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