1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 Address [auto]configuration 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 9 */ 10 11 /* 12 * Changes: 13 * 14 * Janos Farkas : delete timer on ifdown 15 * <chexum@bankinf.banki.hu> 16 * Andi Kleen : kill double kfree on module 17 * unload. 18 * Maciej W. Rozycki : FDDI support 19 * sekiya@USAGI : Don't send too many RS 20 * packets. 21 * yoshfuji@USAGI : Fixed interval between DAD 22 * packets. 23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of 24 * address validation timer. 25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041) 26 * support. 27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6 28 * address on a same interface. 29 * YOSHIFUJI Hideaki @USAGI : ARCnet support 30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to 31 * seq_file. 32 * YOSHIFUJI Hideaki @USAGI : improved source address 33 * selection; consider scope, 34 * status etc. 35 */ 36 37 #define pr_fmt(fmt) "IPv6: " fmt 38 39 #include <crypto/sha1.h> 40 #include <linux/errno.h> 41 #include <linux/types.h> 42 #include <linux/kernel.h> 43 #include <linux/sched/signal.h> 44 #include <linux/socket.h> 45 #include <linux/sockios.h> 46 #include <linux/net.h> 47 #include <linux/inet.h> 48 #include <linux/in6.h> 49 #include <linux/netdevice.h> 50 #include <linux/if_addr.h> 51 #include <linux/if_arp.h> 52 #include <linux/if_arcnet.h> 53 #include <linux/if_infiniband.h> 54 #include <linux/route.h> 55 #include <linux/inetdevice.h> 56 #include <linux/init.h> 57 #include <linux/slab.h> 58 #ifdef CONFIG_SYSCTL 59 #include <linux/sysctl.h> 60 #endif 61 #include <linux/capability.h> 62 #include <linux/delay.h> 63 #include <linux/notifier.h> 64 #include <linux/string.h> 65 #include <linux/hash.h> 66 67 #include <net/ip_tunnels.h> 68 #include <net/net_namespace.h> 69 #include <net/sock.h> 70 #include <net/snmp.h> 71 72 #include <net/6lowpan.h> 73 #include <net/firewire.h> 74 #include <net/ipv6.h> 75 #include <net/protocol.h> 76 #include <net/ndisc.h> 77 #include <net/ip6_route.h> 78 #include <net/addrconf.h> 79 #include <net/tcp.h> 80 #include <net/ip.h> 81 #include <net/netlink.h> 82 #include <net/pkt_sched.h> 83 #include <net/l3mdev.h> 84 #include <net/netdev_lock.h> 85 #include <linux/if_tunnel.h> 86 #include <linux/rtnetlink.h> 87 #include <linux/netconf.h> 88 #include <linux/random.h> 89 #include <linux/uaccess.h> 90 #include <linux/unaligned.h> 91 92 #include <linux/proc_fs.h> 93 #include <linux/seq_file.h> 94 #include <linux/export.h> 95 #include <linux/ioam6.h> 96 97 #define IPV6_MAX_STRLEN \ 98 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255") 99 100 static inline u32 cstamp_delta(unsigned long cstamp) 101 { 102 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ; 103 } 104 105 static inline s32 rfc3315_s14_backoff_init(s32 irt) 106 { 107 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */ 108 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt; 109 do_div(tmp, 1000000); 110 return (s32)tmp; 111 } 112 113 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt) 114 { 115 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */ 116 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt; 117 do_div(tmp, 1000000); 118 if ((s32)tmp > mrt) { 119 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */ 120 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt; 121 do_div(tmp, 1000000); 122 } 123 return (s32)tmp; 124 } 125 126 #ifdef CONFIG_SYSCTL 127 static int addrconf_sysctl_register(struct inet6_dev *idev); 128 static void addrconf_sysctl_unregister(struct inet6_dev *idev); 129 #else 130 static inline int addrconf_sysctl_register(struct inet6_dev *idev) 131 { 132 return 0; 133 } 134 135 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev) 136 { 137 } 138 #endif 139 140 static void ipv6_gen_rnd_iid(struct in6_addr *addr); 141 142 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev); 143 static int ipv6_count_addresses(const struct inet6_dev *idev); 144 static int ipv6_generate_stable_address(struct in6_addr *addr, 145 u8 dad_count, 146 const struct inet6_dev *idev); 147 148 #define IN6_ADDR_HSIZE_SHIFT 8 149 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT) 150 151 static void addrconf_verify(struct net *net); 152 static void addrconf_verify_rtnl(struct net *net); 153 154 static struct workqueue_struct *addrconf_wq; 155 156 static void addrconf_join_anycast(struct inet6_ifaddr *ifp); 157 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp); 158 159 static void addrconf_type_change(struct net_device *dev, 160 unsigned long event); 161 static int addrconf_ifdown(struct net_device *dev, bool unregister); 162 163 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 164 int plen, 165 const struct net_device *dev, 166 u32 flags, u32 noflags, 167 bool no_gw); 168 169 static void addrconf_dad_start(struct inet6_ifaddr *ifp); 170 static void addrconf_dad_work(struct work_struct *w); 171 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 172 bool send_na); 173 static void addrconf_dad_run(struct inet6_dev *idev, bool restart); 174 static void addrconf_rs_timer(struct timer_list *t); 175 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 176 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 177 178 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 179 struct prefix_info *pinfo); 180 181 static struct ipv6_devconf ipv6_devconf __read_mostly = { 182 .forwarding = 0, 183 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 184 .mtu6 = IPV6_MIN_MTU, 185 .accept_ra = 1, 186 .accept_redirects = 1, 187 .autoconf = 1, 188 .force_mld_version = 0, 189 .mldv1_unsolicited_report_interval = 10 * HZ, 190 .mldv2_unsolicited_report_interval = HZ, 191 .dad_transmits = 1, 192 .rtr_solicits = MAX_RTR_SOLICITATIONS, 193 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 194 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 195 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 196 .use_tempaddr = 0, 197 .temp_valid_lft = TEMP_VALID_LIFETIME, 198 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 199 .regen_min_advance = REGEN_MIN_ADVANCE, 200 .regen_max_retry = REGEN_MAX_RETRY, 201 .max_desync_factor = MAX_DESYNC_FACTOR, 202 .max_addresses = IPV6_MAX_ADDRESSES, 203 .accept_ra_defrtr = 1, 204 .ra_defrtr_metric = IP6_RT_PRIO_USER, 205 .accept_ra_from_local = 0, 206 .accept_ra_min_hop_limit= 1, 207 .accept_ra_min_lft = 0, 208 .accept_ra_pinfo = 1, 209 #ifdef CONFIG_IPV6_ROUTER_PREF 210 .accept_ra_rtr_pref = 1, 211 .rtr_probe_interval = 60 * HZ, 212 #ifdef CONFIG_IPV6_ROUTE_INFO 213 .accept_ra_rt_info_min_plen = 0, 214 .accept_ra_rt_info_max_plen = 0, 215 #endif 216 #endif 217 .proxy_ndp = 0, 218 .accept_source_route = 0, /* we do not accept RH0 by default. */ 219 .disable_ipv6 = 0, 220 .accept_dad = 0, 221 .suppress_frag_ndisc = 1, 222 .accept_ra_mtu = 1, 223 .stable_secret = { 224 .initialized = false, 225 }, 226 .use_oif_addrs_only = 0, 227 .ignore_routes_with_linkdown = 0, 228 .keep_addr_on_down = 0, 229 .seg6_enabled = 0, 230 #ifdef CONFIG_IPV6_SEG6_HMAC 231 .seg6_require_hmac = 0, 232 #endif 233 .enhanced_dad = 1, 234 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 235 .disable_policy = 0, 236 .rpl_seg_enabled = 0, 237 .ioam6_enabled = 0, 238 .ioam6_id = IOAM6_DEFAULT_IF_ID, 239 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 240 .ndisc_evict_nocarrier = 1, 241 .ra_honor_pio_life = 0, 242 .ra_honor_pio_pflag = 0, 243 .force_forwarding = 0, 244 }; 245 246 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = { 247 .forwarding = 0, 248 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 249 .mtu6 = IPV6_MIN_MTU, 250 .accept_ra = 1, 251 .accept_redirects = 1, 252 .autoconf = 1, 253 .force_mld_version = 0, 254 .mldv1_unsolicited_report_interval = 10 * HZ, 255 .mldv2_unsolicited_report_interval = HZ, 256 .dad_transmits = 1, 257 .rtr_solicits = MAX_RTR_SOLICITATIONS, 258 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 259 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 260 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 261 .use_tempaddr = 0, 262 .temp_valid_lft = TEMP_VALID_LIFETIME, 263 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 264 .regen_min_advance = REGEN_MIN_ADVANCE, 265 .regen_max_retry = REGEN_MAX_RETRY, 266 .max_desync_factor = MAX_DESYNC_FACTOR, 267 .max_addresses = IPV6_MAX_ADDRESSES, 268 .accept_ra_defrtr = 1, 269 .ra_defrtr_metric = IP6_RT_PRIO_USER, 270 .accept_ra_from_local = 0, 271 .accept_ra_min_hop_limit= 1, 272 .accept_ra_min_lft = 0, 273 .accept_ra_pinfo = 1, 274 #ifdef CONFIG_IPV6_ROUTER_PREF 275 .accept_ra_rtr_pref = 1, 276 .rtr_probe_interval = 60 * HZ, 277 #ifdef CONFIG_IPV6_ROUTE_INFO 278 .accept_ra_rt_info_min_plen = 0, 279 .accept_ra_rt_info_max_plen = 0, 280 #endif 281 #endif 282 .proxy_ndp = 0, 283 .accept_source_route = 0, /* we do not accept RH0 by default. */ 284 .disable_ipv6 = 0, 285 .accept_dad = 1, 286 .suppress_frag_ndisc = 1, 287 .accept_ra_mtu = 1, 288 .stable_secret = { 289 .initialized = false, 290 }, 291 .use_oif_addrs_only = 0, 292 .ignore_routes_with_linkdown = 0, 293 .keep_addr_on_down = 0, 294 .seg6_enabled = 0, 295 #ifdef CONFIG_IPV6_SEG6_HMAC 296 .seg6_require_hmac = 0, 297 #endif 298 .enhanced_dad = 1, 299 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 300 .disable_policy = 0, 301 .rpl_seg_enabled = 0, 302 .ioam6_enabled = 0, 303 .ioam6_id = IOAM6_DEFAULT_IF_ID, 304 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 305 .ndisc_evict_nocarrier = 1, 306 .ra_honor_pio_life = 0, 307 .ra_honor_pio_pflag = 0, 308 .force_forwarding = 0, 309 }; 310 311 /* Check if link is ready: is it up and is a valid qdisc available */ 312 static inline bool addrconf_link_ready(const struct net_device *dev) 313 { 314 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev); 315 } 316 317 static void addrconf_del_rs_timer(struct inet6_dev *idev) 318 { 319 if (timer_delete(&idev->rs_timer)) 320 __in6_dev_put(idev); 321 } 322 323 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp) 324 { 325 if (cancel_delayed_work(&ifp->dad_work)) 326 __in6_ifa_put(ifp); 327 } 328 329 static void addrconf_mod_rs_timer(struct inet6_dev *idev, 330 unsigned long when) 331 { 332 if (!mod_timer(&idev->rs_timer, jiffies + when)) 333 in6_dev_hold(idev); 334 } 335 336 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp, 337 unsigned long delay) 338 { 339 in6_ifa_hold(ifp); 340 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay)) 341 in6_ifa_put(ifp); 342 } 343 344 static int snmp6_alloc_dev(struct inet6_dev *idev) 345 { 346 int i; 347 348 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT); 349 if (!idev->stats.ipv6) 350 goto err_ip; 351 352 for_each_possible_cpu(i) { 353 struct ipstats_mib *addrconf_stats; 354 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i); 355 u64_stats_init(&addrconf_stats->syncp); 356 } 357 358 359 idev->stats.icmpv6dev = kzalloc_obj(struct icmpv6_mib_device); 360 if (!idev->stats.icmpv6dev) 361 goto err_icmp; 362 idev->stats.icmpv6msgdev = kzalloc_obj(struct icmpv6msg_mib_device, 363 GFP_KERNEL_ACCOUNT); 364 if (!idev->stats.icmpv6msgdev) 365 goto err_icmpmsg; 366 367 return 0; 368 369 err_icmpmsg: 370 kfree(idev->stats.icmpv6dev); 371 err_icmp: 372 free_percpu(idev->stats.ipv6); 373 err_ip: 374 return -ENOMEM; 375 } 376 377 static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 378 { 379 struct net *net = dev_net(dev); 380 struct neigh_table *tbl; 381 struct inet6_dev *ndev; 382 int err = -ENOMEM; 383 384 ASSERT_RTNL(); 385 netdev_assert_locked_ops_compat(dev); 386 387 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev) 388 return ERR_PTR(-EINVAL); 389 390 ndev = kzalloc_obj(*ndev, GFP_KERNEL_ACCOUNT); 391 if (!ndev) 392 return ERR_PTR(err); 393 394 rwlock_init(&ndev->lock); 395 ndev->dev = dev; 396 INIT_LIST_HEAD(&ndev->addr_list); 397 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0); 398 memcpy(&ndev->cnf, net->ipv6.devconf_dflt, sizeof(ndev->cnf)); 399 400 if (ndev->cnf.stable_secret.initialized) 401 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY; 402 403 tbl = nd_table(net); 404 ndev->cnf.mtu6 = dev->mtu; 405 ndev->ra_mtu = 0; 406 ndev->nd_parms = neigh_parms_alloc(dev, tbl); 407 if (!ndev->nd_parms) { 408 kfree(ndev); 409 return ERR_PTR(err); 410 } 411 if (ndev->cnf.forwarding) 412 netif_disable_lro(dev); 413 /* We refer to the device */ 414 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL); 415 416 if (snmp6_alloc_dev(ndev) < 0) { 417 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n", 418 __func__); 419 neigh_parms_release(tbl, ndev->nd_parms); 420 netdev_put(dev, &ndev->dev_tracker); 421 kfree(ndev); 422 return ERR_PTR(err); 423 } 424 425 if (dev != blackhole_netdev) { 426 if (snmp6_register_dev(ndev) < 0) { 427 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n", 428 __func__, dev->name); 429 goto err_release; 430 } 431 } 432 /* One reference from device. */ 433 refcount_set(&ndev->refcnt, 1); 434 435 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK)) 436 ndev->cnf.accept_dad = -1; 437 438 #if IS_ENABLED(CONFIG_IPV6_SIT) 439 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) { 440 pr_info("%s: Disabled Multicast RS\n", dev->name); 441 ndev->cnf.rtr_solicits = 0; 442 } 443 #endif 444 445 INIT_LIST_HEAD(&ndev->tempaddr_list); 446 ndev->desync_factor = U32_MAX; 447 if ((dev->flags&IFF_LOOPBACK) || 448 dev->type == ARPHRD_TUNNEL || 449 dev->type == ARPHRD_TUNNEL6 || 450 dev->type == ARPHRD_SIT || 451 dev->type == ARPHRD_NONE) { 452 ndev->cnf.use_tempaddr = -1; 453 } 454 455 ndev->token = in6addr_any; 456 457 if (netif_running(dev) && addrconf_link_ready(dev)) 458 ndev->if_flags |= IF_READY; 459 460 ipv6_mc_init_dev(ndev); 461 ndev->tstamp = jiffies; 462 if (dev != blackhole_netdev) { 463 err = addrconf_sysctl_register(ndev); 464 if (err) { 465 ipv6_mc_destroy_dev(ndev); 466 snmp6_unregister_dev(ndev); 467 goto err_release; 468 } 469 } 470 /* protected by rtnl_lock */ 471 rcu_assign_pointer(dev->ip6_ptr, ndev); 472 473 if (dev != blackhole_netdev) { 474 /* Join interface-local all-node multicast group */ 475 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes); 476 477 /* Join all-node multicast group */ 478 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes); 479 480 /* Join all-router multicast group if forwarding is set */ 481 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST)) 482 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 483 } 484 return ndev; 485 486 err_release: 487 neigh_parms_release(tbl, ndev->nd_parms); 488 ndev->dead = 1; 489 in6_dev_finish_destroy(ndev); 490 return ERR_PTR(err); 491 } 492 493 static struct inet6_dev *ipv6_find_idev(struct net_device *dev) 494 { 495 struct inet6_dev *idev; 496 497 ASSERT_RTNL(); 498 499 idev = __in6_dev_get(dev); 500 if (!idev) { 501 idev = ipv6_add_dev(dev); 502 if (IS_ERR(idev)) 503 return idev; 504 } 505 506 if (dev->flags&IFF_UP) 507 ipv6_mc_up(idev); 508 return idev; 509 } 510 511 static int inet6_netconf_msgsize_devconf(int type) 512 { 513 int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) 514 + nla_total_size(4); /* NETCONFA_IFINDEX */ 515 bool all = false; 516 517 if (type == NETCONFA_ALL) 518 all = true; 519 520 if (all || type == NETCONFA_FORWARDING) 521 size += nla_total_size(4); 522 #ifdef CONFIG_IPV6_MROUTE 523 if (all || type == NETCONFA_MC_FORWARDING) 524 size += nla_total_size(4); 525 #endif 526 if (all || type == NETCONFA_PROXY_NEIGH) 527 size += nla_total_size(4); 528 529 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) 530 size += nla_total_size(4); 531 532 return size; 533 } 534 535 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex, 536 struct ipv6_devconf *devconf, u32 portid, 537 u32 seq, int event, unsigned int flags, 538 int type) 539 { 540 struct nlmsghdr *nlh; 541 struct netconfmsg *ncm; 542 bool all = false; 543 544 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), 545 flags); 546 if (!nlh) 547 return -EMSGSIZE; 548 549 if (type == NETCONFA_ALL) 550 all = true; 551 552 ncm = nlmsg_data(nlh); 553 ncm->ncm_family = AF_INET6; 554 555 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0) 556 goto nla_put_failure; 557 558 if (!devconf) 559 goto out; 560 561 if ((all || type == NETCONFA_FORWARDING) && 562 nla_put_s32(skb, NETCONFA_FORWARDING, 563 READ_ONCE(devconf->forwarding)) < 0) 564 goto nla_put_failure; 565 #ifdef CONFIG_IPV6_MROUTE 566 if ((all || type == NETCONFA_MC_FORWARDING) && 567 nla_put_s32(skb, NETCONFA_MC_FORWARDING, 568 atomic_read(&devconf->mc_forwarding)) < 0) 569 goto nla_put_failure; 570 #endif 571 if ((all || type == NETCONFA_PROXY_NEIGH) && 572 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, 573 READ_ONCE(devconf->proxy_ndp)) < 0) 574 goto nla_put_failure; 575 576 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) && 577 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 578 READ_ONCE(devconf->ignore_routes_with_linkdown)) < 0) 579 goto nla_put_failure; 580 581 out: 582 nlmsg_end(skb, nlh); 583 return 0; 584 585 nla_put_failure: 586 nlmsg_cancel(skb, nlh); 587 return -EMSGSIZE; 588 } 589 590 void inet6_netconf_notify_devconf(struct net *net, int event, int type, 591 int ifindex, struct ipv6_devconf *devconf) 592 { 593 struct sk_buff *skb; 594 int err = -ENOBUFS; 595 596 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL); 597 if (!skb) 598 goto errout; 599 600 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0, 601 event, 0, type); 602 if (err < 0) { 603 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 604 WARN_ON(err == -EMSGSIZE); 605 kfree_skb(skb); 606 goto errout; 607 } 608 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL); 609 return; 610 errout: 611 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err); 612 } 613 614 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = { 615 [NETCONFA_IFINDEX] = { .len = sizeof(int) }, 616 [NETCONFA_FORWARDING] = { .len = sizeof(int) }, 617 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) }, 618 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) }, 619 }; 620 621 static int inet6_netconf_valid_get_req(struct sk_buff *skb, 622 const struct nlmsghdr *nlh, 623 struct nlattr **tb, 624 struct netlink_ext_ack *extack) 625 { 626 int i, err; 627 628 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) { 629 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request"); 630 return -EINVAL; 631 } 632 633 if (!netlink_strict_get_check(skb)) 634 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg), 635 tb, NETCONFA_MAX, 636 devconf_ipv6_policy, extack); 637 638 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg), 639 tb, NETCONFA_MAX, 640 devconf_ipv6_policy, extack); 641 if (err) 642 return err; 643 644 for (i = 0; i <= NETCONFA_MAX; i++) { 645 if (!tb[i]) 646 continue; 647 648 switch (i) { 649 case NETCONFA_IFINDEX: 650 break; 651 default: 652 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request"); 653 return -EINVAL; 654 } 655 } 656 657 return 0; 658 } 659 660 static int inet6_netconf_get_devconf(struct sk_buff *in_skb, 661 struct nlmsghdr *nlh, 662 struct netlink_ext_ack *extack) 663 { 664 struct net *net = sock_net(in_skb->sk); 665 struct nlattr *tb[NETCONFA_MAX+1]; 666 struct inet6_dev *in6_dev = NULL; 667 struct net_device *dev = NULL; 668 struct sk_buff *skb; 669 struct ipv6_devconf *devconf; 670 int ifindex; 671 int err; 672 673 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack); 674 if (err < 0) 675 return err; 676 677 if (!tb[NETCONFA_IFINDEX]) 678 return -EINVAL; 679 680 err = -EINVAL; 681 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); 682 switch (ifindex) { 683 case NETCONFA_IFINDEX_ALL: 684 devconf = net->ipv6.devconf_all; 685 break; 686 case NETCONFA_IFINDEX_DEFAULT: 687 devconf = net->ipv6.devconf_dflt; 688 break; 689 default: 690 dev = dev_get_by_index(net, ifindex); 691 if (!dev) 692 return -EINVAL; 693 in6_dev = in6_dev_get(dev); 694 if (!in6_dev) 695 goto errout; 696 devconf = &in6_dev->cnf; 697 break; 698 } 699 700 err = -ENOBUFS; 701 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); 702 if (!skb) 703 goto errout; 704 705 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 706 NETLINK_CB(in_skb).portid, 707 nlh->nlmsg_seq, RTM_NEWNETCONF, 0, 708 NETCONFA_ALL); 709 if (err < 0) { 710 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 711 WARN_ON(err == -EMSGSIZE); 712 kfree_skb(skb); 713 goto errout; 714 } 715 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 716 errout: 717 if (in6_dev) 718 in6_dev_put(in6_dev); 719 dev_put(dev); 720 return err; 721 } 722 723 /* Combine dev_addr_genid and dev_base_seq to detect changes. 724 */ 725 static u32 inet6_base_seq(const struct net *net) 726 { 727 u32 res = atomic_read(&net->ipv6.dev_addr_genid) + 728 READ_ONCE(net->dev_base_seq); 729 730 /* Must not return 0 (see nl_dump_check_consistent()). 731 * Chose a value far away from 0. 732 */ 733 if (!res) 734 res = 0x80000000; 735 return res; 736 } 737 738 static int inet6_netconf_dump_devconf(struct sk_buff *skb, 739 struct netlink_callback *cb) 740 { 741 const struct nlmsghdr *nlh = cb->nlh; 742 struct net *net = sock_net(skb->sk); 743 struct { 744 unsigned long ifindex; 745 unsigned int all_default; 746 } *ctx = (void *)cb->ctx; 747 struct net_device *dev; 748 struct inet6_dev *idev; 749 int err = 0; 750 751 if (cb->strict_check) { 752 struct netlink_ext_ack *extack = cb->extack; 753 struct netconfmsg *ncm; 754 755 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) { 756 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request"); 757 return -EINVAL; 758 } 759 760 if (nlmsg_attrlen(nlh, sizeof(*ncm))) { 761 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request"); 762 return -EINVAL; 763 } 764 } 765 766 rcu_read_lock(); 767 for_each_netdev_dump(net, dev, ctx->ifindex) { 768 idev = __in6_dev_get(dev); 769 if (!idev) 770 continue; 771 err = inet6_netconf_fill_devconf(skb, dev->ifindex, 772 &idev->cnf, 773 NETLINK_CB(cb->skb).portid, 774 nlh->nlmsg_seq, 775 RTM_NEWNETCONF, 776 NLM_F_MULTI, 777 NETCONFA_ALL); 778 if (err < 0) 779 goto done; 780 } 781 if (ctx->all_default == 0) { 782 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL, 783 net->ipv6.devconf_all, 784 NETLINK_CB(cb->skb).portid, 785 nlh->nlmsg_seq, 786 RTM_NEWNETCONF, NLM_F_MULTI, 787 NETCONFA_ALL); 788 if (err < 0) 789 goto done; 790 ctx->all_default++; 791 } 792 if (ctx->all_default == 1) { 793 err = inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT, 794 net->ipv6.devconf_dflt, 795 NETLINK_CB(cb->skb).portid, 796 nlh->nlmsg_seq, 797 RTM_NEWNETCONF, NLM_F_MULTI, 798 NETCONFA_ALL); 799 if (err < 0) 800 goto done; 801 ctx->all_default++; 802 } 803 done: 804 rcu_read_unlock(); 805 return err; 806 } 807 808 #ifdef CONFIG_SYSCTL 809 static void dev_forward_change(struct inet6_dev *idev) 810 { 811 struct net_device *dev; 812 struct inet6_ifaddr *ifa; 813 LIST_HEAD(tmp_addr_list); 814 815 if (!idev) 816 return; 817 dev = idev->dev; 818 if (idev->cnf.forwarding) 819 dev_disable_lro(dev); 820 if (dev->flags & IFF_MULTICAST) { 821 if (idev->cnf.forwarding) { 822 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 823 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters); 824 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters); 825 } else { 826 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters); 827 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters); 828 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters); 829 } 830 } 831 832 read_lock_bh(&idev->lock); 833 list_for_each_entry(ifa, &idev->addr_list, if_list) { 834 if (ifa->flags&IFA_F_TENTATIVE) 835 continue; 836 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 837 } 838 read_unlock_bh(&idev->lock); 839 840 while (!list_empty(&tmp_addr_list)) { 841 ifa = list_first_entry(&tmp_addr_list, 842 struct inet6_ifaddr, if_list_aux); 843 list_del(&ifa->if_list_aux); 844 if (idev->cnf.forwarding) 845 addrconf_join_anycast(ifa); 846 else 847 addrconf_leave_anycast(ifa); 848 } 849 850 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 851 NETCONFA_FORWARDING, 852 dev->ifindex, &idev->cnf); 853 } 854 855 856 static void addrconf_forward_change(struct net *net, __s32 newf) 857 { 858 struct net_device *dev; 859 struct inet6_dev *idev; 860 861 for_each_netdev(net, dev) { 862 idev = __in6_dev_get_rtnl_net(dev); 863 if (idev) { 864 int changed = (!idev->cnf.forwarding) ^ (!newf); 865 /* Disabling all.forwarding sets 0 to force_forwarding for all interfaces */ 866 if (newf == 0) 867 WRITE_ONCE(idev->cnf.force_forwarding, 0); 868 869 WRITE_ONCE(idev->cnf.forwarding, newf); 870 if (changed) 871 dev_forward_change(idev); 872 } 873 } 874 } 875 876 static int addrconf_fixup_forwarding(const struct ctl_table *table, int *p, int newf) 877 { 878 struct net *net = (struct net *)table->extra2; 879 int old; 880 881 if (!rtnl_net_trylock(net)) 882 return restart_syscall(); 883 884 old = *p; 885 WRITE_ONCE(*p, newf); 886 887 if (p == &net->ipv6.devconf_dflt->forwarding) { 888 if ((!newf) ^ (!old)) 889 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 890 NETCONFA_FORWARDING, 891 NETCONFA_IFINDEX_DEFAULT, 892 net->ipv6.devconf_dflt); 893 rtnl_net_unlock(net); 894 return 0; 895 } 896 897 if (p == &net->ipv6.devconf_all->forwarding) { 898 int old_dflt = net->ipv6.devconf_dflt->forwarding; 899 900 WRITE_ONCE(net->ipv6.devconf_dflt->forwarding, newf); 901 if ((!newf) ^ (!old_dflt)) 902 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 903 NETCONFA_FORWARDING, 904 NETCONFA_IFINDEX_DEFAULT, 905 net->ipv6.devconf_dflt); 906 907 addrconf_forward_change(net, newf); 908 if ((!newf) ^ (!old)) 909 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 910 NETCONFA_FORWARDING, 911 NETCONFA_IFINDEX_ALL, 912 net->ipv6.devconf_all); 913 } else if ((!newf) ^ (!old)) 914 dev_forward_change((struct inet6_dev *)table->extra1); 915 rtnl_net_unlock(net); 916 917 if (newf) 918 rt6_purge_dflt_routers(net); 919 return 0; 920 } 921 922 static void addrconf_linkdown_change(struct net *net, __s32 newf) 923 { 924 struct net_device *dev; 925 struct inet6_dev *idev; 926 927 for_each_netdev(net, dev) { 928 idev = __in6_dev_get_rtnl_net(dev); 929 if (idev) { 930 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf); 931 932 WRITE_ONCE(idev->cnf.ignore_routes_with_linkdown, newf); 933 if (changed) 934 inet6_netconf_notify_devconf(dev_net(dev), 935 RTM_NEWNETCONF, 936 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 937 dev->ifindex, 938 &idev->cnf); 939 } 940 } 941 } 942 943 static int addrconf_fixup_linkdown(const struct ctl_table *table, int *p, int newf) 944 { 945 struct net *net = (struct net *)table->extra2; 946 int old; 947 948 if (!rtnl_net_trylock(net)) 949 return restart_syscall(); 950 951 old = *p; 952 WRITE_ONCE(*p, newf); 953 954 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) { 955 if ((!newf) ^ (!old)) 956 inet6_netconf_notify_devconf(net, 957 RTM_NEWNETCONF, 958 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 959 NETCONFA_IFINDEX_DEFAULT, 960 net->ipv6.devconf_dflt); 961 } else if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) { 962 WRITE_ONCE(net->ipv6.devconf_dflt->ignore_routes_with_linkdown, newf); 963 addrconf_linkdown_change(net, newf); 964 if ((!newf) ^ (!old)) 965 inet6_netconf_notify_devconf(net, 966 RTM_NEWNETCONF, 967 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 968 NETCONFA_IFINDEX_ALL, 969 net->ipv6.devconf_all); 970 } else { 971 if (!newf ^ !old) { 972 struct inet6_dev *idev = table->extra1; 973 974 inet6_netconf_notify_devconf(net, 975 RTM_NEWNETCONF, 976 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 977 idev->dev->ifindex, 978 &idev->cnf); 979 } 980 } 981 982 rtnl_net_unlock(net); 983 984 return 0; 985 } 986 987 #endif 988 989 /* Nobody refers to this ifaddr, destroy it */ 990 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp) 991 { 992 WARN_ON(!hlist_unhashed(&ifp->addr_lst)); 993 994 #ifdef NET_REFCNT_DEBUG 995 pr_debug("%s\n", __func__); 996 #endif 997 998 in6_dev_put(ifp->idev); 999 1000 if (cancel_delayed_work(&ifp->dad_work)) 1001 pr_notice("delayed DAD work was pending while freeing ifa=%p\n", 1002 ifp); 1003 1004 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 1005 pr_warn("Freeing alive inet6 address %p\n", ifp); 1006 return; 1007 } 1008 1009 kfree_rcu(ifp, rcu); 1010 } 1011 1012 static void 1013 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp) 1014 { 1015 struct list_head *p; 1016 int ifp_scope = ipv6_addr_src_scope(&ifp->addr); 1017 1018 /* 1019 * Each device address list is sorted in order of scope - 1020 * global before linklocal. 1021 */ 1022 list_for_each(p, &idev->addr_list) { 1023 struct inet6_ifaddr *ifa 1024 = list_entry(p, struct inet6_ifaddr, if_list); 1025 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr)) 1026 break; 1027 } 1028 1029 list_add_tail_rcu(&ifp->if_list, p); 1030 } 1031 1032 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr) 1033 { 1034 u32 val = __ipv6_addr_jhash(addr, net_hash_mix(net)); 1035 1036 return hash_32(val, IN6_ADDR_HSIZE_SHIFT); 1037 } 1038 1039 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr, 1040 struct net_device *dev, unsigned int hash) 1041 { 1042 struct inet6_ifaddr *ifp; 1043 1044 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 1045 if (ipv6_addr_equal(&ifp->addr, addr)) { 1046 if (!dev || ifp->idev->dev == dev) 1047 return true; 1048 } 1049 } 1050 return false; 1051 } 1052 1053 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa) 1054 { 1055 struct net *net = dev_net(dev); 1056 unsigned int hash = inet6_addr_hash(net, &ifa->addr); 1057 int err = 0; 1058 1059 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1060 1061 /* Ignore adding duplicate addresses on an interface */ 1062 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) { 1063 netdev_dbg(dev, "ipv6_add_addr: already assigned\n"); 1064 err = -EEXIST; 1065 } else { 1066 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]); 1067 } 1068 1069 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1070 1071 return err; 1072 } 1073 1074 /* On success it returns ifp with increased reference count */ 1075 1076 static struct inet6_ifaddr * 1077 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg, 1078 bool can_block, struct netlink_ext_ack *extack) 1079 { 1080 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC; 1081 int addr_type = ipv6_addr_type(cfg->pfx); 1082 struct net *net = dev_net(idev->dev); 1083 struct inet6_ifaddr *ifa = NULL; 1084 struct fib6_info *f6i = NULL; 1085 int err = 0; 1086 1087 if (addr_type == IPV6_ADDR_ANY) { 1088 NL_SET_ERR_MSG_MOD(extack, "Invalid address"); 1089 return ERR_PTR(-EADDRNOTAVAIL); 1090 } else if (addr_type & IPV6_ADDR_MULTICAST && 1091 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) { 1092 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag"); 1093 return ERR_PTR(-EADDRNOTAVAIL); 1094 } else if (!(idev->dev->flags & IFF_LOOPBACK) && 1095 !netif_is_l3_master(idev->dev) && 1096 addr_type & IPV6_ADDR_LOOPBACK) { 1097 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device"); 1098 return ERR_PTR(-EADDRNOTAVAIL); 1099 } 1100 1101 if (idev->dead) { 1102 NL_SET_ERR_MSG_MOD(extack, "device is going away"); 1103 err = -ENODEV; 1104 goto out; 1105 } 1106 1107 if (idev->cnf.disable_ipv6) { 1108 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 1109 err = -EACCES; 1110 goto out; 1111 } 1112 1113 /* validator notifier needs to be blocking; 1114 * do not call in atomic context 1115 */ 1116 if (can_block) { 1117 struct in6_validator_info i6vi = { 1118 .i6vi_addr = *cfg->pfx, 1119 .i6vi_dev = idev, 1120 .extack = extack, 1121 }; 1122 1123 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi); 1124 err = notifier_to_errno(err); 1125 if (err < 0) 1126 goto out; 1127 } 1128 1129 ifa = kzalloc_obj(*ifa, gfp_flags | __GFP_ACCOUNT); 1130 if (!ifa) { 1131 err = -ENOBUFS; 1132 goto out; 1133 } 1134 1135 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack); 1136 if (IS_ERR(f6i)) { 1137 err = PTR_ERR(f6i); 1138 f6i = NULL; 1139 goto out; 1140 } 1141 1142 neigh_parms_data_state_setall(idev->nd_parms); 1143 1144 ifa->addr = *cfg->pfx; 1145 if (cfg->peer_pfx) 1146 ifa->peer_addr = *cfg->peer_pfx; 1147 1148 spin_lock_init(&ifa->lock); 1149 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work); 1150 INIT_HLIST_NODE(&ifa->addr_lst); 1151 ifa->scope = cfg->scope; 1152 ifa->prefix_len = cfg->plen; 1153 ifa->rt_priority = cfg->rt_priority; 1154 ifa->flags = cfg->ifa_flags; 1155 ifa->ifa_proto = cfg->ifa_proto; 1156 /* No need to add the TENTATIVE flag for addresses with NODAD */ 1157 if (!(cfg->ifa_flags & IFA_F_NODAD)) 1158 ifa->flags |= IFA_F_TENTATIVE; 1159 ifa->valid_lft = cfg->valid_lft; 1160 ifa->prefered_lft = cfg->preferred_lft; 1161 ifa->tokenized = false; 1162 1163 ifa->rt = f6i; 1164 ifa->ifpub = cfg->ifpub; 1165 if (cfg->ifpub) { 1166 ifa->cstamp = cfg->cstamp; 1167 ifa->tstamp = cfg->tstamp; 1168 } else { 1169 unsigned long now = jiffies; 1170 1171 ifa->cstamp = now; 1172 ifa->tstamp = now; 1173 } 1174 1175 ifa->idev = idev; 1176 in6_dev_hold(idev); 1177 1178 /* For caller */ 1179 refcount_set(&ifa->refcnt, 1); 1180 1181 rcu_read_lock(); 1182 1183 write_lock_bh(&idev->lock); 1184 1185 if (idev->dead) 1186 err = -ENODEV; 1187 else 1188 err = ipv6_add_addr_hash(idev->dev, ifa); 1189 if (err < 0) { 1190 write_unlock_bh(&idev->lock); 1191 rcu_read_unlock(); 1192 goto out; 1193 } 1194 1195 /* Add to inet6_dev unicast addr list. */ 1196 ipv6_link_dev_addr(idev, ifa); 1197 1198 if (ifa->flags&IFA_F_TEMPORARY) { 1199 /* manage_tempaddrs() relies on addresses being added to the head */ 1200 list_add(&ifa->tmp_list, &idev->tempaddr_list); 1201 in6_ifa_hold(ifa); 1202 } 1203 1204 in6_ifa_hold(ifa); 1205 write_unlock_bh(&idev->lock); 1206 1207 rcu_read_unlock(); 1208 1209 inet6addr_notifier_call_chain(NETDEV_UP, ifa); 1210 out: 1211 if (unlikely(err < 0)) { 1212 fib6_info_release(f6i); 1213 1214 if (ifa) { 1215 if (ifa->idev) 1216 in6_dev_put(ifa->idev); 1217 kfree(ifa); 1218 } 1219 ifa = ERR_PTR(err); 1220 } 1221 1222 return ifa; 1223 } 1224 1225 enum cleanup_prefix_rt_t { 1226 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */ 1227 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */ 1228 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */ 1229 }; 1230 1231 /* 1232 * Check, whether the prefix for ifp would still need a prefix route 1233 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_* 1234 * constants. 1235 * 1236 * 1) we don't purge prefix if address was not permanent. 1237 * prefix is managed by its own lifetime. 1238 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE. 1239 * 3) if there are no addresses, delete prefix. 1240 * 4) if there are still other permanent address(es), 1241 * corresponding prefix is still permanent. 1242 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE, 1243 * don't purge the prefix, assume user space is managing it. 1244 * 6) otherwise, update prefix lifetime to the 1245 * longest valid lifetime among the corresponding 1246 * addresses on the device. 1247 * Note: subsequent RA will update lifetime. 1248 **/ 1249 static enum cleanup_prefix_rt_t 1250 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires) 1251 { 1252 struct inet6_ifaddr *ifa; 1253 struct inet6_dev *idev = ifp->idev; 1254 unsigned long lifetime; 1255 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL; 1256 1257 *expires = jiffies; 1258 1259 list_for_each_entry(ifa, &idev->addr_list, if_list) { 1260 if (ifa == ifp) 1261 continue; 1262 if (ifa->prefix_len != ifp->prefix_len || 1263 !ipv6_prefix_equal(&ifa->addr, &ifp->addr, 1264 ifp->prefix_len)) 1265 continue; 1266 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE)) 1267 return CLEANUP_PREFIX_RT_NOP; 1268 1269 action = CLEANUP_PREFIX_RT_EXPIRE; 1270 1271 spin_lock(&ifa->lock); 1272 1273 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ); 1274 /* 1275 * Note: Because this address is 1276 * not permanent, lifetime < 1277 * LONG_MAX / HZ here. 1278 */ 1279 if (time_before(*expires, ifa->tstamp + lifetime * HZ)) 1280 *expires = ifa->tstamp + lifetime * HZ; 1281 spin_unlock(&ifa->lock); 1282 } 1283 1284 return action; 1285 } 1286 1287 static void 1288 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, 1289 bool del_rt, bool del_peer) 1290 { 1291 struct net *net = dev_net(ifp->idev->dev); 1292 struct fib6_table *table; 1293 struct fib6_info *f6i; 1294 1295 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr, 1296 ifp->prefix_len, 1297 ifp->idev->dev, 0, RTF_DEFAULT, true); 1298 if (f6i) { 1299 if (del_rt) 1300 ip6_del_rt(net, f6i, false); 1301 else { 1302 if (f6i != net->ipv6.fib6_null_entry && 1303 !(f6i->fib6_flags & RTF_EXPIRES)) { 1304 table = f6i->fib6_table; 1305 spin_lock_bh(&table->tb6_lock); 1306 1307 fib6_set_expires(f6i, expires); 1308 fib6_add_gc_list(f6i); 1309 1310 spin_unlock_bh(&table->tb6_lock); 1311 } 1312 fib6_info_release(f6i); 1313 } 1314 } 1315 } 1316 1317 1318 /* This function wants to get referenced ifp and releases it before return */ 1319 1320 static void ipv6_del_addr(struct inet6_ifaddr *ifp) 1321 { 1322 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP; 1323 struct net *net = dev_net(ifp->idev->dev); 1324 unsigned long expires; 1325 int state; 1326 1327 ASSERT_RTNL(); 1328 1329 spin_lock_bh(&ifp->lock); 1330 state = ifp->state; 1331 ifp->state = INET6_IFADDR_STATE_DEAD; 1332 spin_unlock_bh(&ifp->lock); 1333 1334 if (state == INET6_IFADDR_STATE_DEAD) 1335 goto out; 1336 1337 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1338 hlist_del_init_rcu(&ifp->addr_lst); 1339 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1340 1341 write_lock_bh(&ifp->idev->lock); 1342 1343 if (ifp->flags&IFA_F_TEMPORARY) { 1344 list_del(&ifp->tmp_list); 1345 if (ifp->ifpub) { 1346 in6_ifa_put(ifp->ifpub); 1347 ifp->ifpub = NULL; 1348 } 1349 __in6_ifa_put(ifp); 1350 } 1351 1352 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) 1353 action = check_cleanup_prefix_route(ifp, &expires); 1354 1355 list_del_rcu(&ifp->if_list); 1356 __in6_ifa_put(ifp); 1357 1358 write_unlock_bh(&ifp->idev->lock); 1359 1360 addrconf_del_dad_work(ifp); 1361 1362 ipv6_ifa_notify(RTM_DELADDR, ifp); 1363 1364 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp); 1365 1366 if (action != CLEANUP_PREFIX_RT_NOP) { 1367 cleanup_prefix_route(ifp, expires, 1368 action == CLEANUP_PREFIX_RT_DEL, false); 1369 } 1370 1371 /* clean up prefsrc entries */ 1372 rt6_remove_prefsrc(ifp); 1373 out: 1374 in6_ifa_put(ifp); 1375 } 1376 1377 static unsigned long ipv6_get_regen_advance(const struct inet6_dev *idev) 1378 { 1379 return READ_ONCE(idev->cnf.regen_min_advance) + 1380 READ_ONCE(idev->cnf.regen_max_retry) * 1381 READ_ONCE(idev->cnf.dad_transmits) * 1382 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ; 1383 } 1384 1385 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block) 1386 { 1387 struct inet6_dev *idev = ifp->idev; 1388 unsigned long tmp_tstamp, age; 1389 unsigned long regen_advance; 1390 unsigned long now = jiffies; 1391 u32 if_public_preferred_lft; 1392 s32 cnf_temp_preferred_lft; 1393 struct inet6_ifaddr *ift; 1394 struct ifa6_config cfg; 1395 long max_desync_factor; 1396 struct in6_addr addr; 1397 int ret = 0; 1398 1399 write_lock_bh(&idev->lock); 1400 1401 retry: 1402 in6_dev_hold(idev); 1403 if (READ_ONCE(idev->cnf.use_tempaddr) <= 0) { 1404 write_unlock_bh(&idev->lock); 1405 pr_info("%s: use_tempaddr is disabled\n", __func__); 1406 in6_dev_put(idev); 1407 ret = -1; 1408 goto out; 1409 } 1410 spin_lock_bh(&ifp->lock); 1411 if (ifp->regen_count++ >= READ_ONCE(idev->cnf.regen_max_retry)) { 1412 WRITE_ONCE(idev->cnf.use_tempaddr, -1); /*XXX*/ 1413 spin_unlock_bh(&ifp->lock); 1414 write_unlock_bh(&idev->lock); 1415 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n", 1416 __func__); 1417 in6_dev_put(idev); 1418 ret = -1; 1419 goto out; 1420 } 1421 in6_ifa_hold(ifp); 1422 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8); 1423 ipv6_gen_rnd_iid(&addr); 1424 1425 age = (now - ifp->tstamp) / HZ; 1426 1427 regen_advance = ipv6_get_regen_advance(idev); 1428 1429 /* recalculate max_desync_factor each time and update 1430 * idev->desync_factor if it's larger 1431 */ 1432 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft); 1433 max_desync_factor = min_t(long, 1434 READ_ONCE(idev->cnf.max_desync_factor), 1435 cnf_temp_preferred_lft - regen_advance); 1436 1437 if (unlikely(idev->desync_factor > max_desync_factor)) { 1438 if (max_desync_factor > 0) { 1439 get_random_bytes(&idev->desync_factor, 1440 sizeof(idev->desync_factor)); 1441 idev->desync_factor %= max_desync_factor; 1442 } else { 1443 idev->desync_factor = 0; 1444 } 1445 } 1446 1447 if_public_preferred_lft = ifp->prefered_lft; 1448 1449 memset(&cfg, 0, sizeof(cfg)); 1450 cfg.valid_lft = min_t(__u32, ifp->valid_lft, 1451 READ_ONCE(idev->cnf.temp_valid_lft) + age); 1452 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor; 1453 cfg.preferred_lft = min_t(__u32, if_public_preferred_lft, cfg.preferred_lft); 1454 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft); 1455 1456 cfg.plen = ifp->prefix_len; 1457 tmp_tstamp = ifp->tstamp; 1458 spin_unlock_bh(&ifp->lock); 1459 1460 write_unlock_bh(&idev->lock); 1461 1462 /* From RFC 4941: 1463 * 1464 * A temporary address is created only if this calculated Preferred 1465 * Lifetime is greater than REGEN_ADVANCE time units. In 1466 * particular, an implementation must not create a temporary address 1467 * with a zero Preferred Lifetime. 1468 * 1469 * ... 1470 * 1471 * When creating a temporary address, the lifetime values MUST be 1472 * derived from the corresponding prefix as follows: 1473 * 1474 * ... 1475 * 1476 * * Its Preferred Lifetime is the lower of the Preferred Lifetime 1477 * of the public address or TEMP_PREFERRED_LIFETIME - 1478 * DESYNC_FACTOR. 1479 * 1480 * To comply with the RFC's requirements, clamp the preferred lifetime 1481 * to a minimum of regen_advance, unless that would exceed valid_lft or 1482 * ifp->prefered_lft. 1483 * 1484 * Use age calculation as in addrconf_verify to avoid unnecessary 1485 * temporary addresses being generated. 1486 */ 1487 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 1488 if (cfg.preferred_lft <= regen_advance + age) { 1489 cfg.preferred_lft = regen_advance + age + 1; 1490 if (cfg.preferred_lft <= regen_advance + age || 1491 cfg.preferred_lft > cfg.valid_lft || 1492 cfg.preferred_lft > if_public_preferred_lft) { 1493 in6_ifa_put(ifp); 1494 in6_dev_put(idev); 1495 ret = -1; 1496 goto out; 1497 } 1498 } 1499 1500 cfg.ifa_flags = IFA_F_TEMPORARY; 1501 /* set in addrconf_prefix_rcv() */ 1502 if (ifp->flags & IFA_F_OPTIMISTIC) 1503 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 1504 1505 cfg.pfx = &addr; 1506 cfg.scope = ipv6_addr_scope(cfg.pfx); 1507 cfg.ifpub = ifp; 1508 cfg.cstamp = now; 1509 cfg.tstamp = tmp_tstamp; 1510 1511 ift = ipv6_add_addr(idev, &cfg, block, NULL); 1512 if (IS_ERR(ift)) { 1513 in6_ifa_put(ifp); 1514 in6_dev_put(idev); 1515 pr_info("%s: retry temporary address regeneration\n", __func__); 1516 write_lock_bh(&idev->lock); 1517 goto retry; 1518 } 1519 1520 addrconf_dad_start(ift); 1521 in6_ifa_put(ift); 1522 in6_dev_put(idev); 1523 out: 1524 return ret; 1525 } 1526 1527 /* 1528 * Choose an appropriate source address (RFC3484) 1529 */ 1530 enum { 1531 IPV6_SADDR_RULE_INIT = 0, 1532 IPV6_SADDR_RULE_LOCAL, 1533 IPV6_SADDR_RULE_SCOPE, 1534 IPV6_SADDR_RULE_PREFERRED, 1535 #ifdef CONFIG_IPV6_MIP6 1536 IPV6_SADDR_RULE_HOA, 1537 #endif 1538 IPV6_SADDR_RULE_OIF, 1539 IPV6_SADDR_RULE_LABEL, 1540 IPV6_SADDR_RULE_PRIVACY, 1541 IPV6_SADDR_RULE_ORCHID, 1542 IPV6_SADDR_RULE_PREFIX, 1543 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1544 IPV6_SADDR_RULE_NOT_OPTIMISTIC, 1545 #endif 1546 IPV6_SADDR_RULE_MAX 1547 }; 1548 1549 struct ipv6_saddr_score { 1550 int rule; 1551 int addr_type; 1552 struct inet6_ifaddr *ifa; 1553 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX); 1554 int scopedist; 1555 int matchlen; 1556 }; 1557 1558 struct ipv6_saddr_dst { 1559 const struct in6_addr *addr; 1560 int ifindex; 1561 int scope; 1562 int label; 1563 unsigned int prefs; 1564 }; 1565 1566 static inline int ipv6_saddr_preferred(int type) 1567 { 1568 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK)) 1569 return 1; 1570 return 0; 1571 } 1572 1573 static bool ipv6_use_optimistic_addr(const struct net *net, 1574 const struct inet6_dev *idev) 1575 { 1576 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1577 if (!idev) 1578 return false; 1579 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) && 1580 !READ_ONCE(idev->cnf.optimistic_dad)) 1581 return false; 1582 if (!READ_ONCE(net->ipv6.devconf_all->use_optimistic) && 1583 !READ_ONCE(idev->cnf.use_optimistic)) 1584 return false; 1585 1586 return true; 1587 #else 1588 return false; 1589 #endif 1590 } 1591 1592 static bool ipv6_allow_optimistic_dad(const struct net *net, 1593 const struct inet6_dev *idev) 1594 { 1595 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1596 if (!idev) 1597 return false; 1598 if (!READ_ONCE(net->ipv6.devconf_all->optimistic_dad) && 1599 !READ_ONCE(idev->cnf.optimistic_dad)) 1600 return false; 1601 1602 return true; 1603 #else 1604 return false; 1605 #endif 1606 } 1607 1608 static int ipv6_get_saddr_eval(struct net *net, 1609 struct ipv6_saddr_score *score, 1610 struct ipv6_saddr_dst *dst, 1611 int i) 1612 { 1613 int ret; 1614 1615 if (i <= score->rule) { 1616 switch (i) { 1617 case IPV6_SADDR_RULE_SCOPE: 1618 ret = score->scopedist; 1619 break; 1620 case IPV6_SADDR_RULE_PREFIX: 1621 ret = score->matchlen; 1622 break; 1623 default: 1624 ret = !!test_bit(i, score->scorebits); 1625 } 1626 goto out; 1627 } 1628 1629 switch (i) { 1630 case IPV6_SADDR_RULE_INIT: 1631 /* Rule 0: remember if hiscore is not ready yet */ 1632 ret = !!score->ifa; 1633 break; 1634 case IPV6_SADDR_RULE_LOCAL: 1635 /* Rule 1: Prefer same address */ 1636 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr); 1637 break; 1638 case IPV6_SADDR_RULE_SCOPE: 1639 /* Rule 2: Prefer appropriate scope 1640 * 1641 * ret 1642 * ^ 1643 * -1 | d 15 1644 * ---+--+-+---> scope 1645 * | 1646 * | d is scope of the destination. 1647 * B-d | \ 1648 * | \ <- smaller scope is better if 1649 * B-15 | \ if scope is enough for destination. 1650 * | ret = B - scope (-1 <= scope >= d <= 15). 1651 * d-C-1 | / 1652 * |/ <- greater is better 1653 * -C / if scope is not enough for destination. 1654 * /| ret = scope - C (-1 <= d < scope <= 15). 1655 * 1656 * d - C - 1 < B -15 (for all -1 <= d <= 15). 1657 * C > d + 14 - B >= 15 + 14 - B = 29 - B. 1658 * Assume B = 0 and we get C > 29. 1659 */ 1660 ret = __ipv6_addr_src_scope(score->addr_type); 1661 if (ret >= dst->scope) 1662 ret = -ret; 1663 else 1664 ret -= 128; /* 30 is enough */ 1665 score->scopedist = ret; 1666 break; 1667 case IPV6_SADDR_RULE_PREFERRED: 1668 { 1669 /* Rule 3: Avoid deprecated and optimistic addresses */ 1670 u8 avoid = IFA_F_DEPRECATED; 1671 1672 if (!ipv6_use_optimistic_addr(net, score->ifa->idev)) 1673 avoid |= IFA_F_OPTIMISTIC; 1674 ret = ipv6_saddr_preferred(score->addr_type) || 1675 !(score->ifa->flags & avoid); 1676 break; 1677 } 1678 #ifdef CONFIG_IPV6_MIP6 1679 case IPV6_SADDR_RULE_HOA: 1680 { 1681 /* Rule 4: Prefer home address */ 1682 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA); 1683 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome; 1684 break; 1685 } 1686 #endif 1687 case IPV6_SADDR_RULE_OIF: 1688 /* Rule 5: Prefer outgoing interface */ 1689 ret = (!dst->ifindex || 1690 dst->ifindex == score->ifa->idev->dev->ifindex); 1691 break; 1692 case IPV6_SADDR_RULE_LABEL: 1693 /* Rule 6: Prefer matching label */ 1694 ret = ipv6_addr_label(net, 1695 &score->ifa->addr, score->addr_type, 1696 score->ifa->idev->dev->ifindex) == dst->label; 1697 break; 1698 case IPV6_SADDR_RULE_PRIVACY: 1699 { 1700 /* Rule 7: Prefer public address 1701 * Note: prefer temporary address if use_tempaddr >= 2 1702 */ 1703 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ? 1704 !!(dst->prefs & IPV6_PREFER_SRC_TMP) : 1705 READ_ONCE(score->ifa->idev->cnf.use_tempaddr) >= 2; 1706 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp; 1707 break; 1708 } 1709 case IPV6_SADDR_RULE_ORCHID: 1710 /* Rule 8-: Prefer ORCHID vs ORCHID or 1711 * non-ORCHID vs non-ORCHID 1712 */ 1713 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^ 1714 ipv6_addr_orchid(dst->addr)); 1715 break; 1716 case IPV6_SADDR_RULE_PREFIX: 1717 /* Rule 8: Use longest matching prefix */ 1718 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr); 1719 if (ret > score->ifa->prefix_len) 1720 ret = score->ifa->prefix_len; 1721 score->matchlen = ret; 1722 break; 1723 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1724 case IPV6_SADDR_RULE_NOT_OPTIMISTIC: 1725 /* Optimistic addresses still have lower precedence than other 1726 * preferred addresses. 1727 */ 1728 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC); 1729 break; 1730 #endif 1731 default: 1732 ret = 0; 1733 } 1734 1735 if (ret) 1736 __set_bit(i, score->scorebits); 1737 score->rule = i; 1738 out: 1739 return ret; 1740 } 1741 1742 static int __ipv6_dev_get_saddr(struct net *net, 1743 struct ipv6_saddr_dst *dst, 1744 struct inet6_dev *idev, 1745 struct ipv6_saddr_score *scores, 1746 int hiscore_idx) 1747 { 1748 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx]; 1749 1750 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) { 1751 int i; 1752 1753 /* 1754 * - Tentative Address (RFC2462 section 5.4) 1755 * - A tentative address is not considered 1756 * "assigned to an interface" in the traditional 1757 * sense, unless it is also flagged as optimistic. 1758 * - Candidate Source Address (section 4) 1759 * - In any case, anycast addresses, multicast 1760 * addresses, and the unspecified address MUST 1761 * NOT be included in a candidate set. 1762 */ 1763 if ((score->ifa->flags & IFA_F_TENTATIVE) && 1764 (!(score->ifa->flags & IFA_F_OPTIMISTIC))) 1765 continue; 1766 1767 score->addr_type = __ipv6_addr_type(&score->ifa->addr); 1768 1769 if (unlikely(score->addr_type == IPV6_ADDR_ANY || 1770 score->addr_type & IPV6_ADDR_MULTICAST)) { 1771 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s", 1772 idev->dev->name); 1773 continue; 1774 } 1775 1776 score->rule = -1; 1777 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX); 1778 1779 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) { 1780 int minihiscore, miniscore; 1781 1782 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i); 1783 miniscore = ipv6_get_saddr_eval(net, score, dst, i); 1784 1785 if (minihiscore > miniscore) { 1786 if (i == IPV6_SADDR_RULE_SCOPE && 1787 score->scopedist > 0) { 1788 /* 1789 * special case: 1790 * each remaining entry 1791 * has too small (not enough) 1792 * scope, because ifa entries 1793 * are sorted by their scope 1794 * values. 1795 */ 1796 goto out; 1797 } 1798 break; 1799 } else if (minihiscore < miniscore) { 1800 swap(hiscore, score); 1801 hiscore_idx = 1 - hiscore_idx; 1802 1803 /* restore our iterator */ 1804 score->ifa = hiscore->ifa; 1805 1806 break; 1807 } 1808 } 1809 } 1810 out: 1811 return hiscore_idx; 1812 } 1813 1814 static int ipv6_get_saddr_master(struct net *net, 1815 const struct net_device *dst_dev, 1816 const struct net_device *master, 1817 struct ipv6_saddr_dst *dst, 1818 struct ipv6_saddr_score *scores, 1819 int hiscore_idx) 1820 { 1821 struct inet6_dev *idev; 1822 1823 idev = __in6_dev_get(dst_dev); 1824 if (idev) 1825 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1826 scores, hiscore_idx); 1827 1828 idev = __in6_dev_get(master); 1829 if (idev) 1830 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1831 scores, hiscore_idx); 1832 1833 return hiscore_idx; 1834 } 1835 1836 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev, 1837 const struct in6_addr *daddr, unsigned int prefs, 1838 struct in6_addr *saddr) 1839 { 1840 struct ipv6_saddr_score scores[2], *hiscore; 1841 struct ipv6_saddr_dst dst; 1842 struct inet6_dev *idev; 1843 struct net_device *dev; 1844 int dst_type; 1845 bool use_oif_addr = false; 1846 int hiscore_idx = 0; 1847 int ret = 0; 1848 1849 dst_type = __ipv6_addr_type(daddr); 1850 dst.addr = daddr; 1851 dst.ifindex = dst_dev ? dst_dev->ifindex : 0; 1852 dst.scope = __ipv6_addr_src_scope(dst_type); 1853 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex); 1854 dst.prefs = prefs; 1855 1856 scores[hiscore_idx].rule = -1; 1857 scores[hiscore_idx].ifa = NULL; 1858 1859 rcu_read_lock(); 1860 1861 /* Candidate Source Address (section 4) 1862 * - multicast and link-local destination address, 1863 * the set of candidate source address MUST only 1864 * include addresses assigned to interfaces 1865 * belonging to the same link as the outgoing 1866 * interface. 1867 * (- For site-local destination addresses, the 1868 * set of candidate source addresses MUST only 1869 * include addresses assigned to interfaces 1870 * belonging to the same site as the outgoing 1871 * interface.) 1872 * - "It is RECOMMENDED that the candidate source addresses 1873 * be the set of unicast addresses assigned to the 1874 * interface that will be used to send to the destination 1875 * (the 'outgoing' interface)." (RFC 6724) 1876 */ 1877 if (dst_dev) { 1878 idev = __in6_dev_get(dst_dev); 1879 if ((dst_type & IPV6_ADDR_MULTICAST) || 1880 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL || 1881 (idev && READ_ONCE(idev->cnf.use_oif_addrs_only))) { 1882 use_oif_addr = true; 1883 } 1884 } 1885 1886 if (use_oif_addr) { 1887 if (idev) 1888 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1889 } else { 1890 const struct net_device *master; 1891 int master_idx = 0; 1892 1893 /* if dst_dev exists and is enslaved to an L3 device, then 1894 * prefer addresses from dst_dev and then the master over 1895 * any other enslaved devices in the L3 domain. 1896 */ 1897 master = l3mdev_master_dev_rcu(dst_dev); 1898 if (master) { 1899 master_idx = master->ifindex; 1900 1901 hiscore_idx = ipv6_get_saddr_master(net, dst_dev, 1902 master, &dst, 1903 scores, hiscore_idx); 1904 1905 if (scores[hiscore_idx].ifa && 1906 scores[hiscore_idx].scopedist >= 0) 1907 goto out; 1908 } 1909 1910 for_each_netdev_rcu(net, dev) { 1911 /* only consider addresses on devices in the 1912 * same L3 domain 1913 */ 1914 if (l3mdev_master_ifindex_rcu(dev) != master_idx) 1915 continue; 1916 idev = __in6_dev_get(dev); 1917 if (!idev) 1918 continue; 1919 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1920 } 1921 } 1922 1923 out: 1924 hiscore = &scores[hiscore_idx]; 1925 if (!hiscore->ifa) 1926 ret = -EADDRNOTAVAIL; 1927 else 1928 *saddr = hiscore->ifa->addr; 1929 1930 rcu_read_unlock(); 1931 return ret; 1932 } 1933 EXPORT_SYMBOL(ipv6_dev_get_saddr); 1934 1935 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr, 1936 u32 banned_flags) 1937 { 1938 struct inet6_ifaddr *ifp; 1939 int err = -EADDRNOTAVAIL; 1940 1941 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 1942 if (ifp->scope > IFA_LINK) 1943 break; 1944 if (ifp->scope == IFA_LINK && 1945 !(ifp->flags & banned_flags)) { 1946 *addr = ifp->addr; 1947 err = 0; 1948 break; 1949 } 1950 } 1951 return err; 1952 } 1953 1954 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr, 1955 u32 banned_flags) 1956 { 1957 struct inet6_dev *idev; 1958 int err = -EADDRNOTAVAIL; 1959 1960 rcu_read_lock(); 1961 idev = __in6_dev_get(dev); 1962 if (idev) { 1963 read_lock_bh(&idev->lock); 1964 err = __ipv6_get_lladdr(idev, addr, banned_flags); 1965 read_unlock_bh(&idev->lock); 1966 } 1967 rcu_read_unlock(); 1968 return err; 1969 } 1970 1971 static int ipv6_count_addresses(const struct inet6_dev *idev) 1972 { 1973 const struct inet6_ifaddr *ifp; 1974 int cnt = 0; 1975 1976 rcu_read_lock(); 1977 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list) 1978 cnt++; 1979 rcu_read_unlock(); 1980 return cnt; 1981 } 1982 1983 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr, 1984 const struct net_device *dev, int strict) 1985 { 1986 return ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1987 strict, IFA_F_TENTATIVE); 1988 } 1989 EXPORT_SYMBOL(ipv6_chk_addr); 1990 1991 /* device argument is used to find the L3 domain of interest. If 1992 * skip_dev_check is set, then the ifp device is not checked against 1993 * the passed in dev argument. So the 2 cases for addresses checks are: 1994 * 1. does the address exist in the L3 domain that dev is part of 1995 * (skip_dev_check = true), or 1996 * 1997 * 2. does the address exist on the specific device 1998 * (skip_dev_check = false) 1999 */ 2000 static struct net_device * 2001 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 2002 const struct net_device *dev, bool skip_dev_check, 2003 int strict, u32 banned_flags) 2004 { 2005 unsigned int hash = inet6_addr_hash(net, addr); 2006 struct net_device *l3mdev, *ndev; 2007 struct inet6_ifaddr *ifp; 2008 u32 ifp_flags; 2009 2010 rcu_read_lock(); 2011 2012 l3mdev = l3mdev_master_dev_rcu(dev); 2013 if (skip_dev_check) 2014 dev = NULL; 2015 2016 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 2017 ndev = ifp->idev->dev; 2018 2019 if (l3mdev_master_dev_rcu(ndev) != l3mdev) 2020 continue; 2021 2022 /* Decouple optimistic from tentative for evaluation here. 2023 * Ban optimistic addresses explicitly, when required. 2024 */ 2025 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC) 2026 ? (ifp->flags&~IFA_F_TENTATIVE) 2027 : ifp->flags; 2028 if (ipv6_addr_equal(&ifp->addr, addr) && 2029 !(ifp_flags&banned_flags) && 2030 (!dev || ndev == dev || 2031 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) { 2032 rcu_read_unlock(); 2033 return ndev; 2034 } 2035 } 2036 2037 rcu_read_unlock(); 2038 return NULL; 2039 } 2040 2041 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 2042 const struct net_device *dev, bool skip_dev_check, 2043 int strict, u32 banned_flags) 2044 { 2045 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check, 2046 strict, banned_flags) ? 1 : 0; 2047 } 2048 EXPORT_SYMBOL(ipv6_chk_addr_and_flags); 2049 2050 2051 /* Compares an address/prefix_len with addresses on device @dev. 2052 * If one is found it returns true. 2053 */ 2054 bool ipv6_chk_custom_prefix(const struct in6_addr *addr, 2055 const unsigned int prefix_len, struct net_device *dev) 2056 { 2057 const struct inet6_ifaddr *ifa; 2058 const struct inet6_dev *idev; 2059 bool ret = false; 2060 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 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len); 2066 if (ret) 2067 break; 2068 } 2069 } 2070 rcu_read_unlock(); 2071 2072 return ret; 2073 } 2074 EXPORT_SYMBOL(ipv6_chk_custom_prefix); 2075 2076 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev) 2077 { 2078 const struct inet6_ifaddr *ifa; 2079 const struct inet6_dev *idev; 2080 int onlink; 2081 2082 onlink = 0; 2083 rcu_read_lock(); 2084 idev = __in6_dev_get(dev); 2085 if (idev) { 2086 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 2087 onlink = ipv6_prefix_equal(addr, &ifa->addr, 2088 ifa->prefix_len); 2089 if (onlink) 2090 break; 2091 } 2092 } 2093 rcu_read_unlock(); 2094 return onlink; 2095 } 2096 EXPORT_SYMBOL(ipv6_chk_prefix); 2097 2098 /** 2099 * ipv6_dev_find - find the first device with a given source address. 2100 * @net: the net namespace 2101 * @addr: the source address 2102 * @dev: used to find the L3 domain of interest 2103 * 2104 * The caller should be protected by RCU, or RTNL. 2105 */ 2106 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr, 2107 struct net_device *dev) 2108 { 2109 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1, 2110 IFA_F_TENTATIVE); 2111 } 2112 EXPORT_SYMBOL(ipv6_dev_find); 2113 2114 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr, 2115 struct net_device *dev, int strict) 2116 { 2117 unsigned int hash = inet6_addr_hash(net, addr); 2118 struct inet6_ifaddr *ifp, *result = NULL; 2119 2120 rcu_read_lock(); 2121 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 2122 if (ipv6_addr_equal(&ifp->addr, addr)) { 2123 if (!dev || ifp->idev->dev == dev || 2124 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) { 2125 if (in6_ifa_hold_safe(ifp)) { 2126 result = ifp; 2127 break; 2128 } 2129 } 2130 } 2131 } 2132 rcu_read_unlock(); 2133 2134 return result; 2135 } 2136 2137 /* Gets referenced address, destroys ifaddr */ 2138 2139 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed) 2140 { 2141 if (dad_failed) 2142 ifp->flags |= IFA_F_DADFAILED; 2143 2144 if (ifp->flags&IFA_F_TEMPORARY) { 2145 struct inet6_ifaddr *ifpub; 2146 spin_lock_bh(&ifp->lock); 2147 ifpub = ifp->ifpub; 2148 if (ifpub) { 2149 in6_ifa_hold(ifpub); 2150 spin_unlock_bh(&ifp->lock); 2151 ipv6_create_tempaddr(ifpub, true); 2152 in6_ifa_put(ifpub); 2153 } else { 2154 spin_unlock_bh(&ifp->lock); 2155 } 2156 ipv6_del_addr(ifp); 2157 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) { 2158 spin_lock_bh(&ifp->lock); 2159 addrconf_del_dad_work(ifp); 2160 ifp->flags |= IFA_F_TENTATIVE; 2161 if (dad_failed) 2162 ifp->flags &= ~IFA_F_OPTIMISTIC; 2163 spin_unlock_bh(&ifp->lock); 2164 if (dad_failed) 2165 ipv6_ifa_notify(0, ifp); 2166 in6_ifa_put(ifp); 2167 } else { 2168 ipv6_del_addr(ifp); 2169 } 2170 } 2171 2172 static int addrconf_dad_end(struct inet6_ifaddr *ifp) 2173 { 2174 int err = -ENOENT; 2175 2176 spin_lock_bh(&ifp->lock); 2177 if (ifp->state == INET6_IFADDR_STATE_DAD) { 2178 ifp->state = INET6_IFADDR_STATE_POSTDAD; 2179 err = 0; 2180 } 2181 spin_unlock_bh(&ifp->lock); 2182 2183 return err; 2184 } 2185 2186 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp) 2187 { 2188 struct inet6_dev *idev = ifp->idev; 2189 struct net *net = dev_net(idev->dev); 2190 int max_addresses; 2191 2192 spin_lock_bh(&ifp->lock); 2193 2194 if (ifp->state != INET6_IFADDR_STATE_DAD) { 2195 spin_unlock_bh(&ifp->lock); 2196 in6_ifa_put(ifp); 2197 return; 2198 } 2199 ifp->state = INET6_IFADDR_STATE_POSTDAD; 2200 2201 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n", 2202 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source); 2203 2204 if (ifp->flags & IFA_F_STABLE_PRIVACY) { 2205 bool start_dad = false; 2206 struct in6_addr new_addr; 2207 struct inet6_ifaddr *ifp2; 2208 int retries = ifp->stable_privacy_retry + 1; 2209 struct ifa6_config cfg = { 2210 .pfx = &new_addr, 2211 .plen = ifp->prefix_len, 2212 .ifa_flags = ifp->flags, 2213 .valid_lft = ifp->valid_lft, 2214 .preferred_lft = ifp->prefered_lft, 2215 .scope = ifp->scope, 2216 }; 2217 2218 if (retries > net->ipv6.sysctl.idgen_retries) { 2219 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n", 2220 ifp->idev->dev->name); 2221 goto errdad; 2222 } 2223 2224 new_addr = ifp->addr; 2225 if (ipv6_generate_stable_address(&new_addr, retries, 2226 idev)) 2227 goto errdad; 2228 2229 spin_unlock_bh(&ifp->lock); 2230 2231 max_addresses = READ_ONCE(idev->cnf.max_addresses); 2232 if (max_addresses && 2233 ipv6_count_addresses(idev) >= max_addresses) 2234 goto lock_errdad; 2235 2236 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n", 2237 ifp->idev->dev->name); 2238 2239 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL); 2240 if (IS_ERR(ifp2)) 2241 goto lock_errdad; 2242 2243 spin_lock_bh(&ifp2->lock); 2244 if (ifp2->state != INET6_IFADDR_STATE_DEAD) { 2245 ifp2->stable_privacy_retry = retries; 2246 start_dad = true; 2247 } 2248 spin_unlock_bh(&ifp2->lock); 2249 2250 if (start_dad) 2251 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay); 2252 in6_ifa_put(ifp2); 2253 lock_errdad: 2254 spin_lock_bh(&ifp->lock); 2255 if (ifp->state != INET6_IFADDR_STATE_POSTDAD) { 2256 spin_unlock_bh(&ifp->lock); 2257 in6_ifa_put(ifp); 2258 return; 2259 } 2260 } 2261 2262 errdad: 2263 /* transition from _POSTDAD to _ERRDAD */ 2264 ifp->state = INET6_IFADDR_STATE_ERRDAD; 2265 spin_unlock_bh(&ifp->lock); 2266 2267 addrconf_mod_dad_work(ifp, 0); 2268 in6_ifa_put(ifp); 2269 } 2270 2271 /* Join to solicited addr multicast group. */ 2272 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr) 2273 { 2274 struct in6_addr maddr; 2275 2276 if (READ_ONCE(dev->flags) & (IFF_LOOPBACK | IFF_NOARP)) 2277 return; 2278 2279 addrconf_addr_solict_mult(addr, &maddr); 2280 ipv6_dev_mc_inc(dev, &maddr); 2281 } 2282 2283 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr) 2284 { 2285 struct in6_addr maddr; 2286 2287 if (READ_ONCE(idev->dev->flags) & (IFF_LOOPBACK | IFF_NOARP)) 2288 return; 2289 2290 addrconf_addr_solict_mult(addr, &maddr); 2291 __ipv6_dev_mc_dec(idev, &maddr); 2292 } 2293 2294 static void addrconf_join_anycast(struct inet6_ifaddr *ifp) 2295 { 2296 struct in6_addr addr; 2297 2298 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2299 return; 2300 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2301 if (ipv6_addr_any(&addr)) 2302 return; 2303 __ipv6_dev_ac_inc(ifp->idev, &addr); 2304 } 2305 2306 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp) 2307 { 2308 struct in6_addr addr; 2309 2310 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2311 return; 2312 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2313 if (ipv6_addr_any(&addr)) 2314 return; 2315 __ipv6_dev_ac_dec(ifp->idev, &addr); 2316 } 2317 2318 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev) 2319 { 2320 switch (dev->addr_len) { 2321 case ETH_ALEN: 2322 memcpy(eui, dev->dev_addr, 3); 2323 eui[3] = 0xFF; 2324 eui[4] = 0xFE; 2325 memcpy(eui + 5, dev->dev_addr + 3, 3); 2326 break; 2327 case EUI64_ADDR_LEN: 2328 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN); 2329 eui[0] ^= 2; 2330 break; 2331 default: 2332 return -1; 2333 } 2334 2335 return 0; 2336 } 2337 2338 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev) 2339 { 2340 const union fwnet_hwaddr *ha; 2341 2342 if (dev->addr_len != FWNET_ALEN) 2343 return -1; 2344 2345 ha = (const union fwnet_hwaddr *)dev->dev_addr; 2346 2347 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id)); 2348 eui[0] ^= 2; 2349 return 0; 2350 } 2351 2352 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev) 2353 { 2354 /* XXX: inherit EUI-64 from other interface -- yoshfuji */ 2355 if (dev->addr_len != ARCNET_ALEN) 2356 return -1; 2357 memset(eui, 0, 7); 2358 eui[7] = *(u8 *)dev->dev_addr; 2359 return 0; 2360 } 2361 2362 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev) 2363 { 2364 if (dev->addr_len != INFINIBAND_ALEN) 2365 return -1; 2366 memcpy(eui, dev->dev_addr + 12, 8); 2367 eui[0] |= 2; 2368 return 0; 2369 } 2370 2371 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr) 2372 { 2373 if (addr == 0) 2374 return -1; 2375 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) || 2376 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) || 2377 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) || 2378 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) || 2379 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) || 2380 ipv4_is_lbcast(addr)) ? 0x00 : 0x02; 2381 eui[1] = 0; 2382 eui[2] = 0x5E; 2383 eui[3] = 0xFE; 2384 memcpy(eui + 4, &addr, 4); 2385 return 0; 2386 } 2387 2388 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev) 2389 { 2390 if (dev->priv_flags & IFF_ISATAP) 2391 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2392 return -1; 2393 } 2394 2395 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev) 2396 { 2397 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2398 } 2399 2400 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev) 2401 { 2402 memcpy(eui, dev->perm_addr, 3); 2403 memcpy(eui + 5, dev->perm_addr + 3, 3); 2404 eui[3] = 0xFF; 2405 eui[4] = 0xFE; 2406 eui[0] ^= 2; 2407 return 0; 2408 } 2409 2410 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev) 2411 { 2412 switch (dev->type) { 2413 case ARPHRD_ETHER: 2414 case ARPHRD_FDDI: 2415 return addrconf_ifid_eui48(eui, dev); 2416 case ARPHRD_ARCNET: 2417 return addrconf_ifid_arcnet(eui, dev); 2418 case ARPHRD_INFINIBAND: 2419 return addrconf_ifid_infiniband(eui, dev); 2420 case ARPHRD_SIT: 2421 return addrconf_ifid_sit(eui, dev); 2422 case ARPHRD_IPGRE: 2423 case ARPHRD_TUNNEL: 2424 return addrconf_ifid_gre(eui, dev); 2425 case ARPHRD_6LOWPAN: 2426 return addrconf_ifid_6lowpan(eui, dev); 2427 case ARPHRD_IEEE1394: 2428 return addrconf_ifid_ieee1394(eui, dev); 2429 case ARPHRD_TUNNEL6: 2430 case ARPHRD_IP6GRE: 2431 case ARPHRD_RAWIP: 2432 return addrconf_ifid_ip6tnl(eui, dev); 2433 } 2434 return -1; 2435 } 2436 2437 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev) 2438 { 2439 int err = -1; 2440 struct inet6_ifaddr *ifp; 2441 2442 read_lock_bh(&idev->lock); 2443 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 2444 if (ifp->scope > IFA_LINK) 2445 break; 2446 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) { 2447 memcpy(eui, ifp->addr.s6_addr+8, 8); 2448 err = 0; 2449 break; 2450 } 2451 } 2452 read_unlock_bh(&idev->lock); 2453 return err; 2454 } 2455 2456 /* Generation of a randomized Interface Identifier 2457 * draft-ietf-6man-rfc4941bis, Section 3.3.1 2458 */ 2459 2460 static void ipv6_gen_rnd_iid(struct in6_addr *addr) 2461 { 2462 regen: 2463 get_random_bytes(&addr->s6_addr[8], 8); 2464 2465 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1: 2466 * check if generated address is not inappropriate: 2467 * 2468 * - Reserved IPv6 Interface Identifiers 2469 * - XXX: already assigned to an address on the device 2470 */ 2471 2472 /* Subnet-router anycast: 0000:0000:0000:0000 */ 2473 if (!(addr->s6_addr32[2] | addr->s6_addr32[3])) 2474 goto regen; 2475 2476 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212 2477 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213 2478 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF 2479 */ 2480 if (ntohl(addr->s6_addr32[2]) == 0x02005eff && 2481 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000) 2482 goto regen; 2483 2484 /* Reserved subnet anycast addresses */ 2485 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff && 2486 ntohl(addr->s6_addr32[3]) >= 0Xffffff80) 2487 goto regen; 2488 } 2489 2490 /* 2491 * Add prefix route. 2492 */ 2493 2494 static void 2495 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric, 2496 struct net_device *dev, unsigned long expires, 2497 u32 flags, gfp_t gfp_flags) 2498 { 2499 struct fib6_config cfg = { 2500 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX, 2501 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF, 2502 .fc_ifindex = dev->ifindex, 2503 .fc_expires = expires, 2504 .fc_dst_len = plen, 2505 .fc_flags = RTF_UP | flags, 2506 .fc_nlinfo.nl_net = dev_net(dev), 2507 .fc_protocol = RTPROT_KERNEL, 2508 .fc_type = RTN_UNICAST, 2509 }; 2510 2511 cfg.fc_dst = *pfx; 2512 2513 /* Prevent useless cloning on PtP SIT. 2514 This thing is done here expecting that the whole 2515 class of non-broadcast devices need not cloning. 2516 */ 2517 #if IS_ENABLED(CONFIG_IPV6_SIT) 2518 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT)) 2519 cfg.fc_flags |= RTF_NONEXTHOP; 2520 #endif 2521 2522 ip6_route_add(&cfg, gfp_flags, NULL); 2523 } 2524 2525 2526 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 2527 int plen, 2528 const struct net_device *dev, 2529 u32 flags, u32 noflags, 2530 bool no_gw) 2531 { 2532 struct fib6_node *fn; 2533 struct fib6_info *rt = NULL; 2534 struct fib6_table *table; 2535 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX; 2536 2537 table = fib6_get_table(dev_net(dev), tb_id); 2538 if (!table) 2539 return NULL; 2540 2541 rcu_read_lock(); 2542 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true); 2543 if (!fn) 2544 goto out; 2545 2546 for_each_fib6_node_rt_rcu(fn) { 2547 /* prefix routes only use builtin fib6_nh */ 2548 if (rt->nh) 2549 continue; 2550 2551 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex) 2552 continue; 2553 if (no_gw && rt->fib6_nh->fib_nh_gw_family) 2554 continue; 2555 if ((rt->fib6_flags & flags) != flags) 2556 continue; 2557 if ((rt->fib6_flags & noflags) != 0) 2558 continue; 2559 if (!fib6_info_hold_safe(rt)) 2560 continue; 2561 break; 2562 } 2563 out: 2564 rcu_read_unlock(); 2565 return rt; 2566 } 2567 2568 2569 /* Create "default" multicast route to the interface */ 2570 2571 static void addrconf_add_mroute(struct net_device *dev) 2572 { 2573 struct fib6_config cfg = { 2574 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL, 2575 .fc_metric = IP6_RT_PRIO_ADDRCONF, 2576 .fc_ifindex = dev->ifindex, 2577 .fc_dst_len = 8, 2578 .fc_flags = RTF_UP, 2579 .fc_type = RTN_MULTICAST, 2580 .fc_nlinfo.nl_net = dev_net(dev), 2581 .fc_protocol = RTPROT_KERNEL, 2582 }; 2583 2584 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0); 2585 2586 ip6_route_add(&cfg, GFP_KERNEL, NULL); 2587 } 2588 2589 static struct inet6_dev *addrconf_add_dev(struct net_device *dev) 2590 { 2591 struct inet6_dev *idev; 2592 2593 ASSERT_RTNL(); 2594 2595 idev = ipv6_find_idev(dev); 2596 if (IS_ERR(idev)) 2597 return idev; 2598 2599 if (idev->cnf.disable_ipv6) 2600 return ERR_PTR(-EACCES); 2601 2602 /* Add default multicast route */ 2603 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev)) 2604 addrconf_add_mroute(dev); 2605 2606 return idev; 2607 } 2608 2609 static void delete_tempaddrs(struct inet6_dev *idev, 2610 struct inet6_ifaddr *ifp) 2611 { 2612 struct inet6_ifaddr *ift, *tmp; 2613 2614 write_lock_bh(&idev->lock); 2615 list_for_each_entry_safe(ift, tmp, &idev->tempaddr_list, tmp_list) { 2616 if (ift->ifpub != ifp) 2617 continue; 2618 2619 in6_ifa_hold(ift); 2620 write_unlock_bh(&idev->lock); 2621 ipv6_del_addr(ift); 2622 write_lock_bh(&idev->lock); 2623 } 2624 write_unlock_bh(&idev->lock); 2625 } 2626 2627 static void manage_tempaddrs(struct inet6_dev *idev, 2628 struct inet6_ifaddr *ifp, 2629 __u32 valid_lft, __u32 prefered_lft, 2630 bool create, unsigned long now) 2631 { 2632 u32 orig_prefered_lft = prefered_lft; 2633 struct inet6_ifaddr *ift; 2634 bool reset_done = false; 2635 u32 flags; 2636 2637 read_lock_bh(&idev->lock); 2638 /* update all temporary addresses in the list */ 2639 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) { 2640 int age, max_valid, max_prefered; 2641 2642 if (ifp != ift->ifpub) 2643 continue; 2644 2645 /* RFC 4941 section 3.3: 2646 * If a received option will extend the lifetime of a public 2647 * address, the lifetimes of temporary addresses should 2648 * be extended, subject to the overall constraint that no 2649 * temporary addresses should ever remain "valid" or "preferred" 2650 * for a time longer than (TEMP_VALID_LIFETIME) or 2651 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively. 2652 */ 2653 age = (now - ift->cstamp) / HZ; 2654 max_valid = READ_ONCE(idev->cnf.temp_valid_lft) - age; 2655 if (max_valid < 0) 2656 max_valid = 0; 2657 2658 max_prefered = READ_ONCE(idev->cnf.temp_prefered_lft) - 2659 idev->desync_factor - age; 2660 if (max_prefered < 0) 2661 max_prefered = 0; 2662 2663 if (valid_lft > max_valid) 2664 valid_lft = max_valid; 2665 2666 if (prefered_lft > max_prefered) 2667 prefered_lft = max_prefered; 2668 2669 spin_lock(&ift->lock); 2670 /* the first match is the most recent temp address */ 2671 if (!reset_done && orig_prefered_lft > 0) { 2672 ift->regen_count = 0; 2673 reset_done = true; 2674 } 2675 flags = ift->flags; 2676 ift->valid_lft = valid_lft; 2677 ift->prefered_lft = prefered_lft; 2678 ift->tstamp = now; 2679 if (prefered_lft > 0) 2680 ift->flags &= ~IFA_F_DEPRECATED; 2681 2682 spin_unlock(&ift->lock); 2683 if (!(flags&IFA_F_TENTATIVE)) 2684 ipv6_ifa_notify(0, ift); 2685 } 2686 2687 /* Also create a temporary address if it's enabled but no temporary 2688 * address currently exists. 2689 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false 2690 * as part of cleanup (ie. deleting the mngtmpaddr). 2691 * We don't want that to result in creating a new temporary ip address. 2692 */ 2693 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft)) 2694 create = true; 2695 2696 if (create && READ_ONCE(idev->cnf.use_tempaddr) > 0) { 2697 /* When a new public address is created as described 2698 * in [ADDRCONF], also create a new temporary address. 2699 */ 2700 read_unlock_bh(&idev->lock); 2701 ipv6_create_tempaddr(ifp, false); 2702 } else { 2703 read_unlock_bh(&idev->lock); 2704 } 2705 } 2706 2707 static bool is_addr_mode_generate_stable(struct inet6_dev *idev) 2708 { 2709 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY || 2710 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM; 2711 } 2712 2713 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev, 2714 const struct prefix_info *pinfo, 2715 struct inet6_dev *in6_dev, 2716 const struct in6_addr *addr, int addr_type, 2717 u32 addr_flags, bool sllao, bool tokenized, 2718 __u32 valid_lft, u32 prefered_lft) 2719 { 2720 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1); 2721 int create = 0, update_lft = 0; 2722 2723 if (!ifp && valid_lft) { 2724 int max_addresses = READ_ONCE(in6_dev->cnf.max_addresses); 2725 struct ifa6_config cfg = { 2726 .pfx = addr, 2727 .plen = pinfo->prefix_len, 2728 .ifa_flags = addr_flags, 2729 .valid_lft = valid_lft, 2730 .preferred_lft = prefered_lft, 2731 .scope = addr_type & IPV6_ADDR_SCOPE_MASK, 2732 .ifa_proto = IFAPROT_KERNEL_RA 2733 }; 2734 2735 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 2736 if ((READ_ONCE(net->ipv6.devconf_all->optimistic_dad) || 2737 READ_ONCE(in6_dev->cnf.optimistic_dad)) && 2738 !net->ipv6.devconf_all->forwarding && sllao) 2739 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 2740 #endif 2741 2742 /* Do not allow to create too much of autoconfigured 2743 * addresses; this would be too easy way to crash kernel. 2744 */ 2745 if (!max_addresses || 2746 ipv6_count_addresses(in6_dev) < max_addresses) 2747 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL); 2748 2749 if (IS_ERR_OR_NULL(ifp)) 2750 return -1; 2751 2752 create = 1; 2753 spin_lock_bh(&ifp->lock); 2754 ifp->flags |= IFA_F_MANAGETEMPADDR; 2755 ifp->cstamp = jiffies; 2756 ifp->tokenized = tokenized; 2757 spin_unlock_bh(&ifp->lock); 2758 addrconf_dad_start(ifp); 2759 } 2760 2761 if (ifp) { 2762 u32 flags; 2763 unsigned long now; 2764 u32 stored_lft; 2765 2766 /* update lifetime (RFC2462 5.5.3 e) */ 2767 spin_lock_bh(&ifp->lock); 2768 now = jiffies; 2769 if (ifp->valid_lft > (now - ifp->tstamp) / HZ) 2770 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ; 2771 else 2772 stored_lft = 0; 2773 2774 /* RFC4862 Section 5.5.3e: 2775 * "Note that the preferred lifetime of the 2776 * corresponding address is always reset to 2777 * the Preferred Lifetime in the received 2778 * Prefix Information option, regardless of 2779 * whether the valid lifetime is also reset or 2780 * ignored." 2781 * 2782 * So we should always update prefered_lft here. 2783 */ 2784 update_lft = !create && stored_lft; 2785 2786 if (update_lft && !READ_ONCE(in6_dev->cnf.ra_honor_pio_life)) { 2787 const u32 minimum_lft = min_t(u32, 2788 stored_lft, MIN_VALID_LIFETIME); 2789 valid_lft = max(valid_lft, minimum_lft); 2790 } 2791 2792 if (update_lft) { 2793 ifp->valid_lft = valid_lft; 2794 ifp->prefered_lft = prefered_lft; 2795 WRITE_ONCE(ifp->tstamp, now); 2796 flags = ifp->flags; 2797 ifp->flags &= ~IFA_F_DEPRECATED; 2798 spin_unlock_bh(&ifp->lock); 2799 2800 if (!(flags&IFA_F_TENTATIVE)) 2801 ipv6_ifa_notify(0, ifp); 2802 } else 2803 spin_unlock_bh(&ifp->lock); 2804 2805 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft, 2806 create, now); 2807 2808 in6_ifa_put(ifp); 2809 addrconf_verify(net); 2810 } 2811 2812 return 0; 2813 } 2814 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr); 2815 2816 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao) 2817 { 2818 struct prefix_info *pinfo; 2819 struct fib6_table *table; 2820 __u32 valid_lft; 2821 __u32 prefered_lft; 2822 int addr_type, err; 2823 u32 addr_flags = 0; 2824 struct inet6_dev *in6_dev; 2825 struct net *net = dev_net(dev); 2826 bool ignore_autoconf = false; 2827 2828 pinfo = (struct prefix_info *) opt; 2829 2830 if (len < sizeof(struct prefix_info)) { 2831 netdev_dbg(dev, "addrconf: prefix option too short\n"); 2832 return; 2833 } 2834 2835 /* 2836 * Validation checks ([ADDRCONF], page 19) 2837 */ 2838 2839 addr_type = ipv6_addr_type(&pinfo->prefix); 2840 2841 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)) 2842 return; 2843 2844 valid_lft = ntohl(pinfo->valid); 2845 prefered_lft = ntohl(pinfo->prefered); 2846 2847 if (prefered_lft > valid_lft) { 2848 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n"); 2849 return; 2850 } 2851 2852 in6_dev = in6_dev_get(dev); 2853 2854 if (!in6_dev) { 2855 net_dbg_ratelimited("addrconf: device %s not configured\n", 2856 dev->name); 2857 return; 2858 } 2859 2860 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft) 2861 goto put; 2862 2863 /* 2864 * Two things going on here: 2865 * 1) Add routes for on-link prefixes 2866 * 2) Configure prefixes with the auto flag set 2867 */ 2868 2869 if (pinfo->onlink) { 2870 struct fib6_info *rt; 2871 unsigned long rt_expires; 2872 2873 /* Avoid arithmetic overflow. Really, we could 2874 * save rt_expires in seconds, likely valid_lft, 2875 * but it would require division in fib gc, that it 2876 * not good. 2877 */ 2878 if (HZ > USER_HZ) 2879 rt_expires = addrconf_timeout_fixup(valid_lft, HZ); 2880 else 2881 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ); 2882 2883 if (addrconf_finite_timeout(rt_expires)) 2884 rt_expires *= HZ; 2885 2886 rt = addrconf_get_prefix_route(&pinfo->prefix, 2887 pinfo->prefix_len, 2888 dev, 2889 RTF_ADDRCONF | RTF_PREFIX_RT, 2890 RTF_DEFAULT, true); 2891 2892 if (rt) { 2893 /* Autoconf prefix route */ 2894 if (valid_lft == 0) { 2895 ip6_del_rt_reason(net, rt, 2896 RT_DEL_REASON_RA_WITHDRAWN); 2897 rt = NULL; 2898 } else { 2899 table = rt->fib6_table; 2900 spin_lock_bh(&table->tb6_lock); 2901 2902 if (addrconf_finite_timeout(rt_expires)) { 2903 /* not infinity */ 2904 fib6_set_expires(rt, jiffies + rt_expires); 2905 fib6_add_gc_list(rt); 2906 } else { 2907 fib6_clean_expires(rt); 2908 fib6_may_remove_gc_list(net, rt); 2909 } 2910 2911 spin_unlock_bh(&table->tb6_lock); 2912 } 2913 } else if (valid_lft) { 2914 clock_t expires = 0; 2915 int flags = RTF_ADDRCONF | RTF_PREFIX_RT; 2916 if (addrconf_finite_timeout(rt_expires)) { 2917 /* not infinity */ 2918 flags |= RTF_EXPIRES; 2919 expires = jiffies_to_clock_t(rt_expires); 2920 } 2921 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len, 2922 0, dev, expires, flags, 2923 GFP_ATOMIC); 2924 } 2925 fib6_info_release(rt); 2926 } 2927 2928 /* Try to figure out our local address for this prefix */ 2929 2930 ignore_autoconf = READ_ONCE(in6_dev->cnf.ra_honor_pio_pflag) && pinfo->preferpd; 2931 if (pinfo->autoconf && in6_dev->cnf.autoconf && !ignore_autoconf) { 2932 struct in6_addr addr; 2933 bool tokenized = false, dev_addr_generated = false; 2934 2935 if (pinfo->prefix_len == 64) { 2936 memcpy(&addr, &pinfo->prefix, 8); 2937 2938 if (!ipv6_addr_any(&in6_dev->token)) { 2939 read_lock_bh(&in6_dev->lock); 2940 memcpy(addr.s6_addr + 8, 2941 in6_dev->token.s6_addr + 8, 8); 2942 read_unlock_bh(&in6_dev->lock); 2943 tokenized = true; 2944 } else if (is_addr_mode_generate_stable(in6_dev) && 2945 !ipv6_generate_stable_address(&addr, 0, 2946 in6_dev)) { 2947 addr_flags |= IFA_F_STABLE_PRIVACY; 2948 goto ok; 2949 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) && 2950 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) { 2951 goto put; 2952 } else { 2953 dev_addr_generated = true; 2954 } 2955 goto ok; 2956 } 2957 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n", 2958 pinfo->prefix_len); 2959 goto put; 2960 2961 ok: 2962 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, 2963 &addr, addr_type, 2964 addr_flags, sllao, 2965 tokenized, valid_lft, 2966 prefered_lft); 2967 if (err) 2968 goto put; 2969 2970 /* Ignore error case here because previous prefix add addr was 2971 * successful which will be notified. 2972 */ 2973 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr, 2974 addr_type, addr_flags, sllao, 2975 tokenized, valid_lft, 2976 prefered_lft, 2977 dev_addr_generated); 2978 } 2979 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo); 2980 put: 2981 in6_dev_put(in6_dev); 2982 } 2983 2984 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev, 2985 struct in6_ifreq *ireq) 2986 { 2987 struct ip_tunnel_parm_kern p = { }; 2988 int err; 2989 2990 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4)) 2991 return -EADDRNOTAVAIL; 2992 2993 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3]; 2994 p.iph.version = 4; 2995 p.iph.ihl = 5; 2996 p.iph.protocol = IPPROTO_IPV6; 2997 p.iph.ttl = 64; 2998 2999 if (!dev->netdev_ops->ndo_tunnel_ctl) 3000 return -EOPNOTSUPP; 3001 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL); 3002 if (err) 3003 return err; 3004 3005 dev = __dev_get_by_name(net, p.name); 3006 if (!dev) 3007 return -ENOBUFS; 3008 return dev_open(dev, NULL); 3009 } 3010 3011 /* 3012 * Set destination address. 3013 * Special case for SIT interfaces where we create a new "virtual" 3014 * device. 3015 */ 3016 int addrconf_set_dstaddr(struct net *net, void __user *arg) 3017 { 3018 struct net_device *dev; 3019 struct in6_ifreq ireq; 3020 int err = -ENODEV; 3021 3022 if (!IS_ENABLED(CONFIG_IPV6_SIT)) 3023 return -ENODEV; 3024 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3025 return -EFAULT; 3026 3027 rtnl_net_lock(net); 3028 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 3029 if (dev && dev->type == ARPHRD_SIT) 3030 err = addrconf_set_sit_dstaddr(net, dev, &ireq); 3031 rtnl_net_unlock(net); 3032 return err; 3033 } 3034 3035 static int ipv6_mc_config(struct sock *sk, bool join, 3036 const struct in6_addr *addr, int ifindex) 3037 { 3038 int ret; 3039 3040 ASSERT_RTNL(); 3041 3042 lock_sock(sk); 3043 if (join) 3044 ret = ipv6_sock_mc_join(sk, ifindex, addr); 3045 else 3046 ret = ipv6_sock_mc_drop(sk, ifindex, addr); 3047 release_sock(sk); 3048 3049 return ret; 3050 } 3051 3052 /* 3053 * Manual configuration of address on an interface 3054 */ 3055 static int inet6_addr_add(struct net *net, struct net_device *dev, 3056 struct ifa6_config *cfg, clock_t expires, u32 flags, 3057 struct netlink_ext_ack *extack) 3058 { 3059 struct inet6_ifaddr *ifp; 3060 struct inet6_dev *idev; 3061 3062 ASSERT_RTNL_NET(net); 3063 3064 if (cfg->plen > 128) { 3065 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 3066 return -EINVAL; 3067 } 3068 3069 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) { 3070 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64"); 3071 return -EINVAL; 3072 } 3073 3074 idev = addrconf_add_dev(dev); 3075 if (IS_ERR(idev)) { 3076 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 3077 return PTR_ERR(idev); 3078 } 3079 3080 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 3081 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk, 3082 true, cfg->pfx, dev->ifindex); 3083 3084 if (ret < 0) { 3085 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed"); 3086 return ret; 3087 } 3088 } 3089 3090 cfg->scope = ipv6_addr_scope(cfg->pfx); 3091 3092 ifp = ipv6_add_addr(idev, cfg, true, extack); 3093 if (!IS_ERR(ifp)) { 3094 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 3095 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3096 ifp->rt_priority, dev, expires, 3097 flags, GFP_KERNEL); 3098 } 3099 3100 /* Send a netlink notification if DAD is enabled and 3101 * optimistic flag is not set 3102 */ 3103 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD))) 3104 ipv6_ifa_notify(0, ifp); 3105 /* 3106 * Note that section 3.1 of RFC 4429 indicates 3107 * that the Optimistic flag should not be set for 3108 * manually configured addresses 3109 */ 3110 addrconf_dad_start(ifp); 3111 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR) 3112 manage_tempaddrs(idev, ifp, cfg->valid_lft, 3113 cfg->preferred_lft, true, jiffies); 3114 in6_ifa_put(ifp); 3115 addrconf_verify_rtnl(net); 3116 return 0; 3117 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 3118 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false, 3119 cfg->pfx, dev->ifindex); 3120 } 3121 3122 return PTR_ERR(ifp); 3123 } 3124 3125 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags, 3126 const struct in6_addr *pfx, unsigned int plen, 3127 struct netlink_ext_ack *extack) 3128 { 3129 struct inet6_ifaddr *ifp; 3130 struct inet6_dev *idev; 3131 struct net_device *dev; 3132 3133 if (plen > 128) { 3134 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 3135 return -EINVAL; 3136 } 3137 3138 dev = __dev_get_by_index(net, ifindex); 3139 if (!dev) { 3140 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 3141 return -ENODEV; 3142 } 3143 3144 idev = __in6_dev_get_rtnl_net(dev); 3145 if (!idev) { 3146 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 3147 return -ENXIO; 3148 } 3149 3150 read_lock_bh(&idev->lock); 3151 list_for_each_entry(ifp, &idev->addr_list, if_list) { 3152 if (ifp->prefix_len == plen && 3153 ipv6_addr_equal(pfx, &ifp->addr)) { 3154 in6_ifa_hold(ifp); 3155 read_unlock_bh(&idev->lock); 3156 3157 if (!(ifp->flags & IFA_F_TEMPORARY) && 3158 (ifp->flags & IFA_F_MANAGETEMPADDR)) 3159 delete_tempaddrs(idev, ifp); 3160 3161 ipv6_del_addr(ifp); 3162 3163 addrconf_verify_rtnl(net); 3164 if (ipv6_addr_is_multicast(pfx)) { 3165 ipv6_mc_config(net->ipv6.mc_autojoin_sk, 3166 false, pfx, dev->ifindex); 3167 } 3168 return 0; 3169 } 3170 } 3171 read_unlock_bh(&idev->lock); 3172 3173 NL_SET_ERR_MSG_MOD(extack, "address not found"); 3174 return -EADDRNOTAVAIL; 3175 } 3176 3177 3178 int addrconf_add_ifaddr(struct net *net, void __user *arg) 3179 { 3180 struct ifa6_config cfg = { 3181 .ifa_flags = IFA_F_PERMANENT, 3182 .preferred_lft = INFINITY_LIFE_TIME, 3183 .valid_lft = INFINITY_LIFE_TIME, 3184 }; 3185 struct net_device *dev; 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 cfg.pfx = &ireq.ifr6_addr; 3196 cfg.plen = ireq.ifr6_prefixlen; 3197 3198 rtnl_net_lock(net); 3199 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 3200 if (dev) { 3201 netdev_lock_ops(dev); 3202 err = inet6_addr_add(net, dev, &cfg, 0, 0, NULL); 3203 netdev_unlock_ops(dev); 3204 } else { 3205 err = -ENODEV; 3206 } 3207 rtnl_net_unlock(net); 3208 return err; 3209 } 3210 3211 int addrconf_del_ifaddr(struct net *net, void __user *arg) 3212 { 3213 struct in6_ifreq ireq; 3214 int err; 3215 3216 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 3217 return -EPERM; 3218 3219 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3220 return -EFAULT; 3221 3222 rtnl_net_lock(net); 3223 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr, 3224 ireq.ifr6_prefixlen, NULL); 3225 rtnl_net_unlock(net); 3226 return err; 3227 } 3228 3229 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr, 3230 int plen, int scope, u8 proto) 3231 { 3232 struct inet6_ifaddr *ifp; 3233 struct ifa6_config cfg = { 3234 .pfx = addr, 3235 .plen = plen, 3236 .ifa_flags = IFA_F_PERMANENT, 3237 .valid_lft = INFINITY_LIFE_TIME, 3238 .preferred_lft = INFINITY_LIFE_TIME, 3239 .scope = scope, 3240 .ifa_proto = proto 3241 }; 3242 3243 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3244 if (!IS_ERR(ifp)) { 3245 spin_lock_bh(&ifp->lock); 3246 ifp->flags &= ~IFA_F_TENTATIVE; 3247 spin_unlock_bh(&ifp->lock); 3248 rt_genid_bump_ipv6(dev_net(idev->dev)); 3249 ipv6_ifa_notify(RTM_NEWADDR, ifp); 3250 in6_ifa_put(ifp); 3251 } 3252 } 3253 3254 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) 3255 static void add_v4_addrs(struct inet6_dev *idev) 3256 { 3257 struct in6_addr addr; 3258 struct net_device *dev; 3259 struct net *net = dev_net(idev->dev); 3260 int scope, plen; 3261 u32 pflags = 0; 3262 3263 ASSERT_RTNL(); 3264 3265 memset(&addr, 0, sizeof(struct in6_addr)); 3266 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr, 4); 3267 3268 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) { 3269 scope = IPV6_ADDR_COMPATv4; 3270 plen = 96; 3271 pflags |= RTF_NONEXTHOP; 3272 } else { 3273 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE) 3274 return; 3275 3276 addr.s6_addr32[0] = htonl(0xfe800000); 3277 scope = IFA_LINK; 3278 plen = 64; 3279 } 3280 3281 if (addr.s6_addr32[3]) { 3282 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC); 3283 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags, 3284 GFP_KERNEL); 3285 return; 3286 } 3287 3288 for_each_netdev(net, dev) { 3289 struct in_device *in_dev = __in_dev_get_rtnl(dev); 3290 if (in_dev && (dev->flags & IFF_UP)) { 3291 struct in_ifaddr *ifa; 3292 int flag = scope; 3293 3294 in_dev_for_each_ifa_rtnl(ifa, in_dev) { 3295 addr.s6_addr32[3] = ifa->ifa_local; 3296 3297 if (ifa->ifa_scope == RT_SCOPE_LINK) 3298 continue; 3299 if (ifa->ifa_scope >= RT_SCOPE_HOST) { 3300 if (idev->dev->flags&IFF_POINTOPOINT) 3301 continue; 3302 flag |= IFA_HOST; 3303 } 3304 3305 add_addr(idev, &addr, plen, flag, 3306 IFAPROT_UNSPEC); 3307 addrconf_prefix_route(&addr, plen, 0, idev->dev, 3308 0, pflags, GFP_KERNEL); 3309 } 3310 } 3311 } 3312 } 3313 #endif 3314 3315 static void init_loopback(struct net_device *dev) 3316 { 3317 struct inet6_dev *idev; 3318 3319 /* ::1 */ 3320 3321 ASSERT_RTNL(); 3322 3323 idev = ipv6_find_idev(dev); 3324 if (IS_ERR(idev)) { 3325 pr_debug("%s: add_dev failed\n", __func__); 3326 return; 3327 } 3328 3329 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO); 3330 } 3331 3332 void addrconf_add_linklocal(struct inet6_dev *idev, 3333 const struct in6_addr *addr, u32 flags) 3334 { 3335 struct ifa6_config cfg = { 3336 .pfx = addr, 3337 .plen = 64, 3338 .ifa_flags = flags | IFA_F_PERMANENT, 3339 .valid_lft = INFINITY_LIFE_TIME, 3340 .preferred_lft = INFINITY_LIFE_TIME, 3341 .scope = IFA_LINK, 3342 .ifa_proto = IFAPROT_KERNEL_LL 3343 }; 3344 struct inet6_ifaddr *ifp; 3345 3346 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 3347 if ((READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad) || 3348 READ_ONCE(idev->cnf.optimistic_dad)) && 3349 !dev_net(idev->dev)->ipv6.devconf_all->forwarding) 3350 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 3351 #endif 3352 3353 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3354 if (!IS_ERR(ifp)) { 3355 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev, 3356 0, 0, GFP_ATOMIC); 3357 addrconf_dad_start(ifp); 3358 in6_ifa_put(ifp); 3359 } 3360 } 3361 EXPORT_SYMBOL_GPL(addrconf_add_linklocal); 3362 3363 static bool ipv6_reserved_interfaceid(struct in6_addr address) 3364 { 3365 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0) 3366 return true; 3367 3368 if (address.s6_addr32[2] == htonl(0x02005eff) && 3369 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000))) 3370 return true; 3371 3372 if (address.s6_addr32[2] == htonl(0xfdffffff) && 3373 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80))) 3374 return true; 3375 3376 return false; 3377 } 3378 3379 static int ipv6_generate_stable_address(struct in6_addr *address, 3380 u8 dad_count, 3381 const struct inet6_dev *idev) 3382 { 3383 static DEFINE_SPINLOCK(lock); 3384 static struct sha1_ctx sha_ctx; 3385 3386 static union { 3387 u8 __data[SHA1_BLOCK_SIZE]; 3388 struct { 3389 struct in6_addr secret; 3390 __be32 prefix[2]; 3391 unsigned char hwaddr[MAX_ADDR_LEN]; 3392 u8 dad_count; 3393 } __packed; 3394 } data; 3395 3396 struct in6_addr secret; 3397 struct in6_addr temp; 3398 struct net *net = dev_net(idev->dev); 3399 3400 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data)); 3401 3402 if (idev->cnf.stable_secret.initialized) 3403 secret = idev->cnf.stable_secret.secret; 3404 else if (net->ipv6.devconf_dflt->stable_secret.initialized) 3405 secret = net->ipv6.devconf_dflt->stable_secret.secret; 3406 else 3407 return -1; 3408 3409 retry: 3410 spin_lock_bh(&lock); 3411 3412 sha1_init(&sha_ctx); 3413 3414 memset(&data, 0, sizeof(data)); 3415 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len); 3416 data.prefix[0] = address->s6_addr32[0]; 3417 data.prefix[1] = address->s6_addr32[1]; 3418 data.secret = secret; 3419 data.dad_count = dad_count; 3420 3421 sha1_update(&sha_ctx, data.__data, sizeof(data)); 3422 3423 /* 3424 * Note that the SHA-1 finalization is omitted here, and the digest is 3425 * pulled directly from the internal SHA-1 state (making it incompatible 3426 * with standard SHA-1). Unusual, but technically okay since the data 3427 * length is fixed and is a multiple of the SHA-1 block size. 3428 */ 3429 temp = *address; 3430 temp.s6_addr32[2] = (__force __be32)sha_ctx.state.h[0]; 3431 temp.s6_addr32[3] = (__force __be32)sha_ctx.state.h[1]; 3432 3433 spin_unlock_bh(&lock); 3434 3435 if (ipv6_reserved_interfaceid(temp)) { 3436 dad_count++; 3437 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries) 3438 return -1; 3439 goto retry; 3440 } 3441 3442 *address = temp; 3443 return 0; 3444 } 3445 3446 static void ipv6_gen_mode_random_init(struct inet6_dev *idev) 3447 { 3448 struct ipv6_stable_secret *s = &idev->cnf.stable_secret; 3449 3450 if (s->initialized) 3451 return; 3452 s = &idev->cnf.stable_secret; 3453 get_random_bytes(&s->secret, sizeof(s->secret)); 3454 s->initialized = true; 3455 } 3456 3457 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route) 3458 { 3459 struct in6_addr addr; 3460 3461 /* no link local addresses on L3 master devices */ 3462 if (netif_is_l3_master(idev->dev)) 3463 return; 3464 3465 /* no link local addresses on devices flagged as slaves */ 3466 if (idev->dev->priv_flags & IFF_NO_ADDRCONF) 3467 return; 3468 3469 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0); 3470 3471 switch (idev->cnf.addr_gen_mode) { 3472 case IN6_ADDR_GEN_MODE_RANDOM: 3473 ipv6_gen_mode_random_init(idev); 3474 fallthrough; 3475 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY: 3476 if (!ipv6_generate_stable_address(&addr, 0, idev)) 3477 addrconf_add_linklocal(idev, &addr, 3478 IFA_F_STABLE_PRIVACY); 3479 else if (prefix_route) 3480 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3481 0, 0, GFP_KERNEL); 3482 break; 3483 case IN6_ADDR_GEN_MODE_EUI64: 3484 /* addrconf_add_linklocal also adds a prefix_route and we 3485 * only need to care about prefix routes if ipv6_generate_eui64 3486 * couldn't generate one. 3487 */ 3488 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0) 3489 addrconf_add_linklocal(idev, &addr, 0); 3490 else if (prefix_route) 3491 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3492 0, 0, GFP_KERNEL); 3493 break; 3494 case IN6_ADDR_GEN_MODE_NONE: 3495 default: 3496 /* will not add any link local address */ 3497 break; 3498 } 3499 } 3500 3501 static void addrconf_dev_config(struct net_device *dev) 3502 { 3503 struct inet6_dev *idev; 3504 3505 ASSERT_RTNL(); 3506 3507 if ((dev->type != ARPHRD_ETHER) && 3508 (dev->type != ARPHRD_FDDI) && 3509 (dev->type != ARPHRD_ARCNET) && 3510 (dev->type != ARPHRD_INFINIBAND) && 3511 (dev->type != ARPHRD_IEEE1394) && 3512 (dev->type != ARPHRD_TUNNEL6) && 3513 (dev->type != ARPHRD_6LOWPAN) && 3514 (dev->type != ARPHRD_IP6GRE) && 3515 (dev->type != ARPHRD_TUNNEL) && 3516 (dev->type != ARPHRD_NONE) && 3517 (dev->type != ARPHRD_RAWIP)) { 3518 /* Alas, we support only Ethernet autoconfiguration. */ 3519 idev = __in6_dev_get(dev); 3520 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP && 3521 dev->flags & IFF_MULTICAST) 3522 ipv6_mc_up(idev); 3523 return; 3524 } 3525 3526 idev = addrconf_add_dev(dev); 3527 if (IS_ERR(idev)) 3528 return; 3529 3530 /* this device type has no EUI support */ 3531 if (dev->type == ARPHRD_NONE && 3532 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) 3533 WRITE_ONCE(idev->cnf.addr_gen_mode, 3534 IN6_ADDR_GEN_MODE_RANDOM); 3535 3536 addrconf_addr_gen(idev, false); 3537 } 3538 3539 #if IS_ENABLED(CONFIG_IPV6_SIT) 3540 static void addrconf_sit_config(struct net_device *dev) 3541 { 3542 struct inet6_dev *idev; 3543 3544 ASSERT_RTNL(); 3545 3546 /* 3547 * Configure the tunnel with one of our IPv4 3548 * addresses... we should configure all of 3549 * our v4 addrs in the tunnel 3550 */ 3551 3552 idev = ipv6_find_idev(dev); 3553 if (IS_ERR(idev)) { 3554 pr_debug("%s: add_dev failed\n", __func__); 3555 return; 3556 } 3557 3558 if (dev->priv_flags & IFF_ISATAP) { 3559 addrconf_addr_gen(idev, false); 3560 return; 3561 } 3562 3563 add_v4_addrs(idev); 3564 3565 if (dev->flags&IFF_POINTOPOINT) 3566 addrconf_add_mroute(dev); 3567 } 3568 #endif 3569 3570 #if IS_ENABLED(CONFIG_NET_IPGRE) 3571 static void addrconf_gre_config(struct net_device *dev) 3572 { 3573 struct inet6_dev *idev; 3574 3575 ASSERT_RTNL(); 3576 3577 idev = addrconf_add_dev(dev); 3578 if (IS_ERR(idev)) 3579 return; 3580 3581 /* Generate the IPv6 link-local address using addrconf_addr_gen(), 3582 * unless we have an IPv4 GRE device not bound to an IP address and 3583 * which is in EUI64 mode (as __ipv6_isatap_ifid() would fail in this 3584 * case). Such devices fall back to add_v4_addrs() instead. 3585 */ 3586 if (!(*(__be32 *)dev->dev_addr == 0 && 3587 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64)) { 3588 addrconf_addr_gen(idev, true); 3589 return; 3590 } 3591 3592 add_v4_addrs(idev); 3593 } 3594 #endif 3595 3596 static void addrconf_init_auto_addrs(struct net_device *dev) 3597 { 3598 switch (dev->type) { 3599 #if IS_ENABLED(CONFIG_IPV6_SIT) 3600 case ARPHRD_SIT: 3601 addrconf_sit_config(dev); 3602 break; 3603 #endif 3604 #if IS_ENABLED(CONFIG_NET_IPGRE) 3605 case ARPHRD_IPGRE: 3606 addrconf_gre_config(dev); 3607 break; 3608 #endif 3609 case ARPHRD_LOOPBACK: 3610 init_loopback(dev); 3611 break; 3612 3613 default: 3614 addrconf_dev_config(dev); 3615 break; 3616 } 3617 } 3618 3619 static int fixup_permanent_addr(struct net *net, 3620 struct inet6_dev *idev, 3621 struct inet6_ifaddr *ifp) 3622 { 3623 /* !fib6_node means the host route was removed from the 3624 * FIB, for example, if 'lo' device is taken down. In that 3625 * case regenerate the host route. 3626 */ 3627 if (!ifp->rt || !ifp->rt->fib6_node) { 3628 struct fib6_info *f6i, *prev; 3629 3630 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false, 3631 GFP_KERNEL, NULL); 3632 if (IS_ERR(f6i)) 3633 return PTR_ERR(f6i); 3634 3635 /* ifp->rt can be accessed outside of rtnl */ 3636 spin_lock_bh(&ifp->lock); 3637 prev = ifp->rt; 3638 ifp->rt = f6i; 3639 spin_unlock_bh(&ifp->lock); 3640 3641 fib6_info_release(prev); 3642 } 3643 3644 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) { 3645 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3646 ifp->rt_priority, idev->dev, 0, 0, 3647 GFP_KERNEL); 3648 } 3649 3650 if (ifp->state == INET6_IFADDR_STATE_PREDAD) 3651 addrconf_dad_start(ifp); 3652 3653 return 0; 3654 } 3655 3656 static void addrconf_permanent_addr(struct net *net, struct net_device *dev) 3657 { 3658 struct inet6_ifaddr *ifp; 3659 LIST_HEAD(tmp_addr_list); 3660 struct inet6_dev *idev; 3661 3662 /* Mutual exclusion with other if_list_aux users. */ 3663 ASSERT_RTNL(); 3664 3665 idev = __in6_dev_get(dev); 3666 if (!idev) 3667 return; 3668 3669 write_lock_bh(&idev->lock); 3670 list_for_each_entry(ifp, &idev->addr_list, if_list) { 3671 if (ifp->flags & IFA_F_PERMANENT) 3672 list_add_tail(&ifp->if_list_aux, &tmp_addr_list); 3673 } 3674 write_unlock_bh(&idev->lock); 3675 3676 while (!list_empty(&tmp_addr_list)) { 3677 ifp = list_first_entry(&tmp_addr_list, 3678 struct inet6_ifaddr, if_list_aux); 3679 list_del(&ifp->if_list_aux); 3680 3681 if (fixup_permanent_addr(net, idev, ifp) < 0) { 3682 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n", 3683 idev->dev->name, &ifp->addr); 3684 in6_ifa_hold(ifp); 3685 ipv6_del_addr(ifp); 3686 } 3687 } 3688 } 3689 3690 static int addrconf_notify(struct notifier_block *this, unsigned long event, 3691 void *ptr) 3692 { 3693 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 3694 struct netdev_notifier_change_info *change_info; 3695 struct netdev_notifier_changeupper_info *info; 3696 struct inet6_dev *idev = __in6_dev_get(dev); 3697 struct net *net = dev_net(dev); 3698 int run_pending = 0; 3699 int err; 3700 3701 switch (event) { 3702 case NETDEV_REGISTER: 3703 if (!idev && dev->mtu >= IPV6_MIN_MTU) { 3704 idev = ipv6_add_dev(dev); 3705 if (IS_ERR(idev)) 3706 return notifier_from_errno(PTR_ERR(idev)); 3707 } 3708 break; 3709 3710 case NETDEV_CHANGEMTU: 3711 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */ 3712 if (dev->mtu < IPV6_MIN_MTU) { 3713 addrconf_ifdown(dev, dev != net->loopback_dev); 3714 break; 3715 } 3716 3717 if (idev) { 3718 rt6_mtu_change(dev, dev->mtu); 3719 WRITE_ONCE(idev->cnf.mtu6, dev->mtu); 3720 break; 3721 } 3722 3723 /* allocate new idev */ 3724 idev = ipv6_add_dev(dev); 3725 if (IS_ERR(idev)) 3726 break; 3727 3728 /* device is still not ready */ 3729 if (!(idev->if_flags & IF_READY)) 3730 break; 3731 3732 run_pending = 1; 3733 fallthrough; 3734 case NETDEV_UP: 3735 case NETDEV_CHANGE: 3736 if (idev && idev->cnf.disable_ipv6) 3737 break; 3738 3739 if (dev->priv_flags & IFF_NO_ADDRCONF) { 3740 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) && 3741 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST) 3742 ipv6_mc_up(idev); 3743 break; 3744 } 3745 3746 if (event == NETDEV_UP) { 3747 /* restore routes for permanent addresses */ 3748 addrconf_permanent_addr(net, dev); 3749 3750 if (!addrconf_link_ready(dev)) { 3751 /* device is not ready yet. */ 3752 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n", 3753 dev->name); 3754 break; 3755 } 3756 3757 if (!idev && dev->mtu >= IPV6_MIN_MTU) 3758 idev = ipv6_add_dev(dev); 3759 3760 if (!IS_ERR_OR_NULL(idev)) { 3761 idev->if_flags |= IF_READY; 3762 run_pending = 1; 3763 } 3764 } else if (event == NETDEV_CHANGE) { 3765 if (!addrconf_link_ready(dev)) { 3766 /* device is still not ready. */ 3767 rt6_sync_down_dev(dev, event); 3768 break; 3769 } 3770 3771 if (!IS_ERR_OR_NULL(idev)) { 3772 if (idev->if_flags & IF_READY) { 3773 /* device is already configured - 3774 * but resend MLD reports, we might 3775 * have roamed and need to update 3776 * multicast snooping switches 3777 */ 3778 ipv6_mc_up(idev); 3779 change_info = ptr; 3780 if (change_info->flags_changed & IFF_NOARP) 3781 addrconf_dad_run(idev, true); 3782 rt6_sync_up(dev, RTNH_F_LINKDOWN); 3783 break; 3784 } 3785 idev->if_flags |= IF_READY; 3786 } 3787 3788 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n", 3789 dev->name); 3790 3791 run_pending = 1; 3792 } 3793 3794 addrconf_init_auto_addrs(dev); 3795 3796 if (!IS_ERR_OR_NULL(idev)) { 3797 if (run_pending) 3798 addrconf_dad_run(idev, false); 3799 3800 /* Device has an address by now */ 3801 rt6_sync_up(dev, RTNH_F_DEAD); 3802 3803 /* 3804 * If the MTU changed during the interface down, 3805 * when the interface up, the changed MTU must be 3806 * reflected in the idev as well as routers. 3807 */ 3808 if (idev->cnf.mtu6 != dev->mtu && 3809 dev->mtu >= IPV6_MIN_MTU) { 3810 rt6_mtu_change(dev, dev->mtu); 3811 WRITE_ONCE(idev->cnf.mtu6, dev->mtu); 3812 } 3813 WRITE_ONCE(idev->tstamp, jiffies); 3814 inet6_ifinfo_notify(RTM_NEWLINK, idev); 3815 3816 /* 3817 * If the changed mtu during down is lower than 3818 * IPV6_MIN_MTU stop IPv6 on this interface. 3819 */ 3820 if (dev->mtu < IPV6_MIN_MTU) 3821 addrconf_ifdown(dev, dev != net->loopback_dev); 3822 } 3823 break; 3824 3825 case NETDEV_DOWN: 3826 case NETDEV_UNREGISTER: 3827 /* 3828 * Remove all addresses from this interface. 3829 */ 3830 addrconf_ifdown(dev, event != NETDEV_DOWN); 3831 break; 3832 3833 case NETDEV_CHANGENAME: 3834 if (idev) { 3835 snmp6_unregister_dev(idev); 3836 addrconf_sysctl_unregister(idev); 3837 err = addrconf_sysctl_register(idev); 3838 if (err) 3839 return notifier_from_errno(err); 3840 err = snmp6_register_dev(idev); 3841 if (err) { 3842 addrconf_sysctl_unregister(idev); 3843 return notifier_from_errno(err); 3844 } 3845 } 3846 break; 3847 3848 case NETDEV_PRE_TYPE_CHANGE: 3849 case NETDEV_POST_TYPE_CHANGE: 3850 if (idev) 3851 addrconf_type_change(dev, event); 3852 break; 3853 3854 case NETDEV_CHANGEUPPER: 3855 info = ptr; 3856 3857 /* flush all routes if dev is linked to or unlinked from 3858 * an L3 master device (e.g., VRF) 3859 */ 3860 if (info->upper_dev && netif_is_l3_master(info->upper_dev)) 3861 addrconf_ifdown(dev, false); 3862 } 3863 3864 return NOTIFY_OK; 3865 } 3866 3867 /* 3868 * addrconf module should be notified of a device going up 3869 */ 3870 static struct notifier_block ipv6_dev_notf = { 3871 .notifier_call = addrconf_notify, 3872 .priority = ADDRCONF_NOTIFY_PRIORITY, 3873 }; 3874 3875 static void addrconf_type_change(struct net_device *dev, unsigned long event) 3876 { 3877 struct inet6_dev *idev; 3878 ASSERT_RTNL(); 3879 3880 idev = __in6_dev_get(dev); 3881 3882 if (event == NETDEV_POST_TYPE_CHANGE) 3883 ipv6_mc_remap(idev); 3884 else if (event == NETDEV_PRE_TYPE_CHANGE) 3885 ipv6_mc_unmap(idev); 3886 } 3887 3888 static bool addr_is_local(const struct in6_addr *addr) 3889 { 3890 return ipv6_addr_type(addr) & 3891 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK); 3892 } 3893 3894 static int addrconf_ifdown(struct net_device *dev, bool unregister) 3895 { 3896 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN; 3897 struct net *net = dev_net(dev); 3898 struct inet6_dev *idev; 3899 struct inet6_ifaddr *ifa; 3900 LIST_HEAD(tmp_addr_list); 3901 bool keep_addr = false; 3902 bool was_ready; 3903 int state, i; 3904 3905 ASSERT_RTNL(); 3906 3907 rt6_disable_ip(dev, event); 3908 3909 idev = __in6_dev_get(dev); 3910 if (!idev) 3911 return -ENODEV; 3912 3913 /* 3914 * Step 1: remove reference to ipv6 device from parent device. 3915 * Do not dev_put! 3916 */ 3917 if (unregister) { 3918 write_lock_bh(&idev->lock); 3919 WRITE_ONCE(idev->dead, 1); 3920 write_unlock_bh(&idev->lock); 3921 3922 /* protected by rtnl_lock */ 3923 RCU_INIT_POINTER(dev->ip6_ptr, NULL); 3924 3925 /* Step 1.5: remove snmp6 entry */ 3926 snmp6_unregister_dev(idev); 3927 3928 } 3929 3930 /* combine the user config with event to determine if permanent 3931 * addresses are to be removed from address hash table 3932 */ 3933 if (!unregister && !idev->cnf.disable_ipv6) { 3934 /* aggregate the system setting and interface setting */ 3935 int _keep_addr = READ_ONCE(net->ipv6.devconf_all->keep_addr_on_down); 3936 3937 if (!_keep_addr) 3938 _keep_addr = READ_ONCE(idev->cnf.keep_addr_on_down); 3939 3940 keep_addr = (_keep_addr > 0); 3941 } 3942 3943 /* Step 2: clear hash table */ 3944 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 3945 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i]; 3946 3947 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 3948 restart: 3949 hlist_for_each_entry_rcu(ifa, h, addr_lst) { 3950 if (ifa->idev == idev) { 3951 addrconf_del_dad_work(ifa); 3952 /* combined flag + permanent flag decide if 3953 * address is retained on a down event 3954 */ 3955 if (!keep_addr || 3956 !(ifa->flags & IFA_F_PERMANENT) || 3957 addr_is_local(&ifa->addr)) { 3958 hlist_del_init_rcu(&ifa->addr_lst); 3959 goto restart; 3960 } 3961 } 3962 } 3963 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 3964 } 3965 3966 write_lock_bh(&idev->lock); 3967 3968 addrconf_del_rs_timer(idev); 3969 3970 /* Step 2: clear flags for stateless addrconf, repeated down 3971 * detection 3972 */ 3973 was_ready = idev->if_flags & IF_READY; 3974 if (!unregister) 3975 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY); 3976 3977 /* Step 3: clear tempaddr list */ 3978 while (!list_empty(&idev->tempaddr_list)) { 3979 ifa = list_first_entry(&idev->tempaddr_list, 3980 struct inet6_ifaddr, tmp_list); 3981 list_del(&ifa->tmp_list); 3982 write_unlock_bh(&idev->lock); 3983 spin_lock_bh(&ifa->lock); 3984 3985 if (ifa->ifpub) { 3986 in6_ifa_put(ifa->ifpub); 3987 ifa->ifpub = NULL; 3988 } 3989 spin_unlock_bh(&ifa->lock); 3990 in6_ifa_put(ifa); 3991 write_lock_bh(&idev->lock); 3992 } 3993 3994 list_for_each_entry(ifa, &idev->addr_list, if_list) 3995 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 3996 write_unlock_bh(&idev->lock); 3997 3998 while (!list_empty(&tmp_addr_list)) { 3999 struct fib6_info *rt = NULL; 4000 bool keep; 4001 4002 ifa = list_first_entry(&tmp_addr_list, 4003 struct inet6_ifaddr, if_list_aux); 4004 list_del(&ifa->if_list_aux); 4005 4006 addrconf_del_dad_work(ifa); 4007 4008 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) && 4009 !addr_is_local(&ifa->addr); 4010 4011 spin_lock_bh(&ifa->lock); 4012 4013 if (keep) { 4014 /* set state to skip the notifier below */ 4015 state = INET6_IFADDR_STATE_DEAD; 4016 ifa->state = INET6_IFADDR_STATE_PREDAD; 4017 if (!(ifa->flags & IFA_F_NODAD)) 4018 ifa->flags |= IFA_F_TENTATIVE; 4019 4020 rt = ifa->rt; 4021 ifa->rt = NULL; 4022 } else { 4023 state = ifa->state; 4024 ifa->state = INET6_IFADDR_STATE_DEAD; 4025 } 4026 4027 spin_unlock_bh(&ifa->lock); 4028 4029 if (rt) 4030 ip6_del_rt(net, rt, false); 4031 4032 if (state != INET6_IFADDR_STATE_DEAD) { 4033 __ipv6_ifa_notify(RTM_DELADDR, ifa); 4034 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa); 4035 } else { 4036 if (idev->cnf.forwarding) 4037 addrconf_leave_anycast(ifa); 4038 addrconf_leave_solict(ifa->idev, &ifa->addr); 4039 } 4040 4041 if (!keep) { 4042 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 4043 hlist_del_init_rcu(&ifa->addr_lst); 4044 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 4045 4046 write_lock_bh(&idev->lock); 4047 list_del_rcu(&ifa->if_list); 4048 write_unlock_bh(&idev->lock); 4049 in6_ifa_put(ifa); 4050 } 4051 } 4052 4053 /* Step 5: Discard anycast and multicast list */ 4054 if (unregister) { 4055 ipv6_ac_destroy_dev(idev); 4056 ipv6_mc_destroy_dev(idev); 4057 } else if (was_ready) { 4058 ipv6_mc_down(idev); 4059 } 4060 4061 WRITE_ONCE(idev->tstamp, jiffies); 4062 idev->ra_mtu = 0; 4063 4064 /* Last: Shot the device (if unregistered) */ 4065 if (unregister) { 4066 struct neigh_table *tbl = nd_table(net); 4067 4068 addrconf_sysctl_unregister(idev); 4069 neigh_parms_release(tbl, idev->nd_parms); 4070 neigh_ifdown(tbl, dev); 4071 in6_dev_put(idev); 4072 } 4073 return 0; 4074 } 4075 4076 static void addrconf_rs_timer(struct timer_list *t) 4077 { 4078 struct inet6_dev *idev = timer_container_of(idev, t, rs_timer); 4079 struct net_device *dev = idev->dev; 4080 struct in6_addr lladdr; 4081 int rtr_solicits; 4082 4083 write_lock(&idev->lock); 4084 if (idev->dead || !(idev->if_flags & IF_READY)) 4085 goto out; 4086 4087 if (!ipv6_accept_ra(idev)) 4088 goto out; 4089 4090 /* Announcement received after solicitation was sent */ 4091 if (idev->if_flags & IF_RA_RCVD) 4092 goto out; 4093 4094 rtr_solicits = READ_ONCE(idev->cnf.rtr_solicits); 4095 4096 if (idev->rs_probes++ < rtr_solicits || rtr_solicits < 0) { 4097 write_unlock(&idev->lock); 4098 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 4099 ndisc_send_rs(dev, &lladdr, 4100 &in6addr_linklocal_allrouters); 4101 else 4102 goto put; 4103 4104 write_lock(&idev->lock); 4105 idev->rs_interval = rfc3315_s14_backoff_update( 4106 idev->rs_interval, 4107 READ_ONCE(idev->cnf.rtr_solicit_max_interval)); 4108 /* The wait after the last probe can be shorter */ 4109 addrconf_mod_rs_timer(idev, (idev->rs_probes == 4110 READ_ONCE(idev->cnf.rtr_solicits)) ? 4111 READ_ONCE(idev->cnf.rtr_solicit_delay) : 4112 idev->rs_interval); 4113 } else { 4114 /* 4115 * Note: we do not support deprecated "all on-link" 4116 * assumption any longer. 4117 */ 4118 pr_debug("%s: no IPv6 routers present\n", idev->dev->name); 4119 } 4120 4121 out: 4122 write_unlock(&idev->lock); 4123 put: 4124 in6_dev_put(idev); 4125 } 4126 4127 /* 4128 * Duplicate Address Detection 4129 */ 4130 static void addrconf_dad_kick(struct inet6_ifaddr *ifp) 4131 { 4132 struct inet6_dev *idev = ifp->idev; 4133 unsigned long rand_num; 4134 u64 nonce; 4135 4136 if (ifp->flags & IFA_F_OPTIMISTIC) 4137 rand_num = 0; 4138 else 4139 rand_num = get_random_u32_below( 4140 READ_ONCE(idev->cnf.rtr_solicit_delay) ? : 1); 4141 4142 nonce = 0; 4143 if (READ_ONCE(idev->cnf.enhanced_dad) || 4144 READ_ONCE(dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad)) { 4145 do 4146 get_random_bytes(&nonce, 6); 4147 while (nonce == 0); 4148 } 4149 ifp->dad_nonce = nonce; 4150 ifp->dad_probes = READ_ONCE(idev->cnf.dad_transmits); 4151 addrconf_mod_dad_work(ifp, rand_num); 4152 } 4153 4154 static void addrconf_dad_begin(struct inet6_ifaddr *ifp) 4155 { 4156 struct inet6_dev *idev = ifp->idev; 4157 struct net_device *dev = idev->dev; 4158 bool bump_id, notify = false; 4159 struct net *net; 4160 4161 addrconf_join_solict(dev, &ifp->addr); 4162 4163 read_lock_bh(&idev->lock); 4164 spin_lock(&ifp->lock); 4165 if (ifp->state == INET6_IFADDR_STATE_DEAD) 4166 goto out; 4167 4168 net = dev_net(dev); 4169 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) || 4170 (READ_ONCE(net->ipv6.devconf_all->accept_dad) < 1 && 4171 READ_ONCE(idev->cnf.accept_dad) < 1) || 4172 !(ifp->flags&IFA_F_TENTATIVE) || 4173 ifp->flags & IFA_F_NODAD) { 4174 bool send_na = false; 4175 4176 if (ifp->flags & IFA_F_TENTATIVE && 4177 !(ifp->flags & IFA_F_OPTIMISTIC)) 4178 send_na = true; 4179 bump_id = ifp->flags & IFA_F_TENTATIVE; 4180 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4181 spin_unlock(&ifp->lock); 4182 read_unlock_bh(&idev->lock); 4183 4184 addrconf_dad_completed(ifp, bump_id, send_na); 4185 return; 4186 } 4187 4188 if (!(idev->if_flags & IF_READY)) { 4189 spin_unlock(&ifp->lock); 4190 read_unlock_bh(&idev->lock); 4191 /* 4192 * If the device is not ready: 4193 * - keep it tentative if it is a permanent address. 4194 * - otherwise, kill it. 4195 */ 4196 in6_ifa_hold(ifp); 4197 addrconf_dad_stop(ifp, 0); 4198 return; 4199 } 4200 4201 /* 4202 * Optimistic nodes can start receiving 4203 * Frames right away 4204 */ 4205 if (ifp->flags & IFA_F_OPTIMISTIC) { 4206 ip6_ins_rt(net, ifp->rt); 4207 if (ipv6_use_optimistic_addr(net, idev)) { 4208 /* Because optimistic nodes can use this address, 4209 * notify listeners. If DAD fails, RTM_DELADDR is sent. 4210 */ 4211 notify = true; 4212 } 4213 } 4214 4215 addrconf_dad_kick(ifp); 4216 out: 4217 spin_unlock(&ifp->lock); 4218 read_unlock_bh(&idev->lock); 4219 if (notify) 4220 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4221 } 4222 4223 static void addrconf_dad_start(struct inet6_ifaddr *ifp) 4224 { 4225 bool begin_dad = false; 4226 4227 spin_lock_bh(&ifp->lock); 4228 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 4229 ifp->state = INET6_IFADDR_STATE_PREDAD; 4230 begin_dad = true; 4231 } 4232 spin_unlock_bh(&ifp->lock); 4233 4234 if (begin_dad) 4235 addrconf_mod_dad_work(ifp, 0); 4236 } 4237 4238 static void addrconf_dad_work(struct work_struct *w) 4239 { 4240 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w), 4241 struct inet6_ifaddr, 4242 dad_work); 4243 struct inet6_dev *idev = ifp->idev; 4244 bool bump_id, disable_ipv6 = false; 4245 struct in6_addr mcaddr; 4246 struct net *net; 4247 4248 enum { 4249 DAD_PROCESS, 4250 DAD_BEGIN, 4251 DAD_ABORT, 4252 } action = DAD_PROCESS; 4253 4254 net = dev_net(idev->dev); 4255 4256 rtnl_net_lock(net); 4257 4258 spin_lock_bh(&ifp->lock); 4259 if (ifp->state == INET6_IFADDR_STATE_PREDAD) { 4260 action = DAD_BEGIN; 4261 ifp->state = INET6_IFADDR_STATE_DAD; 4262 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) { 4263 action = DAD_ABORT; 4264 ifp->state = INET6_IFADDR_STATE_POSTDAD; 4265 4266 if ((READ_ONCE(net->ipv6.devconf_all->accept_dad) > 1 || 4267 READ_ONCE(idev->cnf.accept_dad) > 1) && 4268 !idev->cnf.disable_ipv6 && 4269 !(ifp->flags & IFA_F_STABLE_PRIVACY)) { 4270 struct in6_addr addr; 4271 4272 addr.s6_addr32[0] = htonl(0xfe800000); 4273 addr.s6_addr32[1] = 0; 4274 4275 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) && 4276 ipv6_addr_equal(&ifp->addr, &addr)) { 4277 /* DAD failed for link-local based on MAC */ 4278 WRITE_ONCE(idev->cnf.disable_ipv6, 1); 4279 4280 pr_info("%s: IPv6 being disabled!\n", 4281 ifp->idev->dev->name); 4282 disable_ipv6 = true; 4283 } 4284 } 4285 } 4286 spin_unlock_bh(&ifp->lock); 4287 4288 if (action == DAD_BEGIN) { 4289 addrconf_dad_begin(ifp); 4290 goto out; 4291 } else if (action == DAD_ABORT) { 4292 in6_ifa_hold(ifp); 4293 addrconf_dad_stop(ifp, 1); 4294 if (disable_ipv6) 4295 addrconf_ifdown(idev->dev, false); 4296 goto out; 4297 } 4298 4299 if (!ifp->dad_probes && addrconf_dad_end(ifp)) 4300 goto out; 4301 4302 write_lock_bh(&idev->lock); 4303 if (idev->dead || !(idev->if_flags & IF_READY)) { 4304 write_unlock_bh(&idev->lock); 4305 goto out; 4306 } 4307 4308 spin_lock(&ifp->lock); 4309 if (ifp->state == INET6_IFADDR_STATE_DEAD) { 4310 spin_unlock(&ifp->lock); 4311 write_unlock_bh(&idev->lock); 4312 goto out; 4313 } 4314 4315 if (ifp->dad_probes == 0) { 4316 bool send_na = false; 4317 4318 /* 4319 * DAD was successful 4320 */ 4321 4322 if (ifp->flags & IFA_F_TENTATIVE && 4323 !(ifp->flags & IFA_F_OPTIMISTIC)) 4324 send_na = true; 4325 bump_id = ifp->flags & IFA_F_TENTATIVE; 4326 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4327 spin_unlock(&ifp->lock); 4328 write_unlock_bh(&idev->lock); 4329 4330 addrconf_dad_completed(ifp, bump_id, send_na); 4331 4332 goto out; 4333 } 4334 4335 ifp->dad_probes--; 4336 addrconf_mod_dad_work(ifp, 4337 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), 4338 HZ/100)); 4339 spin_unlock(&ifp->lock); 4340 write_unlock_bh(&idev->lock); 4341 4342 /* send a neighbour solicitation for our addr */ 4343 addrconf_addr_solict_mult(&ifp->addr, &mcaddr); 4344 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any, 4345 ifp->dad_nonce); 4346 out: 4347 in6_ifa_put(ifp); 4348 rtnl_net_unlock(net); 4349 } 4350 4351 /* ifp->idev must be at least read locked */ 4352 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp) 4353 { 4354 struct inet6_ifaddr *ifpiter; 4355 struct inet6_dev *idev = ifp->idev; 4356 4357 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) { 4358 if (ifpiter->scope > IFA_LINK) 4359 break; 4360 if (ifp != ifpiter && ifpiter->scope == IFA_LINK && 4361 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE| 4362 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) == 4363 IFA_F_PERMANENT) 4364 return false; 4365 } 4366 return true; 4367 } 4368 4369 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 4370 bool send_na) 4371 { 4372 struct net_device *dev = ifp->idev->dev; 4373 struct in6_addr lladdr; 4374 bool send_rs, send_mld; 4375 4376 addrconf_del_dad_work(ifp); 4377 4378 /* 4379 * Configure the address for reception. Now it is valid. 4380 */ 4381 4382 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4383 4384 /* If added prefix is link local and we are prepared to process 4385 router advertisements, start sending router solicitations. 4386 */ 4387 4388 read_lock_bh(&ifp->idev->lock); 4389 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp); 4390 send_rs = send_mld && 4391 ipv6_accept_ra(ifp->idev) && 4392 READ_ONCE(ifp->idev->cnf.rtr_solicits) != 0 && 4393 (dev->flags & IFF_LOOPBACK) == 0 && 4394 (dev->type != ARPHRD_TUNNEL) && 4395 !netif_is_team_port(dev); 4396 read_unlock_bh(&ifp->idev->lock); 4397 4398 /* While dad is in progress mld report's source address is in6_addrany. 4399 * Resend with proper ll now. 4400 */ 4401 if (send_mld) 4402 ipv6_mc_dad_complete(ifp->idev); 4403 4404 /* send unsolicited NA if enabled */ 4405 if (send_na && 4406 (READ_ONCE(ifp->idev->cnf.ndisc_notify) || 4407 READ_ONCE(dev_net(dev)->ipv6.devconf_all->ndisc_notify))) { 4408 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr, 4409 /*router=*/ !!ifp->idev->cnf.forwarding, 4410 /*solicited=*/ false, /*override=*/ true, 4411 /*inc_opt=*/ true); 4412 } 4413 4414 if (send_rs) { 4415 /* 4416 * If a host as already performed a random delay 4417 * [...] as part of DAD [...] there is no need 4418 * to delay again before sending the first RS 4419 */ 4420 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 4421 return; 4422 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters); 4423 4424 write_lock_bh(&ifp->idev->lock); 4425 spin_lock(&ifp->lock); 4426 ifp->idev->rs_interval = rfc3315_s14_backoff_init( 4427 READ_ONCE(ifp->idev->cnf.rtr_solicit_interval)); 4428 ifp->idev->rs_probes = 1; 4429 ifp->idev->if_flags |= IF_RS_SENT; 4430 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval); 4431 spin_unlock(&ifp->lock); 4432 write_unlock_bh(&ifp->idev->lock); 4433 } 4434 4435 if (bump_id) 4436 rt_genid_bump_ipv6(dev_net(dev)); 4437 4438 /* Make sure that a new temporary address will be created 4439 * before this temporary address becomes deprecated. 4440 */ 4441 if (ifp->flags & IFA_F_TEMPORARY) 4442 addrconf_verify_rtnl(dev_net(dev)); 4443 } 4444 4445 static void addrconf_dad_run(struct inet6_dev *idev, bool restart) 4446 { 4447 struct inet6_ifaddr *ifp; 4448 4449 read_lock_bh(&idev->lock); 4450 list_for_each_entry(ifp, &idev->addr_list, if_list) { 4451 spin_lock(&ifp->lock); 4452 if ((ifp->flags & IFA_F_TENTATIVE && 4453 ifp->state == INET6_IFADDR_STATE_DAD) || restart) { 4454 if (restart) 4455 ifp->state = INET6_IFADDR_STATE_PREDAD; 4456 addrconf_dad_kick(ifp); 4457 } 4458 spin_unlock(&ifp->lock); 4459 } 4460 read_unlock_bh(&idev->lock); 4461 } 4462 4463 #ifdef CONFIG_PROC_FS 4464 struct if6_iter_state { 4465 struct seq_net_private p; 4466 int bucket; 4467 int offset; 4468 }; 4469 4470 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos) 4471 { 4472 struct if6_iter_state *state = seq->private; 4473 struct net *net = seq_file_net(seq); 4474 struct inet6_ifaddr *ifa = NULL; 4475 int p = 0; 4476 4477 /* initial bucket if pos is 0 */ 4478 if (pos == 0) { 4479 state->bucket = 0; 4480 state->offset = 0; 4481 } 4482 4483 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) { 4484 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket], 4485 addr_lst) { 4486 /* sync with offset */ 4487 if (p < state->offset) { 4488 p++; 4489 continue; 4490 } 4491 return ifa; 4492 } 4493 4494 /* prepare for next bucket */ 4495 state->offset = 0; 4496 p = 0; 4497 } 4498 return NULL; 4499 } 4500 4501 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, 4502 struct inet6_ifaddr *ifa) 4503 { 4504 struct if6_iter_state *state = seq->private; 4505 struct net *net = seq_file_net(seq); 4506 4507 hlist_for_each_entry_continue_rcu(ifa, addr_lst) { 4508 state->offset++; 4509 return ifa; 4510 } 4511 4512 state->offset = 0; 4513 while (++state->bucket < IN6_ADDR_HSIZE) { 4514 hlist_for_each_entry_rcu(ifa, 4515 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) { 4516 return ifa; 4517 } 4518 } 4519 4520 return NULL; 4521 } 4522 4523 static void *if6_seq_start(struct seq_file *seq, loff_t *pos) 4524 __acquires(rcu) 4525 { 4526 rcu_read_lock(); 4527 return if6_get_first(seq, *pos); 4528 } 4529 4530 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 4531 { 4532 struct inet6_ifaddr *ifa; 4533 4534 ifa = if6_get_next(seq, v); 4535 ++*pos; 4536 return ifa; 4537 } 4538 4539 static void if6_seq_stop(struct seq_file *seq, void *v) 4540 __releases(rcu) 4541 { 4542 rcu_read_unlock(); 4543 } 4544 4545 static int if6_seq_show(struct seq_file *seq, void *v) 4546 { 4547 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v; 4548 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n", 4549 &ifp->addr, 4550 ifp->idev->dev->ifindex, 4551 ifp->prefix_len, 4552 ifp->scope, 4553 (u8) ifp->flags, 4554 ifp->idev->dev->name); 4555 return 0; 4556 } 4557 4558 static const struct seq_operations if6_seq_ops = { 4559 .start = if6_seq_start, 4560 .next = if6_seq_next, 4561 .show = if6_seq_show, 4562 .stop = if6_seq_stop, 4563 }; 4564 4565 static int __net_init if6_proc_net_init(struct net *net) 4566 { 4567 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops, 4568 sizeof(struct if6_iter_state))) 4569 return -ENOMEM; 4570 return 0; 4571 } 4572 4573 static void __net_exit if6_proc_net_exit(struct net *net) 4574 { 4575 remove_proc_entry("if_inet6", net->proc_net); 4576 } 4577 4578 static struct pernet_operations if6_proc_net_ops = { 4579 .init = if6_proc_net_init, 4580 .exit = if6_proc_net_exit, 4581 }; 4582 4583 int __init if6_proc_init(void) 4584 { 4585 return register_pernet_subsys(&if6_proc_net_ops); 4586 } 4587 4588 void if6_proc_exit(void) 4589 { 4590 unregister_pernet_subsys(&if6_proc_net_ops); 4591 } 4592 #endif /* CONFIG_PROC_FS */ 4593 4594 #if IS_ENABLED(CONFIG_IPV6_MIP6) 4595 /* Check if address is a home address configured on any interface. */ 4596 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr) 4597 { 4598 unsigned int hash = inet6_addr_hash(net, addr); 4599 struct inet6_ifaddr *ifp = NULL; 4600 int ret = 0; 4601 4602 rcu_read_lock(); 4603 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4604 if (ipv6_addr_equal(&ifp->addr, addr) && 4605 (ifp->flags & IFA_F_HOMEADDRESS)) { 4606 ret = 1; 4607 break; 4608 } 4609 } 4610 rcu_read_unlock(); 4611 return ret; 4612 } 4613 #endif 4614 4615 /* RFC6554 has some algorithm to avoid loops in segment routing by 4616 * checking if the segments contains any of a local interface address. 4617 * 4618 * Quote: 4619 * 4620 * To detect loops in the SRH, a router MUST determine if the SRH 4621 * includes multiple addresses assigned to any interface on that router. 4622 * If such addresses appear more than once and are separated by at least 4623 * one address not assigned to that router. 4624 */ 4625 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs, 4626 unsigned char nsegs) 4627 { 4628 const struct in6_addr *addr; 4629 int i, ret = 0, found = 0; 4630 struct inet6_ifaddr *ifp; 4631 bool separated = false; 4632 unsigned int hash; 4633 bool hash_found; 4634 4635 rcu_read_lock(); 4636 for (i = 0; i < nsegs; i++) { 4637 addr = &segs[i]; 4638 hash = inet6_addr_hash(net, addr); 4639 4640 hash_found = false; 4641 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4642 4643 if (ipv6_addr_equal(&ifp->addr, addr)) { 4644 hash_found = true; 4645 break; 4646 } 4647 } 4648 4649 if (hash_found) { 4650 if (found > 1 && separated) { 4651 ret = 1; 4652 break; 4653 } 4654 4655 separated = false; 4656 found++; 4657 } else { 4658 separated = true; 4659 } 4660 } 4661 rcu_read_unlock(); 4662 4663 return ret; 4664 } 4665 4666 /* 4667 * Periodic address status verification 4668 */ 4669 4670 static void addrconf_verify_rtnl(struct net *net) 4671 { 4672 unsigned long now, next, next_sec, next_sched; 4673 struct inet6_ifaddr *ifp; 4674 int i; 4675 4676 ASSERT_RTNL(); 4677 4678 rcu_read_lock_bh(); 4679 now = jiffies; 4680 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY); 4681 4682 cancel_delayed_work(&net->ipv6.addr_chk_work); 4683 4684 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 4685 restart: 4686 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) { 4687 unsigned long age; 4688 4689 /* When setting preferred_lft to a value not zero or 4690 * infinity, while valid_lft is infinity 4691 * IFA_F_PERMANENT has a non-infinity life time. 4692 */ 4693 if ((ifp->flags & IFA_F_PERMANENT) && 4694 (ifp->prefered_lft == INFINITY_LIFE_TIME)) 4695 continue; 4696 4697 spin_lock(&ifp->lock); 4698 /* We try to batch several events at once. */ 4699 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 4700 4701 if ((ifp->flags&IFA_F_TEMPORARY) && 4702 !(ifp->flags&IFA_F_TENTATIVE) && 4703 ifp->prefered_lft != INFINITY_LIFE_TIME && 4704 !ifp->regen_count && ifp->ifpub) { 4705 /* This is a non-regenerated temporary addr. */ 4706 4707 unsigned long regen_advance = ipv6_get_regen_advance(ifp->idev); 4708 4709 if (age + regen_advance >= ifp->prefered_lft) { 4710 struct inet6_ifaddr *ifpub = ifp->ifpub; 4711 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4712 next = ifp->tstamp + ifp->prefered_lft * HZ; 4713 4714 ifp->regen_count++; 4715 in6_ifa_hold(ifp); 4716 in6_ifa_hold(ifpub); 4717 spin_unlock(&ifp->lock); 4718 4719 spin_lock(&ifpub->lock); 4720 ifpub->regen_count = 0; 4721 spin_unlock(&ifpub->lock); 4722 rcu_read_unlock_bh(); 4723 ipv6_create_tempaddr(ifpub, true); 4724 in6_ifa_put(ifpub); 4725 in6_ifa_put(ifp); 4726 rcu_read_lock_bh(); 4727 goto restart; 4728 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next)) 4729 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ; 4730 } 4731 4732 if (ifp->valid_lft != INFINITY_LIFE_TIME && 4733 age >= ifp->valid_lft) { 4734 spin_unlock(&ifp->lock); 4735 in6_ifa_hold(ifp); 4736 rcu_read_unlock_bh(); 4737 ipv6_del_addr(ifp); 4738 rcu_read_lock_bh(); 4739 goto restart; 4740 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) { 4741 spin_unlock(&ifp->lock); 4742 continue; 4743 } else if (age >= ifp->prefered_lft) { 4744 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */ 4745 int deprecate = 0; 4746 4747 if (!(ifp->flags&IFA_F_DEPRECATED)) { 4748 deprecate = 1; 4749 ifp->flags |= IFA_F_DEPRECATED; 4750 } 4751 4752 if ((ifp->valid_lft != INFINITY_LIFE_TIME) && 4753 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))) 4754 next = ifp->tstamp + ifp->valid_lft * HZ; 4755 4756 spin_unlock(&ifp->lock); 4757 4758 if (deprecate) { 4759 in6_ifa_hold(ifp); 4760 4761 ipv6_ifa_notify(0, ifp); 4762 in6_ifa_put(ifp); 4763 goto restart; 4764 } 4765 } else { 4766 /* ifp->prefered_lft <= ifp->valid_lft */ 4767 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4768 next = ifp->tstamp + ifp->prefered_lft * HZ; 4769 spin_unlock(&ifp->lock); 4770 } 4771 } 4772 } 4773 4774 next_sec = round_jiffies_up(next); 4775 next_sched = next; 4776 4777 /* If rounded timeout is accurate enough, accept it. */ 4778 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ)) 4779 next_sched = next_sec; 4780 4781 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */ 4782 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX)) 4783 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX; 4784 4785 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n", 4786 now, next, next_sec, next_sched); 4787 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now); 4788 rcu_read_unlock_bh(); 4789 } 4790 4791 static void addrconf_verify_work(struct work_struct *w) 4792 { 4793 struct net *net = container_of(to_delayed_work(w), struct net, 4794 ipv6.addr_chk_work); 4795 4796 rtnl_net_lock(net); 4797 addrconf_verify_rtnl(net); 4798 rtnl_net_unlock(net); 4799 } 4800 4801 static void addrconf_verify(struct net *net) 4802 { 4803 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0); 4804 } 4805 4806 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local, 4807 struct in6_addr **peer_pfx) 4808 { 4809 struct in6_addr *pfx = NULL; 4810 4811 *peer_pfx = NULL; 4812 4813 if (addr) 4814 pfx = nla_data(addr); 4815 4816 if (local) { 4817 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx))) 4818 *peer_pfx = pfx; 4819 pfx = nla_data(local); 4820 } 4821 4822 return pfx; 4823 } 4824 4825 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = { 4826 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) }, 4827 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) }, 4828 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) }, 4829 [IFA_FLAGS] = { .len = sizeof(u32) }, 4830 [IFA_RT_PRIORITY] = { .len = sizeof(u32) }, 4831 [IFA_TARGET_NETNSID] = { .type = NLA_S32 }, 4832 [IFA_PROTO] = { .type = NLA_U8 }, 4833 }; 4834 4835 static int 4836 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, 4837 struct netlink_ext_ack *extack) 4838 { 4839 struct net *net = sock_net(skb->sk); 4840 struct ifaddrmsg *ifm; 4841 struct nlattr *tb[IFA_MAX+1]; 4842 struct in6_addr *pfx, *peer_pfx; 4843 u32 ifa_flags; 4844 int err; 4845 4846 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 4847 ifa_ipv6_policy, extack); 4848 if (err < 0) 4849 return err; 4850 4851 ifm = nlmsg_data(nlh); 4852 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 4853 if (!pfx) 4854 return -EINVAL; 4855 4856 ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags); 4857 4858 /* We ignore other flags so far. */ 4859 ifa_flags &= IFA_F_MANAGETEMPADDR; 4860 4861 rtnl_net_lock(net); 4862 err = inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx, 4863 ifm->ifa_prefixlen, extack); 4864 rtnl_net_unlock(net); 4865 4866 return err; 4867 } 4868 4869 static int modify_prefix_route(struct net *net, struct inet6_ifaddr *ifp, 4870 unsigned long expires, u32 flags, 4871 bool modify_peer) 4872 { 4873 struct fib6_table *table; 4874 struct fib6_info *f6i; 4875 u32 prio; 4876 4877 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4878 ifp->prefix_len, 4879 ifp->idev->dev, 0, RTF_DEFAULT, true); 4880 if (!f6i) 4881 return -ENOENT; 4882 4883 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF; 4884 if (f6i->fib6_metric != prio) { 4885 /* delete old one */ 4886 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false); 4887 4888 /* add new one */ 4889 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4890 ifp->prefix_len, 4891 ifp->rt_priority, ifp->idev->dev, 4892 expires, flags, GFP_KERNEL); 4893 return 0; 4894 } 4895 if (f6i != net->ipv6.fib6_null_entry) { 4896 table = f6i->fib6_table; 4897 spin_lock_bh(&table->tb6_lock); 4898 4899 if (!(flags & RTF_EXPIRES)) { 4900 fib6_clean_expires(f6i); 4901 fib6_may_remove_gc_list(net, f6i); 4902 } else { 4903 fib6_set_expires(f6i, jiffies + clock_t_to_jiffies(expires)); 4904 fib6_add_gc_list(f6i); 4905 } 4906 4907 spin_unlock_bh(&table->tb6_lock); 4908 } 4909 fib6_info_release(f6i); 4910 4911 return 0; 4912 } 4913 4914 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp, 4915 struct ifa6_config *cfg, clock_t expires, 4916 u32 flags) 4917 { 4918 bool was_managetempaddr; 4919 bool new_peer = false; 4920 bool had_prefixroute; 4921 4922 ASSERT_RTNL_NET(net); 4923 4924 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && 4925 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64)) 4926 return -EINVAL; 4927 4928 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED) 4929 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC; 4930 4931 if (cfg->peer_pfx && 4932 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) { 4933 if (!ipv6_addr_any(&ifp->peer_addr)) 4934 cleanup_prefix_route(ifp, expires, true, true); 4935 new_peer = true; 4936 } 4937 4938 spin_lock_bh(&ifp->lock); 4939 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR; 4940 had_prefixroute = ifp->flags & IFA_F_PERMANENT && 4941 !(ifp->flags & IFA_F_NOPREFIXROUTE); 4942 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | 4943 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR | 4944 IFA_F_NOPREFIXROUTE); 4945 ifp->flags |= cfg->ifa_flags; 4946 WRITE_ONCE(ifp->tstamp, jiffies); 4947 WRITE_ONCE(ifp->valid_lft, cfg->valid_lft); 4948 WRITE_ONCE(ifp->prefered_lft, cfg->preferred_lft); 4949 WRITE_ONCE(ifp->ifa_proto, cfg->ifa_proto); 4950 4951 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority) 4952 WRITE_ONCE(ifp->rt_priority, cfg->rt_priority); 4953 4954 if (new_peer) 4955 ifp->peer_addr = *cfg->peer_pfx; 4956 4957 spin_unlock_bh(&ifp->lock); 4958 if (!(ifp->flags&IFA_F_TENTATIVE)) 4959 ipv6_ifa_notify(0, ifp); 4960 4961 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 4962 int rc = -ENOENT; 4963 4964 if (had_prefixroute) 4965 rc = modify_prefix_route(net, ifp, expires, flags, false); 4966 4967 /* prefix route could have been deleted; if so restore it */ 4968 if (rc == -ENOENT) { 4969 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 4970 ifp->rt_priority, ifp->idev->dev, 4971 expires, flags, GFP_KERNEL); 4972 } 4973 4974 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr)) 4975 rc = modify_prefix_route(net, ifp, expires, flags, true); 4976 4977 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) { 4978 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len, 4979 ifp->rt_priority, ifp->idev->dev, 4980 expires, flags, GFP_KERNEL); 4981 } 4982 } else if (had_prefixroute) { 4983 enum cleanup_prefix_rt_t action; 4984 unsigned long rt_expires; 4985 4986 write_lock_bh(&ifp->idev->lock); 4987 action = check_cleanup_prefix_route(ifp, &rt_expires); 4988 write_unlock_bh(&ifp->idev->lock); 4989 4990 if (action != CLEANUP_PREFIX_RT_NOP) { 4991 cleanup_prefix_route(ifp, rt_expires, 4992 action == CLEANUP_PREFIX_RT_DEL, false); 4993 } 4994 } 4995 4996 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) { 4997 if (was_managetempaddr && !(ifp->flags & IFA_F_MANAGETEMPADDR)) 4998 delete_tempaddrs(ifp->idev, ifp); 4999 else 5000 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft, 5001 cfg->preferred_lft, !was_managetempaddr, 5002 jiffies); 5003 } 5004 5005 addrconf_verify_rtnl(net); 5006 5007 return 0; 5008 } 5009 5010 static int 5011 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, 5012 struct netlink_ext_ack *extack) 5013 { 5014 struct net *net = sock_net(skb->sk); 5015 struct nlattr *tb[IFA_MAX+1]; 5016 struct in6_addr *peer_pfx; 5017 struct inet6_ifaddr *ifa; 5018 struct net_device *dev; 5019 struct inet6_dev *idev; 5020 struct ifa6_config cfg; 5021 struct ifaddrmsg *ifm; 5022 unsigned long timeout; 5023 clock_t expires; 5024 u32 flags; 5025 int err; 5026 5027 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 5028 ifa_ipv6_policy, extack); 5029 if (err < 0) 5030 return err; 5031 5032 memset(&cfg, 0, sizeof(cfg)); 5033 5034 ifm = nlmsg_data(nlh); 5035 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 5036 if (!cfg.pfx) 5037 return -EINVAL; 5038 5039 cfg.peer_pfx = peer_pfx; 5040 cfg.plen = ifm->ifa_prefixlen; 5041 if (tb[IFA_RT_PRIORITY]) 5042 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]); 5043 5044 if (tb[IFA_PROTO]) 5045 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]); 5046 5047 cfg.ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags); 5048 5049 /* We ignore other flags so far. */ 5050 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | 5051 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE | 5052 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC; 5053 5054 cfg.ifa_flags |= IFA_F_PERMANENT; 5055 cfg.valid_lft = INFINITY_LIFE_TIME; 5056 cfg.preferred_lft = INFINITY_LIFE_TIME; 5057 expires = 0; 5058 flags = 0; 5059 5060 if (tb[IFA_CACHEINFO]) { 5061 struct ifa_cacheinfo *ci; 5062 5063 ci = nla_data(tb[IFA_CACHEINFO]); 5064 cfg.valid_lft = ci->ifa_valid; 5065 cfg.preferred_lft = ci->ifa_prefered; 5066 5067 if (!cfg.valid_lft || cfg.preferred_lft > cfg.valid_lft) { 5068 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid"); 5069 return -EINVAL; 5070 } 5071 5072 timeout = addrconf_timeout_fixup(cfg.valid_lft, HZ); 5073 if (addrconf_finite_timeout(timeout)) { 5074 cfg.ifa_flags &= ~IFA_F_PERMANENT; 5075 cfg.valid_lft = timeout; 5076 expires = jiffies_to_clock_t(timeout * HZ); 5077 flags = RTF_EXPIRES; 5078 } 5079 5080 timeout = addrconf_timeout_fixup(cfg.preferred_lft, HZ); 5081 if (addrconf_finite_timeout(timeout)) { 5082 if (timeout == 0) 5083 cfg.ifa_flags |= IFA_F_DEPRECATED; 5084 5085 cfg.preferred_lft = timeout; 5086 } 5087 } 5088 5089 rtnl_net_lock(net); 5090 5091 dev = __dev_get_by_index(net, ifm->ifa_index); 5092 if (!dev) { 5093 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 5094 err = -ENODEV; 5095 goto unlock_rtnl; 5096 } 5097 5098 netdev_lock_ops(dev); 5099 idev = ipv6_find_idev(dev); 5100 if (IS_ERR(idev)) { 5101 err = PTR_ERR(idev); 5102 goto unlock; 5103 } 5104 5105 if (!ipv6_allow_optimistic_dad(net, idev)) 5106 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC; 5107 5108 if (cfg.ifa_flags & IFA_F_NODAD && 5109 cfg.ifa_flags & IFA_F_OPTIMISTIC) { 5110 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive"); 5111 err = -EINVAL; 5112 goto unlock; 5113 } 5114 5115 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1); 5116 if (!ifa) { 5117 /* 5118 * It would be best to check for !NLM_F_CREATE here but 5119 * userspace already relies on not having to provide this. 5120 */ 5121 err = inet6_addr_add(net, dev, &cfg, expires, flags, extack); 5122 goto unlock; 5123 } 5124 5125 if (nlh->nlmsg_flags & NLM_F_EXCL || 5126 !(nlh->nlmsg_flags & NLM_F_REPLACE)) { 5127 NL_SET_ERR_MSG_MOD(extack, "address already assigned"); 5128 err = -EEXIST; 5129 } else { 5130 err = inet6_addr_modify(net, ifa, &cfg, expires, flags); 5131 } 5132 5133 in6_ifa_put(ifa); 5134 unlock: 5135 netdev_unlock_ops(dev); 5136 unlock_rtnl: 5137 rtnl_net_unlock(net); 5138 5139 return err; 5140 } 5141 5142 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags, 5143 u8 scope, int ifindex) 5144 { 5145 struct ifaddrmsg *ifm; 5146 5147 ifm = nlmsg_data(nlh); 5148 ifm->ifa_family = AF_INET6; 5149 ifm->ifa_prefixlen = prefixlen; 5150 ifm->ifa_flags = flags; 5151 ifm->ifa_scope = scope; 5152 ifm->ifa_index = ifindex; 5153 } 5154 5155 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp, 5156 unsigned long tstamp, u32 preferred, u32 valid) 5157 { 5158 struct ifa_cacheinfo ci; 5159 5160 ci.cstamp = cstamp_delta(cstamp); 5161 ci.tstamp = cstamp_delta(tstamp); 5162 ci.ifa_prefered = preferred; 5163 ci.ifa_valid = valid; 5164 5165 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci); 5166 } 5167 5168 static inline int rt_scope(int ifa_scope) 5169 { 5170 if (ifa_scope & IFA_HOST) 5171 return RT_SCOPE_HOST; 5172 else if (ifa_scope & IFA_LINK) 5173 return RT_SCOPE_LINK; 5174 else if (ifa_scope & IFA_SITE) 5175 return RT_SCOPE_SITE; 5176 else 5177 return RT_SCOPE_UNIVERSE; 5178 } 5179 5180 static inline int inet6_ifaddr_msgsize(void) 5181 { 5182 return NLMSG_ALIGN(sizeof(struct ifaddrmsg)) 5183 + nla_total_size(16) /* IFA_LOCAL */ 5184 + nla_total_size(16) /* IFA_ADDRESS */ 5185 + nla_total_size(sizeof(struct ifa_cacheinfo)) 5186 + nla_total_size(4) /* IFA_FLAGS */ 5187 + nla_total_size(1) /* IFA_PROTO */ 5188 + nla_total_size(4) /* IFA_RT_PRIORITY */; 5189 } 5190 5191 static int inet6_fill_ifaddr(struct sk_buff *skb, 5192 const struct inet6_ifaddr *ifa, 5193 struct inet6_fill_args *args) 5194 { 5195 struct nlmsghdr *nlh; 5196 u32 preferred, valid; 5197 u32 flags, priority; 5198 u8 proto; 5199 5200 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5201 sizeof(struct ifaddrmsg), args->flags); 5202 if (!nlh) 5203 return -EMSGSIZE; 5204 5205 flags = READ_ONCE(ifa->flags); 5206 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope), 5207 ifa->idev->dev->ifindex); 5208 5209 if (args->netnsid >= 0 && 5210 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) 5211 goto error; 5212 5213 preferred = READ_ONCE(ifa->prefered_lft); 5214 valid = READ_ONCE(ifa->valid_lft); 5215 5216 if (!((flags & IFA_F_PERMANENT) && 5217 (preferred == INFINITY_LIFE_TIME))) { 5218 if (preferred != INFINITY_LIFE_TIME) { 5219 long tval = (jiffies - READ_ONCE(ifa->tstamp)) / HZ; 5220 5221 if (preferred > tval) 5222 preferred -= tval; 5223 else 5224 preferred = 0; 5225 if (valid != INFINITY_LIFE_TIME) { 5226 if (valid > tval) 5227 valid -= tval; 5228 else 5229 valid = 0; 5230 } 5231 } 5232 } else { 5233 preferred = INFINITY_LIFE_TIME; 5234 valid = INFINITY_LIFE_TIME; 5235 } 5236 5237 if (!ipv6_addr_any(&ifa->peer_addr)) { 5238 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 || 5239 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0) 5240 goto error; 5241 } else { 5242 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0) 5243 goto error; 5244 } 5245 5246 priority = READ_ONCE(ifa->rt_priority); 5247 if (priority && nla_put_u32(skb, IFA_RT_PRIORITY, priority)) 5248 goto error; 5249 5250 if (put_cacheinfo(skb, ifa->cstamp, READ_ONCE(ifa->tstamp), 5251 preferred, valid) < 0) 5252 goto error; 5253 5254 if (nla_put_u32(skb, IFA_FLAGS, flags) < 0) 5255 goto error; 5256 5257 proto = READ_ONCE(ifa->ifa_proto); 5258 if (proto && nla_put_u8(skb, IFA_PROTO, proto)) 5259 goto error; 5260 5261 nlmsg_end(skb, nlh); 5262 return 0; 5263 5264 error: 5265 nlmsg_cancel(skb, nlh); 5266 return -EMSGSIZE; 5267 } 5268 5269 int inet6_fill_ifmcaddr(struct sk_buff *skb, 5270 const struct ifmcaddr6 *ifmca, 5271 struct inet6_fill_args *args) 5272 { 5273 int ifindex = ifmca->idev->dev->ifindex; 5274 u8 scope = RT_SCOPE_UNIVERSE; 5275 struct nlmsghdr *nlh; 5276 5277 if (!args->force_rt_scope_universe && 5278 ipv6_addr_scope(&ifmca->mca_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_MULTICAST, &ifmca->mca_addr) < 0 || 5294 nla_put_u32(skb, IFA_MC_USERS, READ_ONCE(ifmca->mca_users)) < 0 || 5295 put_cacheinfo(skb, ifmca->mca_cstamp, READ_ONCE(ifmca->mca_tstamp), 5296 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5297 nlmsg_cancel(skb, nlh); 5298 return -EMSGSIZE; 5299 } 5300 5301 nlmsg_end(skb, nlh); 5302 return 0; 5303 } 5304 5305 int inet6_fill_ifacaddr(struct sk_buff *skb, 5306 const struct ifacaddr6 *ifaca, 5307 struct inet6_fill_args *args) 5308 { 5309 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt); 5310 int ifindex = dev ? dev->ifindex : 1; 5311 u8 scope = RT_SCOPE_UNIVERSE; 5312 struct nlmsghdr *nlh; 5313 5314 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE) 5315 scope = RT_SCOPE_SITE; 5316 5317 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5318 sizeof(struct ifaddrmsg), args->flags); 5319 if (!nlh) 5320 return -EMSGSIZE; 5321 5322 if (args->netnsid >= 0 && 5323 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) { 5324 nlmsg_cancel(skb, nlh); 5325 return -EMSGSIZE; 5326 } 5327 5328 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 5329 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 || 5330 put_cacheinfo(skb, ifaca->aca_cstamp, READ_ONCE(ifaca->aca_tstamp), 5331 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5332 nlmsg_cancel(skb, nlh); 5333 return -EMSGSIZE; 5334 } 5335 5336 nlmsg_end(skb, nlh); 5337 return 0; 5338 } 5339 5340 /* called with rcu_read_lock() */ 5341 static int in6_dump_addrs(const struct inet6_dev *idev, struct sk_buff *skb, 5342 struct netlink_callback *cb, int *s_ip_idx, 5343 struct inet6_fill_args *fillargs) 5344 { 5345 const struct ifmcaddr6 *ifmca; 5346 const struct ifacaddr6 *ifaca; 5347 int ip_idx = 0; 5348 int err = 0; 5349 5350 switch (fillargs->type) { 5351 case UNICAST_ADDR: { 5352 const struct inet6_ifaddr *ifa; 5353 fillargs->event = RTM_NEWADDR; 5354 5355 /* unicast address incl. temp addr */ 5356 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 5357 if (ip_idx < *s_ip_idx) 5358 goto next; 5359 err = inet6_fill_ifaddr(skb, ifa, fillargs); 5360 if (err < 0) 5361 break; 5362 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 5363 next: 5364 ip_idx++; 5365 } 5366 break; 5367 } 5368 case MULTICAST_ADDR: 5369 fillargs->event = RTM_GETMULTICAST; 5370 5371 /* multicast address */ 5372 for (ifmca = rcu_dereference(idev->mc_list); 5373 ifmca; 5374 ifmca = rcu_dereference(ifmca->next), ip_idx++) { 5375 if (ip_idx < *s_ip_idx) 5376 continue; 5377 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs); 5378 if (err < 0) 5379 break; 5380 } 5381 break; 5382 case ANYCAST_ADDR: 5383 fillargs->event = RTM_GETANYCAST; 5384 /* anycast address */ 5385 for (ifaca = rcu_dereference(idev->ac_list); ifaca; 5386 ifaca = rcu_dereference(ifaca->aca_next), ip_idx++) { 5387 if (ip_idx < *s_ip_idx) 5388 continue; 5389 err = inet6_fill_ifacaddr(skb, ifaca, fillargs); 5390 if (err < 0) 5391 break; 5392 } 5393 break; 5394 default: 5395 break; 5396 } 5397 *s_ip_idx = err ? ip_idx : 0; 5398 return err; 5399 } 5400 5401 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh, 5402 struct inet6_fill_args *fillargs, 5403 struct net **tgt_net, struct sock *sk, 5404 struct netlink_callback *cb) 5405 { 5406 struct netlink_ext_ack *extack = cb->extack; 5407 struct nlattr *tb[IFA_MAX+1]; 5408 struct ifaddrmsg *ifm; 5409 int err, i; 5410 5411 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 5412 if (!ifm) { 5413 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request"); 5414 return -EINVAL; 5415 } 5416 5417 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5418 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request"); 5419 return -EINVAL; 5420 } 5421 5422 fillargs->ifindex = ifm->ifa_index; 5423 if (fillargs->ifindex) { 5424 cb->answer_flags |= NLM_F_DUMP_FILTERED; 5425 fillargs->flags |= NLM_F_DUMP_FILTERED; 5426 } 5427 5428 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5429 ifa_ipv6_policy, extack); 5430 if (err < 0) 5431 return err; 5432 5433 for (i = 0; i <= IFA_MAX; ++i) { 5434 if (!tb[i]) 5435 continue; 5436 5437 if (i == IFA_TARGET_NETNSID) { 5438 struct net *net; 5439 5440 fillargs->netnsid = nla_get_s32(tb[i]); 5441 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid); 5442 if (IS_ERR(net)) { 5443 fillargs->netnsid = -1; 5444 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id"); 5445 return PTR_ERR(net); 5446 } 5447 *tgt_net = net; 5448 } else { 5449 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request"); 5450 return -EINVAL; 5451 } 5452 } 5453 5454 return 0; 5455 } 5456 5457 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb, 5458 enum addr_type_t type) 5459 { 5460 struct net *tgt_net = sock_net(skb->sk); 5461 const struct nlmsghdr *nlh = cb->nlh; 5462 struct inet6_fill_args fillargs = { 5463 .portid = NETLINK_CB(cb->skb).portid, 5464 .seq = cb->nlh->nlmsg_seq, 5465 .flags = NLM_F_MULTI, 5466 .netnsid = -1, 5467 .type = type, 5468 .force_rt_scope_universe = false, 5469 }; 5470 struct { 5471 unsigned long ifindex; 5472 int ip_idx; 5473 } *ctx = (void *)cb->ctx; 5474 struct net_device *dev; 5475 struct inet6_dev *idev; 5476 int err = 0; 5477 5478 rcu_read_lock(); 5479 if (cb->strict_check) { 5480 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net, 5481 skb->sk, cb); 5482 if (err < 0) 5483 goto done; 5484 5485 err = 0; 5486 if (fillargs.ifindex) { 5487 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex); 5488 if (!dev) { 5489 err = -ENODEV; 5490 goto done; 5491 } 5492 idev = __in6_dev_get(dev); 5493 if (idev) 5494 err = in6_dump_addrs(idev, skb, cb, 5495 &ctx->ip_idx, 5496 &fillargs); 5497 goto done; 5498 } 5499 } 5500 5501 cb->seq = inet6_base_seq(tgt_net); 5502 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) { 5503 idev = __in6_dev_get(dev); 5504 if (!idev) 5505 continue; 5506 err = in6_dump_addrs(idev, skb, cb, &ctx->ip_idx, 5507 &fillargs); 5508 if (err < 0) 5509 goto done; 5510 } 5511 done: 5512 rcu_read_unlock(); 5513 if (fillargs.netnsid >= 0) 5514 put_net(tgt_net); 5515 5516 return err; 5517 } 5518 5519 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb) 5520 { 5521 enum addr_type_t type = UNICAST_ADDR; 5522 5523 return inet6_dump_addr(skb, cb, type); 5524 } 5525 5526 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb) 5527 { 5528 enum addr_type_t type = MULTICAST_ADDR; 5529 5530 return inet6_dump_addr(skb, cb, type); 5531 } 5532 5533 5534 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb) 5535 { 5536 enum addr_type_t type = ANYCAST_ADDR; 5537 5538 return inet6_dump_addr(skb, cb, type); 5539 } 5540 5541 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb, 5542 const struct nlmsghdr *nlh, 5543 struct nlattr **tb, 5544 struct netlink_ext_ack *extack) 5545 { 5546 struct ifaddrmsg *ifm; 5547 int i, err; 5548 5549 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 5550 if (!ifm) { 5551 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request"); 5552 return -EINVAL; 5553 } 5554 5555 if (!netlink_strict_get_check(skb)) 5556 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 5557 ifa_ipv6_policy, extack); 5558 5559 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5560 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request"); 5561 return -EINVAL; 5562 } 5563 5564 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5565 ifa_ipv6_policy, extack); 5566 if (err) 5567 return err; 5568 5569 for (i = 0; i <= IFA_MAX; i++) { 5570 if (!tb[i]) 5571 continue; 5572 5573 switch (i) { 5574 case IFA_TARGET_NETNSID: 5575 case IFA_ADDRESS: 5576 case IFA_LOCAL: 5577 break; 5578 default: 5579 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request"); 5580 return -EINVAL; 5581 } 5582 } 5583 5584 return 0; 5585 } 5586 5587 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh, 5588 struct netlink_ext_ack *extack) 5589 { 5590 struct net *tgt_net = sock_net(in_skb->sk); 5591 struct inet6_fill_args fillargs = { 5592 .portid = NETLINK_CB(in_skb).portid, 5593 .seq = nlh->nlmsg_seq, 5594 .event = RTM_NEWADDR, 5595 .flags = 0, 5596 .netnsid = -1, 5597 .force_rt_scope_universe = false, 5598 }; 5599 struct ifaddrmsg *ifm; 5600 struct nlattr *tb[IFA_MAX+1]; 5601 struct in6_addr *addr = NULL, *peer; 5602 struct net_device *dev = NULL; 5603 struct inet6_ifaddr *ifa; 5604 struct sk_buff *skb; 5605 int err; 5606 5607 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack); 5608 if (err < 0) 5609 return err; 5610 5611 if (tb[IFA_TARGET_NETNSID]) { 5612 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]); 5613 5614 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk, 5615 fillargs.netnsid); 5616 if (IS_ERR(tgt_net)) 5617 return PTR_ERR(tgt_net); 5618 } 5619 5620 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer); 5621 if (!addr) { 5622 err = -EINVAL; 5623 goto errout; 5624 } 5625 ifm = nlmsg_data(nlh); 5626 if (ifm->ifa_index) 5627 dev = dev_get_by_index(tgt_net, ifm->ifa_index); 5628 5629 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1); 5630 if (!ifa) { 5631 err = -EADDRNOTAVAIL; 5632 goto errout; 5633 } 5634 5635 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL); 5636 if (!skb) { 5637 err = -ENOBUFS; 5638 goto errout_ifa; 5639 } 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_ifa; 5647 } 5648 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid); 5649 errout_ifa: 5650 in6_ifa_put(ifa); 5651 errout: 5652 dev_put(dev); 5653 if (fillargs.netnsid >= 0) 5654 put_net(tgt_net); 5655 5656 return err; 5657 } 5658 5659 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa) 5660 { 5661 struct sk_buff *skb; 5662 struct net *net = dev_net(ifa->idev->dev); 5663 struct inet6_fill_args fillargs = { 5664 .portid = 0, 5665 .seq = 0, 5666 .event = event, 5667 .flags = 0, 5668 .netnsid = -1, 5669 .force_rt_scope_universe = false, 5670 }; 5671 int err = -ENOBUFS; 5672 5673 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC); 5674 if (!skb) 5675 goto errout; 5676 5677 err = inet6_fill_ifaddr(skb, ifa, &fillargs); 5678 if (err < 0) { 5679 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 5680 WARN_ON(err == -EMSGSIZE); 5681 kfree_skb(skb); 5682 goto errout; 5683 } 5684 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC); 5685 return; 5686 errout: 5687 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err); 5688 } 5689 5690 static void ipv6_store_devconf(const struct ipv6_devconf *cnf, 5691 __s32 *array, int bytes) 5692 { 5693 BUG_ON(bytes < (DEVCONF_MAX * 4)); 5694 5695 memset(array, 0, bytes); 5696 array[DEVCONF_FORWARDING] = READ_ONCE(cnf->forwarding); 5697 array[DEVCONF_HOPLIMIT] = READ_ONCE(cnf->hop_limit); 5698 array[DEVCONF_MTU6] = READ_ONCE(cnf->mtu6); 5699 array[DEVCONF_ACCEPT_RA] = READ_ONCE(cnf->accept_ra); 5700 array[DEVCONF_ACCEPT_REDIRECTS] = READ_ONCE(cnf->accept_redirects); 5701 array[DEVCONF_AUTOCONF] = READ_ONCE(cnf->autoconf); 5702 array[DEVCONF_DAD_TRANSMITS] = READ_ONCE(cnf->dad_transmits); 5703 array[DEVCONF_RTR_SOLICITS] = READ_ONCE(cnf->rtr_solicits); 5704 array[DEVCONF_RTR_SOLICIT_INTERVAL] = 5705 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_interval)); 5706 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] = 5707 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_max_interval)); 5708 array[DEVCONF_RTR_SOLICIT_DELAY] = 5709 jiffies_to_msecs(READ_ONCE(cnf->rtr_solicit_delay)); 5710 array[DEVCONF_FORCE_MLD_VERSION] = READ_ONCE(cnf->force_mld_version); 5711 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] = 5712 jiffies_to_msecs(READ_ONCE(cnf->mldv1_unsolicited_report_interval)); 5713 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] = 5714 jiffies_to_msecs(READ_ONCE(cnf->mldv2_unsolicited_report_interval)); 5715 array[DEVCONF_USE_TEMPADDR] = READ_ONCE(cnf->use_tempaddr); 5716 array[DEVCONF_TEMP_VALID_LFT] = READ_ONCE(cnf->temp_valid_lft); 5717 array[DEVCONF_TEMP_PREFERED_LFT] = READ_ONCE(cnf->temp_prefered_lft); 5718 array[DEVCONF_REGEN_MAX_RETRY] = READ_ONCE(cnf->regen_max_retry); 5719 array[DEVCONF_MAX_DESYNC_FACTOR] = READ_ONCE(cnf->max_desync_factor); 5720 array[DEVCONF_MAX_ADDRESSES] = READ_ONCE(cnf->max_addresses); 5721 array[DEVCONF_ACCEPT_RA_DEFRTR] = READ_ONCE(cnf->accept_ra_defrtr); 5722 array[DEVCONF_RA_DEFRTR_METRIC] = READ_ONCE(cnf->ra_defrtr_metric); 5723 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = 5724 READ_ONCE(cnf->accept_ra_min_hop_limit); 5725 array[DEVCONF_ACCEPT_RA_PINFO] = READ_ONCE(cnf->accept_ra_pinfo); 5726 #ifdef CONFIG_IPV6_ROUTER_PREF 5727 array[DEVCONF_ACCEPT_RA_RTR_PREF] = READ_ONCE(cnf->accept_ra_rtr_pref); 5728 array[DEVCONF_RTR_PROBE_INTERVAL] = 5729 jiffies_to_msecs(READ_ONCE(cnf->rtr_probe_interval)); 5730 #ifdef CONFIG_IPV6_ROUTE_INFO 5731 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = 5732 READ_ONCE(cnf->accept_ra_rt_info_min_plen); 5733 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = 5734 READ_ONCE(cnf->accept_ra_rt_info_max_plen); 5735 #endif 5736 #endif 5737 array[DEVCONF_PROXY_NDP] = READ_ONCE(cnf->proxy_ndp); 5738 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = 5739 READ_ONCE(cnf->accept_source_route); 5740 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 5741 array[DEVCONF_OPTIMISTIC_DAD] = READ_ONCE(cnf->optimistic_dad); 5742 array[DEVCONF_USE_OPTIMISTIC] = READ_ONCE(cnf->use_optimistic); 5743 #endif 5744 #ifdef CONFIG_IPV6_MROUTE 5745 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding); 5746 #endif 5747 array[DEVCONF_DISABLE_IPV6] = READ_ONCE(cnf->disable_ipv6); 5748 array[DEVCONF_ACCEPT_DAD] = READ_ONCE(cnf->accept_dad); 5749 array[DEVCONF_FORCE_TLLAO] = READ_ONCE(cnf->force_tllao); 5750 array[DEVCONF_NDISC_NOTIFY] = READ_ONCE(cnf->ndisc_notify); 5751 array[DEVCONF_SUPPRESS_FRAG_NDISC] = 5752 READ_ONCE(cnf->suppress_frag_ndisc); 5753 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = 5754 READ_ONCE(cnf->accept_ra_from_local); 5755 array[DEVCONF_ACCEPT_RA_MTU] = READ_ONCE(cnf->accept_ra_mtu); 5756 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = 5757 READ_ONCE(cnf->ignore_routes_with_linkdown); 5758 /* we omit DEVCONF_STABLE_SECRET for now */ 5759 array[DEVCONF_USE_OIF_ADDRS_ONLY] = READ_ONCE(cnf->use_oif_addrs_only); 5760 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = 5761 READ_ONCE(cnf->drop_unicast_in_l2_multicast); 5762 array[DEVCONF_DROP_UNSOLICITED_NA] = READ_ONCE(cnf->drop_unsolicited_na); 5763 array[DEVCONF_KEEP_ADDR_ON_DOWN] = READ_ONCE(cnf->keep_addr_on_down); 5764 array[DEVCONF_SEG6_ENABLED] = READ_ONCE(cnf->seg6_enabled); 5765 #ifdef CONFIG_IPV6_SEG6_HMAC 5766 array[DEVCONF_SEG6_REQUIRE_HMAC] = READ_ONCE(cnf->seg6_require_hmac); 5767 #endif 5768 array[DEVCONF_ENHANCED_DAD] = READ_ONCE(cnf->enhanced_dad); 5769 array[DEVCONF_ADDR_GEN_MODE] = READ_ONCE(cnf->addr_gen_mode); 5770 array[DEVCONF_DISABLE_POLICY] = READ_ONCE(cnf->disable_policy); 5771 array[DEVCONF_NDISC_TCLASS] = READ_ONCE(cnf->ndisc_tclass); 5772 array[DEVCONF_RPL_SEG_ENABLED] = READ_ONCE(cnf->rpl_seg_enabled); 5773 array[DEVCONF_IOAM6_ENABLED] = READ_ONCE(cnf->ioam6_enabled); 5774 array[DEVCONF_IOAM6_ID] = READ_ONCE(cnf->ioam6_id); 5775 array[DEVCONF_IOAM6_ID_WIDE] = READ_ONCE(cnf->ioam6_id_wide); 5776 array[DEVCONF_NDISC_EVICT_NOCARRIER] = 5777 READ_ONCE(cnf->ndisc_evict_nocarrier); 5778 array[DEVCONF_ACCEPT_UNTRACKED_NA] = 5779 READ_ONCE(cnf->accept_untracked_na); 5780 array[DEVCONF_ACCEPT_RA_MIN_LFT] = READ_ONCE(cnf->accept_ra_min_lft); 5781 array[DEVCONF_FORCE_FORWARDING] = READ_ONCE(cnf->force_forwarding); 5782 } 5783 5784 static inline size_t inet6_ifla6_size(void) 5785 { 5786 return nla_total_size(4) /* IFLA_INET6_FLAGS */ 5787 + nla_total_size(sizeof(struct ifla_cacheinfo)) 5788 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */ 5789 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */ 5790 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */ 5791 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */ 5792 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */ 5793 + nla_total_size(4) /* IFLA_INET6_RA_MTU */ 5794 + 0; 5795 } 5796 5797 static inline size_t inet6_if_nlmsg_size(void) 5798 { 5799 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 5800 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 5801 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 5802 + nla_total_size(4) /* IFLA_MTU */ 5803 + nla_total_size(4) /* IFLA_LINK */ 5804 + nla_total_size(1) /* IFLA_OPERSTATE */ 5805 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */ 5806 } 5807 5808 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib, 5809 int bytes) 5810 { 5811 int i; 5812 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX; 5813 BUG_ON(pad < 0); 5814 5815 /* Use put_unaligned() because stats may not be aligned for u64. */ 5816 put_unaligned(ICMP6_MIB_MAX, &stats[0]); 5817 for (i = 1; i < ICMP6_MIB_MAX; i++) 5818 put_unaligned(atomic_long_read(&mib[i]), &stats[i]); 5819 5820 memset(&stats[ICMP6_MIB_MAX], 0, pad); 5821 } 5822 5823 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib, 5824 int bytes, size_t syncpoff) 5825 { 5826 int i, c; 5827 u64 buff[IPSTATS_MIB_MAX]; 5828 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX; 5829 5830 BUG_ON(pad < 0); 5831 5832 memset(buff, 0, sizeof(buff)); 5833 buff[0] = IPSTATS_MIB_MAX; 5834 5835 for_each_possible_cpu(c) { 5836 for (i = 1; i < IPSTATS_MIB_MAX; i++) 5837 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff); 5838 } 5839 5840 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64)); 5841 memset(&stats[IPSTATS_MIB_MAX], 0, pad); 5842 } 5843 5844 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype, 5845 int bytes) 5846 { 5847 switch (attrtype) { 5848 case IFLA_INET6_STATS: 5849 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes, 5850 offsetof(struct ipstats_mib, syncp)); 5851 break; 5852 case IFLA_INET6_ICMP6STATS: 5853 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes); 5854 break; 5855 } 5856 } 5857 5858 static int inet6_fill_ifla6_stats_attrs(struct sk_buff *skb, 5859 struct inet6_dev *idev) 5860 { 5861 struct nlattr *nla; 5862 5863 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64)); 5864 if (!nla) 5865 goto nla_put_failure; 5866 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla)); 5867 5868 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64)); 5869 if (!nla) 5870 goto nla_put_failure; 5871 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla)); 5872 5873 return 0; 5874 5875 nla_put_failure: 5876 return -EMSGSIZE; 5877 } 5878 5879 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev, 5880 u32 ext_filter_mask) 5881 { 5882 struct ifla_cacheinfo ci; 5883 struct nlattr *nla; 5884 u32 ra_mtu; 5885 5886 if (nla_put_u32(skb, IFLA_INET6_FLAGS, READ_ONCE(idev->if_flags))) 5887 goto nla_put_failure; 5888 ci.max_reasm_len = IPV6_MAXPLEN; 5889 ci.tstamp = cstamp_delta(READ_ONCE(idev->tstamp)); 5890 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time); 5891 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME)); 5892 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci)) 5893 goto nla_put_failure; 5894 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32)); 5895 if (!nla) 5896 goto nla_put_failure; 5897 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla)); 5898 5899 /* XXX - MC not implemented */ 5900 5901 if (!(ext_filter_mask & RTEXT_FILTER_SKIP_STATS)) { 5902 if (inet6_fill_ifla6_stats_attrs(skb, idev) < 0) 5903 goto nla_put_failure; 5904 } 5905 5906 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr)); 5907 if (!nla) 5908 goto nla_put_failure; 5909 read_lock_bh(&idev->lock); 5910 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla)); 5911 read_unlock_bh(&idev->lock); 5912 5913 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, 5914 READ_ONCE(idev->cnf.addr_gen_mode))) 5915 goto nla_put_failure; 5916 5917 ra_mtu = READ_ONCE(idev->ra_mtu); 5918 if (ra_mtu && nla_put_u32(skb, IFLA_INET6_RA_MTU, ra_mtu)) 5919 goto nla_put_failure; 5920 5921 return 0; 5922 5923 nla_put_failure: 5924 return -EMSGSIZE; 5925 } 5926 5927 static size_t inet6_get_link_af_size(const struct net_device *dev, 5928 u32 ext_filter_mask) 5929 { 5930 if (!__in6_dev_get(dev)) 5931 return 0; 5932 5933 return inet6_ifla6_size(); 5934 } 5935 5936 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev, 5937 u32 ext_filter_mask) 5938 { 5939 struct inet6_dev *idev = __in6_dev_get(dev); 5940 5941 if (!idev) 5942 return -ENODATA; 5943 5944 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0) 5945 return -EMSGSIZE; 5946 5947 return 0; 5948 } 5949 5950 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token, 5951 struct netlink_ext_ack *extack) 5952 { 5953 struct inet6_ifaddr *ifp; 5954 struct net_device *dev = idev->dev; 5955 bool clear_token, update_rs = false; 5956 struct in6_addr ll_addr; 5957 5958 ASSERT_RTNL(); 5959 5960 if (!token) 5961 return -EINVAL; 5962 5963 if (dev->flags & IFF_LOOPBACK) { 5964 NL_SET_ERR_MSG_MOD(extack, "Device is loopback"); 5965 return -EINVAL; 5966 } 5967 5968 if (dev->flags & IFF_NOARP) { 5969 NL_SET_ERR_MSG_MOD(extack, 5970 "Device does not do neighbour discovery"); 5971 return -EINVAL; 5972 } 5973 5974 if (!ipv6_accept_ra(idev)) { 5975 NL_SET_ERR_MSG_MOD(extack, 5976 "Router advertisement is disabled on device"); 5977 return -EINVAL; 5978 } 5979 5980 if (READ_ONCE(idev->cnf.rtr_solicits) == 0) { 5981 NL_SET_ERR_MSG(extack, 5982 "Router solicitation is disabled on device"); 5983 return -EINVAL; 5984 } 5985 5986 write_lock_bh(&idev->lock); 5987 5988 BUILD_BUG_ON(sizeof(token->s6_addr) != 16); 5989 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8); 5990 5991 write_unlock_bh(&idev->lock); 5992 5993 clear_token = ipv6_addr_any(token); 5994 if (clear_token) 5995 goto update_lft; 5996 5997 if (!idev->dead && (idev->if_flags & IF_READY) && 5998 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE | 5999 IFA_F_OPTIMISTIC)) { 6000 /* If we're not ready, then normal ifup will take care 6001 * of this. Otherwise, we need to request our rs here. 6002 */ 6003 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters); 6004 update_rs = true; 6005 } 6006 6007 update_lft: 6008 write_lock_bh(&idev->lock); 6009 6010 if (update_rs) { 6011 idev->if_flags |= IF_RS_SENT; 6012 idev->rs_interval = rfc3315_s14_backoff_init( 6013 READ_ONCE(idev->cnf.rtr_solicit_interval)); 6014 idev->rs_probes = 1; 6015 addrconf_mod_rs_timer(idev, idev->rs_interval); 6016 } 6017 6018 /* Well, that's kinda nasty ... */ 6019 list_for_each_entry(ifp, &idev->addr_list, if_list) { 6020 spin_lock(&ifp->lock); 6021 if (ifp->tokenized) { 6022 ifp->valid_lft = 0; 6023 ifp->prefered_lft = 0; 6024 } 6025 spin_unlock(&ifp->lock); 6026 } 6027 6028 write_unlock_bh(&idev->lock); 6029 inet6_ifinfo_notify(RTM_NEWLINK, idev); 6030 addrconf_verify_rtnl(dev_net(dev)); 6031 return 0; 6032 } 6033 6034 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = { 6035 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 }, 6036 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) }, 6037 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT, 6038 .reject_message = 6039 "IFLA_INET6_RA_MTU can not be set" }, 6040 }; 6041 6042 static int check_addr_gen_mode(int mode) 6043 { 6044 if (mode != IN6_ADDR_GEN_MODE_EUI64 && 6045 mode != IN6_ADDR_GEN_MODE_NONE && 6046 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 6047 mode != IN6_ADDR_GEN_MODE_RANDOM) 6048 return -EINVAL; 6049 return 1; 6050 } 6051 6052 static int check_stable_privacy(struct inet6_dev *idev, struct net *net, 6053 int mode) 6054 { 6055 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 6056 !idev->cnf.stable_secret.initialized && 6057 !net->ipv6.devconf_dflt->stable_secret.initialized) 6058 return -EINVAL; 6059 return 1; 6060 } 6061 6062 static int inet6_validate_link_af(const struct net_device *dev, 6063 const struct nlattr *nla, 6064 struct netlink_ext_ack *extack) 6065 { 6066 struct nlattr *tb[IFLA_INET6_MAX + 1]; 6067 struct inet6_dev *idev = NULL; 6068 int err; 6069 6070 if (dev) { 6071 idev = __in6_dev_get(dev); 6072 if (!idev) 6073 return -EAFNOSUPPORT; 6074 } 6075 6076 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, 6077 inet6_af_policy, extack); 6078 if (err) 6079 return err; 6080 6081 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE]) 6082 return -EINVAL; 6083 6084 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 6085 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 6086 6087 if (check_addr_gen_mode(mode) < 0) 6088 return -EINVAL; 6089 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0) 6090 return -EINVAL; 6091 } 6092 6093 return 0; 6094 } 6095 6096 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla, 6097 struct netlink_ext_ack *extack) 6098 { 6099 struct inet6_dev *idev = __in6_dev_get(dev); 6100 struct nlattr *tb[IFLA_INET6_MAX + 1]; 6101 int err; 6102 6103 if (!idev) 6104 return -EAFNOSUPPORT; 6105 6106 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0) 6107 return -EINVAL; 6108 6109 if (tb[IFLA_INET6_TOKEN]) { 6110 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]), 6111 extack); 6112 if (err) 6113 return err; 6114 } 6115 6116 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 6117 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 6118 6119 WRITE_ONCE(idev->cnf.addr_gen_mode, mode); 6120 } 6121 6122 return 0; 6123 } 6124 6125 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, 6126 u32 portid, u32 seq, int event, unsigned int flags) 6127 { 6128 struct net_device *dev = idev->dev; 6129 struct ifinfomsg *hdr; 6130 struct nlmsghdr *nlh; 6131 int ifindex, iflink; 6132 void *protoinfo; 6133 6134 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags); 6135 if (!nlh) 6136 return -EMSGSIZE; 6137 6138 hdr = nlmsg_data(nlh); 6139 hdr->ifi_family = AF_INET6; 6140 hdr->__ifi_pad = 0; 6141 hdr->ifi_type = dev->type; 6142 ifindex = READ_ONCE(dev->ifindex); 6143 hdr->ifi_index = ifindex; 6144 hdr->ifi_flags = netif_get_flags(dev); 6145 hdr->ifi_change = 0; 6146 6147 iflink = dev_get_iflink(dev); 6148 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 6149 (dev->addr_len && 6150 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 6151 nla_put_u32(skb, IFLA_MTU, READ_ONCE(dev->mtu)) || 6152 (ifindex != iflink && 6153 nla_put_u32(skb, IFLA_LINK, iflink)) || 6154 nla_put_u8(skb, IFLA_OPERSTATE, 6155 netif_running(dev) ? READ_ONCE(dev->operstate) : IF_OPER_DOWN)) 6156 goto nla_put_failure; 6157 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO); 6158 if (!protoinfo) 6159 goto nla_put_failure; 6160 6161 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0) 6162 goto nla_put_failure; 6163 6164 nla_nest_end(skb, protoinfo); 6165 nlmsg_end(skb, nlh); 6166 return 0; 6167 6168 nla_put_failure: 6169 nlmsg_cancel(skb, nlh); 6170 return -EMSGSIZE; 6171 } 6172 6173 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh, 6174 struct netlink_ext_ack *extack) 6175 { 6176 struct ifinfomsg *ifm; 6177 6178 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 6179 if (!ifm) { 6180 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request"); 6181 return -EINVAL; 6182 } 6183 6184 if (nlmsg_attrlen(nlh, sizeof(*ifm))) { 6185 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header"); 6186 return -EINVAL; 6187 } 6188 6189 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 6190 ifm->ifi_change || ifm->ifi_index) { 6191 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request"); 6192 return -EINVAL; 6193 } 6194 6195 return 0; 6196 } 6197 6198 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 6199 { 6200 struct net *net = sock_net(skb->sk); 6201 struct { 6202 unsigned long ifindex; 6203 } *ctx = (void *)cb->ctx; 6204 struct net_device *dev; 6205 struct inet6_dev *idev; 6206 int err; 6207 6208 /* only requests using strict checking can pass data to 6209 * influence the dump 6210 */ 6211 if (cb->strict_check) { 6212 err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack); 6213 6214 if (err < 0) 6215 return err; 6216 } 6217 6218 err = 0; 6219 rcu_read_lock(); 6220 for_each_netdev_dump(net, dev, ctx->ifindex) { 6221 idev = __in6_dev_get(dev); 6222 if (!idev) 6223 continue; 6224 err = inet6_fill_ifinfo(skb, idev, 6225 NETLINK_CB(cb->skb).portid, 6226 cb->nlh->nlmsg_seq, 6227 RTM_NEWLINK, NLM_F_MULTI); 6228 if (err < 0) 6229 break; 6230 } 6231 rcu_read_unlock(); 6232 6233 return err; 6234 } 6235 6236 void inet6_ifinfo_notify(int event, struct inet6_dev *idev) 6237 { 6238 struct sk_buff *skb; 6239 struct net *net = dev_net(idev->dev); 6240 int err = -ENOBUFS; 6241 6242 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC); 6243 if (!skb) 6244 goto errout; 6245 6246 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0); 6247 if (err < 0) { 6248 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */ 6249 WARN_ON(err == -EMSGSIZE); 6250 kfree_skb(skb); 6251 goto errout; 6252 } 6253 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC); 6254 return; 6255 errout: 6256 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err); 6257 } 6258 6259 static inline size_t inet6_prefix_nlmsg_size(void) 6260 { 6261 return NLMSG_ALIGN(sizeof(struct prefixmsg)) 6262 + nla_total_size(sizeof(struct in6_addr)) 6263 + nla_total_size(sizeof(struct prefix_cacheinfo)); 6264 } 6265 6266 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev, 6267 struct prefix_info *pinfo, u32 portid, u32 seq, 6268 int event, unsigned int flags) 6269 { 6270 struct prefixmsg *pmsg; 6271 struct nlmsghdr *nlh; 6272 struct prefix_cacheinfo ci; 6273 6274 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags); 6275 if (!nlh) 6276 return -EMSGSIZE; 6277 6278 pmsg = nlmsg_data(nlh); 6279 pmsg->prefix_family = AF_INET6; 6280 pmsg->prefix_pad1 = 0; 6281 pmsg->prefix_pad2 = 0; 6282 pmsg->prefix_ifindex = idev->dev->ifindex; 6283 pmsg->prefix_len = pinfo->prefix_len; 6284 pmsg->prefix_type = pinfo->type; 6285 pmsg->prefix_pad3 = 0; 6286 pmsg->prefix_flags = pinfo->flags; 6287 6288 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix)) 6289 goto nla_put_failure; 6290 ci.preferred_time = ntohl(pinfo->prefered); 6291 ci.valid_time = ntohl(pinfo->valid); 6292 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci)) 6293 goto nla_put_failure; 6294 nlmsg_end(skb, nlh); 6295 return 0; 6296 6297 nla_put_failure: 6298 nlmsg_cancel(skb, nlh); 6299 return -EMSGSIZE; 6300 } 6301 6302 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 6303 struct prefix_info *pinfo) 6304 { 6305 struct sk_buff *skb; 6306 struct net *net = dev_net(idev->dev); 6307 int err = -ENOBUFS; 6308 6309 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC); 6310 if (!skb) 6311 goto errout; 6312 6313 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0); 6314 if (err < 0) { 6315 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */ 6316 WARN_ON(err == -EMSGSIZE); 6317 kfree_skb(skb); 6318 goto errout; 6319 } 6320 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC); 6321 return; 6322 errout: 6323 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err); 6324 } 6325 6326 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6327 { 6328 struct net *net = dev_net(ifp->idev->dev); 6329 6330 if (event) 6331 ASSERT_RTNL(); 6332 6333 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp); 6334 6335 switch (event) { 6336 case RTM_NEWADDR: 6337 /* 6338 * If the address was optimistic we inserted the route at the 6339 * start of our DAD process, so we don't need to do it again. 6340 * If the device was taken down in the middle of the DAD 6341 * cycle there is a race where we could get here without a 6342 * host route, so nothing to insert. That will be fixed when 6343 * the device is brought up. 6344 */ 6345 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) { 6346 ip6_ins_rt(net, ifp->rt); 6347 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) { 6348 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n", 6349 &ifp->addr, ifp->idev->dev->name); 6350 } 6351 6352 if (ifp->idev->cnf.forwarding) 6353 addrconf_join_anycast(ifp); 6354 if (!ipv6_addr_any(&ifp->peer_addr)) 6355 addrconf_prefix_route(&ifp->peer_addr, 128, 6356 ifp->rt_priority, ifp->idev->dev, 6357 0, 0, GFP_ATOMIC); 6358 break; 6359 case RTM_DELADDR: 6360 if (ifp->idev->cnf.forwarding) 6361 addrconf_leave_anycast(ifp); 6362 addrconf_leave_solict(ifp->idev, &ifp->addr); 6363 if (!ipv6_addr_any(&ifp->peer_addr)) { 6364 struct fib6_info *rt; 6365 6366 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128, 6367 ifp->idev->dev, 0, 0, 6368 false); 6369 if (rt) 6370 ip6_del_rt(net, rt, false); 6371 } 6372 if (ifp->rt) { 6373 ip6_del_rt(net, ifp->rt, false); 6374 ifp->rt = NULL; 6375 } 6376 rt_genid_bump_ipv6(net); 6377 break; 6378 } 6379 atomic_inc(&net->ipv6.dev_addr_genid); 6380 } 6381 6382 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6383 { 6384 if (likely(ifp->idev->dead == 0)) 6385 __ipv6_ifa_notify(event, ifp); 6386 } 6387 6388 #ifdef CONFIG_SYSCTL 6389 6390 static int addrconf_sysctl_forward(const struct ctl_table *ctl, int write, 6391 void *buffer, size_t *lenp, loff_t *ppos) 6392 { 6393 struct ctl_table lctl; 6394 int *valp = ctl->data; 6395 int val = *valp; 6396 int ret; 6397 6398 /* 6399 * ctl->data points to idev->cnf.forwarding, we should 6400 * not modify it until we get the rtnl lock. 6401 */ 6402 lctl = *ctl; 6403 lctl.data = &val; 6404 6405 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6406 if (ret) 6407 return ret; 6408 6409 if (write) 6410 ret = addrconf_fixup_forwarding(ctl, valp, val); 6411 return ret; 6412 } 6413 6414 static int addrconf_sysctl_mtu(const struct ctl_table *ctl, int write, 6415 void *buffer, size_t *lenp, loff_t *ppos) 6416 { 6417 struct inet6_dev *idev = ctl->extra1; 6418 int min_mtu = IPV6_MIN_MTU; 6419 struct ctl_table lctl; 6420 6421 lctl = *ctl; 6422 lctl.extra1 = &min_mtu; 6423 lctl.extra2 = idev ? &idev->dev->mtu : NULL; 6424 6425 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos); 6426 } 6427 6428 static void dev_disable_change(struct inet6_dev *idev) 6429 { 6430 struct netdev_notifier_info info; 6431 6432 if (!idev || !idev->dev) 6433 return; 6434 6435 netdev_notifier_info_init(&info, idev->dev); 6436 if (idev->cnf.disable_ipv6) 6437 addrconf_notify(NULL, NETDEV_DOWN, &info); 6438 else 6439 addrconf_notify(NULL, NETDEV_UP, &info); 6440 } 6441 6442 static void addrconf_disable_change(struct net *net, __s32 newf) 6443 { 6444 struct net_device *dev; 6445 struct inet6_dev *idev; 6446 6447 for_each_netdev(net, dev) { 6448 idev = __in6_dev_get_rtnl_net(dev); 6449 if (idev) { 6450 int changed = (!idev->cnf.disable_ipv6) ^ (!newf); 6451 6452 WRITE_ONCE(idev->cnf.disable_ipv6, newf); 6453 if (changed) 6454 dev_disable_change(idev); 6455 } 6456 } 6457 } 6458 6459 static int addrconf_disable_ipv6(const struct ctl_table *table, int *p, int newf) 6460 { 6461 struct net *net = (struct net *)table->extra2; 6462 int old; 6463 6464 if (p == &net->ipv6.devconf_dflt->disable_ipv6) { 6465 WRITE_ONCE(*p, newf); 6466 return 0; 6467 } 6468 6469 if (!rtnl_net_trylock(net)) 6470 return restart_syscall(); 6471 6472 old = *p; 6473 WRITE_ONCE(*p, newf); 6474 6475 if (p == &net->ipv6.devconf_all->disable_ipv6) { 6476 WRITE_ONCE(net->ipv6.devconf_dflt->disable_ipv6, newf); 6477 addrconf_disable_change(net, newf); 6478 } else if ((!newf) ^ (!old)) { 6479 dev_disable_change((struct inet6_dev *)table->extra1); 6480 } 6481 6482 rtnl_net_unlock(net); 6483 return 0; 6484 } 6485 6486 static int addrconf_sysctl_disable(const struct ctl_table *ctl, int write, 6487 void *buffer, size_t *lenp, loff_t *ppos) 6488 { 6489 struct ctl_table lctl; 6490 int *valp = ctl->data; 6491 int val = *valp; 6492 int ret; 6493 6494 /* 6495 * ctl->data points to idev->cnf.disable_ipv6, we should 6496 * not modify it until we get the rtnl lock. 6497 */ 6498 lctl = *ctl; 6499 lctl.data = &val; 6500 6501 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6502 if (ret) 6503 return ret; 6504 6505 if (write) 6506 ret = addrconf_disable_ipv6(ctl, valp, val); 6507 return ret; 6508 } 6509 6510 static int addrconf_sysctl_proxy_ndp(const struct ctl_table *ctl, int write, 6511 void *buffer, size_t *lenp, loff_t *ppos) 6512 { 6513 struct net *net = ctl->extra2; 6514 int *valp = ctl->data; 6515 int old, new; 6516 int ret; 6517 6518 if (write && !rtnl_net_trylock(net)) 6519 return restart_syscall(); 6520 6521 old = *valp; 6522 ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 6523 new = *valp; 6524 6525 if (write && old != new) { 6526 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) { 6527 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6528 NETCONFA_PROXY_NEIGH, 6529 NETCONFA_IFINDEX_DEFAULT, 6530 net->ipv6.devconf_dflt); 6531 } else if (valp == &net->ipv6.devconf_all->proxy_ndp) { 6532 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6533 NETCONFA_PROXY_NEIGH, 6534 NETCONFA_IFINDEX_ALL, 6535 net->ipv6.devconf_all); 6536 } else { 6537 struct inet6_dev *idev = ctl->extra1; 6538 6539 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6540 NETCONFA_PROXY_NEIGH, 6541 idev->dev->ifindex, 6542 &idev->cnf); 6543 } 6544 } 6545 if (write) 6546 rtnl_net_unlock(net); 6547 6548 return ret; 6549 } 6550 6551 static int addrconf_sysctl_addr_gen_mode(const struct ctl_table *ctl, int write, 6552 void *buffer, size_t *lenp, 6553 loff_t *ppos) 6554 { 6555 int ret = 0; 6556 u32 new_val; 6557 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1; 6558 struct net *net = (struct net *)ctl->extra2; 6559 struct ctl_table tmp = { 6560 .data = &new_val, 6561 .maxlen = sizeof(new_val), 6562 .mode = ctl->mode, 6563 }; 6564 6565 if (!rtnl_net_trylock(net)) 6566 return restart_syscall(); 6567 6568 new_val = *((u32 *)ctl->data); 6569 6570 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos); 6571 if (ret != 0) 6572 goto out; 6573 6574 if (write) { 6575 if (check_addr_gen_mode(new_val) < 0) { 6576 ret = -EINVAL; 6577 goto out; 6578 } 6579 6580 if (idev) { 6581 if (check_stable_privacy(idev, net, new_val) < 0) { 6582 ret = -EINVAL; 6583 goto out; 6584 } 6585 6586 if (idev->cnf.addr_gen_mode != new_val) { 6587 WRITE_ONCE(idev->cnf.addr_gen_mode, new_val); 6588 netdev_lock_ops(idev->dev); 6589 addrconf_init_auto_addrs(idev->dev); 6590 netdev_unlock_ops(idev->dev); 6591 } 6592 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) { 6593 struct net_device *dev; 6594 6595 WRITE_ONCE(net->ipv6.devconf_dflt->addr_gen_mode, new_val); 6596 for_each_netdev(net, dev) { 6597 idev = __in6_dev_get_rtnl_net(dev); 6598 if (idev && 6599 idev->cnf.addr_gen_mode != new_val) { 6600 WRITE_ONCE(idev->cnf.addr_gen_mode, 6601 new_val); 6602 netdev_lock_ops(idev->dev); 6603 addrconf_init_auto_addrs(idev->dev); 6604 netdev_unlock_ops(idev->dev); 6605 } 6606 } 6607 } 6608 6609 WRITE_ONCE(*((u32 *)ctl->data), new_val); 6610 } 6611 6612 out: 6613 rtnl_net_unlock(net); 6614 6615 return ret; 6616 } 6617 6618 static int addrconf_sysctl_stable_secret(const struct ctl_table *ctl, int write, 6619 void *buffer, size_t *lenp, 6620 loff_t *ppos) 6621 { 6622 int err; 6623 struct in6_addr addr; 6624 char str[IPV6_MAX_STRLEN]; 6625 struct ctl_table lctl = *ctl; 6626 struct net *net = ctl->extra2; 6627 struct ipv6_stable_secret *secret = ctl->data; 6628 6629 if (&net->ipv6.devconf_all->stable_secret == ctl->data) 6630 return -EIO; 6631 6632 lctl.maxlen = IPV6_MAX_STRLEN; 6633 lctl.data = str; 6634 6635 if (!rtnl_net_trylock(net)) 6636 return restart_syscall(); 6637 6638 if (!write && !secret->initialized) { 6639 err = -EIO; 6640 goto out; 6641 } 6642 6643 err = snprintf(str, sizeof(str), "%pI6", &secret->secret); 6644 if (err >= sizeof(str)) { 6645 err = -EIO; 6646 goto out; 6647 } 6648 6649 err = proc_dostring(&lctl, write, buffer, lenp, ppos); 6650 if (err || !write) 6651 goto out; 6652 6653 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) { 6654 err = -EIO; 6655 goto out; 6656 } 6657 6658 secret->initialized = true; 6659 secret->secret = addr; 6660 6661 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) { 6662 struct net_device *dev; 6663 6664 for_each_netdev(net, dev) { 6665 struct inet6_dev *idev = __in6_dev_get_rtnl_net(dev); 6666 6667 if (idev) { 6668 WRITE_ONCE(idev->cnf.addr_gen_mode, 6669 IN6_ADDR_GEN_MODE_STABLE_PRIVACY); 6670 } 6671 } 6672 } else { 6673 struct inet6_dev *idev = ctl->extra1; 6674 6675 WRITE_ONCE(idev->cnf.addr_gen_mode, 6676 IN6_ADDR_GEN_MODE_STABLE_PRIVACY); 6677 } 6678 6679 out: 6680 rtnl_net_unlock(net); 6681 6682 return err; 6683 } 6684 6685 static 6686 int addrconf_sysctl_ignore_routes_with_linkdown(const struct ctl_table *ctl, 6687 int write, void *buffer, 6688 size_t *lenp, 6689 loff_t *ppos) 6690 { 6691 struct ctl_table lctl; 6692 int *valp = ctl->data; 6693 int val = *valp; 6694 int ret; 6695 6696 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown 6697 * we should not modify it until we get the rtnl lock. 6698 */ 6699 lctl = *ctl; 6700 lctl.data = &val; 6701 6702 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6703 if (ret) 6704 return ret; 6705 6706 if (write) 6707 ret = addrconf_fixup_linkdown(ctl, valp, val); 6708 return ret; 6709 } 6710 6711 static 6712 void addrconf_set_nopolicy(struct rt6_info *rt, int action) 6713 { 6714 if (rt) { 6715 if (action) 6716 rt->dst.flags |= DST_NOPOLICY; 6717 else 6718 rt->dst.flags &= ~DST_NOPOLICY; 6719 } 6720 } 6721 6722 static 6723 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val) 6724 { 6725 struct inet6_ifaddr *ifa; 6726 6727 read_lock_bh(&idev->lock); 6728 list_for_each_entry(ifa, &idev->addr_list, if_list) { 6729 spin_lock(&ifa->lock); 6730 if (ifa->rt) { 6731 /* host routes only use builtin fib6_nh */ 6732 struct fib6_nh *nh = ifa->rt->fib6_nh; 6733 int cpu; 6734 6735 rcu_read_lock(); 6736 ifa->rt->dst_nopolicy = val ? true : false; 6737 if (nh->rt6i_pcpu) { 6738 for_each_possible_cpu(cpu) { 6739 struct rt6_info **rtp; 6740 6741 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu); 6742 addrconf_set_nopolicy(*rtp, val); 6743 } 6744 } 6745 rcu_read_unlock(); 6746 } 6747 spin_unlock(&ifa->lock); 6748 } 6749 read_unlock_bh(&idev->lock); 6750 } 6751 6752 static 6753 int addrconf_disable_policy(const struct ctl_table *ctl, int *valp, int val) 6754 { 6755 struct net *net = (struct net *)ctl->extra2; 6756 struct inet6_dev *idev; 6757 6758 if (valp == &net->ipv6.devconf_dflt->disable_policy) { 6759 WRITE_ONCE(*valp, val); 6760 return 0; 6761 } 6762 6763 if (!rtnl_net_trylock(net)) 6764 return restart_syscall(); 6765 6766 WRITE_ONCE(*valp, val); 6767 6768 if (valp == &net->ipv6.devconf_all->disable_policy) { 6769 struct net_device *dev; 6770 6771 for_each_netdev(net, dev) { 6772 idev = __in6_dev_get_rtnl_net(dev); 6773 if (idev) 6774 addrconf_disable_policy_idev(idev, val); 6775 } 6776 } else { 6777 idev = (struct inet6_dev *)ctl->extra1; 6778 addrconf_disable_policy_idev(idev, val); 6779 } 6780 6781 rtnl_net_unlock(net); 6782 return 0; 6783 } 6784 6785 static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write, 6786 void *buffer, size_t *lenp, loff_t *ppos) 6787 { 6788 struct ctl_table lctl; 6789 int *valp = ctl->data; 6790 int val = *valp; 6791 int ret; 6792 6793 lctl = *ctl; 6794 lctl.data = &val; 6795 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6796 if (ret) 6797 return ret; 6798 6799 if (write && (*valp != val)) 6800 ret = addrconf_disable_policy(ctl, valp, val); 6801 6802 return ret; 6803 } 6804 6805 static void addrconf_force_forward_change(struct net *net, __s32 newf) 6806 { 6807 struct net_device *dev; 6808 struct inet6_dev *idev; 6809 6810 for_each_netdev(net, dev) { 6811 idev = __in6_dev_get_rtnl_net(dev); 6812 if (idev) { 6813 int changed = (!idev->cnf.force_forwarding) ^ (!newf); 6814 6815 WRITE_ONCE(idev->cnf.force_forwarding, newf); 6816 if (changed) 6817 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 6818 NETCONFA_FORCE_FORWARDING, 6819 dev->ifindex, &idev->cnf); 6820 } 6821 } 6822 } 6823 6824 static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write, 6825 void *buffer, size_t *lenp, loff_t *ppos) 6826 { 6827 struct inet6_dev *idev = ctl->extra1; 6828 struct ctl_table tmp_ctl = *ctl; 6829 struct net *net = ctl->extra2; 6830 int *valp = ctl->data; 6831 int new_val = *valp; 6832 int old_val = *valp; 6833 int ret; 6834 6835 tmp_ctl.extra1 = SYSCTL_ZERO; 6836 tmp_ctl.extra2 = SYSCTL_ONE; 6837 tmp_ctl.data = &new_val; 6838 6839 ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos); 6840 6841 if (write && old_val != new_val) { 6842 if (!rtnl_net_trylock(net)) 6843 return restart_syscall(); 6844 6845 WRITE_ONCE(*valp, new_val); 6846 6847 if (valp == &net->ipv6.devconf_dflt->force_forwarding) { 6848 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6849 NETCONFA_FORCE_FORWARDING, 6850 NETCONFA_IFINDEX_DEFAULT, 6851 net->ipv6.devconf_dflt); 6852 } else if (valp == &net->ipv6.devconf_all->force_forwarding) { 6853 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6854 NETCONFA_FORCE_FORWARDING, 6855 NETCONFA_IFINDEX_ALL, 6856 net->ipv6.devconf_all); 6857 6858 addrconf_force_forward_change(net, new_val); 6859 } else { 6860 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6861 NETCONFA_FORCE_FORWARDING, 6862 idev->dev->ifindex, 6863 &idev->cnf); 6864 } 6865 rtnl_net_unlock(net); 6866 } 6867 6868 return ret; 6869 } 6870 6871 static int minus_one = -1; 6872 static const int two_five_five = 255; 6873 static u32 ioam6_if_id_max = U16_MAX; 6874 6875 static const struct ctl_table addrconf_sysctl[] = { 6876 { 6877 .procname = "forwarding", 6878 .data = &ipv6_devconf.forwarding, 6879 .maxlen = sizeof(int), 6880 .mode = 0644, 6881 .proc_handler = addrconf_sysctl_forward, 6882 }, 6883 { 6884 .procname = "hop_limit", 6885 .data = &ipv6_devconf.hop_limit, 6886 .maxlen = sizeof(int), 6887 .mode = 0644, 6888 .proc_handler = proc_dointvec_minmax, 6889 .extra1 = (void *)SYSCTL_ONE, 6890 .extra2 = (void *)&two_five_five, 6891 }, 6892 { 6893 .procname = "mtu", 6894 .data = &ipv6_devconf.mtu6, 6895 .maxlen = sizeof(int), 6896 .mode = 0644, 6897 .proc_handler = addrconf_sysctl_mtu, 6898 }, 6899 { 6900 .procname = "accept_ra", 6901 .data = &ipv6_devconf.accept_ra, 6902 .maxlen = sizeof(int), 6903 .mode = 0644, 6904 .proc_handler = proc_dointvec, 6905 }, 6906 { 6907 .procname = "accept_redirects", 6908 .data = &ipv6_devconf.accept_redirects, 6909 .maxlen = sizeof(int), 6910 .mode = 0644, 6911 .proc_handler = proc_dointvec, 6912 }, 6913 { 6914 .procname = "autoconf", 6915 .data = &ipv6_devconf.autoconf, 6916 .maxlen = sizeof(int), 6917 .mode = 0644, 6918 .proc_handler = proc_dointvec, 6919 }, 6920 { 6921 .procname = "dad_transmits", 6922 .data = &ipv6_devconf.dad_transmits, 6923 .maxlen = sizeof(int), 6924 .mode = 0644, 6925 .proc_handler = proc_dointvec, 6926 }, 6927 { 6928 .procname = "router_solicitations", 6929 .data = &ipv6_devconf.rtr_solicits, 6930 .maxlen = sizeof(int), 6931 .mode = 0644, 6932 .proc_handler = proc_dointvec_minmax, 6933 .extra1 = &minus_one, 6934 }, 6935 { 6936 .procname = "router_solicitation_interval", 6937 .data = &ipv6_devconf.rtr_solicit_interval, 6938 .maxlen = sizeof(int), 6939 .mode = 0644, 6940 .proc_handler = proc_dointvec_jiffies, 6941 }, 6942 { 6943 .procname = "router_solicitation_max_interval", 6944 .data = &ipv6_devconf.rtr_solicit_max_interval, 6945 .maxlen = sizeof(int), 6946 .mode = 0644, 6947 .proc_handler = proc_dointvec_jiffies, 6948 }, 6949 { 6950 .procname = "router_solicitation_delay", 6951 .data = &ipv6_devconf.rtr_solicit_delay, 6952 .maxlen = sizeof(int), 6953 .mode = 0644, 6954 .proc_handler = proc_dointvec_jiffies, 6955 }, 6956 { 6957 .procname = "force_mld_version", 6958 .data = &ipv6_devconf.force_mld_version, 6959 .maxlen = sizeof(int), 6960 .mode = 0644, 6961 .proc_handler = proc_dointvec, 6962 }, 6963 { 6964 .procname = "mldv1_unsolicited_report_interval", 6965 .data = 6966 &ipv6_devconf.mldv1_unsolicited_report_interval, 6967 .maxlen = sizeof(int), 6968 .mode = 0644, 6969 .proc_handler = proc_dointvec_ms_jiffies, 6970 }, 6971 { 6972 .procname = "mldv2_unsolicited_report_interval", 6973 .data = 6974 &ipv6_devconf.mldv2_unsolicited_report_interval, 6975 .maxlen = sizeof(int), 6976 .mode = 0644, 6977 .proc_handler = proc_dointvec_ms_jiffies, 6978 }, 6979 { 6980 .procname = "use_tempaddr", 6981 .data = &ipv6_devconf.use_tempaddr, 6982 .maxlen = sizeof(int), 6983 .mode = 0644, 6984 .proc_handler = proc_dointvec, 6985 }, 6986 { 6987 .procname = "temp_valid_lft", 6988 .data = &ipv6_devconf.temp_valid_lft, 6989 .maxlen = sizeof(int), 6990 .mode = 0644, 6991 .proc_handler = proc_dointvec, 6992 }, 6993 { 6994 .procname = "temp_prefered_lft", 6995 .data = &ipv6_devconf.temp_prefered_lft, 6996 .maxlen = sizeof(int), 6997 .mode = 0644, 6998 .proc_handler = proc_dointvec, 6999 }, 7000 { 7001 .procname = "regen_min_advance", 7002 .data = &ipv6_devconf.regen_min_advance, 7003 .maxlen = sizeof(int), 7004 .mode = 0644, 7005 .proc_handler = proc_dointvec, 7006 }, 7007 { 7008 .procname = "regen_max_retry", 7009 .data = &ipv6_devconf.regen_max_retry, 7010 .maxlen = sizeof(int), 7011 .mode = 0644, 7012 .proc_handler = proc_dointvec, 7013 }, 7014 { 7015 .procname = "max_desync_factor", 7016 .data = &ipv6_devconf.max_desync_factor, 7017 .maxlen = sizeof(int), 7018 .mode = 0644, 7019 .proc_handler = proc_dointvec, 7020 }, 7021 { 7022 .procname = "max_addresses", 7023 .data = &ipv6_devconf.max_addresses, 7024 .maxlen = sizeof(int), 7025 .mode = 0644, 7026 .proc_handler = proc_dointvec, 7027 }, 7028 { 7029 .procname = "accept_ra_defrtr", 7030 .data = &ipv6_devconf.accept_ra_defrtr, 7031 .maxlen = sizeof(int), 7032 .mode = 0644, 7033 .proc_handler = proc_dointvec, 7034 }, 7035 { 7036 .procname = "ra_defrtr_metric", 7037 .data = &ipv6_devconf.ra_defrtr_metric, 7038 .maxlen = sizeof(u32), 7039 .mode = 0644, 7040 .proc_handler = proc_douintvec_minmax, 7041 .extra1 = (void *)SYSCTL_ONE, 7042 }, 7043 { 7044 .procname = "accept_ra_min_hop_limit", 7045 .data = &ipv6_devconf.accept_ra_min_hop_limit, 7046 .maxlen = sizeof(int), 7047 .mode = 0644, 7048 .proc_handler = proc_dointvec, 7049 }, 7050 { 7051 .procname = "accept_ra_min_lft", 7052 .data = &ipv6_devconf.accept_ra_min_lft, 7053 .maxlen = sizeof(int), 7054 .mode = 0644, 7055 .proc_handler = proc_dointvec, 7056 }, 7057 { 7058 .procname = "accept_ra_pinfo", 7059 .data = &ipv6_devconf.accept_ra_pinfo, 7060 .maxlen = sizeof(int), 7061 .mode = 0644, 7062 .proc_handler = proc_dointvec, 7063 }, 7064 { 7065 .procname = "ra_honor_pio_life", 7066 .data = &ipv6_devconf.ra_honor_pio_life, 7067 .maxlen = sizeof(u8), 7068 .mode = 0644, 7069 .proc_handler = proc_dou8vec_minmax, 7070 .extra1 = SYSCTL_ZERO, 7071 .extra2 = SYSCTL_ONE, 7072 }, 7073 { 7074 .procname = "ra_honor_pio_pflag", 7075 .data = &ipv6_devconf.ra_honor_pio_pflag, 7076 .maxlen = sizeof(u8), 7077 .mode = 0644, 7078 .proc_handler = proc_dou8vec_minmax, 7079 .extra1 = SYSCTL_ZERO, 7080 .extra2 = SYSCTL_ONE, 7081 }, 7082 #ifdef CONFIG_IPV6_ROUTER_PREF 7083 { 7084 .procname = "accept_ra_rtr_pref", 7085 .data = &ipv6_devconf.accept_ra_rtr_pref, 7086 .maxlen = sizeof(int), 7087 .mode = 0644, 7088 .proc_handler = proc_dointvec, 7089 }, 7090 { 7091 .procname = "router_probe_interval", 7092 .data = &ipv6_devconf.rtr_probe_interval, 7093 .maxlen = sizeof(int), 7094 .mode = 0644, 7095 .proc_handler = proc_dointvec_jiffies, 7096 }, 7097 #ifdef CONFIG_IPV6_ROUTE_INFO 7098 { 7099 .procname = "accept_ra_rt_info_min_plen", 7100 .data = &ipv6_devconf.accept_ra_rt_info_min_plen, 7101 .maxlen = sizeof(int), 7102 .mode = 0644, 7103 .proc_handler = proc_dointvec, 7104 }, 7105 { 7106 .procname = "accept_ra_rt_info_max_plen", 7107 .data = &ipv6_devconf.accept_ra_rt_info_max_plen, 7108 .maxlen = sizeof(int), 7109 .mode = 0644, 7110 .proc_handler = proc_dointvec, 7111 }, 7112 #endif 7113 #endif 7114 { 7115 .procname = "proxy_ndp", 7116 .data = &ipv6_devconf.proxy_ndp, 7117 .maxlen = sizeof(int), 7118 .mode = 0644, 7119 .proc_handler = addrconf_sysctl_proxy_ndp, 7120 }, 7121 { 7122 .procname = "accept_source_route", 7123 .data = &ipv6_devconf.accept_source_route, 7124 .maxlen = sizeof(int), 7125 .mode = 0644, 7126 .proc_handler = proc_dointvec, 7127 }, 7128 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 7129 { 7130 .procname = "optimistic_dad", 7131 .data = &ipv6_devconf.optimistic_dad, 7132 .maxlen = sizeof(int), 7133 .mode = 0644, 7134 .proc_handler = proc_dointvec, 7135 }, 7136 { 7137 .procname = "use_optimistic", 7138 .data = &ipv6_devconf.use_optimistic, 7139 .maxlen = sizeof(int), 7140 .mode = 0644, 7141 .proc_handler = proc_dointvec, 7142 }, 7143 #endif 7144 #ifdef CONFIG_IPV6_MROUTE 7145 { 7146 .procname = "mc_forwarding", 7147 .data = &ipv6_devconf.mc_forwarding, 7148 .maxlen = sizeof(int), 7149 .mode = 0444, 7150 .proc_handler = proc_dointvec, 7151 }, 7152 #endif 7153 { 7154 .procname = "disable_ipv6", 7155 .data = &ipv6_devconf.disable_ipv6, 7156 .maxlen = sizeof(int), 7157 .mode = 0644, 7158 .proc_handler = addrconf_sysctl_disable, 7159 }, 7160 { 7161 .procname = "accept_dad", 7162 .data = &ipv6_devconf.accept_dad, 7163 .maxlen = sizeof(int), 7164 .mode = 0644, 7165 .proc_handler = proc_dointvec, 7166 }, 7167 { 7168 .procname = "force_tllao", 7169 .data = &ipv6_devconf.force_tllao, 7170 .maxlen = sizeof(int), 7171 .mode = 0644, 7172 .proc_handler = proc_dointvec 7173 }, 7174 { 7175 .procname = "ndisc_notify", 7176 .data = &ipv6_devconf.ndisc_notify, 7177 .maxlen = sizeof(int), 7178 .mode = 0644, 7179 .proc_handler = proc_dointvec 7180 }, 7181 { 7182 .procname = "suppress_frag_ndisc", 7183 .data = &ipv6_devconf.suppress_frag_ndisc, 7184 .maxlen = sizeof(int), 7185 .mode = 0644, 7186 .proc_handler = proc_dointvec 7187 }, 7188 { 7189 .procname = "accept_ra_from_local", 7190 .data = &ipv6_devconf.accept_ra_from_local, 7191 .maxlen = sizeof(int), 7192 .mode = 0644, 7193 .proc_handler = proc_dointvec, 7194 }, 7195 { 7196 .procname = "accept_ra_mtu", 7197 .data = &ipv6_devconf.accept_ra_mtu, 7198 .maxlen = sizeof(int), 7199 .mode = 0644, 7200 .proc_handler = proc_dointvec, 7201 }, 7202 { 7203 .procname = "stable_secret", 7204 .data = &ipv6_devconf.stable_secret, 7205 .maxlen = IPV6_MAX_STRLEN, 7206 .mode = 0600, 7207 .proc_handler = addrconf_sysctl_stable_secret, 7208 }, 7209 { 7210 .procname = "use_oif_addrs_only", 7211 .data = &ipv6_devconf.use_oif_addrs_only, 7212 .maxlen = sizeof(int), 7213 .mode = 0644, 7214 .proc_handler = proc_dointvec, 7215 }, 7216 { 7217 .procname = "ignore_routes_with_linkdown", 7218 .data = &ipv6_devconf.ignore_routes_with_linkdown, 7219 .maxlen = sizeof(int), 7220 .mode = 0644, 7221 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown, 7222 }, 7223 { 7224 .procname = "drop_unicast_in_l2_multicast", 7225 .data = &ipv6_devconf.drop_unicast_in_l2_multicast, 7226 .maxlen = sizeof(int), 7227 .mode = 0644, 7228 .proc_handler = proc_dointvec, 7229 }, 7230 { 7231 .procname = "drop_unsolicited_na", 7232 .data = &ipv6_devconf.drop_unsolicited_na, 7233 .maxlen = sizeof(int), 7234 .mode = 0644, 7235 .proc_handler = proc_dointvec, 7236 }, 7237 { 7238 .procname = "keep_addr_on_down", 7239 .data = &ipv6_devconf.keep_addr_on_down, 7240 .maxlen = sizeof(int), 7241 .mode = 0644, 7242 .proc_handler = proc_dointvec, 7243 7244 }, 7245 { 7246 .procname = "seg6_enabled", 7247 .data = &ipv6_devconf.seg6_enabled, 7248 .maxlen = sizeof(int), 7249 .mode = 0644, 7250 .proc_handler = proc_dointvec, 7251 }, 7252 #ifdef CONFIG_IPV6_SEG6_HMAC 7253 { 7254 .procname = "seg6_require_hmac", 7255 .data = &ipv6_devconf.seg6_require_hmac, 7256 .maxlen = sizeof(int), 7257 .mode = 0644, 7258 .proc_handler = proc_dointvec, 7259 }, 7260 #endif 7261 { 7262 .procname = "enhanced_dad", 7263 .data = &ipv6_devconf.enhanced_dad, 7264 .maxlen = sizeof(int), 7265 .mode = 0644, 7266 .proc_handler = proc_dointvec, 7267 }, 7268 { 7269 .procname = "addr_gen_mode", 7270 .data = &ipv6_devconf.addr_gen_mode, 7271 .maxlen = sizeof(int), 7272 .mode = 0644, 7273 .proc_handler = addrconf_sysctl_addr_gen_mode, 7274 }, 7275 { 7276 .procname = "disable_policy", 7277 .data = &ipv6_devconf.disable_policy, 7278 .maxlen = sizeof(int), 7279 .mode = 0644, 7280 .proc_handler = addrconf_sysctl_disable_policy, 7281 }, 7282 { 7283 .procname = "ndisc_tclass", 7284 .data = &ipv6_devconf.ndisc_tclass, 7285 .maxlen = sizeof(int), 7286 .mode = 0644, 7287 .proc_handler = proc_dointvec_minmax, 7288 .extra1 = (void *)SYSCTL_ZERO, 7289 .extra2 = (void *)&two_five_five, 7290 }, 7291 { 7292 .procname = "rpl_seg_enabled", 7293 .data = &ipv6_devconf.rpl_seg_enabled, 7294 .maxlen = sizeof(int), 7295 .mode = 0644, 7296 .proc_handler = proc_dointvec_minmax, 7297 .extra1 = SYSCTL_ZERO, 7298 .extra2 = SYSCTL_ONE, 7299 }, 7300 { 7301 .procname = "ioam6_enabled", 7302 .data = &ipv6_devconf.ioam6_enabled, 7303 .maxlen = sizeof(u8), 7304 .mode = 0644, 7305 .proc_handler = proc_dou8vec_minmax, 7306 .extra1 = (void *)SYSCTL_ZERO, 7307 .extra2 = (void *)SYSCTL_ONE, 7308 }, 7309 { 7310 .procname = "ioam6_id", 7311 .data = &ipv6_devconf.ioam6_id, 7312 .maxlen = sizeof(u32), 7313 .mode = 0644, 7314 .proc_handler = proc_douintvec_minmax, 7315 .extra1 = (void *)SYSCTL_ZERO, 7316 .extra2 = (void *)&ioam6_if_id_max, 7317 }, 7318 { 7319 .procname = "ioam6_id_wide", 7320 .data = &ipv6_devconf.ioam6_id_wide, 7321 .maxlen = sizeof(u32), 7322 .mode = 0644, 7323 .proc_handler = proc_douintvec, 7324 }, 7325 { 7326 .procname = "ndisc_evict_nocarrier", 7327 .data = &ipv6_devconf.ndisc_evict_nocarrier, 7328 .maxlen = sizeof(u8), 7329 .mode = 0644, 7330 .proc_handler = proc_dou8vec_minmax, 7331 .extra1 = (void *)SYSCTL_ZERO, 7332 .extra2 = (void *)SYSCTL_ONE, 7333 }, 7334 { 7335 .procname = "accept_untracked_na", 7336 .data = &ipv6_devconf.accept_untracked_na, 7337 .maxlen = sizeof(int), 7338 .mode = 0644, 7339 .proc_handler = proc_dointvec_minmax, 7340 .extra1 = SYSCTL_ZERO, 7341 .extra2 = SYSCTL_TWO, 7342 }, 7343 { 7344 .procname = "force_forwarding", 7345 .data = &ipv6_devconf.force_forwarding, 7346 .maxlen = sizeof(int), 7347 .mode = 0644, 7348 .proc_handler = addrconf_sysctl_force_forwarding, 7349 }, 7350 }; 7351 7352 static int __addrconf_sysctl_register(struct net *net, char *dev_name, 7353 struct inet6_dev *idev, struct ipv6_devconf *p) 7354 { 7355 size_t table_size = ARRAY_SIZE(addrconf_sysctl); 7356 int i, ifindex; 7357 struct ctl_table *table; 7358 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ]; 7359 7360 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT); 7361 if (!table) 7362 goto out; 7363 7364 for (i = 0; i < table_size; i++) { 7365 table[i].data += (char *)p - (char *)&ipv6_devconf; 7366 /* If one of these is already set, then it is not safe to 7367 * overwrite either of them: this makes proc_dointvec_minmax 7368 * usable. 7369 */ 7370 if (!table[i].extra1 && !table[i].extra2) { 7371 table[i].extra1 = idev; /* embedded; no ref */ 7372 table[i].extra2 = net; 7373 } 7374 } 7375 7376 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name); 7377 7378 p->sysctl_header = register_net_sysctl_sz(net, path, table, 7379 table_size); 7380 if (!p->sysctl_header) 7381 goto free; 7382 7383 if (!strcmp(dev_name, "all")) 7384 ifindex = NETCONFA_IFINDEX_ALL; 7385 else if (!strcmp(dev_name, "default")) 7386 ifindex = NETCONFA_IFINDEX_DEFAULT; 7387 else 7388 ifindex = idev->dev->ifindex; 7389 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, 7390 ifindex, p); 7391 return 0; 7392 7393 free: 7394 kfree(table); 7395 out: 7396 return -ENOBUFS; 7397 } 7398 7399 static void __addrconf_sysctl_unregister(struct net *net, 7400 struct ipv6_devconf *p, int ifindex) 7401 { 7402 const struct ctl_table *table; 7403 7404 if (!p->sysctl_header) 7405 return; 7406 7407 table = p->sysctl_header->ctl_table_arg; 7408 unregister_net_sysctl_table(p->sysctl_header); 7409 p->sysctl_header = NULL; 7410 kfree(table); 7411 7412 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL); 7413 } 7414 7415 static int addrconf_sysctl_register(struct inet6_dev *idev) 7416 { 7417 int err; 7418 7419 if (!sysctl_dev_name_is_allowed(idev->dev->name)) 7420 return -EINVAL; 7421 7422 err = neigh_sysctl_register(idev->dev, idev->nd_parms, 7423 &ndisc_ifinfo_sysctl_change); 7424 if (err) 7425 return err; 7426 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name, 7427 idev, &idev->cnf); 7428 if (err) 7429 neigh_sysctl_unregister(idev->nd_parms); 7430 7431 return err; 7432 } 7433 7434 static void addrconf_sysctl_unregister(struct inet6_dev *idev) 7435 { 7436 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf, 7437 idev->dev->ifindex); 7438 neigh_sysctl_unregister(idev->nd_parms); 7439 } 7440 7441 7442 #endif 7443 7444 static int __net_init addrconf_init_net(struct net *net) 7445 { 7446 int err = -ENOMEM; 7447 struct ipv6_devconf *all, *dflt; 7448 7449 spin_lock_init(&net->ipv6.addrconf_hash_lock); 7450 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work); 7451 net->ipv6.inet6_addr_lst = kzalloc_objs(struct hlist_head, 7452 IN6_ADDR_HSIZE); 7453 if (!net->ipv6.inet6_addr_lst) 7454 goto err_alloc_addr; 7455 7456 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL); 7457 if (!all) 7458 goto err_alloc_all; 7459 7460 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL); 7461 if (!dflt) 7462 goto err_alloc_dflt; 7463 7464 if (!net_eq(net, &init_net)) { 7465 switch (net_inherit_devconf()) { 7466 case 1: /* copy from init_net */ 7467 memcpy(all, init_net.ipv6.devconf_all, 7468 sizeof(ipv6_devconf)); 7469 memcpy(dflt, init_net.ipv6.devconf_dflt, 7470 sizeof(ipv6_devconf_dflt)); 7471 break; 7472 case 3: /* copy from the current netns */ 7473 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all, 7474 sizeof(ipv6_devconf)); 7475 memcpy(dflt, 7476 current->nsproxy->net_ns->ipv6.devconf_dflt, 7477 sizeof(ipv6_devconf_dflt)); 7478 break; 7479 case 0: 7480 case 2: 7481 /* use compiled values */ 7482 break; 7483 } 7484 } 7485 7486 /* these will be inherited by all namespaces */ 7487 dflt->autoconf = ipv6_defaults.autoconf; 7488 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6; 7489 7490 dflt->stable_secret.initialized = false; 7491 all->stable_secret.initialized = false; 7492 7493 net->ipv6.devconf_all = all; 7494 net->ipv6.devconf_dflt = dflt; 7495 7496 #ifdef CONFIG_SYSCTL 7497 err = __addrconf_sysctl_register(net, "all", NULL, all); 7498 if (err < 0) 7499 goto err_reg_all; 7500 7501 err = __addrconf_sysctl_register(net, "default", NULL, dflt); 7502 if (err < 0) 7503 goto err_reg_dflt; 7504 #endif 7505 return 0; 7506 7507 #ifdef CONFIG_SYSCTL 7508 err_reg_dflt: 7509 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL); 7510 err_reg_all: 7511 kfree(dflt); 7512 net->ipv6.devconf_dflt = NULL; 7513 #endif 7514 err_alloc_dflt: 7515 kfree(all); 7516 net->ipv6.devconf_all = NULL; 7517 err_alloc_all: 7518 kfree(net->ipv6.inet6_addr_lst); 7519 err_alloc_addr: 7520 return err; 7521 } 7522 7523 static void __net_exit addrconf_exit_net(struct net *net) 7524 { 7525 int i; 7526 7527 #ifdef CONFIG_SYSCTL 7528 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt, 7529 NETCONFA_IFINDEX_DEFAULT); 7530 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all, 7531 NETCONFA_IFINDEX_ALL); 7532 #endif 7533 kfree(net->ipv6.devconf_dflt); 7534 net->ipv6.devconf_dflt = NULL; 7535 kfree(net->ipv6.devconf_all); 7536 net->ipv6.devconf_all = NULL; 7537 7538 cancel_delayed_work_sync(&net->ipv6.addr_chk_work); 7539 /* 7540 * Check hash table, then free it. 7541 */ 7542 for (i = 0; i < IN6_ADDR_HSIZE; i++) 7543 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i])); 7544 7545 kfree(net->ipv6.inet6_addr_lst); 7546 net->ipv6.inet6_addr_lst = NULL; 7547 } 7548 7549 static struct pernet_operations addrconf_ops = { 7550 .init = addrconf_init_net, 7551 .exit = addrconf_exit_net, 7552 }; 7553 7554 static struct rtnl_af_ops inet6_ops __read_mostly = { 7555 .family = AF_INET6, 7556 .fill_link_af = inet6_fill_link_af, 7557 .get_link_af_size = inet6_get_link_af_size, 7558 .validate_link_af = inet6_validate_link_af, 7559 .set_link_af = inet6_set_link_af, 7560 }; 7561 7562 static const struct rtnl_msg_handler addrconf_rtnl_msg_handlers[] __initconst_or_module = { 7563 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETLINK, 7564 .dumpit = inet6_dump_ifinfo, .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7565 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_NEWADDR, 7566 .doit = inet6_rtm_newaddr, .flags = RTNL_FLAG_DOIT_PERNET}, 7567 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_DELADDR, 7568 .doit = inet6_rtm_deladdr, .flags = RTNL_FLAG_DOIT_PERNET}, 7569 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETADDR, 7570 .doit = inet6_rtm_getaddr, .dumpit = inet6_dump_ifaddr, 7571 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 7572 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETMULTICAST, 7573 .dumpit = inet6_dump_ifmcaddr, 7574 .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7575 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETANYCAST, 7576 .dumpit = inet6_dump_ifacaddr, 7577 .flags = RTNL_FLAG_DUMP_UNLOCKED}, 7578 {.owner = THIS_MODULE, .protocol = PF_INET6, .msgtype = RTM_GETNETCONF, 7579 .doit = inet6_netconf_get_devconf, .dumpit = inet6_netconf_dump_devconf, 7580 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 7581 }; 7582 7583 /* 7584 * Init / cleanup code 7585 */ 7586 7587 int __init addrconf_init(void) 7588 { 7589 struct inet6_dev *idev; 7590 int err; 7591 7592 err = ipv6_addr_label_init(); 7593 if (err < 0) { 7594 pr_crit("%s: cannot initialize default policy table: %d\n", 7595 __func__, err); 7596 goto out; 7597 } 7598 7599 err = register_pernet_subsys(&addrconf_ops); 7600 if (err < 0) 7601 goto out_addrlabel; 7602 7603 /* All works using addrconf_wq need to lock rtnl. */ 7604 addrconf_wq = create_singlethread_workqueue("ipv6_addrconf"); 7605 if (!addrconf_wq) { 7606 err = -ENOMEM; 7607 goto out_nowq; 7608 } 7609 7610 rtnl_net_lock(&init_net); 7611 idev = ipv6_add_dev(blackhole_netdev); 7612 rtnl_net_unlock(&init_net); 7613 if (IS_ERR(idev)) { 7614 err = PTR_ERR(idev); 7615 goto errlo; 7616 } 7617 7618 ip6_route_init_special_entries(); 7619 7620 register_netdevice_notifier(&ipv6_dev_notf); 7621 7622 addrconf_verify(&init_net); 7623 7624 err = rtnl_af_register(&inet6_ops); 7625 if (err) 7626 goto erraf; 7627 7628 err = rtnl_register_many(addrconf_rtnl_msg_handlers); 7629 if (err) 7630 goto errout; 7631 7632 err = ipv6_addr_label_rtnl_register(); 7633 if (err < 0) 7634 goto errout; 7635 7636 return 0; 7637 errout: 7638 rtnl_unregister_all(PF_INET6); 7639 rtnl_af_unregister(&inet6_ops); 7640 erraf: 7641 unregister_netdevice_notifier(&ipv6_dev_notf); 7642 errlo: 7643 destroy_workqueue(addrconf_wq); 7644 out_nowq: 7645 unregister_pernet_subsys(&addrconf_ops); 7646 out_addrlabel: 7647 ipv6_addr_label_cleanup(); 7648 out: 7649 return err; 7650 } 7651 7652 void addrconf_cleanup(void) 7653 { 7654 struct net_device *dev; 7655 7656 unregister_netdevice_notifier(&ipv6_dev_notf); 7657 unregister_pernet_subsys(&addrconf_ops); 7658 ipv6_addr_label_cleanup(); 7659 7660 rtnl_af_unregister(&inet6_ops); 7661 7662 rtnl_net_lock(&init_net); 7663 7664 /* clean dev list */ 7665 for_each_netdev(&init_net, dev) { 7666 if (!__in6_dev_get_rtnl_net(dev)) 7667 continue; 7668 addrconf_ifdown(dev, true); 7669 } 7670 addrconf_ifdown(init_net.loopback_dev, true); 7671 7672 rtnl_net_unlock(&init_net); 7673 7674 destroy_workqueue(addrconf_wq); 7675 } 7676