority = cfg->rt_priority; 1154 ifa->flags = cfg->ifa_flags; 1155 ifa->ifa_proto = cfg->ifa_proto; 1156 /* No need to add the TENTATIVE flag for addresses with NODAD */ 1157 if (!(cfg->ifa_flags & IFA_F_NODAD)) 1158 ifa->flags |= IFA_F_TENTATIVE; 1159 ifa->valid_lft = cfg->valid_lft; 1160 ifa->prefered_lft = cfg->preferred_lft; 1161 ifa->cstamp = ifa->tstamp = jiffies; 1162 ifa->tokenized = false; 1163 1164 ifa->rt = f6i; 1165 1166 ifa->idev = idev; 1167 in6_dev_hold(idev); 1168 1169 /* For caller */ 1170 refcount_set(&ifa->refcnt, 1); 1171 1172 rcu_read_lock(); 1173 1174 err = ipv6_add_addr_hash(idev->dev, ifa); 1175 if (err < 0) { 1176 rcu_read_unlock(); 1177 goto out; 1178 } 1179 1180 write_lock_bh(&idev->lock); 1181 1182 /* Add to inet6_dev unicast addr list. */ 1183 ipv6_link_dev_addr(idev, ifa); 1184 1185 if (ifa->flags&IFA_F_TEMPORARY) { 1186 /* manage_tempaddrs() relies on addresses being added to the head */ 1187 list_add(&ifa->tmp_list, &idev->tempaddr_list); 1188 in6_ifa_hold(ifa); 1189 } 1190 1191 in6_ifa_hold(ifa); 1192 write_unlock_bh(&idev->lock); 1193 1194 rcu_read_unlock(); 1195 1196 inet6addr_notifier_call_chain(NETDEV_UP, ifa); 1197 out: 1198 if (unlikely(err < 0)) { 1199 fib6_info_release(f6i); 1200 1201 if (ifa) { 1202 if (ifa->idev) 1203 in6_dev_put(ifa->idev); 1204 kfree(ifa); 1205 } 1206 ifa = ERR_PTR(err); 1207 } 1208 1209 return ifa; 1210 } 1211 1212 enum cleanup_prefix_rt_t { 1213 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */ 1214 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */ 1215 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */ 1216 }; 1217 1218 /* 1219 * Check, whether the prefix for ifp would still need a prefix route 1220 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_* 1221 * constants. 1222 * 1223 * 1) we don't purge prefix if address was not permanent. 1224 * prefix is managed by its own lifetime. 1225 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE. 1226 * 3) if there are no addresses, delete prefix. 1227 * 4) if there are still other permanent address(es), 1228 * corresponding prefix is still permanent. 1229 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE, 1230 * don't purge the prefix, assume user space is managing it. 1231 * 6) otherwise, update prefix lifetime to the 1232 * longest valid lifetime among the corresponding 1233 * addresses on the device. 1234 * Note: subsequent RA will update lifetime. 1235 **/ 1236 static enum cleanup_prefix_rt_t 1237 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires) 1238 { 1239 struct inet6_ifaddr *ifa; 1240 struct inet6_dev *idev = ifp->idev; 1241 unsigned long lifetime; 1242 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL; 1243 1244 *expires = jiffies; 1245 1246 list_for_each_entry(ifa, &idev->addr_list, if_list) { 1247 if (ifa == ifp) 1248 continue; 1249 if (ifa->prefix_len != ifp->prefix_len || 1250 !ipv6_prefix_equal(&ifa->addr, &ifp->addr, 1251 ifp->prefix_len)) 1252 continue; 1253 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE)) 1254 return CLEANUP_PREFIX_RT_NOP; 1255 1256 action = CLEANUP_PREFIX_RT_EXPIRE; 1257 1258 spin_lock(&ifa->lock); 1259 1260 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ); 1261 /* 1262 * Note: Because this address is 1263 * not permanent, lifetime < 1264 * LONG_MAX / HZ here. 1265 */ 1266 if (time_before(*expires, ifa->tstamp + lifetime * HZ)) 1267 *expires = ifa->tstamp + lifetime * HZ; 1268 spin_unlock(&ifa->lock); 1269 } 1270 1271 return action; 1272 } 1273 1274 static void 1275 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, 1276 bool del_rt, bool del_peer) 1277 { 1278 struct net *net = dev_net(ifp->idev->dev); 1279 struct fib6_table *table; 1280 struct fib6_info *f6i; 1281 1282 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr, 1283 ifp->prefix_len, 1284 ifp->idev->dev, 0, RTF_DEFAULT, true); 1285 if (f6i) { 1286 if (del_rt) 1287 ip6_del_rt(net, f6i, false); 1288 else { 1289 if (f6i != net->ipv6.fib6_null_entry && 1290 !(f6i->fib6_flags & RTF_EXPIRES)) { 1291 table = f6i->fib6_table; 1292 spin_lock_bh(&table->tb6_lock); 1293 1294 fib6_set_expires(f6i, expires); 1295 fib6_add_gc_list(f6i); 1296 1297 spin_unlock_bh(&table->tb6_lock); 1298 } 1299 fib6_info_release(f6i); 1300 } 1301 } 1302 } 1303 1304 1305 /* This function wants to get referenced ifp and releases it before return */ 1306 1307 static void ipv6_del_addr(struct inet6_ifaddr *ifp) 1308 { 1309 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP; 1310 struct net *net = dev_net(ifp->idev->dev); 1311 unsigned long expires; 1312 int state; 1313 1314 ASSERT_RTNL(); 1315 1316 spin_lock_bh(&ifp->lock); 1317 state = ifp->state; 1318 ifp->state = INET6_IFADDR_STATE_DEAD; 1319 spin_unlock_bh(&ifp->lock); 1320 1321 if (state == INET6_IFADDR_STATE_DEAD) 1322 goto out; 1323 1324 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1325 hlist_del_init_rcu(&ifp->addr_lst); 1326 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1327 1328 write_lock_bh(&ifp->idev->lock); 1329 1330 if (ifp->flags&IFA_F_TEMPORARY) { 1331 list_del(&ifp->tmp_list); 1332 if (ifp->ifpub) { 1333 in6_ifa_put(ifp->ifpub); 1334 ifp->ifpub = NULL; 1335 } 1336 __in6_ifa_put(ifp); 1337 } 1338 1339 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) 1340 action = check_cleanup_prefix_route(ifp, &expires); 1341 1342 list_del_rcu(&ifp->if_list); 1343 __in6_ifa_put(ifp); 1344 1345 write_unlock_bh(&ifp->idev->lock); 1346 1347 addrconf_del_dad_work(ifp); 1348 1349 ipv6_ifa_notify(RTM_DELADDR, ifp); 1350 1351 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp); 1352 1353 if (action != CLEANUP_PREFIX_RT_NOP) { 1354 cleanup_prefix_route(ifp, expires, 1355 action == CLEANUP_PREFIX_RT_DEL, false); 1356 } 1357 1358 /* clean up prefsrc entries */ 1359 rt6_remove_prefsrc(ifp); 1360 out: 1361 in6_ifa_put(ifp); 1362 } 1363 1364 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev) 1365 { 1366 return READ_ONCE(idev->cnf.regen_min_advance) + 1367 READ_ONCE(idev->cnf.regen_max_retry) * 1368 READ_ONCE(idev->cnf.dad_transmits) * 1369 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ; 1370 } 1371 1372 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block) 1373 { 1374 struct inet6_dev *idev = ifp->idev; 1375 unsigned long tmp_tstamp, age; 1376 unsigned long regen_advance; 1377 unsigned long now = jiffies; 1378 u32 if_public_preferred_lft; 1379 s32 cnf_temp_preferred_lft; 1380 struct inet6_ifaddr *ift; 1381 struct ifa6_config cfg; 1382 long max_desync_factor; 1383 struct in6_addr addr; 1384 int ret = 0; 1385 1386 write_lock_bh(&idev->lock); 1387 1388 retry: 1389 in6_dev_hold(idev); 1390 if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) { 1391 write_unlock_bh(&idev->lock); 1392 pr_info("%s: use_tempaddr is disabled\n", __func__); 1393 in6_dev_put(idev); 1394 ret = -1; 1395 goto out; 1396 } 1397 spin_lock_bh(&ifp->lock); 1398 if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) { 1399 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/ 1400 spin_unlock_bh(&ifp->lock); 1401 write_unlock_bh(&idev->lock); 1402 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n", 1403 __func__); 1404 in6_dev_put(idev); 1405 ret = -1; 1406 goto out; 1407 } 1408 in6_ifa_hold(ifp); 1409 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8); 1410 ipv6_gen_rnd_iid(&addr); 1411 1412 age = (now - ifp->tstamp) / HZ; 1413 1414 regen_advance = ipv6_get_regen_advance(idev); 1415 1416 /* recalculate max_desync_factor each time and update 1417 * idev->desync_factor if it's larger 1418 */ 1419 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft); 1420 max_desync_factor = min_t(long, 1421 READ_ONCE(idev->cnf.max_desync_factor), 1422 cnf_temp_preferred_lft - regen_advance); 1423 1424 if (unlikely(idev->desync_factor > max_desync_factor)) { 1425 if (max_desync_factor > 0) { 1426 get_random_bytes(&idev->desync_factor, 1427 sizeof(idev->desync_factor)); 1428 idev->desync_factor %= max_desync_factor; 1429 } else { 1430 idev->desync_factor = 0; 1431 } 1432 } 1433 1434 if_public_preferred_lft = ifp->prefered_lft; 1435 1436 memset(&cfg, 0, sizeof(cfg)); 1437 cfg.valid_lft = min_t(__u32, ifp->valid_lft, 1438 READ_ONCE(idev->cnf.temp_valid_lft) + age); 1439 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor; 1440 cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft); 1441 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft); 1442 1443 cfg.plen = ifp->prefix_len; 1444 tmp_tstamp = ifp->tstamp; 1445 spin_unlock_bh(&ifp->lock); 1446 1447 write_unlock_bh(&idev->lock); 1448 1449 /* From RFC 4941: 1450 * 1451 * A temporary address is created only if this calculated Preferred 1452 * Lifetime is greater than REGEN_ADVANCE time units. In 1453 * particular, an implementation must not create a temporary address 1454 * with a zero Preferred Lifetime. 1455 * 1456 * ... 1457 * 1458 * When creating a temporary address, the lifetime values MUST be 1459 * derived from the corresponding prefix as follows: 1460 * 1461 * ... 1462 * 1463 * * Its Preferred Lifetime is the lower of the Preferred Lifetime 1464 * of the public address or TEMP_PREFERRED_LIFETIME - 1465 * DESYNC_FACTOR. 1466 * 1467 * To comply with the RFC's requirements, clamp the preferred lifetime 1468 * to a minimum of regen_advance, unless that would exceed valid_lft or 1469 * ifp->prefered_lft. 1470 * 1471 * Use age calculation as in addrconf_verify to avoid unnecessary 1472 * temporary addresses being generated. 1473 */ 1474 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 1475 if (cfg.preferred_lft <= regen_advance + age) { 1476 cfg.preferred_lft = regen_advance + age + 1; 1477 if (cfg.preferred_lft > cfg.valid_lft || 1478 cfg.preferred_lft > if_public_preferred_lft) { 1479 in6_ifa_put(ifp); 1480 in6_dev_put(idev); 1481 ret = -1; 1482 goto out; 1483 } 1484 } 1485 1486 cfg.ifa_flags = IFA_F_TEMPORARY; 1487 /* set in addrconf_prefix_rcv() */ 1488 if (ifp->flags & IFA_F_OPTIMISTIC) 1489 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 1490 1491 cfg.pfx = &addr; 1492 cfg.scope = ipv6_addr_scope(cfg.pfx); 1493 1494 ift = ipv6_add_addr(idev, &cfg, block, NULL); 1495 if (IS_ERR(ift)) { 1496 in6_ifa_put(ifp); 1497 in6_dev_put(idev); 1498 pr_info("%s: retry temporary address regeneration\n", __func__); 1499 write_lock_bh(&idev->lock); 1500 goto retry; 1501 } 1502 1503 spin_lock_bh(&ift->lock); 1504 ift->ifpub = ifp; 1505 ift->cstamp = now; 1506 ift->tstamp = tmp_tstamp; 1507 spin_unlock_bh(&ift->lock); 1508 1509 addrconf_dad_start(ift); 1510 in6_ifa_put(ift); 1511 in6_dev_put(idev); 1512 out: 1513 return ret; 1514 } 1515 1516 /* 1517 * Choose an appropriate source address (RFC3484) 1518 */ 1519 enum { 1520 IPV6_SADDR_RULE_INIT = 0, 1521 IPV6_SADDR_RULE_LOCAL, 1522 IPV6_SADDR_RULE_SCOPE, 1523 IPV6_SADDR_RULE_PREFERRED, 1524 #ifdef CONFIG_IPV6_MIP6 1525 IPV6_SADDR_RULE_HOA, 1526 #endif 1527 IPV6_SADDR_RULE_OIF, 1528 IPV6_SADDR_RULE_LABEL, 1529 IPV6_SADDR_RULE_PRIVACY, 1530 IPV6_SADDR_RULE_ORCHID, 1531 IPV6_SADDR_RULE_PREFIX, 1532 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1533 IPV6_SADDR_RULE_NOT_OPTIMISTIC, 1534 #endif 1535 IPV6_SADDR_RULE_MAX 1536 }; 1537 1538 struct ipv6_saddr_score { 1539 int rule; 1540 int addr_type; 1541 struct inet6_ifaddr *ifa; 1542 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX); 1543 int scopedist; 1544 int matchlen; 1545 }; 1546 1547 struct ipv6_saddr_dst { 1548 const struct in6_addr *addr; 1549 int ifindex; 1550 int scope; 1551 int label; 1552 unsigned int prefs; 1553 }; 1554 1555 static inline int ipv6_saddr_preferred(int type) 1556 { 1557 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK)) 1558 return 1; 1559 return 0; 1560 } 1561 1562 static bool ipv6_use_optimistic_addr(const struct net *net, 1563 const struct inet6_dev *idev) 1564 { 1565 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1566 if (!idev) 1567 return false; 1568 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) && 1569 !READ_ONCE(idev->cnf.optimistic_dad)) 1570 return false; 1571 if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) && 1572 !READ_ONCE(idev->cnf.use_optimistic)) 1573 return false; 1574 1575 return true; 1576 #else 1577 return false; 1578 #endif 1579 } 1580 1581 static bool ipv6_allow_optimistic_dad(const struct net *net, 1582 const struct inet6_dev *idev) 1583 { 1584 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1585 if (!idev) 1586 return false; 1587 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) && 1588 !READ_ONCE(idev->cnf.optimistic_dad)) 1589 return false; 1590 1591 return true; 1592 #else 1593 return false; 1594 #endif 1595 } 1596 1597 static int ipv6_get_saddr_eval(struct net *net, 1598 struct ipv6_saddr_score *score, 1599 struct ipv6_saddr_dst *dst, 1600 int i) 1601 { 1602 int ret; 1603 1604 if (i <= score->rule) { 1605 switch (i) { 1606 case IPV6_SADDR_RULE_SCOPE: 1607 ret = score->scopedist; 1608 break; 1609 case IPV6_SADDR_RULE_PREFIX: 1610 ret = score->matchlen; 1611 break; 1612 default: 1613 ret = !!test_bit(i, score->scorebits); 1614 } 1615 goto out; 1616 } 1617 1618 switch (i) { 1619 case IPV6_SADDR_RULE_INIT: 1620 /* Rule 0: remember if hiscore is not ready yet */ 1621 ret = !!score->ifa; 1622 break; 1623 case IPV6_SADDR_RULE_LOCAL: 1624 /* Rule 1: Prefer same address */ 1625 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr); 1626 break; 1627 case IPV6_SADDR_RULE_SCOPE: 1628 /* Rule 2: Prefer appropriate scope 1629 * 1630 * ret 1631 * ^ 1632 * -1 | d 15 1633 * ---+--+-+---> scope 1634 * | 1635 * | d is scope of the destination. 1636 * B-d | \ 1637 * | \ <- smaller scope is better if 1638 * B-15 | \ if scope is enough for destination. 1639 * | ret = B - scope (-1 <= scope >= d <= 15). 1640 * d-C-1 | / 1641 * |/ <- greater is better 1642 * -C / if scope is not enough for destination. 1643 * /| ret = scope - C (-1 <= d < scope <= 15). 1644 * 1645 * d - C - 1 < B -15 (for all -1 <= d <= 15). 1646 * C > d + 14 - B >= 15 + 14 - B = 29 - B. 1647 * Assume B = 0 and we get C > 29. 1648 */ 1649 ret = __ipv6_addr_src_scope(score->addr_type); 1650 if (ret >= dst->scope) 1651 ret = -ret; 1652 else 1653 ret -= 128; /* 30 is enough */ 1654 score->scopedist = ret; 1655 break; 1656 case IPV6_SADDR_RULE_PREFERRED: 1657 { 1658 /* Rule 3: Avoid deprecated and optimistic addresses */ 1659 u8 avoid = IFA_F_DEPRECATED; 1660 1661 if (!ipv6_use_optimistic_addr(net, score->ifa->idev)) 1662 avoid |= IFA_F_OPTIMISTIC; 1663 ret = ipv6_saddr_preferred(score->addr_type) || 1664 !(score->ifa->flags & avoid); 1665 break; 1666 } 1667 #ifdef CONFIG_IPV6_MIP6 1668 case IPV6_SADDR_RULE_HOA: 1669 { 1670 /* Rule 4: Prefer home address */ 1671 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA); 1672 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome; 1673 break; 1674 } 1675 #endif 1676 case IPV6_SADDR_RULE_OIF: 1677 /* Rule 5: Prefer outgoing interface */ 1678 ret = (!dst->ifindex || 1679 dst->ifindex == score->ifa->idev->dev->ifindex); 1680 break; 1681 case IPV6_SADDR_RULE_LABEL: 1682 /* Rule 6: Prefer matching label */ 1683 ret = ipv6_addr_label(net, 1684 &score->ifa->addr, score->addr_type, 1685 score->ifa->idev->dev->ifindex) == dst->label; 1686 break; 1687 case IPV6_SADDR_RULE_PRIVACY: 1688 { 1689 /* Rule 7: Prefer public address 1690 * Note: prefer temporary address if use_tempaddr >= 2 1691 */ 1692 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ? 1693 !!(dst->prefs & IPV6_PREFER_SRC_TMP) : 1694 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2; 1695 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp; 1696 break; 1697 } 1698 case IPV6_SADDR_RULE_ORCHID: 1699 /* Rule 8-: Prefer ORCHID vs ORCHID or 1700 * non-ORCHID vs non-ORCHID 1701 */ 1702 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^ 1703 ipv6_addr_orchid(dst->addr)); 1704 break; 1705 case IPV6_SADDR_RULE_PREFIX: 1706 /* Rule 8: Use longest matching prefix */ 1707 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr); 1708 if (ret > score->ifa->prefix_len) 1709 ret = score->ifa->prefix_len; 1710 score->matchlen = ret; 1711 break; 1712 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1713 case IPV6_SADDR_RULE_NOT_OPTIMISTIC: 1714 /* Optimistic addresses still have lower precedence than other 1715 * preferred addresses. 1716 */ 1717 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC); 1718 break; 1719 #endif 1720 default: 1721 ret = 0; 1722 } 1723 1724 if (ret) 1725 __set_bit(i, score->scorebits); 1726 score->rule = i; 1727 out: 1728 return ret; 1729 } 1730 1731 static int __ipv6_dev_get_saddr(struct net *net, 1732 struct ipv6_saddr_dst *dst, 1733 struct inet6_dev *idev, 1734 struct ipv6_saddr_score *scores, 1735 int hiscore_idx) 1736 { 1737 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx]; 1738 1739 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) { 1740 int i; 1741 1742 /* 1743 * - Tentative Address (RFC2462 section 5.4) 1744 * - A tentative address is not considered 1745 * "assigned to an interface" in the traditional 1746 * sense, unless it is also flagged as optimistic. 1747 * - Candidate Source Address (section 4) 1748 * - In any case, anycast addresses, multicast 1749 * addresses, and the unspecified address MUST 1750 * NOT be included in a candidate set. 1751 */ 1752 if ((score->ifa->flags & IFA_F_TENTATIVE) && 1753 (!(score->ifa->flags & IFA_F_OPTIMISTIC))) 1754 continue; 1755 1756 score->addr_type = __ipv6_addr_type(&score->ifa->addr); 1757 1758 if (unlikely(score->addr_type == IPV6_ADDR_ANY || 1759 score->addr_type & IPV6_ADDR_MULTICAST)) { 1760 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s", 1761 idev->dev->name); 1762 continue; 1763 } 1764 1765 score->rule = -1; 1766 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX); 1767 1768 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) { 1769 int minihiscore, miniscore; 1770 1771 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i); 1772 miniscore = ipv6_get_saddr_eval(net, score, dst, i); 1773 1774 if (minihiscore > miniscore) { 1775 if (i == IPV6_SADDR_RULE_SCOPE && 1776 score->scopedist > 0) { 1777 /* 1778 * special case: 1779 * each remaining entry 1780 * has too small (not enough) 1781 * scope, because ifa entries 1782 * are sorted by their scope 1783 * values. 1784 */ 1785 goto out; 1786 } 1787 break; 1788 } else if (minihiscore < miniscore) { 1789 swap(hiscore, score); 1790 hiscore_idx = 1 - hiscore_idx; 1791 1792 /* restore our iterator */ 1793 score->ifa = hiscore->ifa; 1794 1795 break; 1796 } 1797 } 1798 } 1799 out: 1800 return hiscore_idx; 1801 } 1802 1803 static int ipv6_get_saddr_master(struct net *net, 1804 const struct net_device *dst_dev, 1805 const struct net_device *master, 1806 struct ipv6_saddr_dst *dst, 1807 struct ipv6_saddr_score *scores, 1808 int hiscore_idx) 1809 { 1810 struct inet6_dev *idev; 1811 1812 idev = __in6_dev_get(dst_dev); 1813 if (idev) 1814 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1815 scores, hiscore_idx); 1816 1817 idev = __in6_dev_get(master); 1818 if (idev) 1819 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1820 scores, hiscore_idx); 1821 1822 return hiscore_idx; 1823 } 1824 1825 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev, 1826 const struct in6_addr *daddr, unsigned int prefs, 1827 struct in6_addr *saddr) 1828 { 1829 struct ipv6_saddr_score scores[2], *hiscore; 1830 struct ipv6_saddr_dst dst; 1831 struct inet6_dev *idev; 1832 struct net_device *dev; 1833 int dst_type; 1834 bool use_oif_addr = false; 1835 int hiscore_idx = 0; 1836 int ret = 0; 1837 1838 dst_type = __ipv6_addr_type(daddr); 1839 dst.addr = daddr; 1840 dst.ifindex = dst_dev ? dst_dev->ifindex : 0; 1841 dst.scope = __ipv6_addr_src_scope(dst_type); 1842 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex); 1843 dst.prefs = prefs; 1844 1845 scores[hiscore_idx].rule = -1; 1846 scores[hiscore_idx].ifa = NULL; 1847 1848 rcu_read_lock(); 1849 1850 /* Candidate Source Address (section 4) 1851 * - multicast and link-local destination address, 1852 * the set of candidate source address MUST only 1853 * include addresses assigned to interfaces 1854 * belonging to the same link as the outgoing 1855 * interface. 1856 * (- For site-local destination addresses, the 1857 * set of candidate source addresses MUST only 1858 * include addresses assigned to interfaces 1859 * belonging to the same site as the outgoing 1860 * interface.) 1861 * - "It is RECOMMENDED that the candidate source addresses 1862 * be the set of unicast addresses assigned to the 1863 * interface that will be used to send to the destination 1864 * (the 'outgoing' interface)." (RFC 6724) 1865 */ 1866 if (dst_dev) { 1867 idev = __in6_dev_get(dst_dev); 1868 if ((dst_type & IPV6_ADDR_MULTICAST) || 1869 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL || 1870 (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) { 1871 use_oif_addr = true; 1872 } 1873 } 1874 1875 if (use_oif_addr) { 1876 if (idev) 1877 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1878 } else { 1879 const struct net_device *master; 1880 int master_idx = 0; 1881 1882 /* if dst_dev exists and is enslaved to an L3 device, then 1883 * prefer addresses from dst_dev and then the master over 1884 * any other enslaved devices in the L3 domain. 1885 */ 1886 master = l3mdev_master_dev_rcu(dst_dev); 1887 if (master) { 1888 master_idx = master->ifindex; 1889 1890 hiscore_idx = ipv6_get_saddr_master(net, dst_dev, 1891 master, &dst, 1892 scores, hiscore_idx); 1893 1894 if (scores[hiscore_idx].ifa && 1895 scores[hiscore_idx].scopedist >= 0) 1896 goto out; 1897 } 1898 1899 for_each_netdev_rcu(net, dev) { 1900 /* only consider addresses on devices in the 1901 * same L3 domain 1902 */ 1903 if (l3mdev_master_ifindex_rcu(dev) != master_idx) 1904 continue; 1905 idev = __in6_dev_get(dev); 1906 if (!idev) 1907 continue; 1908 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1909 } 1910 } 1911 1912 out: 1913 hiscore = &scores[hiscore_idx]; 1914 if (!hiscore->ifa) 1915 ret = -EADDRNOTAVAIL; 1916 else 1917 *saddr = hiscore->ifa->addr; 1918 1919 rcu_read_unlock(); 1920 return ret; 1921 } 1922 EXPORT_SYMBOL(ipv6_dev_get_saddr); 1923 1924 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr, 1925 u32 banned_flags) 1926 { 1927 struct inet6_ifaddr *ifp; 1928 int err = -EADDRNOTAVAIL; 1929 1930 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 1931 if (ifp->scope > IFA_LINK) 1932 break; 1933 if (ifp->scope == IFA_LINK && 1934 !(ifp->flags & banned_flags)) { 1935 *addr = ifp->addr; 1936 err = 0; 1937 break; 1938 } 1939 } 1940 return err; 1941 } 1942 1943 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr, 1944 u32 banned_flags) 1945 { 1946 struct inet6_dev *idev; 1947 int err = -EADDRNOTAVAIL; 1948 1949 rcu_read_lock(); 1950 idev = __in6_dev_get(dev); 1951 if (idev) { 1952 read_lock_bh(&idev->lock); 1953 err = __ipv6_get_lladdr(idev, addr, banned_flags); 1954 read_unlock_bh(&idev->lock); 1955 } 1956 rcu_read_unlock(); 1957 return err; 1958 } 1959 1960 static int ipv6_count_addresses(const struct inet6_dev *idev) 1961 { 1962 const struct inet6_ifaddr *ifp; 1963 int cnt = 0; 1964 1965 rcu_read_lock(); 1966 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list) 1967 cnt++; 1968 rcu_read_unlock(); 1969 return cnt; 1970 } 1971 1972 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr, 1973 const struct net_device *dev, int strict) 1974 { 1975 return ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1976 strict, IFA_F_TENTATIVE); 1977 } 1978 EXPORT_SYMBOL(ipv6_chk_addr); 1979 1980 /* device argument is used to find the L3 domain of interest. If 1981 * skip_dev_check is set, then the ifp device is not checked against 1982 * the passed in dev argument. So the 2 cases for addresses checks are: 1983 * 1. does the address exist in the L3 domain that dev is part of 1984 * (skip_dev_check = true), or 1985 * 1986 * 2. does the address exist on the specific device 1987 * (skip_dev_check = false) 1988 */ 1989 static struct net_device * 1990 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 1991 const struct net_device *dev, bool skip_dev_check, 1992 int strict, u32 banned_flags) 1993 { 1994 unsigned int hash = inet6_addr_hash(net, addr); 1995 struct net_device *l3mdev, *ndev; 1996 struct inet6_ifaddr *ifp; 1997 u32 ifp_flags; 1998 1999 rcu_read_lock(); 2000 2001 l3mdev = l3mdev_master_dev_rcu(dev); 2002 if (skip_dev_check) 2003 dev = NULL; 2004 2005 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 2006 ndev = ifp->idev->dev; 2007 2008 if (l3mdev_master_dev_rcu(ndev) != l3mdev) 2009 continue; 2010 2011 /* Decouple optimistic from tentative for evaluation here. 2012 * Ban optimistic addresses explicitly, when required. 2013 */ 2014 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC) 2015 ? (ifp->flags&~IFA_F_TENTATIVE) 2016 : ifp->flags; 2017 if (ipv6_addr_equal(&ifp->addr, addr) && 2018 !(ifp_flags&banned_flags) && 2019 (!dev || ndev == dev || 2020 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) { 2021 rcu_read_unlock(); 2022 return ndev; 2023 } 2024 } 2025 2026 rcu_read_unlock(); 2027 return NULL; 2028 } 2029 2030 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 2031 const struct net_device *dev, bool skip_dev_check, 2032 int strict, u32 banned_flags) 2033 { 2034 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check, 2035 strict, banned_flags) ? 1 : 0; 2036 } 2037 EXPORT_SYMBOL(ipv6_chk_addr_and_flags); 2038 2039 2040 /* Compares an address/prefix_len with addresses on device @dev. 2041 * If one is found it returns true. 2042 */ 2043 bool ipv6_chk_custom_prefix(const struct in6_addr *addr, 2044 const unsigned int prefix_len, struct net_device *dev) 2045 { 2046 const struct inet6_ifaddr *ifa; 2047 const struct inet6_dev *idev; 2048 bool ret = false; 2049 2050 rcu_read_lock(); 2051 idev = __in6_dev_get(dev); 2052 if (idev) { 2053 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 2054 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len); 2055 if (ret) 2056 break; 2057 } 2058 } 2059 rcu_read_unlock(); 2060 2061 return ret; 2062 } 2063 EXPORT_SYMBOL(ipv6_chk_custom_prefix); 2064 2065 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev) 2066 { 2067 const struct inet6_ifaddr *ifa; 2068 const struct inet6_dev *idev; 2069 int onlink; 2070 2071 onlink = 0; 2072 rcu_read_lock(); 2073 idev = __in6_dev_get(dev); 2074 if (idev) { 2075 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 2076 onlink = ipv6_prefix_equal(addr, &ifa->addr, 2077 ifa->prefix_len); 2078 if (onlink) 2079 break; 2080 } 2081 } 2082 rcu_read_unlock(); 2083 return onlink; 2084 } 2085 EXPORT_SYMBOL(ipv6_chk_prefix); 2086 2087 /** 2088 * ipv6_dev_find - find the first device with a given source address. 2089 * @net: the net namespace 2090 * @addr: the source address 2091 * @dev: used to find the L3 domain of interest 2092 * 2093 * The caller should be protected by RCU, or RTNL. 2094 */ 2095 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr, 2096 struct net_device *dev) 2097 { 2098 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1, 2099 IFA_F_TENTATIVE); 2100 } 2101 EXPORT_SYMBOL(ipv6_dev_find); 2102 2103 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr, 2104 struct net_device *dev, int strict) 2105 { 2106 unsigned int hash = inet6_addr_hash(net, addr); 2107 struct inet6_ifaddr *ifp, *result = NULL; 2108 2109 rcu_read_lock(); 2110 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 2111 if (ipv6_addr_equal(&ifp->addr, addr)) { 2112 if (!dev || ifp->idev->dev == dev || 2113 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) { 2114 if (in6_ifa_hold_safe(ifp)) { 2115 result = ifp; 2116 break; 2117 } 2118 } 2119 } 2120 } 2121 rcu_read_unlock(); 2122 2123 return result; 2124 } 2125 2126 /* Gets referenced address, destroys ifaddr */ 2127 2128 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed) 2129 { 2130 if (dad_failed) 2131 ifp->flags |= IFA_F_DADFAILED; 2132 2133 if (ifp->flags&IFA_F_TEMPORARY) { 2134 struct inet6_ifaddr *ifpub; 2135 spin_lock_bh(&ifp->lock); 2136 ifpub = ifp->ifpub; 2137 if (ifpub) { 2138 in6_ifa_hold(ifpub); 2139 spin_unlock_bh(&ifp->lock); 2140 ipv6_create_tempaddr(ifpub, true); 2141 in6_ifa_put(ifpub); 2142 } else { 2143 spin_unlock_bh(&ifp->lock); 2144 } 2145 ipv6_del_addr(ifp); 2146 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) { 2147 spin_lock_bh(&ifp->lock); 2148 addrconf_del_dad_work(ifp); 2149 ifp->flags |= IFA_F_TENTATIVE; 2150 if (dad_failed) 2151 ifp->flags &= ~IFA_F_OPTIMISTIC; 2152 spin_unlock_bh(&ifp->lock); 2153 if (dad_failed) 2154 ipv6_ifa_notify(0, ifp); 2155 in6_ifa_put(ifp); 2156 } else { 2157 ipv6_del_addr(ifp); 2158 } 2159 } 2160 2161 static int addrconf_dad_end(struct inet6_ifaddr *ifp) 2162 { 2163 int err = -ENOENT; 2164 2165 spin_lock_bh(&ifp->lock); 2166 if (ifp->state == INET6_IFADDR_STATE_DAD) { 2167 ifp->state = INET6_IFADDR_STATE_POSTDAD; 2168 err = 0; 2169 } 2170 spin_unlock_bh(&ifp->lock); 2171 2172 return err; 2173 } 2174 2175 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp) 2176 { 2177 struct inet6_dev *idev = ifp->idev; 2178 struct net *net = dev_net(idev->dev); 2179 int max_addresses; 2180 2181 spin_lock_bh(&ifp->lock); 2182 2183 if (ifp->state != INET6_IFADDR_STATE_DAD) { 2184 spin_unlock_bh(&ifp->lock); 2185 in6_ifa_put(ifp); 2186 return; 2187 } 2188 ifp->state = INET6_IFADDR_STATE_POSTDAD; 2189 2190 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n", 2191 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source); 2192 2193 if (ifp->flags & IFA_F_STABLE_PRIVACY) { 2194 struct in6_addr new_addr; 2195 struct inet6_ifaddr *ifp2; 2196 int retries = ifp->stable_privacy_retry + 1; 2197 struct ifa6_config cfg = { 2198 .pfx = &new_addr, 2199 .plen = ifp->prefix_len, 2200 .ifa_flags = ifp->flags, 2201 .valid_lft = ifp->valid_lft, 2202 .preferred_lft = ifp->prefered_lft, 2203 .scope = ifp->scope, 2204 }; 2205 2206 if (retries > net->ipv6.sysctl.idgen_retries) { 2207 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n", 2208 ifp->idev->dev->name); 2209 goto errdad; 2210 } 2211 2212 new_addr = ifp->addr; 2213 if (ipv6_generate_stable_address(&new_addr, retries, 2214 idev)) 2215 goto errdad; 2216 2217 spin_unlock_bh(&ifp->lock); 2218 2219 max_addresses = READ_ONCE(idev->cnf.max_addresses); 2220 if (max_addresses && 2221 ipv6_count_addresses(idev) >= max_addresses) 2222 goto lock_errdad; 2223 2224 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n", 2225 ifp->idev->dev->name); 2226 2227 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL); 2228 if (IS_ERR(ifp2)) 2229 goto lock_errdad; 2230 2231 spin_lock_bh(&ifp2->lock); 2232 ifp2->stable_privacy_retry = retries; 2233 ifp2->state = INET6_IFADDR_STATE_PREDAD; 2234 spin_unlock_bh(&ifp2->lock); 2235 2236 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay); 2237 in6_ifa_put(ifp2); 2238 lock_errdad: 2239 spin_lock_bh(&ifp->lock); 2240 if (ifp->state != INET6_IFADDR_STATE_POSTDAD) { 2241 spin_unlock_bh(&ifp->lock); 2242 in6_ifa_put(ifp); 2243 return; 2244 } 2245 } 2246 2247 errdad: 2248 /* transition from _POSTDAD to _ERRDAD */ 2249 ifp->state = INET6_IFADDR_STATE_ERRDAD; 2250 spin_unlock_bh(&ifp->lock); 2251 2252 addrconf_mod_dad_work(ifp, 0); 2253 in6_ifa_put(ifp); 2254 } 2255 2256 /* Join to solicited addr multicast group. */ 2257 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr) 2258 { 2259 struct in6_addr maddr; 2260 2261 if (READ_ONCE(dev->flags) & (IFF_LOOPBACK | IFF_NOARP)) 2262 return; 2263 2264 addrconf_addr_solict_mult(addr, &maddr); 2265 ipv6_dev_mc_inc(dev, &maddr); 2266 } 2267 2268 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr) 2269 { 2270 struct in6_addr maddr; 2271 2272 if (READ_ONCE(idev->dev->flags) & (IFF_LOOPBACK | IFF_NOARP)) 2273 return; 2274 2275 addrconf_addr_solict_mult(addr, &maddr); 2276 __ipv6_dev_mc_dec(idev, &maddr); 2277 } 2278 2279 static void addrconf_join_anycast(struct inet6_ifaddr *ifp) 2280 { 2281 struct in6_addr addr; 2282 2283 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2284 return; 2285 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2286 if (ipv6_addr_any(&addr)) 2287 return; 2288 __ipv6_dev_ac_inc(ifp->idev, &addr); 2289 } 2290 2291 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp) 2292 { 2293 struct in6_addr addr; 2294 2295 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2296 return; 2297 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2298 if (ipv6_addr_any(&addr)) 2299 return; 2300 __ipv6_dev_ac_dec(ifp->idev, &addr); 2301 } 2302 2303 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev) 2304 { 2305 switch (dev->addr_len) { 2306 case ETH_ALEN: 2307 memcpy(eui, dev->dev_addr, 3); 2308 eui[3] = 0xFF; 2309 eui[4] = 0xFE; 2310 memcpy(eui + 5, dev->dev_addr + 3, 3); 2311 break; 2312 case EUI64_ADDR_LEN: 2313 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN); 2314 eui[0] ^= 2; 2315 break; 2316 default: 2317 return -1; 2318 } 2319 2320 return 0; 2321 } 2322 2323 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev) 2324 { 2325 const union fwnet_hwaddr *ha; 2326 2327 if (dev->addr_len != FWNET_ALEN) 2328 return -1; 2329 2330 ha = (const union fwnet_hwaddr *)dev->dev_addr; 2331 2332 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id)); 2333 eui[0] ^= 2; 2334 return 0; 2335 } 2336 2337 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev) 2338 { 2339 /* XXX: inherit EUI-64 from other interface -- yoshfuji */ 2340 if (dev->addr_len != ARCNET_ALEN) 2341 return -1; 2342 memset(eui, 0, 7); 2343 eui[7] = *(u8 *)dev->dev_addr; 2344 return 0; 2345 } 2346 2347 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev) 2348 { 2349 if (dev->addr_len != INFINIBAND_ALEN) 2350 return -1; 2351 memcpy(eui, dev->dev_addr + 12, 8); 2352 eui[0] |= 2; 2353 return 0; 2354 } 2355 2356 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr) 2357 { 2358 if (addr == 0) 2359 return -1; 2360 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) || 2361 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) || 2362 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) || 2363 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) || 2364 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) || 2365 ipv4_is_lbcast(addr)) ? 0x00 : 0x02; 2366 eui[1] = 0; 2367 eui[2] = 0x5E; 2368 eui[3] = 0xFE; 2369 memcpy(eui + 4, &addr, 4); 2370 return 0; 2371 } 2372 2373 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev) 2374 { 2375 if (dev->priv_flags & IFF_ISATAP) 2376 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2377 return -1; 2378 } 2379 2380 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev) 2381 { 2382 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2383 } 2384 2385 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev) 2386 { 2387 memcpy(eui, dev->perm_addr, 3); 2388 memcpy(eui + 5, dev->perm_addr + 3, 3); 2389 eui[3] = 0xFF; 2390 eui[4] = 0xFE; 2391 eui[0] ^= 2; 2392 return 0; 2393 } 2394 2395 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev) 2396 { 2397 switch (dev->type) { 2398 case ARPHRD_ETHER: 2399 case ARPHRD_FDDI: 2400 return addrconf_ifid_eui48(eui, dev); 2401 case ARPHRD_ARCNET: 2402 return addrconf_ifid_arcnet(eui, dev); 2403 case ARPHRD_INFINIBAND: 2404 return addrconf_ifid_infiniband(eui, dev); 2405 case ARPHRD_SIT: 2406 return addrconf_ifid_sit(eui, dev); 2407 case ARPHRD_IPGRE: 2408 case ARPHRD_TUNNEL: 2409 return addrconf_ifid_gre(eui, dev); 2410 case ARPHRD_6LOWPAN: 2411 return addrconf_ifid_6lowpan(eui, dev); 2412 case ARPHRD_IEEE1394: 2413 return addrconf_ifid_ieee1394(eui, dev); 2414 case ARPHRD_TUNNEL6: 2415 case ARPHRD_IP6GRE: 2416 case ARPHRD_RAWIP: 2417 return addrconf_ifid_ip6tnl(eui, dev); 2418 } 2419 return -1; 2420 } 2421 2422 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev) 2423 { 2424 int err = -1; 2425 struct inet6_ifaddr *ifp; 2426 2427 read_lock_bh(&idev->lock); 2428 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 2429 if (ifp->scope > IFA_LINK) 2430 break; 2431 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) { 2432 memcpy(eui, ifp->addr.s6_addr+8, 8); 2433 err = 0; 2434 break; 2435 } 2436 } 2437 read_unlock_bh(&idev->lock); 2438 return err; 2439 } 2440 2441 /* Generation of a randomized Interface Identifier 2442 * draft-ietf-6man-rfc4941bis, Section 3.3.1 2443 */ 2444 2445 static void ipv6_gen_rnd_iid(struct in6_addr *addr) 2446 { 2447 regen: 2448 get_random_bytes(&addr->s6_addr[8], 8); 2449 2450 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1: 2451 * check if generated address is not inappropriate: 2452 * 2453 * - Reserved IPv6 Interface Identifiers 2454 * - XXX: already assigned to an address on the device 2455 */ 2456 2457 /* Subnet-router anycast: 0000:0000:0000:0000 */ 2458 if (!(addr->s6_addr32[2] | addr->s6_addr32[3])) 2459 goto regen; 2460 2461 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212 2462 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213 2463 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF 2464 */ 2465 if (ntohl(addr->s6_addr32[2]) == 0x02005eff && 2466 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000) 2467 goto regen; 2468 2469 /* Reserved subnet anycast addresses */ 2470 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff && 2471 ntohl(addr->s6_addr32[3]) >= 0Xffffff80) 2472 goto regen; 2473 } 2474 2475 /* 2476 * Add prefix route. 2477 */ 2478 2479 static void 2480 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric, 2481 struct net_device *dev, unsigned long expires, 2482 u32 flags, gfp_t gfp_flags) 2483 { 2484 struct fib6_config cfg = { 2485 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX, 2486 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF, 2487 .fc_ifindex = dev->ifindex, 2488 .fc_expires = expires, 2489 .fc_dst_len = plen, 2490 .fc_flags = RTF_UP | flags, 2491 .fc_nlinfo.nl_net = dev_net(dev), 2492 .fc_protocol = RTPROT_KERNEL, 2493 .fc_type = RTN_UNICAST, 2494 }; 2495 2496 cfg.fc_dst = *pfx; 2497 2498 /* Prevent useless cloning on PtP SIT. 2499 This thing is done here expecting that the whole 2500 class of non-broadcast devices need not cloning. 2501 */ 2502 #if IS_ENABLED(CONFIG_IPV6_SIT) 2503 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT)) 2504 cfg.fc_flags |= RTF_NONEXTHOP; 2505 #endif 2506 2507 ip6_route_add(&cfg, gfp_flags, NULL); 2508 } 2509 2510 2511 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 2512 int plen, 2513 const struct net_device *dev, 2514 u32 flags, u32 noflags, 2515 bool no_gw) 2516 { 2517 struct fib6_node *fn; 2518 struct fib6_info *rt = NULL; 2519 struct fib6_table *table; 2520 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX; 2521 2522 table = fib6_get_table(dev_net(dev), tb_id); 2523 if (!table) 2524 return NULL; 2525 2526 rcu_read_lock(); 2527 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true); 2528 if (!fn) 2529 goto out; 2530 2531 for_each_fib6_node_rt_rcu(fn) { 2532 /* prefix routes only use builtin fib6_nh */ 2533 if (rt->nh) 2534 continue; 2535 2536 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex) 2537 continue; 2538 if (no_gw && rt->fib6_nh->fib_nh_gw_family) 2539 continue; 2540 if ((rt->fib6_flags & flags) != flags) 2541 continue; 2542 if ((rt->fib6_flags & noflags) != 0) 2543 continue; 2544 if (!fib6_info_hold_safe(rt)) 2545 continue; 2546 break; 2547 } 2548 out: 2549 rcu_read_unlock(); 2550 return rt; 2551 } 2552 2553 2554 /* Create "default" multicast route to the interface */ 2555 2556 static void addrconf_add_mroute(struct net_device *dev) 2557 { 2558 struct fib6_config cfg = { 2559 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL, 2560 .fc_metric = IP6_RT_PRIO_ADDRCONF, 2561 .fc_ifindex = dev->ifindex, 2562 .fc_dst_len = 8, 2563 .fc_flags = RTF_UP, 2564 .fc_type = RTN_MULTICAST, 2565 .fc_nlinfo.nl_net = dev_net(dev), 2566 .fc_protocol = RTPROT_KERNEL, 2567 }; 2568 2569 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0); 2570 2571 ip6_route_add(&cfg, GFP_KERNEL, NULL); 2572 } 2573 2574 static struct inet6_dev *addrconf_add_dev(struct net_device *dev) 2575 { 2576 struct inet6_dev *idev; 2577 2578 ASSERT_RTNL(); 2579 2580 idev = ipv6_find_idev(dev); 2581 if (IS_ERR(idev)) 2582 return idev; 2583 2584 if (idev->cnf.disable_ipv6) 2585 return ERR_PTR(-EACCES); 2586 2587 /* Add default multicast route */ 2588 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev)) 2589 addrconf_add_mroute(dev); 2590 2591 return idev; 2592 } 2593 2594 static void delete_tempaddrs(struct inet6_dev *idev, 2595 struct inet6_ifaddr *ifp) 2596 { 2597 struct inet6_ifaddr *ift, *tmp; 2598 2599 write_lock_bh(&idev->lock); 2600 list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) { 2601 if (ift->ifpub != ifp) 2602 continue; 2603 2604 in6_ifa_hold(ift); 2605 write_unlock_bh(&idev->lock); 2606 ipv6_del_addr(ift); 2607 write_lock_bh(&idev->lock); 2608 } 2609 write_unlock_bh(&idev->lock); 2610 } 2611 2612 static void manage_tempaddrs(struct inet6_dev *idev, 2613 struct inet6_ifaddr *ifp, 2614 __u32 valid_lft, __u32 prefered_lft, 2615 bool create, unsigned long now) 2616 { 2617 u32 orig_prefered_lft = prefered_lft; 2618 struct inet6_ifaddr *ift; 2619 bool reset_done = false; 2620 u32 flags; 2621 2622 read_lock_bh(&idev->lock); 2623 /* update all temporary addresses in the list */ 2624 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) { 2625 int age, max_valid, max_prefered; 2626 2627 if (ifp != ift->ifpub) 2628 continue; 2629 2630 /* RFC 4941 section 3.3: 2631 * If a received option will extend the lifetime of a public 2632 * address, the lifetimes of temporary addresses should 2633 * be extended, subject to the overall constraint that no 2634 * temporary addresses should ever remain "valid" or "preferred" 2635 * for a time longer than (TEMP_VALID_LIFETIME) or 2636 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively. 2637 */ 2638 age = (now - ift->cstamp) / HZ; 2639 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age; 2640 if (max_valid < 0) 2641 max_valid = 0; 2642 2643 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) - 2644 idev->desync_factor - age; 2645 if (max_prefered < 0) 2646 max_prefered = 0; 2647 2648 if (valid_lft > max_valid) 2649 valid_lft = max_valid; 2650 2651 if (prefered_lft > max_prefered) 2652 prefered_lft = max_prefered; 2653 2654 spin_lock(&ift->lock); 2655 /* the first match is the most recent temp address */ 2656 if (!reset_done && orig_prefered_lft > 0) { 2657 ift->regen_count = 0; 2658 reset_done = true; 2659 } 2660 flags = ift->flags; 2661 ift->valid_lft = valid_lft; 2662 ift->prefered_lft = prefered_lft; 2663 ift->tstamp = now; 2664 if (prefered_lft > 0) 2665 ift->flags &= ~IFA_F_DEPRECATED; 2666 2667 spin_unlock(&ift->lock); 2668 if (!(flags&IFA_F_TENTATIVE)) 2669 ipv6_ifa_notify(0, ift); 2670 } 2671 2672 /* Also create a temporary address if it's enabled but no temporary 2673 * address currently exists. 2674 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false 2675 * as part of cleanup (ie. deleting the mngtmpaddr). 2676 * We don't want that to result in creating a new temporary ip address. 2677 */ 2678 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft)) 2679 create = true; 2680 2681 if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) { 2682 /* When a new public address is created as described 2683 * in [ADDRCONF], also create a new temporary address. 2684 */ 2685 read_unlock_bh(&idev->lock); 2686 ipv6_create_tempaddr(ifp, false); 2687 } else { 2688 read_unlock_bh(&idev->lock); 2689 } 2690 } 2691 2692 static bool is_addr_mode_generate_stable(struct inet6_dev *idev) 2693 { 2694 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY || 2695 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM; 2696 } 2697 2698 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev, 2699 const struct prefix_info *pinfo, 2700 struct inet6_dev *in6_dev, 2701 const struct in6_addr *addr, int addr_type, 2702 u32 addr_flags, bool sllao, bool tokenized, 2703 __u32 valid_lft, u32 prefered_lft) 2704 { 2705 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1); 2706 int create = 0, update_lft = 0; 2707 2708 if (!ifp && valid_lft) { 2709 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses); 2710 struct ifa6_config cfg = { 2711 .pfx = addr, 2712 .plen = pinfo->prefix_len, 2713 .ifa_flags = addr_flags, 2714 .valid_lft = valid_lft, 2715 .preferred_lft = prefered_lft, 2716 .scope = addr_type & IPV6_ADDR_SCOPE_MASK, 2717 .ifa_proto = IFAPROT_KERNEL_RA 2718 }; 2719 2720 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 2721 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) || 2722 READ_ONCE(in6_dev->cnf.optimistic_dad)) && 2723 !net->ipv6.devconf_all->forwarding && sllao) 2724 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 2725 #endif 2726 2727 /* Do not allow to create too much of autoconfigured 2728 * addresses; this would be too easy way to crash kernel. 2729 */ 2730 if (!max_addresses || 2731 ipv6_count_addresses(in6_dev) < max_addresses) 2732 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL); 2733 2734 if (IS_ERR_OR_NULL(ifp)) 2735 return -1; 2736 2737 create = 1; 2738 spin_lock_bh(&ifp->lock); 2739 ifp->flags |= IFA_F_MANAGETEMPADDR; 2740 ifp->cstamp = jiffies; 2741 ifp->tokenized = tokenized; 2742 spin_unlock_bh(&ifp->lock); 2743 addrconf_dad_start(ifp); 2744 } 2745 2746 if (ifp) { 2747 u32 flags; 2748 unsigned long now; 2749 u32 stored_lft; 2750 2751 /* update lifetime (RFC2462 5.5.3 e) */ 2752 spin_lock_bh(&ifp->lock); 2753 now = jiffies; 2754 if (ifp->valid_lft > (now - ifp->tstamp) / HZ) 2755 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ; 2756 else 2757 stored_lft = 0; 2758 2759 /* RFC4862 Section 5.5.3e: 2760 * "Note that the preferred lifetime of the 2761 * corresponding address is always reset to 2762 * the Preferred Lifetime in the received 2763 * Prefix Information option, regardless of 2764 * whether the valid lifetime is also reset or 2765 * ignored." 2766 * 2767 * So we should always update prefered_lft here. 2768 */ 2769 update_lft = !create && stored_lft; 2770 2771 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) { 2772 const u32 minimum_lft = min_t(u32, 2773 stored_lft, MIN_VALID_LIFETIME); 2774 valid_lft = max(valid_lft, minimum_lft); 2775 } 2776 2777 if (update_lft) { 2778 ifp->valid_lft = valid_lft; 2779 ifp->prefered_lft = prefered_lft; 2780 WRITE_ONCE(ifp->tstamp, now); 2781 flags = ifp->flags; 2782 ifp->flags &= ~IFA_F_DEPRECATED; 2783 spin_unlock_bh(&ifp->lock); 2784 2785 if (!(flags&IFA_F_TENTATIVE)) 2786 ipv6_ifa_notify(0, ifp); 2787 } else 2788 spin_unlock_bh(&ifp->lock); 2789 2790 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft, 2791 create, now); 2792 2793 in6_ifa_put(ifp); 2794 addrconf_verify(net); 2795 } 2796 2797 return 0; 2798 } 2799 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr); 2800 2801 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao) 2802 { 2803 struct prefix_info *pinfo; 2804 struct fib6_table *table; 2805 __u32 valid_lft; 2806 __u32 prefered_lft; 2807 int addr_type, err; 2808 u32 addr_flags = 0; 2809 struct inet6_dev *in6_dev; 2810 struct net *net = dev_net(dev); 2811 bool ignore_autoconf = false; 2812 2813 pinfo = (struct prefix_info *) opt; 2814 2815 if (len < sizeof(struct prefix_info)) { 2816 netdev_dbg(dev, "addrconf: prefix option too short\n"); 2817 return; 2818 } 2819 2820 /* 2821 * Validation checks ([ADDRCONF], page 19) 2822 */ 2823 2824 addr_type = ipv6_addr_type(&pinfo->prefix); 2825 2826 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)) 2827 return; 2828 2829 valid_lft = ntohl(pinfo->valid); 2830 prefered_lft = ntohl(pinfo->prefered); 2831 2832 if (prefered_lft > valid_lft) { 2833 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n"); 2834 return; 2835 } 2836 2837 in6_dev = in6_dev_get(dev); 2838 2839 if (!in6_dev) { 2840 net_dbg_ratelimited("addrconf: device %s not configured\n", 2841 dev->name); 2842 return; 2843 } 2844 2845 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft) 2846 goto put; 2847 2848 /* 2849 * Two things going on here: 2850 * 1) Add routes for on-link prefixes 2851 * 2) Configure prefixes with the auto flag set 2852 */ 2853 2854 if (pinfo->onlink) { 2855 struct fib6_info *rt; 2856 unsigned long rt_expires; 2857 2858 /* Avoid arithmetic overflow. Really, we could 2859 * save rt_expires in seconds, likely valid_lft, 2860 * but it would require division in fib gc, that it 2861 * not good. 2862 */ 2863 if (HZ > USER_HZ) 2864 rt_expires = addrconf_timeout_fixup(valid_lft, HZ); 2865 else 2866 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ); 2867 2868 if (addrconf_finite_timeout(rt_expires)) 2869 rt_expires *= HZ; 2870 2871 rt = addrconf_get_prefix_route(&pinfo->prefix, 2872 pinfo->prefix_len, 2873 dev, 2874 RTF_ADDRCONF | RTF_PREFIX_RT, 2875 RTF_DEFAULT, true); 2876 2877 if (rt) { 2878 /* Autoconf prefix route */ 2879 if (valid_lft == 0) { 2880 ip6_del_rt_reason(net, rt, 2881 RT_DEL_REASON_RA_WITHDRAWN); 2882 rt = NULL; 2883 } else { 2884 table = rt->fib6_table; 2885 spin_lock_bh(&table->tb6_lock); 2886 2887 if (addrconf_finite_timeout(rt_expires)) { 2888 /* not infinity */ 2889 fib6_set_expires(rt, jiffies + rt_expires); 2890 fib6_add_gc_list(rt); 2891 } else { 2892 fib6_clean_expires(rt); 2893 fib6_may_remove_gc_list(net, rt); 2894 } 2895 2896 spin_unlock_bh(&table->tb6_lock); 2897 } 2898 } else if (valid_lft) { 2899 clock_t expires = 0; 2900 int flags = RTF_ADDRCONF | RTF_PREFIX_RT; 2901 if (addrconf_finite_timeout(rt_expires)) { 2902 /* not infinity */ 2903 flags |= RTF_EXPIRES; 2904 expires = jiffies_to_clock_t(rt_expires); 2905 } 2906 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len, 2907 0, dev, expires, flags, 2908 GFP_ATOMIC); 2909 } 2910 fib6_info_release(rt); 2911 } 2912 2913 /* Try to figure out our local address for this prefix */ 2914 2915 ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd; 2916 if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) { 2917 struct in6_addr addr; 2918 bool tokenized = false, dev_addr_generated = false; 2919 2920 if (pinfo->prefix_len == 64) { 2921 memcpy(&addr, &pinfo->prefix, 8); 2922 2923 if (!ipv6_addr_any(&in6_dev->token)) { 2924 read_lock_bh(&in6_dev->lock); 2925 memcpy(addr.s6_addr + 8, 2926 in6_dev->token.s6_addr + 8, 8); 2927 read_unlock_bh(&in6_dev->lock); 2928 tokenized = true; 2929 } else if (is_addr_mode_generate_stable(in6_dev) && 2930 !ipv6_generate_stable_address(&addr, 0, 2931 in6_dev)) { 2932 addr_flags |= IFA_F_STABLE_PRIVACY; 2933 goto ok; 2934 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) && 2935 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) { 2936 goto put; 2937 } else { 2938 dev_addr_generated = true; 2939 } 2940 goto ok; 2941 } 2942 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n", 2943 pinfo->prefix_len); 2944 goto put; 2945 2946 ok: 2947 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, 2948 &addr, addr_type, 2949 addr_flags, sllao, 2950 tokenized, valid_lft, 2951 prefered_lft); 2952 if (err) 2953 goto put; 2954 2955 /* Ignore error case here because previous prefix add addr was 2956 * successful which will be notified. 2957 */ 2958 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr, 2959 addr_type, addr_flags, sllao, 2960 tokenized, valid_lft, 2961 prefered_lft, 2962 dev_addr_generated); 2963 } 2964 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo); 2965 put: 2966 in6_dev_put(in6_dev); 2967 } 2968 2969 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev, 2970 struct in6_ifreq *ireq) 2971 { 2972 struct ip_tunnel_parm_kern p = { }; 2973 int err; 2974 2975 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4)) 2976 return -EADDRNOTAVAIL; 2977 2978 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3]; 2979 p.iph.version = 4; 2980 p.iph.ihl = 5; 2981 p.iph.protocol = IPPROTO_IPV6; 2982 p.iph.ttl = 64; 2983 2984 if (!dev->netdev_ops->ndo_tunnel_ctl) 2985 return -EOPNOTSUPP; 2986 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL); 2987 if (err) 2988 return err; 2989 2990 dev = __dev_get_by_name(net, p.name); 2991 if (!dev) 2992 return -ENOBUFS; 2993 return dev_open(dev, NULL); 2994 } 2995 2996 /* 2997 * Set destination address. 2998 * Special case for SIT interfaces where we create a new "virtual" 2999 * device. 3000 */ 3001 int addrconf_set_dstaddr(struct net *net, void __user *arg) 3002 { 3003 struct net_device *dev; 3004 struct in6_ifreq ireq; 3005 int err = -ENODEV; 3006 3007 if (!IS_ENABLED(CONFIG_IPV6_SIT)) 3008 return -ENODEV; 3009 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3010 return -EFAULT; 3011 3012 rtnl_net_lock(net); 3013 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 3014 if (dev && dev->type == ARPHRD_SIT) 3015 err = addrconf_set_sit_dstaddr(net, dev, &ireq); 3016 rtnl_net_unlock(net); 3017 return err; 3018 } 3019 3020 static int ipv6_mc_config(struct sock *sk, bool join, 3021 const struct in6_addr *addr, int ifindex) 3022 { 3023 int ret; 3024 3025 ASSERT_RTNL(); 3026 3027 lock_sock(sk); 3028 if (join) 3029 ret = ipv6_sock_mc_join(sk, ifindex, addr); 3030 else 3031 ret = ipv6_sock_mc_drop(sk, ifindex, addr); 3032 release_sock(sk); 3033 3034 return ret; 3035 } 3036 3037 /* 3038 * Manual configuration of address on an interface 3039 */ 3040 static int inet6_addr_add(struct net *net, struct net_device *dev, 3041 struct ifa6_config *cfg, clock_t expires, u32 flags, 3042 struct netlink_ext_ack *extack) 3043 { 3044 struct inet6_ifaddr *ifp; 3045 struct inet6_dev *idev; 3046 3047 ASSERT_RTNL_NET(net); 3048 3049 if (cfg->plen > 128) { 3050 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 3051 return -EINVAL; 3052 } 3053 3054 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) { 3055 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64"); 3056 return -EINVAL; 3057 } 3058 3059 idev = addrconf_add_dev(dev); 3060 if (IS_ERR(idev)) { 3061 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 3062 return PTR_ERR(idev); 3063 } 3064 3065 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 3066 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk, 3067 true, cfg->pfx, dev->ifindex); 3068 3069 if (ret < 0) { 3070 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed"); 3071 return ret; 3072 } 3073 } 3074 3075 cfg->scope = ipv6_addr_scope(cfg->pfx); 3076 3077 ifp = ipv6_add_addr(idev, cfg, true, extack); 3078 if (!IS_ERR(ifp)) { 3079 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 3080 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3081 ifp->rt_priority, dev, expires, 3082 flags, GFP_KERNEL); 3083 } 3084 3085 /* Send a netlink notification if DAD is enabled and 3086 * optimistic flag is not set 3087 */ 3088 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD))) 3089 ipv6_ifa_notify(0, ifp); 3090 /* 3091 * Note that section 3.1 of RFC 4429 indicates 3092 * that the Optimistic flag should not be set for 3093 * manually configured addresses 3094 */ 3095 addrconf_dad_start(ifp); 3096 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR) 3097 manage_tempaddrs(idev, ifp, cfg->valid_lft, 3098 cfg->preferred_lft, true, jiffies); 3099 in6_ifa_put(ifp); 3100 addrconf_verify_rtnl(net); 3101 return 0; 3102 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 3103 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false, 3104 cfg->pfx, dev->ifindex); 3105 } 3106 3107 return PTR_ERR(ifp); 3108 } 3109 3110 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags, 3111 const struct in6_addr *pfx, unsigned int plen, 3112 struct netlink_ext_ack *extack) 3113 { 3114 struct inet6_ifaddr *ifp; 3115 struct inet6_dev *idev; 3116 struct net_device *dev; 3117 3118 if (plen > 128) { 3119 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 3120 return -EINVAL; 3121 } 3122 3123 dev = __dev_get_by_index(net, ifindex); 3124 if (!dev) { 3125 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 3126 return -ENODEV; 3127 } 3128 3129 idev = __in6_dev_get_rtnl_net(dev); 3130 if (!idev) { 3131 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 3132 return -ENXIO; 3133 } 3134 3135 read_lock_bh(&idev->lock); 3136 list_for_each_entry(ifp, &idev->addr_list, if_list) { 3137 if (ifp->prefix_len == plen && 3138 ipv6_addr_equal(pfx, &ifp->addr)) { 3139 in6_ifa_hold(ifp); 3140 read_unlock_bh(&idev->lock); 3141 3142 if (!(ifp->flags & IFA_F_TEMPORARY) && 3143 (ifp->flags & IFA_F_MANAGETEMPADDR)) 3144 delete_tempaddrs(idev, ifp); 3145 3146 ipv6_del_addr(ifp); 3147 3148 addrconf_verify_rtnl(net); 3149 if (ipv6_addr_is_multicast(pfx)) { 3150 ipv6_mc_config(net->ipv6.mc_autojoin_sk, 3151 false, pfx, dev->ifindex); 3152 } 3153 return 0; 3154 } 3155 } 3156 read_unlock_bh(&idev->lock); 3157 3158 NL_SET_ERR_MSG_MOD(extack, "address not found"); 3159 return -EADDRNOTAVAIL; 3160 } 3161 3162 3163 int addrconf_add_ifaddr(struct net *net, void __user *arg) 3164 { 3165 struct ifa6_config cfg = { 3166 .ifa_flags = IFA_F_PERMANENT, 3167 .preferred_lft = INFINITY_LIFE_TIME, 3168 .valid_lft = INFINITY_LIFE_TIME, 3169 }; 3170 struct net_device *dev; 3171 struct in6_ifreq ireq; 3172 int err; 3173 3174 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 3175 return -EPERM; 3176 3177 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3178 return -EFAULT; 3179 3180 cfg.pfx = &ireq.ifr6_addr; 3181 cfg.plen = ireq.ifr6_prefixlen; 3182 3183 rtnl_net_lock(net); 3184 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 3185 if (dev) { 3186 netdev_lock_ops(dev); 3187 err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL); 3188 netdev_unlock_ops(dev); 3189 } else { 3190 err = -ENODEV; 3191 } 3192 rtnl_net_unlock(net); 3193 return err; 3194 } 3195 3196 int addrconf_del_ifaddr(struct net *net, void __user *arg) 3197 { 3198 struct in6_ifreq ireq; 3199 int err; 3200 3201 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 3202 return -EPERM; 3203 3204 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3205 return -EFAULT; 3206 3207 rtnl_net_lock(net); 3208 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr, 3209 ireq.ifr6_prefixlen, NULL); 3210 rtnl_net_unlock(net); 3211 return err; 3212 } 3213 3214 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr, 3215 int plen, int scope, u8 proto) 3216 { 3217 struct inet6_ifaddr *ifp; 3218 struct ifa6_config cfg = { 3219 .pfx = addr, 3220 .plen = plen, 3221 .ifa_flags = IFA_F_PERMANENT, 3222 .valid_lft = INFINITY_LIFE_TIME, 3223 .preferred_lft = INFINITY_LIFE_TIME, 3224 .scope = scope, 3225 .ifa_proto = proto 3226 }; 3227 3228 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3229 if (!IS_ERR(ifp)) { 3230 spin_lock_bh(&ifp->lock); 3231 ifp->flags &= ~IFA_F_TENTATIVE; 3232 spin_unlock_bh(&ifp->lock); 3233 rt_genid_bump_ipv6(dev_net(idev->dev)); 3234 ipv6_ifa_notify(RTM_NEWADDR, ifp); 3235 in6_ifa_put(ifp); 3236 } 3237 } 3238 3239 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) 3240 static void add_v4_addrs(struct inet6_dev *idev) 3241 { 3242 struct in6_addr addr; 3243 struct net_device *dev; 3244 struct net *net = dev_net(idev->dev); 3245 int scope, plen; 3246 u32 pflags = 0; 3247 3248 ASSERT_RTNL(); 3249 3250 memset(&addr, 0, sizeof(struct in6_addr)); 3251 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4); 3252 3253 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) { 3254 scope = IPV6_ADDR_COMPATv4; 3255 plen = 96; 3256 pflags |= RTF_NONEXTHOP; 3257 } else { 3258 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE) 3259 return; 3260 3261 addr.s6_addr32[0] = htonl(0xfe800000); 3262 scope = IFA_LINK; 3263 plen = 64; 3264 } 3265 3266 if (addr.s6_addr32[3]) { 3267 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC); 3268 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags, 3269 GFP_KERNEL); 3270 return; 3271 } 3272 3273 for_each_netdev(net, dev) { 3274 struct in_device *in_dev = __in_dev_get_rtnl(dev); 3275 if (in_dev && (dev->flags & IFF_UP)) { 3276 struct in_ifaddr *ifa; 3277 int flag = scope; 3278 3279 in_dev_for_each_ifa_rtnl(ifa, in_dev) { 3280 addr.s6_addr32[3] = ifa->ifa_local; 3281 3282 if (ifa->ifa_scope == RT_SCOPE_LINK) 3283 continue; 3284 if (ifa->ifa_scope >= RT_SCOPE_HOST) { 3285 if (idev->dev->flags&IFF_POINTOPOINT) 3286 continue; 3287 flag |= IFA_HOST; 3288 } 3289 3290 add_addr(idev, &addr, plen, flag, 3291 IFAPROT_UNSPEC); 3292 addrconf_prefix_route(&addr, plen, 0, idev->dev, 3293 0, pflags, GFP_KERNEL); 3294 } 3295 } 3296 } 3297 } 3298 #endif 3299 3300 static void init_loopback(struct net_device *dev) 3301 { 3302 struct inet6_dev *idev; 3303 3304 /* ::1 */ 3305 3306 ASSERT_RTNL(); 3307 3308 idev = ipv6_find_idev(dev); 3309 if (IS_ERR(idev)) { 3310 pr_debug("%s: add_dev failed\n", __func__); 3311 return; 3312 } 3313 3314 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO); 3315 } 3316 3317 void addrconf_add_linklocal(struct inet6_dev *idev, 3318 const struct in6_addr *addr, u32 flags) 3319 { 3320 struct ifa6_config cfg = { 3321 .pfx = addr, 3322 .plen = 64, 3323 .ifa_flags = flags | IFA_F_PERMANENT, 3324 .valid_lft = INFINITY_LIFE_TIME, 3325 .preferred_lft = INFINITY_LIFE_TIME, 3326 .scope = IFA_LINK, 3327 .ifa_proto = IFAPROT_KERNEL_LL 3328 }; 3329 struct inet6_ifaddr *ifp; 3330 3331 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 3332 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) || 3333 READ_ONCE(idev->cnf.optimistic_dad)) && 3334 !dev_net(idev->dev)->ipv6.devconf_all->forwarding) 3335 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 3336 #endif 3337 3338 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3339 if (!IS_ERR(ifp)) { 3340 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev, 3341 0, 0, GFP_ATOMIC); 3342 addrconf_dad_start(ifp); 3343 in6_ifa_put(ifp); 3344 } 3345 } 3346 EXPORT_SYMBOL_GPL(addrconf_add_linklocal); 3347 3348 static bool ipv6_reserved_interfaceid(struct in6_addr address) 3349 { 3350 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0) 3351 return true; 3352 3353 if (address.s6_addr32[2] == htonl(0x02005eff) && 3354 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000))) 3355 return true; 3356 3357 if (address.s6_addr32[2] == htonl(0xfdffffff) && 3358 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80))) 3359 return true; 3360 3361 return false; 3362 } 3363 3364 static int ipv6_generate_stable_address(struct in6_addr *address, 3365 u8 dad_count, 3366 const struct inet6_dev *idev) 3367 { 3368 static DEFINE_SPINLOCK(lock); 3369 static struct sha1_ctx sha_ctx; 3370 3371 static union { 3372 u8 __data[SHA1_BLOCK_SIZE]; 3373 struct { 3374 struct in6_addr secret; 3375 __be32 prefix[2]; 3376 unsigned char hwaddr[MAX_ADDR_LEN]; 3377 u8 dad_count; 3378 } __packed; 3379 } data; 3380 3381 struct in6_addr secret; 3382 struct in6_addr temp; 3383 struct net *net = dev_net(idev->dev); 3384 3385 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data)); 3386 3387 if (idev->cnf.stable_secret.initialized) 3388 secret = idev->cnf.stable_secret.secret; 3389 else if (net->ipv6.devconf_dflt->stable_secret.initialized) 3390 secret = net->ipv6.devconf_dflt->stable_secret.secret; 3391 else 3392 return -1; 3393 3394 retry: 3395 spin_lock_bh(&lock); 3396 3397 sha1_init(&sha_ctx); 3398 3399 memset(&data, 0, sizeof(data)); 3400 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len); 3401 data.prefix[0] = address->s6_addr32[0]; 3402 data.prefix[1] = address->s6_addr32[1]; 3403 data.secret = secret; 3404 data.dad_count = dad_count; 3405 3406 sha1_update(&sha_ctx, data.__data, sizeof(data)); 3407 3408 /* 3409 * Note that the SHA-1 finalization is omitted here, and the digest is 3410 * pulled directly from the internal SHA-1 state (making it incompatible 3411 * with standard SHA-1). Unusual, but technically okay since the data 3412 * length is fixed and is a multiple of the SHA-1 block size. 3413 */ 3414 temp = *address; 3415 temp.s6_addr32[2] = (__force __be32)sha_ctx.state.h[0]; 3416 temp.s6_addr32[3] = (__force __be32)sha_ctx.state.h[1]; 3417 3418 spin_unlock_bh(&lock); 3419 3420 if (ipv6_reserved_interfaceid(temp)) { 3421 dad_count++; 3422 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries) 3423 return -1; 3424 goto retry; 3425 } 3426 3427 *address = temp; 3428 return 0; 3429 } 3430 3431 static void ipv6_gen_mode_random_init(struct inet6_dev *idev) 3432 { 3433 struct ipv6_stable_secret *s = &idev->cnf.stable_secret; 3434 3435 if (s->initialized) 3436 return; 3437 s = &idev->cnf.stable_secret; 3438 get_random_bytes(&s->secret, sizeof(s->secret)); 3439 s->initialized = true; 3440 } 3441 3442 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route) 3443 { 3444 struct in6_addr addr; 3445 3446 /* no link local addresses on L3 master devices */ 3447 if (netif_is_l3_master(idev->dev)) 3448 return; 3449 3450 /* no link local addresses on devices flagged as slaves */ 3451 if (idev->dev->priv_flags & IFF_NO_ADDRCONF) 3452 return; 3453 3454 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0); 3455 3456 switch (idev->cnf.addr_gen_mode) { 3457 case IN6_ADDR_GEN_MODE_RANDOM: 3458 ipv6_gen_mode_random_init(idev); 3459 fallthrough; 3460 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY: 3461 if (!ipv6_generate_stable_address(&addr, 0, idev)) 3462 addrconf_add_linklocal(idev, &addr, 3463 IFA_F_STABLE_PRIVACY); 3464 else if (prefix_route) 3465 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3466 0, 0, GFP_KERNEL); 3467 break; 3468 case IN6_ADDR_GEN_MODE_EUI64: 3469 /* addrconf_add_linklocal also adds a prefix_route and we 3470 * only need to care about prefix routes if ipv6_generate_eui64 3471 * couldn't generate one. 3472 */ 3473 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0) 3474 addrconf_add_linklocal(idev, &addr, 0); 3475 else if (prefix_route) 3476 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3477 0, 0, GFP_KERNEL); 3478 break; 3479 case IN6_ADDR_GEN_MODE_NONE: 3480 default: 3481 /* will not add any link local address */ 3482 break; 3483 } 3484 } 3485 3486 static void addrconf_dev_config(struct net_device *dev) 3487 { 3488 struct inet6_dev *idev; 3489 3490 ASSERT_RTNL(); 3491 3492 if ((dev->type != ARPHRD_ETHER) && 3493 (dev->type != ARPHRD_FDDI) && 3494 (dev->type != ARPHRD_ARCNET) && 3495 (dev->type != ARPHRD_INFINIBAND) && 3496 (dev->type != ARPHRD_IEEE1394) && 3497 (dev->type != ARPHRD_TUNNEL6) && 3498 (dev->type != ARPHRD_6LOWPAN) && 3499 (dev->type != ARPHRD_IP6GRE) && 3500 (dev->type != ARPHRD_TUNNEL) && 3501 (dev->type != ARPHRD_NONE) && 3502 (dev->type != ARPHRD_RAWIP)) { 3503 /* Alas, we support only Ethernet autoconfiguration. */ 3504 idev = __in6_dev_get(dev); 3505 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP && 3506 dev->flags & IFF_MULTICAST) 3507 ipv6_mc_up(idev); 3508 return; 3509 } 3510 3511 idev = addrconf_add_dev(dev); 3512 if (IS_ERR(idev)) 3513 return; 3514 3515 /* this device type has no EUI support */ 3516 if (dev->type == ARPHRD_NONE && 3517 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) 3518 WRITE_ONCE(idev->cnf.addr_gen_mode, 3519 IN6_ADDR_GEN_MODE_RANDOM); 3520 3521 addrconf_addr_gen(idev, false); 3522 } 3523 3524 #if IS_ENABLED(CONFIG_IPV6_SIT) 3525 static void addrconf_sit_config(struct net_device *dev) 3526 { 3527 struct inet6_dev *idev; 3528 3529 ASSERT_RTNL(); 3530 3531 /* 3532 * Configure the tunnel with one of our IPv4 3533 * addresses... we should configure all of 3534 * our v4 addrs in the tunnel 3535 */ 3536 3537 idev = ipv6_find_idev(dev); 3538 if (IS_ERR(idev)) { 3539 pr_debug("%s: add_dev failed\n", __func__); 3540 return; 3541 } 3542 3543 if (dev->priv_flags & IFF_ISATAP) { 3544 addrconf_addr_gen(idev, false); 3545 return; 3546 } 3547 3548 add_v4_addrs(idev); 3549 3550 if (dev->flags&IFF_POINTOPOINT) 3551 addrconf_add_mroute(dev); 3552 } 3553 #endif 3554 3555 #if IS_ENABLED(CONFIG_NET_IPGRE) 3556 static void addrconf_gre_config(struct net_device *dev) 3557 { 3558 struct inet6_dev *idev; 3559 3560 ASSERT_RTNL(); 3561 3562 idev = addrconf_add_dev(dev); 3563 if (IS_ERR(idev)) 3564 return; 3565 3566 /* Generate the IPv6 link-local address using addrconf_addr_gen(), 3567 * unless we have an IPv4 GRE device not bound to an IP address and 3568 * which is in EUI64 mode (as __ipv6_isatap_ifid() would fail in this 3569 * case). Such devices fall back to add_v4_addrs() instead. 3570 */ 3571 if (!(*(__be32 *)dev->dev_addr == 0 && 3572 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)) { 3573 addrconf_addr_gen(idev, true); 3574 return; 3575 } 3576 3577 add_v4_addrs(idev); 3578 } 3579 #endif 3580 3581 static void addrconf_init_auto_addrs(struct net_device *dev) 3582 { 3583 switch (dev->type) { 3584 #if IS_ENABLED(CONFIG_IPV6_SIT) 3585 case ARPHRD_SIT: 3586 addrconf_sit_config(dev); 3587 break; 3588 #endif 3589 #if IS_ENABLED(CONFIG_NET_IPGRE) 3590 case ARPHRD_IPGRE: 3591 addrconf_gre_config(dev); 3592 break; 3593 #endif 3594 case ARPHRD_LOOPBACK: 3595 init_loopback(dev); 3596 break; 3597 3598 default: 3599 addrconf_dev_config(dev); 3600 break; 3601 } 3602 } 3603 3604 static int fixup_permanent_addr(struct net *net, 3605 struct inet6_dev *idev, 3606 struct inet6_ifaddr *ifp) 3607 { 3608 /* !fib6_node means the host route was removed from the 3609 * FIB, for example, if 'lo' device is taken down. In that 3610 * case regenerate the host route. 3611 */ 3612 if (!ifp->rt || !ifp->rt->fib6_node) { 3613 struct fib6_info *f6i, *prev; 3614 3615 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false, 3616 GFP_KERNEL, NULL); 3617 if (IS_ERR(f6i)) 3618 return PTR_ERR(f6i); 3619 3620 /* ifp->rt can be accessed outside of rtnl */ 3621 spin_lock_bh(&ifp->lock); 3622 prev = ifp->rt; 3623 ifp->rt = f6i; 3624 spin_unlock_bh(&ifp->lock); 3625 3626 fib6_info_release(prev); 3627 } 3628 3629 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) { 3630 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3631 ifp->rt_priority, idev->dev, 0, 0, 3632 GFP_KERNEL); 3633 } 3634 3635 if (ifp->state == INET6_IFADDR_STATE_PREDAD) 3636 addrconf_dad_start(ifp); 3637 3638 return 0; 3639 } 3640 3641 static void addrconf_permanent_addr(struct net *net, struct net_device *dev) 3642 { 3643 struct inet6_ifaddr *ifp; 3644 LIST_HEAD(tmp_addr_list); 3645 struct inet6_dev *idev; 3646 3647 /* Mutual exclusion with other if_list_aux users. */ 3648 ASSERT_RTNL(); 3649 3650 idev = __in6_dev_get(dev); 3651 if (!idev) 3652 return; 3653 3654 write_lock_bh(&idev->lock); 3655 list_for_each_entry(ifp, &idev->addr_list, if_list) { 3656 if (ifp->flags & IFA_F_PERMANENT) 3657 list_add_tail(&ifp->if_list_aux, &tmp_addr_list); 3658 } 3659 write_unlock_bh(&idev->lock); 3660 3661 while (!list_empty(&tmp_addr_list)) { 3662 ifp = list_first_entry(&tmp_addr_list, 3663 struct inet6_ifaddr, if_list_aux); 3664 list_del(&ifp->if_list_aux); 3665 3666 if (fixup_permanent_addr(net, idev, ifp) < 0) { 3667 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n", 3668 idev->dev->name, &ifp->addr); 3669 in6_ifa_hold(ifp); 3670 ipv6_del_addr(ifp); 3671 } 3672 } 3673 } 3674 3675 static int addrconf_notify(struct notifier_block *this, unsigned long event, 3676 void *ptr) 3677 { 3678 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 3679 struct netdev_notifier_change_info *change_info; 3680 struct netdev_notifier_changeupper_info *info; 3681 struct inet6_dev *idev = __in6_dev_get(dev); 3682 struct net *net = dev_net(dev); 3683 int run_pending = 0; 3684 int err; 3685 3686 switch (event) { 3687 case NETDEV_REGISTER: 3688 if (!idev && dev->mtu >= IPV6_MIN_MTU) { 3689 idev = ipv6_add_dev(dev); 3690 if (IS_ERR(idev)) 3691 return notifier_from_errno(PTR_ERR(idev)); 3692 } 3693 break; 3694 3695 case NETDEV_CHANGEMTU: 3696 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */ 3697 if (dev->mtu < IPV6_MIN_MTU) { 3698 addrconf_ifdown(dev, dev != net->loopback_dev); 3699 break; 3700 } 3701 3702 if (idev) { 3703 rt6_mtu_change(dev, dev->mtu); 3704 WRITE_ONCE(idev->cnf.mtu6, dev->mtu); 3705 break; 3706 } 3707 3708 /* allocate new idev */ 3709 idev = ipv6_add_dev(dev); 3710 if (IS_ERR(idev)) 3711 break; 3712 3713 /* device is still not ready */ 3714 if (!(idev->if_flags & IF_READY)) 3715 break; 3716 3717 run_pending = 1; 3718 fallthrough; 3719 case NETDEV_UP: 3720 case NETDEV_CHANGE: 3721 if (idev && idev->cnf.disable_ipv6) 3722 break; 3723 3724 if (dev->priv_flags & IFF_NO_ADDRCONF) { 3725 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) && 3726 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST) 3727 ipv6_mc_up(idev); 3728 break; 3729 } 3730 3731 if (event == NETDEV_UP) { 3732 /* restore routes for permanent addresses */ 3733 addrconf_permanent_addr(net, dev); 3734 3735 if (!addrconf_link_ready(dev)) { 3736 /* device is not ready yet. */ 3737 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n", 3738 dev->name); 3739 break; 3740 } 3741 3742 if (!idev && dev->mtu >= IPV6_MIN_MTU) 3743 idev = ipv6_add_dev(dev); 3744 3745 if (!IS_ERR_OR_NULL(idev)) { 3746 idev->if_flags |= IF_READY; 3747 run_pending = 1; 3748 } 3749 } else if (event == NETDEV_CHANGE) { 3750 if (!addrconf_link_ready(dev)) { 3751 /* device is still not ready. */ 3752 rt6_sync_down_dev(dev, event); 3753 break; 3754 } 3755 3756 if (!IS_ERR_OR_NULL(idev)) { 3757 if (idev->if_flags & IF_READY) { 3758 /* device is already configured - 3759 * but resend MLD reports, we might 3760 * have roamed and need to update 3761 * multicast snooping switches 3762 */ 3763 ipv6_mc_up(idev); 3764 change_info = ptr; 3765 if (change_info->flags_changed & IFF_NOARP) 3766 addrconf_dad_run(idev, true); 3767 rt6_sync_up(dev, RTNH_F_LINKDOWN); 3768 break; 3769 } 3770 idev->if_flags |= IF_READY; 3771 } 3772 3773 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n", 3774 dev->name); 3775 3776 run_pending = 1; 3777 } 3778 3779 addrconf_init_auto_addrs(dev); 3780 3781 if (!IS_ERR_OR_NULL(idev)) { 3782 if (run_pending) 3783 addrconf_dad_run(idev, false); 3784 3785 /* Device has an address by now */ 3786 rt6_sync_up(dev, RTNH_F_DEAD); 3787 3788 /* 3789 * If the MTU changed during the interface down, 3790 * when the interface up, the changed MTU must be 3791 * reflected in the idev as well as routers. 3792 */ 3793 if (idev->cnf.mtu6 != dev->mtu && 3794 dev->mtu >= IPV6_MIN_MTU) { 3795 rt6_mtu_change(dev, dev->mtu); 3796 WRITE_ONCE(idev->cnf.mtu6, dev->mtu); 3797 } 3798 WRITE_ONCE(idev->tstamp, jiffies); 3799 inet6_ifinfo_notify(RTM_NEWLINK, idev); 3800 3801 /* 3802 * If the changed mtu during down is lower than 3803 * IPV6_MIN_MTU stop IPv6 on this interface. 3804 */ 3805 if (dev->mtu < IPV6_MIN_MTU) 3806 addrconf_ifdown(dev, dev != net->loopback_dev); 3807 } 3808 break; 3809 3810 case NETDEV_DOWN: 3811 case NETDEV_UNREGISTER: 3812 /* 3813 * Remove all addresses from this interface. 3814 */ 3815 addrconf_ifdown(dev, event != NETDEV_DOWN); 3816 break; 3817 3818 case NETDEV_CHANGENAME: 3819 if (idev) { 3820 snmp6_unregister_dev(idev); 3821 addrconf_sysctl_unregister(idev); 3822 err = addrconf_sysctl_register(idev); 3823 if (err) 3824 return notifier_from_errno(err); 3825 err = snmp6_register_dev(idev); 3826 if (err) { 3827 addrconf_sysctl_unregister(idev); 3828 return notifier_from_errno(err); 3829 } 3830 } 3831 break; 3832 3833 case NETDEV_PRE_TYPE_CHANGE: 3834 case NETDEV_POST_TYPE_CHANGE: 3835 if (idev) 3836 addrconf_type_change(dev, event); 3837 break; 3838 3839 case NETDEV_CHANGEUPPER: 3840 info = ptr; 3841 3842 /* flush all routes if dev is linked to or unlinked from 3843 * an L3 master device (e.g., VRF) 3844 */ 3845 if (info->upper_dev && netif_is_l3_master(info->upper_dev)) 3846 addrconf_ifdown(dev, false); 3847 } 3848 3849 return NOTIFY_OK; 3850 } 3851 3852 /* 3853 * addrconf module should be notified of a device going up 3854 */ 3855 static struct notifier_block ipv6_dev_notf = { 3856 .notifier_call = addrconf_notify, 3857 .priority = ADDRCONF_NOTIFY_PRIORITY, 3858 }; 3859 3860 static void addrconf_type_change(struct net_device *dev, unsigned long event) 3861 { 3862 struct inet6_dev *idev; 3863 ASSERT_RTNL(); 3864 3865 idev = __in6_dev_get(dev); 3866 3867 if (event == NETDEV_POST_TYPE_CHANGE) 3868 ipv6_mc_remap(idev); 3869 else if (event == NETDEV_PRE_TYPE_CHANGE) 3870 ipv6_mc_unmap(idev); 3871 } 3872 3873 static bool addr_is_local(const struct in6_addr *addr) 3874 { 3875 return ipv6_addr_type(addr) & 3876 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK); 3877 } 3878 3879 static int addrconf_ifdown(struct net_device *dev, bool unregister) 3880 { 3881 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN; 3882 struct net *net = dev_net(dev); 3883 struct inet6_dev *idev; 3884 struct inet6_ifaddr *ifa; 3885 LIST_HEAD(tmp_addr_list); 3886 bool keep_addr = false; 3887 bool was_ready; 3888 int state, i; 3889 3890 ASSERT_RTNL(); 3891 3892 rt6_disable_ip(dev, event); 3893 3894 idev = __in6_dev_get(dev); 3895 if (!idev) 3896 return -ENODEV; 3897 3898 /* 3899 * Step 1: remove reference to ipv6 device from parent device. 3900 * Do not dev_put! 3901 */ 3902 if (unregister) { 3903 WRITE_ONCE(idev->dead, 1); 3904 3905 /* protected by rtnl_lock */ 3906 RCU_INIT_POINTER(dev->ip6_ptr, NULL); 3907 3908 /* Step 1.5: remove snmp6 entry */ 3909 snmp6_unregister_dev(idev); 3910 3911 } 3912 3913 /* combine the user config with event to determine if permanent 3914 * addresses are to be removed from address hash table 3915 */ 3916 if (!unregister && !idev->cnf.disable_ipv6) { 3917 /* aggregate the system setting and interface setting */ 3918 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down); 3919 3920 if (!_keep_addr) 3921 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down); 3922 3923 keep_addr = (_keep_addr > 0); 3924 } 3925 3926 /* Step 2: clear hash table */ 3927 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 3928 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i]; 3929 3930 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 3931 restart: 3932 hlist_for_each_entry_rcu(ifa, h, addr_lst) { 3933 if (ifa->idev == idev) { 3934 addrconf_del_dad_work(ifa); 3935 /* combined flag + permanent flag decide if 3936 * address is retained on a down event 3937 */ 3938 if (!keep_addr || 3939 !(ifa->flags & IFA_F_PERMANENT) || 3940 addr_is_local(&ifa->addr)) { 3941 hlist_del_init_rcu(&ifa->addr_lst); 3942 goto restart; 3943 } 3944 } 3945 } 3946 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 3947 } 3948 3949 write_lock_bh(&idev->lock); 3950 3951 addrconf_del_rs_timer(idev); 3952 3953 /* Step 2: clear flags for stateless addrconf, repeated down 3954 * detection 3955 */ 3956 was_ready = idev->if_flags & IF_READY; 3957 if (!unregister) 3958 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY); 3959 3960 /* Step 3: clear tempaddr list */ 3961 while (!list_empty(&idev->tempaddr_list)) { 3962 ifa = list_first_entry(&idev->tempaddr_list, 3963 struct inet6_ifaddr, tmp_list); 3964 list_del(&ifa->tmp_list); 3965 write_unlock_bh(&idev->lock); 3966 spin_lock_bh(&ifa->lock); 3967 3968 if (ifa->ifpub) { 3969 in6_ifa_put(ifa->ifpub); 3970 ifa->ifpub = NULL; 3971 } 3972 spin_unlock_bh(&ifa->lock); 3973 in6_ifa_put(ifa); 3974 write_lock_bh(&idev->lock); 3975 } 3976 3977 list_for_each_entry(ifa, &idev->addr_list, if_list) 3978 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 3979 write_unlock_bh(&idev->lock); 3980 3981 while (!list_empty(&tmp_addr_list)) { 3982 struct fib6_info *rt = NULL; 3983 bool keep; 3984 3985 ifa = list_first_entry(&tmp_addr_list, 3986 struct inet6_ifaddr, if_list_aux); 3987 list_del(&ifa->if_list_aux); 3988 3989 addrconf_del_dad_work(ifa); 3990 3991 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) && 3992 !addr_is_local(&ifa->addr); 3993 3994 spin_lock_bh(&ifa->lock); 3995 3996 if (keep) { 3997 /* set state to skip the notifier below */ 3998 state = INET6_IFADDR_STATE_DEAD; 3999 ifa->state = INET6_IFADDR_STATE_PREDAD; 4000 if (!(ifa->flags & IFA_F_NODAD)) 4001 ifa->flags |= IFA_F_TENTATIVE; 4002 4003 rt = ifa->rt; 4004 ifa->rt = NULL; 4005 } else { 4006 state = ifa->state; 4007 ifa->state = INET6_IFADDR_STATE_DEAD; 4008 } 4009 4010 spin_unlock_bh(&ifa->lock); 4011 4012 if (rt) 4013 ip6_del_rt(net, rt, false); 4014 4015 if (state != INET6_IFADDR_STATE_DEAD) { 4016 __ipv6_ifa_notify(RTM_DELADDR, ifa); 4017 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa); 4018 } else { 4019 if (idev->cnf.forwarding) 4020 addrconf_leave_anycast(ifa); 4021 addrconf_leave_solict(ifa->idev, &ifa->addr); 4022 } 4023 4024 if (!keep) { 4025 write_lock_bh(&idev->lock); 4026 list_del_rcu(&ifa->if_list); 4027 write_unlock_bh(&idev->lock); 4028 in6_ifa_put(ifa); 4029 } 4030 } 4031 4032 /* Step 5: Discard anycast and multicast list */ 4033 if (unregister) { 4034 ipv6_ac_destroy_dev(idev); 4035 ipv6_mc_destroy_dev(idev); 4036 } else if (was_ready) { 4037 ipv6_mc_down(idev); 4038 } 4039 4040 WRITE_ONCE(idev->tstamp, jiffies); 4041 idev->ra_mtu = 0; 4042 4043 /* Last: Shot the device (if unregistered) */ 4044 if (unregister) { 4045 struct neigh_table *tbl = nd_table(net); 4046 4047 addrconf_sysctl_unregister(idev); 4048 neigh_parms_release(tbl, idev->nd_parms); 4049 neigh_ifdown(tbl, dev); 4050 in6_dev_put(idev); 4051 } 4052 return 0; 4053 } 4054 4055 static void addrconf_rs_timer(struct timer_list *t) 4056 { 4057 struct inet6_dev *idev = timer_container_of(idev, t, rs_timer); 4058 struct net_device *dev = idev->dev; 4059 struct in6_addr lladdr; 4060 int rtr_solicits; 4061 4062 write_lock(&idev->lock); 4063 if (idev->dead || !(idev->if_flags & IF_READY)) 4064 goto out; 4065 4066 if (!ipv6_accept_ra(idev)) 4067 goto out; 4068 4069 /* Announcement received after solicitation was sent */ 4070 if (idev->if_flags & IF_RA_RCVD) 4071 goto out; 4072 4073 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits); 4074 4075 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) { 4076 write_unlock(&idev->lock); 4077 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 4078 ndisc_send_rs(dev, &lladdr, 4079 &in6addr_linklocal_allrouters); 4080 else 4081 goto put; 4082 4083 write_lock(&idev->lock); 4084 idev->rs_interval = rfc3315_s14_backoff_update( 4085 idev->rs_interval, 4086 READ_ONCE(idev->cnf.rtr_solicit_max_interval)); 4087 /* The wait after the last probe can be shorter */ 4088 addrconf_mod_rs_timer(idev, (idev->rs_probes == 4089 READ_ONCE(idev->cnf.rtr_solicits)) ? 4090 READ_ONCE(idev->cnf.rtr_solicit_delay) : 4091 idev->rs_interval); 4092 } else { 4093 /* 4094 * Note: we do not support deprecated "all on-link" 4095 * assumption any longer. 4096 */ 4097 pr_debug("%s: no IPv6 routers present\n", idev->dev->name); 4098 } 4099 4100 out: 4101 write_unlock(&idev->lock); 4102 put: 4103 in6_dev_put(idev); 4104 } 4105 4106 /* 4107 * Duplicate Address Detection 4108 */ 4109 static void addrconf_dad_kick(struct inet6_ifaddr *ifp) 4110 { 4111 struct inet6_dev *idev = ifp->idev; 4112 unsigned long rand_num; 4113 u64 nonce; 4114 4115 if (ifp->flags & IFA_F_OPTIMISTIC) 4116 rand_num = 0; 4117 else 4118 rand_num = get_random_u32_below( 4119 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1); 4120 4121 nonce = 0; 4122 if (READ_ONCE(idev->cnf.enhanced_dad) || 4123 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) { 4124 do 4125 get_random_bytes(&nonce, 6); 4126 while (nonce == 0); 4127 } 4128 ifp->dad_nonce = nonce; 4129 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits); 4130 addrconf_mod_dad_work(ifp, rand_num); 4131 } 4132 4133 static void addrconf_dad_begin(struct inet6_ifaddr *ifp) 4134 { 4135 struct inet6_dev *idev = ifp->idev; 4136 struct net_device *dev = idev->dev; 4137 bool bump_id, notify = false; 4138 struct net *net; 4139 4140 addrconf_join_solict(dev, &ifp->addr); 4141 4142 read_lock_bh(&idev->lock); 4143 spin_lock(&ifp->lock); 4144 if (ifp->state == INET6_IFADDR_STATE_DEAD) 4145 goto out; 4146 4147 net = dev_net(dev); 4148 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) || 4149 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 && 4150 READ_ONCE(idev->cnf.accept_dad) < 1) || 4151 !(ifp->flags&IFA_F_TENTATIVE) || 4152 ifp->flags & IFA_F_NODAD) { 4153 bool send_na = false; 4154 4155 if (ifp->flags & IFA_F_TENTATIVE && 4156 !(ifp->flags & IFA_F_OPTIMISTIC)) 4157 send_na = true; 4158 bump_id = ifp->flags & IFA_F_TENTATIVE; 4159 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4160 spin_unlock(&ifp->lock); 4161 read_unlock_bh(&idev->lock); 4162 4163 addrconf_dad_completed(ifp, bump_id, send_na); 4164 return; 4165 } 4166 4167 if (!(idev->if_flags & IF_READY)) { 4168 spin_unlock(&ifp->lock); 4169 read_unlock_bh(&idev->lock); 4170 /* 4171 * If the device is not ready: 4172 * - keep it tentative if it is a permanent address. 4173 * - otherwise, kill it. 4174 */ 4175 in6_ifa_hold(ifp); 4176 addrconf_dad_stop(ifp, 0); 4177 return; 4178 } 4179 4180 /* 4181 * Optimistic nodes can start receiving 4182 * Frames right away 4183 */ 4184 if (ifp->flags & IFA_F_OPTIMISTIC) { 4185 ip6_ins_rt(net, ifp->rt); 4186 if (ipv6_use_optimistic_addr(net, idev)) { 4187 /* Because optimistic nodes can use this address, 4188 * notify listeners. If DAD fails, RTM_DELADDR is sent. 4189 */ 4190 notify = true; 4191 } 4192 } 4193 4194 addrconf_dad_kick(ifp); 4195 out: 4196 spin_unlock(&ifp->lock); 4197 read_unlock_bh(&idev->lock); 4198 if (notify) 4199 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4200 } 4201 4202 static void addrconf_dad_start(struct inet6_ifaddr *ifp) 4203 { 4204 bool begin_dad = false; 4205 4206 spin_lock_bh(&ifp->lock); 4207 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 4208 ifp->state = INET6_IFADDR_STATE_PREDAD; 4209 begin_dad = true; 4210 } 4211 spin_unlock_bh(&ifp->lock); 4212 4213 if (begin_dad) 4214 addrconf_mod_dad_work(ifp, 0); 4215 } 4216 4217 static void addrconf_dad_work(struct work_struct *w) 4218 { 4219 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w), 4220 struct inet6_ifaddr, 4221 dad_work); 4222 struct inet6_dev *idev = ifp->idev; 4223 bool bump_id, disable_ipv6 = false; 4224 struct in6_addr mcaddr; 4225 struct net *net; 4226 4227 enum { 4228 DAD_PROCESS, 4229 DAD_BEGIN, 4230 DAD_ABORT, 4231 } action = DAD_PROCESS; 4232 4233 net = dev_net(idev->dev); 4234 4235 rtnl_net_lock(net); 4236 4237 spin_lock_bh(&ifp->lock); 4238 if (ifp->state == INET6_IFADDR_STATE_PREDAD) { 4239 action = DAD_BEGIN; 4240 ifp->state = INET6_IFADDR_STATE_DAD; 4241 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) { 4242 action = DAD_ABORT; 4243 ifp->state = INET6_IFADDR_STATE_POSTDAD; 4244 4245 if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 || 4246 READ_ONCE(idev->cnf.accept_dad) > 1) && 4247 !idev->cnf.disable_ipv6 && 4248 !(ifp->flags & IFA_F_STABLE_PRIVACY)) { 4249 struct in6_addr addr; 4250 4251 addr.s6_addr32[0] = htonl(0xfe800000); 4252 addr.s6_addr32[1] = 0; 4253 4254 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) && 4255 ipv6_addr_equal(&ifp->addr, &addr)) { 4256 /* DAD failed for link-local based on MAC */ 4257 WRITE_ONCE(idev->cnf.disable_ipv6, 1); 4258 4259 pr_info("%s: IPv6 being disabled!\n", 4260 ifp->idev->dev->name); 4261 disable_ipv6 = true; 4262 } 4263 } 4264 } 4265 spin_unlock_bh(&ifp->lock); 4266 4267 if (action == DAD_BEGIN) { 4268 addrconf_dad_begin(ifp); 4269 goto out; 4270 } else if (action == DAD_ABORT) { 4271 in6_ifa_hold(ifp); 4272 addrconf_dad_stop(ifp, 1); 4273 if (disable_ipv6) 4274 addrconf_ifdown(idev->dev, false); 4275 goto out; 4276 } 4277 4278 if (!ifp->dad_probes && addrconf_dad_end(ifp)) 4279 goto out; 4280 4281 write_lock_bh(&idev->lock); 4282 if (idev->dead || !(idev->if_flags & IF_READY)) { 4283 write_unlock_bh(&idev->lock); 4284 goto out; 4285 } 4286 4287 spin_lock(&ifp->lock); 4288 if (ifp->state == INET6_IFADDR_STATE_DEAD) { 4289 spin_unlock(&ifp->lock); 4290 write_unlock_bh(&idev->lock); 4291 goto out; 4292 } 4293 4294 if (ifp->dad_probes == 0) { 4295 bool send_na = false; 4296 4297 /* 4298 * DAD was successful 4299 */ 4300 4301 if (ifp->flags & IFA_F_TENTATIVE && 4302 !(ifp->flags & IFA_F_OPTIMISTIC)) 4303 send_na = true; 4304 bump_id = ifp->flags & IFA_F_TENTATIVE; 4305 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4306 spin_unlock(&ifp->lock); 4307 write_unlock_bh(&idev->lock); 4308 4309 addrconf_dad_completed(ifp, bump_id, send_na); 4310 4311 goto out; 4312 } 4313 4314 ifp->dad_probes--; 4315 addrconf_mod_dad_work(ifp, 4316 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), 4317 HZ/100)); 4318 spin_unlock(&ifp->lock); 4319 write_unlock_bh(&idev->lock); 4320 4321 /* send a neighbour solicitation for our addr */ 4322 addrconf_addr_solict_mult(&ifp->addr, &mcaddr); 4323 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any, 4324 ifp->dad_nonce); 4325 out: 4326 in6_ifa_put(ifp); 4327 rtnl_net_unlock(net); 4328 } 4329 4330 /* ifp->idev must be at least read locked */ 4331 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp) 4332 { 4333 struct inet6_ifaddr *ifpiter; 4334 struct inet6_dev *idev = ifp->idev; 4335 4336 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) { 4337 if (ifpiter->scope > IFA_LINK) 4338 break; 4339 if (ifp != ifpiter && ifpiter->scope == IFA_LINK && 4340 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE| 4341 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) == 4342 IFA_F_PERMANENT) 4343 return false; 4344 } 4345 return true; 4346 } 4347 4348 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 4349 bool send_na) 4350 { 4351 struct net_device *dev = ifp->idev->dev; 4352 struct in6_addr lladdr; 4353 bool send_rs, send_mld; 4354 4355 addrconf_del_dad_work(ifp); 4356 4357 /* 4358 * Configure the address for reception. Now it is valid. 4359 */ 4360 4361 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4362 4363 /* If added prefix is link local and we are prepared to process 4364 router advertisements, start sending router solicitations. 4365 */ 4366 4367 read_lock_bh(&ifp->idev->lock); 4368 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp); 4369 send_rs = send_mld && 4370 ipv6_accept_ra(ifp->idev) && 4371 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 && 4372 (dev->flags & IFF_LOOPBACK) == 0 && 4373 (dev->type != ARPHRD_TUNNEL) && 4374 !netif_is_team_port(dev); 4375 read_unlock_bh(&ifp->idev->lock); 4376 4377 /* While dad is in progress mld report's source address is in6_addrany. 4378 * Resend with proper ll now. 4379 */ 4380 if (send_mld) 4381 ipv6_mc_dad_complete(ifp->idev); 4382 4383 /* send unsolicited NA if enabled */ 4384 if (send_na && 4385 (READ_ONCE(ifp->idev->cnf.ndisc_notify) || 4386 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) { 4387 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr, 4388 /*router=*/ !!ifp->idev->cnf.forwarding, 4389 /*solicited=*/ false, /*override=*/ true, 4390 /*inc_opt=*/ true); 4391 } 4392 4393 if (send_rs) { 4394 /* 4395 * If a host as already performed a random delay 4396 * [...] as part of DAD [...] there is no need 4397 * to delay again before sending the first RS 4398 */ 4399 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 4400 return; 4401 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters); 4402 4403 write_lock_bh(&ifp->idev->lock); 4404 spin_lock(&ifp->lock); 4405 ifp->idev->rs_interval = rfc3315_s14_backoff_init( 4406 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval)); 4407 ifp->idev->rs_probes = 1; 4408 ifp->idev->if_flags |= IF_RS_SENT; 4409 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval); 4410 spin_unlock(&ifp->lock); 4411 write_unlock_bh(&ifp->idev->lock); 4412 } 4413 4414 if (bump_id) 4415 rt_genid_bump_ipv6(dev_net(dev)); 4416 4417 /* Make sure that a new temporary address will be created 4418 * before this temporary address becomes deprecated. 4419 */ 4420 if (ifp->flags & IFA_F_TEMPORARY) 4421 addrconf_verify_rtnl(dev_net(dev)); 4422 } 4423 4424 static void addrconf_dad_run(struct inet6_dev *idev, bool restart) 4425 { 4426 struct inet6_ifaddr *ifp; 4427 4428 read_lock_bh(&idev->lock); 4429 list_for_each_entry(ifp, &idev->addr_list, if_list) { 4430 spin_lock(&ifp->lock); 4431 if ((ifp->flags & IFA_F_TENTATIVE && 4432 ifp->state == INET6_IFADDR_STATE_DAD) || restart) { 4433 if (restart) 4434 ifp->state = INET6_IFADDR_STATE_PREDAD; 4435 addrconf_dad_kick(ifp); 4436 } 4437 spin_unlock(&ifp->lock); 4438 } 4439 read_unlock_bh(&idev->lock); 4440 } 4441 4442 #ifdef CONFIG_PROC_FS 4443 struct if6_iter_state { 4444 struct seq_net_private p; 4445 int bucket; 4446 int offset; 4447 }; 4448 4449 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos) 4450 { 4451 struct if6_iter_state *state = seq->private; 4452 struct net *net = seq_file_net(seq); 4453 struct inet6_ifaddr *ifa = NULL; 4454 int p = 0; 4455 4456 /* initial bucket if pos is 0 */ 4457 if (pos == 0) { 4458 state->bucket = 0; 4459 state->offset = 0; 4460 } 4461 4462 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) { 4463 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket], 4464 addr_lst) { 4465 /* sync with offset */ 4466 if (p < state->offset) { 4467 p++; 4468 continue; 4469 } 4470 return ifa; 4471 } 4472 4473 /* prepare for next bucket */ 4474 state->offset = 0; 4475 p = 0; 4476 } 4477 return NULL; 4478 } 4479 4480 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, 4481 struct inet6_ifaddr *ifa) 4482 { 4483 struct if6_iter_state *state = seq->private; 4484 struct net *net = seq_file_net(seq); 4485 4486 hlist_for_each_entry_continue_rcu(ifa, addr_lst) { 4487 state->offset++; 4488 return ifa; 4489 } 4490 4491 state->offset = 0; 4492 while (++state->bucket < IN6_ADDR_HSIZE) { 4493 hlist_for_each_entry_rcu(ifa, 4494 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) { 4495 return ifa; 4496 } 4497 } 4498 4499 return NULL; 4500 } 4501 4502 static void *if6_seq_start(struct seq_file *seq, loff_t *pos) 4503 __acquires(rcu) 4504 { 4505 rcu_read_lock(); 4506 return if6_get_first(seq, *pos); 4507 } 4508 4509 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 4510 { 4511 struct inet6_ifaddr *ifa; 4512 4513 ifa = if6_get_next(seq, v); 4514 ++*pos; 4515 return ifa; 4516 } 4517 4518 static void if6_seq_stop(struct seq_file *seq, void *v) 4519 __releases(rcu) 4520 { 4521 rcu_read_unlock(); 4522 } 4523 4524 static int if6_seq_show(struct seq_file *seq, void *v) 4525 { 4526 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v; 4527 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n", 4528 &ifp->addr, 4529 ifp->idev->dev->ifindex, 4530 ifp->prefix_len, 4531 ifp->scope, 4532 (u8) ifp->flags, 4533 ifp->idev->dev->name); 4534 return 0; 4535 } 4536 4537 static const struct seq_operations if6_seq_ops = { 4538 .start = if6_seq_start, 4539 .next = if6_seq_next, 4540 .show = if6_seq_show, 4541 .stop = if6_seq_stop, 4542 }; 4543 4544 static int __net_init if6_proc_net_init(struct net *net) 4545 { 4546 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops, 4547 sizeof(struct if6_iter_state))) 4548 return -ENOMEM; 4549 return 0; 4550 } 4551 4552 static void __net_exit if6_proc_net_exit(struct net *net) 4553 { 4554 remove_proc_entry("if_inet6", net->proc_net); 4555 } 4556 4557 static struct pernet_operations if6_proc_net_ops = { 4558 .init = if6_proc_net_init, 4559 .exit = if6_proc_net_exit, 4560 }; 4561 4562 int __init if6_proc_init(void) 4563 { 4564 return register_pernet_subsys(&if6_proc_net_ops); 4565 } 4566 4567 void if6_proc_exit(void) 4568 { 4569 unregister_pernet_subsys(&if6_proc_net_ops); 4570 } 4571 #endif /* CONFIG_PROC_FS */ 4572 4573 #if IS_ENABLED(CONFIG_IPV6_MIP6) 4574 /* Check if address is a home address configured on any interface. */ 4575 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr) 4576 { 4577 unsigned int hash = inet6_addr_hash(net, addr); 4578 struct inet6_ifaddr *ifp = NULL; 4579 int ret = 0; 4580 4581 rcu_read_lock(); 4582 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4583 if (ipv6_addr_equal(&ifp->addr, addr) && 4584 (ifp->flags & IFA_F_HOMEADDRESS)) { 4585 ret = 1; 4586 break; 4587 } 4588 } 4589 rcu_read_unlock(); 4590 return ret; 4591 } 4592 #endif 4593 4594 /* RFC6554 has some algorithm to avoid loops in segment routing by 4595 * checking if the segments contains any of a local interface address. 4596 * 4597 * Quote: 4598 * 4599 * To detect loops in the SRH, a router MUST determine if the SRH 4600 * includes multiple addresses assigned to any interface on that router. 4601 * If such addresses appear more than once and are separated by at least 4602 * one address not assigned to that router. 4603 */ 4604 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs, 4605 unsigned char nsegs) 4606 { 4607 const struct in6_addr *addr; 4608 int i, ret = 0, found = 0; 4609 struct inet6_ifaddr *ifp; 4610 bool separated = false; 4611 unsigned int hash; 4612 bool hash_found; 4613 4614 rcu_read_lock(); 4615 for (i = 0; i < nsegs; i++) { 4616 addr = &segs[i]; 4617 hash = inet6_addr_hash(net, addr); 4618 4619 hash_found = false; 4620 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4621 4622 if (ipv6_addr_equal(&ifp->addr, addr)) { 4623 hash_found = true; 4624 break; 4625 } 4626 } 4627 4628 if (hash_found) { 4629 if (found > 1 && separated) { 4630 ret = 1; 4631 break; 4632 } 4633 4634 separated = false; 4635 found++; 4636 } else { 4637 separated = true; 4638 } 4639 } 4640 rcu_read_unlock(); 4641 4642 return ret; 4643 } 4644 4645 /* 4646 * Periodic address status verification 4647 */ 4648 4649 static void addrconf_verify_rtnl(struct net *net) 4650 { 4651 unsigned long now, next, next_sec, next_sched; 4652 struct inet6_ifaddr *ifp; 4653 int i; 4654 4655 ASSERT_RTNL(); 4656 4657 rcu_read_lock_bh(); 4658 now = jiffies; 4659 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY); 4660 4661 cancel_delayed_work(&net->ipv6.addr_chk_work); 4662 4663 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 4664 restart: 4665 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) { 4666 unsigned long age; 4667 4668 /* When setting preferred_lft to a value not zero or 4669 * infinity, while valid_lft is infinity 4670 * IFA_F_PERMANENT has a non-infinity life time. 4671 */ 4672 if ((ifp->flags & IFA_F_PERMANENT) && 4673 (ifp->prefered_lft == INFINITY_LIFE_TIME)) 4674 continue; 4675 4676 spin_lock(&ifp->lock); 4677 /* We try to batch several events at once. */ 4678 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 4679 4680 if ((ifp->flags&IFA_F_TEMPORARY) && 4681 !(ifp->flags&IFA_F_TENTATIVE) && 4682 ifp->prefered_lft != INFINITY_LIFE_TIME && 4683 !ifp->regen_count && ifp->ifpub) { 4684 /* This is a non-regenerated temporary addr. */ 4685 4686 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev); 4687 4688 if (age + regen_advance >= ifp->prefered_lft) { 4689 struct inet6_ifaddr *ifpub = ifp->ifpub; 4690 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4691 next = ifp->tstamp + ifp->prefered_lft * HZ; 4692 4693 ifp->regen_count++; 4694 in6_ifa_hold(ifp); 4695 in6_ifa_hold(ifpub); 4696 spin_unlock(&ifp->lock); 4697 4698 spin_lock(&ifpub->lock); 4699 ifpub->regen_count = 0; 4700 spin_unlock(&ifpub->lock); 4701 rcu_read_unlock_bh(); 4702 ipv6_create_tempaddr(ifpub, true); 4703 in6_ifa_put(ifpub); 4704 in6_ifa_put(ifp); 4705 rcu_read_lock_bh(); 4706 goto restart; 4707 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next)) 4708 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ; 4709 } 4710 4711 if (ifp->valid_lft != INFINITY_LIFE_TIME && 4712 age >= ifp->valid_lft) { 4713 spin_unlock(&ifp->lock); 4714 in6_ifa_hold(ifp); 4715 rcu_read_unlock_bh(); 4716 ipv6_del_addr(ifp); 4717 rcu_read_lock_bh(); 4718 goto restart; 4719 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) { 4720 spin_unlock(&ifp->lock); 4721 continue; 4722 } else if (age >= ifp->prefered_lft) { 4723 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */ 4724 int deprecate = 0; 4725 4726 if (!(ifp->flags&IFA_F_DEPRECATED)) { 4727 deprecate = 1; 4728 ifp->flags |= IFA_F_DEPRECATED; 4729 } 4730 4731 if ((ifp->valid_lft != INFINITY_LIFE_TIME) && 4732 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))) 4733 next = ifp->tstamp + ifp->valid_lft * HZ; 4734 4735 spin_unlock(&ifp->lock); 4736 4737 if (deprecate) { 4738 in6_ifa_hold(ifp); 4739 4740 ipv6_ifa_notify(0, ifp); 4741 in6_ifa_put(ifp); 4742 goto restart; 4743 } 4744 } else { 4745 /* ifp->prefered_lft <= ifp->valid_lft */ 4746 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4747 next = ifp->tstamp + ifp->prefered_lft * HZ; 4748 spin_unlock(&ifp->lock); 4749 } 4750 } 4751 } 4752 4753 next_sec = round_jiffies_up(next); 4754 next_sched = next; 4755 4756 /* If rounded timeout is accurate enough, accept it. */ 4757 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ)) 4758 next_sched = next_sec; 4759 4760 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */ 4761 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX)) 4762 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX; 4763 4764 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n", 4765 now, next, next_sec, next_sched); 4766 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now); 4767 rcu_read_unlock_bh(); 4768 } 4769 4770 static void addrconf_verify_work(struct work_struct *w) 4771 { 4772 struct net *net = container_of(to_delayed_work(w), struct net, 4773 ipv6.addr_chk_work); 4774 4775 rtnl_net_lock(net); 4776 addrconf_verify_rtnl(net); 4777 rtnl_net_unlock(net); 4778 } 4779 4780 static void addrconf_verify(struct net *net) 4781 { 4782 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0); 4783 } 4784 4785 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local, 4786 struct in6_addr **peer_pfx) 4787 { 4788 struct in6_addr *pfx = NULL; 4789 4790 *peer_pfx = NULL; 4791 4792 if (addr) 4793 pfx = nla_data(addr); 4794 4795 if (local) { 4796 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx))) 4797 *peer_pfx = pfx; 4798 pfx = nla_data(local); 4799 } 4800 4801 return pfx; 4802 } 4803 4804 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = { 4805 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) }, 4806 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) }, 4807 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) }, 4808 [IFA_FLAGS] = { .len = sizeof(u32) }, 4809 [IFA_RT_PRIORITY] = { .len = sizeof(u32) }, 4810 [IFA_TARGET_NETNSID] = { .type = NLA_S32 }, 4811 [IFA_PROTO] = { .type = NLA_U8 }, 4812 }; 4813 4814 static int 4815 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, 4816 struct netlink_ext_ack *extack) 4817 { 4818 struct net *net = sock_net(skb->sk); 4819 struct ifaddrmsg *ifm; 4820 struct nlattr *tb[IFA_MAX+1]; 4821 struct in6_addr *pfx, *peer_pfx; 4822 u32 ifa_flags; 4823 int err; 4824 4825 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 4826 ifa_ipv6_policy, extack); 4827 if (err < 0) 4828 return err; 4829 4830 ifm = nlmsg_data(nlh); 4831 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 4832 if (!pfx) 4833 return -EINVAL; 4834 4835 ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags); 4836 4837 /* We ignore other flags so far. */ 4838 ifa_flags &= IFA_F_MANAGETEMPADDR; 4839 4840 rtnl_net_lock(net); 4841 err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx, 4842 ifm->ifa_prefixlen, extack); 4843 rtnl_net_unlock(net); 4844 4845 return err; 4846 } 4847 4848 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp, 4849 unsigned long expires, u32 flags, 4850 bool modify_peer) 4851 { 4852 struct fib6_table *table; 4853 struct fib6_info *f6i; 4854 u32 prio; 4855 4856 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4857 ifp->prefix_len, 4858 ifp->idev->dev, 0, RTF_DEFAULT, true); 4859 if (!f6i) 4860 return -ENOENT; 4861 4862 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF; 4863 if (f6i->fib6_metric != prio) { 4864 /* delete old one */ 4865 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false); 4866 4867 /* add new one */ 4868 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4869 ifp->prefix_len, 4870 ifp->rt_priority, ifp->idev->dev, 4871 expires, flags, GFP_KERNEL); 4872 return 0; 4873 } 4874 if (f6i != net->ipv6.fib6_null_entry) { 4875 table = f6i->fib6_table; 4876 spin_lock_bh(&table->tb6_lock); 4877 4878 if (!(flags & RTF_EXPIRES)) { 4879 fib6_clean_expires(f6i); 4880 fib6_may_remove_gc_list(net, f6i); 4881 } else { 4882 fib6_set_expires(f6i, jiffies + clock_t_to_jiffies(expires)); 4883 fib6_add_gc_list(f6i); 4884 } 4885 4886 spin_unlock_bh(&table->tb6_lock); 4887 } 4888 fib6_info_release(f6i); 4889 4890 return 0; 4891 } 4892 4893 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp, 4894 struct ifa6_config *cfg, clock_t expires, 4895 u32 flags) 4896 { 4897 bool was_managetempaddr; 4898 bool new_peer = false; 4899 bool had_prefixroute; 4900 4901 ASSERT_RTNL_NET(net); 4902 4903 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && 4904 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64)) 4905 return -EINVAL; 4906 4907 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED) 4908 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC; 4909 4910 if (cfg->peer_pfx && 4911 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) { 4912 if (!ipv6_addr_any(&ifp->peer_addr)) 4913 cleanup_prefix_route(ifp, expires, true, true); 4914 new_peer = true; 4915 } 4916 4917 spin_lock_bh(&ifp->lock); 4918 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR; 4919 had_prefixroute = ifp->flags & IFA_F_PERMANENT && 4920 !(ifp->flags & IFA_F_NOPREFIXROUTE); 4921 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | 4922 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR | 4923 IFA_F_NOPREFIXROUTE); 4924 ifp->flags |= cfg->ifa_flags; 4925 WRITE_ONCE(ifp->tstamp, jiffies); 4926 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft); 4927 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft); 4928 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto); 4929 4930 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority) 4931 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority); 4932 4933 if (new_peer) 4934 ifp->peer_addr = *cfg->peer_pfx; 4935 4936 spin_unlock_bh(&ifp->lock); 4937 if (!(ifp->flags&IFA_F_TENTATIVE)) 4938 ipv6_ifa_notify(0, ifp); 4939 4940 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 4941 int rc = -ENOENT; 4942 4943 if (had_prefixroute) 4944 rc = modify_prefix_route(net, ifp, expires, flags, false); 4945 4946 /* prefix route could have been deleted; if so restore it */ 4947 if (rc == -ENOENT) { 4948 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 4949 ifp->rt_priority, ifp->idev->dev, 4950 expires, flags, GFP_KERNEL); 4951 } 4952 4953 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr)) 4954 rc = modify_prefix_route(net, ifp, expires, flags, true); 4955 4956 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) { 4957 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len, 4958 ifp->rt_priority, ifp->idev->dev, 4959 expires, flags, GFP_KERNEL); 4960 } 4961 } else if (had_prefixroute) { 4962 enum cleanup_prefix_rt_t action; 4963 unsigned long rt_expires; 4964 4965 write_lock_bh(&ifp->idev->lock); 4966 action = check_cleanup_prefix_route(ifp, &rt_expires); 4967 write_unlock_bh(&ifp->idev->lock); 4968 4969 if (action != CLEANUP_PREFIX_RT_NOP) { 4970 cleanup_prefix_route(ifp, rt_expires, 4971 action == CLEANUP_PREFIX_RT_DEL, false); 4972 } 4973 } 4974 4975 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) { 4976 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR)) 4977 delete_tempaddrs(ifp->idev, ifp); 4978 else 4979 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft, 4980 cfg->preferred_lft, !was_managetempaddr, 4981 jiffies); 4982 } 4983 4984 addrconf_verify_rtnl(net); 4985 4986 return 0; 4987 } 4988 4989 static int 4990 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, 4991 struct netlink_ext_ack *extack) 4992 { 4993 struct net *net = sock_net(skb->sk); 4994 struct nlattr *tb[IFA_MAX+1]; 4995 struct in6_addr *peer_pfx; 4996 struct inet6_ifaddr *ifa; 4997 struct net_device *dev; 4998 struct inet6_dev *idev; 4999 struct ifa6_config cfg; 5000 struct ifaddrmsg *ifm; 5001 unsigned long timeout; 5002 clock_t expires; 5003 u32 flags; 5004 int err; 5005 5006 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 5007 ifa_ipv6_policy, extack); 5008 if (err < 0) 5009 return err; 5010 5011 memset(&cfg, 0, sizeof(cfg)); 5012 5013 ifm = nlmsg_data(nlh); 5014 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 5015 if (!cfg.pfx) 5016 return -EINVAL; 5017 5018 cfg.peer_pfx = peer_pfx; 5019 cfg.plen = ifm->ifa_prefixlen; 5020 if (tb[IFA_RT_PRIORITY]) 5021 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]); 5022 5023 if (tb[IFA_PROTO]) 5024 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]); 5025 5026 cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags); 5027 5028 /* We ignore other flags so far. */ 5029 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | 5030 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE | 5031 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC; 5032 5033 cfg.ifa_flags |= IFA_F_PERMANENT; 5034 cfg.valid_lft = INFINITY_LIFE_TIME; 5035 cfg.preferred_lft = INFINITY_LIFE_TIME; 5036 expires = 0; 5037 flags = 0; 5038 5039 if (tb[IFA_CACHEINFO]) { 5040 struct ifa_cacheinfo *ci; 5041 5042 ci = nla_data(tb[IFA_CACHEINFO]); 5043 cfg.valid_lft = ci->ifa_valid; 5044 cfg.preferred_lft = ci->ifa_prefered; 5045 5046 if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) { 5047 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid"); 5048 return -EINVAL; 5049 } 5050 5051 timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ); 5052 if (addrconf_finite_timeout(timeout)) { 5053 cfg.ifa_flags &= ~IFA_F_PERMANENT; 5054 cfg.valid_lft = timeout; 5055 expires = jiffies_to_clock_t(timeout * HZ); 5056 flags = RTF_EXPIRES; 5057 } 5058 5059 timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ); 5060 if (addrconf_finite_timeout(timeout)) { 5061 if (timeout == 0) 5062 cfg.ifa_flags |= IFA_F_DEPRECATED; 5063 5064 cfg.preferred_lft = timeout; 5065 } 5066 } 5067 5068 rtnl_net_lock(net); 5069 5070 dev = __dev_get_by_index(net, ifm->ifa_index); 5071 if (!dev) { 5072 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 5073 err = -ENODEV; 5074 goto unlock_rtnl; 5075 } 5076 5077 netdev_lock_ops(dev); 5078 idev = ipv6_find_idev(dev); 5079 if (IS_ERR(idev)) { 5080 err = PTR_ERR(idev); 5081 goto unlock; 5082 } 5083 5084 if (!ipv6_allow_optimistic_dad(net, idev)) 5085 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC; 5086 5087 if (cfg.ifa_flags & IFA_F_NODAD && 5088 cfg.ifa_flags & IFA_F_OPTIMISTIC) { 5089 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive"); 5090 err = -EINVAL; 5091 goto unlock; 5092 } 5093 5094 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1); 5095 if (!ifa) { 5096 /* 5097 * It would be best to check for !NLM_F_CREATE here but 5098 * userspace already relies on not having to provide this. 5099 */ 5100 err = inet6_addr_add(net, dev, &cfg, expires, flags, extack); 5101 goto unlock; 5102 } 5103 5104 if (nlh->nlmsg_flags & NLM_F_EXCL || 5105 !(nlh->nlmsg_flags & NLM_F_REPLACE)) { 5106 NL_SET_ERR_MSG_MOD(extack, "address already assigned"); 5107 err = -EEXIST; 5108 } else { 5109 err = inet6_addr_modify(net, ifa, &cfg, expires, flags); 5110 } 5111 5112 in6_ifa_put(ifa); 5113 unlock: 5114 netdev_unlock_ops(dev); 5115 unlock_rtnl: 5116 rtnl_net_unlock(net); 5117 5118 return err; 5119 } 5120 5121 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags, 5122 u8 scope, int ifindex) 5123 { 5124 struct ifaddrmsg *ifm; 5125 5126 ifm = nlmsg_data(nlh); 5127 ifm->ifa_family = AF_INET6; 5128 ifm->ifa_prefixlen = prefixlen; 5129 ifm->ifa_flags = flags; 5130 ifm->ifa_scope = scope; 5131 ifm->ifa_index = ifindex; 5132 } 5133 5134 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp, 5135 unsigned long tstamp, u32 preferred, u32 valid) 5136 { 5137 struct ifa_cacheinfo ci; 5138 5139 ci.cstamp = cstamp_delta(cstamp); 5140 ci.tstamp = cstamp_delta(tstamp); 5141 ci.ifa_prefered = preferred; 5142 ci.ifa_valid = valid; 5143 5144 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci); 5145 } 5146 5147 static inline int rt_scope(int ifa_scope) 5148 { 5149 if (ifa_scope & IFA_HOST) 5150 return RT_SCOPE_HOST; 5151 else if (ifa_scope & IFA_LINK) 5152 return RT_SCOPE_LINK; 5153 else if (ifa_scope & IFA_SITE) 5154 return RT_SCOPE_SITE; 5155 else 5156 return RT_SCOPE_UNIVERSE; 5157 } 5158 5159 static inline int inet6_ifaddr_msgsize(void) 5160 { 5161 return NLMSG_ALIGN(sizeof(struct ifaddrmsg)) 5162 + nla_total_size(16) /* IFA_LOCAL */ 5163 + nla_total_size(16) /* IFA_ADDRESS */ 5164 + nla_total_size(sizeof(struct ifa_cacheinfo)) 5165 + nla_total_size(4) /* IFA_FLAGS */ 5166 + nla_total_size(1) /* IFA_PROTO */ 5167 + nla_total_size(4) /* IFA_RT_PRIORITY */; 5168 } 5169 5170 static int inet6_fill_ifaddr(struct sk_buff *skb, 5171 const struct inet6_ifaddr *ifa, 5172 struct inet6_fill_args *args) 5173 { 5174 struct nlmsghdr *nlh; 5175 u32 preferred, valid; 5176 u32 flags, priority; 5177 u8 proto; 5178 5179 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5180 sizeof(struct ifaddrmsg), args->flags); 5181 if (!nlh) 5182 return -EMSGSIZE; 5183 5184 flags = READ_ONCE(ifa->flags); 5185 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope), 5186 ifa->idev->dev->ifindex); 5187 5188 if (args->netnsid >= 0 && 5189 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) 5190 goto error; 5191 5192 preferred = READ_ONCE(ifa->prefered_lft); 5193 valid = READ_ONCE(ifa->valid_lft); 5194 5195 if (!((flags & IFA_F_PERMANENT) && 5196 (preferred == INFINITY_LIFE_TIME))) { 5197 if (preferred != INFINITY_LIFE_TIME) { 5198 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ; 5199 5200 if (preferred > tval) 5201 preferred -= tval; 5202 else 5203 preferred = 0; 5204 if (valid != INFINITY_LIFE_TIME) { 5205 if (valid > tval) 5206 valid -= tval; 5207 else 5208 valid = 0; 5209 } 5210 } 5211 } else { 5212 preferred = INFINITY_LIFE_TIME; 5213 valid = INFINITY_LIFE_TIME; 5214 } 5215 5216 if (!ipv6_addr_any(&ifa->peer_addr)) { 5217 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 || 5218 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0) 5219 goto error; 5220 } else { 5221 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0) 5222 goto error; 5223 } 5224 5225 priority = READ_ONCE(ifa->rt_priority); 5226 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority)) 5227 goto error; 5228 5229 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp), 5230 preferred, valid) < 0) 5231 goto error; 5232 5233 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0) 5234 goto error; 5235 5236 proto = READ_ONCE(ifa->ifa_proto); 5237 if (proto && nla_put_u8(skb, IFA_PROTO, proto)) 5238 goto error; 5239 5240 nlmsg_end(skb, nlh); 5241 return 0; 5242 5243 error: 5244 nlmsg_cancel(skb, nlh); 5245 return -EMSGSIZE; 5246 } 5247 5248 int inet6_fill_ifmcaddr(struct sk_buff *skb, 5249 const struct ifmcaddr6 *ifmca, 5250 struct inet6_fill_args *args) 5251 { 5252 int ifindex = ifmca->idev->dev->ifindex; 5253 u8 scope = RT_SCOPE_UNIVERSE; 5254 struct nlmsghdr *nlh; 5255 5256 if (!args->force_rt_scope_universe && 5257 ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE) 5258 scope = RT_SCOPE_SITE; 5259 5260 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5261 sizeof(struct ifaddrmsg), args->flags); 5262 if (!nlh) 5263 return -EMSGSIZE; 5264 5265 if (args->netnsid >= 0 && 5266 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) { 5267 nlmsg_cancel(skb, nlh); 5268 return -EMSGSIZE; 5269 } 5270 5271 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 5272 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 || 5273 nla_put_u32(skb, IFA_MC_USERS, READ_ONCE(ifmca->mca_users)) < 0 || 5274 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp), 5275 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5276 nlmsg_cancel(skb, nlh); 5277 return -EMSGSIZE; 5278 } 5279 5280 nlmsg_end(skb, nlh); 5281 return 0; 5282 } 5283 5284 int inet6_fill_ifacaddr(struct sk_buff *skb, 5285 const struct ifacaddr6 *ifaca, 5286 struct inet6_fill_args *args) 5287 { 5288 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt); 5289 int ifindex = dev ? dev->ifindex : 1; 5290 u8 scope = RT_SCOPE_UNIVERSE; 5291 struct nlmsghdr *nlh; 5292 5293 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE) 5294 scope = RT_SCOPE_SITE; 5295 5296 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5297 sizeof(struct ifaddrmsg), args->flags); 5298 if (!nlh) 5299 return -EMSGSIZE; 5300 5301 if (args->netnsid >= 0 && 5302 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) { 5303 nlmsg_cancel(skb, nlh); 5304 return -EMSGSIZE; 5305 } 5306 5307 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 5308 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 || 5309 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp), 5310 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5311 nlmsg_cancel(skb, nlh); 5312 return -EMSGSIZE; 5313 } 5314 5315 nlmsg_end(skb, nlh); 5316 return 0; 5317 } 5318 5319 /* called with rcu_read_lock() */ 5320 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb, 5321 struct netlink_callback *cb, int *s_ip_idx, 5322 struct inet6_fill_args *fillargs) 5323 { 5324 const struct ifmcaddr6 *ifmca; 5325 const struct ifacaddr6 *ifaca; 5326 int ip_idx = 0; 5327 int err = 0; 5328 5329 switch (fillargs->type) { 5330 case UNICAST_ADDR: { 5331 const struct inet6_ifaddr *ifa; 5332 fillargs->event = RTM_NEWADDR; 5333 5334 /* unicast address incl. temp addr */ 5335 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 5336 if (ip_idx < *s_ip_idx) 5337 goto next; 5338 err = inet6_fill_ifaddr(skb, ifa, fillargs); 5339 if (err < 0) 5340 break; 5341 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 5342 next: 5343 ip_idx++; 5344 } 5345 break; 5346 } 5347 case MULTICAST_ADDR: 5348 fillargs->event = RTM_GETMULTICAST; 5349 5350 /* multicast address */ 5351 for (ifmca = rcu_dereference(idev->mc_list); 5352 ifmca; 5353 ifmca = rcu_dereference(ifmca->next), ip_idx++) { 5354 if (ip_idx < *s_ip_idx) 5355 continue; 5356 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs); 5357 if (err < 0) 5358 break; 5359 } 5360 break; 5361 case ANYCAST_ADDR: 5362 fillargs->event = RTM_GETANYCAST; 5363 /* anycast address */ 5364 for (ifaca = rcu_dereference(idev->ac_list); ifaca; 5365 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) { 5366 if (ip_idx < *s_ip_idx) 5367 continue; 5368 err = inet6_fill_ifacaddr(skb, ifaca, fillargs); 5369 if (err < 0) 5370 break; 5371 } 5372 break; 5373 default: 5374 break; 5375 } 5376 *s_ip_idx = err ? ip_idx : 0; 5377 return err; 5378 } 5379 5380 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh, 5381 struct inet6_fill_args *fillargs, 5382 struct net **tgt_net, struct sock *sk, 5383 struct netlink_callback *cb) 5384 { 5385 struct netlink_ext_ack *extack = cb->extack; 5386 struct nlattr *tb[IFA_MAX+1]; 5387 struct ifaddrmsg *ifm; 5388 int err, i; 5389 5390 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 5391 if (!ifm) { 5392 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request"); 5393 return -EINVAL; 5394 } 5395 5396 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5397 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request"); 5398 return -EINVAL; 5399 } 5400 5401 fillargs->ifindex = ifm->ifa_index; 5402 if (fillargs->ifindex) { 5403 cb->answer_flags |= NLM_F_DUMP_FILTERED; 5404 fillargs->flags |= NLM_F_DUMP_FILTERED; 5405 } 5406 5407 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5408 ifa_ipv6_policy, extack); 5409 if (err < 0) 5410 return err; 5411 5412 for (i = 0; i <= IFA_MAX; ++i) { 5413 if (!tb[i]) 5414 continue; 5415 5416 if (i == IFA_TARGET_NETNSID) { 5417 struct net *net; 5418 5419 fillargs->netnsid = nla_get_s32(tb[i]); 5420 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid); 5421 if (IS_ERR(net)) { 5422 fillargs->netnsid = -1; 5423 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id"); 5424 return PTR_ERR(net); 5425 } 5426 *tgt_net = net; 5427 } else { 5428 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request"); 5429 return -EINVAL; 5430 } 5431 } 5432 5433 return 0; 5434 } 5435 5436 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb, 5437 enum addr_type_t type) 5438 { 5439 struct net *tgt_net = sock_net(skb->sk); 5440 const struct nlmsghdr *nlh = cb->nlh; 5441 struct inet6_fill_args fillargs = { 5442 .portid = NETLINK_CB(cb->skb).portid, 5443 .seq = cb->nlh->nlmsg_seq, 5444 .flags = NLM_F_MULTI, 5445 .netnsid = -1, 5446 .type = type, 5447 .force_rt_scope_universe = false, 5448 }; 5449 struct { 5450 unsigned long ifindex; 5451 int ip_idx; 5452 } *ctx = (void *)cb->ctx; 5453 struct net_device *dev; 5454 struct inet6_dev *idev; 5455 int err = 0; 5456 5457 rcu_read_lock(); 5458 if (cb->strict_check) { 5459 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net, 5460 skb->sk, cb); 5461 if (err < 0) 5462 goto done; 5463 5464 err = 0; 5465 if (fillargs.ifindex) { 5466 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex); 5467 if (!dev) { 5468 err = -ENODEV; 5469 goto done; 5470 } 5471 idev = __in6_dev_get(dev); 5472 if (idev) 5473 err = in6_dump_addrs(idev, skb, cb, 5474 &ctx->ip_idx, 5475 &fillargs); 5476 goto done; 5477 } 5478 } 5479 5480 cb->seq = inet6_base_seq(tgt_net); 5481 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) { 5482 idev = __in6_dev_get(dev); 5483 if (!idev) 5484 continue; 5485 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx, 5486 &fillargs); 5487 if (err < 0) 5488 goto done; 5489 } 5490 done: 5491 rcu_read_unlock(); 5492 if (fillargs.netnsid >= 0) 5493 put_net(tgt_net); 5494 5495 return err; 5496 } 5497 5498 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb) 5499 { 5500 enum addr_type_t type = UNICAST_ADDR; 5501 5502 return inet6_dump_addr(skb, cb, type); 5503 } 5504 5505 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb) 5506 { 5507 enum addr_type_t type = MULTICAST_ADDR; 5508 5509 return inet6_dump_addr(skb, cb, type); 5510 } 5511 5512 5513 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb) 5514 { 5515 enum addr_type_t type = ANYCAST_ADDR; 5516 5517 return inet6_dump_addr(skb, cb, type); 5518 } 5519 5520 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb, 5521 const struct nlmsghdr *nlh, 5522 struct nlattr **tb, 5523 struct netlink_ext_ack *extack) 5524 { 5525 struct ifaddrmsg *ifm; 5526 int i, err; 5527 5528 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 5529 if (!ifm) { 5530 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request"); 5531 return -EINVAL; 5532 } 5533 5534 if (!netlink_strict_get_check(skb)) 5535 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 5536 ifa_ipv6_policy, extack); 5537 5538 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5539 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request"); 5540 return -EINVAL; 5541 } 5542 5543 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5544 ifa_ipv6_policy, extack); 5545 if (err) 5546 return err; 5547 5548 for (i = 0; i <= IFA_MAX; i++) { 5549 if (!tb[i]) 5550 continue; 5551 5552 switch (i) { 5553 case IFA_TARGET_NETNSID: 5554 case IFA_ADDRESS: 5555 case IFA_LOCAL: 5556 break; 5557 default: 5558 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request"); 5559 return -EINVAL; 5560 } 5561 } 5562 5563 return 0; 5564 } 5565 5566 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh, 5567 struct netlink_ext_ack *extack) 5568 { 5569 struct net *tgt_net = sock_net(in_skb->sk); 5570 struct inet6_fill_args fillargs = { 5571 .portid = NETLINK_CB(in_skb).portid, 5572 .seq = nlh->nlmsg_seq, 5573 .event = RTM_NEWADDR, 5574 .flags = 0, 5575 .netnsid = -1, 5576 .force_rt_scope_universe = false, 5577 }; 5578 struct ifaddrmsg *ifm; 5579 struct nlattr *tb[IFA_MAX+1]; 5580 struct in6_addr *addr = NULL, *peer; 5581 struct net_device *dev = NULL; 5582 struct inet6_ifaddr *ifa; 5583 struct sk_buff *skb; 5584 int err; 5585 5586 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack); 5587 if (err < 0) 5588 return err; 5589 5590 if (tb[IFA_TARGET_NETNSID]) { 5591 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]); 5592 5593 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk, 5594 fillargs.netnsid); 5595 if (IS_ERR(tgt_net)) 5596 return PTR_ERR(tgt_net); 5597 } 5598 5599 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer); 5600 if (!addr) { 5601 err = -EINVAL; 5602 goto errout; 5603 } 5604 ifm = nlmsg_data(nlh); 5605 if (ifm->ifa_index) 5606 dev = dev_get_by_index(tgt_net, ifm->ifa_index); 5607 5608 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1); 5609 if (!ifa) { 5610 err = -EADDRNOTAVAIL; 5611 goto errout; 5612 } 5613 5614 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL); 5615 if (!skb) { 5616 err = -ENOBUFS; 5617 goto errout_ifa; 5618 } 5619 5620 err = inet6_fill_ifaddr(skb, ifa, &fillargs); 5621 if (err < 0) { 5622 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 5623 WARN_ON(err == -EMSGSIZE); 5624 kfree_skb(skb); 5625 goto errout_ifa; 5626 } 5627 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid); 5628 errout_ifa: 5629 in6_ifa_put(ifa); 5630 errout: 5631 dev_put(dev); 5632 if (fillargs.netnsid >= 0) 5633 put_net(tgt_net); 5634 5635 return err; 5636 } 5637 5638 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa) 5639 { 5640 struct sk_buff *skb; 5641 struct net *net = dev_net(ifa->idev->dev); 5642 struct inet6_fill_args fillargs = { 5643 .portid = 0, 5644 .seq = 0, 5645 .event = event, 5646 .flags = 0, 5647 .netnsid = -1, 5648 .force_rt_scope_universe = false, 5649 }; 5650 int err = -ENOBUFS; 5651 5652 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC); 5653 if (!skb) 5654 goto errout; 5655 5656 err = inet6_fill_ifaddr(skb, ifa, &fillargs); 5657 if (err < 0) { 5658 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 5659 WARN_ON(err == -EMSGSIZE); 5660 kfree_skb(skb); 5661 goto errout; 5662 } 5663 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC); 5664 return; 5665 errout: 5666 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err); 5667 } 5668 5669 static void ipv6_store_devconf(const struct ipv6_devconf *cnf, 5670 __s32 *array, int bytes) 5671 { 5672 BUG_ON(bytes < (DEVCONF_MAX * 4)); 5673 5674 memset(array, 0, bytes); 5675 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding); 5676 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit); 5677 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6); 5678 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra); 5679 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects); 5680 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf); 5681 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits); 5682 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits); 5683 array[DEVCONF_RTR_SOLICIT_INTERVAL] = 5684 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval)); 5685 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] = 5686 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval)); 5687 array[DEVCONF_RTR_SOLICIT_DELAY] = 5688 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay)); 5689 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version); 5690 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] = 5691 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval)); 5692 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] = 5693 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval)); 5694 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr); 5695 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft); 5696 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft); 5697 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry); 5698 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor); 5699 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses); 5700 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr); 5701 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric); 5702 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = 5703 READ_ONCE(cnf->accept_ra_min_hop_limit); 5704 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo); 5705 #ifdef CONFIG_IPV6_ROUTER_PREF 5706 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref); 5707 array[DEVCONF_RTR_PROBE_INTERVAL] = 5708 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval)); 5709 #ifdef CONFIG_IPV6_ROUTE_INFO 5710 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = 5711 READ_ONCE(cnf->accept_ra_rt_info_min_plen); 5712 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = 5713 READ_ONCE(cnf->accept_ra_rt_info_max_plen); 5714 #endif 5715 #endif 5716 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp); 5717 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = 5718 READ_ONCE(cnf->accept_source_route); 5719 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 5720 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad); 5721 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic); 5722 #endif 5723 #ifdef CONFIG_IPV6_MROUTE 5724 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding); 5725 #endif 5726 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6); 5727 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad); 5728 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao); 5729 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify); 5730 array[DEVCONF_SUPPRESS_FRAG_NDISC] = 5731 READ_ONCE(cnf->suppress_frag_ndisc); 5732 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = 5733 READ_ONCE(cnf->accept_ra_from_local); 5734 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu); 5735 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = 5736 READ_ONCE(cnf->ignore_routes_with_linkdown); 5737 /* we omit DEVCONF_STABLE_SECRET for now */ 5738 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only); 5739 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = 5740 READ_ONCE(cnf->drop_unicast_in_l2_multicast); 5741 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na); 5742 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down); 5743 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled); 5744 #ifdef CONFIG_IPV6_SEG6_HMAC 5745 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac); 5746 #endif 5747 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad); 5748 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode); 5749 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy); 5750 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass); 5751 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled); 5752 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled); 5753 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id); 5754 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide); 5755 array[DEVCONF_NDISC_EVICT_NOCARRIER] = 5756 READ_ONCE(cnf->ndisc_evict_nocarrier); 5757 array[DEVCONF_ACCEPT_UNTRACKED_NA] = 5758 READ_ONCE(cnf->accept_untracked_na); 5759 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft); 5760 array[DEVCONF_FORCE_FORWARDING] = READ_ONCE(cnf->force_forwarding); 5761 } 5762 5763 static inline size_t inet6_ifla6_size(void) 5764 { 5765 return nla_total_size(4) /* IFLA_INET6_FLAGS */ 5766 + nla_total_size(sizeof(struct ifla_cacheinfo)) 5767 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */ 5768 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */ 5769 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */ 5770 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */ 5771 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */ 5772 + nla_total_size(4) /* IFLA_INET6_RA_MTU */ 5773 + 0; 5774 } 5775 5776 static inline size_t inet6_if_nlmsg_size(void) 5777 { 5778 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 5779 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 5780 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 5781 + nla_total_size(4) /* IFLA_MTU */ 5782 + nla_total_size(4) /* IFLA_LINK */ 5783 + nla_total_size(1) /* IFLA_OPERSTATE */ 5784 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */ 5785 } 5786 5787 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib, 5788 int bytes) 5789 { 5790 int i; 5791 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX; 5792 BUG_ON(pad < 0); 5793 5794 /* Use put_unaligned() because stats may not be aligned for u64. */ 5795 put_unaligned(ICMP6_MIB_MAX, &stats[0]); 5796 for (i = 1; i < ICMP6_MIB_MAX; i++) 5797 put_unaligned(atomic_long_read(&mib[i]), &stats[i]); 5798 5799 memset(&stats[ICMP6_MIB_MAX], 0, pad); 5800 } 5801 5802 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib, 5803 int bytes, size_t syncpoff) 5804 { 5805 int i, c; 5806 u64 buff[IPSTATS_MIB_MAX]; 5807 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX; 5808 5809 BUG_ON(pad < 0); 5810 5811 memset(buff, 0, sizeof(buff)); 5812 buff[0] = IPSTATS_MIB_MAX; 5813 5814 for_each_possible_cpu(c) { 5815 for (i = 1; i < IPSTATS_MIB_MAX; i++) 5816 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff); 5817 } 5818 5819 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64)); 5820 memset(&stats[IPSTATS_MIB_MAX], 0, pad); 5821 } 5822 5823 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype, 5824 int bytes) 5825 { 5826 switch (attrtype) { 5827 case IFLA_INET6_STATS: 5828 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes, 5829 offsetof(struct ipstats_mib, syncp)); 5830 break; 5831 case IFLA_INET6_ICMP6STATS: 5832 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes); 5833 break; 5834 } 5835 } 5836 5837 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb, 5838 struct inet6_dev *idev) 5839 { 5840 struct nlattr *nla; 5841 5842 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64)); 5843 if (!nla) 5844 goto nla_put_failure; 5845 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla)); 5846 5847 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64)); 5848 if (!nla) 5849 goto nla_put_failure; 5850 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla)); 5851 5852 return 0; 5853 5854 nla_put_failure: 5855 return -EMSGSIZE; 5856 } 5857 5858 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev, 5859 u32 ext_filter_mask) 5860 { 5861 struct ifla_cacheinfo ci; 5862 struct nlattr *nla; 5863 u32 ra_mtu; 5864 5865 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags))) 5866 goto nla_put_failure; 5867 ci.max_reasm_len = IPV6_MAXPLEN; 5868 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp)); 5869 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time); 5870 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME)); 5871 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci)) 5872 goto nla_put_failure; 5873 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32)); 5874 if (!nla) 5875 goto nla_put_failure; 5876 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla)); 5877 5878 /* XXX - MC not implemented */ 5879 5880 if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) { 5881 if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0) 5882 goto nla_put_failure; 5883 } 5884 5885 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr)); 5886 if (!nla) 5887 goto nla_put_failure; 5888 read_lock_bh(&idev->lock); 5889 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla)); 5890 read_unlock_bh(&idev->lock); 5891 5892 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, 5893 READ_ONCE(idev->cnf.addr_gen_mode))) 5894 goto nla_put_failure; 5895 5896 ra_mtu = READ_ONCE(idev->ra_mtu); 5897 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu)) 5898 goto nla_put_failure; 5899 5900 return 0; 5901 5902 nla_put_failure: 5903 return -EMSGSIZE; 5904 } 5905 5906 static size_t inet6_get_link_af_size(const struct net_device *dev, 5907 u32 ext_filter_mask) 5908 { 5909 if (!__in6_dev_get(dev)) 5910 return 0; 5911 5912 return inet6_ifla6_size(); 5913 } 5914 5915 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev, 5916 u32 ext_filter_mask) 5917 { 5918 struct inet6_dev *idev = __in6_dev_get(dev); 5919 5920 if (!idev) 5921 return -ENODATA; 5922 5923 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0) 5924 return -EMSGSIZE; 5925 5926 return 0; 5927 } 5928 5929 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token, 5930 struct netlink_ext_ack *extack) 5931 { 5932 struct inet6_ifaddr *ifp; 5933 struct net_device *dev = idev->dev; 5934 bool clear_token, update_rs = false; 5935 struct in6_addr ll_addr; 5936 5937 ASSERT_RTNL(); 5938 5939 if (!token) 5940 return -EINVAL; 5941 5942 if (dev->flags & IFF_LOOPBACK) { 5943 NL_SET_ERR_MSG_MOD(extack, "Device is loopback"); 5944 return -EINVAL; 5945 } 5946 5947 if (dev->flags & IFF_NOARP) { 5948 NL_SET_ERR_MSG_MOD(extack, 5949 "Device does not do neighbour discovery"); 5950 return -EINVAL; 5951 } 5952 5953 if (!ipv6_accept_ra(idev)) { 5954 NL_SET_ERR_MSG_MOD(extack, 5955 "Router advertisement is disabled on device"); 5956 return -EINVAL; 5957 } 5958 5959 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) { 5960 NL_SET_ERR_MSG(extack, 5961 "Router solicitation is disabled on device"); 5962 return -EINVAL; 5963 } 5964 5965 write_lock_bh(&idev->lock); 5966 5967 BUILD_BUG_ON(sizeof(token->s6_addr) != 16); 5968 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8); 5969 5970 write_unlock_bh(&idev->lock); 5971 5972 clear_token = ipv6_addr_any(token); 5973 if (clear_token) 5974 goto update_lft; 5975 5976 if (!idev->dead && (idev->if_flags & IF_READY) && 5977 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE | 5978 IFA_F_OPTIMISTIC)) { 5979 /* If we're not ready, then normal ifup will take care 5980 * of this. Otherwise, we need to request our rs here. 5981 */ 5982 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters); 5983 update_rs = true; 5984 } 5985 5986 update_lft: 5987 write_lock_bh(&idev->lock); 5988 5989 if (update_rs) { 5990 idev->if_flags |= IF_RS_SENT; 5991 idev->rs_interval = rfc3315_s14_backoff_init( 5992 READ_ONCE(idev->cnf.rtr_solicit_interval)); 5993 idev->rs_probes = 1; 5994 addrconf_mod_rs_timer(idev, idev->rs_interval); 5995 } 5996 5997 /* Well, that's kinda nasty ... */ 5998 list_for_each_entry(ifp, &idev->addr_list, if_list) { 5999 spin_lock(&ifp->lock); 6000 if (ifp->tokenized) { 6001 ifp->valid_lft = 0; 6002 ifp->prefered_lft = 0; 6003 } 6004 spin_unlock(&ifp->lock); 6005 } 6006 6007 write_unlock_bh(&idev->lock); 6008 inet6_ifinfo_notify(RTM_NEWLINK, idev); 6009 addrconf_verify_rtnl(dev_net(dev)); 6010 return 0; 6011 } 6012 6013 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = { 6014 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 }, 6015 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) }, 6016 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT, 6017 .reject_message = 6018 "IFLA_INET6_RA_MTU can not be set" }, 6019 }; 6020 6021 static int check_addr_gen_mode(int mode) 6022 { 6023 if (mode != IN6_ADDR_GEN_MODE_EUI64 && 6024 mode != IN6_ADDR_GEN_MODE_NONE && 6025 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 6026 mode != IN6_ADDR_GEN_MODE_RANDOM) 6027 return -EINVAL; 6028 return 1; 6029 } 6030 6031 static int check_stable_privacy(struct inet6_dev *idev, struct net *net, 6032 int mode) 6033 { 6034 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 6035 !idev->cnf.stable_secret.initialized && 6036 !net->ipv6.devconf_dflt->stable_secret.initialized) 6037 return -EINVAL; 6038 return 1; 6039 } 6040 6041 static int inet6_validate_link_af(const struct net_device *dev, 6042 const struct nlattr *nla, 6043 struct netlink_ext_ack *extack) 6044 { 6045 struct nlattr *tb[IFLA_INET6_MAX + 1]; 6046 struct inet6_dev *idev = NULL; 6047 int err; 6048 6049 if (dev) { 6050 idev = __in6_dev_get(dev); 6051 if (!idev) 6052 return -EAFNOSUPPORT; 6053 } 6054 6055 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, 6056 inet6_af_policy, extack); 6057 if (err) 6058 return err; 6059 6060 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE]) 6061 return -EINVAL; 6062 6063 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 6064 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 6065 6066 if (check_addr_gen_mode(mode) < 0) 6067 return -EINVAL; 6068 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0) 6069 return -EINVAL; 6070 } 6071 6072 return 0; 6073 } 6074 6075 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla, 6076 struct netlink_ext_ack *extack) 6077 { 6078 struct inet6_dev *idev = __in6_dev_get(dev); 6079 struct nlattr *tb[IFLA_INET6_MAX + 1]; 6080 int err; 6081 6082 if (!idev) 6083 return -EAFNOSUPPORT; 6084 6085 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0) 6086 return -EINVAL; 6087 6088 if (tb[IFLA_INET6_TOKEN]) { 6089 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]), 6090 extack); 6091 if (err) 6092 return err; 6093 } 6094 6095 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 6096 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 6097 6098 WRITE_ONCE(idev->cnf.addr_gen_mode, mode); 6099 } 6100 6101 return 0; 6102 } 6103 6104 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, 6105 u32 portid, u32 seq, int event, unsigned int flags) 6106 { 6107 struct net_device *dev = idev->dev; 6108 struct ifinfomsg *hdr; 6109 struct nlmsghdr *nlh; 6110 int ifindex, iflink; 6111 void *protoinfo; 6112 6113 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags); 6114 if (!nlh) 6115 return -EMSGSIZE; 6116 6117 hdr = nlmsg_data(nlh); 6118 hdr->ifi_family = AF_INET6; 6119 hdr->__ifi_pad = 0; 6120 hdr->ifi_type = dev->type; 6121 ifindex = READ_ONCE(dev->ifindex); 6122 hdr->ifi_index = ifindex; 6123 hdr->ifi_flags = netif_get_flags(dev); 6124 hdr->ifi_change = 0; 6125 6126 iflink = dev_get_iflink(dev); 6127 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 6128 (dev->addr_len && 6129 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 6130 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) || 6131 (ifindex != iflink && 6132 nla_put_u32(skb, IFLA_LINK, iflink)) || 6133 nla_put_u8(skb, IFLA_OPERSTATE, 6134 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN)) 6135 goto nla_put_failure; 6136 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO); 6137 if (!protoinfo) 6138 goto nla_put_failure; 6139 6140 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0) 6141 goto nla_put_failure; 6142 6143 nla_nest_end(skb, protoinfo); 6144 nlmsg_end(skb, nlh); 6145 return 0; 6146 6147 nla_put_failure: 6148 nlmsg_cancel(skb, nlh); 6149 return -EMSGSIZE; 6150 } 6151 6152 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh, 6153 struct netlink_ext_ack *extack) 6154 { 6155 struct ifinfomsg *ifm; 6156 6157 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 6158 if (!ifm) { 6159 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request"); 6160 return -EINVAL; 6161 } 6162 6163 if (nlmsg_attrlen(nlh, sizeof(*ifm))) { 6164 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header"); 6165 return -EINVAL; 6166 } 6167 6168 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 6169 ifm->ifi_change || ifm->ifi_index) { 6170 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request"); 6171 return -EINVAL; 6172 } 6173 6174 return 0; 6175 } 6176 6177 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 6178 { 6179 struct net *net = sock_net(skb->sk); 6180 struct { 6181 unsigned long ifindex; 6182 } *ctx = (void *)cb->ctx; 6183 struct net_device *dev; 6184 struct inet6_dev *idev; 6185 int err; 6186 6187 /* only requests using strict checking can pass data to 6188 * influence the dump 6189 */ 6190 if (cb->strict_check) { 6191 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack); 6192 6193 if (err < 0) 6194 return err; 6195 } 6196 6197 err = 0; 6198 rcu_read_lock(); 6199 for_each_netdev_dump(net, dev, ctx->ifindex) { 6200 idev = __in6_dev_get(dev); 6201 if (!idev) 6202 continue; 6203 err = inet6_fill_ifinfo(skb, idev, 6204 NETLINK_CB(cb->skb).portid, 6205 cb->nlh->nlmsg_seq, 6206 RTM_NEWLINK, NLM_F_MULTI); 6207 if (err < 0) 6208 break; 6209 } 6210 rcu_read_unlock(); 6211 6212 return err; 6213 } 6214 6215 void inet6_ifinfo_notify(int event, struct inet6_dev *idev) 6216 { 6217 struct sk_buff *skb; 6218 struct net *net = dev_net(idev->dev); 6219 int err = -ENOBUFS; 6220 6221 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC); 6222 if (!skb) 6223 goto errout; 6224 6225 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0); 6226 if (err < 0) { 6227 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */ 6228 WARN_ON(err == -EMSGSIZE); 6229 kfree_skb(skb); 6230 goto errout; 6231 } 6232 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC); 6233 return; 6234 errout: 6235 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err); 6236 } 6237 6238 static inline size_t inet6_prefix_nlmsg_size(void) 6239 { 6240 return NLMSG_ALIGN(sizeof(struct prefixmsg)) 6241 + nla_total_size(sizeof(struct in6_addr)) 6242 + nla_total_size(sizeof(struct prefix_cacheinfo)); 6243 } 6244 6245 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev, 6246 struct prefix_info *pinfo, u32 portid, u32 seq, 6247 int event, unsigned int flags) 6248 { 6249 struct prefixmsg *pmsg; 6250 struct nlmsghdr *nlh; 6251 struct prefix_cacheinfo ci; 6252 6253 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags); 6254 if (!nlh) 6255 return -EMSGSIZE; 6256 6257 pmsg = nlmsg_data(nlh); 6258 pmsg->prefix_family = AF_INET6; 6259 pmsg->prefix_pad1 = 0; 6260 pmsg->prefix_pad2 = 0; 6261 pmsg->prefix_ifindex = idev->dev->ifindex; 6262 pmsg->prefix_len = pinfo->prefix_len; 6263 pmsg->prefix_type = pinfo->type; 6264 pmsg->prefix_pad3 = 0; 6265 pmsg->prefix_flags = pinfo->flags; 6266 6267 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix)) 6268 goto nla_put_failure; 6269 ci.preferred_time = ntohl(pinfo->prefered); 6270 ci.valid_time = ntohl(pinfo->valid); 6271 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci)) 6272 goto nla_put_failure; 6273 nlmsg_end(skb, nlh); 6274 return 0; 6275 6276 nla_put_failure: 6277 nlmsg_cancel(skb, nlh); 6278 return -EMSGSIZE; 6279 } 6280 6281 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 6282 struct prefix_info *pinfo) 6283 { 6284 struct sk_buff *skb; 6285 struct net *net = dev_net(idev->dev); 6286 int err = -ENOBUFS; 6287 6288 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC); 6289 if (!skb) 6290 goto errout; 6291 6292 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0); 6293 if (err < 0) { 6294 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */ 6295 WARN_ON(err == -EMSGSIZE); 6296 kfree_skb(skb); 6297 goto errout; 6298 } 6299 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC); 6300 return; 6301 errout: 6302 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err); 6303 } 6304 6305 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6306 { 6307 struct net *net = dev_net(ifp->idev->dev); 6308 6309 if (event) 6310 ASSERT_RTNL(); 6311 6312 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp); 6313 6314 switch (event) { 6315 case RTM_NEWADDR: 6316 /* 6317 * If the address was optimistic we inserted the route at the 6318 * start of our DAD process, so we don't need to do it again. 6319 * If the device was taken down in the middle of the DAD 6320 * cycle there is a race where we could get here without a 6321 * host route, so nothing to insert. That will be fixed when 6322 * the device is brought up. 6323 */ 6324 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) { 6325 ip6_ins_rt(net, ifp->rt); 6326 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) { 6327 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n", 6328 &ifp->addr, ifp->idev->dev->name); 6329 } 6330 6331 if (ifp->idev->cnf.forwarding) 6332 addrconf_join_anycast(ifp); 6333 if (!ipv6_addr_any(&ifp->peer_addr)) 6334 addrconf_prefix_route(&ifp->peer_addr, 128, 6335 ifp->rt_priority, ifp->idev->dev, 6336 0, 0, GFP_ATOMIC); 6337 break; 6338 case RTM_DELADDR: 6339 if (ifp->idev->cnf.forwarding) 6340 addrconf_leave_anycast(ifp); 6341 addrconf_leave_solict(ifp->idev, &ifp->addr); 6342 if (!ipv6_addr_any(&ifp->peer_addr)) { 6343 struct fib6_info *rt; 6344 6345 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128, 6346 ifp->idev->dev, 0, 0, 6347 false); 6348 if (rt) 6349 ip6_del_rt(net, rt, false); 6350 } 6351 if (ifp->rt) { 6352 ip6_del_rt(net, ifp->rt, false); 6353 ifp->rt = NULL; 6354 } 6355 rt_genid_bump_ipv6(net); 6356 break; 6357 } 6358 atomic_inc(&net->ipv6.dev_addr_genid); 6359 } 6360 6361 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6362 { 6363 if (likely(ifp->idev->dead == 0)) 6364 __ipv6_ifa_notify(event, ifp); 6365 } 6366 6367 #ifdef CONFIG_SYSCTL 6368 6369 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write, 6370 void *buffer, size_t *lenp, loff_t *ppos) 6371 { 6372 struct ctl_table lctl; 6373 int *valp = ctl->data; 6374 int val = *valp; 6375 int ret; 6376 6377 /* 6378 * ctl->data points to idev->cnf.forwarding, we should 6379 * not modify it until we get the rtnl lock. 6380 */ 6381 lctl = *ctl; 6382 lctl.data = &val; 6383 6384 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6385 if (ret) 6386 return ret; 6387 6388 if (write) 6389 ret = addrconf_fixup_forwarding(ctl, valp, val); 6390 return ret; 6391 } 6392 6393 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write, 6394 void *buffer, size_t *lenp, loff_t *ppos) 6395 { 6396 struct inet6_dev *idev = ctl->extra1; 6397 int min_mtu = IPV6_MIN_MTU; 6398 struct ctl_table lctl; 6399 6400 lctl = *ctl; 6401 lctl.extra1 = &min_mtu; 6402 lctl.extra2 = idev ? &idev->dev->mtu : NULL; 6403 6404 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos); 6405 } 6406 6407 static void dev_disable_change(struct inet6_dev *idev) 6408 { 6409 struct netdev_notifier_info info; 6410 6411 if (!idev || !idev->dev) 6412 return; 6413 6414 netdev_notifier_info_init(&info, idev->dev); 6415 if (idev->cnf.disable_ipv6) 6416 addrconf_notify(NULL, NETDEV_DOWN, &info); 6417 else 6418 addrconf_notify(NULL, NETDEV_UP, &info); 6419 } 6420 6421 static void addrconf_disable_change(struct net *net, __s32 newf) 6422 { 6423 struct net_device *dev; 6424 struct inet6_dev *idev; 6425 6426 for_each_netdev(net, dev) { 6427 idev = __in6_dev_get_rtnl_net(dev); 6428 if (idev) { 6429 int changed = (!idev->cnf.disable_ipv6) ^ (!newf); 6430 6431 WRITE_ONCE(idev->cnf.disable_ipv6, newf); 6432 if (changed) 6433 dev_disable_change(idev); 6434 } 6435 } 6436 } 6437 6438 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf) 6439 { 6440 struct net *net = (struct net *)table->extra2; 6441 int old; 6442 6443 if (p == &net->ipv6.devconf_dflt->disable_ipv6) { 6444 WRITE_ONCE(*p, newf); 6445 return 0; 6446 } 6447 6448 if (!rtnl_net_trylock(net)) 6449 return restart_syscall(); 6450 6451 old = *p; 6452 WRITE_ONCE(*p, newf); 6453 6454 if (p == &net->ipv6.devconf_all->disable_ipv6) { 6455 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf); 6456 addrconf_disable_change(net, newf); 6457 } else if ((!newf) ^ (!old)) { 6458 dev_disable_change((struct inet6_dev *)table->extra1); 6459 } 6460 6461 rtnl_net_unlock(net); 6462 return 0; 6463 } 6464 6465 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write, 6466 void *buffer, size_t *lenp, loff_t *ppos) 6467 { 6468 struct ctl_table lctl; 6469 int *valp = ctl->data; 6470 int val = *valp; 6471 int ret; 6472 6473 /* 6474 * ctl->data points to idev->cnf.disable_ipv6, we should 6475 * not modify it until we get the rtnl lock. 6476 */ 6477 lctl = *ctl; 6478 lctl.data = &val; 6479 6480 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6481 if (ret) 6482 return ret; 6483 6484 if (write) 6485 ret = addrconf_disable_ipv6(ctl, valp, val); 6486 return ret; 6487 } 6488 6489 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write, 6490 void *buffer, size_t *lenp, loff_t *ppos) 6491 { 6492 struct net *net = ctl->extra2; 6493 int *valp = ctl->data; 6494 int old, new; 6495 int ret; 6496 6497 if (write && !rtnl_net_trylock(net)) 6498 return restart_syscall(); 6499 6500 old = *valp; 6501 ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 6502 new = *valp; 6503 6504 if (write && old != new) { 6505 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) { 6506 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6507 NETCONFA_PROXY_NEIGH, 6508 NETCONFA_IFINDEX_DEFAULT, 6509 net->ipv6.devconf_dflt); 6510 } else if (valp == &net->ipv6.devconf_all->proxy_ndp) { 6511 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6512 NETCONFA_PROXY_NEIGH, 6513 NETCONFA_IFINDEX_ALL, 6514 net->ipv6.devconf_all); 6515 } else { 6516 struct inet6_dev *idev = ctl->extra1; 6517 6518 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6519 NETCONFA_PROXY_NEIGH, 6520 idev->dev->ifindex, 6521 &idev->cnf); 6522 } 6523 } 6524 if (write) 6525 rtnl_net_unlock(net); 6526 6527 return ret; 6528 } 6529 6530 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write, 6531 void *buffer, size_t *lenp, 6532 loff_t *ppos) 6533 { 6534 int ret = 0; 6535 u32 new_val; 6536 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1; 6537 struct net *net = (struct net *)ctl->extra2; 6538 struct ctl_table tmp = { 6539 .data = &new_val, 6540 .maxlen = sizeof(new_val), 6541 .mode = ctl->mode, 6542 }; 6543 6544 if (!rtnl_net_trylock(net)) 6545 return restart_syscall(); 6546 6547 new_val = *((u32 *)ctl->data); 6548 6549 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos); 6550 if (ret != 0) 6551 goto out; 6552 6553 if (write) { 6554 if (check_addr_gen_mode(new_val) < 0) { 6555 ret = -EINVAL; 6556 goto out; 6557 } 6558 6559 if (idev) { 6560 if (check_stable_privacy(idev, net, new_val) < 0) { 6561 ret = -EINVAL; 6562 goto out; 6563 } 6564 6565 if (idev->cnf.addr_gen_mode != new_val) { 6566 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val); 6567 netdev_lock_ops(idev->dev); 6568 addrconf_init_auto_addrs(idev->dev); 6569 netdev_unlock_ops(idev->dev); 6570 } 6571 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) { 6572 struct net_device *dev; 6573 6574 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val); 6575 for_each_netdev(net, dev) { 6576 idev = __in6_dev_get_rtnl_net(dev); 6577 if (idev && 6578 idev->cnf.addr_gen_mode != new_val) { 6579 WRITE_ONCE(idev->cnf.addr_gen_mode, 6580 new_val); 6581 netdev_lock_ops(idev->dev); 6582 addrconf_init_auto_addrs(idev->dev); 6583 netdev_unlock_ops(idev->dev); 6584 } 6585 } 6586 } 6587 6588 WRITE_ONCE(*((u32 *)ctl->data), new_val); 6589 } 6590 6591 out: 6592 rtnl_net_unlock(net); 6593 6594 return ret; 6595 } 6596 6597 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write, 6598 void *buffer, size_t *lenp, 6599 loff_t *ppos) 6600 { 6601 int err; 6602 struct in6_addr addr; 6603 char str[IPV6_MAX_STRLEN]; 6604 struct ctl_table lctl = *ctl; 6605 struct net *net = ctl->extra2; 6606 struct ipv6_stable_secret *secret = ctl->data; 6607 6608 if (&net->ipv6.devconf_all->stable_secret == ctl->data) 6609 return -EIO; 6610 6611 lctl.maxlen = IPV6_MAX_STRLEN; 6612 lctl.data = str; 6613 6614 if (!rtnl_net_trylock(net)) 6615 return restart_syscall(); 6616 6617 if (!write && !secret->initialized) { 6618 err = -EIO; 6619 goto out; 6620 } 6621 6622 err = snprintf(str, sizeof(str), "%pI6", &secret->secret); 6623 if (err >= sizeof(str)) { 6624 err = -EIO; 6625 goto out; 6626 } 6627 6628 err = proc_dostring(&lctl, write, buffer, lenp, ppos); 6629 if (err || !write) 6630 goto out; 6631 6632 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) { 6633 err = -EIO; 6634 goto out; 6635 } 6636 6637 secret->initialized = true; 6638 secret->secret = addr; 6639 6640 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) { 6641 struct net_device *dev; 6642 6643 for_each_netdev(net, dev) { 6644 struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev); 6645 6646 if (idev) { 6647 WRITE_ONCE(idev->cnf.addr_gen_mode, 6648 IN6_ADDR_GEN_MODE_STABLE_PRIVACY); 6649 } 6650 } 6651 } else { 6652 struct inet6_dev *idev = ctl->extra1; 6653 6654 WRITE_ONCE(idev->cnf.addr_gen_mode, 6655 IN6_ADDR_GEN_MODE_STABLE_PRIVACY); 6656 } 6657 6658 out: 6659 rtnl_net_unlock(net); 6660 6661 return err; 6662 } 6663 6664 static 6665 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl, 6666 int write, void *buffer, 6667 size_t *lenp, 6668 loff_t *ppos) 6669 { 6670 struct ctl_table lctl; 6671 int *valp = ctl->data; 6672 int val = *valp; 6673 int ret; 6674 6675 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown 6676 * we should not modify it until we get the rtnl lock. 6677 */ 6678 lctl = *ctl; 6679 lctl.data = &val; 6680 6681 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6682 if (ret) 6683 return ret; 6684 6685 if (write) 6686 ret = addrconf_fixup_linkdown(ctl, valp, val); 6687 return ret; 6688 } 6689 6690 static 6691 void addrconf_set_nopolicy(struct rt6_info *rt, int action) 6692 { 6693 if (rt) { 6694 if (action) 6695 rt->dst.flags |= DST_NOPOLICY; 6696 else 6697 rt->dst.flags &= ~DST_NOPOLICY; 6698 } 6699 } 6700 6701 static 6702 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val) 6703 { 6704 struct inet6_ifaddr *ifa; 6705 6706 read_lock_bh(&idev->lock); 6707 list_for_each_entry(ifa, &idev->addr_list, if_list) { 6708 spin_lock(&ifa->lock); 6709 if (ifa->rt) { 6710 /* host routes only use builtin fib6_nh */ 6711 struct fib6_nh *nh = ifa->rt->fib6_nh; 6712 int cpu; 6713 6714 rcu_read_lock(); 6715 ifa->rt->dst_nopolicy = val ? true : false; 6716 if (nh->rt6i_pcpu) { 6717 for_each_possible_cpu(cpu) { 6718 struct rt6_info **rtp; 6719 6720 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu); 6721 addrconf_set_nopolicy(*rtp, val); 6722 } 6723 } 6724 rcu_read_unlock(); 6725 } 6726 spin_unlock(&ifa->lock); 6727 } 6728 read_unlock_bh(&idev->lock); 6729 } 6730 6731 static 6732 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val) 6733 { 6734 struct net *net = (struct net *)ctl->extra2; 6735 struct inet6_dev *idev; 6736 6737 if (valp == &net->ipv6.devconf_dflt->disable_policy) { 6738 WRITE_ONCE(*valp, val); 6739 return 0; 6740 } 6741 6742 if (!rtnl_net_trylock(net)) 6743 return restart_syscall(); 6744 6745 WRITE_ONCE(*valp, val); 6746 6747 if (valp == &net->ipv6.devconf_all->disable_policy) { 6748 struct net_device *dev; 6749 6750 for_each_netdev(net, dev) { 6751 idev = __in6_dev_get_rtnl_net(dev); 6752 if (idev) 6753 addrconf_disable_policy_idev(idev, val); 6754 } 6755 } else { 6756 idev = (struct inet6_dev *)ctl->extra1; 6757 addrconf_disable_policy_idev(idev, val); 6758 } 6759 6760 rtnl_net_unlock(net); 6761 return 0; 6762 } 6763 6764 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write, 6765 void *buffer, size_t *lenp, loff_t *ppos) 6766 { 6767 struct ctl_table lctl; 6768 int *valp = ctl->data; 6769 int val = *valp; 6770 int ret; 6771 6772 lctl = *ctl; 6773 lctl.data = &val; 6774 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6775 if (ret) 6776 return ret; 6777 6778 if (write && (*valp != val)) 6779 ret = addrconf_disable_policy(ctl, valp, val); 6780 6781 return ret; 6782 } 6783 6784 static void addrconf_force_forward_change(struct net *net, __s32 newf) 6785 { 6786 struct net_device *dev; 6787 struct inet6_dev *idev; 6788 6789 for_each_netdev(net, dev) { 6790 idev = __in6_dev_get_rtnl_net(dev); 6791 if (idev) { 6792 int changed = (!idev->cnf.force_forwarding) ^ (!newf); 6793 6794 WRITE_ONCE(idev->cnf.force_forwarding, newf); 6795 if (changed) 6796 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 6797 NETCONFA_FORCE_FORWARDING, 6798 dev->ifindex, &idev->cnf); 6799 } 6800 } 6801 } 6802 6803 static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write, 6804 void *buffer, size_t *lenp, loff_t *ppos) 6805 { 6806 struct inet6_dev *idev = ctl->extra1; 6807 struct ctl_table tmp_ctl = *ctl; 6808 struct net *net = ctl->extra2; 6809 int *valp = ctl->data; 6810 int new_val = *valp; 6811 int old_val = *valp; 6812 int ret; 6813 6814 tmp_ctl.extra1 = SYSCTL_ZERO; 6815 tmp_ctl.extra2 = SYSCTL_ONE; 6816 tmp_ctl.data = &new_val; 6817 6818 ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos); 6819 6820 if (write && old_val != new_val) { 6821 if (!rtnl_net_trylock(net)) 6822 return restart_syscall(); 6823 6824 WRITE_ONCE(*valp, new_val); 6825 6826 if (valp == &net->ipv6.devconf_dflt->force_forwarding) { 6827 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6828 NETCONFA_FORCE_FORWARDING, 6829 NETCONFA_IFINDEX_DEFAULT, 6830 net->ipv6.devconf_dflt); 6831 } else if (valp == &net->ipv6.devconf_all->force_forwarding) { 6832 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6833 NETCONFA_FORCE_FORWARDING, 6834 NETCONFA_IFINDEX_ALL, 6835 net->ipv6.devconf_all); 6836 6837 addrconf_force_forward_change(net, new_val); 6838 } else { 6839 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6840 NETCONFA_FORCE_FORWARDING, 6841 idev->dev->ifindex, 6842 &idev->cnf); 6843 } 6844 rtnl_net_unlock(net); 6845 } 6846 6847 return ret; 6848 } 6849 6850 static int minus_one = -1; 6851 static const int two_five_five = 255; 6852 static u32 ioam6_if_id_max = U16_MAX; 6853 6854 static const struct ctl_table addrconf_sysctl[] = { 6855 { 6856 .procname = "forwarding", 6857 .data = &ipv6_devconf.forwarding, 6858 .maxlen = sizeof(int), 6859 .mode = 0644, 6860 .proc_handler = addrconf_sysctl_forward, 6861 }, 6862 { 6863 .procname = "hop_limit", 6864 .data = &ipv6_devconf.hop_limit, 6865 .maxlen = sizeof(int), 6866 .mode = 0644, 6867 .proc_handler = proc_dointvec_minmax, 6868 .extra1 = (void *)SYSCTL_ONE, 6869 .extra2 = (void *)&two_five_five, 6870 }, 6871 { 6872 .procname = "mtu", 6873 .data = &ipv6_devconf.mtu6, 6874 .maxlen = sizeof(int), 6875 .mode = 0644, 6876 .proc_handler = addrconf_sysctl_mtu, 6877 }, 6878 { 6879 .procname = "accept_ra", 6880 .data = &ipv6_devconf.accept_ra, 6881 .maxlen = sizeof(int), 6882 .mode = 0644, 6883 .proc_handler = proc_dointvec, 6884 }, 6885 { 6886 .procname = "accept_redirects", 6887 .data = &ipv6_devconf.accept_redirects, 6888 .maxlen = sizeof(int), 6889 .mode = 0644, 6890 .proc_handler = proc_dointvec, 6891 }, 6892 { 6893 .procname = "autoconf", 6894 .data = &ipv6_devconf.autoconf, 6895 .maxlen = sizeof(int), 6896 .mode = 0644, 6897 .proc_handler = proc_dointvec, 6898 }, 6899 { 6900 .procname = "dad_transmits", 6901 .data = &ipv6_devconf.dad_transmits, 6902 .maxlen = sizeof(int), 6903 .mode = 0644, 6904 .proc_handler = proc_dointvec, 6905 }, 6906 { 6907 .procname = "router_solicitations", 6908 .data = &ipv6_devconf.rtr_solicits, 6909 .maxlen = sizeof(int), 6910 .mode = 0644, 6911 .proc_handler = proc_dointvec_minmax, 6912 .extra1 = &minus_one, 6913 }, 6914 { 6915 .procname = "router_solicitation_interval", 6916 .data = &ipv6_devconf.rtr_solicit_interval, 6917 .maxlen = sizeof(int), 6918 .mode = 0644, 6919 .proc_handler = proc_dointvec_jiffies, 6920 }, 6921 { 6922 .procname = "router_solicitation_max_interval", 6923 .data = &ipv6_devconf.rtr_solicit_max_interval, 6924 .maxlen = sizeof(int), 6925 .mode = 0644, 6926 .proc_handler = proc_dointvec_jiffies, 6927 }, 6928 { 6929 .procname = "router_solicitation_delay", 6930 .data = &ipv6_devconf.rtr_solicit_delay, 6931 .maxlen = sizeof(int), 6932 .mode = 0644, 6933 .proc_handler = proc_dointvec_jiffies, 6934 }, 6935 { 6936 .procname = "force_mld_version", 6937 .data = &ipv6_devconf.force_mld_version, 6938 .maxlen = sizeof(int), 6939 .mode = 0644, 6940 .proc_handler = proc_dointvec, 6941 }, 6942 { 6943 .procname = "mldv1_unsolicited_report_interval", 6944 .data = 6945 &ipv6_devconf.mldv1_unsolicited_report_interval, 6946 .maxlen = sizeof(int), 6947 .mode = 0644, 6948 .proc_handler = proc_dointvec_ms_jiffies, 6949 }, 6950 { 6951 .procname = "mldv2_unsolicited_report_interval", 6952 .data = 6953 &ipv6_devconf.mldv2_unsolicited_report_interval, 6954 .maxlen = sizeof(int), 6955 .mode = 0644, 6956 .proc_handler = proc_dointvec_ms_jiffies, 6957 }, 6958 { 6959 .procname = "use_tempaddr", 6960 .data = &ipv6_devconf.use_tempaddr, 6961 .maxlen = sizeof(int), 6962 .mode = 0644, 6963 .proc_handler = proc_dointvec, 6964 }, 6965 { 6966 .procname = "temp_valid_lft", 6967 .data = &ipv6_devconf.temp_valid_lft, 6968 .maxlen = sizeof(int), 6969 .mode = 0644, 6970 .proc_handler = proc_dointvec, 6971 }, 6972 { 6973 .procname = "temp_prefered_lft", 6974 .data = &ipv6_devconf.temp_prefered_lft, 6975 .maxlen = sizeof(int), 6976 .mode = 0644, 6977 .proc_handler = proc_dointvec, 6978 }, 6979 { 6980 .procname = "regen_min_advance", 6981 .data = &ipv6_devconf.regen_min_advance, 6982 .maxlen = sizeof(int), 6983 .mode = 0644, 6984 .proc_handler = proc_dointvec, 6985 }, 6986 { 6987 .procname = "regen_max_retry", 6988 .data = &ipv6_devconf.regen_max_retry, 6989 .maxlen = sizeof(int), 6990 .mode = 0644, 6991 .proc_handler = proc_dointvec, 6992 }, 6993 { 6994 .procname = "max_desync_factor", 6995 .data = &ipv6_devconf.max_desync_factor, 6996 .maxlen = sizeof(int), 6997 .mode = 0644, 6998 .proc_handler = proc_dointvec, 6999 }, 7000 { 7001 .procname = "max_addresses", 7002 .data = &ipv6_devconf.max_addresses, 7003 .maxlen = sizeof(int), 7004 .mode = 0644, 7005 .proc_handler = proc_dointvec, 7006 }, 7007 { 7008 .procname = "accept_ra_defrtr", 7009 .data = &ipv6_devconf.accept_ra_defrtr, 7010 .maxlen = sizeof(int), 7011 .mode = 0644, 7012 .proc_handler = proc_dointvec, 7013 }, 7014 { 7015 .procname = "ra_defrtr_metric", 7016 .data = &ipv6_devconf.ra_defrtr_metric, 7017 .maxlen = sizeof(u32), 7018 .mode = 0644, 7019 .proc_handler = proc_douintvec_minmax, 7020 .extra1 = (void *)SYSCTL_ONE, 7021 }, 7022 { 7023 .procname = "accept_ra_min_hop_limit", 7024 .data = &ipv6_devconf.accept_ra_min_hop_limit, 7025 .maxlen = sizeof(int), 7026 .mode = 0644, 7027 .proc_handler = proc_dointvec, 7028 }, 7029 { 7030 .procname = "accept_ra_min_lft", 7031 .data = &ipv6_devconf.accept_ra_min_lft, 7032 .maxlen = sizeof(int), 7033 .mode = 0644, 7034 .proc_handler = proc_dointvec, 7035 }, 7036 { 7037 .procname = "accept_ra_pinfo", 7038 .data = &ipv6_devconf.accept_ra_pinfo, 7039 .maxlen = sizeof(int), 7040 .mode = 0644, 7041 .proc_handler = proc_dointvec, 7042 }, 7043 { 7044 .procname = "ra_honor_pio_life", 7045 .data = &ipv6_devconf.ra_honor_pio_life, 7046 .maxlen = sizeof(u8), 7047 .mode = 0644, 7048 .proc_handler = proc_dou8vec_minmax, 7049 .extra1 = SYSCTL_ZERO, 7050 .extra2 = SYSCTL_ONE, 7051 }, 7052 { 7053 .procname = "ra_honor_pio_pflag", 7054 .data = &ipv6_devconf.ra_honor_pio_pflag, 7055 .maxlen = sizeof(u8), 7056 .mode = 0644, 7057 .proc_handler = proc_dou8vec_minmax, 7058 .extra1 = SYSCTL_ZERO, 7059 .extra2 = SYSCTL_ONE, 7060 }, 7061 #ifdef CONFIG_IPV6_ROUTER_PREF 7062 { 7063 .procname = "accept_ra_rtr_pref", 7064 .data = &ipv6_devconf.accept_ra_rtr_pref, 7065 .maxlen = sizeof(int), 7066 .mode = 0644, 7067 .proc_handler = proc_dointvec, 7068 }, 7069 { 7070 .procname = "router_probe_interval", 7071 .data = &ipv6_devconf.rtr_probe_interval, 7072 .maxlen = sizeof(int), 7073 .mode = 0644, 7074 .proc_handler = proc_dointvec_jiffies, 7075 }, 7076 #ifdef CONFIG_IPV6_ROUTE_INFO 7077 { 7078 .procname = "accept_ra_rt_info_min_plen", 7079 .data = &ipv6_devconf.accept_ra_rt_info_min_plen, 7080 .maxlen = sizeof(int), 7081 .mode = 0644, 7082 .proc_handler = proc_dointvec, 7083 }, 7084 { 7085 .procname = "accept_ra_rt_info_max_plen", 7086 .data = &ipv6_devconf.accept_ra_rt_info_max_plen, 7087 .maxlen = sizeof(int), 7088 .mode = 0644, 7089 .proc_handler = proc_dointvec, 7090 }, 7091 #endif 7092 #endif 7093 { 7094 .procname = "proxy_ndp", 7095 .data = &ipv6_devconf.proxy_ndp, 7096 .maxlen = sizeof(int), 7097 .mode = 0644, 7098 .proc_handler = addrconf_sysctl_proxy_ndp, 7099 }, 7100 { 7101 .procname = "accept_source_route", 7102 .data = &ipv6_devconf.accept_source_route, 7103 .maxlen = sizeof(int), 7104 .mode = 0644, 7105 .proc_handler = proc_dointvec, 7106 }, 7107 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 7108 { 7109 .procname = "optimistic_dad", 7110 .data = &ipv6_devconf.optimistic_dad, 7111 .maxlen = sizeof(int), 7112 .mode = 0644, 7113 .proc_handler = proc_dointvec, 7114 }, 7115 { 7116 .procname = "use_optimistic", 7117 .data = &ipv6_devconf.use_optimistic, 7118 .maxlen = sizeof(int), 7119 .mode = 0644, 7120 .proc_handler = proc_dointvec, 7121 }, 7122 #endif 7123 #ifdef CONFIG_IPV6_MROUTE 7124 { 7125 .procname = "mc_forwarding", 7126 .data = &ipv6_devconf.mc_forwarding, 7127 .maxlen = sizeof(int), 7128 .mode = 0444, 7129 .proc_handler = proc_dointvec, 7130 }, 7131 #endif 7132 { 7133 .procname = "disable_ipv6", 7134 .data = &ipv6_devconf.disable_ipv6, 7135 .maxlen = sizeof(int), 7136 .mode = 0644, 7137 .proc_handler = addrconf_sysctl_disable, 7138 }, 7139 { 7140 .procname = "accept_dad", 7141 .data = &ipv6_devconf.accept_dad, 7142 .maxlen = sizeof(int), 7143 .mode = 0644, 7144 .proc_handler = proc_dointvec, 7145 }, 7146 { 7147 .procname = "force_tllao", 7148 .data = &ipv6_devconf.force_tllao, 7149 .maxlen = sizeof(int), 7150 .mode = 0644, 7151 .proc_handler = proc_dointvec 7152 }, 7153 { 7154 .procname = "ndisc_notify", 7155 .data = &ipv6_devconf.ndisc_notify, 7156 .maxlen = sizeof(int), 7157 .mode = 0644, 7158 .proc_handler = proc_dointvec 7159 }, 7160 { 7161 .procname = "suppress_frag_ndisc", 7162 .data = &ipv6_devconf.suppress_frag_ndisc, 7163 .maxlen = sizeof(int), 7164 .mode = 0644, 7165 .proc_handler = proc_dointvec 7166 }, 7167 { 7168 .procname = "accept_ra_from_local", 7169 .data = &ipv6_devconf.accept_ra_from_local, 7170 .maxlen = sizeof(int), 7171 .mode = 0644, 7172 .proc_handler = proc_dointvec, 7173 }, 7174 { 7175 .procname = "accept_ra_mtu", 7176 .data = &ipv6_devconf.accept_ra_mtu, 7177 .maxlen = sizeof(int), 7178 .mode = 0644, 7179 .proc_handler = proc_dointvec, 7180 }, 7181 { 7182 .procname = "stable_secret", 7183 .data = &ipv6_devconf.stable_secret, 7184 .maxlen = IPV6_MAX_STRLEN, 7185 .mode = 0600, 7186 .proc_handler = addrconf_sysctl_stable_secret, 7187 }, 7188 { 7189 .procname = "use_oif_addrs_only", 7190 .data = &ipv6_devconf.use_oif_addrs_only, 7191 .maxlen = sizeof(int), 7192 .mode = 0644, 7193 .proc_handler = proc_dointvec, 7194 }, 7195 { 7196 .procname = "ignore_routes_with_linkdown", 7197 .data = &ipv6_devconf.ignore_routes_with_linkdown, 7198 .maxlen = sizeof(int), 7199 .mode = 0644, 7200 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown, 7201 }, 7202 { 7203 .procname = "drop_unicast_in_l2_multicast", 7204 .data = &ipv6_devconf.drop_unicast_in_l2_multicast, 7205 .maxlen = sizeof(int), 7206 .mode = 0644, 7207 .proc_handler = proc_dointvec, 7208 }, 7209 { 7210 .procname = "drop_unsolicited_na", 7211 .data = &ipv6_devconf.drop_unsolicited_na, 7212 .maxlen = sizeof(int), 7213 .mode = 0644, 7214 .proc_handler = proc_dointvec, 7215 }, 7216 { 7217 .procname = "keep_addr_on_down", 7218 .data = &ipv6_devconf.keep_addr_on_down, 7219 .maxlen = sizeof(int), 7220 .mode = 0644, 7221 .proc_handler = proc_dointvec, 7222 7223 }, 7224 { 7225 .procname = "seg6_enabled", 7226 .data = &ipv6_devconf.seg6_enabled, 7227 .maxlen = sizeof(int), 7228 .mode = 0644, 7229 .proc_handler = proc_dointvec, 7230 }, 7231 #ifdef CONFIG_IPV6_SEG6_HMAC 7232 { 7233 .procname = "seg6_require_hmac", 7234 .data = &ipv6_devconf.seg6_require_hmac, 7235 .maxlen = sizeof(int), 7236 .mode = 0644, 7237 .proc_handler = proc_dointvec, 7238 }, 7239 #endif 7240 { 7241 .procname = "enhanced_dad", 7242 .data = &ipv6_devconf.enhanced_dad, 7243 .maxlen = sizeof(int), 7244 .mode = 0644, 7245 .proc_handler = proc_dointvec, 7246 }, 7247 { 7248 .procname = "addr_gen_mode", 7249 .data = &ipv6_devconf.addr_gen_mode, 7250 .maxlen = sizeof(int), 7251 .mode = 0644, 7252 .proc_handler = addrconf_sysctl_addr_gen_mode, 7253 }, 7254 { 7255 .procname = "disable_policy", 7256 .data = &ipv6_devconf.disable_policy, 7257 .maxlen = sizeof(int), 7258 .mode = 0644, 7259 .proc_handler = addrconf_sysctl_disable_policy, 7260 }, 7261 { 7262 .procname = "ndisc_tclass", 7263 .data = &ipv6_devconf.ndisc_tclass, 7264 .maxlen = sizeof(int), 7265 .mode = 0644, 7266 .proc_handler = proc_dointvec_minmax, 7267 .extra1 = (void *)SYSCTL_ZERO, 7268 .extra2 = (void *)&two_five_five, 7269 }, 7270 { 7271 .procname = "rpl_seg_enabled", 7272 .data = &ipv6_devconf.rpl_seg_enabled, 7273 .maxlen = sizeof(int), 7274 .mode = 0644, 7275 .proc_handler = proc_dointvec_minmax, 7276 .extra1 = SYSCTL_ZERO, 7277 .extra2 = SYSCTL_ONE, 7278 }, 7279 { 7280 .procname = "ioam6_enabled", 7281 .data = &ipv6_devconf.ioam6_enabled, 7282 .maxlen = sizeof(u8), 7283 .mode = 0644, 7284 .proc_handler = proc_dou8vec_minmax, 7285 .extra1 = (void *)SYSCTL_ZERO, 7286 .extra2 = (void *)SYSCTL_ONE, 7287 }, 7288 { 7289 .procname = "ioam6_id", 7290 .data = &ipv6_devconf.ioam6_id, 7291 .maxlen = sizeof(u32), 7292 .mode = 0644, 7293 .proc_handler = proc_douintvec_minmax, 7294 .extra1 = (void *)SYSCTL_ZERO, 7295 .extra2 = (void *)&ioam6_if_id_max, 7296 }, 7297 { 7298 .procname = "ioam6_id_wide", 7299 .data = &ipv6_devconf.ioam6_id_wide, 7300 .maxlen = sizeof(u32), 7301 .mode = 0644, 7302 .proc_handler = proc_douintvec, 7303 }, 7304 { 7305 .procname = "ndisc_evict_nocarrier", 7306 .data = &ipv6_devconf.ndisc_evict_nocarrier, 7307 .maxlen = sizeof(u8), 7308 .mode = 0644, 7309 .proc_handler = proc_dou8vec_minmax, 7310 .extra1 = (void *)SYSCTL_ZERO, 7311 .extra2 = (void *)SYSCTL_ONE, 7312 }, 7313 { 7314 .procname = "accept_untracked_na", 7315 .data = &ipv6_devconf.accept_untracked_na, 7316 .maxlen = sizeof(int), 7317 .mode = 0644, 7318 .proc_handler = proc_dointvec_minmax, 7319 .extra1 = SYSCTL_ZERO, 7320 .extra2 = SYSCTL_TWO, 7321 }, 7322 { 7323 .procname = "force_forwarding", 7324 .data = &ipv6_devconf.force_forwarding, 7325 .maxlen = sizeof(int), 7326 .mode = 0644, 7327 .proc_handler = addrconf_sysctl_force_forwarding, 7328 }, 7329 }; 7330 7331 static int __addrconf_sysctl_register(struct net *net, char *dev_name, 7332 struct inet6_dev *idev, struct ipv6_devconf *p) 7333 { 7334 size_t table_size = ARRAY_SIZE(addrconf_sysctl); 7335 int i, ifindex; 7336 struct ctl_table *table; 7337 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ]; 7338 7339 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT); 7340 if (!table) 7341 goto out; 7342 7343 for (i = 0; i < table_size; i++) { 7344 table[i].data += (char *)p - (char *)&ipv6_devconf; 7345 /* If one of these is already set, then it is not safe to 7346 * overwrite either of them: this makes proc_dointvec_minmax 7347 * usable. 7348 */ 7349 if (!table[i].extra1 && !table[i].extra2) { 7350 table[i].extra1 = idev; /* embedded; no ref */ 7351 table[i].extra2 = net; 7352 } 7353 } 7354 7355 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name); 7356 7357 p->sysctl_header = register_net_sysctl_sz(net, path, table, 7358 table_size); 7359 if (!p->sysctl_header) 7360 goto free; 7361 7362 if (!strcmp(dev_name, "all")) 7363 ifindex = NETCONFA_IFINDEX_ALL; 7364 else if (!strcmp(dev_name, "default")) 7365 ifindex = NETCONFA_IFINDEX_DEFAULT; 7366 else 7367 ifindex = idev->dev->ifindex; 7368 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, 7369 ifindex, p); 7370 return 0; 7371 7372 free: 7373 kfree(table); 7374 out: 7375 return -ENOBUFS; 7376 } 7377 7378 static void __addrconf_sysctl_unregister(struct net *net, 7379 struct ipv6_devconf *p, int ifindex) 7380 { 7381 const struct ctl_table *table; 7382 7383 if (!p->sysctl_header) 7384 return; 7385 7386 table = p->sysctl_header->ctl_table_arg; 7387 unregister_net_sysctl_table(p->sysctl_header); 7388 p->sysctl_header = NULL; 7389 kfree(table); 7390 7391 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL); 7392 } 7393 7394 static int addrconf_sysctl_register(struct inet6_dev *idev) 7395 { 7396 int err; 7397 7398 if (!sysctl_dev_name_is_allowed(idev->dev->name)) 7399 return -EINVAL; 7400 7401 err = neigh_sysctl_register(idev->dev, idev->nd_parms, 7402 &ndisc_ifinfo_sysctl_change); 7403 if (err) 7404 return err; 7405 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name, 7406 idev, &idev->cnf); 7407 if (err) 7408 neigh_sysctl_unregister(idev->nd_parms); 7409 7410 return err; 7411 } 7412 7413 static void addrconf_sysctl_unregister(struct inet6_dev *idev) 7414 { 7415 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf, 7416 idev->dev->ifindex); 7417 neigh_sysctl_unregister(idev->nd_parms); 7418 } 7419 7420 7421 #endif 7422 7423 static int __net_init addrconf_init_net(struct net *net) 7424 { 7425 int err = -ENOMEM; 7426 struct ipv6_devconf *all, *dflt; 7427 7428 spin_lock_init(&net->ipv6.addrconf_hash_lock); 7429 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work); 7430 net->ipv6.inet6_addr_lst = kzalloc_objs(struct hlist_head, 7431 IN6_ADDR_HSIZE); 7432 if (!net->ipv6.inet6_addr_lst) 7433 goto err_alloc_addr; 7434 7435 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL); 7436 if (!all) 7437 goto err_alloc_all; 7438 7439 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL); 7440 if (!dflt) 7441 goto err_alloc_dflt; 7442 7443 if (!net_eq(net, &init_net)) { 7444 switch (net_inherit_devconf()) { 7445 case 1: /* copy from init_net */ 7446 memcpy(all, init_net.ipv6.devconf_all, 7447 sizeof(ipv6_devconf)); 7448 memcpy(dflt, init_net.ipv6.devconf_dflt, 7449 sizeof(ipv6_devconf_dflt)); 7450 break; 7451 case 3: /* copy from the current netns */ 7452 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all, 7453 sizeof(ipv6_devconf)); 7454 memcpy(dflt, 7455 current->nsproxy->net_ns->ipv6.devconf_dflt, 7456 sizeof(ipv6_devconf_dflt)); 7457 break; 7458 case 0: 7459 case 2: 7460 /* use compiled values */ 7461 break; 7462 } 7463 } 7464 7465 /* these will be inherited by all namespaces */ 7466 dflt->autoconf = ipv6_defaults.autoconf; 7467 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6; 7468 7469 dflt->stable_secret.initialized = false; 7470 all->stable_secret.initialized = false; 7471 7472 net->ipv6.devconf_all = all; 7473 net->ipv6.devconf_dflt = dflt; 7474 7475 #ifdef CONFIG_SYSCTL 7476 err = __addrconf_sysctl_register(net, "all", NULL, all); 7477 if (err < 0) 7478 goto err_reg_all; 7479 7480 err = __addrconf_sysctl_register(net, "default", NULL, dflt); 7481 if (err < 0) 7482 goto err_reg_dflt; 7483 #endif 7484 return 0; 7485 7486 #ifdef CONFIG_SYSCTL 7487 err_reg_dflt: 7488 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL); 7489 err_reg_all: 7490 kfree(dflt); 7491 net->ipv6.devconf_dflt = NULL; 7492 #endif 7493 err_alloc_dflt: 7494 kfree(all); 7495 net->ipv6.devconf_all = NULL; 7496 err_alloc_all: 7497 kfree(net->ipv6.inet6_addr_lst); 7498 err_alloc_addr: 7499 return err; 7500 } 7501 7502 static void __net_exit addrconf_exit_net(struct net *net) 7503 { 7504 int i; 7505 7506 #ifdef CONFIG_SYSCTL 7507 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt, 7508 NETCONFA_IFINDEX_DEFAULT); 7509 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all, 7510 NETCONFA_IFINDEX_ALL); 7511 #endif 7512 kfree(net->ipv6.devconf_dflt); 7513 net->ipv6.devconf_dflt = NULL; 7514 kfree(net->ipv6.devconf_all); 7515 net->ipv6.devconf_all = NULL; 7516 7517 cancel_delayed_work_sync(&net->ipv6.addr_chk_work); 7518 /* 7519 * Check hash table, then free it. 7520 */ 7521 for (i = 0; i < IN6_ADDR_HSIZE; i++) 7522 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i])); 7523 7524 kfree(net->ipv6.inet6_addr_lst); 7525 net->ipv6.inet6_addr_lst = NULL; 7526 } 7527 7528 static struct pernet_operations addrconf_ops = { 7529 .init = addrconf_init_net, 7530 .exit = addrconf_exit_net, 7531 }; 7532 7533 static struct rtnl_af_ops inet6_ops __read_mostly = { 7534 .family = AF_INET6, 7535 .fill_link_af = inet6_fill_link_af, 7536 .get_link_af_size = inet6_get_link_af_size, 7537 .validate_link_af = inet6_validate_link_af, 7538 .set_link_af = inet6_set_link_af, 7539 }; 7540 7541 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = { 7542 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK, 7543 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7544 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR, 7545 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET}, 7546 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR, 7547 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET}, 7548 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR, 7549 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr, 7550 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 7551 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST, 7552 .dumpit = inet6_dump_ifmcaddr, 7553 .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7554 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST, 7555 .dumpit = inet6_dump_ifacaddr, 7556 .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7557 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF, 7558 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf, 7559 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 7560 }; 7561 7562 /* 7563 * Init / cleanup code 7564 */ 7565 7566 int __init addrconf_init(void) 7567 { 7568 struct inet6_dev *idev; 7569 int err; 7570 7571 err = ipv6_addr_label_init(); 7572 if (err < 0) { 7573 pr_crit("%s: cannot initialize default policy table: %d\n", 7574 __func__, err); 7575 goto out; 7576 } 7577 7578 err = register_pernet_subsys(&addrconf_ops); 7579 if (err < 0) 7580 goto out_addrlabel; 7581 7582 /* All works using addrconf_wq need to lock rtnl. */ 7583 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf"); 7584 if (!addrconf_wq) { 7585 err = -ENOMEM; 7586 goto out_nowq; 7587 } 7588 7589 rtnl_net_lock(&init_net); 7590 idev = ipv6_add_dev(blackhole_netdev); 7591 rtnl_net_unlock(&init_net); 7592 if (IS_ERR(idev)) { 7593 err = PTR_ERR(idev); 7594 goto errlo; 7595 } 7596 7597 ip6_route_init_special_entries(); 7598 7599 register_netdevice_notifier(&ipv6_dev_notf); 7600 7601 addrconf_verify(&init_net); 7602 7603 err = rtnl_af_register(&inet6_ops); 7604 if (err) 7605 goto erraf; 7606 7607 err = rtnl_register_many(addrconf_rtnl_msg_handlers); 7608 if (err) 7609 goto errout; 7610 7611 err = ipv6_addr_label_rtnl_register(); 7612 if (err < 0) 7613 goto errout; 7614 7615 return 0; 7616 errout: 7617 rtnl_unregister_all(PF_INET6); 7618 rtnl_af_unregister(&inet6_ops); 7619 erraf: 7620 unregister_netdevice_notifier(&ipv6_dev_notf); 7621 errlo: 7622 destroy_workqueue(addrconf_wq); 7623 out_nowq: 7624 unregister_pernet_subsys(&addrconf_ops); 7625 out_addrlabel: 7626 ipv6_addr_label_cleanup(); 7627 out: 7628 return err; 7629 } 7630 7631 void addrconf_cleanup(void) 7632 { 7633 struct net_device *dev; 7634 7635 unregister_netdevice_notifier(&ipv6_dev_notf); 7636 unregister_pernet_subsys(&addrconf_ops); 7637 ipv6_addr_label_cleanup(); 7638 7639 rtnl_af_unregister(&inet6_ops); 7640 7641 rtnl_net_lock(&init_net); 7642 7643 /* clean dev list */ 7644 for_each_netdev(&init_net, dev) { 7645 if (!__in6_dev_get_rtnl_net(dev)) 7646 continue; 7647 addrconf_ifdown(dev, true); 7648 } 7649 addrconf_ifdown(init_net.loopback_dev, true); 7650 7651 rtnl_net_unlock(&init_net); 7652 7653 destroy_workqueue(addrconf_wq); 7654 } 7655