1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * IPv6 Address [auto]configuration 4 * Linux INET6 implementation 5 * 6 * Authors: 7 * Pedro Roque <roque@di.fc.ul.pt> 8 * Alexey Kuznetsov <kuznet@ms2.inr.ac.ru> 9 */ 10 11 /* 12 * Changes: 13 * 14 * Janos Farkas : delete timer on ifdown 15 * <chexum@bankinf.banki.hu> 16 * Andi Kleen : kill double kfree on module 17 * unload. 18 * Maciej W. Rozycki : FDDI support 19 * sekiya@USAGI : Don't send too many RS 20 * packets. 21 * yoshfuji@USAGI : Fixed interval between DAD 22 * packets. 23 * YOSHIFUJI Hideaki @USAGI : improved accuracy of 24 * address validation timer. 25 * YOSHIFUJI Hideaki @USAGI : Privacy Extensions (RFC3041) 26 * support. 27 * Yuji SEKIYA @USAGI : Don't assign a same IPv6 28 * address on a same interface. 29 * YOSHIFUJI Hideaki @USAGI : ARCnet support 30 * YOSHIFUJI Hideaki @USAGI : convert /proc/net/if_inet6 to 31 * seq_file. 32 * YOSHIFUJI Hideaki @USAGI : improved source address 33 * selection; consider scope, 34 * status etc. 35 */ 36 37 #define pr_fmt(fmt) "IPv6: " fmt 38 39 #include <linux/errno.h> 40 #include <linux/types.h> 41 #include <linux/kernel.h> 42 #include <linux/sched/signal.h> 43 #include <linux/socket.h> 44 #include <linux/sockios.h> 45 #include <linux/net.h> 46 #include <linux/inet.h> 47 #include <linux/in6.h> 48 #include <linux/netdevice.h> 49 #include <linux/if_addr.h> 50 #include <linux/if_arp.h> 51 #include <linux/if_arcnet.h> 52 #include <linux/if_infiniband.h> 53 #include <linux/route.h> 54 #include <linux/inetdevice.h> 55 #include <linux/init.h> 56 #include <linux/slab.h> 57 #ifdef CONFIG_SYSCTL 58 #include <linux/sysctl.h> 59 #endif 60 #include <linux/capability.h> 61 #include <linux/delay.h> 62 #include <linux/notifier.h> 63 #include <linux/string.h> 64 #include <linux/hash.h> 65 66 #include <net/net_namespace.h> 67 #include <net/sock.h> 68 #include <net/snmp.h> 69 70 #include <net/6lowpan.h> 71 #include <net/firewire.h> 72 #include <net/ipv6.h> 73 #include <net/protocol.h> 74 #include <net/ndisc.h> 75 #include <net/ip6_route.h> 76 #include <net/addrconf.h> 77 #include <net/tcp.h> 78 #include <net/ip.h> 79 #include <net/netlink.h> 80 #include <net/pkt_sched.h> 81 #include <net/l3mdev.h> 82 #include <linux/if_tunnel.h> 83 #include <linux/rtnetlink.h> 84 #include <linux/netconf.h> 85 #include <linux/random.h> 86 #include <linux/uaccess.h> 87 #include <asm/unaligned.h> 88 89 #include <linux/proc_fs.h> 90 #include <linux/seq_file.h> 91 #include <linux/export.h> 92 #include <linux/ioam6.h> 93 94 #define INFINITY_LIFE_TIME 0xFFFFFFFF 95 96 #define IPV6_MAX_STRLEN \ 97 sizeof("ffff:ffff:ffff:ffff:ffff:ffff:255.255.255.255") 98 99 static inline u32 cstamp_delta(unsigned long cstamp) 100 { 101 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ; 102 } 103 104 static inline s32 rfc3315_s14_backoff_init(s32 irt) 105 { 106 /* multiply 'initial retransmission time' by 0.9 .. 1.1 */ 107 u64 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)irt; 108 do_div(tmp, 1000000); 109 return (s32)tmp; 110 } 111 112 static inline s32 rfc3315_s14_backoff_update(s32 rt, s32 mrt) 113 { 114 /* multiply 'retransmission timeout' by 1.9 .. 2.1 */ 115 u64 tmp = get_random_u32_inclusive(1900000, 2100000) * (u64)rt; 116 do_div(tmp, 1000000); 117 if ((s32)tmp > mrt) { 118 /* multiply 'maximum retransmission time' by 0.9 .. 1.1 */ 119 tmp = get_random_u32_inclusive(900000, 1100000) * (u64)mrt; 120 do_div(tmp, 1000000); 121 } 122 return (s32)tmp; 123 } 124 125 #ifdef CONFIG_SYSCTL 126 static int addrconf_sysctl_register(struct inet6_dev *idev); 127 static void addrconf_sysctl_unregister(struct inet6_dev *idev); 128 #else 129 static inline int addrconf_sysctl_register(struct inet6_dev *idev) 130 { 131 return 0; 132 } 133 134 static inline void addrconf_sysctl_unregister(struct inet6_dev *idev) 135 { 136 } 137 #endif 138 139 static void ipv6_gen_rnd_iid(struct in6_addr *addr); 140 141 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev); 142 static int ipv6_count_addresses(const struct inet6_dev *idev); 143 static int ipv6_generate_stable_address(struct in6_addr *addr, 144 u8 dad_count, 145 const struct inet6_dev *idev); 146 147 #define IN6_ADDR_HSIZE_SHIFT 8 148 #define IN6_ADDR_HSIZE (1 << IN6_ADDR_HSIZE_SHIFT) 149 150 static void addrconf_verify(struct net *net); 151 static void addrconf_verify_rtnl(struct net *net); 152 153 static struct workqueue_struct *addrconf_wq; 154 155 static void addrconf_join_anycast(struct inet6_ifaddr *ifp); 156 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp); 157 158 static void addrconf_type_change(struct net_device *dev, 159 unsigned long event); 160 static int addrconf_ifdown(struct net_device *dev, bool unregister); 161 162 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 163 int plen, 164 const struct net_device *dev, 165 u32 flags, u32 noflags, 166 bool no_gw); 167 168 static void addrconf_dad_start(struct inet6_ifaddr *ifp); 169 static void addrconf_dad_work(struct work_struct *w); 170 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 171 bool send_na); 172 static void addrconf_dad_run(struct inet6_dev *idev, bool restart); 173 static void addrconf_rs_timer(struct timer_list *t); 174 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 175 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifa); 176 177 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 178 struct prefix_info *pinfo); 179 180 static struct ipv6_devconf ipv6_devconf __read_mostly = { 181 .forwarding = 0, 182 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 183 .mtu6 = IPV6_MIN_MTU, 184 .accept_ra = 1, 185 .accept_redirects = 1, 186 .autoconf = 1, 187 .force_mld_version = 0, 188 .mldv1_unsolicited_report_interval = 10 * HZ, 189 .mldv2_unsolicited_report_interval = HZ, 190 .dad_transmits = 1, 191 .rtr_solicits = MAX_RTR_SOLICITATIONS, 192 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 193 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 194 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 195 .use_tempaddr = 0, 196 .temp_valid_lft = TEMP_VALID_LIFETIME, 197 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 198 .regen_max_retry = REGEN_MAX_RETRY, 199 .max_desync_factor = MAX_DESYNC_FACTOR, 200 .max_addresses = IPV6_MAX_ADDRESSES, 201 .accept_ra_defrtr = 1, 202 .ra_defrtr_metric = IP6_RT_PRIO_USER, 203 .accept_ra_from_local = 0, 204 .accept_ra_min_hop_limit= 1, 205 .accept_ra_min_lft = 0, 206 .accept_ra_pinfo = 1, 207 #ifdef CONFIG_IPV6_ROUTER_PREF 208 .accept_ra_rtr_pref = 1, 209 .rtr_probe_interval = 60 * HZ, 210 #ifdef CONFIG_IPV6_ROUTE_INFO 211 .accept_ra_rt_info_min_plen = 0, 212 .accept_ra_rt_info_max_plen = 0, 213 #endif 214 #endif 215 .proxy_ndp = 0, 216 .accept_source_route = 0, /* we do not accept RH0 by default. */ 217 .disable_ipv6 = 0, 218 .accept_dad = 0, 219 .suppress_frag_ndisc = 1, 220 .accept_ra_mtu = 1, 221 .stable_secret = { 222 .initialized = false, 223 }, 224 .use_oif_addrs_only = 0, 225 .ignore_routes_with_linkdown = 0, 226 .keep_addr_on_down = 0, 227 .seg6_enabled = 0, 228 #ifdef CONFIG_IPV6_SEG6_HMAC 229 .seg6_require_hmac = 0, 230 #endif 231 .enhanced_dad = 1, 232 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 233 .disable_policy = 0, 234 .rpl_seg_enabled = 0, 235 .ioam6_enabled = 0, 236 .ioam6_id = IOAM6_DEFAULT_IF_ID, 237 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 238 .ndisc_evict_nocarrier = 1, 239 .ra_honor_pio_life = 0, 240 }; 241 242 static struct ipv6_devconf ipv6_devconf_dflt __read_mostly = { 243 .forwarding = 0, 244 .hop_limit = IPV6_DEFAULT_HOPLIMIT, 245 .mtu6 = IPV6_MIN_MTU, 246 .accept_ra = 1, 247 .accept_redirects = 1, 248 .autoconf = 1, 249 .force_mld_version = 0, 250 .mldv1_unsolicited_report_interval = 10 * HZ, 251 .mldv2_unsolicited_report_interval = HZ, 252 .dad_transmits = 1, 253 .rtr_solicits = MAX_RTR_SOLICITATIONS, 254 .rtr_solicit_interval = RTR_SOLICITATION_INTERVAL, 255 .rtr_solicit_max_interval = RTR_SOLICITATION_MAX_INTERVAL, 256 .rtr_solicit_delay = MAX_RTR_SOLICITATION_DELAY, 257 .use_tempaddr = 0, 258 .temp_valid_lft = TEMP_VALID_LIFETIME, 259 .temp_prefered_lft = TEMP_PREFERRED_LIFETIME, 260 .regen_max_retry = REGEN_MAX_RETRY, 261 .max_desync_factor = MAX_DESYNC_FACTOR, 262 .max_addresses = IPV6_MAX_ADDRESSES, 263 .accept_ra_defrtr = 1, 264 .ra_defrtr_metric = IP6_RT_PRIO_USER, 265 .accept_ra_from_local = 0, 266 .accept_ra_min_hop_limit= 1, 267 .accept_ra_min_lft = 0, 268 .accept_ra_pinfo = 1, 269 #ifdef CONFIG_IPV6_ROUTER_PREF 270 .accept_ra_rtr_pref = 1, 271 .rtr_probe_interval = 60 * HZ, 272 #ifdef CONFIG_IPV6_ROUTE_INFO 273 .accept_ra_rt_info_min_plen = 0, 274 .accept_ra_rt_info_max_plen = 0, 275 #endif 276 #endif 277 .proxy_ndp = 0, 278 .accept_source_route = 0, /* we do not accept RH0 by default. */ 279 .disable_ipv6 = 0, 280 .accept_dad = 1, 281 .suppress_frag_ndisc = 1, 282 .accept_ra_mtu = 1, 283 .stable_secret = { 284 .initialized = false, 285 }, 286 .use_oif_addrs_only = 0, 287 .ignore_routes_with_linkdown = 0, 288 .keep_addr_on_down = 0, 289 .seg6_enabled = 0, 290 #ifdef CONFIG_IPV6_SEG6_HMAC 291 .seg6_require_hmac = 0, 292 #endif 293 .enhanced_dad = 1, 294 .addr_gen_mode = IN6_ADDR_GEN_MODE_EUI64, 295 .disable_policy = 0, 296 .rpl_seg_enabled = 0, 297 .ioam6_enabled = 0, 298 .ioam6_id = IOAM6_DEFAULT_IF_ID, 299 .ioam6_id_wide = IOAM6_DEFAULT_IF_ID_WIDE, 300 .ndisc_evict_nocarrier = 1, 301 .ra_honor_pio_life = 0, 302 }; 303 304 /* Check if link is ready: is it up and is a valid qdisc available */ 305 static inline bool addrconf_link_ready(const struct net_device *dev) 306 { 307 return netif_oper_up(dev) && !qdisc_tx_is_noop(dev); 308 } 309 310 static void addrconf_del_rs_timer(struct inet6_dev *idev) 311 { 312 if (del_timer(&idev->rs_timer)) 313 __in6_dev_put(idev); 314 } 315 316 static void addrconf_del_dad_work(struct inet6_ifaddr *ifp) 317 { 318 if (cancel_delayed_work(&ifp->dad_work)) 319 __in6_ifa_put(ifp); 320 } 321 322 static void addrconf_mod_rs_timer(struct inet6_dev *idev, 323 unsigned long when) 324 { 325 if (!mod_timer(&idev->rs_timer, jiffies + when)) 326 in6_dev_hold(idev); 327 } 328 329 static void addrconf_mod_dad_work(struct inet6_ifaddr *ifp, 330 unsigned long delay) 331 { 332 in6_ifa_hold(ifp); 333 if (mod_delayed_work(addrconf_wq, &ifp->dad_work, delay)) 334 in6_ifa_put(ifp); 335 } 336 337 static int snmp6_alloc_dev(struct inet6_dev *idev) 338 { 339 int i; 340 341 idev->stats.ipv6 = alloc_percpu_gfp(struct ipstats_mib, GFP_KERNEL_ACCOUNT); 342 if (!idev->stats.ipv6) 343 goto err_ip; 344 345 for_each_possible_cpu(i) { 346 struct ipstats_mib *addrconf_stats; 347 addrconf_stats = per_cpu_ptr(idev->stats.ipv6, i); 348 u64_stats_init(&addrconf_stats->syncp); 349 } 350 351 352 idev->stats.icmpv6dev = kzalloc(sizeof(struct icmpv6_mib_device), 353 GFP_KERNEL); 354 if (!idev->stats.icmpv6dev) 355 goto err_icmp; 356 idev->stats.icmpv6msgdev = kzalloc(sizeof(struct icmpv6msg_mib_device), 357 GFP_KERNEL_ACCOUNT); 358 if (!idev->stats.icmpv6msgdev) 359 goto err_icmpmsg; 360 361 return 0; 362 363 err_icmpmsg: 364 kfree(idev->stats.icmpv6dev); 365 err_icmp: 366 free_percpu(idev->stats.ipv6); 367 err_ip: 368 return -ENOMEM; 369 } 370 371 static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 372 { 373 struct inet6_dev *ndev; 374 int err = -ENOMEM; 375 376 ASSERT_RTNL(); 377 378 if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev) 379 return ERR_PTR(-EINVAL); 380 381 ndev = kzalloc(sizeof(*ndev), GFP_KERNEL_ACCOUNT); 382 if (!ndev) 383 return ERR_PTR(err); 384 385 rwlock_init(&ndev->lock); 386 ndev->dev = dev; 387 INIT_LIST_HEAD(&ndev->addr_list); 388 timer_setup(&ndev->rs_timer, addrconf_rs_timer, 0); 389 memcpy(&ndev->cnf, dev_net(dev)->ipv6.devconf_dflt, sizeof(ndev->cnf)); 390 391 if (ndev->cnf.stable_secret.initialized) 392 ndev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY; 393 394 ndev->cnf.mtu6 = dev->mtu; 395 ndev->ra_mtu = 0; 396 ndev->nd_parms = neigh_parms_alloc(dev, &nd_tbl); 397 if (!ndev->nd_parms) { 398 kfree(ndev); 399 return ERR_PTR(err); 400 } 401 if (ndev->cnf.forwarding) 402 dev_disable_lro(dev); 403 /* We refer to the device */ 404 netdev_hold(dev, &ndev->dev_tracker, GFP_KERNEL); 405 406 if (snmp6_alloc_dev(ndev) < 0) { 407 netdev_dbg(dev, "%s: cannot allocate memory for statistics\n", 408 __func__); 409 neigh_parms_release(&nd_tbl, ndev->nd_parms); 410 netdev_put(dev, &ndev->dev_tracker); 411 kfree(ndev); 412 return ERR_PTR(err); 413 } 414 415 if (dev != blackhole_netdev) { 416 if (snmp6_register_dev(ndev) < 0) { 417 netdev_dbg(dev, "%s: cannot create /proc/net/dev_snmp6/%s\n", 418 __func__, dev->name); 419 goto err_release; 420 } 421 } 422 /* One reference from device. */ 423 refcount_set(&ndev->refcnt, 1); 424 425 if (dev->flags & (IFF_NOARP | IFF_LOOPBACK)) 426 ndev->cnf.accept_dad = -1; 427 428 #if IS_ENABLED(CONFIG_IPV6_SIT) 429 if (dev->type == ARPHRD_SIT && (dev->priv_flags & IFF_ISATAP)) { 430 pr_info("%s: Disabled Multicast RS\n", dev->name); 431 ndev->cnf.rtr_solicits = 0; 432 } 433 #endif 434 435 INIT_LIST_HEAD(&ndev->tempaddr_list); 436 ndev->desync_factor = U32_MAX; 437 if ((dev->flags&IFF_LOOPBACK) || 438 dev->type == ARPHRD_TUNNEL || 439 dev->type == ARPHRD_TUNNEL6 || 440 dev->type == ARPHRD_SIT || 441 dev->type == ARPHRD_NONE) { 442 ndev->cnf.use_tempaddr = -1; 443 } 444 445 ndev->token = in6addr_any; 446 447 if (netif_running(dev) && addrconf_link_ready(dev)) 448 ndev->if_flags |= IF_READY; 449 450 ipv6_mc_init_dev(ndev); 451 ndev->tstamp = jiffies; 452 if (dev != blackhole_netdev) { 453 err = addrconf_sysctl_register(ndev); 454 if (err) { 455 ipv6_mc_destroy_dev(ndev); 456 snmp6_unregister_dev(ndev); 457 goto err_release; 458 } 459 } 460 /* protected by rtnl_lock */ 461 rcu_assign_pointer(dev->ip6_ptr, ndev); 462 463 if (dev != blackhole_netdev) { 464 /* Join interface-local all-node multicast group */ 465 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes); 466 467 /* Join all-node multicast group */ 468 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allnodes); 469 470 /* Join all-router multicast group if forwarding is set */ 471 if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST)) 472 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 473 } 474 return ndev; 475 476 err_release: 477 neigh_parms_release(&nd_tbl, ndev->nd_parms); 478 ndev->dead = 1; 479 in6_dev_finish_destroy(ndev); 480 return ERR_PTR(err); 481 } 482 483 static struct inet6_dev *ipv6_find_idev(struct net_device *dev) 484 { 485 struct inet6_dev *idev; 486 487 ASSERT_RTNL(); 488 489 idev = __in6_dev_get(dev); 490 if (!idev) { 491 idev = ipv6_add_dev(dev); 492 if (IS_ERR(idev)) 493 return idev; 494 } 495 496 if (dev->flags&IFF_UP) 497 ipv6_mc_up(idev); 498 return idev; 499 } 500 501 static int inet6_netconf_msgsize_devconf(int type) 502 { 503 int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) 504 + nla_total_size(4); /* NETCONFA_IFINDEX */ 505 bool all = false; 506 507 if (type == NETCONFA_ALL) 508 all = true; 509 510 if (all || type == NETCONFA_FORWARDING) 511 size += nla_total_size(4); 512 #ifdef CONFIG_IPV6_MROUTE 513 if (all || type == NETCONFA_MC_FORWARDING) 514 size += nla_total_size(4); 515 #endif 516 if (all || type == NETCONFA_PROXY_NEIGH) 517 size += nla_total_size(4); 518 519 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) 520 size += nla_total_size(4); 521 522 return size; 523 } 524 525 static int inet6_netconf_fill_devconf(struct sk_buff *skb, int ifindex, 526 struct ipv6_devconf *devconf, u32 portid, 527 u32 seq, int event, unsigned int flags, 528 int type) 529 { 530 struct nlmsghdr *nlh; 531 struct netconfmsg *ncm; 532 bool all = false; 533 534 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), 535 flags); 536 if (!nlh) 537 return -EMSGSIZE; 538 539 if (type == NETCONFA_ALL) 540 all = true; 541 542 ncm = nlmsg_data(nlh); 543 ncm->ncm_family = AF_INET6; 544 545 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0) 546 goto nla_put_failure; 547 548 if (!devconf) 549 goto out; 550 551 if ((all || type == NETCONFA_FORWARDING) && 552 nla_put_s32(skb, NETCONFA_FORWARDING, devconf->forwarding) < 0) 553 goto nla_put_failure; 554 #ifdef CONFIG_IPV6_MROUTE 555 if ((all || type == NETCONFA_MC_FORWARDING) && 556 nla_put_s32(skb, NETCONFA_MC_FORWARDING, 557 atomic_read(&devconf->mc_forwarding)) < 0) 558 goto nla_put_failure; 559 #endif 560 if ((all || type == NETCONFA_PROXY_NEIGH) && 561 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, devconf->proxy_ndp) < 0) 562 goto nla_put_failure; 563 564 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) && 565 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 566 devconf->ignore_routes_with_linkdown) < 0) 567 goto nla_put_failure; 568 569 out: 570 nlmsg_end(skb, nlh); 571 return 0; 572 573 nla_put_failure: 574 nlmsg_cancel(skb, nlh); 575 return -EMSGSIZE; 576 } 577 578 void inet6_netconf_notify_devconf(struct net *net, int event, int type, 579 int ifindex, struct ipv6_devconf *devconf) 580 { 581 struct sk_buff *skb; 582 int err = -ENOBUFS; 583 584 skb = nlmsg_new(inet6_netconf_msgsize_devconf(type), GFP_KERNEL); 585 if (!skb) 586 goto errout; 587 588 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 0, 0, 589 event, 0, type); 590 if (err < 0) { 591 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 592 WARN_ON(err == -EMSGSIZE); 593 kfree_skb(skb); 594 goto errout; 595 } 596 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_NETCONF, NULL, GFP_KERNEL); 597 return; 598 errout: 599 rtnl_set_sk_err(net, RTNLGRP_IPV6_NETCONF, err); 600 } 601 602 static const struct nla_policy devconf_ipv6_policy[NETCONFA_MAX+1] = { 603 [NETCONFA_IFINDEX] = { .len = sizeof(int) }, 604 [NETCONFA_FORWARDING] = { .len = sizeof(int) }, 605 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) }, 606 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) }, 607 }; 608 609 static int inet6_netconf_valid_get_req(struct sk_buff *skb, 610 const struct nlmsghdr *nlh, 611 struct nlattr **tb, 612 struct netlink_ext_ack *extack) 613 { 614 int i, err; 615 616 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) { 617 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf get request"); 618 return -EINVAL; 619 } 620 621 if (!netlink_strict_get_check(skb)) 622 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg), 623 tb, NETCONFA_MAX, 624 devconf_ipv6_policy, extack); 625 626 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg), 627 tb, NETCONFA_MAX, 628 devconf_ipv6_policy, extack); 629 if (err) 630 return err; 631 632 for (i = 0; i <= NETCONFA_MAX; i++) { 633 if (!tb[i]) 634 continue; 635 636 switch (i) { 637 case NETCONFA_IFINDEX: 638 break; 639 default: 640 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in netconf get request"); 641 return -EINVAL; 642 } 643 } 644 645 return 0; 646 } 647 648 static int inet6_netconf_get_devconf(struct sk_buff *in_skb, 649 struct nlmsghdr *nlh, 650 struct netlink_ext_ack *extack) 651 { 652 struct net *net = sock_net(in_skb->sk); 653 struct nlattr *tb[NETCONFA_MAX+1]; 654 struct inet6_dev *in6_dev = NULL; 655 struct net_device *dev = NULL; 656 struct sk_buff *skb; 657 struct ipv6_devconf *devconf; 658 int ifindex; 659 int err; 660 661 err = inet6_netconf_valid_get_req(in_skb, nlh, tb, extack); 662 if (err < 0) 663 return err; 664 665 if (!tb[NETCONFA_IFINDEX]) 666 return -EINVAL; 667 668 err = -EINVAL; 669 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); 670 switch (ifindex) { 671 case NETCONFA_IFINDEX_ALL: 672 devconf = net->ipv6.devconf_all; 673 break; 674 case NETCONFA_IFINDEX_DEFAULT: 675 devconf = net->ipv6.devconf_dflt; 676 break; 677 default: 678 dev = dev_get_by_index(net, ifindex); 679 if (!dev) 680 return -EINVAL; 681 in6_dev = in6_dev_get(dev); 682 if (!in6_dev) 683 goto errout; 684 devconf = &in6_dev->cnf; 685 break; 686 } 687 688 err = -ENOBUFS; 689 skb = nlmsg_new(inet6_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); 690 if (!skb) 691 goto errout; 692 693 err = inet6_netconf_fill_devconf(skb, ifindex, devconf, 694 NETLINK_CB(in_skb).portid, 695 nlh->nlmsg_seq, RTM_NEWNETCONF, 0, 696 NETCONFA_ALL); 697 if (err < 0) { 698 /* -EMSGSIZE implies BUG in inet6_netconf_msgsize_devconf() */ 699 WARN_ON(err == -EMSGSIZE); 700 kfree_skb(skb); 701 goto errout; 702 } 703 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 704 errout: 705 if (in6_dev) 706 in6_dev_put(in6_dev); 707 dev_put(dev); 708 return err; 709 } 710 711 static int inet6_netconf_dump_devconf(struct sk_buff *skb, 712 struct netlink_callback *cb) 713 { 714 const struct nlmsghdr *nlh = cb->nlh; 715 struct net *net = sock_net(skb->sk); 716 int h, s_h; 717 int idx, s_idx; 718 struct net_device *dev; 719 struct inet6_dev *idev; 720 struct hlist_head *head; 721 722 if (cb->strict_check) { 723 struct netlink_ext_ack *extack = cb->extack; 724 struct netconfmsg *ncm; 725 726 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) { 727 NL_SET_ERR_MSG_MOD(extack, "Invalid header for netconf dump request"); 728 return -EINVAL; 729 } 730 731 if (nlmsg_attrlen(nlh, sizeof(*ncm))) { 732 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header in netconf dump request"); 733 return -EINVAL; 734 } 735 } 736 737 s_h = cb->args[0]; 738 s_idx = idx = cb->args[1]; 739 740 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 741 idx = 0; 742 head = &net->dev_index_head[h]; 743 rcu_read_lock(); 744 cb->seq = atomic_read(&net->ipv6.dev_addr_genid) ^ 745 net->dev_base_seq; 746 hlist_for_each_entry_rcu(dev, head, index_hlist) { 747 if (idx < s_idx) 748 goto cont; 749 idev = __in6_dev_get(dev); 750 if (!idev) 751 goto cont; 752 753 if (inet6_netconf_fill_devconf(skb, dev->ifindex, 754 &idev->cnf, 755 NETLINK_CB(cb->skb).portid, 756 nlh->nlmsg_seq, 757 RTM_NEWNETCONF, 758 NLM_F_MULTI, 759 NETCONFA_ALL) < 0) { 760 rcu_read_unlock(); 761 goto done; 762 } 763 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 764 cont: 765 idx++; 766 } 767 rcu_read_unlock(); 768 } 769 if (h == NETDEV_HASHENTRIES) { 770 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL, 771 net->ipv6.devconf_all, 772 NETLINK_CB(cb->skb).portid, 773 nlh->nlmsg_seq, 774 RTM_NEWNETCONF, NLM_F_MULTI, 775 NETCONFA_ALL) < 0) 776 goto done; 777 else 778 h++; 779 } 780 if (h == NETDEV_HASHENTRIES + 1) { 781 if (inet6_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT, 782 net->ipv6.devconf_dflt, 783 NETLINK_CB(cb->skb).portid, 784 nlh->nlmsg_seq, 785 RTM_NEWNETCONF, NLM_F_MULTI, 786 NETCONFA_ALL) < 0) 787 goto done; 788 else 789 h++; 790 } 791 done: 792 cb->args[0] = h; 793 cb->args[1] = idx; 794 795 return skb->len; 796 } 797 798 #ifdef CONFIG_SYSCTL 799 static void dev_forward_change(struct inet6_dev *idev) 800 { 801 struct net_device *dev; 802 struct inet6_ifaddr *ifa; 803 LIST_HEAD(tmp_addr_list); 804 805 if (!idev) 806 return; 807 dev = idev->dev; 808 if (idev->cnf.forwarding) 809 dev_disable_lro(dev); 810 if (dev->flags & IFF_MULTICAST) { 811 if (idev->cnf.forwarding) { 812 ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters); 813 ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allrouters); 814 ipv6_dev_mc_inc(dev, &in6addr_sitelocal_allrouters); 815 } else { 816 ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters); 817 ipv6_dev_mc_dec(dev, &in6addr_interfacelocal_allrouters); 818 ipv6_dev_mc_dec(dev, &in6addr_sitelocal_allrouters); 819 } 820 } 821 822 read_lock_bh(&idev->lock); 823 list_for_each_entry(ifa, &idev->addr_list, if_list) { 824 if (ifa->flags&IFA_F_TENTATIVE) 825 continue; 826 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 827 } 828 read_unlock_bh(&idev->lock); 829 830 while (!list_empty(&tmp_addr_list)) { 831 ifa = list_first_entry(&tmp_addr_list, 832 struct inet6_ifaddr, if_list_aux); 833 list_del(&ifa->if_list_aux); 834 if (idev->cnf.forwarding) 835 addrconf_join_anycast(ifa); 836 else 837 addrconf_leave_anycast(ifa); 838 } 839 840 inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF, 841 NETCONFA_FORWARDING, 842 dev->ifindex, &idev->cnf); 843 } 844 845 846 static void addrconf_forward_change(struct net *net, __s32 newf) 847 { 848 struct net_device *dev; 849 struct inet6_dev *idev; 850 851 for_each_netdev(net, dev) { 852 idev = __in6_dev_get(dev); 853 if (idev) { 854 int changed = (!idev->cnf.forwarding) ^ (!newf); 855 idev->cnf.forwarding = newf; 856 if (changed) 857 dev_forward_change(idev); 858 } 859 } 860 } 861 862 static int addrconf_fixup_forwarding(struct ctl_table *table, int *p, int newf) 863 { 864 struct net *net; 865 int old; 866 867 if (!rtnl_trylock()) 868 return restart_syscall(); 869 870 net = (struct net *)table->extra2; 871 old = *p; 872 *p = newf; 873 874 if (p == &net->ipv6.devconf_dflt->forwarding) { 875 if ((!newf) ^ (!old)) 876 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 877 NETCONFA_FORWARDING, 878 NETCONFA_IFINDEX_DEFAULT, 879 net->ipv6.devconf_dflt); 880 rtnl_unlock(); 881 return 0; 882 } 883 884 if (p == &net->ipv6.devconf_all->forwarding) { 885 int old_dflt = net->ipv6.devconf_dflt->forwarding; 886 887 net->ipv6.devconf_dflt->forwarding = newf; 888 if ((!newf) ^ (!old_dflt)) 889 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 890 NETCONFA_FORWARDING, 891 NETCONFA_IFINDEX_DEFAULT, 892 net->ipv6.devconf_dflt); 893 894 addrconf_forward_change(net, newf); 895 if ((!newf) ^ (!old)) 896 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 897 NETCONFA_FORWARDING, 898 NETCONFA_IFINDEX_ALL, 899 net->ipv6.devconf_all); 900 } else if ((!newf) ^ (!old)) 901 dev_forward_change((struct inet6_dev *)table->extra1); 902 rtnl_unlock(); 903 904 if (newf) 905 rt6_purge_dflt_routers(net); 906 return 1; 907 } 908 909 static void addrconf_linkdown_change(struct net *net, __s32 newf) 910 { 911 struct net_device *dev; 912 struct inet6_dev *idev; 913 914 for_each_netdev(net, dev) { 915 idev = __in6_dev_get(dev); 916 if (idev) { 917 int changed = (!idev->cnf.ignore_routes_with_linkdown) ^ (!newf); 918 919 idev->cnf.ignore_routes_with_linkdown = newf; 920 if (changed) 921 inet6_netconf_notify_devconf(dev_net(dev), 922 RTM_NEWNETCONF, 923 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 924 dev->ifindex, 925 &idev->cnf); 926 } 927 } 928 } 929 930 static int addrconf_fixup_linkdown(struct ctl_table *table, int *p, int newf) 931 { 932 struct net *net; 933 int old; 934 935 if (!rtnl_trylock()) 936 return restart_syscall(); 937 938 net = (struct net *)table->extra2; 939 old = *p; 940 *p = newf; 941 942 if (p == &net->ipv6.devconf_dflt->ignore_routes_with_linkdown) { 943 if ((!newf) ^ (!old)) 944 inet6_netconf_notify_devconf(net, 945 RTM_NEWNETCONF, 946 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 947 NETCONFA_IFINDEX_DEFAULT, 948 net->ipv6.devconf_dflt); 949 rtnl_unlock(); 950 return 0; 951 } 952 953 if (p == &net->ipv6.devconf_all->ignore_routes_with_linkdown) { 954 net->ipv6.devconf_dflt->ignore_routes_with_linkdown = newf; 955 addrconf_linkdown_change(net, newf); 956 if ((!newf) ^ (!old)) 957 inet6_netconf_notify_devconf(net, 958 RTM_NEWNETCONF, 959 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 960 NETCONFA_IFINDEX_ALL, 961 net->ipv6.devconf_all); 962 } 963 rtnl_unlock(); 964 965 return 1; 966 } 967 968 #endif 969 970 /* Nobody refers to this ifaddr, destroy it */ 971 void inet6_ifa_finish_destroy(struct inet6_ifaddr *ifp) 972 { 973 WARN_ON(!hlist_unhashed(&ifp->addr_lst)); 974 975 #ifdef NET_REFCNT_DEBUG 976 pr_debug("%s\n", __func__); 977 #endif 978 979 in6_dev_put(ifp->idev); 980 981 if (cancel_delayed_work(&ifp->dad_work)) 982 pr_notice("delayed DAD work was pending while freeing ifa=%p\n", 983 ifp); 984 985 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 986 pr_warn("Freeing alive inet6 address %p\n", ifp); 987 return; 988 } 989 990 kfree_rcu(ifp, rcu); 991 } 992 993 static void 994 ipv6_link_dev_addr(struct inet6_dev *idev, struct inet6_ifaddr *ifp) 995 { 996 struct list_head *p; 997 int ifp_scope = ipv6_addr_src_scope(&ifp->addr); 998 999 /* 1000 * Each device address list is sorted in order of scope - 1001 * global before linklocal. 1002 */ 1003 list_for_each(p, &idev->addr_list) { 1004 struct inet6_ifaddr *ifa 1005 = list_entry(p, struct inet6_ifaddr, if_list); 1006 if (ifp_scope >= ipv6_addr_src_scope(&ifa->addr)) 1007 break; 1008 } 1009 1010 list_add_tail_rcu(&ifp->if_list, p); 1011 } 1012 1013 static u32 inet6_addr_hash(const struct net *net, const struct in6_addr *addr) 1014 { 1015 u32 val = ipv6_addr_hash(addr) ^ net_hash_mix(net); 1016 1017 return hash_32(val, IN6_ADDR_HSIZE_SHIFT); 1018 } 1019 1020 static bool ipv6_chk_same_addr(struct net *net, const struct in6_addr *addr, 1021 struct net_device *dev, unsigned int hash) 1022 { 1023 struct inet6_ifaddr *ifp; 1024 1025 hlist_for_each_entry(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 1026 if (ipv6_addr_equal(&ifp->addr, addr)) { 1027 if (!dev || ifp->idev->dev == dev) 1028 return true; 1029 } 1030 } 1031 return false; 1032 } 1033 1034 static int ipv6_add_addr_hash(struct net_device *dev, struct inet6_ifaddr *ifa) 1035 { 1036 struct net *net = dev_net(dev); 1037 unsigned int hash = inet6_addr_hash(net, &ifa->addr); 1038 int err = 0; 1039 1040 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1041 1042 /* Ignore adding duplicate addresses on an interface */ 1043 if (ipv6_chk_same_addr(net, &ifa->addr, dev, hash)) { 1044 netdev_dbg(dev, "ipv6_add_addr: already assigned\n"); 1045 err = -EEXIST; 1046 } else { 1047 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv6.inet6_addr_lst[hash]); 1048 } 1049 1050 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1051 1052 return err; 1053 } 1054 1055 /* On success it returns ifp with increased reference count */ 1056 1057 static struct inet6_ifaddr * 1058 ipv6_add_addr(struct inet6_dev *idev, struct ifa6_config *cfg, 1059 bool can_block, struct netlink_ext_ack *extack) 1060 { 1061 gfp_t gfp_flags = can_block ? GFP_KERNEL : GFP_ATOMIC; 1062 int addr_type = ipv6_addr_type(cfg->pfx); 1063 struct net *net = dev_net(idev->dev); 1064 struct inet6_ifaddr *ifa = NULL; 1065 struct fib6_info *f6i = NULL; 1066 int err = 0; 1067 1068 if (addr_type == IPV6_ADDR_ANY) { 1069 NL_SET_ERR_MSG_MOD(extack, "Invalid address"); 1070 return ERR_PTR(-EADDRNOTAVAIL); 1071 } else if (addr_type & IPV6_ADDR_MULTICAST && 1072 !(cfg->ifa_flags & IFA_F_MCAUTOJOIN)) { 1073 NL_SET_ERR_MSG_MOD(extack, "Cannot assign multicast address without \"IFA_F_MCAUTOJOIN\" flag"); 1074 return ERR_PTR(-EADDRNOTAVAIL); 1075 } else if (!(idev->dev->flags & IFF_LOOPBACK) && 1076 !netif_is_l3_master(idev->dev) && 1077 addr_type & IPV6_ADDR_LOOPBACK) { 1078 NL_SET_ERR_MSG_MOD(extack, "Cannot assign loopback address on this device"); 1079 return ERR_PTR(-EADDRNOTAVAIL); 1080 } 1081 1082 if (idev->dead) { 1083 NL_SET_ERR_MSG_MOD(extack, "device is going away"); 1084 err = -ENODEV; 1085 goto out; 1086 } 1087 1088 if (idev->cnf.disable_ipv6) { 1089 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 1090 err = -EACCES; 1091 goto out; 1092 } 1093 1094 /* validator notifier needs to be blocking; 1095 * do not call in atomic context 1096 */ 1097 if (can_block) { 1098 struct in6_validator_info i6vi = { 1099 .i6vi_addr = *cfg->pfx, 1100 .i6vi_dev = idev, 1101 .extack = extack, 1102 }; 1103 1104 err = inet6addr_validator_notifier_call_chain(NETDEV_UP, &i6vi); 1105 err = notifier_to_errno(err); 1106 if (err < 0) 1107 goto out; 1108 } 1109 1110 ifa = kzalloc(sizeof(*ifa), gfp_flags | __GFP_ACCOUNT); 1111 if (!ifa) { 1112 err = -ENOBUFS; 1113 goto out; 1114 } 1115 1116 f6i = addrconf_f6i_alloc(net, idev, cfg->pfx, false, gfp_flags, extack); 1117 if (IS_ERR(f6i)) { 1118 err = PTR_ERR(f6i); 1119 f6i = NULL; 1120 goto out; 1121 } 1122 1123 neigh_parms_data_state_setall(idev->nd_parms); 1124 1125 ifa->addr = *cfg->pfx; 1126 if (cfg->peer_pfx) 1127 ifa->peer_addr = *cfg->peer_pfx; 1128 1129 spin_lock_init(&ifa->lock); 1130 INIT_DELAYED_WORK(&ifa->dad_work, addrconf_dad_work); 1131 INIT_HLIST_NODE(&ifa->addr_lst); 1132 ifa->scope = cfg->scope; 1133 ifa->prefix_len = cfg->plen; 1134 ifa->rt_priority = cfg->rt_priority; 1135 ifa->flags = cfg->ifa_flags; 1136 ifa->ifa_proto = cfg->ifa_proto; 1137 /* No need to add the TENTATIVE flag for addresses with NODAD */ 1138 if (!(cfg->ifa_flags & IFA_F_NODAD)) 1139 ifa->flags |= IFA_F_TENTATIVE; 1140 ifa->valid_lft = cfg->valid_lft; 1141 ifa->prefered_lft = cfg->preferred_lft; 1142 ifa->cstamp = ifa->tstamp = jiffies; 1143 ifa->tokenized = false; 1144 1145 ifa->rt = f6i; 1146 1147 ifa->idev = idev; 1148 in6_dev_hold(idev); 1149 1150 /* For caller */ 1151 refcount_set(&ifa->refcnt, 1); 1152 1153 rcu_read_lock(); 1154 1155 err = ipv6_add_addr_hash(idev->dev, ifa); 1156 if (err < 0) { 1157 rcu_read_unlock(); 1158 goto out; 1159 } 1160 1161 write_lock_bh(&idev->lock); 1162 1163 /* Add to inet6_dev unicast addr list. */ 1164 ipv6_link_dev_addr(idev, ifa); 1165 1166 if (ifa->flags&IFA_F_TEMPORARY) { 1167 list_add(&ifa->tmp_list, &idev->tempaddr_list); 1168 in6_ifa_hold(ifa); 1169 } 1170 1171 in6_ifa_hold(ifa); 1172 write_unlock_bh(&idev->lock); 1173 1174 rcu_read_unlock(); 1175 1176 inet6addr_notifier_call_chain(NETDEV_UP, ifa); 1177 out: 1178 if (unlikely(err < 0)) { 1179 fib6_info_release(f6i); 1180 1181 if (ifa) { 1182 if (ifa->idev) 1183 in6_dev_put(ifa->idev); 1184 kfree(ifa); 1185 } 1186 ifa = ERR_PTR(err); 1187 } 1188 1189 return ifa; 1190 } 1191 1192 enum cleanup_prefix_rt_t { 1193 CLEANUP_PREFIX_RT_NOP, /* no cleanup action for prefix route */ 1194 CLEANUP_PREFIX_RT_DEL, /* delete the prefix route */ 1195 CLEANUP_PREFIX_RT_EXPIRE, /* update the lifetime of the prefix route */ 1196 }; 1197 1198 /* 1199 * Check, whether the prefix for ifp would still need a prefix route 1200 * after deleting ifp. The function returns one of the CLEANUP_PREFIX_RT_* 1201 * constants. 1202 * 1203 * 1) we don't purge prefix if address was not permanent. 1204 * prefix is managed by its own lifetime. 1205 * 2) we also don't purge, if the address was IFA_F_NOPREFIXROUTE. 1206 * 3) if there are no addresses, delete prefix. 1207 * 4) if there are still other permanent address(es), 1208 * corresponding prefix is still permanent. 1209 * 5) if there are still other addresses with IFA_F_NOPREFIXROUTE, 1210 * don't purge the prefix, assume user space is managing it. 1211 * 6) otherwise, update prefix lifetime to the 1212 * longest valid lifetime among the corresponding 1213 * addresses on the device. 1214 * Note: subsequent RA will update lifetime. 1215 **/ 1216 static enum cleanup_prefix_rt_t 1217 check_cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long *expires) 1218 { 1219 struct inet6_ifaddr *ifa; 1220 struct inet6_dev *idev = ifp->idev; 1221 unsigned long lifetime; 1222 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_DEL; 1223 1224 *expires = jiffies; 1225 1226 list_for_each_entry(ifa, &idev->addr_list, if_list) { 1227 if (ifa == ifp) 1228 continue; 1229 if (ifa->prefix_len != ifp->prefix_len || 1230 !ipv6_prefix_equal(&ifa->addr, &ifp->addr, 1231 ifp->prefix_len)) 1232 continue; 1233 if (ifa->flags & (IFA_F_PERMANENT | IFA_F_NOPREFIXROUTE)) 1234 return CLEANUP_PREFIX_RT_NOP; 1235 1236 action = CLEANUP_PREFIX_RT_EXPIRE; 1237 1238 spin_lock(&ifa->lock); 1239 1240 lifetime = addrconf_timeout_fixup(ifa->valid_lft, HZ); 1241 /* 1242 * Note: Because this address is 1243 * not permanent, lifetime < 1244 * LONG_MAX / HZ here. 1245 */ 1246 if (time_before(*expires, ifa->tstamp + lifetime * HZ)) 1247 *expires = ifa->tstamp + lifetime * HZ; 1248 spin_unlock(&ifa->lock); 1249 } 1250 1251 return action; 1252 } 1253 1254 static void 1255 cleanup_prefix_route(struct inet6_ifaddr *ifp, unsigned long expires, 1256 bool del_rt, bool del_peer) 1257 { 1258 struct fib6_info *f6i; 1259 1260 f6i = addrconf_get_prefix_route(del_peer ? &ifp->peer_addr : &ifp->addr, 1261 ifp->prefix_len, 1262 ifp->idev->dev, 0, RTF_DEFAULT, true); 1263 if (f6i) { 1264 if (del_rt) 1265 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false); 1266 else { 1267 if (!(f6i->fib6_flags & RTF_EXPIRES)) 1268 fib6_set_expires(f6i, expires); 1269 fib6_info_release(f6i); 1270 } 1271 } 1272 } 1273 1274 1275 /* This function wants to get referenced ifp and releases it before return */ 1276 1277 static void ipv6_del_addr(struct inet6_ifaddr *ifp) 1278 { 1279 enum cleanup_prefix_rt_t action = CLEANUP_PREFIX_RT_NOP; 1280 struct net *net = dev_net(ifp->idev->dev); 1281 unsigned long expires; 1282 int state; 1283 1284 ASSERT_RTNL(); 1285 1286 spin_lock_bh(&ifp->lock); 1287 state = ifp->state; 1288 ifp->state = INET6_IFADDR_STATE_DEAD; 1289 spin_unlock_bh(&ifp->lock); 1290 1291 if (state == INET6_IFADDR_STATE_DEAD) 1292 goto out; 1293 1294 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 1295 hlist_del_init_rcu(&ifp->addr_lst); 1296 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 1297 1298 write_lock_bh(&ifp->idev->lock); 1299 1300 if (ifp->flags&IFA_F_TEMPORARY) { 1301 list_del(&ifp->tmp_list); 1302 if (ifp->ifpub) { 1303 in6_ifa_put(ifp->ifpub); 1304 ifp->ifpub = NULL; 1305 } 1306 __in6_ifa_put(ifp); 1307 } 1308 1309 if (ifp->flags & IFA_F_PERMANENT && !(ifp->flags & IFA_F_NOPREFIXROUTE)) 1310 action = check_cleanup_prefix_route(ifp, &expires); 1311 1312 list_del_rcu(&ifp->if_list); 1313 __in6_ifa_put(ifp); 1314 1315 write_unlock_bh(&ifp->idev->lock); 1316 1317 addrconf_del_dad_work(ifp); 1318 1319 ipv6_ifa_notify(RTM_DELADDR, ifp); 1320 1321 inet6addr_notifier_call_chain(NETDEV_DOWN, ifp); 1322 1323 if (action != CLEANUP_PREFIX_RT_NOP) { 1324 cleanup_prefix_route(ifp, expires, 1325 action == CLEANUP_PREFIX_RT_DEL, false); 1326 } 1327 1328 /* clean up prefsrc entries */ 1329 rt6_remove_prefsrc(ifp); 1330 out: 1331 in6_ifa_put(ifp); 1332 } 1333 1334 static int ipv6_create_tempaddr(struct inet6_ifaddr *ifp, bool block) 1335 { 1336 struct inet6_dev *idev = ifp->idev; 1337 unsigned long tmp_tstamp, age; 1338 unsigned long regen_advance; 1339 unsigned long now = jiffies; 1340 s32 cnf_temp_preferred_lft; 1341 struct inet6_ifaddr *ift; 1342 struct ifa6_config cfg; 1343 long max_desync_factor; 1344 struct in6_addr addr; 1345 int ret = 0; 1346 1347 write_lock_bh(&idev->lock); 1348 1349 retry: 1350 in6_dev_hold(idev); 1351 if (idev->cnf.use_tempaddr <= 0) { 1352 write_unlock_bh(&idev->lock); 1353 pr_info("%s: use_tempaddr is disabled\n", __func__); 1354 in6_dev_put(idev); 1355 ret = -1; 1356 goto out; 1357 } 1358 spin_lock_bh(&ifp->lock); 1359 if (ifp->regen_count++ >= idev->cnf.regen_max_retry) { 1360 idev->cnf.use_tempaddr = -1; /*XXX*/ 1361 spin_unlock_bh(&ifp->lock); 1362 write_unlock_bh(&idev->lock); 1363 pr_warn("%s: regeneration time exceeded - disabled temporary address support\n", 1364 __func__); 1365 in6_dev_put(idev); 1366 ret = -1; 1367 goto out; 1368 } 1369 in6_ifa_hold(ifp); 1370 memcpy(addr.s6_addr, ifp->addr.s6_addr, 8); 1371 ipv6_gen_rnd_iid(&addr); 1372 1373 age = (now - ifp->tstamp) / HZ; 1374 1375 regen_advance = idev->cnf.regen_max_retry * 1376 idev->cnf.dad_transmits * 1377 max(NEIGH_VAR(idev->nd_parms, RETRANS_TIME), HZ/100) / HZ; 1378 1379 /* recalculate max_desync_factor each time and update 1380 * idev->desync_factor if it's larger 1381 */ 1382 cnf_temp_preferred_lft = READ_ONCE(idev->cnf.temp_prefered_lft); 1383 max_desync_factor = min_t(long, 1384 idev->cnf.max_desync_factor, 1385 cnf_temp_preferred_lft - regen_advance); 1386 1387 if (unlikely(idev->desync_factor > max_desync_factor)) { 1388 if (max_desync_factor > 0) { 1389 get_random_bytes(&idev->desync_factor, 1390 sizeof(idev->desync_factor)); 1391 idev->desync_factor %= max_desync_factor; 1392 } else { 1393 idev->desync_factor = 0; 1394 } 1395 } 1396 1397 memset(&cfg, 0, sizeof(cfg)); 1398 cfg.valid_lft = min_t(__u32, ifp->valid_lft, 1399 idev->cnf.temp_valid_lft + age); 1400 cfg.preferred_lft = cnf_temp_preferred_lft + age - idev->desync_factor; 1401 cfg.preferred_lft = min_t(__u32, ifp->prefered_lft, cfg.preferred_lft); 1402 cfg.preferred_lft = min_t(__u32, cfg.valid_lft, cfg.preferred_lft); 1403 1404 cfg.plen = ifp->prefix_len; 1405 tmp_tstamp = ifp->tstamp; 1406 spin_unlock_bh(&ifp->lock); 1407 1408 write_unlock_bh(&idev->lock); 1409 1410 /* From RFC 4941: 1411 * 1412 * A temporary address is created only if this calculated Preferred 1413 * Lifetime is greater than REGEN_ADVANCE time units. In 1414 * particular, an implementation must not create a temporary address 1415 * with a zero Preferred Lifetime. 1416 * 1417 * Clamp the preferred lifetime to a minimum of regen_advance, unless 1418 * that would exceed valid_lft. 1419 * 1420 * Use age calculation as in addrconf_verify to avoid unnecessary 1421 * temporary addresses being generated. 1422 */ 1423 age = (now - tmp_tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 1424 if (cfg.preferred_lft <= regen_advance + age) 1425 cfg.preferred_lft = regen_advance + age + 1; 1426 if (cfg.preferred_lft > cfg.valid_lft) { 1427 in6_ifa_put(ifp); 1428 in6_dev_put(idev); 1429 ret = -1; 1430 goto out; 1431 } 1432 1433 cfg.ifa_flags = IFA_F_TEMPORARY; 1434 /* set in addrconf_prefix_rcv() */ 1435 if (ifp->flags & IFA_F_OPTIMISTIC) 1436 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 1437 1438 cfg.pfx = &addr; 1439 cfg.scope = ipv6_addr_scope(cfg.pfx); 1440 1441 ift = ipv6_add_addr(idev, &cfg, block, NULL); 1442 if (IS_ERR(ift)) { 1443 in6_ifa_put(ifp); 1444 in6_dev_put(idev); 1445 pr_info("%s: retry temporary address regeneration\n", __func__); 1446 write_lock_bh(&idev->lock); 1447 goto retry; 1448 } 1449 1450 spin_lock_bh(&ift->lock); 1451 ift->ifpub = ifp; 1452 ift->cstamp = now; 1453 ift->tstamp = tmp_tstamp; 1454 spin_unlock_bh(&ift->lock); 1455 1456 addrconf_dad_start(ift); 1457 in6_ifa_put(ift); 1458 in6_dev_put(idev); 1459 out: 1460 return ret; 1461 } 1462 1463 /* 1464 * Choose an appropriate source address (RFC3484) 1465 */ 1466 enum { 1467 IPV6_SADDR_RULE_INIT = 0, 1468 IPV6_SADDR_RULE_LOCAL, 1469 IPV6_SADDR_RULE_SCOPE, 1470 IPV6_SADDR_RULE_PREFERRED, 1471 #ifdef CONFIG_IPV6_MIP6 1472 IPV6_SADDR_RULE_HOA, 1473 #endif 1474 IPV6_SADDR_RULE_OIF, 1475 IPV6_SADDR_RULE_LABEL, 1476 IPV6_SADDR_RULE_PRIVACY, 1477 IPV6_SADDR_RULE_ORCHID, 1478 IPV6_SADDR_RULE_PREFIX, 1479 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1480 IPV6_SADDR_RULE_NOT_OPTIMISTIC, 1481 #endif 1482 IPV6_SADDR_RULE_MAX 1483 }; 1484 1485 struct ipv6_saddr_score { 1486 int rule; 1487 int addr_type; 1488 struct inet6_ifaddr *ifa; 1489 DECLARE_BITMAP(scorebits, IPV6_SADDR_RULE_MAX); 1490 int scopedist; 1491 int matchlen; 1492 }; 1493 1494 struct ipv6_saddr_dst { 1495 const struct in6_addr *addr; 1496 int ifindex; 1497 int scope; 1498 int label; 1499 unsigned int prefs; 1500 }; 1501 1502 static inline int ipv6_saddr_preferred(int type) 1503 { 1504 if (type & (IPV6_ADDR_MAPPED|IPV6_ADDR_COMPATv4|IPV6_ADDR_LOOPBACK)) 1505 return 1; 1506 return 0; 1507 } 1508 1509 static bool ipv6_use_optimistic_addr(struct net *net, 1510 struct inet6_dev *idev) 1511 { 1512 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1513 if (!idev) 1514 return false; 1515 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad) 1516 return false; 1517 if (!net->ipv6.devconf_all->use_optimistic && !idev->cnf.use_optimistic) 1518 return false; 1519 1520 return true; 1521 #else 1522 return false; 1523 #endif 1524 } 1525 1526 static bool ipv6_allow_optimistic_dad(struct net *net, 1527 struct inet6_dev *idev) 1528 { 1529 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1530 if (!idev) 1531 return false; 1532 if (!net->ipv6.devconf_all->optimistic_dad && !idev->cnf.optimistic_dad) 1533 return false; 1534 1535 return true; 1536 #else 1537 return false; 1538 #endif 1539 } 1540 1541 static int ipv6_get_saddr_eval(struct net *net, 1542 struct ipv6_saddr_score *score, 1543 struct ipv6_saddr_dst *dst, 1544 int i) 1545 { 1546 int ret; 1547 1548 if (i <= score->rule) { 1549 switch (i) { 1550 case IPV6_SADDR_RULE_SCOPE: 1551 ret = score->scopedist; 1552 break; 1553 case IPV6_SADDR_RULE_PREFIX: 1554 ret = score->matchlen; 1555 break; 1556 default: 1557 ret = !!test_bit(i, score->scorebits); 1558 } 1559 goto out; 1560 } 1561 1562 switch (i) { 1563 case IPV6_SADDR_RULE_INIT: 1564 /* Rule 0: remember if hiscore is not ready yet */ 1565 ret = !!score->ifa; 1566 break; 1567 case IPV6_SADDR_RULE_LOCAL: 1568 /* Rule 1: Prefer same address */ 1569 ret = ipv6_addr_equal(&score->ifa->addr, dst->addr); 1570 break; 1571 case IPV6_SADDR_RULE_SCOPE: 1572 /* Rule 2: Prefer appropriate scope 1573 * 1574 * ret 1575 * ^ 1576 * -1 | d 15 1577 * ---+--+-+---> scope 1578 * | 1579 * | d is scope of the destination. 1580 * B-d | \ 1581 * | \ <- smaller scope is better if 1582 * B-15 | \ if scope is enough for destination. 1583 * | ret = B - scope (-1 <= scope >= d <= 15). 1584 * d-C-1 | / 1585 * |/ <- greater is better 1586 * -C / if scope is not enough for destination. 1587 * /| ret = scope - C (-1 <= d < scope <= 15). 1588 * 1589 * d - C - 1 < B -15 (for all -1 <= d <= 15). 1590 * C > d + 14 - B >= 15 + 14 - B = 29 - B. 1591 * Assume B = 0 and we get C > 29. 1592 */ 1593 ret = __ipv6_addr_src_scope(score->addr_type); 1594 if (ret >= dst->scope) 1595 ret = -ret; 1596 else 1597 ret -= 128; /* 30 is enough */ 1598 score->scopedist = ret; 1599 break; 1600 case IPV6_SADDR_RULE_PREFERRED: 1601 { 1602 /* Rule 3: Avoid deprecated and optimistic addresses */ 1603 u8 avoid = IFA_F_DEPRECATED; 1604 1605 if (!ipv6_use_optimistic_addr(net, score->ifa->idev)) 1606 avoid |= IFA_F_OPTIMISTIC; 1607 ret = ipv6_saddr_preferred(score->addr_type) || 1608 !(score->ifa->flags & avoid); 1609 break; 1610 } 1611 #ifdef CONFIG_IPV6_MIP6 1612 case IPV6_SADDR_RULE_HOA: 1613 { 1614 /* Rule 4: Prefer home address */ 1615 int prefhome = !(dst->prefs & IPV6_PREFER_SRC_COA); 1616 ret = !(score->ifa->flags & IFA_F_HOMEADDRESS) ^ prefhome; 1617 break; 1618 } 1619 #endif 1620 case IPV6_SADDR_RULE_OIF: 1621 /* Rule 5: Prefer outgoing interface */ 1622 ret = (!dst->ifindex || 1623 dst->ifindex == score->ifa->idev->dev->ifindex); 1624 break; 1625 case IPV6_SADDR_RULE_LABEL: 1626 /* Rule 6: Prefer matching label */ 1627 ret = ipv6_addr_label(net, 1628 &score->ifa->addr, score->addr_type, 1629 score->ifa->idev->dev->ifindex) == dst->label; 1630 break; 1631 case IPV6_SADDR_RULE_PRIVACY: 1632 { 1633 /* Rule 7: Prefer public address 1634 * Note: prefer temporary address if use_tempaddr >= 2 1635 */ 1636 int preftmp = dst->prefs & (IPV6_PREFER_SRC_PUBLIC|IPV6_PREFER_SRC_TMP) ? 1637 !!(dst->prefs & IPV6_PREFER_SRC_TMP) : 1638 score->ifa->idev->cnf.use_tempaddr >= 2; 1639 ret = (!(score->ifa->flags & IFA_F_TEMPORARY)) ^ preftmp; 1640 break; 1641 } 1642 case IPV6_SADDR_RULE_ORCHID: 1643 /* Rule 8-: Prefer ORCHID vs ORCHID or 1644 * non-ORCHID vs non-ORCHID 1645 */ 1646 ret = !(ipv6_addr_orchid(&score->ifa->addr) ^ 1647 ipv6_addr_orchid(dst->addr)); 1648 break; 1649 case IPV6_SADDR_RULE_PREFIX: 1650 /* Rule 8: Use longest matching prefix */ 1651 ret = ipv6_addr_diff(&score->ifa->addr, dst->addr); 1652 if (ret > score->ifa->prefix_len) 1653 ret = score->ifa->prefix_len; 1654 score->matchlen = ret; 1655 break; 1656 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 1657 case IPV6_SADDR_RULE_NOT_OPTIMISTIC: 1658 /* Optimistic addresses still have lower precedence than other 1659 * preferred addresses. 1660 */ 1661 ret = !(score->ifa->flags & IFA_F_OPTIMISTIC); 1662 break; 1663 #endif 1664 default: 1665 ret = 0; 1666 } 1667 1668 if (ret) 1669 __set_bit(i, score->scorebits); 1670 score->rule = i; 1671 out: 1672 return ret; 1673 } 1674 1675 static int __ipv6_dev_get_saddr(struct net *net, 1676 struct ipv6_saddr_dst *dst, 1677 struct inet6_dev *idev, 1678 struct ipv6_saddr_score *scores, 1679 int hiscore_idx) 1680 { 1681 struct ipv6_saddr_score *score = &scores[1 - hiscore_idx], *hiscore = &scores[hiscore_idx]; 1682 1683 list_for_each_entry_rcu(score->ifa, &idev->addr_list, if_list) { 1684 int i; 1685 1686 /* 1687 * - Tentative Address (RFC2462 section 5.4) 1688 * - A tentative address is not considered 1689 * "assigned to an interface" in the traditional 1690 * sense, unless it is also flagged as optimistic. 1691 * - Candidate Source Address (section 4) 1692 * - In any case, anycast addresses, multicast 1693 * addresses, and the unspecified address MUST 1694 * NOT be included in a candidate set. 1695 */ 1696 if ((score->ifa->flags & IFA_F_TENTATIVE) && 1697 (!(score->ifa->flags & IFA_F_OPTIMISTIC))) 1698 continue; 1699 1700 score->addr_type = __ipv6_addr_type(&score->ifa->addr); 1701 1702 if (unlikely(score->addr_type == IPV6_ADDR_ANY || 1703 score->addr_type & IPV6_ADDR_MULTICAST)) { 1704 net_dbg_ratelimited("ADDRCONF: unspecified / multicast address assigned as unicast address on %s", 1705 idev->dev->name); 1706 continue; 1707 } 1708 1709 score->rule = -1; 1710 bitmap_zero(score->scorebits, IPV6_SADDR_RULE_MAX); 1711 1712 for (i = 0; i < IPV6_SADDR_RULE_MAX; i++) { 1713 int minihiscore, miniscore; 1714 1715 minihiscore = ipv6_get_saddr_eval(net, hiscore, dst, i); 1716 miniscore = ipv6_get_saddr_eval(net, score, dst, i); 1717 1718 if (minihiscore > miniscore) { 1719 if (i == IPV6_SADDR_RULE_SCOPE && 1720 score->scopedist > 0) { 1721 /* 1722 * special case: 1723 * each remaining entry 1724 * has too small (not enough) 1725 * scope, because ifa entries 1726 * are sorted by their scope 1727 * values. 1728 */ 1729 goto out; 1730 } 1731 break; 1732 } else if (minihiscore < miniscore) { 1733 swap(hiscore, score); 1734 hiscore_idx = 1 - hiscore_idx; 1735 1736 /* restore our iterator */ 1737 score->ifa = hiscore->ifa; 1738 1739 break; 1740 } 1741 } 1742 } 1743 out: 1744 return hiscore_idx; 1745 } 1746 1747 static int ipv6_get_saddr_master(struct net *net, 1748 const struct net_device *dst_dev, 1749 const struct net_device *master, 1750 struct ipv6_saddr_dst *dst, 1751 struct ipv6_saddr_score *scores, 1752 int hiscore_idx) 1753 { 1754 struct inet6_dev *idev; 1755 1756 idev = __in6_dev_get(dst_dev); 1757 if (idev) 1758 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1759 scores, hiscore_idx); 1760 1761 idev = __in6_dev_get(master); 1762 if (idev) 1763 hiscore_idx = __ipv6_dev_get_saddr(net, dst, idev, 1764 scores, hiscore_idx); 1765 1766 return hiscore_idx; 1767 } 1768 1769 int ipv6_dev_get_saddr(struct net *net, const struct net_device *dst_dev, 1770 const struct in6_addr *daddr, unsigned int prefs, 1771 struct in6_addr *saddr) 1772 { 1773 struct ipv6_saddr_score scores[2], *hiscore; 1774 struct ipv6_saddr_dst dst; 1775 struct inet6_dev *idev; 1776 struct net_device *dev; 1777 int dst_type; 1778 bool use_oif_addr = false; 1779 int hiscore_idx = 0; 1780 int ret = 0; 1781 1782 dst_type = __ipv6_addr_type(daddr); 1783 dst.addr = daddr; 1784 dst.ifindex = dst_dev ? dst_dev->ifindex : 0; 1785 dst.scope = __ipv6_addr_src_scope(dst_type); 1786 dst.label = ipv6_addr_label(net, daddr, dst_type, dst.ifindex); 1787 dst.prefs = prefs; 1788 1789 scores[hiscore_idx].rule = -1; 1790 scores[hiscore_idx].ifa = NULL; 1791 1792 rcu_read_lock(); 1793 1794 /* Candidate Source Address (section 4) 1795 * - multicast and link-local destination address, 1796 * the set of candidate source address MUST only 1797 * include addresses assigned to interfaces 1798 * belonging to the same link as the outgoing 1799 * interface. 1800 * (- For site-local destination addresses, the 1801 * set of candidate source addresses MUST only 1802 * include addresses assigned to interfaces 1803 * belonging to the same site as the outgoing 1804 * interface.) 1805 * - "It is RECOMMENDED that the candidate source addresses 1806 * be the set of unicast addresses assigned to the 1807 * interface that will be used to send to the destination 1808 * (the 'outgoing' interface)." (RFC 6724) 1809 */ 1810 if (dst_dev) { 1811 idev = __in6_dev_get(dst_dev); 1812 if ((dst_type & IPV6_ADDR_MULTICAST) || 1813 dst.scope <= IPV6_ADDR_SCOPE_LINKLOCAL || 1814 (idev && idev->cnf.use_oif_addrs_only)) { 1815 use_oif_addr = true; 1816 } 1817 } 1818 1819 if (use_oif_addr) { 1820 if (idev) 1821 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1822 } else { 1823 const struct net_device *master; 1824 int master_idx = 0; 1825 1826 /* if dst_dev exists and is enslaved to an L3 device, then 1827 * prefer addresses from dst_dev and then the master over 1828 * any other enslaved devices in the L3 domain. 1829 */ 1830 master = l3mdev_master_dev_rcu(dst_dev); 1831 if (master) { 1832 master_idx = master->ifindex; 1833 1834 hiscore_idx = ipv6_get_saddr_master(net, dst_dev, 1835 master, &dst, 1836 scores, hiscore_idx); 1837 1838 if (scores[hiscore_idx].ifa) 1839 goto out; 1840 } 1841 1842 for_each_netdev_rcu(net, dev) { 1843 /* only consider addresses on devices in the 1844 * same L3 domain 1845 */ 1846 if (l3mdev_master_ifindex_rcu(dev) != master_idx) 1847 continue; 1848 idev = __in6_dev_get(dev); 1849 if (!idev) 1850 continue; 1851 hiscore_idx = __ipv6_dev_get_saddr(net, &dst, idev, scores, hiscore_idx); 1852 } 1853 } 1854 1855 out: 1856 hiscore = &scores[hiscore_idx]; 1857 if (!hiscore->ifa) 1858 ret = -EADDRNOTAVAIL; 1859 else 1860 *saddr = hiscore->ifa->addr; 1861 1862 rcu_read_unlock(); 1863 return ret; 1864 } 1865 EXPORT_SYMBOL(ipv6_dev_get_saddr); 1866 1867 static int __ipv6_get_lladdr(struct inet6_dev *idev, struct in6_addr *addr, 1868 u32 banned_flags) 1869 { 1870 struct inet6_ifaddr *ifp; 1871 int err = -EADDRNOTAVAIL; 1872 1873 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 1874 if (ifp->scope > IFA_LINK) 1875 break; 1876 if (ifp->scope == IFA_LINK && 1877 !(ifp->flags & banned_flags)) { 1878 *addr = ifp->addr; 1879 err = 0; 1880 break; 1881 } 1882 } 1883 return err; 1884 } 1885 1886 int ipv6_get_lladdr(struct net_device *dev, struct in6_addr *addr, 1887 u32 banned_flags) 1888 { 1889 struct inet6_dev *idev; 1890 int err = -EADDRNOTAVAIL; 1891 1892 rcu_read_lock(); 1893 idev = __in6_dev_get(dev); 1894 if (idev) { 1895 read_lock_bh(&idev->lock); 1896 err = __ipv6_get_lladdr(idev, addr, banned_flags); 1897 read_unlock_bh(&idev->lock); 1898 } 1899 rcu_read_unlock(); 1900 return err; 1901 } 1902 1903 static int ipv6_count_addresses(const struct inet6_dev *idev) 1904 { 1905 const struct inet6_ifaddr *ifp; 1906 int cnt = 0; 1907 1908 rcu_read_lock(); 1909 list_for_each_entry_rcu(ifp, &idev->addr_list, if_list) 1910 cnt++; 1911 rcu_read_unlock(); 1912 return cnt; 1913 } 1914 1915 int ipv6_chk_addr(struct net *net, const struct in6_addr *addr, 1916 const struct net_device *dev, int strict) 1917 { 1918 return ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1919 strict, IFA_F_TENTATIVE); 1920 } 1921 EXPORT_SYMBOL(ipv6_chk_addr); 1922 1923 /* device argument is used to find the L3 domain of interest. If 1924 * skip_dev_check is set, then the ifp device is not checked against 1925 * the passed in dev argument. So the 2 cases for addresses checks are: 1926 * 1. does the address exist in the L3 domain that dev is part of 1927 * (skip_dev_check = true), or 1928 * 1929 * 2. does the address exist on the specific device 1930 * (skip_dev_check = false) 1931 */ 1932 static struct net_device * 1933 __ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 1934 const struct net_device *dev, bool skip_dev_check, 1935 int strict, u32 banned_flags) 1936 { 1937 unsigned int hash = inet6_addr_hash(net, addr); 1938 struct net_device *l3mdev, *ndev; 1939 struct inet6_ifaddr *ifp; 1940 u32 ifp_flags; 1941 1942 rcu_read_lock(); 1943 1944 l3mdev = l3mdev_master_dev_rcu(dev); 1945 if (skip_dev_check) 1946 dev = NULL; 1947 1948 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 1949 ndev = ifp->idev->dev; 1950 1951 if (l3mdev_master_dev_rcu(ndev) != l3mdev) 1952 continue; 1953 1954 /* Decouple optimistic from tentative for evaluation here. 1955 * Ban optimistic addresses explicitly, when required. 1956 */ 1957 ifp_flags = (ifp->flags&IFA_F_OPTIMISTIC) 1958 ? (ifp->flags&~IFA_F_TENTATIVE) 1959 : ifp->flags; 1960 if (ipv6_addr_equal(&ifp->addr, addr) && 1961 !(ifp_flags&banned_flags) && 1962 (!dev || ndev == dev || 1963 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict))) { 1964 rcu_read_unlock(); 1965 return ndev; 1966 } 1967 } 1968 1969 rcu_read_unlock(); 1970 return NULL; 1971 } 1972 1973 int ipv6_chk_addr_and_flags(struct net *net, const struct in6_addr *addr, 1974 const struct net_device *dev, bool skip_dev_check, 1975 int strict, u32 banned_flags) 1976 { 1977 return __ipv6_chk_addr_and_flags(net, addr, dev, skip_dev_check, 1978 strict, banned_flags) ? 1 : 0; 1979 } 1980 EXPORT_SYMBOL(ipv6_chk_addr_and_flags); 1981 1982 1983 /* Compares an address/prefix_len with addresses on device @dev. 1984 * If one is found it returns true. 1985 */ 1986 bool ipv6_chk_custom_prefix(const struct in6_addr *addr, 1987 const unsigned int prefix_len, struct net_device *dev) 1988 { 1989 const struct inet6_ifaddr *ifa; 1990 const struct inet6_dev *idev; 1991 bool ret = false; 1992 1993 rcu_read_lock(); 1994 idev = __in6_dev_get(dev); 1995 if (idev) { 1996 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 1997 ret = ipv6_prefix_equal(addr, &ifa->addr, prefix_len); 1998 if (ret) 1999 break; 2000 } 2001 } 2002 rcu_read_unlock(); 2003 2004 return ret; 2005 } 2006 EXPORT_SYMBOL(ipv6_chk_custom_prefix); 2007 2008 int ipv6_chk_prefix(const struct in6_addr *addr, struct net_device *dev) 2009 { 2010 const struct inet6_ifaddr *ifa; 2011 const struct inet6_dev *idev; 2012 int onlink; 2013 2014 onlink = 0; 2015 rcu_read_lock(); 2016 idev = __in6_dev_get(dev); 2017 if (idev) { 2018 list_for_each_entry_rcu(ifa, &idev->addr_list, if_list) { 2019 onlink = ipv6_prefix_equal(addr, &ifa->addr, 2020 ifa->prefix_len); 2021 if (onlink) 2022 break; 2023 } 2024 } 2025 rcu_read_unlock(); 2026 return onlink; 2027 } 2028 EXPORT_SYMBOL(ipv6_chk_prefix); 2029 2030 /** 2031 * ipv6_dev_find - find the first device with a given source address. 2032 * @net: the net namespace 2033 * @addr: the source address 2034 * @dev: used to find the L3 domain of interest 2035 * 2036 * The caller should be protected by RCU, or RTNL. 2037 */ 2038 struct net_device *ipv6_dev_find(struct net *net, const struct in6_addr *addr, 2039 struct net_device *dev) 2040 { 2041 return __ipv6_chk_addr_and_flags(net, addr, dev, !dev, 1, 2042 IFA_F_TENTATIVE); 2043 } 2044 EXPORT_SYMBOL(ipv6_dev_find); 2045 2046 struct inet6_ifaddr *ipv6_get_ifaddr(struct net *net, const struct in6_addr *addr, 2047 struct net_device *dev, int strict) 2048 { 2049 unsigned int hash = inet6_addr_hash(net, addr); 2050 struct inet6_ifaddr *ifp, *result = NULL; 2051 2052 rcu_read_lock(); 2053 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 2054 if (ipv6_addr_equal(&ifp->addr, addr)) { 2055 if (!dev || ifp->idev->dev == dev || 2056 !(ifp->scope&(IFA_LINK|IFA_HOST) || strict)) { 2057 result = ifp; 2058 in6_ifa_hold(ifp); 2059 break; 2060 } 2061 } 2062 } 2063 rcu_read_unlock(); 2064 2065 return result; 2066 } 2067 2068 /* Gets referenced address, destroys ifaddr */ 2069 2070 static void addrconf_dad_stop(struct inet6_ifaddr *ifp, int dad_failed) 2071 { 2072 if (dad_failed) 2073 ifp->flags |= IFA_F_DADFAILED; 2074 2075 if (ifp->flags&IFA_F_TEMPORARY) { 2076 struct inet6_ifaddr *ifpub; 2077 spin_lock_bh(&ifp->lock); 2078 ifpub = ifp->ifpub; 2079 if (ifpub) { 2080 in6_ifa_hold(ifpub); 2081 spin_unlock_bh(&ifp->lock); 2082 ipv6_create_tempaddr(ifpub, true); 2083 in6_ifa_put(ifpub); 2084 } else { 2085 spin_unlock_bh(&ifp->lock); 2086 } 2087 ipv6_del_addr(ifp); 2088 } else if (ifp->flags&IFA_F_PERMANENT || !dad_failed) { 2089 spin_lock_bh(&ifp->lock); 2090 addrconf_del_dad_work(ifp); 2091 ifp->flags |= IFA_F_TENTATIVE; 2092 if (dad_failed) 2093 ifp->flags &= ~IFA_F_OPTIMISTIC; 2094 spin_unlock_bh(&ifp->lock); 2095 if (dad_failed) 2096 ipv6_ifa_notify(0, ifp); 2097 in6_ifa_put(ifp); 2098 } else { 2099 ipv6_del_addr(ifp); 2100 } 2101 } 2102 2103 static int addrconf_dad_end(struct inet6_ifaddr *ifp) 2104 { 2105 int err = -ENOENT; 2106 2107 spin_lock_bh(&ifp->lock); 2108 if (ifp->state == INET6_IFADDR_STATE_DAD) { 2109 ifp->state = INET6_IFADDR_STATE_POSTDAD; 2110 err = 0; 2111 } 2112 spin_unlock_bh(&ifp->lock); 2113 2114 return err; 2115 } 2116 2117 void addrconf_dad_failure(struct sk_buff *skb, struct inet6_ifaddr *ifp) 2118 { 2119 struct inet6_dev *idev = ifp->idev; 2120 struct net *net = dev_net(idev->dev); 2121 2122 if (addrconf_dad_end(ifp)) { 2123 in6_ifa_put(ifp); 2124 return; 2125 } 2126 2127 net_info_ratelimited("%s: IPv6 duplicate address %pI6c used by %pM detected!\n", 2128 ifp->idev->dev->name, &ifp->addr, eth_hdr(skb)->h_source); 2129 2130 spin_lock_bh(&ifp->lock); 2131 2132 if (ifp->flags & IFA_F_STABLE_PRIVACY) { 2133 struct in6_addr new_addr; 2134 struct inet6_ifaddr *ifp2; 2135 int retries = ifp->stable_privacy_retry + 1; 2136 struct ifa6_config cfg = { 2137 .pfx = &new_addr, 2138 .plen = ifp->prefix_len, 2139 .ifa_flags = ifp->flags, 2140 .valid_lft = ifp->valid_lft, 2141 .preferred_lft = ifp->prefered_lft, 2142 .scope = ifp->scope, 2143 }; 2144 2145 if (retries > net->ipv6.sysctl.idgen_retries) { 2146 net_info_ratelimited("%s: privacy stable address generation failed because of DAD conflicts!\n", 2147 ifp->idev->dev->name); 2148 goto errdad; 2149 } 2150 2151 new_addr = ifp->addr; 2152 if (ipv6_generate_stable_address(&new_addr, retries, 2153 idev)) 2154 goto errdad; 2155 2156 spin_unlock_bh(&ifp->lock); 2157 2158 if (idev->cnf.max_addresses && 2159 ipv6_count_addresses(idev) >= 2160 idev->cnf.max_addresses) 2161 goto lock_errdad; 2162 2163 net_info_ratelimited("%s: generating new stable privacy address because of DAD conflict\n", 2164 ifp->idev->dev->name); 2165 2166 ifp2 = ipv6_add_addr(idev, &cfg, false, NULL); 2167 if (IS_ERR(ifp2)) 2168 goto lock_errdad; 2169 2170 spin_lock_bh(&ifp2->lock); 2171 ifp2->stable_privacy_retry = retries; 2172 ifp2->state = INET6_IFADDR_STATE_PREDAD; 2173 spin_unlock_bh(&ifp2->lock); 2174 2175 addrconf_mod_dad_work(ifp2, net->ipv6.sysctl.idgen_delay); 2176 in6_ifa_put(ifp2); 2177 lock_errdad: 2178 spin_lock_bh(&ifp->lock); 2179 } 2180 2181 errdad: 2182 /* transition from _POSTDAD to _ERRDAD */ 2183 ifp->state = INET6_IFADDR_STATE_ERRDAD; 2184 spin_unlock_bh(&ifp->lock); 2185 2186 addrconf_mod_dad_work(ifp, 0); 2187 in6_ifa_put(ifp); 2188 } 2189 2190 /* Join to solicited addr multicast group. 2191 * caller must hold RTNL */ 2192 void addrconf_join_solict(struct net_device *dev, const struct in6_addr *addr) 2193 { 2194 struct in6_addr maddr; 2195 2196 if (dev->flags&(IFF_LOOPBACK|IFF_NOARP)) 2197 return; 2198 2199 addrconf_addr_solict_mult(addr, &maddr); 2200 ipv6_dev_mc_inc(dev, &maddr); 2201 } 2202 2203 /* caller must hold RTNL */ 2204 void addrconf_leave_solict(struct inet6_dev *idev, const struct in6_addr *addr) 2205 { 2206 struct in6_addr maddr; 2207 2208 if (idev->dev->flags&(IFF_LOOPBACK|IFF_NOARP)) 2209 return; 2210 2211 addrconf_addr_solict_mult(addr, &maddr); 2212 __ipv6_dev_mc_dec(idev, &maddr); 2213 } 2214 2215 /* caller must hold RTNL */ 2216 static void addrconf_join_anycast(struct inet6_ifaddr *ifp) 2217 { 2218 struct in6_addr addr; 2219 2220 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2221 return; 2222 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2223 if (ipv6_addr_any(&addr)) 2224 return; 2225 __ipv6_dev_ac_inc(ifp->idev, &addr); 2226 } 2227 2228 /* caller must hold RTNL */ 2229 static void addrconf_leave_anycast(struct inet6_ifaddr *ifp) 2230 { 2231 struct in6_addr addr; 2232 2233 if (ifp->prefix_len >= 127) /* RFC 6164 */ 2234 return; 2235 ipv6_addr_prefix(&addr, &ifp->addr, ifp->prefix_len); 2236 if (ipv6_addr_any(&addr)) 2237 return; 2238 __ipv6_dev_ac_dec(ifp->idev, &addr); 2239 } 2240 2241 static int addrconf_ifid_6lowpan(u8 *eui, struct net_device *dev) 2242 { 2243 switch (dev->addr_len) { 2244 case ETH_ALEN: 2245 memcpy(eui, dev->dev_addr, 3); 2246 eui[3] = 0xFF; 2247 eui[4] = 0xFE; 2248 memcpy(eui + 5, dev->dev_addr + 3, 3); 2249 break; 2250 case EUI64_ADDR_LEN: 2251 memcpy(eui, dev->dev_addr, EUI64_ADDR_LEN); 2252 eui[0] ^= 2; 2253 break; 2254 default: 2255 return -1; 2256 } 2257 2258 return 0; 2259 } 2260 2261 static int addrconf_ifid_ieee1394(u8 *eui, struct net_device *dev) 2262 { 2263 const union fwnet_hwaddr *ha; 2264 2265 if (dev->addr_len != FWNET_ALEN) 2266 return -1; 2267 2268 ha = (const union fwnet_hwaddr *)dev->dev_addr; 2269 2270 memcpy(eui, &ha->uc.uniq_id, sizeof(ha->uc.uniq_id)); 2271 eui[0] ^= 2; 2272 return 0; 2273 } 2274 2275 static int addrconf_ifid_arcnet(u8 *eui, struct net_device *dev) 2276 { 2277 /* XXX: inherit EUI-64 from other interface -- yoshfuji */ 2278 if (dev->addr_len != ARCNET_ALEN) 2279 return -1; 2280 memset(eui, 0, 7); 2281 eui[7] = *(u8 *)dev->dev_addr; 2282 return 0; 2283 } 2284 2285 static int addrconf_ifid_infiniband(u8 *eui, struct net_device *dev) 2286 { 2287 if (dev->addr_len != INFINIBAND_ALEN) 2288 return -1; 2289 memcpy(eui, dev->dev_addr + 12, 8); 2290 eui[0] |= 2; 2291 return 0; 2292 } 2293 2294 static int __ipv6_isatap_ifid(u8 *eui, __be32 addr) 2295 { 2296 if (addr == 0) 2297 return -1; 2298 eui[0] = (ipv4_is_zeronet(addr) || ipv4_is_private_10(addr) || 2299 ipv4_is_loopback(addr) || ipv4_is_linklocal_169(addr) || 2300 ipv4_is_private_172(addr) || ipv4_is_test_192(addr) || 2301 ipv4_is_anycast_6to4(addr) || ipv4_is_private_192(addr) || 2302 ipv4_is_test_198(addr) || ipv4_is_multicast(addr) || 2303 ipv4_is_lbcast(addr)) ? 0x00 : 0x02; 2304 eui[1] = 0; 2305 eui[2] = 0x5E; 2306 eui[3] = 0xFE; 2307 memcpy(eui + 4, &addr, 4); 2308 return 0; 2309 } 2310 2311 static int addrconf_ifid_sit(u8 *eui, struct net_device *dev) 2312 { 2313 if (dev->priv_flags & IFF_ISATAP) 2314 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2315 return -1; 2316 } 2317 2318 static int addrconf_ifid_gre(u8 *eui, struct net_device *dev) 2319 { 2320 return __ipv6_isatap_ifid(eui, *(__be32 *)dev->dev_addr); 2321 } 2322 2323 static int addrconf_ifid_ip6tnl(u8 *eui, struct net_device *dev) 2324 { 2325 memcpy(eui, dev->perm_addr, 3); 2326 memcpy(eui + 5, dev->perm_addr + 3, 3); 2327 eui[3] = 0xFF; 2328 eui[4] = 0xFE; 2329 eui[0] ^= 2; 2330 return 0; 2331 } 2332 2333 static int ipv6_generate_eui64(u8 *eui, struct net_device *dev) 2334 { 2335 switch (dev->type) { 2336 case ARPHRD_ETHER: 2337 case ARPHRD_FDDI: 2338 return addrconf_ifid_eui48(eui, dev); 2339 case ARPHRD_ARCNET: 2340 return addrconf_ifid_arcnet(eui, dev); 2341 case ARPHRD_INFINIBAND: 2342 return addrconf_ifid_infiniband(eui, dev); 2343 case ARPHRD_SIT: 2344 return addrconf_ifid_sit(eui, dev); 2345 case ARPHRD_IPGRE: 2346 case ARPHRD_TUNNEL: 2347 return addrconf_ifid_gre(eui, dev); 2348 case ARPHRD_6LOWPAN: 2349 return addrconf_ifid_6lowpan(eui, dev); 2350 case ARPHRD_IEEE1394: 2351 return addrconf_ifid_ieee1394(eui, dev); 2352 case ARPHRD_TUNNEL6: 2353 case ARPHRD_IP6GRE: 2354 case ARPHRD_RAWIP: 2355 return addrconf_ifid_ip6tnl(eui, dev); 2356 } 2357 return -1; 2358 } 2359 2360 static int ipv6_inherit_eui64(u8 *eui, struct inet6_dev *idev) 2361 { 2362 int err = -1; 2363 struct inet6_ifaddr *ifp; 2364 2365 read_lock_bh(&idev->lock); 2366 list_for_each_entry_reverse(ifp, &idev->addr_list, if_list) { 2367 if (ifp->scope > IFA_LINK) 2368 break; 2369 if (ifp->scope == IFA_LINK && !(ifp->flags&IFA_F_TENTATIVE)) { 2370 memcpy(eui, ifp->addr.s6_addr+8, 8); 2371 err = 0; 2372 break; 2373 } 2374 } 2375 read_unlock_bh(&idev->lock); 2376 return err; 2377 } 2378 2379 /* Generation of a randomized Interface Identifier 2380 * draft-ietf-6man-rfc4941bis, Section 3.3.1 2381 */ 2382 2383 static void ipv6_gen_rnd_iid(struct in6_addr *addr) 2384 { 2385 regen: 2386 get_random_bytes(&addr->s6_addr[8], 8); 2387 2388 /* <draft-ietf-6man-rfc4941bis-08.txt>, Section 3.3.1: 2389 * check if generated address is not inappropriate: 2390 * 2391 * - Reserved IPv6 Interface Identifiers 2392 * - XXX: already assigned to an address on the device 2393 */ 2394 2395 /* Subnet-router anycast: 0000:0000:0000:0000 */ 2396 if (!(addr->s6_addr32[2] | addr->s6_addr32[3])) 2397 goto regen; 2398 2399 /* IANA Ethernet block: 0200:5EFF:FE00:0000-0200:5EFF:FE00:5212 2400 * Proxy Mobile IPv6: 0200:5EFF:FE00:5213 2401 * IANA Ethernet block: 0200:5EFF:FE00:5214-0200:5EFF:FEFF:FFFF 2402 */ 2403 if (ntohl(addr->s6_addr32[2]) == 0x02005eff && 2404 (ntohl(addr->s6_addr32[3]) & 0Xff000000) == 0xfe000000) 2405 goto regen; 2406 2407 /* Reserved subnet anycast addresses */ 2408 if (ntohl(addr->s6_addr32[2]) == 0xfdffffff && 2409 ntohl(addr->s6_addr32[3]) >= 0Xffffff80) 2410 goto regen; 2411 } 2412 2413 /* 2414 * Add prefix route. 2415 */ 2416 2417 static void 2418 addrconf_prefix_route(struct in6_addr *pfx, int plen, u32 metric, 2419 struct net_device *dev, unsigned long expires, 2420 u32 flags, gfp_t gfp_flags) 2421 { 2422 struct fib6_config cfg = { 2423 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX, 2424 .fc_metric = metric ? : IP6_RT_PRIO_ADDRCONF, 2425 .fc_ifindex = dev->ifindex, 2426 .fc_expires = expires, 2427 .fc_dst_len = plen, 2428 .fc_flags = RTF_UP | flags, 2429 .fc_nlinfo.nl_net = dev_net(dev), 2430 .fc_protocol = RTPROT_KERNEL, 2431 .fc_type = RTN_UNICAST, 2432 }; 2433 2434 cfg.fc_dst = *pfx; 2435 2436 /* Prevent useless cloning on PtP SIT. 2437 This thing is done here expecting that the whole 2438 class of non-broadcast devices need not cloning. 2439 */ 2440 #if IS_ENABLED(CONFIG_IPV6_SIT) 2441 if (dev->type == ARPHRD_SIT && (dev->flags & IFF_POINTOPOINT)) 2442 cfg.fc_flags |= RTF_NONEXTHOP; 2443 #endif 2444 2445 ip6_route_add(&cfg, gfp_flags, NULL); 2446 } 2447 2448 2449 static struct fib6_info *addrconf_get_prefix_route(const struct in6_addr *pfx, 2450 int plen, 2451 const struct net_device *dev, 2452 u32 flags, u32 noflags, 2453 bool no_gw) 2454 { 2455 struct fib6_node *fn; 2456 struct fib6_info *rt = NULL; 2457 struct fib6_table *table; 2458 u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_PREFIX; 2459 2460 table = fib6_get_table(dev_net(dev), tb_id); 2461 if (!table) 2462 return NULL; 2463 2464 rcu_read_lock(); 2465 fn = fib6_locate(&table->tb6_root, pfx, plen, NULL, 0, true); 2466 if (!fn) 2467 goto out; 2468 2469 for_each_fib6_node_rt_rcu(fn) { 2470 /* prefix routes only use builtin fib6_nh */ 2471 if (rt->nh) 2472 continue; 2473 2474 if (rt->fib6_nh->fib_nh_dev->ifindex != dev->ifindex) 2475 continue; 2476 if (no_gw && rt->fib6_nh->fib_nh_gw_family) 2477 continue; 2478 if ((rt->fib6_flags & flags) != flags) 2479 continue; 2480 if ((rt->fib6_flags & noflags) != 0) 2481 continue; 2482 if (!fib6_info_hold_safe(rt)) 2483 continue; 2484 break; 2485 } 2486 out: 2487 rcu_read_unlock(); 2488 return rt; 2489 } 2490 2491 2492 /* Create "default" multicast route to the interface */ 2493 2494 static void addrconf_add_mroute(struct net_device *dev) 2495 { 2496 struct fib6_config cfg = { 2497 .fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_LOCAL, 2498 .fc_metric = IP6_RT_PRIO_ADDRCONF, 2499 .fc_ifindex = dev->ifindex, 2500 .fc_dst_len = 8, 2501 .fc_flags = RTF_UP, 2502 .fc_type = RTN_MULTICAST, 2503 .fc_nlinfo.nl_net = dev_net(dev), 2504 .fc_protocol = RTPROT_KERNEL, 2505 }; 2506 2507 ipv6_addr_set(&cfg.fc_dst, htonl(0xFF000000), 0, 0, 0); 2508 2509 ip6_route_add(&cfg, GFP_KERNEL, NULL); 2510 } 2511 2512 static struct inet6_dev *addrconf_add_dev(struct net_device *dev) 2513 { 2514 struct inet6_dev *idev; 2515 2516 ASSERT_RTNL(); 2517 2518 idev = ipv6_find_idev(dev); 2519 if (IS_ERR(idev)) 2520 return idev; 2521 2522 if (idev->cnf.disable_ipv6) 2523 return ERR_PTR(-EACCES); 2524 2525 /* Add default multicast route */ 2526 if (!(dev->flags & IFF_LOOPBACK) && !netif_is_l3_master(dev)) 2527 addrconf_add_mroute(dev); 2528 2529 return idev; 2530 } 2531 2532 static void manage_tempaddrs(struct inet6_dev *idev, 2533 struct inet6_ifaddr *ifp, 2534 __u32 valid_lft, __u32 prefered_lft, 2535 bool create, unsigned long now) 2536 { 2537 u32 flags; 2538 struct inet6_ifaddr *ift; 2539 2540 read_lock_bh(&idev->lock); 2541 /* update all temporary addresses in the list */ 2542 list_for_each_entry(ift, &idev->tempaddr_list, tmp_list) { 2543 int age, max_valid, max_prefered; 2544 2545 if (ifp != ift->ifpub) 2546 continue; 2547 2548 /* RFC 4941 section 3.3: 2549 * If a received option will extend the lifetime of a public 2550 * address, the lifetimes of temporary addresses should 2551 * be extended, subject to the overall constraint that no 2552 * temporary addresses should ever remain "valid" or "preferred" 2553 * for a time longer than (TEMP_VALID_LIFETIME) or 2554 * (TEMP_PREFERRED_LIFETIME - DESYNC_FACTOR), respectively. 2555 */ 2556 age = (now - ift->cstamp) / HZ; 2557 max_valid = idev->cnf.temp_valid_lft - age; 2558 if (max_valid < 0) 2559 max_valid = 0; 2560 2561 max_prefered = idev->cnf.temp_prefered_lft - 2562 idev->desync_factor - age; 2563 if (max_prefered < 0) 2564 max_prefered = 0; 2565 2566 if (valid_lft > max_valid) 2567 valid_lft = max_valid; 2568 2569 if (prefered_lft > max_prefered) 2570 prefered_lft = max_prefered; 2571 2572 spin_lock(&ift->lock); 2573 flags = ift->flags; 2574 ift->valid_lft = valid_lft; 2575 ift->prefered_lft = prefered_lft; 2576 ift->tstamp = now; 2577 if (prefered_lft > 0) 2578 ift->flags &= ~IFA_F_DEPRECATED; 2579 2580 spin_unlock(&ift->lock); 2581 if (!(flags&IFA_F_TENTATIVE)) 2582 ipv6_ifa_notify(0, ift); 2583 } 2584 2585 /* Also create a temporary address if it's enabled but no temporary 2586 * address currently exists. 2587 * However, we get called with valid_lft == 0, prefered_lft == 0, create == false 2588 * as part of cleanup (ie. deleting the mngtmpaddr). 2589 * We don't want that to result in creating a new temporary ip address. 2590 */ 2591 if (list_empty(&idev->tempaddr_list) && (valid_lft || prefered_lft)) 2592 create = true; 2593 2594 if (create && idev->cnf.use_tempaddr > 0) { 2595 /* When a new public address is created as described 2596 * in [ADDRCONF], also create a new temporary address. 2597 */ 2598 read_unlock_bh(&idev->lock); 2599 ipv6_create_tempaddr(ifp, false); 2600 } else { 2601 read_unlock_bh(&idev->lock); 2602 } 2603 } 2604 2605 static bool is_addr_mode_generate_stable(struct inet6_dev *idev) 2606 { 2607 return idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY || 2608 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_RANDOM; 2609 } 2610 2611 int addrconf_prefix_rcv_add_addr(struct net *net, struct net_device *dev, 2612 const struct prefix_info *pinfo, 2613 struct inet6_dev *in6_dev, 2614 const struct in6_addr *addr, int addr_type, 2615 u32 addr_flags, bool sllao, bool tokenized, 2616 __u32 valid_lft, u32 prefered_lft) 2617 { 2618 struct inet6_ifaddr *ifp = ipv6_get_ifaddr(net, addr, dev, 1); 2619 int create = 0, update_lft = 0; 2620 2621 if (!ifp && valid_lft) { 2622 int max_addresses = in6_dev->cnf.max_addresses; 2623 struct ifa6_config cfg = { 2624 .pfx = addr, 2625 .plen = pinfo->prefix_len, 2626 .ifa_flags = addr_flags, 2627 .valid_lft = valid_lft, 2628 .preferred_lft = prefered_lft, 2629 .scope = addr_type & IPV6_ADDR_SCOPE_MASK, 2630 .ifa_proto = IFAPROT_KERNEL_RA 2631 }; 2632 2633 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 2634 if ((net->ipv6.devconf_all->optimistic_dad || 2635 in6_dev->cnf.optimistic_dad) && 2636 !net->ipv6.devconf_all->forwarding && sllao) 2637 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 2638 #endif 2639 2640 /* Do not allow to create too much of autoconfigured 2641 * addresses; this would be too easy way to crash kernel. 2642 */ 2643 if (!max_addresses || 2644 ipv6_count_addresses(in6_dev) < max_addresses) 2645 ifp = ipv6_add_addr(in6_dev, &cfg, false, NULL); 2646 2647 if (IS_ERR_OR_NULL(ifp)) 2648 return -1; 2649 2650 create = 1; 2651 spin_lock_bh(&ifp->lock); 2652 ifp->flags |= IFA_F_MANAGETEMPADDR; 2653 ifp->cstamp = jiffies; 2654 ifp->tokenized = tokenized; 2655 spin_unlock_bh(&ifp->lock); 2656 addrconf_dad_start(ifp); 2657 } 2658 2659 if (ifp) { 2660 u32 flags; 2661 unsigned long now; 2662 u32 stored_lft; 2663 2664 /* update lifetime (RFC2462 5.5.3 e) */ 2665 spin_lock_bh(&ifp->lock); 2666 now = jiffies; 2667 if (ifp->valid_lft > (now - ifp->tstamp) / HZ) 2668 stored_lft = ifp->valid_lft - (now - ifp->tstamp) / HZ; 2669 else 2670 stored_lft = 0; 2671 2672 /* RFC4862 Section 5.5.3e: 2673 * "Note that the preferred lifetime of the 2674 * corresponding address is always reset to 2675 * the Preferred Lifetime in the received 2676 * Prefix Information option, regardless of 2677 * whether the valid lifetime is also reset or 2678 * ignored." 2679 * 2680 * So we should always update prefered_lft here. 2681 */ 2682 update_lft = !create && stored_lft; 2683 2684 if (update_lft && !in6_dev->cnf.ra_honor_pio_life) { 2685 const u32 minimum_lft = min_t(u32, 2686 stored_lft, MIN_VALID_LIFETIME); 2687 valid_lft = max(valid_lft, minimum_lft); 2688 } 2689 2690 if (update_lft) { 2691 ifp->valid_lft = valid_lft; 2692 ifp->prefered_lft = prefered_lft; 2693 ifp->tstamp = now; 2694 flags = ifp->flags; 2695 ifp->flags &= ~IFA_F_DEPRECATED; 2696 spin_unlock_bh(&ifp->lock); 2697 2698 if (!(flags&IFA_F_TENTATIVE)) 2699 ipv6_ifa_notify(0, ifp); 2700 } else 2701 spin_unlock_bh(&ifp->lock); 2702 2703 manage_tempaddrs(in6_dev, ifp, valid_lft, prefered_lft, 2704 create, now); 2705 2706 in6_ifa_put(ifp); 2707 addrconf_verify(net); 2708 } 2709 2710 return 0; 2711 } 2712 EXPORT_SYMBOL_GPL(addrconf_prefix_rcv_add_addr); 2713 2714 void addrconf_prefix_rcv(struct net_device *dev, u8 *opt, int len, bool sllao) 2715 { 2716 struct prefix_info *pinfo; 2717 __u32 valid_lft; 2718 __u32 prefered_lft; 2719 int addr_type, err; 2720 u32 addr_flags = 0; 2721 struct inet6_dev *in6_dev; 2722 struct net *net = dev_net(dev); 2723 2724 pinfo = (struct prefix_info *) opt; 2725 2726 if (len < sizeof(struct prefix_info)) { 2727 netdev_dbg(dev, "addrconf: prefix option too short\n"); 2728 return; 2729 } 2730 2731 /* 2732 * Validation checks ([ADDRCONF], page 19) 2733 */ 2734 2735 addr_type = ipv6_addr_type(&pinfo->prefix); 2736 2737 if (addr_type & (IPV6_ADDR_MULTICAST|IPV6_ADDR_LINKLOCAL)) 2738 return; 2739 2740 valid_lft = ntohl(pinfo->valid); 2741 prefered_lft = ntohl(pinfo->prefered); 2742 2743 if (prefered_lft > valid_lft) { 2744 net_warn_ratelimited("addrconf: prefix option has invalid lifetime\n"); 2745 return; 2746 } 2747 2748 in6_dev = in6_dev_get(dev); 2749 2750 if (!in6_dev) { 2751 net_dbg_ratelimited("addrconf: device %s not configured\n", 2752 dev->name); 2753 return; 2754 } 2755 2756 if (valid_lft != 0 && valid_lft < in6_dev->cnf.accept_ra_min_lft) 2757 goto put; 2758 2759 /* 2760 * Two things going on here: 2761 * 1) Add routes for on-link prefixes 2762 * 2) Configure prefixes with the auto flag set 2763 */ 2764 2765 if (pinfo->onlink) { 2766 struct fib6_info *rt; 2767 unsigned long rt_expires; 2768 2769 /* Avoid arithmetic overflow. Really, we could 2770 * save rt_expires in seconds, likely valid_lft, 2771 * but it would require division in fib gc, that it 2772 * not good. 2773 */ 2774 if (HZ > USER_HZ) 2775 rt_expires = addrconf_timeout_fixup(valid_lft, HZ); 2776 else 2777 rt_expires = addrconf_timeout_fixup(valid_lft, USER_HZ); 2778 2779 if (addrconf_finite_timeout(rt_expires)) 2780 rt_expires *= HZ; 2781 2782 rt = addrconf_get_prefix_route(&pinfo->prefix, 2783 pinfo->prefix_len, 2784 dev, 2785 RTF_ADDRCONF | RTF_PREFIX_RT, 2786 RTF_DEFAULT, true); 2787 2788 if (rt) { 2789 /* Autoconf prefix route */ 2790 if (valid_lft == 0) { 2791 ip6_del_rt(net, rt, false); 2792 rt = NULL; 2793 } else if (addrconf_finite_timeout(rt_expires)) { 2794 /* not infinity */ 2795 fib6_set_expires(rt, jiffies + rt_expires); 2796 } else { 2797 fib6_clean_expires(rt); 2798 } 2799 } else if (valid_lft) { 2800 clock_t expires = 0; 2801 int flags = RTF_ADDRCONF | RTF_PREFIX_RT; 2802 if (addrconf_finite_timeout(rt_expires)) { 2803 /* not infinity */ 2804 flags |= RTF_EXPIRES; 2805 expires = jiffies_to_clock_t(rt_expires); 2806 } 2807 addrconf_prefix_route(&pinfo->prefix, pinfo->prefix_len, 2808 0, dev, expires, flags, 2809 GFP_ATOMIC); 2810 } 2811 fib6_info_release(rt); 2812 } 2813 2814 /* Try to figure out our local address for this prefix */ 2815 2816 if (pinfo->autoconf && in6_dev->cnf.autoconf) { 2817 struct in6_addr addr; 2818 bool tokenized = false, dev_addr_generated = false; 2819 2820 if (pinfo->prefix_len == 64) { 2821 memcpy(&addr, &pinfo->prefix, 8); 2822 2823 if (!ipv6_addr_any(&in6_dev->token)) { 2824 read_lock_bh(&in6_dev->lock); 2825 memcpy(addr.s6_addr + 8, 2826 in6_dev->token.s6_addr + 8, 8); 2827 read_unlock_bh(&in6_dev->lock); 2828 tokenized = true; 2829 } else if (is_addr_mode_generate_stable(in6_dev) && 2830 !ipv6_generate_stable_address(&addr, 0, 2831 in6_dev)) { 2832 addr_flags |= IFA_F_STABLE_PRIVACY; 2833 goto ok; 2834 } else if (ipv6_generate_eui64(addr.s6_addr + 8, dev) && 2835 ipv6_inherit_eui64(addr.s6_addr + 8, in6_dev)) { 2836 goto put; 2837 } else { 2838 dev_addr_generated = true; 2839 } 2840 goto ok; 2841 } 2842 net_dbg_ratelimited("IPv6 addrconf: prefix with wrong length %d\n", 2843 pinfo->prefix_len); 2844 goto put; 2845 2846 ok: 2847 err = addrconf_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, 2848 &addr, addr_type, 2849 addr_flags, sllao, 2850 tokenized, valid_lft, 2851 prefered_lft); 2852 if (err) 2853 goto put; 2854 2855 /* Ignore error case here because previous prefix add addr was 2856 * successful which will be notified. 2857 */ 2858 ndisc_ops_prefix_rcv_add_addr(net, dev, pinfo, in6_dev, &addr, 2859 addr_type, addr_flags, sllao, 2860 tokenized, valid_lft, 2861 prefered_lft, 2862 dev_addr_generated); 2863 } 2864 inet6_prefix_notify(RTM_NEWPREFIX, in6_dev, pinfo); 2865 put: 2866 in6_dev_put(in6_dev); 2867 } 2868 2869 static int addrconf_set_sit_dstaddr(struct net *net, struct net_device *dev, 2870 struct in6_ifreq *ireq) 2871 { 2872 struct ip_tunnel_parm p = { }; 2873 int err; 2874 2875 if (!(ipv6_addr_type(&ireq->ifr6_addr) & IPV6_ADDR_COMPATv4)) 2876 return -EADDRNOTAVAIL; 2877 2878 p.iph.daddr = ireq->ifr6_addr.s6_addr32[3]; 2879 p.iph.version = 4; 2880 p.iph.ihl = 5; 2881 p.iph.protocol = IPPROTO_IPV6; 2882 p.iph.ttl = 64; 2883 2884 if (!dev->netdev_ops->ndo_tunnel_ctl) 2885 return -EOPNOTSUPP; 2886 err = dev->netdev_ops->ndo_tunnel_ctl(dev, &p, SIOCADDTUNNEL); 2887 if (err) 2888 return err; 2889 2890 dev = __dev_get_by_name(net, p.name); 2891 if (!dev) 2892 return -ENOBUFS; 2893 return dev_open(dev, NULL); 2894 } 2895 2896 /* 2897 * Set destination address. 2898 * Special case for SIT interfaces where we create a new "virtual" 2899 * device. 2900 */ 2901 int addrconf_set_dstaddr(struct net *net, void __user *arg) 2902 { 2903 struct net_device *dev; 2904 struct in6_ifreq ireq; 2905 int err = -ENODEV; 2906 2907 if (!IS_ENABLED(CONFIG_IPV6_SIT)) 2908 return -ENODEV; 2909 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 2910 return -EFAULT; 2911 2912 rtnl_lock(); 2913 dev = __dev_get_by_index(net, ireq.ifr6_ifindex); 2914 if (dev && dev->type == ARPHRD_SIT) 2915 err = addrconf_set_sit_dstaddr(net, dev, &ireq); 2916 rtnl_unlock(); 2917 return err; 2918 } 2919 2920 static int ipv6_mc_config(struct sock *sk, bool join, 2921 const struct in6_addr *addr, int ifindex) 2922 { 2923 int ret; 2924 2925 ASSERT_RTNL(); 2926 2927 lock_sock(sk); 2928 if (join) 2929 ret = ipv6_sock_mc_join(sk, ifindex, addr); 2930 else 2931 ret = ipv6_sock_mc_drop(sk, ifindex, addr); 2932 release_sock(sk); 2933 2934 return ret; 2935 } 2936 2937 /* 2938 * Manual configuration of address on an interface 2939 */ 2940 static int inet6_addr_add(struct net *net, int ifindex, 2941 struct ifa6_config *cfg, 2942 struct netlink_ext_ack *extack) 2943 { 2944 struct inet6_ifaddr *ifp; 2945 struct inet6_dev *idev; 2946 struct net_device *dev; 2947 unsigned long timeout; 2948 clock_t expires; 2949 u32 flags; 2950 2951 ASSERT_RTNL(); 2952 2953 if (cfg->plen > 128) { 2954 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 2955 return -EINVAL; 2956 } 2957 2958 /* check the lifetime */ 2959 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) { 2960 NL_SET_ERR_MSG_MOD(extack, "address lifetime invalid"); 2961 return -EINVAL; 2962 } 2963 2964 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && cfg->plen != 64) { 2965 NL_SET_ERR_MSG_MOD(extack, "address with \"mngtmpaddr\" flag must have a prefix length of 64"); 2966 return -EINVAL; 2967 } 2968 2969 dev = __dev_get_by_index(net, ifindex); 2970 if (!dev) 2971 return -ENODEV; 2972 2973 idev = addrconf_add_dev(dev); 2974 if (IS_ERR(idev)) { 2975 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 2976 return PTR_ERR(idev); 2977 } 2978 2979 if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 2980 int ret = ipv6_mc_config(net->ipv6.mc_autojoin_sk, 2981 true, cfg->pfx, ifindex); 2982 2983 if (ret < 0) { 2984 NL_SET_ERR_MSG_MOD(extack, "Multicast auto join failed"); 2985 return ret; 2986 } 2987 } 2988 2989 cfg->scope = ipv6_addr_scope(cfg->pfx); 2990 2991 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ); 2992 if (addrconf_finite_timeout(timeout)) { 2993 expires = jiffies_to_clock_t(timeout * HZ); 2994 cfg->valid_lft = timeout; 2995 flags = RTF_EXPIRES; 2996 } else { 2997 expires = 0; 2998 flags = 0; 2999 cfg->ifa_flags |= IFA_F_PERMANENT; 3000 } 3001 3002 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ); 3003 if (addrconf_finite_timeout(timeout)) { 3004 if (timeout == 0) 3005 cfg->ifa_flags |= IFA_F_DEPRECATED; 3006 cfg->preferred_lft = timeout; 3007 } 3008 3009 ifp = ipv6_add_addr(idev, cfg, true, extack); 3010 if (!IS_ERR(ifp)) { 3011 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 3012 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3013 ifp->rt_priority, dev, expires, 3014 flags, GFP_KERNEL); 3015 } 3016 3017 /* Send a netlink notification if DAD is enabled and 3018 * optimistic flag is not set 3019 */ 3020 if (!(ifp->flags & (IFA_F_OPTIMISTIC | IFA_F_NODAD))) 3021 ipv6_ifa_notify(0, ifp); 3022 /* 3023 * Note that section 3.1 of RFC 4429 indicates 3024 * that the Optimistic flag should not be set for 3025 * manually configured addresses 3026 */ 3027 addrconf_dad_start(ifp); 3028 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR) 3029 manage_tempaddrs(idev, ifp, cfg->valid_lft, 3030 cfg->preferred_lft, true, jiffies); 3031 in6_ifa_put(ifp); 3032 addrconf_verify_rtnl(net); 3033 return 0; 3034 } else if (cfg->ifa_flags & IFA_F_MCAUTOJOIN) { 3035 ipv6_mc_config(net->ipv6.mc_autojoin_sk, false, 3036 cfg->pfx, ifindex); 3037 } 3038 3039 return PTR_ERR(ifp); 3040 } 3041 3042 static int inet6_addr_del(struct net *net, int ifindex, u32 ifa_flags, 3043 const struct in6_addr *pfx, unsigned int plen, 3044 struct netlink_ext_ack *extack) 3045 { 3046 struct inet6_ifaddr *ifp; 3047 struct inet6_dev *idev; 3048 struct net_device *dev; 3049 3050 if (plen > 128) { 3051 NL_SET_ERR_MSG_MOD(extack, "Invalid prefix length"); 3052 return -EINVAL; 3053 } 3054 3055 dev = __dev_get_by_index(net, ifindex); 3056 if (!dev) { 3057 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 3058 return -ENODEV; 3059 } 3060 3061 idev = __in6_dev_get(dev); 3062 if (!idev) { 3063 NL_SET_ERR_MSG_MOD(extack, "IPv6 is disabled on this device"); 3064 return -ENXIO; 3065 } 3066 3067 read_lock_bh(&idev->lock); 3068 list_for_each_entry(ifp, &idev->addr_list, if_list) { 3069 if (ifp->prefix_len == plen && 3070 ipv6_addr_equal(pfx, &ifp->addr)) { 3071 in6_ifa_hold(ifp); 3072 read_unlock_bh(&idev->lock); 3073 3074 if (!(ifp->flags & IFA_F_TEMPORARY) && 3075 (ifa_flags & IFA_F_MANAGETEMPADDR)) 3076 manage_tempaddrs(idev, ifp, 0, 0, false, 3077 jiffies); 3078 ipv6_del_addr(ifp); 3079 addrconf_verify_rtnl(net); 3080 if (ipv6_addr_is_multicast(pfx)) { 3081 ipv6_mc_config(net->ipv6.mc_autojoin_sk, 3082 false, pfx, dev->ifindex); 3083 } 3084 return 0; 3085 } 3086 } 3087 read_unlock_bh(&idev->lock); 3088 3089 NL_SET_ERR_MSG_MOD(extack, "address not found"); 3090 return -EADDRNOTAVAIL; 3091 } 3092 3093 3094 int addrconf_add_ifaddr(struct net *net, void __user *arg) 3095 { 3096 struct ifa6_config cfg = { 3097 .ifa_flags = IFA_F_PERMANENT, 3098 .preferred_lft = INFINITY_LIFE_TIME, 3099 .valid_lft = INFINITY_LIFE_TIME, 3100 }; 3101 struct in6_ifreq ireq; 3102 int err; 3103 3104 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 3105 return -EPERM; 3106 3107 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3108 return -EFAULT; 3109 3110 cfg.pfx = &ireq.ifr6_addr; 3111 cfg.plen = ireq.ifr6_prefixlen; 3112 3113 rtnl_lock(); 3114 err = inet6_addr_add(net, ireq.ifr6_ifindex, &cfg, NULL); 3115 rtnl_unlock(); 3116 return err; 3117 } 3118 3119 int addrconf_del_ifaddr(struct net *net, void __user *arg) 3120 { 3121 struct in6_ifreq ireq; 3122 int err; 3123 3124 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 3125 return -EPERM; 3126 3127 if (copy_from_user(&ireq, arg, sizeof(struct in6_ifreq))) 3128 return -EFAULT; 3129 3130 rtnl_lock(); 3131 err = inet6_addr_del(net, ireq.ifr6_ifindex, 0, &ireq.ifr6_addr, 3132 ireq.ifr6_prefixlen, NULL); 3133 rtnl_unlock(); 3134 return err; 3135 } 3136 3137 static void add_addr(struct inet6_dev *idev, const struct in6_addr *addr, 3138 int plen, int scope, u8 proto) 3139 { 3140 struct inet6_ifaddr *ifp; 3141 struct ifa6_config cfg = { 3142 .pfx = addr, 3143 .plen = plen, 3144 .ifa_flags = IFA_F_PERMANENT, 3145 .valid_lft = INFINITY_LIFE_TIME, 3146 .preferred_lft = INFINITY_LIFE_TIME, 3147 .scope = scope, 3148 .ifa_proto = proto 3149 }; 3150 3151 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3152 if (!IS_ERR(ifp)) { 3153 spin_lock_bh(&ifp->lock); 3154 ifp->flags &= ~IFA_F_TENTATIVE; 3155 spin_unlock_bh(&ifp->lock); 3156 rt_genid_bump_ipv6(dev_net(idev->dev)); 3157 ipv6_ifa_notify(RTM_NEWADDR, ifp); 3158 in6_ifa_put(ifp); 3159 } 3160 } 3161 3162 #if IS_ENABLED(CONFIG_IPV6_SIT) || IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE) 3163 static void add_v4_addrs(struct inet6_dev *idev) 3164 { 3165 struct in6_addr addr; 3166 struct net_device *dev; 3167 struct net *net = dev_net(idev->dev); 3168 int scope, plen, offset = 0; 3169 u32 pflags = 0; 3170 3171 ASSERT_RTNL(); 3172 3173 memset(&addr, 0, sizeof(struct in6_addr)); 3174 /* in case of IP6GRE the dev_addr is an IPv6 and therefore we use only the last 4 bytes */ 3175 if (idev->dev->addr_len == sizeof(struct in6_addr)) 3176 offset = sizeof(struct in6_addr) - 4; 3177 memcpy(&addr.s6_addr32[3], idev->dev->dev_addr + offset, 4); 3178 3179 if (!(idev->dev->flags & IFF_POINTOPOINT) && idev->dev->type == ARPHRD_SIT) { 3180 scope = IPV6_ADDR_COMPATv4; 3181 plen = 96; 3182 pflags |= RTF_NONEXTHOP; 3183 } else { 3184 if (idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_NONE) 3185 return; 3186 3187 addr.s6_addr32[0] = htonl(0xfe800000); 3188 scope = IFA_LINK; 3189 plen = 64; 3190 } 3191 3192 if (addr.s6_addr32[3]) { 3193 add_addr(idev, &addr, plen, scope, IFAPROT_UNSPEC); 3194 addrconf_prefix_route(&addr, plen, 0, idev->dev, 0, pflags, 3195 GFP_KERNEL); 3196 return; 3197 } 3198 3199 for_each_netdev(net, dev) { 3200 struct in_device *in_dev = __in_dev_get_rtnl(dev); 3201 if (in_dev && (dev->flags & IFF_UP)) { 3202 struct in_ifaddr *ifa; 3203 int flag = scope; 3204 3205 in_dev_for_each_ifa_rtnl(ifa, in_dev) { 3206 addr.s6_addr32[3] = ifa->ifa_local; 3207 3208 if (ifa->ifa_scope == RT_SCOPE_LINK) 3209 continue; 3210 if (ifa->ifa_scope >= RT_SCOPE_HOST) { 3211 if (idev->dev->flags&IFF_POINTOPOINT) 3212 continue; 3213 flag |= IFA_HOST; 3214 } 3215 3216 add_addr(idev, &addr, plen, flag, 3217 IFAPROT_UNSPEC); 3218 addrconf_prefix_route(&addr, plen, 0, idev->dev, 3219 0, pflags, GFP_KERNEL); 3220 } 3221 } 3222 } 3223 } 3224 #endif 3225 3226 static void init_loopback(struct net_device *dev) 3227 { 3228 struct inet6_dev *idev; 3229 3230 /* ::1 */ 3231 3232 ASSERT_RTNL(); 3233 3234 idev = ipv6_find_idev(dev); 3235 if (IS_ERR(idev)) { 3236 pr_debug("%s: add_dev failed\n", __func__); 3237 return; 3238 } 3239 3240 add_addr(idev, &in6addr_loopback, 128, IFA_HOST, IFAPROT_KERNEL_LO); 3241 } 3242 3243 void addrconf_add_linklocal(struct inet6_dev *idev, 3244 const struct in6_addr *addr, u32 flags) 3245 { 3246 struct ifa6_config cfg = { 3247 .pfx = addr, 3248 .plen = 64, 3249 .ifa_flags = flags | IFA_F_PERMANENT, 3250 .valid_lft = INFINITY_LIFE_TIME, 3251 .preferred_lft = INFINITY_LIFE_TIME, 3252 .scope = IFA_LINK, 3253 .ifa_proto = IFAPROT_KERNEL_LL 3254 }; 3255 struct inet6_ifaddr *ifp; 3256 3257 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 3258 if ((dev_net(idev->dev)->ipv6.devconf_all->optimistic_dad || 3259 idev->cnf.optimistic_dad) && 3260 !dev_net(idev->dev)->ipv6.devconf_all->forwarding) 3261 cfg.ifa_flags |= IFA_F_OPTIMISTIC; 3262 #endif 3263 3264 ifp = ipv6_add_addr(idev, &cfg, true, NULL); 3265 if (!IS_ERR(ifp)) { 3266 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 0, idev->dev, 3267 0, 0, GFP_ATOMIC); 3268 addrconf_dad_start(ifp); 3269 in6_ifa_put(ifp); 3270 } 3271 } 3272 EXPORT_SYMBOL_GPL(addrconf_add_linklocal); 3273 3274 static bool ipv6_reserved_interfaceid(struct in6_addr address) 3275 { 3276 if ((address.s6_addr32[2] | address.s6_addr32[3]) == 0) 3277 return true; 3278 3279 if (address.s6_addr32[2] == htonl(0x02005eff) && 3280 ((address.s6_addr32[3] & htonl(0xfe000000)) == htonl(0xfe000000))) 3281 return true; 3282 3283 if (address.s6_addr32[2] == htonl(0xfdffffff) && 3284 ((address.s6_addr32[3] & htonl(0xffffff80)) == htonl(0xffffff80))) 3285 return true; 3286 3287 return false; 3288 } 3289 3290 static int ipv6_generate_stable_address(struct in6_addr *address, 3291 u8 dad_count, 3292 const struct inet6_dev *idev) 3293 { 3294 static DEFINE_SPINLOCK(lock); 3295 static __u32 digest[SHA1_DIGEST_WORDS]; 3296 static __u32 workspace[SHA1_WORKSPACE_WORDS]; 3297 3298 static union { 3299 char __data[SHA1_BLOCK_SIZE]; 3300 struct { 3301 struct in6_addr secret; 3302 __be32 prefix[2]; 3303 unsigned char hwaddr[MAX_ADDR_LEN]; 3304 u8 dad_count; 3305 } __packed; 3306 } data; 3307 3308 struct in6_addr secret; 3309 struct in6_addr temp; 3310 struct net *net = dev_net(idev->dev); 3311 3312 BUILD_BUG_ON(sizeof(data.__data) != sizeof(data)); 3313 3314 if (idev->cnf.stable_secret.initialized) 3315 secret = idev->cnf.stable_secret.secret; 3316 else if (net->ipv6.devconf_dflt->stable_secret.initialized) 3317 secret = net->ipv6.devconf_dflt->stable_secret.secret; 3318 else 3319 return -1; 3320 3321 retry: 3322 spin_lock_bh(&lock); 3323 3324 sha1_init(digest); 3325 memset(&data, 0, sizeof(data)); 3326 memset(workspace, 0, sizeof(workspace)); 3327 memcpy(data.hwaddr, idev->dev->perm_addr, idev->dev->addr_len); 3328 data.prefix[0] = address->s6_addr32[0]; 3329 data.prefix[1] = address->s6_addr32[1]; 3330 data.secret = secret; 3331 data.dad_count = dad_count; 3332 3333 sha1_transform(digest, data.__data, workspace); 3334 3335 temp = *address; 3336 temp.s6_addr32[2] = (__force __be32)digest[0]; 3337 temp.s6_addr32[3] = (__force __be32)digest[1]; 3338 3339 spin_unlock_bh(&lock); 3340 3341 if (ipv6_reserved_interfaceid(temp)) { 3342 dad_count++; 3343 if (dad_count > dev_net(idev->dev)->ipv6.sysctl.idgen_retries) 3344 return -1; 3345 goto retry; 3346 } 3347 3348 *address = temp; 3349 return 0; 3350 } 3351 3352 static void ipv6_gen_mode_random_init(struct inet6_dev *idev) 3353 { 3354 struct ipv6_stable_secret *s = &idev->cnf.stable_secret; 3355 3356 if (s->initialized) 3357 return; 3358 s = &idev->cnf.stable_secret; 3359 get_random_bytes(&s->secret, sizeof(s->secret)); 3360 s->initialized = true; 3361 } 3362 3363 static void addrconf_addr_gen(struct inet6_dev *idev, bool prefix_route) 3364 { 3365 struct in6_addr addr; 3366 3367 /* no link local addresses on L3 master devices */ 3368 if (netif_is_l3_master(idev->dev)) 3369 return; 3370 3371 /* no link local addresses on devices flagged as slaves */ 3372 if (idev->dev->priv_flags & IFF_NO_ADDRCONF) 3373 return; 3374 3375 ipv6_addr_set(&addr, htonl(0xFE800000), 0, 0, 0); 3376 3377 switch (idev->cnf.addr_gen_mode) { 3378 case IN6_ADDR_GEN_MODE_RANDOM: 3379 ipv6_gen_mode_random_init(idev); 3380 fallthrough; 3381 case IN6_ADDR_GEN_MODE_STABLE_PRIVACY: 3382 if (!ipv6_generate_stable_address(&addr, 0, idev)) 3383 addrconf_add_linklocal(idev, &addr, 3384 IFA_F_STABLE_PRIVACY); 3385 else if (prefix_route) 3386 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3387 0, 0, GFP_KERNEL); 3388 break; 3389 case IN6_ADDR_GEN_MODE_EUI64: 3390 /* addrconf_add_linklocal also adds a prefix_route and we 3391 * only need to care about prefix routes if ipv6_generate_eui64 3392 * couldn't generate one. 3393 */ 3394 if (ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) == 0) 3395 addrconf_add_linklocal(idev, &addr, 0); 3396 else if (prefix_route) 3397 addrconf_prefix_route(&addr, 64, 0, idev->dev, 3398 0, 0, GFP_KERNEL); 3399 break; 3400 case IN6_ADDR_GEN_MODE_NONE: 3401 default: 3402 /* will not add any link local address */ 3403 break; 3404 } 3405 } 3406 3407 static void addrconf_dev_config(struct net_device *dev) 3408 { 3409 struct inet6_dev *idev; 3410 3411 ASSERT_RTNL(); 3412 3413 if ((dev->type != ARPHRD_ETHER) && 3414 (dev->type != ARPHRD_FDDI) && 3415 (dev->type != ARPHRD_ARCNET) && 3416 (dev->type != ARPHRD_INFINIBAND) && 3417 (dev->type != ARPHRD_IEEE1394) && 3418 (dev->type != ARPHRD_TUNNEL6) && 3419 (dev->type != ARPHRD_6LOWPAN) && 3420 (dev->type != ARPHRD_TUNNEL) && 3421 (dev->type != ARPHRD_NONE) && 3422 (dev->type != ARPHRD_RAWIP)) { 3423 /* Alas, we support only Ethernet autoconfiguration. */ 3424 idev = __in6_dev_get(dev); 3425 if (!IS_ERR_OR_NULL(idev) && dev->flags & IFF_UP && 3426 dev->flags & IFF_MULTICAST) 3427 ipv6_mc_up(idev); 3428 return; 3429 } 3430 3431 idev = addrconf_add_dev(dev); 3432 if (IS_ERR(idev)) 3433 return; 3434 3435 /* this device type has no EUI support */ 3436 if (dev->type == ARPHRD_NONE && 3437 idev->cnf.addr_gen_mode == IN6_ADDR_GEN_MODE_EUI64) 3438 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_RANDOM; 3439 3440 addrconf_addr_gen(idev, false); 3441 } 3442 3443 #if IS_ENABLED(CONFIG_IPV6_SIT) 3444 static void addrconf_sit_config(struct net_device *dev) 3445 { 3446 struct inet6_dev *idev; 3447 3448 ASSERT_RTNL(); 3449 3450 /* 3451 * Configure the tunnel with one of our IPv4 3452 * addresses... we should configure all of 3453 * our v4 addrs in the tunnel 3454 */ 3455 3456 idev = ipv6_find_idev(dev); 3457 if (IS_ERR(idev)) { 3458 pr_debug("%s: add_dev failed\n", __func__); 3459 return; 3460 } 3461 3462 if (dev->priv_flags & IFF_ISATAP) { 3463 addrconf_addr_gen(idev, false); 3464 return; 3465 } 3466 3467 add_v4_addrs(idev); 3468 3469 if (dev->flags&IFF_POINTOPOINT) 3470 addrconf_add_mroute(dev); 3471 } 3472 #endif 3473 3474 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE) 3475 static void addrconf_gre_config(struct net_device *dev) 3476 { 3477 struct inet6_dev *idev; 3478 3479 ASSERT_RTNL(); 3480 3481 idev = ipv6_find_idev(dev); 3482 if (IS_ERR(idev)) { 3483 pr_debug("%s: add_dev failed\n", __func__); 3484 return; 3485 } 3486 3487 if (dev->type == ARPHRD_ETHER) { 3488 addrconf_addr_gen(idev, true); 3489 return; 3490 } 3491 3492 add_v4_addrs(idev); 3493 3494 if (dev->flags & IFF_POINTOPOINT) 3495 addrconf_add_mroute(dev); 3496 } 3497 #endif 3498 3499 static void addrconf_init_auto_addrs(struct net_device *dev) 3500 { 3501 switch (dev->type) { 3502 #if IS_ENABLED(CONFIG_IPV6_SIT) 3503 case ARPHRD_SIT: 3504 addrconf_sit_config(dev); 3505 break; 3506 #endif 3507 #if IS_ENABLED(CONFIG_NET_IPGRE) || IS_ENABLED(CONFIG_IPV6_GRE) 3508 case ARPHRD_IP6GRE: 3509 case ARPHRD_IPGRE: 3510 addrconf_gre_config(dev); 3511 break; 3512 #endif 3513 case ARPHRD_LOOPBACK: 3514 init_loopback(dev); 3515 break; 3516 3517 default: 3518 addrconf_dev_config(dev); 3519 break; 3520 } 3521 } 3522 3523 static int fixup_permanent_addr(struct net *net, 3524 struct inet6_dev *idev, 3525 struct inet6_ifaddr *ifp) 3526 { 3527 /* !fib6_node means the host route was removed from the 3528 * FIB, for example, if 'lo' device is taken down. In that 3529 * case regenerate the host route. 3530 */ 3531 if (!ifp->rt || !ifp->rt->fib6_node) { 3532 struct fib6_info *f6i, *prev; 3533 3534 f6i = addrconf_f6i_alloc(net, idev, &ifp->addr, false, 3535 GFP_ATOMIC, NULL); 3536 if (IS_ERR(f6i)) 3537 return PTR_ERR(f6i); 3538 3539 /* ifp->rt can be accessed outside of rtnl */ 3540 spin_lock(&ifp->lock); 3541 prev = ifp->rt; 3542 ifp->rt = f6i; 3543 spin_unlock(&ifp->lock); 3544 3545 fib6_info_release(prev); 3546 } 3547 3548 if (!(ifp->flags & IFA_F_NOPREFIXROUTE)) { 3549 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 3550 ifp->rt_priority, idev->dev, 0, 0, 3551 GFP_ATOMIC); 3552 } 3553 3554 if (ifp->state == INET6_IFADDR_STATE_PREDAD) 3555 addrconf_dad_start(ifp); 3556 3557 return 0; 3558 } 3559 3560 static void addrconf_permanent_addr(struct net *net, struct net_device *dev) 3561 { 3562 struct inet6_ifaddr *ifp, *tmp; 3563 struct inet6_dev *idev; 3564 3565 idev = __in6_dev_get(dev); 3566 if (!idev) 3567 return; 3568 3569 write_lock_bh(&idev->lock); 3570 3571 list_for_each_entry_safe(ifp, tmp, &idev->addr_list, if_list) { 3572 if ((ifp->flags & IFA_F_PERMANENT) && 3573 fixup_permanent_addr(net, idev, ifp) < 0) { 3574 write_unlock_bh(&idev->lock); 3575 in6_ifa_hold(ifp); 3576 ipv6_del_addr(ifp); 3577 write_lock_bh(&idev->lock); 3578 3579 net_info_ratelimited("%s: Failed to add prefix route for address %pI6c; dropping\n", 3580 idev->dev->name, &ifp->addr); 3581 } 3582 } 3583 3584 write_unlock_bh(&idev->lock); 3585 } 3586 3587 static int addrconf_notify(struct notifier_block *this, unsigned long event, 3588 void *ptr) 3589 { 3590 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 3591 struct netdev_notifier_change_info *change_info; 3592 struct netdev_notifier_changeupper_info *info; 3593 struct inet6_dev *idev = __in6_dev_get(dev); 3594 struct net *net = dev_net(dev); 3595 int run_pending = 0; 3596 int err; 3597 3598 switch (event) { 3599 case NETDEV_REGISTER: 3600 if (!idev && dev->mtu >= IPV6_MIN_MTU) { 3601 idev = ipv6_add_dev(dev); 3602 if (IS_ERR(idev)) 3603 return notifier_from_errno(PTR_ERR(idev)); 3604 } 3605 break; 3606 3607 case NETDEV_CHANGEMTU: 3608 /* if MTU under IPV6_MIN_MTU stop IPv6 on this interface. */ 3609 if (dev->mtu < IPV6_MIN_MTU) { 3610 addrconf_ifdown(dev, dev != net->loopback_dev); 3611 break; 3612 } 3613 3614 if (idev) { 3615 rt6_mtu_change(dev, dev->mtu); 3616 idev->cnf.mtu6 = dev->mtu; 3617 break; 3618 } 3619 3620 /* allocate new idev */ 3621 idev = ipv6_add_dev(dev); 3622 if (IS_ERR(idev)) 3623 break; 3624 3625 /* device is still not ready */ 3626 if (!(idev->if_flags & IF_READY)) 3627 break; 3628 3629 run_pending = 1; 3630 fallthrough; 3631 case NETDEV_UP: 3632 case NETDEV_CHANGE: 3633 if (idev && idev->cnf.disable_ipv6) 3634 break; 3635 3636 if (dev->priv_flags & IFF_NO_ADDRCONF) { 3637 if (event == NETDEV_UP && !IS_ERR_OR_NULL(idev) && 3638 dev->flags & IFF_UP && dev->flags & IFF_MULTICAST) 3639 ipv6_mc_up(idev); 3640 break; 3641 } 3642 3643 if (event == NETDEV_UP) { 3644 /* restore routes for permanent addresses */ 3645 addrconf_permanent_addr(net, dev); 3646 3647 if (!addrconf_link_ready(dev)) { 3648 /* device is not ready yet. */ 3649 pr_debug("ADDRCONF(NETDEV_UP): %s: link is not ready\n", 3650 dev->name); 3651 break; 3652 } 3653 3654 if (!idev && dev->mtu >= IPV6_MIN_MTU) 3655 idev = ipv6_add_dev(dev); 3656 3657 if (!IS_ERR_OR_NULL(idev)) { 3658 idev->if_flags |= IF_READY; 3659 run_pending = 1; 3660 } 3661 } else if (event == NETDEV_CHANGE) { 3662 if (!addrconf_link_ready(dev)) { 3663 /* device is still not ready. */ 3664 rt6_sync_down_dev(dev, event); 3665 break; 3666 } 3667 3668 if (!IS_ERR_OR_NULL(idev)) { 3669 if (idev->if_flags & IF_READY) { 3670 /* device is already configured - 3671 * but resend MLD reports, we might 3672 * have roamed and need to update 3673 * multicast snooping switches 3674 */ 3675 ipv6_mc_up(idev); 3676 change_info = ptr; 3677 if (change_info->flags_changed & IFF_NOARP) 3678 addrconf_dad_run(idev, true); 3679 rt6_sync_up(dev, RTNH_F_LINKDOWN); 3680 break; 3681 } 3682 idev->if_flags |= IF_READY; 3683 } 3684 3685 pr_debug("ADDRCONF(NETDEV_CHANGE): %s: link becomes ready\n", 3686 dev->name); 3687 3688 run_pending = 1; 3689 } 3690 3691 addrconf_init_auto_addrs(dev); 3692 3693 if (!IS_ERR_OR_NULL(idev)) { 3694 if (run_pending) 3695 addrconf_dad_run(idev, false); 3696 3697 /* Device has an address by now */ 3698 rt6_sync_up(dev, RTNH_F_DEAD); 3699 3700 /* 3701 * If the MTU changed during the interface down, 3702 * when the interface up, the changed MTU must be 3703 * reflected in the idev as well as routers. 3704 */ 3705 if (idev->cnf.mtu6 != dev->mtu && 3706 dev->mtu >= IPV6_MIN_MTU) { 3707 rt6_mtu_change(dev, dev->mtu); 3708 idev->cnf.mtu6 = dev->mtu; 3709 } 3710 idev->tstamp = jiffies; 3711 inet6_ifinfo_notify(RTM_NEWLINK, idev); 3712 3713 /* 3714 * If the changed mtu during down is lower than 3715 * IPV6_MIN_MTU stop IPv6 on this interface. 3716 */ 3717 if (dev->mtu < IPV6_MIN_MTU) 3718 addrconf_ifdown(dev, dev != net->loopback_dev); 3719 } 3720 break; 3721 3722 case NETDEV_DOWN: 3723 case NETDEV_UNREGISTER: 3724 /* 3725 * Remove all addresses from this interface. 3726 */ 3727 addrconf_ifdown(dev, event != NETDEV_DOWN); 3728 break; 3729 3730 case NETDEV_CHANGENAME: 3731 if (idev) { 3732 snmp6_unregister_dev(idev); 3733 addrconf_sysctl_unregister(idev); 3734 err = addrconf_sysctl_register(idev); 3735 if (err) 3736 return notifier_from_errno(err); 3737 err = snmp6_register_dev(idev); 3738 if (err) { 3739 addrconf_sysctl_unregister(idev); 3740 return notifier_from_errno(err); 3741 } 3742 } 3743 break; 3744 3745 case NETDEV_PRE_TYPE_CHANGE: 3746 case NETDEV_POST_TYPE_CHANGE: 3747 if (idev) 3748 addrconf_type_change(dev, event); 3749 break; 3750 3751 case NETDEV_CHANGEUPPER: 3752 info = ptr; 3753 3754 /* flush all routes if dev is linked to or unlinked from 3755 * an L3 master device (e.g., VRF) 3756 */ 3757 if (info->upper_dev && netif_is_l3_master(info->upper_dev)) 3758 addrconf_ifdown(dev, false); 3759 } 3760 3761 return NOTIFY_OK; 3762 } 3763 3764 /* 3765 * addrconf module should be notified of a device going up 3766 */ 3767 static struct notifier_block ipv6_dev_notf = { 3768 .notifier_call = addrconf_notify, 3769 .priority = ADDRCONF_NOTIFY_PRIORITY, 3770 }; 3771 3772 static void addrconf_type_change(struct net_device *dev, unsigned long event) 3773 { 3774 struct inet6_dev *idev; 3775 ASSERT_RTNL(); 3776 3777 idev = __in6_dev_get(dev); 3778 3779 if (event == NETDEV_POST_TYPE_CHANGE) 3780 ipv6_mc_remap(idev); 3781 else if (event == NETDEV_PRE_TYPE_CHANGE) 3782 ipv6_mc_unmap(idev); 3783 } 3784 3785 static bool addr_is_local(const struct in6_addr *addr) 3786 { 3787 return ipv6_addr_type(addr) & 3788 (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_LOOPBACK); 3789 } 3790 3791 static int addrconf_ifdown(struct net_device *dev, bool unregister) 3792 { 3793 unsigned long event = unregister ? NETDEV_UNREGISTER : NETDEV_DOWN; 3794 struct net *net = dev_net(dev); 3795 struct inet6_dev *idev; 3796 struct inet6_ifaddr *ifa; 3797 LIST_HEAD(tmp_addr_list); 3798 bool keep_addr = false; 3799 bool was_ready; 3800 int state, i; 3801 3802 ASSERT_RTNL(); 3803 3804 rt6_disable_ip(dev, event); 3805 3806 idev = __in6_dev_get(dev); 3807 if (!idev) 3808 return -ENODEV; 3809 3810 /* 3811 * Step 1: remove reference to ipv6 device from parent device. 3812 * Do not dev_put! 3813 */ 3814 if (unregister) { 3815 idev->dead = 1; 3816 3817 /* protected by rtnl_lock */ 3818 RCU_INIT_POINTER(dev->ip6_ptr, NULL); 3819 3820 /* Step 1.5: remove snmp6 entry */ 3821 snmp6_unregister_dev(idev); 3822 3823 } 3824 3825 /* combine the user config with event to determine if permanent 3826 * addresses are to be removed from address hash table 3827 */ 3828 if (!unregister && !idev->cnf.disable_ipv6) { 3829 /* aggregate the system setting and interface setting */ 3830 int _keep_addr = net->ipv6.devconf_all->keep_addr_on_down; 3831 3832 if (!_keep_addr) 3833 _keep_addr = idev->cnf.keep_addr_on_down; 3834 3835 keep_addr = (_keep_addr > 0); 3836 } 3837 3838 /* Step 2: clear hash table */ 3839 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 3840 struct hlist_head *h = &net->ipv6.inet6_addr_lst[i]; 3841 3842 spin_lock_bh(&net->ipv6.addrconf_hash_lock); 3843 restart: 3844 hlist_for_each_entry_rcu(ifa, h, addr_lst) { 3845 if (ifa->idev == idev) { 3846 addrconf_del_dad_work(ifa); 3847 /* combined flag + permanent flag decide if 3848 * address is retained on a down event 3849 */ 3850 if (!keep_addr || 3851 !(ifa->flags & IFA_F_PERMANENT) || 3852 addr_is_local(&ifa->addr)) { 3853 hlist_del_init_rcu(&ifa->addr_lst); 3854 goto restart; 3855 } 3856 } 3857 } 3858 spin_unlock_bh(&net->ipv6.addrconf_hash_lock); 3859 } 3860 3861 write_lock_bh(&idev->lock); 3862 3863 addrconf_del_rs_timer(idev); 3864 3865 /* Step 2: clear flags for stateless addrconf, repeated down 3866 * detection 3867 */ 3868 was_ready = idev->if_flags & IF_READY; 3869 if (!unregister) 3870 idev->if_flags &= ~(IF_RS_SENT|IF_RA_RCVD|IF_READY); 3871 3872 /* Step 3: clear tempaddr list */ 3873 while (!list_empty(&idev->tempaddr_list)) { 3874 ifa = list_first_entry(&idev->tempaddr_list, 3875 struct inet6_ifaddr, tmp_list); 3876 list_del(&ifa->tmp_list); 3877 write_unlock_bh(&idev->lock); 3878 spin_lock_bh(&ifa->lock); 3879 3880 if (ifa->ifpub) { 3881 in6_ifa_put(ifa->ifpub); 3882 ifa->ifpub = NULL; 3883 } 3884 spin_unlock_bh(&ifa->lock); 3885 in6_ifa_put(ifa); 3886 write_lock_bh(&idev->lock); 3887 } 3888 3889 list_for_each_entry(ifa, &idev->addr_list, if_list) 3890 list_add_tail(&ifa->if_list_aux, &tmp_addr_list); 3891 write_unlock_bh(&idev->lock); 3892 3893 while (!list_empty(&tmp_addr_list)) { 3894 struct fib6_info *rt = NULL; 3895 bool keep; 3896 3897 ifa = list_first_entry(&tmp_addr_list, 3898 struct inet6_ifaddr, if_list_aux); 3899 list_del(&ifa->if_list_aux); 3900 3901 addrconf_del_dad_work(ifa); 3902 3903 keep = keep_addr && (ifa->flags & IFA_F_PERMANENT) && 3904 !addr_is_local(&ifa->addr); 3905 3906 spin_lock_bh(&ifa->lock); 3907 3908 if (keep) { 3909 /* set state to skip the notifier below */ 3910 state = INET6_IFADDR_STATE_DEAD; 3911 ifa->state = INET6_IFADDR_STATE_PREDAD; 3912 if (!(ifa->flags & IFA_F_NODAD)) 3913 ifa->flags |= IFA_F_TENTATIVE; 3914 3915 rt = ifa->rt; 3916 ifa->rt = NULL; 3917 } else { 3918 state = ifa->state; 3919 ifa->state = INET6_IFADDR_STATE_DEAD; 3920 } 3921 3922 spin_unlock_bh(&ifa->lock); 3923 3924 if (rt) 3925 ip6_del_rt(net, rt, false); 3926 3927 if (state != INET6_IFADDR_STATE_DEAD) { 3928 __ipv6_ifa_notify(RTM_DELADDR, ifa); 3929 inet6addr_notifier_call_chain(NETDEV_DOWN, ifa); 3930 } else { 3931 if (idev->cnf.forwarding) 3932 addrconf_leave_anycast(ifa); 3933 addrconf_leave_solict(ifa->idev, &ifa->addr); 3934 } 3935 3936 if (!keep) { 3937 write_lock_bh(&idev->lock); 3938 list_del_rcu(&ifa->if_list); 3939 write_unlock_bh(&idev->lock); 3940 in6_ifa_put(ifa); 3941 } 3942 } 3943 3944 /* Step 5: Discard anycast and multicast list */ 3945 if (unregister) { 3946 ipv6_ac_destroy_dev(idev); 3947 ipv6_mc_destroy_dev(idev); 3948 } else if (was_ready) { 3949 ipv6_mc_down(idev); 3950 } 3951 3952 idev->tstamp = jiffies; 3953 idev->ra_mtu = 0; 3954 3955 /* Last: Shot the device (if unregistered) */ 3956 if (unregister) { 3957 addrconf_sysctl_unregister(idev); 3958 neigh_parms_release(&nd_tbl, idev->nd_parms); 3959 neigh_ifdown(&nd_tbl, dev); 3960 in6_dev_put(idev); 3961 } 3962 return 0; 3963 } 3964 3965 static void addrconf_rs_timer(struct timer_list *t) 3966 { 3967 struct inet6_dev *idev = from_timer(idev, t, rs_timer); 3968 struct net_device *dev = idev->dev; 3969 struct in6_addr lladdr; 3970 3971 write_lock(&idev->lock); 3972 if (idev->dead || !(idev->if_flags & IF_READY)) 3973 goto out; 3974 3975 if (!ipv6_accept_ra(idev)) 3976 goto out; 3977 3978 /* Announcement received after solicitation was sent */ 3979 if (idev->if_flags & IF_RA_RCVD) 3980 goto out; 3981 3982 if (idev->rs_probes++ < idev->cnf.rtr_solicits || idev->cnf.rtr_solicits < 0) { 3983 write_unlock(&idev->lock); 3984 if (!ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 3985 ndisc_send_rs(dev, &lladdr, 3986 &in6addr_linklocal_allrouters); 3987 else 3988 goto put; 3989 3990 write_lock(&idev->lock); 3991 idev->rs_interval = rfc3315_s14_backoff_update( 3992 idev->rs_interval, idev->cnf.rtr_solicit_max_interval); 3993 /* The wait after the last probe can be shorter */ 3994 addrconf_mod_rs_timer(idev, (idev->rs_probes == 3995 idev->cnf.rtr_solicits) ? 3996 idev->cnf.rtr_solicit_delay : 3997 idev->rs_interval); 3998 } else { 3999 /* 4000 * Note: we do not support deprecated "all on-link" 4001 * assumption any longer. 4002 */ 4003 pr_debug("%s: no IPv6 routers present\n", idev->dev->name); 4004 } 4005 4006 out: 4007 write_unlock(&idev->lock); 4008 put: 4009 in6_dev_put(idev); 4010 } 4011 4012 /* 4013 * Duplicate Address Detection 4014 */ 4015 static void addrconf_dad_kick(struct inet6_ifaddr *ifp) 4016 { 4017 unsigned long rand_num; 4018 struct inet6_dev *idev = ifp->idev; 4019 u64 nonce; 4020 4021 if (ifp->flags & IFA_F_OPTIMISTIC) 4022 rand_num = 0; 4023 else 4024 rand_num = get_random_u32_below(idev->cnf.rtr_solicit_delay ? : 1); 4025 4026 nonce = 0; 4027 if (idev->cnf.enhanced_dad || 4028 dev_net(idev->dev)->ipv6.devconf_all->enhanced_dad) { 4029 do 4030 get_random_bytes(&nonce, 6); 4031 while (nonce == 0); 4032 } 4033 ifp->dad_nonce = nonce; 4034 ifp->dad_probes = idev->cnf.dad_transmits; 4035 addrconf_mod_dad_work(ifp, rand_num); 4036 } 4037 4038 static void addrconf_dad_begin(struct inet6_ifaddr *ifp) 4039 { 4040 struct inet6_dev *idev = ifp->idev; 4041 struct net_device *dev = idev->dev; 4042 bool bump_id, notify = false; 4043 struct net *net; 4044 4045 addrconf_join_solict(dev, &ifp->addr); 4046 4047 read_lock_bh(&idev->lock); 4048 spin_lock(&ifp->lock); 4049 if (ifp->state == INET6_IFADDR_STATE_DEAD) 4050 goto out; 4051 4052 net = dev_net(dev); 4053 if (dev->flags&(IFF_NOARP|IFF_LOOPBACK) || 4054 (net->ipv6.devconf_all->accept_dad < 1 && 4055 idev->cnf.accept_dad < 1) || 4056 !(ifp->flags&IFA_F_TENTATIVE) || 4057 ifp->flags & IFA_F_NODAD) { 4058 bool send_na = false; 4059 4060 if (ifp->flags & IFA_F_TENTATIVE && 4061 !(ifp->flags & IFA_F_OPTIMISTIC)) 4062 send_na = true; 4063 bump_id = ifp->flags & IFA_F_TENTATIVE; 4064 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4065 spin_unlock(&ifp->lock); 4066 read_unlock_bh(&idev->lock); 4067 4068 addrconf_dad_completed(ifp, bump_id, send_na); 4069 return; 4070 } 4071 4072 if (!(idev->if_flags & IF_READY)) { 4073 spin_unlock(&ifp->lock); 4074 read_unlock_bh(&idev->lock); 4075 /* 4076 * If the device is not ready: 4077 * - keep it tentative if it is a permanent address. 4078 * - otherwise, kill it. 4079 */ 4080 in6_ifa_hold(ifp); 4081 addrconf_dad_stop(ifp, 0); 4082 return; 4083 } 4084 4085 /* 4086 * Optimistic nodes can start receiving 4087 * Frames right away 4088 */ 4089 if (ifp->flags & IFA_F_OPTIMISTIC) { 4090 ip6_ins_rt(net, ifp->rt); 4091 if (ipv6_use_optimistic_addr(net, idev)) { 4092 /* Because optimistic nodes can use this address, 4093 * notify listeners. If DAD fails, RTM_DELADDR is sent. 4094 */ 4095 notify = true; 4096 } 4097 } 4098 4099 addrconf_dad_kick(ifp); 4100 out: 4101 spin_unlock(&ifp->lock); 4102 read_unlock_bh(&idev->lock); 4103 if (notify) 4104 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4105 } 4106 4107 static void addrconf_dad_start(struct inet6_ifaddr *ifp) 4108 { 4109 bool begin_dad = false; 4110 4111 spin_lock_bh(&ifp->lock); 4112 if (ifp->state != INET6_IFADDR_STATE_DEAD) { 4113 ifp->state = INET6_IFADDR_STATE_PREDAD; 4114 begin_dad = true; 4115 } 4116 spin_unlock_bh(&ifp->lock); 4117 4118 if (begin_dad) 4119 addrconf_mod_dad_work(ifp, 0); 4120 } 4121 4122 static void addrconf_dad_work(struct work_struct *w) 4123 { 4124 struct inet6_ifaddr *ifp = container_of(to_delayed_work(w), 4125 struct inet6_ifaddr, 4126 dad_work); 4127 struct inet6_dev *idev = ifp->idev; 4128 bool bump_id, disable_ipv6 = false; 4129 struct in6_addr mcaddr; 4130 4131 enum { 4132 DAD_PROCESS, 4133 DAD_BEGIN, 4134 DAD_ABORT, 4135 } action = DAD_PROCESS; 4136 4137 rtnl_lock(); 4138 4139 spin_lock_bh(&ifp->lock); 4140 if (ifp->state == INET6_IFADDR_STATE_PREDAD) { 4141 action = DAD_BEGIN; 4142 ifp->state = INET6_IFADDR_STATE_DAD; 4143 } else if (ifp->state == INET6_IFADDR_STATE_ERRDAD) { 4144 action = DAD_ABORT; 4145 ifp->state = INET6_IFADDR_STATE_POSTDAD; 4146 4147 if ((dev_net(idev->dev)->ipv6.devconf_all->accept_dad > 1 || 4148 idev->cnf.accept_dad > 1) && 4149 !idev->cnf.disable_ipv6 && 4150 !(ifp->flags & IFA_F_STABLE_PRIVACY)) { 4151 struct in6_addr addr; 4152 4153 addr.s6_addr32[0] = htonl(0xfe800000); 4154 addr.s6_addr32[1] = 0; 4155 4156 if (!ipv6_generate_eui64(addr.s6_addr + 8, idev->dev) && 4157 ipv6_addr_equal(&ifp->addr, &addr)) { 4158 /* DAD failed for link-local based on MAC */ 4159 idev->cnf.disable_ipv6 = 1; 4160 4161 pr_info("%s: IPv6 being disabled!\n", 4162 ifp->idev->dev->name); 4163 disable_ipv6 = true; 4164 } 4165 } 4166 } 4167 spin_unlock_bh(&ifp->lock); 4168 4169 if (action == DAD_BEGIN) { 4170 addrconf_dad_begin(ifp); 4171 goto out; 4172 } else if (action == DAD_ABORT) { 4173 in6_ifa_hold(ifp); 4174 addrconf_dad_stop(ifp, 1); 4175 if (disable_ipv6) 4176 addrconf_ifdown(idev->dev, false); 4177 goto out; 4178 } 4179 4180 if (!ifp->dad_probes && addrconf_dad_end(ifp)) 4181 goto out; 4182 4183 write_lock_bh(&idev->lock); 4184 if (idev->dead || !(idev->if_flags & IF_READY)) { 4185 write_unlock_bh(&idev->lock); 4186 goto out; 4187 } 4188 4189 spin_lock(&ifp->lock); 4190 if (ifp->state == INET6_IFADDR_STATE_DEAD) { 4191 spin_unlock(&ifp->lock); 4192 write_unlock_bh(&idev->lock); 4193 goto out; 4194 } 4195 4196 if (ifp->dad_probes == 0) { 4197 bool send_na = false; 4198 4199 /* 4200 * DAD was successful 4201 */ 4202 4203 if (ifp->flags & IFA_F_TENTATIVE && 4204 !(ifp->flags & IFA_F_OPTIMISTIC)) 4205 send_na = true; 4206 bump_id = ifp->flags & IFA_F_TENTATIVE; 4207 ifp->flags &= ~(IFA_F_TENTATIVE|IFA_F_OPTIMISTIC|IFA_F_DADFAILED); 4208 spin_unlock(&ifp->lock); 4209 write_unlock_bh(&idev->lock); 4210 4211 addrconf_dad_completed(ifp, bump_id, send_na); 4212 4213 goto out; 4214 } 4215 4216 ifp->dad_probes--; 4217 addrconf_mod_dad_work(ifp, 4218 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), 4219 HZ/100)); 4220 spin_unlock(&ifp->lock); 4221 write_unlock_bh(&idev->lock); 4222 4223 /* send a neighbour solicitation for our addr */ 4224 addrconf_addr_solict_mult(&ifp->addr, &mcaddr); 4225 ndisc_send_ns(ifp->idev->dev, &ifp->addr, &mcaddr, &in6addr_any, 4226 ifp->dad_nonce); 4227 out: 4228 in6_ifa_put(ifp); 4229 rtnl_unlock(); 4230 } 4231 4232 /* ifp->idev must be at least read locked */ 4233 static bool ipv6_lonely_lladdr(struct inet6_ifaddr *ifp) 4234 { 4235 struct inet6_ifaddr *ifpiter; 4236 struct inet6_dev *idev = ifp->idev; 4237 4238 list_for_each_entry_reverse(ifpiter, &idev->addr_list, if_list) { 4239 if (ifpiter->scope > IFA_LINK) 4240 break; 4241 if (ifp != ifpiter && ifpiter->scope == IFA_LINK && 4242 (ifpiter->flags & (IFA_F_PERMANENT|IFA_F_TENTATIVE| 4243 IFA_F_OPTIMISTIC|IFA_F_DADFAILED)) == 4244 IFA_F_PERMANENT) 4245 return false; 4246 } 4247 return true; 4248 } 4249 4250 static void addrconf_dad_completed(struct inet6_ifaddr *ifp, bool bump_id, 4251 bool send_na) 4252 { 4253 struct net_device *dev = ifp->idev->dev; 4254 struct in6_addr lladdr; 4255 bool send_rs, send_mld; 4256 4257 addrconf_del_dad_work(ifp); 4258 4259 /* 4260 * Configure the address for reception. Now it is valid. 4261 */ 4262 4263 ipv6_ifa_notify(RTM_NEWADDR, ifp); 4264 4265 /* If added prefix is link local and we are prepared to process 4266 router advertisements, start sending router solicitations. 4267 */ 4268 4269 read_lock_bh(&ifp->idev->lock); 4270 send_mld = ifp->scope == IFA_LINK && ipv6_lonely_lladdr(ifp); 4271 send_rs = send_mld && 4272 ipv6_accept_ra(ifp->idev) && 4273 ifp->idev->cnf.rtr_solicits != 0 && 4274 (dev->flags & IFF_LOOPBACK) == 0 && 4275 (dev->type != ARPHRD_TUNNEL) && 4276 !netif_is_team_port(dev); 4277 read_unlock_bh(&ifp->idev->lock); 4278 4279 /* While dad is in progress mld report's source address is in6_addrany. 4280 * Resend with proper ll now. 4281 */ 4282 if (send_mld) 4283 ipv6_mc_dad_complete(ifp->idev); 4284 4285 /* send unsolicited NA if enabled */ 4286 if (send_na && 4287 (ifp->idev->cnf.ndisc_notify || 4288 dev_net(dev)->ipv6.devconf_all->ndisc_notify)) { 4289 ndisc_send_na(dev, &in6addr_linklocal_allnodes, &ifp->addr, 4290 /*router=*/ !!ifp->idev->cnf.forwarding, 4291 /*solicited=*/ false, /*override=*/ true, 4292 /*inc_opt=*/ true); 4293 } 4294 4295 if (send_rs) { 4296 /* 4297 * If a host as already performed a random delay 4298 * [...] as part of DAD [...] there is no need 4299 * to delay again before sending the first RS 4300 */ 4301 if (ipv6_get_lladdr(dev, &lladdr, IFA_F_TENTATIVE)) 4302 return; 4303 ndisc_send_rs(dev, &lladdr, &in6addr_linklocal_allrouters); 4304 4305 write_lock_bh(&ifp->idev->lock); 4306 spin_lock(&ifp->lock); 4307 ifp->idev->rs_interval = rfc3315_s14_backoff_init( 4308 ifp->idev->cnf.rtr_solicit_interval); 4309 ifp->idev->rs_probes = 1; 4310 ifp->idev->if_flags |= IF_RS_SENT; 4311 addrconf_mod_rs_timer(ifp->idev, ifp->idev->rs_interval); 4312 spin_unlock(&ifp->lock); 4313 write_unlock_bh(&ifp->idev->lock); 4314 } 4315 4316 if (bump_id) 4317 rt_genid_bump_ipv6(dev_net(dev)); 4318 4319 /* Make sure that a new temporary address will be created 4320 * before this temporary address becomes deprecated. 4321 */ 4322 if (ifp->flags & IFA_F_TEMPORARY) 4323 addrconf_verify_rtnl(dev_net(dev)); 4324 } 4325 4326 static void addrconf_dad_run(struct inet6_dev *idev, bool restart) 4327 { 4328 struct inet6_ifaddr *ifp; 4329 4330 read_lock_bh(&idev->lock); 4331 list_for_each_entry(ifp, &idev->addr_list, if_list) { 4332 spin_lock(&ifp->lock); 4333 if ((ifp->flags & IFA_F_TENTATIVE && 4334 ifp->state == INET6_IFADDR_STATE_DAD) || restart) { 4335 if (restart) 4336 ifp->state = INET6_IFADDR_STATE_PREDAD; 4337 addrconf_dad_kick(ifp); 4338 } 4339 spin_unlock(&ifp->lock); 4340 } 4341 read_unlock_bh(&idev->lock); 4342 } 4343 4344 #ifdef CONFIG_PROC_FS 4345 struct if6_iter_state { 4346 struct seq_net_private p; 4347 int bucket; 4348 int offset; 4349 }; 4350 4351 static struct inet6_ifaddr *if6_get_first(struct seq_file *seq, loff_t pos) 4352 { 4353 struct if6_iter_state *state = seq->private; 4354 struct net *net = seq_file_net(seq); 4355 struct inet6_ifaddr *ifa = NULL; 4356 int p = 0; 4357 4358 /* initial bucket if pos is 0 */ 4359 if (pos == 0) { 4360 state->bucket = 0; 4361 state->offset = 0; 4362 } 4363 4364 for (; state->bucket < IN6_ADDR_HSIZE; ++state->bucket) { 4365 hlist_for_each_entry_rcu(ifa, &net->ipv6.inet6_addr_lst[state->bucket], 4366 addr_lst) { 4367 /* sync with offset */ 4368 if (p < state->offset) { 4369 p++; 4370 continue; 4371 } 4372 return ifa; 4373 } 4374 4375 /* prepare for next bucket */ 4376 state->offset = 0; 4377 p = 0; 4378 } 4379 return NULL; 4380 } 4381 4382 static struct inet6_ifaddr *if6_get_next(struct seq_file *seq, 4383 struct inet6_ifaddr *ifa) 4384 { 4385 struct if6_iter_state *state = seq->private; 4386 struct net *net = seq_file_net(seq); 4387 4388 hlist_for_each_entry_continue_rcu(ifa, addr_lst) { 4389 state->offset++; 4390 return ifa; 4391 } 4392 4393 state->offset = 0; 4394 while (++state->bucket < IN6_ADDR_HSIZE) { 4395 hlist_for_each_entry_rcu(ifa, 4396 &net->ipv6.inet6_addr_lst[state->bucket], addr_lst) { 4397 return ifa; 4398 } 4399 } 4400 4401 return NULL; 4402 } 4403 4404 static void *if6_seq_start(struct seq_file *seq, loff_t *pos) 4405 __acquires(rcu) 4406 { 4407 rcu_read_lock(); 4408 return if6_get_first(seq, *pos); 4409 } 4410 4411 static void *if6_seq_next(struct seq_file *seq, void *v, loff_t *pos) 4412 { 4413 struct inet6_ifaddr *ifa; 4414 4415 ifa = if6_get_next(seq, v); 4416 ++*pos; 4417 return ifa; 4418 } 4419 4420 static void if6_seq_stop(struct seq_file *seq, void *v) 4421 __releases(rcu) 4422 { 4423 rcu_read_unlock(); 4424 } 4425 4426 static int if6_seq_show(struct seq_file *seq, void *v) 4427 { 4428 struct inet6_ifaddr *ifp = (struct inet6_ifaddr *)v; 4429 seq_printf(seq, "%pi6 %02x %02x %02x %02x %8s\n", 4430 &ifp->addr, 4431 ifp->idev->dev->ifindex, 4432 ifp->prefix_len, 4433 ifp->scope, 4434 (u8) ifp->flags, 4435 ifp->idev->dev->name); 4436 return 0; 4437 } 4438 4439 static const struct seq_operations if6_seq_ops = { 4440 .start = if6_seq_start, 4441 .next = if6_seq_next, 4442 .show = if6_seq_show, 4443 .stop = if6_seq_stop, 4444 }; 4445 4446 static int __net_init if6_proc_net_init(struct net *net) 4447 { 4448 if (!proc_create_net("if_inet6", 0444, net->proc_net, &if6_seq_ops, 4449 sizeof(struct if6_iter_state))) 4450 return -ENOMEM; 4451 return 0; 4452 } 4453 4454 static void __net_exit if6_proc_net_exit(struct net *net) 4455 { 4456 remove_proc_entry("if_inet6", net->proc_net); 4457 } 4458 4459 static struct pernet_operations if6_proc_net_ops = { 4460 .init = if6_proc_net_init, 4461 .exit = if6_proc_net_exit, 4462 }; 4463 4464 int __init if6_proc_init(void) 4465 { 4466 return register_pernet_subsys(&if6_proc_net_ops); 4467 } 4468 4469 void if6_proc_exit(void) 4470 { 4471 unregister_pernet_subsys(&if6_proc_net_ops); 4472 } 4473 #endif /* CONFIG_PROC_FS */ 4474 4475 #if IS_ENABLED(CONFIG_IPV6_MIP6) 4476 /* Check if address is a home address configured on any interface. */ 4477 int ipv6_chk_home_addr(struct net *net, const struct in6_addr *addr) 4478 { 4479 unsigned int hash = inet6_addr_hash(net, addr); 4480 struct inet6_ifaddr *ifp = NULL; 4481 int ret = 0; 4482 4483 rcu_read_lock(); 4484 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4485 if (ipv6_addr_equal(&ifp->addr, addr) && 4486 (ifp->flags & IFA_F_HOMEADDRESS)) { 4487 ret = 1; 4488 break; 4489 } 4490 } 4491 rcu_read_unlock(); 4492 return ret; 4493 } 4494 #endif 4495 4496 /* RFC6554 has some algorithm to avoid loops in segment routing by 4497 * checking if the segments contains any of a local interface address. 4498 * 4499 * Quote: 4500 * 4501 * To detect loops in the SRH, a router MUST determine if the SRH 4502 * includes multiple addresses assigned to any interface on that router. 4503 * If such addresses appear more than once and are separated by at least 4504 * one address not assigned to that router. 4505 */ 4506 int ipv6_chk_rpl_srh_loop(struct net *net, const struct in6_addr *segs, 4507 unsigned char nsegs) 4508 { 4509 const struct in6_addr *addr; 4510 int i, ret = 0, found = 0; 4511 struct inet6_ifaddr *ifp; 4512 bool separated = false; 4513 unsigned int hash; 4514 bool hash_found; 4515 4516 rcu_read_lock(); 4517 for (i = 0; i < nsegs; i++) { 4518 addr = &segs[i]; 4519 hash = inet6_addr_hash(net, addr); 4520 4521 hash_found = false; 4522 hlist_for_each_entry_rcu(ifp, &net->ipv6.inet6_addr_lst[hash], addr_lst) { 4523 4524 if (ipv6_addr_equal(&ifp->addr, addr)) { 4525 hash_found = true; 4526 break; 4527 } 4528 } 4529 4530 if (hash_found) { 4531 if (found > 1 && separated) { 4532 ret = 1; 4533 break; 4534 } 4535 4536 separated = false; 4537 found++; 4538 } else { 4539 separated = true; 4540 } 4541 } 4542 rcu_read_unlock(); 4543 4544 return ret; 4545 } 4546 4547 /* 4548 * Periodic address status verification 4549 */ 4550 4551 static void addrconf_verify_rtnl(struct net *net) 4552 { 4553 unsigned long now, next, next_sec, next_sched; 4554 struct inet6_ifaddr *ifp; 4555 int i; 4556 4557 ASSERT_RTNL(); 4558 4559 rcu_read_lock_bh(); 4560 now = jiffies; 4561 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY); 4562 4563 cancel_delayed_work(&net->ipv6.addr_chk_work); 4564 4565 for (i = 0; i < IN6_ADDR_HSIZE; i++) { 4566 restart: 4567 hlist_for_each_entry_rcu_bh(ifp, &net->ipv6.inet6_addr_lst[i], addr_lst) { 4568 unsigned long age; 4569 4570 /* When setting preferred_lft to a value not zero or 4571 * infinity, while valid_lft is infinity 4572 * IFA_F_PERMANENT has a non-infinity life time. 4573 */ 4574 if ((ifp->flags & IFA_F_PERMANENT) && 4575 (ifp->prefered_lft == INFINITY_LIFE_TIME)) 4576 continue; 4577 4578 spin_lock(&ifp->lock); 4579 /* We try to batch several events at once. */ 4580 age = (now - ifp->tstamp + ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 4581 4582 if ((ifp->flags&IFA_F_TEMPORARY) && 4583 !(ifp->flags&IFA_F_TENTATIVE) && 4584 ifp->prefered_lft != INFINITY_LIFE_TIME && 4585 !ifp->regen_count && ifp->ifpub) { 4586 /* This is a non-regenerated temporary addr. */ 4587 4588 unsigned long regen_advance = ifp->idev->cnf.regen_max_retry * 4589 ifp->idev->cnf.dad_transmits * 4590 max(NEIGH_VAR(ifp->idev->nd_parms, RETRANS_TIME), HZ/100) / HZ; 4591 4592 if (age + regen_advance >= ifp->prefered_lft) { 4593 struct inet6_ifaddr *ifpub = ifp->ifpub; 4594 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4595 next = ifp->tstamp + ifp->prefered_lft * HZ; 4596 4597 ifp->regen_count++; 4598 in6_ifa_hold(ifp); 4599 in6_ifa_hold(ifpub); 4600 spin_unlock(&ifp->lock); 4601 4602 spin_lock(&ifpub->lock); 4603 ifpub->regen_count = 0; 4604 spin_unlock(&ifpub->lock); 4605 rcu_read_unlock_bh(); 4606 ipv6_create_tempaddr(ifpub, true); 4607 in6_ifa_put(ifpub); 4608 in6_ifa_put(ifp); 4609 rcu_read_lock_bh(); 4610 goto restart; 4611 } else if (time_before(ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ, next)) 4612 next = ifp->tstamp + ifp->prefered_lft * HZ - regen_advance * HZ; 4613 } 4614 4615 if (ifp->valid_lft != INFINITY_LIFE_TIME && 4616 age >= ifp->valid_lft) { 4617 spin_unlock(&ifp->lock); 4618 in6_ifa_hold(ifp); 4619 rcu_read_unlock_bh(); 4620 ipv6_del_addr(ifp); 4621 rcu_read_lock_bh(); 4622 goto restart; 4623 } else if (ifp->prefered_lft == INFINITY_LIFE_TIME) { 4624 spin_unlock(&ifp->lock); 4625 continue; 4626 } else if (age >= ifp->prefered_lft) { 4627 /* jiffies - ifp->tstamp > age >= ifp->prefered_lft */ 4628 int deprecate = 0; 4629 4630 if (!(ifp->flags&IFA_F_DEPRECATED)) { 4631 deprecate = 1; 4632 ifp->flags |= IFA_F_DEPRECATED; 4633 } 4634 4635 if ((ifp->valid_lft != INFINITY_LIFE_TIME) && 4636 (time_before(ifp->tstamp + ifp->valid_lft * HZ, next))) 4637 next = ifp->tstamp + ifp->valid_lft * HZ; 4638 4639 spin_unlock(&ifp->lock); 4640 4641 if (deprecate) { 4642 in6_ifa_hold(ifp); 4643 4644 ipv6_ifa_notify(0, ifp); 4645 in6_ifa_put(ifp); 4646 goto restart; 4647 } 4648 } else { 4649 /* ifp->prefered_lft <= ifp->valid_lft */ 4650 if (time_before(ifp->tstamp + ifp->prefered_lft * HZ, next)) 4651 next = ifp->tstamp + ifp->prefered_lft * HZ; 4652 spin_unlock(&ifp->lock); 4653 } 4654 } 4655 } 4656 4657 next_sec = round_jiffies_up(next); 4658 next_sched = next; 4659 4660 /* If rounded timeout is accurate enough, accept it. */ 4661 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ)) 4662 next_sched = next_sec; 4663 4664 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */ 4665 if (time_before(next_sched, jiffies + ADDRCONF_TIMER_FUZZ_MAX)) 4666 next_sched = jiffies + ADDRCONF_TIMER_FUZZ_MAX; 4667 4668 pr_debug("now = %lu, schedule = %lu, rounded schedule = %lu => %lu\n", 4669 now, next, next_sec, next_sched); 4670 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, next_sched - now); 4671 rcu_read_unlock_bh(); 4672 } 4673 4674 static void addrconf_verify_work(struct work_struct *w) 4675 { 4676 struct net *net = container_of(to_delayed_work(w), struct net, 4677 ipv6.addr_chk_work); 4678 4679 rtnl_lock(); 4680 addrconf_verify_rtnl(net); 4681 rtnl_unlock(); 4682 } 4683 4684 static void addrconf_verify(struct net *net) 4685 { 4686 mod_delayed_work(addrconf_wq, &net->ipv6.addr_chk_work, 0); 4687 } 4688 4689 static struct in6_addr *extract_addr(struct nlattr *addr, struct nlattr *local, 4690 struct in6_addr **peer_pfx) 4691 { 4692 struct in6_addr *pfx = NULL; 4693 4694 *peer_pfx = NULL; 4695 4696 if (addr) 4697 pfx = nla_data(addr); 4698 4699 if (local) { 4700 if (pfx && nla_memcmp(local, pfx, sizeof(*pfx))) 4701 *peer_pfx = pfx; 4702 pfx = nla_data(local); 4703 } 4704 4705 return pfx; 4706 } 4707 4708 static const struct nla_policy ifa_ipv6_policy[IFA_MAX+1] = { 4709 [IFA_ADDRESS] = { .len = sizeof(struct in6_addr) }, 4710 [IFA_LOCAL] = { .len = sizeof(struct in6_addr) }, 4711 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) }, 4712 [IFA_FLAGS] = { .len = sizeof(u32) }, 4713 [IFA_RT_PRIORITY] = { .len = sizeof(u32) }, 4714 [IFA_TARGET_NETNSID] = { .type = NLA_S32 }, 4715 [IFA_PROTO] = { .type = NLA_U8 }, 4716 }; 4717 4718 static int 4719 inet6_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, 4720 struct netlink_ext_ack *extack) 4721 { 4722 struct net *net = sock_net(skb->sk); 4723 struct ifaddrmsg *ifm; 4724 struct nlattr *tb[IFA_MAX+1]; 4725 struct in6_addr *pfx, *peer_pfx; 4726 u32 ifa_flags; 4727 int err; 4728 4729 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 4730 ifa_ipv6_policy, extack); 4731 if (err < 0) 4732 return err; 4733 4734 ifm = nlmsg_data(nlh); 4735 pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 4736 if (!pfx) 4737 return -EINVAL; 4738 4739 ifa_flags = tb[IFA_FLAGS] ? nla_get_u32(tb[IFA_FLAGS]) : ifm->ifa_flags; 4740 4741 /* We ignore other flags so far. */ 4742 ifa_flags &= IFA_F_MANAGETEMPADDR; 4743 4744 return inet6_addr_del(net, ifm->ifa_index, ifa_flags, pfx, 4745 ifm->ifa_prefixlen, extack); 4746 } 4747 4748 static int modify_prefix_route(struct inet6_ifaddr *ifp, 4749 unsigned long expires, u32 flags, 4750 bool modify_peer) 4751 { 4752 struct fib6_info *f6i; 4753 u32 prio; 4754 4755 f6i = addrconf_get_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4756 ifp->prefix_len, 4757 ifp->idev->dev, 0, RTF_DEFAULT, true); 4758 if (!f6i) 4759 return -ENOENT; 4760 4761 prio = ifp->rt_priority ? : IP6_RT_PRIO_ADDRCONF; 4762 if (f6i->fib6_metric != prio) { 4763 /* delete old one */ 4764 ip6_del_rt(dev_net(ifp->idev->dev), f6i, false); 4765 4766 /* add new one */ 4767 addrconf_prefix_route(modify_peer ? &ifp->peer_addr : &ifp->addr, 4768 ifp->prefix_len, 4769 ifp->rt_priority, ifp->idev->dev, 4770 expires, flags, GFP_KERNEL); 4771 } else { 4772 if (!expires) 4773 fib6_clean_expires(f6i); 4774 else 4775 fib6_set_expires(f6i, expires); 4776 4777 fib6_info_release(f6i); 4778 } 4779 4780 return 0; 4781 } 4782 4783 static int inet6_addr_modify(struct net *net, struct inet6_ifaddr *ifp, 4784 struct ifa6_config *cfg) 4785 { 4786 u32 flags; 4787 clock_t expires; 4788 unsigned long timeout; 4789 bool was_managetempaddr; 4790 bool had_prefixroute; 4791 bool new_peer = false; 4792 4793 ASSERT_RTNL(); 4794 4795 if (!cfg->valid_lft || cfg->preferred_lft > cfg->valid_lft) 4796 return -EINVAL; 4797 4798 if (cfg->ifa_flags & IFA_F_MANAGETEMPADDR && 4799 (ifp->flags & IFA_F_TEMPORARY || ifp->prefix_len != 64)) 4800 return -EINVAL; 4801 4802 if (!(ifp->flags & IFA_F_TENTATIVE) || ifp->flags & IFA_F_DADFAILED) 4803 cfg->ifa_flags &= ~IFA_F_OPTIMISTIC; 4804 4805 timeout = addrconf_timeout_fixup(cfg->valid_lft, HZ); 4806 if (addrconf_finite_timeout(timeout)) { 4807 expires = jiffies_to_clock_t(timeout * HZ); 4808 cfg->valid_lft = timeout; 4809 flags = RTF_EXPIRES; 4810 } else { 4811 expires = 0; 4812 flags = 0; 4813 cfg->ifa_flags |= IFA_F_PERMANENT; 4814 } 4815 4816 timeout = addrconf_timeout_fixup(cfg->preferred_lft, HZ); 4817 if (addrconf_finite_timeout(timeout)) { 4818 if (timeout == 0) 4819 cfg->ifa_flags |= IFA_F_DEPRECATED; 4820 cfg->preferred_lft = timeout; 4821 } 4822 4823 if (cfg->peer_pfx && 4824 memcmp(&ifp->peer_addr, cfg->peer_pfx, sizeof(struct in6_addr))) { 4825 if (!ipv6_addr_any(&ifp->peer_addr)) 4826 cleanup_prefix_route(ifp, expires, true, true); 4827 new_peer = true; 4828 } 4829 4830 spin_lock_bh(&ifp->lock); 4831 was_managetempaddr = ifp->flags & IFA_F_MANAGETEMPADDR; 4832 had_prefixroute = ifp->flags & IFA_F_PERMANENT && 4833 !(ifp->flags & IFA_F_NOPREFIXROUTE); 4834 ifp->flags &= ~(IFA_F_DEPRECATED | IFA_F_PERMANENT | IFA_F_NODAD | 4835 IFA_F_HOMEADDRESS | IFA_F_MANAGETEMPADDR | 4836 IFA_F_NOPREFIXROUTE); 4837 ifp->flags |= cfg->ifa_flags; 4838 ifp->tstamp = jiffies; 4839 ifp->valid_lft = cfg->valid_lft; 4840 ifp->prefered_lft = cfg->preferred_lft; 4841 ifp->ifa_proto = cfg->ifa_proto; 4842 4843 if (cfg->rt_priority && cfg->rt_priority != ifp->rt_priority) 4844 ifp->rt_priority = cfg->rt_priority; 4845 4846 if (new_peer) 4847 ifp->peer_addr = *cfg->peer_pfx; 4848 4849 spin_unlock_bh(&ifp->lock); 4850 if (!(ifp->flags&IFA_F_TENTATIVE)) 4851 ipv6_ifa_notify(0, ifp); 4852 4853 if (!(cfg->ifa_flags & IFA_F_NOPREFIXROUTE)) { 4854 int rc = -ENOENT; 4855 4856 if (had_prefixroute) 4857 rc = modify_prefix_route(ifp, expires, flags, false); 4858 4859 /* prefix route could have been deleted; if so restore it */ 4860 if (rc == -ENOENT) { 4861 addrconf_prefix_route(&ifp->addr, ifp->prefix_len, 4862 ifp->rt_priority, ifp->idev->dev, 4863 expires, flags, GFP_KERNEL); 4864 } 4865 4866 if (had_prefixroute && !ipv6_addr_any(&ifp->peer_addr)) 4867 rc = modify_prefix_route(ifp, expires, flags, true); 4868 4869 if (rc == -ENOENT && !ipv6_addr_any(&ifp->peer_addr)) { 4870 addrconf_prefix_route(&ifp->peer_addr, ifp->prefix_len, 4871 ifp->rt_priority, ifp->idev->dev, 4872 expires, flags, GFP_KERNEL); 4873 } 4874 } else if (had_prefixroute) { 4875 enum cleanup_prefix_rt_t action; 4876 unsigned long rt_expires; 4877 4878 write_lock_bh(&ifp->idev->lock); 4879 action = check_cleanup_prefix_route(ifp, &rt_expires); 4880 write_unlock_bh(&ifp->idev->lock); 4881 4882 if (action != CLEANUP_PREFIX_RT_NOP) { 4883 cleanup_prefix_route(ifp, rt_expires, 4884 action == CLEANUP_PREFIX_RT_DEL, false); 4885 } 4886 } 4887 4888 if (was_managetempaddr || ifp->flags & IFA_F_MANAGETEMPADDR) { 4889 if (was_managetempaddr && 4890 !(ifp->flags & IFA_F_MANAGETEMPADDR)) { 4891 cfg->valid_lft = 0; 4892 cfg->preferred_lft = 0; 4893 } 4894 manage_tempaddrs(ifp->idev, ifp, cfg->valid_lft, 4895 cfg->preferred_lft, !was_managetempaddr, 4896 jiffies); 4897 } 4898 4899 addrconf_verify_rtnl(net); 4900 4901 return 0; 4902 } 4903 4904 static int 4905 inet6_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, 4906 struct netlink_ext_ack *extack) 4907 { 4908 struct net *net = sock_net(skb->sk); 4909 struct ifaddrmsg *ifm; 4910 struct nlattr *tb[IFA_MAX+1]; 4911 struct in6_addr *peer_pfx; 4912 struct inet6_ifaddr *ifa; 4913 struct net_device *dev; 4914 struct inet6_dev *idev; 4915 struct ifa6_config cfg; 4916 int err; 4917 4918 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 4919 ifa_ipv6_policy, extack); 4920 if (err < 0) 4921 return err; 4922 4923 memset(&cfg, 0, sizeof(cfg)); 4924 4925 ifm = nlmsg_data(nlh); 4926 cfg.pfx = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer_pfx); 4927 if (!cfg.pfx) 4928 return -EINVAL; 4929 4930 cfg.peer_pfx = peer_pfx; 4931 cfg.plen = ifm->ifa_prefixlen; 4932 if (tb[IFA_RT_PRIORITY]) 4933 cfg.rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]); 4934 4935 if (tb[IFA_PROTO]) 4936 cfg.ifa_proto = nla_get_u8(tb[IFA_PROTO]); 4937 4938 cfg.valid_lft = INFINITY_LIFE_TIME; 4939 cfg.preferred_lft = INFINITY_LIFE_TIME; 4940 4941 if (tb[IFA_CACHEINFO]) { 4942 struct ifa_cacheinfo *ci; 4943 4944 ci = nla_data(tb[IFA_CACHEINFO]); 4945 cfg.valid_lft = ci->ifa_valid; 4946 cfg.preferred_lft = ci->ifa_prefered; 4947 } 4948 4949 dev = __dev_get_by_index(net, ifm->ifa_index); 4950 if (!dev) { 4951 NL_SET_ERR_MSG_MOD(extack, "Unable to find the interface"); 4952 return -ENODEV; 4953 } 4954 4955 if (tb[IFA_FLAGS]) 4956 cfg.ifa_flags = nla_get_u32(tb[IFA_FLAGS]); 4957 else 4958 cfg.ifa_flags = ifm->ifa_flags; 4959 4960 /* We ignore other flags so far. */ 4961 cfg.ifa_flags &= IFA_F_NODAD | IFA_F_HOMEADDRESS | 4962 IFA_F_MANAGETEMPADDR | IFA_F_NOPREFIXROUTE | 4963 IFA_F_MCAUTOJOIN | IFA_F_OPTIMISTIC; 4964 4965 idev = ipv6_find_idev(dev); 4966 if (IS_ERR(idev)) 4967 return PTR_ERR(idev); 4968 4969 if (!ipv6_allow_optimistic_dad(net, idev)) 4970 cfg.ifa_flags &= ~IFA_F_OPTIMISTIC; 4971 4972 if (cfg.ifa_flags & IFA_F_NODAD && 4973 cfg.ifa_flags & IFA_F_OPTIMISTIC) { 4974 NL_SET_ERR_MSG(extack, "IFA_F_NODAD and IFA_F_OPTIMISTIC are mutually exclusive"); 4975 return -EINVAL; 4976 } 4977 4978 ifa = ipv6_get_ifaddr(net, cfg.pfx, dev, 1); 4979 if (!ifa) { 4980 /* 4981 * It would be best to check for !NLM_F_CREATE here but 4982 * userspace already relies on not having to provide this. 4983 */ 4984 return inet6_addr_add(net, ifm->ifa_index, &cfg, extack); 4985 } 4986 4987 if (nlh->nlmsg_flags & NLM_F_EXCL || 4988 !(nlh->nlmsg_flags & NLM_F_REPLACE)) { 4989 NL_SET_ERR_MSG_MOD(extack, "address already assigned"); 4990 err = -EEXIST; 4991 } else { 4992 err = inet6_addr_modify(net, ifa, &cfg); 4993 } 4994 4995 in6_ifa_put(ifa); 4996 4997 return err; 4998 } 4999 5000 static void put_ifaddrmsg(struct nlmsghdr *nlh, u8 prefixlen, u32 flags, 5001 u8 scope, int ifindex) 5002 { 5003 struct ifaddrmsg *ifm; 5004 5005 ifm = nlmsg_data(nlh); 5006 ifm->ifa_family = AF_INET6; 5007 ifm->ifa_prefixlen = prefixlen; 5008 ifm->ifa_flags = flags; 5009 ifm->ifa_scope = scope; 5010 ifm->ifa_index = ifindex; 5011 } 5012 5013 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp, 5014 unsigned long tstamp, u32 preferred, u32 valid) 5015 { 5016 struct ifa_cacheinfo ci; 5017 5018 ci.cstamp = cstamp_delta(cstamp); 5019 ci.tstamp = cstamp_delta(tstamp); 5020 ci.ifa_prefered = preferred; 5021 ci.ifa_valid = valid; 5022 5023 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci); 5024 } 5025 5026 static inline int rt_scope(int ifa_scope) 5027 { 5028 if (ifa_scope & IFA_HOST) 5029 return RT_SCOPE_HOST; 5030 else if (ifa_scope & IFA_LINK) 5031 return RT_SCOPE_LINK; 5032 else if (ifa_scope & IFA_SITE) 5033 return RT_SCOPE_SITE; 5034 else 5035 return RT_SCOPE_UNIVERSE; 5036 } 5037 5038 static inline int inet6_ifaddr_msgsize(void) 5039 { 5040 return NLMSG_ALIGN(sizeof(struct ifaddrmsg)) 5041 + nla_total_size(16) /* IFA_LOCAL */ 5042 + nla_total_size(16) /* IFA_ADDRESS */ 5043 + nla_total_size(sizeof(struct ifa_cacheinfo)) 5044 + nla_total_size(4) /* IFA_FLAGS */ 5045 + nla_total_size(1) /* IFA_PROTO */ 5046 + nla_total_size(4) /* IFA_RT_PRIORITY */; 5047 } 5048 5049 enum addr_type_t { 5050 UNICAST_ADDR, 5051 MULTICAST_ADDR, 5052 ANYCAST_ADDR, 5053 }; 5054 5055 struct inet6_fill_args { 5056 u32 portid; 5057 u32 seq; 5058 int event; 5059 unsigned int flags; 5060 int netnsid; 5061 int ifindex; 5062 enum addr_type_t type; 5063 }; 5064 5065 static int inet6_fill_ifaddr(struct sk_buff *skb, struct inet6_ifaddr *ifa, 5066 struct inet6_fill_args *args) 5067 { 5068 struct nlmsghdr *nlh; 5069 u32 preferred, valid; 5070 5071 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5072 sizeof(struct ifaddrmsg), args->flags); 5073 if (!nlh) 5074 return -EMSGSIZE; 5075 5076 put_ifaddrmsg(nlh, ifa->prefix_len, ifa->flags, rt_scope(ifa->scope), 5077 ifa->idev->dev->ifindex); 5078 5079 if (args->netnsid >= 0 && 5080 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) 5081 goto error; 5082 5083 spin_lock_bh(&ifa->lock); 5084 if (!((ifa->flags&IFA_F_PERMANENT) && 5085 (ifa->prefered_lft == INFINITY_LIFE_TIME))) { 5086 preferred = ifa->prefered_lft; 5087 valid = ifa->valid_lft; 5088 if (preferred != INFINITY_LIFE_TIME) { 5089 long tval = (jiffies - ifa->tstamp)/HZ; 5090 if (preferred > tval) 5091 preferred -= tval; 5092 else 5093 preferred = 0; 5094 if (valid != INFINITY_LIFE_TIME) { 5095 if (valid > tval) 5096 valid -= tval; 5097 else 5098 valid = 0; 5099 } 5100 } 5101 } else { 5102 preferred = INFINITY_LIFE_TIME; 5103 valid = INFINITY_LIFE_TIME; 5104 } 5105 spin_unlock_bh(&ifa->lock); 5106 5107 if (!ipv6_addr_any(&ifa->peer_addr)) { 5108 if (nla_put_in6_addr(skb, IFA_LOCAL, &ifa->addr) < 0 || 5109 nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->peer_addr) < 0) 5110 goto error; 5111 } else 5112 if (nla_put_in6_addr(skb, IFA_ADDRESS, &ifa->addr) < 0) 5113 goto error; 5114 5115 if (ifa->rt_priority && 5116 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->rt_priority)) 5117 goto error; 5118 5119 if (put_cacheinfo(skb, ifa->cstamp, ifa->tstamp, preferred, valid) < 0) 5120 goto error; 5121 5122 if (nla_put_u32(skb, IFA_FLAGS, ifa->flags) < 0) 5123 goto error; 5124 5125 if (ifa->ifa_proto && 5126 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto)) 5127 goto error; 5128 5129 nlmsg_end(skb, nlh); 5130 return 0; 5131 5132 error: 5133 nlmsg_cancel(skb, nlh); 5134 return -EMSGSIZE; 5135 } 5136 5137 static int inet6_fill_ifmcaddr(struct sk_buff *skb, struct ifmcaddr6 *ifmca, 5138 struct inet6_fill_args *args) 5139 { 5140 struct nlmsghdr *nlh; 5141 u8 scope = RT_SCOPE_UNIVERSE; 5142 int ifindex = ifmca->idev->dev->ifindex; 5143 5144 if (ipv6_addr_scope(&ifmca->mca_addr) & IFA_SITE) 5145 scope = RT_SCOPE_SITE; 5146 5147 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5148 sizeof(struct ifaddrmsg), args->flags); 5149 if (!nlh) 5150 return -EMSGSIZE; 5151 5152 if (args->netnsid >= 0 && 5153 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) { 5154 nlmsg_cancel(skb, nlh); 5155 return -EMSGSIZE; 5156 } 5157 5158 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 5159 if (nla_put_in6_addr(skb, IFA_MULTICAST, &ifmca->mca_addr) < 0 || 5160 put_cacheinfo(skb, ifmca->mca_cstamp, ifmca->mca_tstamp, 5161 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5162 nlmsg_cancel(skb, nlh); 5163 return -EMSGSIZE; 5164 } 5165 5166 nlmsg_end(skb, nlh); 5167 return 0; 5168 } 5169 5170 static int inet6_fill_ifacaddr(struct sk_buff *skb, struct ifacaddr6 *ifaca, 5171 struct inet6_fill_args *args) 5172 { 5173 struct net_device *dev = fib6_info_nh_dev(ifaca->aca_rt); 5174 int ifindex = dev ? dev->ifindex : 1; 5175 struct nlmsghdr *nlh; 5176 u8 scope = RT_SCOPE_UNIVERSE; 5177 5178 if (ipv6_addr_scope(&ifaca->aca_addr) & IFA_SITE) 5179 scope = RT_SCOPE_SITE; 5180 5181 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, 5182 sizeof(struct ifaddrmsg), args->flags); 5183 if (!nlh) 5184 return -EMSGSIZE; 5185 5186 if (args->netnsid >= 0 && 5187 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) { 5188 nlmsg_cancel(skb, nlh); 5189 return -EMSGSIZE; 5190 } 5191 5192 put_ifaddrmsg(nlh, 128, IFA_F_PERMANENT, scope, ifindex); 5193 if (nla_put_in6_addr(skb, IFA_ANYCAST, &ifaca->aca_addr) < 0 || 5194 put_cacheinfo(skb, ifaca->aca_cstamp, ifaca->aca_tstamp, 5195 INFINITY_LIFE_TIME, INFINITY_LIFE_TIME) < 0) { 5196 nlmsg_cancel(skb, nlh); 5197 return -EMSGSIZE; 5198 } 5199 5200 nlmsg_end(skb, nlh); 5201 return 0; 5202 } 5203 5204 /* called with rcu_read_lock() */ 5205 static int in6_dump_addrs(struct inet6_dev *idev, struct sk_buff *skb, 5206 struct netlink_callback *cb, int s_ip_idx, 5207 struct inet6_fill_args *fillargs) 5208 { 5209 struct ifmcaddr6 *ifmca; 5210 struct ifacaddr6 *ifaca; 5211 int ip_idx = 0; 5212 int err = 1; 5213 5214 read_lock_bh(&idev->lock); 5215 switch (fillargs->type) { 5216 case UNICAST_ADDR: { 5217 struct inet6_ifaddr *ifa; 5218 fillargs->event = RTM_NEWADDR; 5219 5220 /* unicast address incl. temp addr */ 5221 list_for_each_entry(ifa, &idev->addr_list, if_list) { 5222 if (ip_idx < s_ip_idx) 5223 goto next; 5224 err = inet6_fill_ifaddr(skb, ifa, fillargs); 5225 if (err < 0) 5226 break; 5227 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 5228 next: 5229 ip_idx++; 5230 } 5231 break; 5232 } 5233 case MULTICAST_ADDR: 5234 read_unlock_bh(&idev->lock); 5235 fillargs->event = RTM_GETMULTICAST; 5236 5237 /* multicast address */ 5238 for (ifmca = rtnl_dereference(idev->mc_list); 5239 ifmca; 5240 ifmca = rtnl_dereference(ifmca->next), ip_idx++) { 5241 if (ip_idx < s_ip_idx) 5242 continue; 5243 err = inet6_fill_ifmcaddr(skb, ifmca, fillargs); 5244 if (err < 0) 5245 break; 5246 } 5247 read_lock_bh(&idev->lock); 5248 break; 5249 case ANYCAST_ADDR: 5250 fillargs->event = RTM_GETANYCAST; 5251 /* anycast address */ 5252 for (ifaca = idev->ac_list; ifaca; 5253 ifaca = ifaca->aca_next, ip_idx++) { 5254 if (ip_idx < s_ip_idx) 5255 continue; 5256 err = inet6_fill_ifacaddr(skb, ifaca, fillargs); 5257 if (err < 0) 5258 break; 5259 } 5260 break; 5261 default: 5262 break; 5263 } 5264 read_unlock_bh(&idev->lock); 5265 cb->args[2] = ip_idx; 5266 return err; 5267 } 5268 5269 static int inet6_valid_dump_ifaddr_req(const struct nlmsghdr *nlh, 5270 struct inet6_fill_args *fillargs, 5271 struct net **tgt_net, struct sock *sk, 5272 struct netlink_callback *cb) 5273 { 5274 struct netlink_ext_ack *extack = cb->extack; 5275 struct nlattr *tb[IFA_MAX+1]; 5276 struct ifaddrmsg *ifm; 5277 int err, i; 5278 5279 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) { 5280 NL_SET_ERR_MSG_MOD(extack, "Invalid header for address dump request"); 5281 return -EINVAL; 5282 } 5283 5284 ifm = nlmsg_data(nlh); 5285 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5286 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for address dump request"); 5287 return -EINVAL; 5288 } 5289 5290 fillargs->ifindex = ifm->ifa_index; 5291 if (fillargs->ifindex) { 5292 cb->answer_flags |= NLM_F_DUMP_FILTERED; 5293 fillargs->flags |= NLM_F_DUMP_FILTERED; 5294 } 5295 5296 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5297 ifa_ipv6_policy, extack); 5298 if (err < 0) 5299 return err; 5300 5301 for (i = 0; i <= IFA_MAX; ++i) { 5302 if (!tb[i]) 5303 continue; 5304 5305 if (i == IFA_TARGET_NETNSID) { 5306 struct net *net; 5307 5308 fillargs->netnsid = nla_get_s32(tb[i]); 5309 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid); 5310 if (IS_ERR(net)) { 5311 fillargs->netnsid = -1; 5312 NL_SET_ERR_MSG_MOD(extack, "Invalid target network namespace id"); 5313 return PTR_ERR(net); 5314 } 5315 *tgt_net = net; 5316 } else { 5317 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in dump request"); 5318 return -EINVAL; 5319 } 5320 } 5321 5322 return 0; 5323 } 5324 5325 static int inet6_dump_addr(struct sk_buff *skb, struct netlink_callback *cb, 5326 enum addr_type_t type) 5327 { 5328 const struct nlmsghdr *nlh = cb->nlh; 5329 struct inet6_fill_args fillargs = { 5330 .portid = NETLINK_CB(cb->skb).portid, 5331 .seq = cb->nlh->nlmsg_seq, 5332 .flags = NLM_F_MULTI, 5333 .netnsid = -1, 5334 .type = type, 5335 }; 5336 struct net *tgt_net = sock_net(skb->sk); 5337 int idx, s_idx, s_ip_idx; 5338 int h, s_h; 5339 struct net_device *dev; 5340 struct inet6_dev *idev; 5341 struct hlist_head *head; 5342 int err = 0; 5343 5344 s_h = cb->args[0]; 5345 s_idx = idx = cb->args[1]; 5346 s_ip_idx = cb->args[2]; 5347 5348 if (cb->strict_check) { 5349 err = inet6_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net, 5350 skb->sk, cb); 5351 if (err < 0) 5352 goto put_tgt_net; 5353 5354 err = 0; 5355 if (fillargs.ifindex) { 5356 dev = __dev_get_by_index(tgt_net, fillargs.ifindex); 5357 if (!dev) { 5358 err = -ENODEV; 5359 goto put_tgt_net; 5360 } 5361 idev = __in6_dev_get(dev); 5362 if (idev) { 5363 err = in6_dump_addrs(idev, skb, cb, s_ip_idx, 5364 &fillargs); 5365 if (err > 0) 5366 err = 0; 5367 } 5368 goto put_tgt_net; 5369 } 5370 } 5371 5372 rcu_read_lock(); 5373 cb->seq = atomic_read(&tgt_net->ipv6.dev_addr_genid) ^ tgt_net->dev_base_seq; 5374 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 5375 idx = 0; 5376 head = &tgt_net->dev_index_head[h]; 5377 hlist_for_each_entry_rcu(dev, head, index_hlist) { 5378 if (idx < s_idx) 5379 goto cont; 5380 if (h > s_h || idx > s_idx) 5381 s_ip_idx = 0; 5382 idev = __in6_dev_get(dev); 5383 if (!idev) 5384 goto cont; 5385 5386 if (in6_dump_addrs(idev, skb, cb, s_ip_idx, 5387 &fillargs) < 0) 5388 goto done; 5389 cont: 5390 idx++; 5391 } 5392 } 5393 done: 5394 rcu_read_unlock(); 5395 cb->args[0] = h; 5396 cb->args[1] = idx; 5397 put_tgt_net: 5398 if (fillargs.netnsid >= 0) 5399 put_net(tgt_net); 5400 5401 return skb->len ? : err; 5402 } 5403 5404 static int inet6_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb) 5405 { 5406 enum addr_type_t type = UNICAST_ADDR; 5407 5408 return inet6_dump_addr(skb, cb, type); 5409 } 5410 5411 static int inet6_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb) 5412 { 5413 enum addr_type_t type = MULTICAST_ADDR; 5414 5415 return inet6_dump_addr(skb, cb, type); 5416 } 5417 5418 5419 static int inet6_dump_ifacaddr(struct sk_buff *skb, struct netlink_callback *cb) 5420 { 5421 enum addr_type_t type = ANYCAST_ADDR; 5422 5423 return inet6_dump_addr(skb, cb, type); 5424 } 5425 5426 static int inet6_rtm_valid_getaddr_req(struct sk_buff *skb, 5427 const struct nlmsghdr *nlh, 5428 struct nlattr **tb, 5429 struct netlink_ext_ack *extack) 5430 { 5431 struct ifaddrmsg *ifm; 5432 int i, err; 5433 5434 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) { 5435 NL_SET_ERR_MSG_MOD(extack, "Invalid header for get address request"); 5436 return -EINVAL; 5437 } 5438 5439 if (!netlink_strict_get_check(skb)) 5440 return nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 5441 ifa_ipv6_policy, extack); 5442 5443 ifm = nlmsg_data(nlh); 5444 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 5445 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get address request"); 5446 return -EINVAL; 5447 } 5448 5449 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 5450 ifa_ipv6_policy, extack); 5451 if (err) 5452 return err; 5453 5454 for (i = 0; i <= IFA_MAX; i++) { 5455 if (!tb[i]) 5456 continue; 5457 5458 switch (i) { 5459 case IFA_TARGET_NETNSID: 5460 case IFA_ADDRESS: 5461 case IFA_LOCAL: 5462 break; 5463 default: 5464 NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get address request"); 5465 return -EINVAL; 5466 } 5467 } 5468 5469 return 0; 5470 } 5471 5472 static int inet6_rtm_getaddr(struct sk_buff *in_skb, struct nlmsghdr *nlh, 5473 struct netlink_ext_ack *extack) 5474 { 5475 struct net *tgt_net = sock_net(in_skb->sk); 5476 struct inet6_fill_args fillargs = { 5477 .portid = NETLINK_CB(in_skb).portid, 5478 .seq = nlh->nlmsg_seq, 5479 .event = RTM_NEWADDR, 5480 .flags = 0, 5481 .netnsid = -1, 5482 }; 5483 struct ifaddrmsg *ifm; 5484 struct nlattr *tb[IFA_MAX+1]; 5485 struct in6_addr *addr = NULL, *peer; 5486 struct net_device *dev = NULL; 5487 struct inet6_ifaddr *ifa; 5488 struct sk_buff *skb; 5489 int err; 5490 5491 err = inet6_rtm_valid_getaddr_req(in_skb, nlh, tb, extack); 5492 if (err < 0) 5493 return err; 5494 5495 if (tb[IFA_TARGET_NETNSID]) { 5496 fillargs.netnsid = nla_get_s32(tb[IFA_TARGET_NETNSID]); 5497 5498 tgt_net = rtnl_get_net_ns_capable(NETLINK_CB(in_skb).sk, 5499 fillargs.netnsid); 5500 if (IS_ERR(tgt_net)) 5501 return PTR_ERR(tgt_net); 5502 } 5503 5504 addr = extract_addr(tb[IFA_ADDRESS], tb[IFA_LOCAL], &peer); 5505 if (!addr) 5506 return -EINVAL; 5507 5508 ifm = nlmsg_data(nlh); 5509 if (ifm->ifa_index) 5510 dev = dev_get_by_index(tgt_net, ifm->ifa_index); 5511 5512 ifa = ipv6_get_ifaddr(tgt_net, addr, dev, 1); 5513 if (!ifa) { 5514 err = -EADDRNOTAVAIL; 5515 goto errout; 5516 } 5517 5518 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_KERNEL); 5519 if (!skb) { 5520 err = -ENOBUFS; 5521 goto errout_ifa; 5522 } 5523 5524 err = inet6_fill_ifaddr(skb, ifa, &fillargs); 5525 if (err < 0) { 5526 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 5527 WARN_ON(err == -EMSGSIZE); 5528 kfree_skb(skb); 5529 goto errout_ifa; 5530 } 5531 err = rtnl_unicast(skb, tgt_net, NETLINK_CB(in_skb).portid); 5532 errout_ifa: 5533 in6_ifa_put(ifa); 5534 errout: 5535 dev_put(dev); 5536 if (fillargs.netnsid >= 0) 5537 put_net(tgt_net); 5538 5539 return err; 5540 } 5541 5542 static void inet6_ifa_notify(int event, struct inet6_ifaddr *ifa) 5543 { 5544 struct sk_buff *skb; 5545 struct net *net = dev_net(ifa->idev->dev); 5546 struct inet6_fill_args fillargs = { 5547 .portid = 0, 5548 .seq = 0, 5549 .event = event, 5550 .flags = 0, 5551 .netnsid = -1, 5552 }; 5553 int err = -ENOBUFS; 5554 5555 skb = nlmsg_new(inet6_ifaddr_msgsize(), GFP_ATOMIC); 5556 if (!skb) 5557 goto errout; 5558 5559 err = inet6_fill_ifaddr(skb, ifa, &fillargs); 5560 if (err < 0) { 5561 /* -EMSGSIZE implies BUG in inet6_ifaddr_msgsize() */ 5562 WARN_ON(err == -EMSGSIZE); 5563 kfree_skb(skb); 5564 goto errout; 5565 } 5566 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFADDR, NULL, GFP_ATOMIC); 5567 return; 5568 errout: 5569 if (err < 0) 5570 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFADDR, err); 5571 } 5572 5573 static inline void ipv6_store_devconf(struct ipv6_devconf *cnf, 5574 __s32 *array, int bytes) 5575 { 5576 BUG_ON(bytes < (DEVCONF_MAX * 4)); 5577 5578 memset(array, 0, bytes); 5579 array[DEVCONF_FORWARDING] = cnf->forwarding; 5580 array[DEVCONF_HOPLIMIT] = cnf->hop_limit; 5581 array[DEVCONF_MTU6] = cnf->mtu6; 5582 array[DEVCONF_ACCEPT_RA] = cnf->accept_ra; 5583 array[DEVCONF_ACCEPT_REDIRECTS] = cnf->accept_redirects; 5584 array[DEVCONF_AUTOCONF] = cnf->autoconf; 5585 array[DEVCONF_DAD_TRANSMITS] = cnf->dad_transmits; 5586 array[DEVCONF_RTR_SOLICITS] = cnf->rtr_solicits; 5587 array[DEVCONF_RTR_SOLICIT_INTERVAL] = 5588 jiffies_to_msecs(cnf->rtr_solicit_interval); 5589 array[DEVCONF_RTR_SOLICIT_MAX_INTERVAL] = 5590 jiffies_to_msecs(cnf->rtr_solicit_max_interval); 5591 array[DEVCONF_RTR_SOLICIT_DELAY] = 5592 jiffies_to_msecs(cnf->rtr_solicit_delay); 5593 array[DEVCONF_FORCE_MLD_VERSION] = cnf->force_mld_version; 5594 array[DEVCONF_MLDV1_UNSOLICITED_REPORT_INTERVAL] = 5595 jiffies_to_msecs(cnf->mldv1_unsolicited_report_interval); 5596 array[DEVCONF_MLDV2_UNSOLICITED_REPORT_INTERVAL] = 5597 jiffies_to_msecs(cnf->mldv2_unsolicited_report_interval); 5598 array[DEVCONF_USE_TEMPADDR] = cnf->use_tempaddr; 5599 array[DEVCONF_TEMP_VALID_LFT] = cnf->temp_valid_lft; 5600 array[DEVCONF_TEMP_PREFERED_LFT] = cnf->temp_prefered_lft; 5601 array[DEVCONF_REGEN_MAX_RETRY] = cnf->regen_max_retry; 5602 array[DEVCONF_MAX_DESYNC_FACTOR] = cnf->max_desync_factor; 5603 array[DEVCONF_MAX_ADDRESSES] = cnf->max_addresses; 5604 array[DEVCONF_ACCEPT_RA_DEFRTR] = cnf->accept_ra_defrtr; 5605 array[DEVCONF_RA_DEFRTR_METRIC] = cnf->ra_defrtr_metric; 5606 array[DEVCONF_ACCEPT_RA_MIN_HOP_LIMIT] = cnf->accept_ra_min_hop_limit; 5607 array[DEVCONF_ACCEPT_RA_PINFO] = cnf->accept_ra_pinfo; 5608 #ifdef CONFIG_IPV6_ROUTER_PREF 5609 array[DEVCONF_ACCEPT_RA_RTR_PREF] = cnf->accept_ra_rtr_pref; 5610 array[DEVCONF_RTR_PROBE_INTERVAL] = 5611 jiffies_to_msecs(cnf->rtr_probe_interval); 5612 #ifdef CONFIG_IPV6_ROUTE_INFO 5613 array[DEVCONF_ACCEPT_RA_RT_INFO_MIN_PLEN] = cnf->accept_ra_rt_info_min_plen; 5614 array[DEVCONF_ACCEPT_RA_RT_INFO_MAX_PLEN] = cnf->accept_ra_rt_info_max_plen; 5615 #endif 5616 #endif 5617 array[DEVCONF_PROXY_NDP] = cnf->proxy_ndp; 5618 array[DEVCONF_ACCEPT_SOURCE_ROUTE] = cnf->accept_source_route; 5619 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 5620 array[DEVCONF_OPTIMISTIC_DAD] = cnf->optimistic_dad; 5621 array[DEVCONF_USE_OPTIMISTIC] = cnf->use_optimistic; 5622 #endif 5623 #ifdef CONFIG_IPV6_MROUTE 5624 array[DEVCONF_MC_FORWARDING] = atomic_read(&cnf->mc_forwarding); 5625 #endif 5626 array[DEVCONF_DISABLE_IPV6] = cnf->disable_ipv6; 5627 array[DEVCONF_ACCEPT_DAD] = cnf->accept_dad; 5628 array[DEVCONF_FORCE_TLLAO] = cnf->force_tllao; 5629 array[DEVCONF_NDISC_NOTIFY] = cnf->ndisc_notify; 5630 array[DEVCONF_SUPPRESS_FRAG_NDISC] = cnf->suppress_frag_ndisc; 5631 array[DEVCONF_ACCEPT_RA_FROM_LOCAL] = cnf->accept_ra_from_local; 5632 array[DEVCONF_ACCEPT_RA_MTU] = cnf->accept_ra_mtu; 5633 array[DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN] = cnf->ignore_routes_with_linkdown; 5634 /* we omit DEVCONF_STABLE_SECRET for now */ 5635 array[DEVCONF_USE_OIF_ADDRS_ONLY] = cnf->use_oif_addrs_only; 5636 array[DEVCONF_DROP_UNICAST_IN_L2_MULTICAST] = cnf->drop_unicast_in_l2_multicast; 5637 array[DEVCONF_DROP_UNSOLICITED_NA] = cnf->drop_unsolicited_na; 5638 array[DEVCONF_KEEP_ADDR_ON_DOWN] = cnf->keep_addr_on_down; 5639 array[DEVCONF_SEG6_ENABLED] = cnf->seg6_enabled; 5640 #ifdef CONFIG_IPV6_SEG6_HMAC 5641 array[DEVCONF_SEG6_REQUIRE_HMAC] = cnf->seg6_require_hmac; 5642 #endif 5643 array[DEVCONF_ENHANCED_DAD] = cnf->enhanced_dad; 5644 array[DEVCONF_ADDR_GEN_MODE] = cnf->addr_gen_mode; 5645 array[DEVCONF_DISABLE_POLICY] = cnf->disable_policy; 5646 array[DEVCONF_NDISC_TCLASS] = cnf->ndisc_tclass; 5647 array[DEVCONF_RPL_SEG_ENABLED] = cnf->rpl_seg_enabled; 5648 array[DEVCONF_IOAM6_ENABLED] = cnf->ioam6_enabled; 5649 array[DEVCONF_IOAM6_ID] = cnf->ioam6_id; 5650 array[DEVCONF_IOAM6_ID_WIDE] = cnf->ioam6_id_wide; 5651 array[DEVCONF_NDISC_EVICT_NOCARRIER] = cnf->ndisc_evict_nocarrier; 5652 array[DEVCONF_ACCEPT_UNTRACKED_NA] = cnf->accept_untracked_na; 5653 array[DEVCONF_ACCEPT_RA_MIN_LFT] = cnf->accept_ra_min_lft; 5654 } 5655 5656 static inline size_t inet6_ifla6_size(void) 5657 { 5658 return nla_total_size(4) /* IFLA_INET6_FLAGS */ 5659 + nla_total_size(sizeof(struct ifla_cacheinfo)) 5660 + nla_total_size(DEVCONF_MAX * 4) /* IFLA_INET6_CONF */ 5661 + nla_total_size(IPSTATS_MIB_MAX * 8) /* IFLA_INET6_STATS */ 5662 + nla_total_size(ICMP6_MIB_MAX * 8) /* IFLA_INET6_ICMP6STATS */ 5663 + nla_total_size(sizeof(struct in6_addr)) /* IFLA_INET6_TOKEN */ 5664 + nla_total_size(1) /* IFLA_INET6_ADDR_GEN_MODE */ 5665 + nla_total_size(4) /* IFLA_INET6_RA_MTU */ 5666 + 0; 5667 } 5668 5669 static inline size_t inet6_if_nlmsg_size(void) 5670 { 5671 return NLMSG_ALIGN(sizeof(struct ifinfomsg)) 5672 + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ 5673 + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ 5674 + nla_total_size(4) /* IFLA_MTU */ 5675 + nla_total_size(4) /* IFLA_LINK */ 5676 + nla_total_size(1) /* IFLA_OPERSTATE */ 5677 + nla_total_size(inet6_ifla6_size()); /* IFLA_PROTINFO */ 5678 } 5679 5680 static inline void __snmp6_fill_statsdev(u64 *stats, atomic_long_t *mib, 5681 int bytes) 5682 { 5683 int i; 5684 int pad = bytes - sizeof(u64) * ICMP6_MIB_MAX; 5685 BUG_ON(pad < 0); 5686 5687 /* Use put_unaligned() because stats may not be aligned for u64. */ 5688 put_unaligned(ICMP6_MIB_MAX, &stats[0]); 5689 for (i = 1; i < ICMP6_MIB_MAX; i++) 5690 put_unaligned(atomic_long_read(&mib[i]), &stats[i]); 5691 5692 memset(&stats[ICMP6_MIB_MAX], 0, pad); 5693 } 5694 5695 static inline void __snmp6_fill_stats64(u64 *stats, void __percpu *mib, 5696 int bytes, size_t syncpoff) 5697 { 5698 int i, c; 5699 u64 buff[IPSTATS_MIB_MAX]; 5700 int pad = bytes - sizeof(u64) * IPSTATS_MIB_MAX; 5701 5702 BUG_ON(pad < 0); 5703 5704 memset(buff, 0, sizeof(buff)); 5705 buff[0] = IPSTATS_MIB_MAX; 5706 5707 for_each_possible_cpu(c) { 5708 for (i = 1; i < IPSTATS_MIB_MAX; i++) 5709 buff[i] += snmp_get_cpu_field64(mib, c, i, syncpoff); 5710 } 5711 5712 memcpy(stats, buff, IPSTATS_MIB_MAX * sizeof(u64)); 5713 memset(&stats[IPSTATS_MIB_MAX], 0, pad); 5714 } 5715 5716 static void snmp6_fill_stats(u64 *stats, struct inet6_dev *idev, int attrtype, 5717 int bytes) 5718 { 5719 switch (attrtype) { 5720 case IFLA_INET6_STATS: 5721 __snmp6_fill_stats64(stats, idev->stats.ipv6, bytes, 5722 offsetof(struct ipstats_mib, syncp)); 5723 break; 5724 case IFLA_INET6_ICMP6STATS: 5725 __snmp6_fill_statsdev(stats, idev->stats.icmpv6dev->mibs, bytes); 5726 break; 5727 } 5728 } 5729 5730 static int inet6_fill_ifla6_attrs(struct sk_buff *skb, struct inet6_dev *idev, 5731 u32 ext_filter_mask) 5732 { 5733 struct nlattr *nla; 5734 struct ifla_cacheinfo ci; 5735 5736 if (nla_put_u32(skb, IFLA_INET6_FLAGS, idev->if_flags)) 5737 goto nla_put_failure; 5738 ci.max_reasm_len = IPV6_MAXPLEN; 5739 ci.tstamp = cstamp_delta(idev->tstamp); 5740 ci.reachable_time = jiffies_to_msecs(idev->nd_parms->reachable_time); 5741 ci.retrans_time = jiffies_to_msecs(NEIGH_VAR(idev->nd_parms, RETRANS_TIME)); 5742 if (nla_put(skb, IFLA_INET6_CACHEINFO, sizeof(ci), &ci)) 5743 goto nla_put_failure; 5744 nla = nla_reserve(skb, IFLA_INET6_CONF, DEVCONF_MAX * sizeof(s32)); 5745 if (!nla) 5746 goto nla_put_failure; 5747 ipv6_store_devconf(&idev->cnf, nla_data(nla), nla_len(nla)); 5748 5749 /* XXX - MC not implemented */ 5750 5751 if (ext_filter_mask & RTEXT_FILTER_SKIP_STATS) 5752 return 0; 5753 5754 nla = nla_reserve(skb, IFLA_INET6_STATS, IPSTATS_MIB_MAX * sizeof(u64)); 5755 if (!nla) 5756 goto nla_put_failure; 5757 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_STATS, nla_len(nla)); 5758 5759 nla = nla_reserve(skb, IFLA_INET6_ICMP6STATS, ICMP6_MIB_MAX * sizeof(u64)); 5760 if (!nla) 5761 goto nla_put_failure; 5762 snmp6_fill_stats(nla_data(nla), idev, IFLA_INET6_ICMP6STATS, nla_len(nla)); 5763 5764 nla = nla_reserve(skb, IFLA_INET6_TOKEN, sizeof(struct in6_addr)); 5765 if (!nla) 5766 goto nla_put_failure; 5767 read_lock_bh(&idev->lock); 5768 memcpy(nla_data(nla), idev->token.s6_addr, nla_len(nla)); 5769 read_unlock_bh(&idev->lock); 5770 5771 if (nla_put_u8(skb, IFLA_INET6_ADDR_GEN_MODE, idev->cnf.addr_gen_mode)) 5772 goto nla_put_failure; 5773 5774 if (idev->ra_mtu && 5775 nla_put_u32(skb, IFLA_INET6_RA_MTU, idev->ra_mtu)) 5776 goto nla_put_failure; 5777 5778 return 0; 5779 5780 nla_put_failure: 5781 return -EMSGSIZE; 5782 } 5783 5784 static size_t inet6_get_link_af_size(const struct net_device *dev, 5785 u32 ext_filter_mask) 5786 { 5787 if (!__in6_dev_get(dev)) 5788 return 0; 5789 5790 return inet6_ifla6_size(); 5791 } 5792 5793 static int inet6_fill_link_af(struct sk_buff *skb, const struct net_device *dev, 5794 u32 ext_filter_mask) 5795 { 5796 struct inet6_dev *idev = __in6_dev_get(dev); 5797 5798 if (!idev) 5799 return -ENODATA; 5800 5801 if (inet6_fill_ifla6_attrs(skb, idev, ext_filter_mask) < 0) 5802 return -EMSGSIZE; 5803 5804 return 0; 5805 } 5806 5807 static int inet6_set_iftoken(struct inet6_dev *idev, struct in6_addr *token, 5808 struct netlink_ext_ack *extack) 5809 { 5810 struct inet6_ifaddr *ifp; 5811 struct net_device *dev = idev->dev; 5812 bool clear_token, update_rs = false; 5813 struct in6_addr ll_addr; 5814 5815 ASSERT_RTNL(); 5816 5817 if (!token) 5818 return -EINVAL; 5819 5820 if (dev->flags & IFF_LOOPBACK) { 5821 NL_SET_ERR_MSG_MOD(extack, "Device is loopback"); 5822 return -EINVAL; 5823 } 5824 5825 if (dev->flags & IFF_NOARP) { 5826 NL_SET_ERR_MSG_MOD(extack, 5827 "Device does not do neighbour discovery"); 5828 return -EINVAL; 5829 } 5830 5831 if (!ipv6_accept_ra(idev)) { 5832 NL_SET_ERR_MSG_MOD(extack, 5833 "Router advertisement is disabled on device"); 5834 return -EINVAL; 5835 } 5836 5837 if (idev->cnf.rtr_solicits == 0) { 5838 NL_SET_ERR_MSG(extack, 5839 "Router solicitation is disabled on device"); 5840 return -EINVAL; 5841 } 5842 5843 write_lock_bh(&idev->lock); 5844 5845 BUILD_BUG_ON(sizeof(token->s6_addr) != 16); 5846 memcpy(idev->token.s6_addr + 8, token->s6_addr + 8, 8); 5847 5848 write_unlock_bh(&idev->lock); 5849 5850 clear_token = ipv6_addr_any(token); 5851 if (clear_token) 5852 goto update_lft; 5853 5854 if (!idev->dead && (idev->if_flags & IF_READY) && 5855 !ipv6_get_lladdr(dev, &ll_addr, IFA_F_TENTATIVE | 5856 IFA_F_OPTIMISTIC)) { 5857 /* If we're not ready, then normal ifup will take care 5858 * of this. Otherwise, we need to request our rs here. 5859 */ 5860 ndisc_send_rs(dev, &ll_addr, &in6addr_linklocal_allrouters); 5861 update_rs = true; 5862 } 5863 5864 update_lft: 5865 write_lock_bh(&idev->lock); 5866 5867 if (update_rs) { 5868 idev->if_flags |= IF_RS_SENT; 5869 idev->rs_interval = rfc3315_s14_backoff_init( 5870 idev->cnf.rtr_solicit_interval); 5871 idev->rs_probes = 1; 5872 addrconf_mod_rs_timer(idev, idev->rs_interval); 5873 } 5874 5875 /* Well, that's kinda nasty ... */ 5876 list_for_each_entry(ifp, &idev->addr_list, if_list) { 5877 spin_lock(&ifp->lock); 5878 if (ifp->tokenized) { 5879 ifp->valid_lft = 0; 5880 ifp->prefered_lft = 0; 5881 } 5882 spin_unlock(&ifp->lock); 5883 } 5884 5885 write_unlock_bh(&idev->lock); 5886 inet6_ifinfo_notify(RTM_NEWLINK, idev); 5887 addrconf_verify_rtnl(dev_net(dev)); 5888 return 0; 5889 } 5890 5891 static const struct nla_policy inet6_af_policy[IFLA_INET6_MAX + 1] = { 5892 [IFLA_INET6_ADDR_GEN_MODE] = { .type = NLA_U8 }, 5893 [IFLA_INET6_TOKEN] = { .len = sizeof(struct in6_addr) }, 5894 [IFLA_INET6_RA_MTU] = { .type = NLA_REJECT, 5895 .reject_message = 5896 "IFLA_INET6_RA_MTU can not be set" }, 5897 }; 5898 5899 static int check_addr_gen_mode(int mode) 5900 { 5901 if (mode != IN6_ADDR_GEN_MODE_EUI64 && 5902 mode != IN6_ADDR_GEN_MODE_NONE && 5903 mode != IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 5904 mode != IN6_ADDR_GEN_MODE_RANDOM) 5905 return -EINVAL; 5906 return 1; 5907 } 5908 5909 static int check_stable_privacy(struct inet6_dev *idev, struct net *net, 5910 int mode) 5911 { 5912 if (mode == IN6_ADDR_GEN_MODE_STABLE_PRIVACY && 5913 !idev->cnf.stable_secret.initialized && 5914 !net->ipv6.devconf_dflt->stable_secret.initialized) 5915 return -EINVAL; 5916 return 1; 5917 } 5918 5919 static int inet6_validate_link_af(const struct net_device *dev, 5920 const struct nlattr *nla, 5921 struct netlink_ext_ack *extack) 5922 { 5923 struct nlattr *tb[IFLA_INET6_MAX + 1]; 5924 struct inet6_dev *idev = NULL; 5925 int err; 5926 5927 if (dev) { 5928 idev = __in6_dev_get(dev); 5929 if (!idev) 5930 return -EAFNOSUPPORT; 5931 } 5932 5933 err = nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, 5934 inet6_af_policy, extack); 5935 if (err) 5936 return err; 5937 5938 if (!tb[IFLA_INET6_TOKEN] && !tb[IFLA_INET6_ADDR_GEN_MODE]) 5939 return -EINVAL; 5940 5941 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 5942 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 5943 5944 if (check_addr_gen_mode(mode) < 0) 5945 return -EINVAL; 5946 if (dev && check_stable_privacy(idev, dev_net(dev), mode) < 0) 5947 return -EINVAL; 5948 } 5949 5950 return 0; 5951 } 5952 5953 static int inet6_set_link_af(struct net_device *dev, const struct nlattr *nla, 5954 struct netlink_ext_ack *extack) 5955 { 5956 struct inet6_dev *idev = __in6_dev_get(dev); 5957 struct nlattr *tb[IFLA_INET6_MAX + 1]; 5958 int err; 5959 5960 if (!idev) 5961 return -EAFNOSUPPORT; 5962 5963 if (nla_parse_nested_deprecated(tb, IFLA_INET6_MAX, nla, NULL, NULL) < 0) 5964 return -EINVAL; 5965 5966 if (tb[IFLA_INET6_TOKEN]) { 5967 err = inet6_set_iftoken(idev, nla_data(tb[IFLA_INET6_TOKEN]), 5968 extack); 5969 if (err) 5970 return err; 5971 } 5972 5973 if (tb[IFLA_INET6_ADDR_GEN_MODE]) { 5974 u8 mode = nla_get_u8(tb[IFLA_INET6_ADDR_GEN_MODE]); 5975 5976 idev->cnf.addr_gen_mode = mode; 5977 } 5978 5979 return 0; 5980 } 5981 5982 static int inet6_fill_ifinfo(struct sk_buff *skb, struct inet6_dev *idev, 5983 u32 portid, u32 seq, int event, unsigned int flags) 5984 { 5985 struct net_device *dev = idev->dev; 5986 struct ifinfomsg *hdr; 5987 struct nlmsghdr *nlh; 5988 void *protoinfo; 5989 5990 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*hdr), flags); 5991 if (!nlh) 5992 return -EMSGSIZE; 5993 5994 hdr = nlmsg_data(nlh); 5995 hdr->ifi_family = AF_INET6; 5996 hdr->__ifi_pad = 0; 5997 hdr->ifi_type = dev->type; 5998 hdr->ifi_index = dev->ifindex; 5999 hdr->ifi_flags = dev_get_flags(dev); 6000 hdr->ifi_change = 0; 6001 6002 if (nla_put_string(skb, IFLA_IFNAME, dev->name) || 6003 (dev->addr_len && 6004 nla_put(skb, IFLA_ADDRESS, dev->addr_len, dev->dev_addr)) || 6005 nla_put_u32(skb, IFLA_MTU, dev->mtu) || 6006 (dev->ifindex != dev_get_iflink(dev) && 6007 nla_put_u32(skb, IFLA_LINK, dev_get_iflink(dev))) || 6008 nla_put_u8(skb, IFLA_OPERSTATE, 6009 netif_running(dev) ? dev->operstate : IF_OPER_DOWN)) 6010 goto nla_put_failure; 6011 protoinfo = nla_nest_start_noflag(skb, IFLA_PROTINFO); 6012 if (!protoinfo) 6013 goto nla_put_failure; 6014 6015 if (inet6_fill_ifla6_attrs(skb, idev, 0) < 0) 6016 goto nla_put_failure; 6017 6018 nla_nest_end(skb, protoinfo); 6019 nlmsg_end(skb, nlh); 6020 return 0; 6021 6022 nla_put_failure: 6023 nlmsg_cancel(skb, nlh); 6024 return -EMSGSIZE; 6025 } 6026 6027 static int inet6_valid_dump_ifinfo(const struct nlmsghdr *nlh, 6028 struct netlink_ext_ack *extack) 6029 { 6030 struct ifinfomsg *ifm; 6031 6032 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ifm))) { 6033 NL_SET_ERR_MSG_MOD(extack, "Invalid header for link dump request"); 6034 return -EINVAL; 6035 } 6036 6037 if (nlmsg_attrlen(nlh, sizeof(*ifm))) { 6038 NL_SET_ERR_MSG_MOD(extack, "Invalid data after header"); 6039 return -EINVAL; 6040 } 6041 6042 ifm = nlmsg_data(nlh); 6043 if (ifm->__ifi_pad || ifm->ifi_type || ifm->ifi_flags || 6044 ifm->ifi_change || ifm->ifi_index) { 6045 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for dump request"); 6046 return -EINVAL; 6047 } 6048 6049 return 0; 6050 } 6051 6052 static int inet6_dump_ifinfo(struct sk_buff *skb, struct netlink_callback *cb) 6053 { 6054 struct net *net = sock_net(skb->sk); 6055 int h, s_h; 6056 int idx = 0, s_idx; 6057 struct net_device *dev; 6058 struct inet6_dev *idev; 6059 struct hlist_head *head; 6060 6061 /* only requests using strict checking can pass data to 6062 * influence the dump 6063 */ 6064 if (cb->strict_check) { 6065 int err = inet6_valid_dump_ifinfo(cb->nlh, cb->extack); 6066 6067 if (err < 0) 6068 return err; 6069 } 6070 6071 s_h = cb->args[0]; 6072 s_idx = cb->args[1]; 6073 6074 rcu_read_lock(); 6075 for (h = s_h; h < NETDEV_HASHENTRIES; h++, s_idx = 0) { 6076 idx = 0; 6077 head = &net->dev_index_head[h]; 6078 hlist_for_each_entry_rcu(dev, head, index_hlist) { 6079 if (idx < s_idx) 6080 goto cont; 6081 idev = __in6_dev_get(dev); 6082 if (!idev) 6083 goto cont; 6084 if (inet6_fill_ifinfo(skb, idev, 6085 NETLINK_CB(cb->skb).portid, 6086 cb->nlh->nlmsg_seq, 6087 RTM_NEWLINK, NLM_F_MULTI) < 0) 6088 goto out; 6089 cont: 6090 idx++; 6091 } 6092 } 6093 out: 6094 rcu_read_unlock(); 6095 cb->args[1] = idx; 6096 cb->args[0] = h; 6097 6098 return skb->len; 6099 } 6100 6101 void inet6_ifinfo_notify(int event, struct inet6_dev *idev) 6102 { 6103 struct sk_buff *skb; 6104 struct net *net = dev_net(idev->dev); 6105 int err = -ENOBUFS; 6106 6107 skb = nlmsg_new(inet6_if_nlmsg_size(), GFP_ATOMIC); 6108 if (!skb) 6109 goto errout; 6110 6111 err = inet6_fill_ifinfo(skb, idev, 0, 0, event, 0); 6112 if (err < 0) { 6113 /* -EMSGSIZE implies BUG in inet6_if_nlmsg_size() */ 6114 WARN_ON(err == -EMSGSIZE); 6115 kfree_skb(skb); 6116 goto errout; 6117 } 6118 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_IFINFO, NULL, GFP_ATOMIC); 6119 return; 6120 errout: 6121 if (err < 0) 6122 rtnl_set_sk_err(net, RTNLGRP_IPV6_IFINFO, err); 6123 } 6124 6125 static inline size_t inet6_prefix_nlmsg_size(void) 6126 { 6127 return NLMSG_ALIGN(sizeof(struct prefixmsg)) 6128 + nla_total_size(sizeof(struct in6_addr)) 6129 + nla_total_size(sizeof(struct prefix_cacheinfo)); 6130 } 6131 6132 static int inet6_fill_prefix(struct sk_buff *skb, struct inet6_dev *idev, 6133 struct prefix_info *pinfo, u32 portid, u32 seq, 6134 int event, unsigned int flags) 6135 { 6136 struct prefixmsg *pmsg; 6137 struct nlmsghdr *nlh; 6138 struct prefix_cacheinfo ci; 6139 6140 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*pmsg), flags); 6141 if (!nlh) 6142 return -EMSGSIZE; 6143 6144 pmsg = nlmsg_data(nlh); 6145 pmsg->prefix_family = AF_INET6; 6146 pmsg->prefix_pad1 = 0; 6147 pmsg->prefix_pad2 = 0; 6148 pmsg->prefix_ifindex = idev->dev->ifindex; 6149 pmsg->prefix_len = pinfo->prefix_len; 6150 pmsg->prefix_type = pinfo->type; 6151 pmsg->prefix_pad3 = 0; 6152 pmsg->prefix_flags = 0; 6153 if (pinfo->onlink) 6154 pmsg->prefix_flags |= IF_PREFIX_ONLINK; 6155 if (pinfo->autoconf) 6156 pmsg->prefix_flags |= IF_PREFIX_AUTOCONF; 6157 6158 if (nla_put(skb, PREFIX_ADDRESS, sizeof(pinfo->prefix), &pinfo->prefix)) 6159 goto nla_put_failure; 6160 ci.preferred_time = ntohl(pinfo->prefered); 6161 ci.valid_time = ntohl(pinfo->valid); 6162 if (nla_put(skb, PREFIX_CACHEINFO, sizeof(ci), &ci)) 6163 goto nla_put_failure; 6164 nlmsg_end(skb, nlh); 6165 return 0; 6166 6167 nla_put_failure: 6168 nlmsg_cancel(skb, nlh); 6169 return -EMSGSIZE; 6170 } 6171 6172 static void inet6_prefix_notify(int event, struct inet6_dev *idev, 6173 struct prefix_info *pinfo) 6174 { 6175 struct sk_buff *skb; 6176 struct net *net = dev_net(idev->dev); 6177 int err = -ENOBUFS; 6178 6179 skb = nlmsg_new(inet6_prefix_nlmsg_size(), GFP_ATOMIC); 6180 if (!skb) 6181 goto errout; 6182 6183 err = inet6_fill_prefix(skb, idev, pinfo, 0, 0, event, 0); 6184 if (err < 0) { 6185 /* -EMSGSIZE implies BUG in inet6_prefix_nlmsg_size() */ 6186 WARN_ON(err == -EMSGSIZE); 6187 kfree_skb(skb); 6188 goto errout; 6189 } 6190 rtnl_notify(skb, net, 0, RTNLGRP_IPV6_PREFIX, NULL, GFP_ATOMIC); 6191 return; 6192 errout: 6193 if (err < 0) 6194 rtnl_set_sk_err(net, RTNLGRP_IPV6_PREFIX, err); 6195 } 6196 6197 static void __ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6198 { 6199 struct net *net = dev_net(ifp->idev->dev); 6200 6201 if (event) 6202 ASSERT_RTNL(); 6203 6204 inet6_ifa_notify(event ? : RTM_NEWADDR, ifp); 6205 6206 switch (event) { 6207 case RTM_NEWADDR: 6208 /* 6209 * If the address was optimistic we inserted the route at the 6210 * start of our DAD process, so we don't need to do it again. 6211 * If the device was taken down in the middle of the DAD 6212 * cycle there is a race where we could get here without a 6213 * host route, so nothing to insert. That will be fixed when 6214 * the device is brought up. 6215 */ 6216 if (ifp->rt && !rcu_access_pointer(ifp->rt->fib6_node)) { 6217 ip6_ins_rt(net, ifp->rt); 6218 } else if (!ifp->rt && (ifp->idev->dev->flags & IFF_UP)) { 6219 pr_warn("BUG: Address %pI6c on device %s is missing its host route.\n", 6220 &ifp->addr, ifp->idev->dev->name); 6221 } 6222 6223 if (ifp->idev->cnf.forwarding) 6224 addrconf_join_anycast(ifp); 6225 if (!ipv6_addr_any(&ifp->peer_addr)) 6226 addrconf_prefix_route(&ifp->peer_addr, 128, 6227 ifp->rt_priority, ifp->idev->dev, 6228 0, 0, GFP_ATOMIC); 6229 break; 6230 case RTM_DELADDR: 6231 if (ifp->idev->cnf.forwarding) 6232 addrconf_leave_anycast(ifp); 6233 addrconf_leave_solict(ifp->idev, &ifp->addr); 6234 if (!ipv6_addr_any(&ifp->peer_addr)) { 6235 struct fib6_info *rt; 6236 6237 rt = addrconf_get_prefix_route(&ifp->peer_addr, 128, 6238 ifp->idev->dev, 0, 0, 6239 false); 6240 if (rt) 6241 ip6_del_rt(net, rt, false); 6242 } 6243 if (ifp->rt) { 6244 ip6_del_rt(net, ifp->rt, false); 6245 ifp->rt = NULL; 6246 } 6247 rt_genid_bump_ipv6(net); 6248 break; 6249 } 6250 atomic_inc(&net->ipv6.dev_addr_genid); 6251 } 6252 6253 static void ipv6_ifa_notify(int event, struct inet6_ifaddr *ifp) 6254 { 6255 if (likely(ifp->idev->dead == 0)) 6256 __ipv6_ifa_notify(event, ifp); 6257 } 6258 6259 #ifdef CONFIG_SYSCTL 6260 6261 static int addrconf_sysctl_forward(struct ctl_table *ctl, int write, 6262 void *buffer, size_t *lenp, loff_t *ppos) 6263 { 6264 int *valp = ctl->data; 6265 int val = *valp; 6266 loff_t pos = *ppos; 6267 struct ctl_table lctl; 6268 int ret; 6269 6270 /* 6271 * ctl->data points to idev->cnf.forwarding, we should 6272 * not modify it until we get the rtnl lock. 6273 */ 6274 lctl = *ctl; 6275 lctl.data = &val; 6276 6277 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6278 6279 if (write) 6280 ret = addrconf_fixup_forwarding(ctl, valp, val); 6281 if (ret) 6282 *ppos = pos; 6283 return ret; 6284 } 6285 6286 static int addrconf_sysctl_mtu(struct ctl_table *ctl, int write, 6287 void *buffer, size_t *lenp, loff_t *ppos) 6288 { 6289 struct inet6_dev *idev = ctl->extra1; 6290 int min_mtu = IPV6_MIN_MTU; 6291 struct ctl_table lctl; 6292 6293 lctl = *ctl; 6294 lctl.extra1 = &min_mtu; 6295 lctl.extra2 = idev ? &idev->dev->mtu : NULL; 6296 6297 return proc_dointvec_minmax(&lctl, write, buffer, lenp, ppos); 6298 } 6299 6300 static void dev_disable_change(struct inet6_dev *idev) 6301 { 6302 struct netdev_notifier_info info; 6303 6304 if (!idev || !idev->dev) 6305 return; 6306 6307 netdev_notifier_info_init(&info, idev->dev); 6308 if (idev->cnf.disable_ipv6) 6309 addrconf_notify(NULL, NETDEV_DOWN, &info); 6310 else 6311 addrconf_notify(NULL, NETDEV_UP, &info); 6312 } 6313 6314 static void addrconf_disable_change(struct net *net, __s32 newf) 6315 { 6316 struct net_device *dev; 6317 struct inet6_dev *idev; 6318 6319 for_each_netdev(net, dev) { 6320 idev = __in6_dev_get(dev); 6321 if (idev) { 6322 int changed = (!idev->cnf.disable_ipv6) ^ (!newf); 6323 idev->cnf.disable_ipv6 = newf; 6324 if (changed) 6325 dev_disable_change(idev); 6326 } 6327 } 6328 } 6329 6330 static int addrconf_disable_ipv6(struct ctl_table *table, int *p, int newf) 6331 { 6332 struct net *net; 6333 int old; 6334 6335 if (!rtnl_trylock()) 6336 return restart_syscall(); 6337 6338 net = (struct net *)table->extra2; 6339 old = *p; 6340 *p = newf; 6341 6342 if (p == &net->ipv6.devconf_dflt->disable_ipv6) { 6343 rtnl_unlock(); 6344 return 0; 6345 } 6346 6347 if (p == &net->ipv6.devconf_all->disable_ipv6) { 6348 net->ipv6.devconf_dflt->disable_ipv6 = newf; 6349 addrconf_disable_change(net, newf); 6350 } else if ((!newf) ^ (!old)) 6351 dev_disable_change((struct inet6_dev *)table->extra1); 6352 6353 rtnl_unlock(); 6354 return 0; 6355 } 6356 6357 static int addrconf_sysctl_disable(struct ctl_table *ctl, int write, 6358 void *buffer, size_t *lenp, loff_t *ppos) 6359 { 6360 int *valp = ctl->data; 6361 int val = *valp; 6362 loff_t pos = *ppos; 6363 struct ctl_table lctl; 6364 int ret; 6365 6366 /* 6367 * ctl->data points to idev->cnf.disable_ipv6, we should 6368 * not modify it until we get the rtnl lock. 6369 */ 6370 lctl = *ctl; 6371 lctl.data = &val; 6372 6373 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6374 6375 if (write) 6376 ret = addrconf_disable_ipv6(ctl, valp, val); 6377 if (ret) 6378 *ppos = pos; 6379 return ret; 6380 } 6381 6382 static int addrconf_sysctl_proxy_ndp(struct ctl_table *ctl, int write, 6383 void *buffer, size_t *lenp, loff_t *ppos) 6384 { 6385 int *valp = ctl->data; 6386 int ret; 6387 int old, new; 6388 6389 old = *valp; 6390 ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 6391 new = *valp; 6392 6393 if (write && old != new) { 6394 struct net *net = ctl->extra2; 6395 6396 if (!rtnl_trylock()) 6397 return restart_syscall(); 6398 6399 if (valp == &net->ipv6.devconf_dflt->proxy_ndp) 6400 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6401 NETCONFA_PROXY_NEIGH, 6402 NETCONFA_IFINDEX_DEFAULT, 6403 net->ipv6.devconf_dflt); 6404 else if (valp == &net->ipv6.devconf_all->proxy_ndp) 6405 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6406 NETCONFA_PROXY_NEIGH, 6407 NETCONFA_IFINDEX_ALL, 6408 net->ipv6.devconf_all); 6409 else { 6410 struct inet6_dev *idev = ctl->extra1; 6411 6412 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, 6413 NETCONFA_PROXY_NEIGH, 6414 idev->dev->ifindex, 6415 &idev->cnf); 6416 } 6417 rtnl_unlock(); 6418 } 6419 6420 return ret; 6421 } 6422 6423 static int addrconf_sysctl_addr_gen_mode(struct ctl_table *ctl, int write, 6424 void *buffer, size_t *lenp, 6425 loff_t *ppos) 6426 { 6427 int ret = 0; 6428 u32 new_val; 6429 struct inet6_dev *idev = (struct inet6_dev *)ctl->extra1; 6430 struct net *net = (struct net *)ctl->extra2; 6431 struct ctl_table tmp = { 6432 .data = &new_val, 6433 .maxlen = sizeof(new_val), 6434 .mode = ctl->mode, 6435 }; 6436 6437 if (!rtnl_trylock()) 6438 return restart_syscall(); 6439 6440 new_val = *((u32 *)ctl->data); 6441 6442 ret = proc_douintvec(&tmp, write, buffer, lenp, ppos); 6443 if (ret != 0) 6444 goto out; 6445 6446 if (write) { 6447 if (check_addr_gen_mode(new_val) < 0) { 6448 ret = -EINVAL; 6449 goto out; 6450 } 6451 6452 if (idev) { 6453 if (check_stable_privacy(idev, net, new_val) < 0) { 6454 ret = -EINVAL; 6455 goto out; 6456 } 6457 6458 if (idev->cnf.addr_gen_mode != new_val) { 6459 idev->cnf.addr_gen_mode = new_val; 6460 addrconf_init_auto_addrs(idev->dev); 6461 } 6462 } else if (&net->ipv6.devconf_all->addr_gen_mode == ctl->data) { 6463 struct net_device *dev; 6464 6465 net->ipv6.devconf_dflt->addr_gen_mode = new_val; 6466 for_each_netdev(net, dev) { 6467 idev = __in6_dev_get(dev); 6468 if (idev && 6469 idev->cnf.addr_gen_mode != new_val) { 6470 idev->cnf.addr_gen_mode = new_val; 6471 addrconf_init_auto_addrs(idev->dev); 6472 } 6473 } 6474 } 6475 6476 *((u32 *)ctl->data) = new_val; 6477 } 6478 6479 out: 6480 rtnl_unlock(); 6481 6482 return ret; 6483 } 6484 6485 static int addrconf_sysctl_stable_secret(struct ctl_table *ctl, int write, 6486 void *buffer, size_t *lenp, 6487 loff_t *ppos) 6488 { 6489 int err; 6490 struct in6_addr addr; 6491 char str[IPV6_MAX_STRLEN]; 6492 struct ctl_table lctl = *ctl; 6493 struct net *net = ctl->extra2; 6494 struct ipv6_stable_secret *secret = ctl->data; 6495 6496 if (&net->ipv6.devconf_all->stable_secret == ctl->data) 6497 return -EIO; 6498 6499 lctl.maxlen = IPV6_MAX_STRLEN; 6500 lctl.data = str; 6501 6502 if (!rtnl_trylock()) 6503 return restart_syscall(); 6504 6505 if (!write && !secret->initialized) { 6506 err = -EIO; 6507 goto out; 6508 } 6509 6510 err = snprintf(str, sizeof(str), "%pI6", &secret->secret); 6511 if (err >= sizeof(str)) { 6512 err = -EIO; 6513 goto out; 6514 } 6515 6516 err = proc_dostring(&lctl, write, buffer, lenp, ppos); 6517 if (err || !write) 6518 goto out; 6519 6520 if (in6_pton(str, -1, addr.in6_u.u6_addr8, -1, NULL) != 1) { 6521 err = -EIO; 6522 goto out; 6523 } 6524 6525 secret->initialized = true; 6526 secret->secret = addr; 6527 6528 if (&net->ipv6.devconf_dflt->stable_secret == ctl->data) { 6529 struct net_device *dev; 6530 6531 for_each_netdev(net, dev) { 6532 struct inet6_dev *idev = __in6_dev_get(dev); 6533 6534 if (idev) { 6535 idev->cnf.addr_gen_mode = 6536 IN6_ADDR_GEN_MODE_STABLE_PRIVACY; 6537 } 6538 } 6539 } else { 6540 struct inet6_dev *idev = ctl->extra1; 6541 6542 idev->cnf.addr_gen_mode = IN6_ADDR_GEN_MODE_STABLE_PRIVACY; 6543 } 6544 6545 out: 6546 rtnl_unlock(); 6547 6548 return err; 6549 } 6550 6551 static 6552 int addrconf_sysctl_ignore_routes_with_linkdown(struct ctl_table *ctl, 6553 int write, void *buffer, 6554 size_t *lenp, 6555 loff_t *ppos) 6556 { 6557 int *valp = ctl->data; 6558 int val = *valp; 6559 loff_t pos = *ppos; 6560 struct ctl_table lctl; 6561 int ret; 6562 6563 /* ctl->data points to idev->cnf.ignore_routes_when_linkdown 6564 * we should not modify it until we get the rtnl lock. 6565 */ 6566 lctl = *ctl; 6567 lctl.data = &val; 6568 6569 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6570 6571 if (write) 6572 ret = addrconf_fixup_linkdown(ctl, valp, val); 6573 if (ret) 6574 *ppos = pos; 6575 return ret; 6576 } 6577 6578 static 6579 void addrconf_set_nopolicy(struct rt6_info *rt, int action) 6580 { 6581 if (rt) { 6582 if (action) 6583 rt->dst.flags |= DST_NOPOLICY; 6584 else 6585 rt->dst.flags &= ~DST_NOPOLICY; 6586 } 6587 } 6588 6589 static 6590 void addrconf_disable_policy_idev(struct inet6_dev *idev, int val) 6591 { 6592 struct inet6_ifaddr *ifa; 6593 6594 read_lock_bh(&idev->lock); 6595 list_for_each_entry(ifa, &idev->addr_list, if_list) { 6596 spin_lock(&ifa->lock); 6597 if (ifa->rt) { 6598 /* host routes only use builtin fib6_nh */ 6599 struct fib6_nh *nh = ifa->rt->fib6_nh; 6600 int cpu; 6601 6602 rcu_read_lock(); 6603 ifa->rt->dst_nopolicy = val ? true : false; 6604 if (nh->rt6i_pcpu) { 6605 for_each_possible_cpu(cpu) { 6606 struct rt6_info **rtp; 6607 6608 rtp = per_cpu_ptr(nh->rt6i_pcpu, cpu); 6609 addrconf_set_nopolicy(*rtp, val); 6610 } 6611 } 6612 rcu_read_unlock(); 6613 } 6614 spin_unlock(&ifa->lock); 6615 } 6616 read_unlock_bh(&idev->lock); 6617 } 6618 6619 static 6620 int addrconf_disable_policy(struct ctl_table *ctl, int *valp, int val) 6621 { 6622 struct inet6_dev *idev; 6623 struct net *net; 6624 6625 if (!rtnl_trylock()) 6626 return restart_syscall(); 6627 6628 *valp = val; 6629 6630 net = (struct net *)ctl->extra2; 6631 if (valp == &net->ipv6.devconf_dflt->disable_policy) { 6632 rtnl_unlock(); 6633 return 0; 6634 } 6635 6636 if (valp == &net->ipv6.devconf_all->disable_policy) { 6637 struct net_device *dev; 6638 6639 for_each_netdev(net, dev) { 6640 idev = __in6_dev_get(dev); 6641 if (idev) 6642 addrconf_disable_policy_idev(idev, val); 6643 } 6644 } else { 6645 idev = (struct inet6_dev *)ctl->extra1; 6646 addrconf_disable_policy_idev(idev, val); 6647 } 6648 6649 rtnl_unlock(); 6650 return 0; 6651 } 6652 6653 static int addrconf_sysctl_disable_policy(struct ctl_table *ctl, int write, 6654 void *buffer, size_t *lenp, loff_t *ppos) 6655 { 6656 int *valp = ctl->data; 6657 int val = *valp; 6658 loff_t pos = *ppos; 6659 struct ctl_table lctl; 6660 int ret; 6661 6662 lctl = *ctl; 6663 lctl.data = &val; 6664 ret = proc_dointvec(&lctl, write, buffer, lenp, ppos); 6665 6666 if (write && (*valp != val)) 6667 ret = addrconf_disable_policy(ctl, valp, val); 6668 6669 if (ret) 6670 *ppos = pos; 6671 6672 return ret; 6673 } 6674 6675 static int minus_one = -1; 6676 static const int two_five_five = 255; 6677 static u32 ioam6_if_id_max = U16_MAX; 6678 6679 static const struct ctl_table addrconf_sysctl[] = { 6680 { 6681 .procname = "forwarding", 6682 .data = &ipv6_devconf.forwarding, 6683 .maxlen = sizeof(int), 6684 .mode = 0644, 6685 .proc_handler = addrconf_sysctl_forward, 6686 }, 6687 { 6688 .procname = "hop_limit", 6689 .data = &ipv6_devconf.hop_limit, 6690 .maxlen = sizeof(int), 6691 .mode = 0644, 6692 .proc_handler = proc_dointvec_minmax, 6693 .extra1 = (void *)SYSCTL_ONE, 6694 .extra2 = (void *)&two_five_five, 6695 }, 6696 { 6697 .procname = "mtu", 6698 .data = &ipv6_devconf.mtu6, 6699 .maxlen = sizeof(int), 6700 .mode = 0644, 6701 .proc_handler = addrconf_sysctl_mtu, 6702 }, 6703 { 6704 .procname = "accept_ra", 6705 .data = &ipv6_devconf.accept_ra, 6706 .maxlen = sizeof(int), 6707 .mode = 0644, 6708 .proc_handler = proc_dointvec, 6709 }, 6710 { 6711 .procname = "accept_redirects", 6712 .data = &ipv6_devconf.accept_redirects, 6713 .maxlen = sizeof(int), 6714 .mode = 0644, 6715 .proc_handler = proc_dointvec, 6716 }, 6717 { 6718 .procname = "autoconf", 6719 .data = &ipv6_devconf.autoconf, 6720 .maxlen = sizeof(int), 6721 .mode = 0644, 6722 .proc_handler = proc_dointvec, 6723 }, 6724 { 6725 .procname = "dad_transmits", 6726 .data = &ipv6_devconf.dad_transmits, 6727 .maxlen = sizeof(int), 6728 .mode = 0644, 6729 .proc_handler = proc_dointvec, 6730 }, 6731 { 6732 .procname = "router_solicitations", 6733 .data = &ipv6_devconf.rtr_solicits, 6734 .maxlen = sizeof(int), 6735 .mode = 0644, 6736 .proc_handler = proc_dointvec_minmax, 6737 .extra1 = &minus_one, 6738 }, 6739 { 6740 .procname = "router_solicitation_interval", 6741 .data = &ipv6_devconf.rtr_solicit_interval, 6742 .maxlen = sizeof(int), 6743 .mode = 0644, 6744 .proc_handler = proc_dointvec_jiffies, 6745 }, 6746 { 6747 .procname = "router_solicitation_max_interval", 6748 .data = &ipv6_devconf.rtr_solicit_max_interval, 6749 .maxlen = sizeof(int), 6750 .mode = 0644, 6751 .proc_handler = proc_dointvec_jiffies, 6752 }, 6753 { 6754 .procname = "router_solicitation_delay", 6755 .data = &ipv6_devconf.rtr_solicit_delay, 6756 .maxlen = sizeof(int), 6757 .mode = 0644, 6758 .proc_handler = proc_dointvec_jiffies, 6759 }, 6760 { 6761 .procname = "force_mld_version", 6762 .data = &ipv6_devconf.force_mld_version, 6763 .maxlen = sizeof(int), 6764 .mode = 0644, 6765 .proc_handler = proc_dointvec, 6766 }, 6767 { 6768 .procname = "mldv1_unsolicited_report_interval", 6769 .data = 6770 &ipv6_devconf.mldv1_unsolicited_report_interval, 6771 .maxlen = sizeof(int), 6772 .mode = 0644, 6773 .proc_handler = proc_dointvec_ms_jiffies, 6774 }, 6775 { 6776 .procname = "mldv2_unsolicited_report_interval", 6777 .data = 6778 &ipv6_devconf.mldv2_unsolicited_report_interval, 6779 .maxlen = sizeof(int), 6780 .mode = 0644, 6781 .proc_handler = proc_dointvec_ms_jiffies, 6782 }, 6783 { 6784 .procname = "use_tempaddr", 6785 .data = &ipv6_devconf.use_tempaddr, 6786 .maxlen = sizeof(int), 6787 .mode = 0644, 6788 .proc_handler = proc_dointvec, 6789 }, 6790 { 6791 .procname = "temp_valid_lft", 6792 .data = &ipv6_devconf.temp_valid_lft, 6793 .maxlen = sizeof(int), 6794 .mode = 0644, 6795 .proc_handler = proc_dointvec, 6796 }, 6797 { 6798 .procname = "temp_prefered_lft", 6799 .data = &ipv6_devconf.temp_prefered_lft, 6800 .maxlen = sizeof(int), 6801 .mode = 0644, 6802 .proc_handler = proc_dointvec, 6803 }, 6804 { 6805 .procname = "regen_max_retry", 6806 .data = &ipv6_devconf.regen_max_retry, 6807 .maxlen = sizeof(int), 6808 .mode = 0644, 6809 .proc_handler = proc_dointvec, 6810 }, 6811 { 6812 .procname = "max_desync_factor", 6813 .data = &ipv6_devconf.max_desync_factor, 6814 .maxlen = sizeof(int), 6815 .mode = 0644, 6816 .proc_handler = proc_dointvec, 6817 }, 6818 { 6819 .procname = "max_addresses", 6820 .data = &ipv6_devconf.max_addresses, 6821 .maxlen = sizeof(int), 6822 .mode = 0644, 6823 .proc_handler = proc_dointvec, 6824 }, 6825 { 6826 .procname = "accept_ra_defrtr", 6827 .data = &ipv6_devconf.accept_ra_defrtr, 6828 .maxlen = sizeof(int), 6829 .mode = 0644, 6830 .proc_handler = proc_dointvec, 6831 }, 6832 { 6833 .procname = "ra_defrtr_metric", 6834 .data = &ipv6_devconf.ra_defrtr_metric, 6835 .maxlen = sizeof(u32), 6836 .mode = 0644, 6837 .proc_handler = proc_douintvec_minmax, 6838 .extra1 = (void *)SYSCTL_ONE, 6839 }, 6840 { 6841 .procname = "accept_ra_min_hop_limit", 6842 .data = &ipv6_devconf.accept_ra_min_hop_limit, 6843 .maxlen = sizeof(int), 6844 .mode = 0644, 6845 .proc_handler = proc_dointvec, 6846 }, 6847 { 6848 .procname = "accept_ra_min_lft", 6849 .data = &ipv6_devconf.accept_ra_min_lft, 6850 .maxlen = sizeof(int), 6851 .mode = 0644, 6852 .proc_handler = proc_dointvec, 6853 }, 6854 { 6855 .procname = "accept_ra_pinfo", 6856 .data = &ipv6_devconf.accept_ra_pinfo, 6857 .maxlen = sizeof(int), 6858 .mode = 0644, 6859 .proc_handler = proc_dointvec, 6860 }, 6861 { 6862 .procname = "ra_honor_pio_life", 6863 .data = &ipv6_devconf.ra_honor_pio_life, 6864 .maxlen = sizeof(u8), 6865 .mode = 0644, 6866 .proc_handler = proc_dou8vec_minmax, 6867 .extra1 = SYSCTL_ZERO, 6868 .extra2 = SYSCTL_ONE, 6869 }, 6870 #ifdef CONFIG_IPV6_ROUTER_PREF 6871 { 6872 .procname = "accept_ra_rtr_pref", 6873 .data = &ipv6_devconf.accept_ra_rtr_pref, 6874 .maxlen = sizeof(int), 6875 .mode = 0644, 6876 .proc_handler = proc_dointvec, 6877 }, 6878 { 6879 .procname = "router_probe_interval", 6880 .data = &ipv6_devconf.rtr_probe_interval, 6881 .maxlen = sizeof(int), 6882 .mode = 0644, 6883 .proc_handler = proc_dointvec_jiffies, 6884 }, 6885 #ifdef CONFIG_IPV6_ROUTE_INFO 6886 { 6887 .procname = "accept_ra_rt_info_min_plen", 6888 .data = &ipv6_devconf.accept_ra_rt_info_min_plen, 6889 .maxlen = sizeof(int), 6890 .mode = 0644, 6891 .proc_handler = proc_dointvec, 6892 }, 6893 { 6894 .procname = "accept_ra_rt_info_max_plen", 6895 .data = &ipv6_devconf.accept_ra_rt_info_max_plen, 6896 .maxlen = sizeof(int), 6897 .mode = 0644, 6898 .proc_handler = proc_dointvec, 6899 }, 6900 #endif 6901 #endif 6902 { 6903 .procname = "proxy_ndp", 6904 .data = &ipv6_devconf.proxy_ndp, 6905 .maxlen = sizeof(int), 6906 .mode = 0644, 6907 .proc_handler = addrconf_sysctl_proxy_ndp, 6908 }, 6909 { 6910 .procname = "accept_source_route", 6911 .data = &ipv6_devconf.accept_source_route, 6912 .maxlen = sizeof(int), 6913 .mode = 0644, 6914 .proc_handler = proc_dointvec, 6915 }, 6916 #ifdef CONFIG_IPV6_OPTIMISTIC_DAD 6917 { 6918 .procname = "optimistic_dad", 6919 .data = &ipv6_devconf.optimistic_dad, 6920 .maxlen = sizeof(int), 6921 .mode = 0644, 6922 .proc_handler = proc_dointvec, 6923 }, 6924 { 6925 .procname = "use_optimistic", 6926 .data = &ipv6_devconf.use_optimistic, 6927 .maxlen = sizeof(int), 6928 .mode = 0644, 6929 .proc_handler = proc_dointvec, 6930 }, 6931 #endif 6932 #ifdef CONFIG_IPV6_MROUTE 6933 { 6934 .procname = "mc_forwarding", 6935 .data = &ipv6_devconf.mc_forwarding, 6936 .maxlen = sizeof(int), 6937 .mode = 0444, 6938 .proc_handler = proc_dointvec, 6939 }, 6940 #endif 6941 { 6942 .procname = "disable_ipv6", 6943 .data = &ipv6_devconf.disable_ipv6, 6944 .maxlen = sizeof(int), 6945 .mode = 0644, 6946 .proc_handler = addrconf_sysctl_disable, 6947 }, 6948 { 6949 .procname = "accept_dad", 6950 .data = &ipv6_devconf.accept_dad, 6951 .maxlen = sizeof(int), 6952 .mode = 0644, 6953 .proc_handler = proc_dointvec, 6954 }, 6955 { 6956 .procname = "force_tllao", 6957 .data = &ipv6_devconf.force_tllao, 6958 .maxlen = sizeof(int), 6959 .mode = 0644, 6960 .proc_handler = proc_dointvec 6961 }, 6962 { 6963 .procname = "ndisc_notify", 6964 .data = &ipv6_devconf.ndisc_notify, 6965 .maxlen = sizeof(int), 6966 .mode = 0644, 6967 .proc_handler = proc_dointvec 6968 }, 6969 { 6970 .procname = "suppress_frag_ndisc", 6971 .data = &ipv6_devconf.suppress_frag_ndisc, 6972 .maxlen = sizeof(int), 6973 .mode = 0644, 6974 .proc_handler = proc_dointvec 6975 }, 6976 { 6977 .procname = "accept_ra_from_local", 6978 .data = &ipv6_devconf.accept_ra_from_local, 6979 .maxlen = sizeof(int), 6980 .mode = 0644, 6981 .proc_handler = proc_dointvec, 6982 }, 6983 { 6984 .procname = "accept_ra_mtu", 6985 .data = &ipv6_devconf.accept_ra_mtu, 6986 .maxlen = sizeof(int), 6987 .mode = 0644, 6988 .proc_handler = proc_dointvec, 6989 }, 6990 { 6991 .procname = "stable_secret", 6992 .data = &ipv6_devconf.stable_secret, 6993 .maxlen = IPV6_MAX_STRLEN, 6994 .mode = 0600, 6995 .proc_handler = addrconf_sysctl_stable_secret, 6996 }, 6997 { 6998 .procname = "use_oif_addrs_only", 6999 .data = &ipv6_devconf.use_oif_addrs_only, 7000 .maxlen = sizeof(int), 7001 .mode = 0644, 7002 .proc_handler = proc_dointvec, 7003 }, 7004 { 7005 .procname = "ignore_routes_with_linkdown", 7006 .data = &ipv6_devconf.ignore_routes_with_linkdown, 7007 .maxlen = sizeof(int), 7008 .mode = 0644, 7009 .proc_handler = addrconf_sysctl_ignore_routes_with_linkdown, 7010 }, 7011 { 7012 .procname = "drop_unicast_in_l2_multicast", 7013 .data = &ipv6_devconf.drop_unicast_in_l2_multicast, 7014 .maxlen = sizeof(int), 7015 .mode = 0644, 7016 .proc_handler = proc_dointvec, 7017 }, 7018 { 7019 .procname = "drop_unsolicited_na", 7020 .data = &ipv6_devconf.drop_unsolicited_na, 7021 .maxlen = sizeof(int), 7022 .mode = 0644, 7023 .proc_handler = proc_dointvec, 7024 }, 7025 { 7026 .procname = "keep_addr_on_down", 7027 .data = &ipv6_devconf.keep_addr_on_down, 7028 .maxlen = sizeof(int), 7029 .mode = 0644, 7030 .proc_handler = proc_dointvec, 7031 7032 }, 7033 { 7034 .procname = "seg6_enabled", 7035 .data = &ipv6_devconf.seg6_enabled, 7036 .maxlen = sizeof(int), 7037 .mode = 0644, 7038 .proc_handler = proc_dointvec, 7039 }, 7040 #ifdef CONFIG_IPV6_SEG6_HMAC 7041 { 7042 .procname = "seg6_require_hmac", 7043 .data = &ipv6_devconf.seg6_require_hmac, 7044 .maxlen = sizeof(int), 7045 .mode = 0644, 7046 .proc_handler = proc_dointvec, 7047 }, 7048 #endif 7049 { 7050 .procname = "enhanced_dad", 7051 .data = &ipv6_devconf.enhanced_dad, 7052 .maxlen = sizeof(int), 7053 .mode = 0644, 7054 .proc_handler = proc_dointvec, 7055 }, 7056 { 7057 .procname = "addr_gen_mode", 7058 .data = &ipv6_devconf.addr_gen_mode, 7059 .maxlen = sizeof(int), 7060 .mode = 0644, 7061 .proc_handler = addrconf_sysctl_addr_gen_mode, 7062 }, 7063 { 7064 .procname = "disable_policy", 7065 .data = &ipv6_devconf.disable_policy, 7066 .maxlen = sizeof(int), 7067 .mode = 0644, 7068 .proc_handler = addrconf_sysctl_disable_policy, 7069 }, 7070 { 7071 .procname = "ndisc_tclass", 7072 .data = &ipv6_devconf.ndisc_tclass, 7073 .maxlen = sizeof(int), 7074 .mode = 0644, 7075 .proc_handler = proc_dointvec_minmax, 7076 .extra1 = (void *)SYSCTL_ZERO, 7077 .extra2 = (void *)&two_five_five, 7078 }, 7079 { 7080 .procname = "rpl_seg_enabled", 7081 .data = &ipv6_devconf.rpl_seg_enabled, 7082 .maxlen = sizeof(int), 7083 .mode = 0644, 7084 .proc_handler = proc_dointvec, 7085 }, 7086 { 7087 .procname = "ioam6_enabled", 7088 .data = &ipv6_devconf.ioam6_enabled, 7089 .maxlen = sizeof(u8), 7090 .mode = 0644, 7091 .proc_handler = proc_dou8vec_minmax, 7092 .extra1 = (void *)SYSCTL_ZERO, 7093 .extra2 = (void *)SYSCTL_ONE, 7094 }, 7095 { 7096 .procname = "ioam6_id", 7097 .data = &ipv6_devconf.ioam6_id, 7098 .maxlen = sizeof(u32), 7099 .mode = 0644, 7100 .proc_handler = proc_douintvec_minmax, 7101 .extra1 = (void *)SYSCTL_ZERO, 7102 .extra2 = (void *)&ioam6_if_id_max, 7103 }, 7104 { 7105 .procname = "ioam6_id_wide", 7106 .data = &ipv6_devconf.ioam6_id_wide, 7107 .maxlen = sizeof(u32), 7108 .mode = 0644, 7109 .proc_handler = proc_douintvec, 7110 }, 7111 { 7112 .procname = "ndisc_evict_nocarrier", 7113 .data = &ipv6_devconf.ndisc_evict_nocarrier, 7114 .maxlen = sizeof(u8), 7115 .mode = 0644, 7116 .proc_handler = proc_dou8vec_minmax, 7117 .extra1 = (void *)SYSCTL_ZERO, 7118 .extra2 = (void *)SYSCTL_ONE, 7119 }, 7120 { 7121 .procname = "accept_untracked_na", 7122 .data = &ipv6_devconf.accept_untracked_na, 7123 .maxlen = sizeof(int), 7124 .mode = 0644, 7125 .proc_handler = proc_dointvec_minmax, 7126 .extra1 = SYSCTL_ZERO, 7127 .extra2 = SYSCTL_TWO, 7128 }, 7129 { 7130 /* sentinel */ 7131 } 7132 }; 7133 7134 static int __addrconf_sysctl_register(struct net *net, char *dev_name, 7135 struct inet6_dev *idev, struct ipv6_devconf *p) 7136 { 7137 int i, ifindex; 7138 struct ctl_table *table; 7139 char path[sizeof("net/ipv6/conf/") + IFNAMSIZ]; 7140 7141 table = kmemdup(addrconf_sysctl, sizeof(addrconf_sysctl), GFP_KERNEL_ACCOUNT); 7142 if (!table) 7143 goto out; 7144 7145 for (i = 0; table[i].data; i++) { 7146 table[i].data += (char *)p - (char *)&ipv6_devconf; 7147 /* If one of these is already set, then it is not safe to 7148 * overwrite either of them: this makes proc_dointvec_minmax 7149 * usable. 7150 */ 7151 if (!table[i].extra1 && !table[i].extra2) { 7152 table[i].extra1 = idev; /* embedded; no ref */ 7153 table[i].extra2 = net; 7154 } 7155 } 7156 7157 snprintf(path, sizeof(path), "net/ipv6/conf/%s", dev_name); 7158 7159 p->sysctl_header = register_net_sysctl_sz(net, path, table, 7160 ARRAY_SIZE(addrconf_sysctl)); 7161 if (!p->sysctl_header) 7162 goto free; 7163 7164 if (!strcmp(dev_name, "all")) 7165 ifindex = NETCONFA_IFINDEX_ALL; 7166 else if (!strcmp(dev_name, "default")) 7167 ifindex = NETCONFA_IFINDEX_DEFAULT; 7168 else 7169 ifindex = idev->dev->ifindex; 7170 inet6_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, 7171 ifindex, p); 7172 return 0; 7173 7174 free: 7175 kfree(table); 7176 out: 7177 return -ENOBUFS; 7178 } 7179 7180 static void __addrconf_sysctl_unregister(struct net *net, 7181 struct ipv6_devconf *p, int ifindex) 7182 { 7183 struct ctl_table *table; 7184 7185 if (!p->sysctl_header) 7186 return; 7187 7188 table = p->sysctl_header->ctl_table_arg; 7189 unregister_net_sysctl_table(p->sysctl_header); 7190 p->sysctl_header = NULL; 7191 kfree(table); 7192 7193 inet6_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL); 7194 } 7195 7196 static int addrconf_sysctl_register(struct inet6_dev *idev) 7197 { 7198 int err; 7199 7200 if (!sysctl_dev_name_is_allowed(idev->dev->name)) 7201 return -EINVAL; 7202 7203 err = neigh_sysctl_register(idev->dev, idev->nd_parms, 7204 &ndisc_ifinfo_sysctl_change); 7205 if (err) 7206 return err; 7207 err = __addrconf_sysctl_register(dev_net(idev->dev), idev->dev->name, 7208 idev, &idev->cnf); 7209 if (err) 7210 neigh_sysctl_unregister(idev->nd_parms); 7211 7212 return err; 7213 } 7214 7215 static void addrconf_sysctl_unregister(struct inet6_dev *idev) 7216 { 7217 __addrconf_sysctl_unregister(dev_net(idev->dev), &idev->cnf, 7218 idev->dev->ifindex); 7219 neigh_sysctl_unregister(idev->nd_parms); 7220 } 7221 7222 7223 #endif 7224 7225 static int __net_init addrconf_init_net(struct net *net) 7226 { 7227 int err = -ENOMEM; 7228 struct ipv6_devconf *all, *dflt; 7229 7230 spin_lock_init(&net->ipv6.addrconf_hash_lock); 7231 INIT_DEFERRABLE_WORK(&net->ipv6.addr_chk_work, addrconf_verify_work); 7232 net->ipv6.inet6_addr_lst = kcalloc(IN6_ADDR_HSIZE, 7233 sizeof(struct hlist_head), 7234 GFP_KERNEL); 7235 if (!net->ipv6.inet6_addr_lst) 7236 goto err_alloc_addr; 7237 7238 all = kmemdup(&ipv6_devconf, sizeof(ipv6_devconf), GFP_KERNEL); 7239 if (!all) 7240 goto err_alloc_all; 7241 7242 dflt = kmemdup(&ipv6_devconf_dflt, sizeof(ipv6_devconf_dflt), GFP_KERNEL); 7243 if (!dflt) 7244 goto err_alloc_dflt; 7245 7246 if (!net_eq(net, &init_net)) { 7247 switch (net_inherit_devconf()) { 7248 case 1: /* copy from init_net */ 7249 memcpy(all, init_net.ipv6.devconf_all, 7250 sizeof(ipv6_devconf)); 7251 memcpy(dflt, init_net.ipv6.devconf_dflt, 7252 sizeof(ipv6_devconf_dflt)); 7253 break; 7254 case 3: /* copy from the current netns */ 7255 memcpy(all, current->nsproxy->net_ns->ipv6.devconf_all, 7256 sizeof(ipv6_devconf)); 7257 memcpy(dflt, 7258 current->nsproxy->net_ns->ipv6.devconf_dflt, 7259 sizeof(ipv6_devconf_dflt)); 7260 break; 7261 case 0: 7262 case 2: 7263 /* use compiled values */ 7264 break; 7265 } 7266 } 7267 7268 /* these will be inherited by all namespaces */ 7269 dflt->autoconf = ipv6_defaults.autoconf; 7270 dflt->disable_ipv6 = ipv6_defaults.disable_ipv6; 7271 7272 dflt->stable_secret.initialized = false; 7273 all->stable_secret.initialized = false; 7274 7275 net->ipv6.devconf_all = all; 7276 net->ipv6.devconf_dflt = dflt; 7277 7278 #ifdef CONFIG_SYSCTL 7279 err = __addrconf_sysctl_register(net, "all", NULL, all); 7280 if (err < 0) 7281 goto err_reg_all; 7282 7283 err = __addrconf_sysctl_register(net, "default", NULL, dflt); 7284 if (err < 0) 7285 goto err_reg_dflt; 7286 #endif 7287 return 0; 7288 7289 #ifdef CONFIG_SYSCTL 7290 err_reg_dflt: 7291 __addrconf_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL); 7292 err_reg_all: 7293 kfree(dflt); 7294 net->ipv6.devconf_dflt = NULL; 7295 #endif 7296 err_alloc_dflt: 7297 kfree(all); 7298 net->ipv6.devconf_all = NULL; 7299 err_alloc_all: 7300 kfree(net->ipv6.inet6_addr_lst); 7301 err_alloc_addr: 7302 return err; 7303 } 7304 7305 static void __net_exit addrconf_exit_net(struct net *net) 7306 { 7307 int i; 7308 7309 #ifdef CONFIG_SYSCTL 7310 __addrconf_sysctl_unregister(net, net->ipv6.devconf_dflt, 7311 NETCONFA_IFINDEX_DEFAULT); 7312 __addrconf_sysctl_unregister(net, net->ipv6.devconf_all, 7313 NETCONFA_IFINDEX_ALL); 7314 #endif 7315 kfree(net->ipv6.devconf_dflt); 7316 net->ipv6.devconf_dflt = NULL; 7317 kfree(net->ipv6.devconf_all); 7318 net->ipv6.devconf_all = NULL; 7319 7320 cancel_delayed_work_sync(&net->ipv6.addr_chk_work); 7321 /* 7322 * Check hash table, then free it. 7323 */ 7324 for (i = 0; i < IN6_ADDR_HSIZE; i++) 7325 WARN_ON_ONCE(!hlist_empty(&net->ipv6.inet6_addr_lst[i])); 7326 7327 kfree(net->ipv6.inet6_addr_lst); 7328 net->ipv6.inet6_addr_lst = NULL; 7329 } 7330 7331 static struct pernet_operations addrconf_ops = { 7332 .init = addrconf_init_net, 7333 .exit = addrconf_exit_net, 7334 }; 7335 7336 static struct rtnl_af_ops inet6_ops __read_mostly = { 7337 .family = AF_INET6, 7338 .fill_link_af = inet6_fill_link_af, 7339 .get_link_af_size = inet6_get_link_af_size, 7340 .validate_link_af = inet6_validate_link_af, 7341 .set_link_af = inet6_set_link_af, 7342 }; 7343 7344 /* 7345 * Init / cleanup code 7346 */ 7347 7348 int __init addrconf_init(void) 7349 { 7350 struct inet6_dev *idev; 7351 int err; 7352 7353 err = ipv6_addr_label_init(); 7354 if (err < 0) { 7355 pr_crit("%s: cannot initialize default policy table: %d\n", 7356 __func__, err); 7357 goto out; 7358 } 7359 7360 err = register_pernet_subsys(&addrconf_ops); 7361 if (err < 0) 7362 goto out_addrlabel; 7363 7364 addrconf_wq = create_workqueue("ipv6_addrconf"); 7365 if (!addrconf_wq) { 7366 err = -ENOMEM; 7367 goto out_nowq; 7368 } 7369 7370 rtnl_lock(); 7371 idev = ipv6_add_dev(blackhole_netdev); 7372 rtnl_unlock(); 7373 if (IS_ERR(idev)) { 7374 err = PTR_ERR(idev); 7375 goto errlo; 7376 } 7377 7378 ip6_route_init_special_entries(); 7379 7380 register_netdevice_notifier(&ipv6_dev_notf); 7381 7382 addrconf_verify(&init_net); 7383 7384 rtnl_af_register(&inet6_ops); 7385 7386 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETLINK, 7387 NULL, inet6_dump_ifinfo, 0); 7388 if (err < 0) 7389 goto errout; 7390 7391 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWADDR, 7392 inet6_rtm_newaddr, NULL, 0); 7393 if (err < 0) 7394 goto errout; 7395 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELADDR, 7396 inet6_rtm_deladdr, NULL, 0); 7397 if (err < 0) 7398 goto errout; 7399 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETADDR, 7400 inet6_rtm_getaddr, inet6_dump_ifaddr, 7401 RTNL_FLAG_DOIT_UNLOCKED); 7402 if (err < 0) 7403 goto errout; 7404 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETMULTICAST, 7405 NULL, inet6_dump_ifmcaddr, 0); 7406 if (err < 0) 7407 goto errout; 7408 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETANYCAST, 7409 NULL, inet6_dump_ifacaddr, 0); 7410 if (err < 0) 7411 goto errout; 7412 err = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETNETCONF, 7413 inet6_netconf_get_devconf, 7414 inet6_netconf_dump_devconf, 7415 RTNL_FLAG_DOIT_UNLOCKED); 7416 if (err < 0) 7417 goto errout; 7418 err = ipv6_addr_label_rtnl_register(); 7419 if (err < 0) 7420 goto errout; 7421 7422 return 0; 7423 errout: 7424 rtnl_unregister_all(PF_INET6); 7425 rtnl_af_unregister(&inet6_ops); 7426 unregister_netdevice_notifier(&ipv6_dev_notf); 7427 errlo: 7428 destroy_workqueue(addrconf_wq); 7429 out_nowq: 7430 unregister_pernet_subsys(&addrconf_ops); 7431 out_addrlabel: 7432 ipv6_addr_label_cleanup(); 7433 out: 7434 return err; 7435 } 7436 7437 void addrconf_cleanup(void) 7438 { 7439 struct net_device *dev; 7440 7441 unregister_netdevice_notifier(&ipv6_dev_notf); 7442 unregister_pernet_subsys(&addrconf_ops); 7443 ipv6_addr_label_cleanup(); 7444 7445 rtnl_af_unregister(&inet6_ops); 7446 7447 rtnl_lock(); 7448 7449 /* clean dev list */ 7450 for_each_netdev(&init_net, dev) { 7451 if (__in6_dev_get(dev) == NULL) 7452 continue; 7453 addrconf_ifdown(dev, true); 7454 } 7455 addrconf_ifdown(init_net.loopback_dev, true); 7456 7457 rtnl_unlock(); 7458 7459 destroy_workqueue(addrconf_wq); 7460 } 7461