1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * NET3 IP device support routines. 4 * 5 * Derived from the IP parts of dev.c 1.0.19 6 * Authors: Ross Biro 7 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> 8 * Mark Evans, <evansmp@uhura.aston.ac.uk> 9 * 10 * Additional Authors: 11 * Alan Cox, <gw4pts@gw4pts.ampr.org> 12 * Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru> 13 * 14 * Changes: 15 * Alexey Kuznetsov: pa_* fields are replaced with ifaddr 16 * lists. 17 * Cyrus Durgin: updated for kmod 18 * Matthias Andree: in devinet_ioctl, compare label and 19 * address (4.4BSD alias style support), 20 * fall back to comparing just the label 21 * if no match found. 22 */ 23 24 25 #include <linux/uaccess.h> 26 #include <linux/bitops.h> 27 #include <linux/capability.h> 28 #include <linux/module.h> 29 #include <linux/types.h> 30 #include <linux/kernel.h> 31 #include <linux/sched/signal.h> 32 #include <linux/string.h> 33 #include <linux/mm.h> 34 #include <linux/socket.h> 35 #include <linux/sockios.h> 36 #include <linux/in.h> 37 #include <linux/errno.h> 38 #include <linux/interrupt.h> 39 #include <linux/if_addr.h> 40 #include <linux/if_ether.h> 41 #include <linux/inet.h> 42 #include <linux/netdevice.h> 43 #include <linux/etherdevice.h> 44 #include <linux/skbuff.h> 45 #include <linux/init.h> 46 #include <linux/notifier.h> 47 #include <linux/inetdevice.h> 48 #include <linux/igmp.h> 49 #include "igmp_internal.h" 50 #include <linux/slab.h> 51 #include <linux/hash.h> 52 #ifdef CONFIG_SYSCTL 53 #include <linux/sysctl.h> 54 #endif 55 #include <linux/kmod.h> 56 #include <linux/netconf.h> 57 58 #include <net/arp.h> 59 #include <net/ip.h> 60 #include <net/route.h> 61 #include <net/ip_fib.h> 62 #include <net/rtnetlink.h> 63 #include <net/net_namespace.h> 64 #include <net/addrconf.h> 65 66 #define IPV6ONLY_FLAGS \ 67 (IFA_F_NODAD | IFA_F_OPTIMISTIC | IFA_F_DADFAILED | \ 68 IFA_F_HOMEADDRESS | IFA_F_TENTATIVE | \ 69 IFA_F_MANAGETEMPADDR | IFA_F_STABLE_PRIVACY) 70 71 static struct ipv4_devconf ipv4_devconf = { 72 .data = { 73 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1, 74 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1, 75 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1, 76 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1, 77 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/, 78 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/, 79 [IPV4_DEVCONF_ARP_EVICT_NOCARRIER - 1] = 1, 80 }, 81 }; 82 83 static struct ipv4_devconf ipv4_devconf_dflt = { 84 .data = { 85 [IPV4_DEVCONF_ACCEPT_REDIRECTS - 1] = 1, 86 [IPV4_DEVCONF_SEND_REDIRECTS - 1] = 1, 87 [IPV4_DEVCONF_SECURE_REDIRECTS - 1] = 1, 88 [IPV4_DEVCONF_SHARED_MEDIA - 1] = 1, 89 [IPV4_DEVCONF_ACCEPT_SOURCE_ROUTE - 1] = 1, 90 [IPV4_DEVCONF_IGMPV2_UNSOLICITED_REPORT_INTERVAL - 1] = 10000 /*ms*/, 91 [IPV4_DEVCONF_IGMPV3_UNSOLICITED_REPORT_INTERVAL - 1] = 1000 /*ms*/, 92 [IPV4_DEVCONF_ARP_EVICT_NOCARRIER - 1] = 1, 93 }, 94 }; 95 96 #define IPV4_DEVCONF_DFLT(net, attr) \ 97 IPV4_DEVCONF((*net->ipv4.devconf_dflt), attr) 98 99 static const struct nla_policy ifa_ipv4_policy[IFA_MAX+1] = { 100 [IFA_LOCAL] = { .type = NLA_U32 }, 101 [IFA_ADDRESS] = { .type = NLA_U32 }, 102 [IFA_BROADCAST] = { .type = NLA_U32 }, 103 [IFA_LABEL] = { .type = NLA_STRING, .len = IFNAMSIZ - 1 }, 104 [IFA_CACHEINFO] = { .len = sizeof(struct ifa_cacheinfo) }, 105 [IFA_FLAGS] = { .type = NLA_U32 }, 106 [IFA_RT_PRIORITY] = { .type = NLA_U32 }, 107 [IFA_TARGET_NETNSID] = { .type = NLA_S32 }, 108 [IFA_PROTO] = { .type = NLA_U8 }, 109 }; 110 111 #define IN4_ADDR_HSIZE_SHIFT 8 112 #define IN4_ADDR_HSIZE (1U << IN4_ADDR_HSIZE_SHIFT) 113 114 static u32 inet_addr_hash(const struct net *net, __be32 addr) 115 { 116 u32 val = __ipv4_addr_hash(addr, net_hash_mix(net)); 117 118 return hash_32(val, IN4_ADDR_HSIZE_SHIFT); 119 } 120 121 static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa) 122 { 123 u32 hash = inet_addr_hash(net, ifa->ifa_local); 124 125 ASSERT_RTNL(); 126 hlist_add_head_rcu(&ifa->addr_lst, &net->ipv4.inet_addr_lst[hash]); 127 } 128 129 static void inet_hash_remove(struct in_ifaddr *ifa) 130 { 131 ASSERT_RTNL(); 132 hlist_del_init_rcu(&ifa->addr_lst); 133 } 134 135 /** 136 * __ip_dev_find - find the first device with a given source address. 137 * @net: the net namespace 138 * @addr: the source address 139 * @devref: if true, take a reference on the found device 140 * 141 * If a caller uses devref=false, it should be protected by RCU, or RTNL 142 */ 143 struct net_device *__ip_dev_find(struct net *net, __be32 addr, bool devref) 144 { 145 struct net_device *result = NULL; 146 struct in_ifaddr *ifa; 147 148 rcu_read_lock(); 149 ifa = inet_lookup_ifaddr_rcu(net, addr); 150 if (!ifa) { 151 struct flowi4 fl4 = { .daddr = addr }; 152 struct fib_result res = { 0 }; 153 struct fib_table *local; 154 155 /* Fallback to FIB local table so that communication 156 * over loopback subnets work. 157 */ 158 local = fib_get_table(net, RT_TABLE_LOCAL); 159 if (local && 160 !fib_table_lookup(local, &fl4, &res, FIB_LOOKUP_NOREF) && 161 res.type == RTN_LOCAL) 162 result = FIB_RES_DEV(res); 163 } else { 164 result = ifa->ifa_dev->dev; 165 } 166 if (result && devref) 167 dev_hold(result); 168 rcu_read_unlock(); 169 return result; 170 } 171 EXPORT_SYMBOL(__ip_dev_find); 172 173 /* called under RCU lock */ 174 struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr) 175 { 176 u32 hash = inet_addr_hash(net, addr); 177 struct in_ifaddr *ifa; 178 179 hlist_for_each_entry_rcu(ifa, &net->ipv4.inet_addr_lst[hash], addr_lst) 180 if (ifa->ifa_local == addr) 181 return ifa; 182 183 return NULL; 184 } 185 186 static void rtmsg_ifa(int event, struct in_ifaddr *, struct nlmsghdr *, u32); 187 188 static BLOCKING_NOTIFIER_HEAD(inetaddr_chain); 189 static BLOCKING_NOTIFIER_HEAD(inetaddr_validator_chain); 190 static void inet_del_ifa(struct in_device *in_dev, 191 struct in_ifaddr __rcu **ifap, 192 int destroy); 193 #ifdef CONFIG_SYSCTL 194 static int devinet_sysctl_register(struct in_device *idev); 195 static void devinet_sysctl_unregister(struct in_device *idev); 196 #else 197 static int devinet_sysctl_register(struct in_device *idev) 198 { 199 return 0; 200 } 201 static void devinet_sysctl_unregister(struct in_device *idev) 202 { 203 } 204 #endif 205 206 /* Locks all the inet devices. */ 207 208 static struct in_ifaddr *inet_alloc_ifa(struct in_device *in_dev) 209 { 210 struct in_ifaddr *ifa; 211 212 ifa = kzalloc(sizeof(*ifa), GFP_KERNEL_ACCOUNT); 213 if (!ifa) 214 return NULL; 215 216 in_dev_hold(in_dev); 217 ifa->ifa_dev = in_dev; 218 219 INIT_HLIST_NODE(&ifa->addr_lst); 220 221 return ifa; 222 } 223 224 static void inet_rcu_free_ifa(struct rcu_head *head) 225 { 226 struct in_ifaddr *ifa = container_of(head, struct in_ifaddr, rcu_head); 227 228 in_dev_put(ifa->ifa_dev); 229 kfree(ifa); 230 } 231 232 static void inet_free_ifa(struct in_ifaddr *ifa) 233 { 234 /* Our reference to ifa->ifa_dev must be freed ASAP 235 * to release the reference to the netdev the same way. 236 * in_dev_put() -> in_dev_finish_destroy() -> netdev_put() 237 */ 238 call_rcu_hurry(&ifa->rcu_head, inet_rcu_free_ifa); 239 } 240 241 static void in_dev_free_rcu(struct rcu_head *head) 242 { 243 struct in_device *idev = container_of(head, struct in_device, rcu_head); 244 245 kfree(rcu_dereference_protected(idev->mc_hash, 1)); 246 kfree(idev); 247 } 248 249 void in_dev_finish_destroy(struct in_device *idev) 250 { 251 struct net_device *dev = idev->dev; 252 253 WARN_ON(idev->ifa_list); 254 WARN_ON(idev->mc_list); 255 #ifdef NET_REFCNT_DEBUG 256 pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL"); 257 #endif 258 netdev_put(dev, &idev->dev_tracker); 259 if (!idev->dead) 260 pr_err("Freeing alive in_device %p\n", idev); 261 else 262 call_rcu(&idev->rcu_head, in_dev_free_rcu); 263 } 264 EXPORT_SYMBOL(in_dev_finish_destroy); 265 266 static struct in_device *inetdev_init(struct net_device *dev) 267 { 268 struct in_device *in_dev; 269 int err = -ENOMEM; 270 271 ASSERT_RTNL(); 272 273 in_dev = kzalloc(sizeof(*in_dev), GFP_KERNEL); 274 if (!in_dev) 275 goto out; 276 memcpy(&in_dev->cnf, dev_net(dev)->ipv4.devconf_dflt, 277 sizeof(in_dev->cnf)); 278 in_dev->cnf.sysctl = NULL; 279 in_dev->dev = dev; 280 in_dev->arp_parms = neigh_parms_alloc(dev, &arp_tbl); 281 if (!in_dev->arp_parms) 282 goto out_kfree; 283 if (IPV4_DEVCONF(in_dev->cnf, FORWARDING)) 284 netif_disable_lro(dev); 285 /* Reference in_dev->dev */ 286 netdev_hold(dev, &in_dev->dev_tracker, GFP_KERNEL); 287 /* Account for reference dev->ip_ptr (below) */ 288 refcount_set(&in_dev->refcnt, 1); 289 290 if (dev != blackhole_netdev) { 291 err = devinet_sysctl_register(in_dev); 292 if (err) { 293 in_dev->dead = 1; 294 neigh_parms_release(&arp_tbl, in_dev->arp_parms); 295 in_dev_put(in_dev); 296 in_dev = NULL; 297 goto out; 298 } 299 ip_mc_init_dev(in_dev); 300 if (dev->flags & IFF_UP) 301 ip_mc_up(in_dev); 302 } 303 304 /* we can receive as soon as ip_ptr is set -- do this last */ 305 rcu_assign_pointer(dev->ip_ptr, in_dev); 306 out: 307 return in_dev ?: ERR_PTR(err); 308 out_kfree: 309 kfree(in_dev); 310 in_dev = NULL; 311 goto out; 312 } 313 314 static void inetdev_destroy(struct in_device *in_dev) 315 { 316 struct net_device *dev; 317 struct in_ifaddr *ifa; 318 319 ASSERT_RTNL(); 320 321 dev = in_dev->dev; 322 323 in_dev->dead = 1; 324 325 ip_mc_destroy_dev(in_dev); 326 327 while ((ifa = rtnl_dereference(in_dev->ifa_list)) != NULL) { 328 inet_del_ifa(in_dev, &in_dev->ifa_list, 0); 329 inet_free_ifa(ifa); 330 } 331 332 RCU_INIT_POINTER(dev->ip_ptr, NULL); 333 334 devinet_sysctl_unregister(in_dev); 335 neigh_parms_release(&arp_tbl, in_dev->arp_parms); 336 arp_ifdown(dev); 337 338 in_dev_put(in_dev); 339 } 340 341 static int __init inet_blackhole_dev_init(void) 342 { 343 struct in_device *in_dev; 344 345 rtnl_lock(); 346 in_dev = inetdev_init(blackhole_netdev); 347 rtnl_unlock(); 348 349 return PTR_ERR_OR_ZERO(in_dev); 350 } 351 late_initcall(inet_blackhole_dev_init); 352 353 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b) 354 { 355 const struct in_ifaddr *ifa; 356 357 rcu_read_lock(); 358 in_dev_for_each_ifa_rcu(ifa, in_dev) { 359 if (inet_ifa_match(a, ifa)) { 360 if (!b || inet_ifa_match(b, ifa)) { 361 rcu_read_unlock(); 362 return 1; 363 } 364 } 365 } 366 rcu_read_unlock(); 367 return 0; 368 } 369 370 static void __inet_del_ifa(struct in_device *in_dev, 371 struct in_ifaddr __rcu **ifap, 372 int destroy, struct nlmsghdr *nlh, u32 portid) 373 { 374 struct in_ifaddr *promote = NULL; 375 struct in_ifaddr *ifa, *ifa1; 376 struct in_ifaddr __rcu **last_prim; 377 struct in_ifaddr *prev_prom = NULL; 378 int do_promote = IN_DEV_PROMOTE_SECONDARIES(in_dev); 379 380 ASSERT_RTNL(); 381 382 ifa1 = rtnl_dereference(*ifap); 383 last_prim = ifap; 384 if (in_dev->dead) 385 goto no_promotions; 386 387 /* 1. Deleting primary ifaddr forces deletion all secondaries 388 * unless alias promotion is set 389 **/ 390 391 if (!(ifa1->ifa_flags & IFA_F_SECONDARY)) { 392 struct in_ifaddr __rcu **ifap1 = &ifa1->ifa_next; 393 394 while ((ifa = rtnl_dereference(*ifap1)) != NULL) { 395 if (!(ifa->ifa_flags & IFA_F_SECONDARY) && 396 ifa1->ifa_scope <= ifa->ifa_scope) 397 last_prim = &ifa->ifa_next; 398 399 if (!(ifa->ifa_flags & IFA_F_SECONDARY) || 400 ifa1->ifa_mask != ifa->ifa_mask || 401 !inet_ifa_match(ifa1->ifa_address, ifa)) { 402 ifap1 = &ifa->ifa_next; 403 prev_prom = ifa; 404 continue; 405 } 406 407 if (!do_promote) { 408 inet_hash_remove(ifa); 409 *ifap1 = ifa->ifa_next; 410 411 rtmsg_ifa(RTM_DELADDR, ifa, nlh, portid); 412 blocking_notifier_call_chain(&inetaddr_chain, 413 NETDEV_DOWN, ifa); 414 inet_free_ifa(ifa); 415 } else { 416 promote = ifa; 417 break; 418 } 419 } 420 } 421 422 /* On promotion all secondaries from subnet are changing 423 * the primary IP, we must remove all their routes silently 424 * and later to add them back with new prefsrc. Do this 425 * while all addresses are on the device list. 426 */ 427 for (ifa = promote; ifa; ifa = rtnl_dereference(ifa->ifa_next)) { 428 if (ifa1->ifa_mask == ifa->ifa_mask && 429 inet_ifa_match(ifa1->ifa_address, ifa)) 430 fib_del_ifaddr(ifa, ifa1); 431 } 432 433 no_promotions: 434 /* 2. Unlink it */ 435 436 *ifap = ifa1->ifa_next; 437 inet_hash_remove(ifa1); 438 439 /* 3. Announce address deletion */ 440 441 /* Send message first, then call notifier. 442 At first sight, FIB update triggered by notifier 443 will refer to already deleted ifaddr, that could confuse 444 netlink listeners. It is not true: look, gated sees 445 that route deleted and if it still thinks that ifaddr 446 is valid, it will try to restore deleted routes... Grr. 447 So that, this order is correct. 448 */ 449 rtmsg_ifa(RTM_DELADDR, ifa1, nlh, portid); 450 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_DOWN, ifa1); 451 452 if (promote) { 453 struct in_ifaddr *next_sec; 454 455 next_sec = rtnl_dereference(promote->ifa_next); 456 if (prev_prom) { 457 struct in_ifaddr *last_sec; 458 459 rcu_assign_pointer(prev_prom->ifa_next, next_sec); 460 461 last_sec = rtnl_dereference(*last_prim); 462 rcu_assign_pointer(promote->ifa_next, last_sec); 463 rcu_assign_pointer(*last_prim, promote); 464 } 465 466 promote->ifa_flags &= ~IFA_F_SECONDARY; 467 rtmsg_ifa(RTM_NEWADDR, promote, nlh, portid); 468 blocking_notifier_call_chain(&inetaddr_chain, 469 NETDEV_UP, promote); 470 for (ifa = next_sec; ifa; 471 ifa = rtnl_dereference(ifa->ifa_next)) { 472 if (ifa1->ifa_mask != ifa->ifa_mask || 473 !inet_ifa_match(ifa1->ifa_address, ifa)) 474 continue; 475 fib_add_ifaddr(ifa); 476 } 477 478 } 479 if (destroy) 480 inet_free_ifa(ifa1); 481 } 482 483 static void inet_del_ifa(struct in_device *in_dev, 484 struct in_ifaddr __rcu **ifap, 485 int destroy) 486 { 487 __inet_del_ifa(in_dev, ifap, destroy, NULL, 0); 488 } 489 490 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh, 491 u32 portid, struct netlink_ext_ack *extack) 492 { 493 struct in_ifaddr __rcu **last_primary, **ifap; 494 struct in_device *in_dev = ifa->ifa_dev; 495 struct net *net = dev_net(in_dev->dev); 496 struct in_validator_info ivi; 497 struct in_ifaddr *ifa1; 498 int ret; 499 500 ASSERT_RTNL(); 501 502 ifa->ifa_flags &= ~IFA_F_SECONDARY; 503 last_primary = &in_dev->ifa_list; 504 505 /* Don't set IPv6 only flags to IPv4 addresses */ 506 ifa->ifa_flags &= ~IPV6ONLY_FLAGS; 507 508 ifap = &in_dev->ifa_list; 509 ifa1 = rtnl_dereference(*ifap); 510 511 while (ifa1) { 512 if (!(ifa1->ifa_flags & IFA_F_SECONDARY) && 513 ifa->ifa_scope <= ifa1->ifa_scope) 514 last_primary = &ifa1->ifa_next; 515 if (ifa1->ifa_mask == ifa->ifa_mask && 516 inet_ifa_match(ifa1->ifa_address, ifa)) { 517 if (ifa1->ifa_local == ifa->ifa_local) { 518 inet_free_ifa(ifa); 519 return -EEXIST; 520 } 521 if (ifa1->ifa_scope != ifa->ifa_scope) { 522 NL_SET_ERR_MSG(extack, "ipv4: Invalid scope value"); 523 inet_free_ifa(ifa); 524 return -EINVAL; 525 } 526 ifa->ifa_flags |= IFA_F_SECONDARY; 527 } 528 529 ifap = &ifa1->ifa_next; 530 ifa1 = rtnl_dereference(*ifap); 531 } 532 533 /* Allow any devices that wish to register ifaddr validtors to weigh 534 * in now, before changes are committed. The rntl lock is serializing 535 * access here, so the state should not change between a validator call 536 * and a final notify on commit. This isn't invoked on promotion under 537 * the assumption that validators are checking the address itself, and 538 * not the flags. 539 */ 540 ivi.ivi_addr = ifa->ifa_address; 541 ivi.ivi_dev = ifa->ifa_dev; 542 ivi.extack = extack; 543 ret = blocking_notifier_call_chain(&inetaddr_validator_chain, 544 NETDEV_UP, &ivi); 545 ret = notifier_to_errno(ret); 546 if (ret) { 547 inet_free_ifa(ifa); 548 return ret; 549 } 550 551 if (!(ifa->ifa_flags & IFA_F_SECONDARY)) 552 ifap = last_primary; 553 554 rcu_assign_pointer(ifa->ifa_next, *ifap); 555 rcu_assign_pointer(*ifap, ifa); 556 557 inet_hash_insert(dev_net(in_dev->dev), ifa); 558 559 cancel_delayed_work(&net->ipv4.addr_chk_work); 560 queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work, 0); 561 562 /* Send message first, then call notifier. 563 Notifier will trigger FIB update, so that 564 listeners of netlink will know about new ifaddr */ 565 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, portid); 566 blocking_notifier_call_chain(&inetaddr_chain, NETDEV_UP, ifa); 567 568 return 0; 569 } 570 571 static int inet_insert_ifa(struct in_ifaddr *ifa) 572 { 573 if (!ifa->ifa_local) { 574 inet_free_ifa(ifa); 575 return 0; 576 } 577 578 return __inet_insert_ifa(ifa, NULL, 0, NULL); 579 } 580 581 static int inet_set_ifa(struct net_device *dev, struct in_ifaddr *ifa) 582 { 583 struct in_device *in_dev = __in_dev_get_rtnl_net(dev); 584 585 ipv4_devconf_setall(in_dev); 586 neigh_parms_data_state_setall(in_dev->arp_parms); 587 588 if (ipv4_is_loopback(ifa->ifa_local)) 589 ifa->ifa_scope = RT_SCOPE_HOST; 590 return inet_insert_ifa(ifa); 591 } 592 593 /* Caller must hold RCU or RTNL : 594 * We dont take a reference on found in_device 595 */ 596 struct in_device *inetdev_by_index(struct net *net, int ifindex) 597 { 598 struct net_device *dev; 599 struct in_device *in_dev = NULL; 600 601 rcu_read_lock(); 602 dev = dev_get_by_index_rcu(net, ifindex); 603 if (dev) 604 in_dev = rcu_dereference_rtnl(dev->ip_ptr); 605 rcu_read_unlock(); 606 return in_dev; 607 } 608 EXPORT_SYMBOL(inetdev_by_index); 609 610 /* Called only from RTNL semaphored context. No locks. */ 611 612 struct in_ifaddr *inet_ifa_byprefix(struct in_device *in_dev, __be32 prefix, 613 __be32 mask) 614 { 615 struct in_ifaddr *ifa; 616 617 ASSERT_RTNL(); 618 619 in_dev_for_each_ifa_rtnl(ifa, in_dev) { 620 if (ifa->ifa_mask == mask && inet_ifa_match(prefix, ifa)) 621 return ifa; 622 } 623 return NULL; 624 } 625 626 static int ip_mc_autojoin_config(struct net *net, bool join, 627 const struct in_ifaddr *ifa) 628 { 629 #if defined(CONFIG_IP_MULTICAST) 630 struct ip_mreqn mreq = { 631 .imr_multiaddr.s_addr = ifa->ifa_address, 632 .imr_ifindex = ifa->ifa_dev->dev->ifindex, 633 }; 634 struct sock *sk = net->ipv4.mc_autojoin_sk; 635 int ret; 636 637 ASSERT_RTNL_NET(net); 638 639 lock_sock(sk); 640 if (join) 641 ret = ip_mc_join_group(sk, &mreq); 642 else 643 ret = ip_mc_leave_group(sk, &mreq); 644 release_sock(sk); 645 646 return ret; 647 #else 648 return -EOPNOTSUPP; 649 #endif 650 } 651 652 static int inet_rtm_deladdr(struct sk_buff *skb, struct nlmsghdr *nlh, 653 struct netlink_ext_ack *extack) 654 { 655 struct net *net = sock_net(skb->sk); 656 struct in_ifaddr __rcu **ifap; 657 struct nlattr *tb[IFA_MAX+1]; 658 struct in_device *in_dev; 659 struct ifaddrmsg *ifm; 660 struct in_ifaddr *ifa; 661 int err; 662 663 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 664 ifa_ipv4_policy, extack); 665 if (err < 0) 666 goto out; 667 668 ifm = nlmsg_data(nlh); 669 670 rtnl_net_lock(net); 671 672 in_dev = inetdev_by_index(net, ifm->ifa_index); 673 if (!in_dev) { 674 NL_SET_ERR_MSG(extack, "ipv4: Device not found"); 675 err = -ENODEV; 676 goto unlock; 677 } 678 679 for (ifap = &in_dev->ifa_list; 680 (ifa = rtnl_net_dereference(net, *ifap)) != NULL; 681 ifap = &ifa->ifa_next) { 682 if (tb[IFA_LOCAL] && 683 ifa->ifa_local != nla_get_in_addr(tb[IFA_LOCAL])) 684 continue; 685 686 if (tb[IFA_LABEL] && nla_strcmp(tb[IFA_LABEL], ifa->ifa_label)) 687 continue; 688 689 if (tb[IFA_ADDRESS] && 690 (ifm->ifa_prefixlen != ifa->ifa_prefixlen || 691 !inet_ifa_match(nla_get_in_addr(tb[IFA_ADDRESS]), ifa))) 692 continue; 693 694 if (ipv4_is_multicast(ifa->ifa_address)) 695 ip_mc_autojoin_config(net, false, ifa); 696 697 __inet_del_ifa(in_dev, ifap, 1, nlh, NETLINK_CB(skb).portid); 698 goto unlock; 699 } 700 701 NL_SET_ERR_MSG(extack, "ipv4: Address not found"); 702 err = -EADDRNOTAVAIL; 703 unlock: 704 rtnl_net_unlock(net); 705 out: 706 return err; 707 } 708 709 static void check_lifetime(struct work_struct *work) 710 { 711 unsigned long now, next, next_sec, next_sched; 712 struct in_ifaddr *ifa; 713 struct hlist_node *n; 714 struct net *net; 715 int i; 716 717 net = container_of(to_delayed_work(work), struct net, ipv4.addr_chk_work); 718 now = jiffies; 719 next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY); 720 721 for (i = 0; i < IN4_ADDR_HSIZE; i++) { 722 struct hlist_head *head = &net->ipv4.inet_addr_lst[i]; 723 bool change_needed = false; 724 725 rcu_read_lock(); 726 hlist_for_each_entry_rcu(ifa, head, addr_lst) { 727 unsigned long age, tstamp; 728 u32 preferred_lft; 729 u32 valid_lft; 730 u32 flags; 731 732 flags = READ_ONCE(ifa->ifa_flags); 733 if (flags & IFA_F_PERMANENT) 734 continue; 735 736 preferred_lft = READ_ONCE(ifa->ifa_preferred_lft); 737 valid_lft = READ_ONCE(ifa->ifa_valid_lft); 738 tstamp = READ_ONCE(ifa->ifa_tstamp); 739 /* We try to batch several events at once. */ 740 age = (now - tstamp + 741 ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 742 743 if (valid_lft != INFINITY_LIFE_TIME && 744 age >= valid_lft) { 745 change_needed = true; 746 } else if (preferred_lft == 747 INFINITY_LIFE_TIME) { 748 continue; 749 } else if (age >= preferred_lft) { 750 if (time_before(tstamp + valid_lft * HZ, next)) 751 next = tstamp + valid_lft * HZ; 752 753 if (!(flags & IFA_F_DEPRECATED)) 754 change_needed = true; 755 } else if (time_before(tstamp + preferred_lft * HZ, 756 next)) { 757 next = tstamp + preferred_lft * HZ; 758 } 759 } 760 rcu_read_unlock(); 761 if (!change_needed) 762 continue; 763 764 rtnl_net_lock(net); 765 hlist_for_each_entry_safe(ifa, n, head, addr_lst) { 766 unsigned long age; 767 768 if (ifa->ifa_flags & IFA_F_PERMANENT) 769 continue; 770 771 /* We try to batch several events at once. */ 772 age = (now - ifa->ifa_tstamp + 773 ADDRCONF_TIMER_FUZZ_MINUS) / HZ; 774 775 if (ifa->ifa_valid_lft != INFINITY_LIFE_TIME && 776 age >= ifa->ifa_valid_lft) { 777 struct in_ifaddr __rcu **ifap; 778 struct in_ifaddr *tmp; 779 780 ifap = &ifa->ifa_dev->ifa_list; 781 tmp = rtnl_net_dereference(net, *ifap); 782 while (tmp) { 783 if (tmp == ifa) { 784 inet_del_ifa(ifa->ifa_dev, 785 ifap, 1); 786 break; 787 } 788 ifap = &tmp->ifa_next; 789 tmp = rtnl_net_dereference(net, *ifap); 790 } 791 } else if (ifa->ifa_preferred_lft != 792 INFINITY_LIFE_TIME && 793 age >= ifa->ifa_preferred_lft && 794 !(ifa->ifa_flags & IFA_F_DEPRECATED)) { 795 ifa->ifa_flags |= IFA_F_DEPRECATED; 796 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0); 797 } 798 } 799 rtnl_net_unlock(net); 800 } 801 802 next_sec = round_jiffies_up(next); 803 next_sched = next; 804 805 /* If rounded timeout is accurate enough, accept it. */ 806 if (time_before(next_sec, next + ADDRCONF_TIMER_FUZZ)) 807 next_sched = next_sec; 808 809 now = jiffies; 810 /* And minimum interval is ADDRCONF_TIMER_FUZZ_MAX. */ 811 if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX)) 812 next_sched = now + ADDRCONF_TIMER_FUZZ_MAX; 813 814 queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work, 815 next_sched - now); 816 } 817 818 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft, 819 __u32 prefered_lft) 820 { 821 unsigned long timeout; 822 u32 flags; 823 824 flags = ifa->ifa_flags & ~(IFA_F_PERMANENT | IFA_F_DEPRECATED); 825 826 timeout = addrconf_timeout_fixup(valid_lft, HZ); 827 if (addrconf_finite_timeout(timeout)) 828 WRITE_ONCE(ifa->ifa_valid_lft, timeout); 829 else 830 flags |= IFA_F_PERMANENT; 831 832 timeout = addrconf_timeout_fixup(prefered_lft, HZ); 833 if (addrconf_finite_timeout(timeout)) { 834 if (timeout == 0) 835 flags |= IFA_F_DEPRECATED; 836 WRITE_ONCE(ifa->ifa_preferred_lft, timeout); 837 } 838 WRITE_ONCE(ifa->ifa_flags, flags); 839 WRITE_ONCE(ifa->ifa_tstamp, jiffies); 840 if (!ifa->ifa_cstamp) 841 WRITE_ONCE(ifa->ifa_cstamp, ifa->ifa_tstamp); 842 } 843 844 static int inet_validate_rtm(struct nlmsghdr *nlh, struct nlattr **tb, 845 struct netlink_ext_ack *extack, 846 __u32 *valid_lft, __u32 *prefered_lft) 847 { 848 struct ifaddrmsg *ifm = nlmsg_data(nlh); 849 int err; 850 851 err = nlmsg_parse_deprecated(nlh, sizeof(*ifm), tb, IFA_MAX, 852 ifa_ipv4_policy, extack); 853 if (err < 0) 854 return err; 855 856 if (ifm->ifa_prefixlen > 32) { 857 NL_SET_ERR_MSG(extack, "ipv4: Invalid prefix length"); 858 return -EINVAL; 859 } 860 861 if (!tb[IFA_LOCAL]) { 862 NL_SET_ERR_MSG(extack, "ipv4: Local address is not supplied"); 863 return -EINVAL; 864 } 865 866 if (tb[IFA_CACHEINFO]) { 867 struct ifa_cacheinfo *ci; 868 869 ci = nla_data(tb[IFA_CACHEINFO]); 870 if (!ci->ifa_valid || ci->ifa_prefered > ci->ifa_valid) { 871 NL_SET_ERR_MSG(extack, "ipv4: address lifetime invalid"); 872 return -EINVAL; 873 } 874 875 *valid_lft = ci->ifa_valid; 876 *prefered_lft = ci->ifa_prefered; 877 } 878 879 return 0; 880 } 881 882 static struct in_ifaddr *inet_rtm_to_ifa(struct net *net, struct nlmsghdr *nlh, 883 struct nlattr **tb, 884 struct netlink_ext_ack *extack) 885 { 886 struct ifaddrmsg *ifm = nlmsg_data(nlh); 887 struct in_device *in_dev; 888 struct net_device *dev; 889 struct in_ifaddr *ifa; 890 int err; 891 892 dev = __dev_get_by_index(net, ifm->ifa_index); 893 err = -ENODEV; 894 if (!dev) { 895 NL_SET_ERR_MSG(extack, "ipv4: Device not found"); 896 goto errout; 897 } 898 899 in_dev = __in_dev_get_rtnl_net(dev); 900 err = -ENOBUFS; 901 if (!in_dev) 902 goto errout; 903 904 ifa = inet_alloc_ifa(in_dev); 905 if (!ifa) 906 /* 907 * A potential indev allocation can be left alive, it stays 908 * assigned to its device and is destroy with it. 909 */ 910 goto errout; 911 912 ipv4_devconf_setall(in_dev); 913 neigh_parms_data_state_setall(in_dev->arp_parms); 914 915 if (!tb[IFA_ADDRESS]) 916 tb[IFA_ADDRESS] = tb[IFA_LOCAL]; 917 918 ifa->ifa_prefixlen = ifm->ifa_prefixlen; 919 ifa->ifa_mask = inet_make_mask(ifm->ifa_prefixlen); 920 ifa->ifa_flags = nla_get_u32_default(tb[IFA_FLAGS], ifm->ifa_flags); 921 ifa->ifa_scope = ifm->ifa_scope; 922 ifa->ifa_local = nla_get_in_addr(tb[IFA_LOCAL]); 923 ifa->ifa_address = nla_get_in_addr(tb[IFA_ADDRESS]); 924 925 if (tb[IFA_BROADCAST]) 926 ifa->ifa_broadcast = nla_get_in_addr(tb[IFA_BROADCAST]); 927 928 if (tb[IFA_LABEL]) 929 nla_strscpy(ifa->ifa_label, tb[IFA_LABEL], IFNAMSIZ); 930 else 931 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 932 933 if (tb[IFA_RT_PRIORITY]) 934 ifa->ifa_rt_priority = nla_get_u32(tb[IFA_RT_PRIORITY]); 935 936 if (tb[IFA_PROTO]) 937 ifa->ifa_proto = nla_get_u8(tb[IFA_PROTO]); 938 939 return ifa; 940 941 errout: 942 return ERR_PTR(err); 943 } 944 945 static struct in_ifaddr *find_matching_ifa(struct net *net, struct in_ifaddr *ifa) 946 { 947 struct in_device *in_dev = ifa->ifa_dev; 948 struct in_ifaddr *ifa1; 949 950 in_dev_for_each_ifa_rtnl_net(net, ifa1, in_dev) { 951 if (ifa1->ifa_mask == ifa->ifa_mask && 952 inet_ifa_match(ifa1->ifa_address, ifa) && 953 ifa1->ifa_local == ifa->ifa_local) 954 return ifa1; 955 } 956 957 return NULL; 958 } 959 960 static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh, 961 struct netlink_ext_ack *extack) 962 { 963 __u32 prefered_lft = INFINITY_LIFE_TIME; 964 __u32 valid_lft = INFINITY_LIFE_TIME; 965 struct net *net = sock_net(skb->sk); 966 struct in_ifaddr *ifa_existing; 967 struct nlattr *tb[IFA_MAX + 1]; 968 struct in_ifaddr *ifa; 969 int ret; 970 971 ret = inet_validate_rtm(nlh, tb, extack, &valid_lft, &prefered_lft); 972 if (ret < 0) 973 return ret; 974 975 if (!nla_get_in_addr(tb[IFA_LOCAL])) 976 return 0; 977 978 rtnl_net_lock(net); 979 980 ifa = inet_rtm_to_ifa(net, nlh, tb, extack); 981 if (IS_ERR(ifa)) { 982 ret = PTR_ERR(ifa); 983 goto unlock; 984 } 985 986 ifa_existing = find_matching_ifa(net, ifa); 987 if (!ifa_existing) { 988 /* It would be best to check for !NLM_F_CREATE here but 989 * userspace already relies on not having to provide this. 990 */ 991 set_ifa_lifetime(ifa, valid_lft, prefered_lft); 992 if (ifa->ifa_flags & IFA_F_MCAUTOJOIN) { 993 ret = ip_mc_autojoin_config(net, true, ifa); 994 if (ret < 0) { 995 NL_SET_ERR_MSG(extack, "ipv4: Multicast auto join failed"); 996 inet_free_ifa(ifa); 997 goto unlock; 998 } 999 } 1000 1001 ret = __inet_insert_ifa(ifa, nlh, NETLINK_CB(skb).portid, extack); 1002 } else { 1003 u32 new_metric = ifa->ifa_rt_priority; 1004 u8 new_proto = ifa->ifa_proto; 1005 1006 inet_free_ifa(ifa); 1007 1008 if (nlh->nlmsg_flags & NLM_F_EXCL || 1009 !(nlh->nlmsg_flags & NLM_F_REPLACE)) { 1010 NL_SET_ERR_MSG(extack, "ipv4: Address already assigned"); 1011 ret = -EEXIST; 1012 goto unlock; 1013 } 1014 ifa = ifa_existing; 1015 1016 if (ifa->ifa_rt_priority != new_metric) { 1017 fib_modify_prefix_metric(ifa, new_metric); 1018 ifa->ifa_rt_priority = new_metric; 1019 } 1020 1021 ifa->ifa_proto = new_proto; 1022 1023 set_ifa_lifetime(ifa, valid_lft, prefered_lft); 1024 cancel_delayed_work(&net->ipv4.addr_chk_work); 1025 queue_delayed_work(system_power_efficient_wq, 1026 &net->ipv4.addr_chk_work, 0); 1027 rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid); 1028 } 1029 1030 unlock: 1031 rtnl_net_unlock(net); 1032 1033 return ret; 1034 } 1035 1036 /* 1037 * Determine a default network mask, based on the IP address. 1038 */ 1039 1040 static int inet_abc_len(__be32 addr) 1041 { 1042 int rc = -1; /* Something else, probably a multicast. */ 1043 1044 if (ipv4_is_zeronet(addr) || ipv4_is_lbcast(addr)) 1045 rc = 0; 1046 else { 1047 __u32 haddr = ntohl(addr); 1048 if (IN_CLASSA(haddr)) 1049 rc = 8; 1050 else if (IN_CLASSB(haddr)) 1051 rc = 16; 1052 else if (IN_CLASSC(haddr)) 1053 rc = 24; 1054 else if (IN_CLASSE(haddr)) 1055 rc = 32; 1056 } 1057 1058 return rc; 1059 } 1060 1061 1062 int devinet_ioctl(struct net *net, unsigned int cmd, struct ifreq *ifr) 1063 { 1064 struct sockaddr_in sin_orig; 1065 struct sockaddr_in *sin = (struct sockaddr_in *)&ifr->ifr_addr; 1066 struct in_ifaddr __rcu **ifap = NULL; 1067 struct in_device *in_dev; 1068 struct in_ifaddr *ifa = NULL; 1069 struct net_device *dev; 1070 char *colon; 1071 int ret = -EFAULT; 1072 int tryaddrmatch = 0; 1073 1074 ifr->ifr_name[IFNAMSIZ - 1] = 0; 1075 1076 /* save original address for comparison */ 1077 memcpy(&sin_orig, sin, sizeof(*sin)); 1078 1079 colon = strchr(ifr->ifr_name, ':'); 1080 if (colon) 1081 *colon = 0; 1082 1083 dev_load(net, ifr->ifr_name); 1084 1085 switch (cmd) { 1086 case SIOCGIFADDR: /* Get interface address */ 1087 case SIOCGIFBRDADDR: /* Get the broadcast address */ 1088 case SIOCGIFDSTADDR: /* Get the destination address */ 1089 case SIOCGIFNETMASK: /* Get the netmask for the interface */ 1090 /* Note that these ioctls will not sleep, 1091 so that we do not impose a lock. 1092 One day we will be forced to put shlock here (I mean SMP) 1093 */ 1094 tryaddrmatch = (sin_orig.sin_family == AF_INET); 1095 memset(sin, 0, sizeof(*sin)); 1096 sin->sin_family = AF_INET; 1097 break; 1098 1099 case SIOCSIFFLAGS: 1100 ret = -EPERM; 1101 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1102 goto out; 1103 break; 1104 case SIOCSIFADDR: /* Set interface address (and family) */ 1105 case SIOCSIFBRDADDR: /* Set the broadcast address */ 1106 case SIOCSIFDSTADDR: /* Set the destination address */ 1107 case SIOCSIFNETMASK: /* Set the netmask for the interface */ 1108 ret = -EPERM; 1109 if (!ns_capable(net->user_ns, CAP_NET_ADMIN)) 1110 goto out; 1111 ret = -EINVAL; 1112 if (sin->sin_family != AF_INET) 1113 goto out; 1114 break; 1115 default: 1116 ret = -EINVAL; 1117 goto out; 1118 } 1119 1120 rtnl_net_lock(net); 1121 1122 ret = -ENODEV; 1123 dev = __dev_get_by_name(net, ifr->ifr_name); 1124 if (!dev) 1125 goto done; 1126 1127 if (colon) 1128 *colon = ':'; 1129 1130 in_dev = __in_dev_get_rtnl_net(dev); 1131 if (in_dev) { 1132 if (tryaddrmatch) { 1133 /* Matthias Andree */ 1134 /* compare label and address (4.4BSD style) */ 1135 /* note: we only do this for a limited set of ioctls 1136 and only if the original address family was AF_INET. 1137 This is checked above. */ 1138 1139 for (ifap = &in_dev->ifa_list; 1140 (ifa = rtnl_net_dereference(net, *ifap)) != NULL; 1141 ifap = &ifa->ifa_next) { 1142 if (!strcmp(ifr->ifr_name, ifa->ifa_label) && 1143 sin_orig.sin_addr.s_addr == 1144 ifa->ifa_local) { 1145 break; /* found */ 1146 } 1147 } 1148 } 1149 /* we didn't get a match, maybe the application is 1150 4.3BSD-style and passed in junk so we fall back to 1151 comparing just the label */ 1152 if (!ifa) { 1153 for (ifap = &in_dev->ifa_list; 1154 (ifa = rtnl_net_dereference(net, *ifap)) != NULL; 1155 ifap = &ifa->ifa_next) 1156 if (!strcmp(ifr->ifr_name, ifa->ifa_label)) 1157 break; 1158 } 1159 } 1160 1161 ret = -EADDRNOTAVAIL; 1162 if (!ifa && cmd != SIOCSIFADDR && cmd != SIOCSIFFLAGS) 1163 goto done; 1164 1165 switch (cmd) { 1166 case SIOCGIFADDR: /* Get interface address */ 1167 ret = 0; 1168 sin->sin_addr.s_addr = ifa->ifa_local; 1169 break; 1170 1171 case SIOCGIFBRDADDR: /* Get the broadcast address */ 1172 ret = 0; 1173 sin->sin_addr.s_addr = ifa->ifa_broadcast; 1174 break; 1175 1176 case SIOCGIFDSTADDR: /* Get the destination address */ 1177 ret = 0; 1178 sin->sin_addr.s_addr = ifa->ifa_address; 1179 break; 1180 1181 case SIOCGIFNETMASK: /* Get the netmask for the interface */ 1182 ret = 0; 1183 sin->sin_addr.s_addr = ifa->ifa_mask; 1184 break; 1185 1186 case SIOCSIFFLAGS: 1187 if (colon) { 1188 ret = -EADDRNOTAVAIL; 1189 if (!ifa) 1190 break; 1191 ret = 0; 1192 if (!(ifr->ifr_flags & IFF_UP)) 1193 inet_del_ifa(in_dev, ifap, 1); 1194 break; 1195 } 1196 1197 /* NETDEV_UP/DOWN/CHANGE could touch a peer dev */ 1198 ASSERT_RTNL(); 1199 ret = dev_change_flags(dev, ifr->ifr_flags, NULL); 1200 break; 1201 1202 case SIOCSIFADDR: /* Set interface address (and family) */ 1203 ret = -EINVAL; 1204 if (inet_abc_len(sin->sin_addr.s_addr) < 0) 1205 break; 1206 1207 if (!ifa) { 1208 ret = -ENOBUFS; 1209 if (!in_dev) 1210 break; 1211 ifa = inet_alloc_ifa(in_dev); 1212 if (!ifa) 1213 break; 1214 1215 if (colon) 1216 memcpy(ifa->ifa_label, ifr->ifr_name, IFNAMSIZ); 1217 else 1218 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 1219 } else { 1220 ret = 0; 1221 if (ifa->ifa_local == sin->sin_addr.s_addr) 1222 break; 1223 inet_del_ifa(in_dev, ifap, 0); 1224 ifa->ifa_broadcast = 0; 1225 ifa->ifa_scope = 0; 1226 } 1227 1228 ifa->ifa_address = ifa->ifa_local = sin->sin_addr.s_addr; 1229 1230 if (!(dev->flags & IFF_POINTOPOINT)) { 1231 ifa->ifa_prefixlen = inet_abc_len(ifa->ifa_address); 1232 ifa->ifa_mask = inet_make_mask(ifa->ifa_prefixlen); 1233 if ((dev->flags & IFF_BROADCAST) && 1234 ifa->ifa_prefixlen < 31) 1235 ifa->ifa_broadcast = ifa->ifa_address | 1236 ~ifa->ifa_mask; 1237 } else { 1238 ifa->ifa_prefixlen = 32; 1239 ifa->ifa_mask = inet_make_mask(32); 1240 } 1241 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, INFINITY_LIFE_TIME); 1242 ret = inet_set_ifa(dev, ifa); 1243 break; 1244 1245 case SIOCSIFBRDADDR: /* Set the broadcast address */ 1246 ret = 0; 1247 if (ifa->ifa_broadcast != sin->sin_addr.s_addr) { 1248 inet_del_ifa(in_dev, ifap, 0); 1249 ifa->ifa_broadcast = sin->sin_addr.s_addr; 1250 inet_insert_ifa(ifa); 1251 } 1252 break; 1253 1254 case SIOCSIFDSTADDR: /* Set the destination address */ 1255 ret = 0; 1256 if (ifa->ifa_address == sin->sin_addr.s_addr) 1257 break; 1258 ret = -EINVAL; 1259 if (inet_abc_len(sin->sin_addr.s_addr) < 0) 1260 break; 1261 ret = 0; 1262 inet_del_ifa(in_dev, ifap, 0); 1263 ifa->ifa_address = sin->sin_addr.s_addr; 1264 inet_insert_ifa(ifa); 1265 break; 1266 1267 case SIOCSIFNETMASK: /* Set the netmask for the interface */ 1268 1269 /* 1270 * The mask we set must be legal. 1271 */ 1272 ret = -EINVAL; 1273 if (bad_mask(sin->sin_addr.s_addr, 0)) 1274 break; 1275 ret = 0; 1276 if (ifa->ifa_mask != sin->sin_addr.s_addr) { 1277 __be32 old_mask = ifa->ifa_mask; 1278 inet_del_ifa(in_dev, ifap, 0); 1279 ifa->ifa_mask = sin->sin_addr.s_addr; 1280 ifa->ifa_prefixlen = inet_mask_len(ifa->ifa_mask); 1281 1282 /* See if current broadcast address matches 1283 * with current netmask, then recalculate 1284 * the broadcast address. Otherwise it's a 1285 * funny address, so don't touch it since 1286 * the user seems to know what (s)he's doing... 1287 */ 1288 if ((dev->flags & IFF_BROADCAST) && 1289 (ifa->ifa_prefixlen < 31) && 1290 (ifa->ifa_broadcast == 1291 (ifa->ifa_local|~old_mask))) { 1292 ifa->ifa_broadcast = (ifa->ifa_local | 1293 ~sin->sin_addr.s_addr); 1294 } 1295 inet_insert_ifa(ifa); 1296 } 1297 break; 1298 } 1299 done: 1300 rtnl_net_unlock(net); 1301 out: 1302 return ret; 1303 } 1304 1305 int inet_gifconf(struct net_device *dev, char __user *buf, int len, int size) 1306 { 1307 struct in_device *in_dev = __in_dev_get_rtnl_net(dev); 1308 const struct in_ifaddr *ifa; 1309 struct ifreq ifr; 1310 int done = 0; 1311 1312 if (WARN_ON(size > sizeof(struct ifreq))) 1313 goto out; 1314 1315 if (!in_dev) 1316 goto out; 1317 1318 in_dev_for_each_ifa_rtnl_net(dev_net(dev), ifa, in_dev) { 1319 if (!buf) { 1320 done += size; 1321 continue; 1322 } 1323 if (len < size) 1324 break; 1325 memset(&ifr, 0, sizeof(struct ifreq)); 1326 strcpy(ifr.ifr_name, ifa->ifa_label); 1327 1328 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_family = AF_INET; 1329 (*(struct sockaddr_in *)&ifr.ifr_addr).sin_addr.s_addr = 1330 ifa->ifa_local; 1331 1332 if (copy_to_user(buf + done, &ifr, size)) { 1333 done = -EFAULT; 1334 break; 1335 } 1336 len -= size; 1337 done += size; 1338 } 1339 out: 1340 return done; 1341 } 1342 1343 static __be32 in_dev_select_addr(const struct in_device *in_dev, 1344 int scope) 1345 { 1346 const struct in_ifaddr *ifa; 1347 1348 in_dev_for_each_ifa_rcu(ifa, in_dev) { 1349 if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY) 1350 continue; 1351 if (ifa->ifa_scope != RT_SCOPE_LINK && 1352 ifa->ifa_scope <= scope) 1353 return ifa->ifa_local; 1354 } 1355 1356 return 0; 1357 } 1358 1359 __be32 inet_select_addr(const struct net_device *dev, __be32 dst, int scope) 1360 { 1361 const struct in_ifaddr *ifa; 1362 __be32 addr = 0; 1363 unsigned char localnet_scope = RT_SCOPE_HOST; 1364 struct in_device *in_dev; 1365 struct net *net; 1366 int master_idx; 1367 1368 rcu_read_lock(); 1369 net = dev_net_rcu(dev); 1370 in_dev = __in_dev_get_rcu(dev); 1371 if (!in_dev) 1372 goto no_in_dev; 1373 1374 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev))) 1375 localnet_scope = RT_SCOPE_LINK; 1376 1377 in_dev_for_each_ifa_rcu(ifa, in_dev) { 1378 if (READ_ONCE(ifa->ifa_flags) & IFA_F_SECONDARY) 1379 continue; 1380 if (min(ifa->ifa_scope, localnet_scope) > scope) 1381 continue; 1382 if (!dst || inet_ifa_match(dst, ifa)) { 1383 addr = ifa->ifa_local; 1384 break; 1385 } 1386 if (!addr) 1387 addr = ifa->ifa_local; 1388 } 1389 1390 if (addr) 1391 goto out_unlock; 1392 no_in_dev: 1393 master_idx = l3mdev_master_ifindex_rcu(dev); 1394 1395 /* For VRFs, the VRF device takes the place of the loopback device, 1396 * with addresses on it being preferred. Note in such cases the 1397 * loopback device will be among the devices that fail the master_idx 1398 * equality check in the loop below. 1399 */ 1400 if (master_idx && 1401 (dev = dev_get_by_index_rcu(net, master_idx)) && 1402 (in_dev = __in_dev_get_rcu(dev))) { 1403 addr = in_dev_select_addr(in_dev, scope); 1404 if (addr) 1405 goto out_unlock; 1406 } 1407 1408 /* Not loopback addresses on loopback should be preferred 1409 in this case. It is important that lo is the first interface 1410 in dev_base list. 1411 */ 1412 for_each_netdev_rcu(net, dev) { 1413 if (l3mdev_master_ifindex_rcu(dev) != master_idx) 1414 continue; 1415 1416 in_dev = __in_dev_get_rcu(dev); 1417 if (!in_dev) 1418 continue; 1419 1420 addr = in_dev_select_addr(in_dev, scope); 1421 if (addr) 1422 goto out_unlock; 1423 } 1424 out_unlock: 1425 rcu_read_unlock(); 1426 return addr; 1427 } 1428 EXPORT_SYMBOL(inet_select_addr); 1429 1430 static __be32 confirm_addr_indev(struct in_device *in_dev, __be32 dst, 1431 __be32 local, int scope) 1432 { 1433 unsigned char localnet_scope = RT_SCOPE_HOST; 1434 const struct in_ifaddr *ifa; 1435 __be32 addr = 0; 1436 int same = 0; 1437 1438 if (unlikely(IN_DEV_ROUTE_LOCALNET(in_dev))) 1439 localnet_scope = RT_SCOPE_LINK; 1440 1441 in_dev_for_each_ifa_rcu(ifa, in_dev) { 1442 unsigned char min_scope = min(ifa->ifa_scope, localnet_scope); 1443 1444 if (!addr && 1445 (local == ifa->ifa_local || !local) && 1446 min_scope <= scope) { 1447 addr = ifa->ifa_local; 1448 if (same) 1449 break; 1450 } 1451 if (!same) { 1452 same = (!local || inet_ifa_match(local, ifa)) && 1453 (!dst || inet_ifa_match(dst, ifa)); 1454 if (same && addr) { 1455 if (local || !dst) 1456 break; 1457 /* Is the selected addr into dst subnet? */ 1458 if (inet_ifa_match(addr, ifa)) 1459 break; 1460 /* No, then can we use new local src? */ 1461 if (min_scope <= scope) { 1462 addr = ifa->ifa_local; 1463 break; 1464 } 1465 /* search for large dst subnet for addr */ 1466 same = 0; 1467 } 1468 } 1469 } 1470 1471 return same ? addr : 0; 1472 } 1473 1474 /* 1475 * Confirm that local IP address exists using wildcards: 1476 * - net: netns to check, cannot be NULL 1477 * - in_dev: only on this interface, NULL=any interface 1478 * - dst: only in the same subnet as dst, 0=any dst 1479 * - local: address, 0=autoselect the local address 1480 * - scope: maximum allowed scope value for the local address 1481 */ 1482 __be32 inet_confirm_addr(struct net *net, struct in_device *in_dev, 1483 __be32 dst, __be32 local, int scope) 1484 { 1485 __be32 addr = 0; 1486 struct net_device *dev; 1487 1488 if (in_dev) 1489 return confirm_addr_indev(in_dev, dst, local, scope); 1490 1491 rcu_read_lock(); 1492 for_each_netdev_rcu(net, dev) { 1493 in_dev = __in_dev_get_rcu(dev); 1494 if (in_dev) { 1495 addr = confirm_addr_indev(in_dev, dst, local, scope); 1496 if (addr) 1497 break; 1498 } 1499 } 1500 rcu_read_unlock(); 1501 1502 return addr; 1503 } 1504 EXPORT_SYMBOL(inet_confirm_addr); 1505 1506 /* 1507 * Device notifier 1508 */ 1509 1510 int register_inetaddr_notifier(struct notifier_block *nb) 1511 { 1512 return blocking_notifier_chain_register(&inetaddr_chain, nb); 1513 } 1514 EXPORT_SYMBOL(register_inetaddr_notifier); 1515 1516 int unregister_inetaddr_notifier(struct notifier_block *nb) 1517 { 1518 return blocking_notifier_chain_unregister(&inetaddr_chain, nb); 1519 } 1520 EXPORT_SYMBOL(unregister_inetaddr_notifier); 1521 1522 int register_inetaddr_validator_notifier(struct notifier_block *nb) 1523 { 1524 return blocking_notifier_chain_register(&inetaddr_validator_chain, nb); 1525 } 1526 EXPORT_SYMBOL(register_inetaddr_validator_notifier); 1527 1528 int unregister_inetaddr_validator_notifier(struct notifier_block *nb) 1529 { 1530 return blocking_notifier_chain_unregister(&inetaddr_validator_chain, 1531 nb); 1532 } 1533 EXPORT_SYMBOL(unregister_inetaddr_validator_notifier); 1534 1535 /* Rename ifa_labels for a device name change. Make some effort to preserve 1536 * existing alias numbering and to create unique labels if possible. 1537 */ 1538 static void inetdev_changename(struct net_device *dev, struct in_device *in_dev) 1539 { 1540 struct in_ifaddr *ifa; 1541 int named = 0; 1542 1543 in_dev_for_each_ifa_rtnl(ifa, in_dev) { 1544 char old[IFNAMSIZ], *dot; 1545 1546 memcpy(old, ifa->ifa_label, IFNAMSIZ); 1547 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 1548 if (named++ == 0) 1549 goto skip; 1550 dot = strchr(old, ':'); 1551 if (!dot) { 1552 sprintf(old, ":%d", named); 1553 dot = old; 1554 } 1555 if (strlen(dot) + strlen(dev->name) < IFNAMSIZ) 1556 strcat(ifa->ifa_label, dot); 1557 else 1558 strcpy(ifa->ifa_label + (IFNAMSIZ - strlen(dot) - 1), dot); 1559 skip: 1560 rtmsg_ifa(RTM_NEWADDR, ifa, NULL, 0); 1561 } 1562 } 1563 1564 static void inetdev_send_gratuitous_arp(struct net_device *dev, 1565 struct in_device *in_dev) 1566 1567 { 1568 const struct in_ifaddr *ifa; 1569 1570 in_dev_for_each_ifa_rtnl(ifa, in_dev) { 1571 arp_send(ARPOP_REQUEST, ETH_P_ARP, 1572 ifa->ifa_local, dev, 1573 ifa->ifa_local, NULL, 1574 dev->dev_addr, NULL); 1575 } 1576 } 1577 1578 /* Called only under RTNL semaphore */ 1579 1580 static int inetdev_event(struct notifier_block *this, unsigned long event, 1581 void *ptr) 1582 { 1583 struct net_device *dev = netdev_notifier_info_to_dev(ptr); 1584 struct in_device *in_dev = __in_dev_get_rtnl(dev); 1585 1586 ASSERT_RTNL(); 1587 1588 if (!in_dev) { 1589 if (event == NETDEV_REGISTER) { 1590 in_dev = inetdev_init(dev); 1591 if (IS_ERR(in_dev)) 1592 return notifier_from_errno(PTR_ERR(in_dev)); 1593 if (dev->flags & IFF_LOOPBACK) { 1594 IN_DEV_CONF_SET(in_dev, NOXFRM, 1); 1595 IN_DEV_CONF_SET(in_dev, NOPOLICY, 1); 1596 } 1597 } else if (event == NETDEV_CHANGEMTU) { 1598 /* Re-enabling IP */ 1599 if (inetdev_valid_mtu(dev->mtu)) 1600 in_dev = inetdev_init(dev); 1601 } 1602 goto out; 1603 } 1604 1605 switch (event) { 1606 case NETDEV_REGISTER: 1607 pr_debug("%s: bug\n", __func__); 1608 RCU_INIT_POINTER(dev->ip_ptr, NULL); 1609 break; 1610 case NETDEV_UP: 1611 if (!inetdev_valid_mtu(dev->mtu)) 1612 break; 1613 if (dev->flags & IFF_LOOPBACK) { 1614 struct in_ifaddr *ifa = inet_alloc_ifa(in_dev); 1615 1616 if (ifa) { 1617 ifa->ifa_local = 1618 ifa->ifa_address = htonl(INADDR_LOOPBACK); 1619 ifa->ifa_prefixlen = 8; 1620 ifa->ifa_mask = inet_make_mask(8); 1621 ifa->ifa_scope = RT_SCOPE_HOST; 1622 memcpy(ifa->ifa_label, dev->name, IFNAMSIZ); 1623 set_ifa_lifetime(ifa, INFINITY_LIFE_TIME, 1624 INFINITY_LIFE_TIME); 1625 ipv4_devconf_setall(in_dev); 1626 neigh_parms_data_state_setall(in_dev->arp_parms); 1627 inet_insert_ifa(ifa); 1628 } 1629 } 1630 ip_mc_up(in_dev); 1631 fallthrough; 1632 case NETDEV_CHANGEADDR: 1633 if (!IN_DEV_ARP_NOTIFY(in_dev)) 1634 break; 1635 fallthrough; 1636 case NETDEV_NOTIFY_PEERS: 1637 /* Send gratuitous ARP to notify of link change */ 1638 inetdev_send_gratuitous_arp(dev, in_dev); 1639 break; 1640 case NETDEV_DOWN: 1641 ip_mc_down(in_dev); 1642 break; 1643 case NETDEV_PRE_TYPE_CHANGE: 1644 ip_mc_unmap(in_dev); 1645 break; 1646 case NETDEV_POST_TYPE_CHANGE: 1647 ip_mc_remap(in_dev); 1648 break; 1649 case NETDEV_CHANGEMTU: 1650 if (inetdev_valid_mtu(dev->mtu)) 1651 break; 1652 /* disable IP when MTU is not enough */ 1653 fallthrough; 1654 case NETDEV_UNREGISTER: 1655 inetdev_destroy(in_dev); 1656 break; 1657 case NETDEV_CHANGENAME: 1658 /* Do not notify about label change, this event is 1659 * not interesting to applications using netlink. 1660 */ 1661 inetdev_changename(dev, in_dev); 1662 1663 devinet_sysctl_unregister(in_dev); 1664 devinet_sysctl_register(in_dev); 1665 break; 1666 } 1667 out: 1668 return NOTIFY_DONE; 1669 } 1670 1671 static struct notifier_block ip_netdev_notifier = { 1672 .notifier_call = inetdev_event, 1673 }; 1674 1675 static size_t inet_nlmsg_size(void) 1676 { 1677 return NLMSG_ALIGN(sizeof(struct ifaddrmsg)) 1678 + nla_total_size(4) /* IFA_ADDRESS */ 1679 + nla_total_size(4) /* IFA_LOCAL */ 1680 + nla_total_size(4) /* IFA_BROADCAST */ 1681 + nla_total_size(IFNAMSIZ) /* IFA_LABEL */ 1682 + nla_total_size(4) /* IFA_FLAGS */ 1683 + nla_total_size(1) /* IFA_PROTO */ 1684 + nla_total_size(4) /* IFA_RT_PRIORITY */ 1685 + nla_total_size(sizeof(struct ifa_cacheinfo)); /* IFA_CACHEINFO */ 1686 } 1687 1688 static inline u32 cstamp_delta(unsigned long cstamp) 1689 { 1690 return (cstamp - INITIAL_JIFFIES) * 100UL / HZ; 1691 } 1692 1693 static int put_cacheinfo(struct sk_buff *skb, unsigned long cstamp, 1694 unsigned long tstamp, u32 preferred, u32 valid) 1695 { 1696 struct ifa_cacheinfo ci; 1697 1698 ci.cstamp = cstamp_delta(cstamp); 1699 ci.tstamp = cstamp_delta(tstamp); 1700 ci.ifa_prefered = preferred; 1701 ci.ifa_valid = valid; 1702 1703 return nla_put(skb, IFA_CACHEINFO, sizeof(ci), &ci); 1704 } 1705 1706 static int inet_fill_ifaddr(struct sk_buff *skb, const struct in_ifaddr *ifa, 1707 struct inet_fill_args *args) 1708 { 1709 struct ifaddrmsg *ifm; 1710 struct nlmsghdr *nlh; 1711 unsigned long tstamp; 1712 u32 preferred, valid; 1713 u32 flags; 1714 1715 nlh = nlmsg_put(skb, args->portid, args->seq, args->event, sizeof(*ifm), 1716 args->flags); 1717 if (!nlh) 1718 return -EMSGSIZE; 1719 1720 ifm = nlmsg_data(nlh); 1721 ifm->ifa_family = AF_INET; 1722 ifm->ifa_prefixlen = ifa->ifa_prefixlen; 1723 1724 flags = READ_ONCE(ifa->ifa_flags); 1725 /* Warning : ifm->ifa_flags is an __u8, it holds only 8 bits. 1726 * The 32bit value is given in IFA_FLAGS attribute. 1727 */ 1728 ifm->ifa_flags = (__u8)flags; 1729 1730 ifm->ifa_scope = ifa->ifa_scope; 1731 ifm->ifa_index = ifa->ifa_dev->dev->ifindex; 1732 1733 if (args->netnsid >= 0 && 1734 nla_put_s32(skb, IFA_TARGET_NETNSID, args->netnsid)) 1735 goto nla_put_failure; 1736 1737 tstamp = READ_ONCE(ifa->ifa_tstamp); 1738 if (!(flags & IFA_F_PERMANENT)) { 1739 preferred = READ_ONCE(ifa->ifa_preferred_lft); 1740 valid = READ_ONCE(ifa->ifa_valid_lft); 1741 if (preferred != INFINITY_LIFE_TIME) { 1742 long tval = (jiffies - tstamp) / HZ; 1743 1744 if (preferred > tval) 1745 preferred -= tval; 1746 else 1747 preferred = 0; 1748 if (valid != INFINITY_LIFE_TIME) { 1749 if (valid > tval) 1750 valid -= tval; 1751 else 1752 valid = 0; 1753 } 1754 } 1755 } else { 1756 preferred = INFINITY_LIFE_TIME; 1757 valid = INFINITY_LIFE_TIME; 1758 } 1759 if ((ifa->ifa_address && 1760 nla_put_in_addr(skb, IFA_ADDRESS, ifa->ifa_address)) || 1761 (ifa->ifa_local && 1762 nla_put_in_addr(skb, IFA_LOCAL, ifa->ifa_local)) || 1763 (ifa->ifa_broadcast && 1764 nla_put_in_addr(skb, IFA_BROADCAST, ifa->ifa_broadcast)) || 1765 (ifa->ifa_label[0] && 1766 nla_put_string(skb, IFA_LABEL, ifa->ifa_label)) || 1767 (ifa->ifa_proto && 1768 nla_put_u8(skb, IFA_PROTO, ifa->ifa_proto)) || 1769 nla_put_u32(skb, IFA_FLAGS, flags) || 1770 (ifa->ifa_rt_priority && 1771 nla_put_u32(skb, IFA_RT_PRIORITY, ifa->ifa_rt_priority)) || 1772 put_cacheinfo(skb, READ_ONCE(ifa->ifa_cstamp), tstamp, 1773 preferred, valid)) 1774 goto nla_put_failure; 1775 1776 nlmsg_end(skb, nlh); 1777 return 0; 1778 1779 nla_put_failure: 1780 nlmsg_cancel(skb, nlh); 1781 return -EMSGSIZE; 1782 } 1783 1784 static int inet_valid_dump_ifaddr_req(const struct nlmsghdr *nlh, 1785 struct inet_fill_args *fillargs, 1786 struct net **tgt_net, struct sock *sk, 1787 struct netlink_callback *cb) 1788 { 1789 struct netlink_ext_ack *extack = cb->extack; 1790 struct nlattr *tb[IFA_MAX+1]; 1791 struct ifaddrmsg *ifm; 1792 int err, i; 1793 1794 ifm = nlmsg_payload(nlh, sizeof(*ifm)); 1795 if (!ifm) { 1796 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for address dump request"); 1797 return -EINVAL; 1798 } 1799 1800 if (ifm->ifa_prefixlen || ifm->ifa_flags || ifm->ifa_scope) { 1801 NL_SET_ERR_MSG(extack, "ipv4: Invalid values in header for address dump request"); 1802 return -EINVAL; 1803 } 1804 1805 fillargs->ifindex = ifm->ifa_index; 1806 if (fillargs->ifindex) { 1807 cb->answer_flags |= NLM_F_DUMP_FILTERED; 1808 fillargs->flags |= NLM_F_DUMP_FILTERED; 1809 } 1810 1811 err = nlmsg_parse_deprecated_strict(nlh, sizeof(*ifm), tb, IFA_MAX, 1812 ifa_ipv4_policy, extack); 1813 if (err < 0) 1814 return err; 1815 1816 for (i = 0; i <= IFA_MAX; ++i) { 1817 if (!tb[i]) 1818 continue; 1819 1820 if (i == IFA_TARGET_NETNSID) { 1821 struct net *net; 1822 1823 fillargs->netnsid = nla_get_s32(tb[i]); 1824 1825 net = rtnl_get_net_ns_capable(sk, fillargs->netnsid); 1826 if (IS_ERR(net)) { 1827 fillargs->netnsid = -1; 1828 NL_SET_ERR_MSG(extack, "ipv4: Invalid target network namespace id"); 1829 return PTR_ERR(net); 1830 } 1831 *tgt_net = net; 1832 } else { 1833 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in dump request"); 1834 return -EINVAL; 1835 } 1836 } 1837 1838 return 0; 1839 } 1840 1841 static int in_dev_dump_ifmcaddr(struct in_device *in_dev, struct sk_buff *skb, 1842 struct netlink_callback *cb, int *s_ip_idx, 1843 struct inet_fill_args *fillargs) 1844 { 1845 struct ip_mc_list *im; 1846 int ip_idx = 0; 1847 int err; 1848 1849 for (im = rcu_dereference(in_dev->mc_list); 1850 im; 1851 im = rcu_dereference(im->next_rcu)) { 1852 if (ip_idx < *s_ip_idx) { 1853 ip_idx++; 1854 continue; 1855 } 1856 err = inet_fill_ifmcaddr(skb, in_dev->dev, im, fillargs); 1857 if (err < 0) 1858 goto done; 1859 1860 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1861 ip_idx++; 1862 } 1863 err = 0; 1864 ip_idx = 0; 1865 done: 1866 *s_ip_idx = ip_idx; 1867 return err; 1868 } 1869 1870 static int in_dev_dump_ifaddr(struct in_device *in_dev, struct sk_buff *skb, 1871 struct netlink_callback *cb, int *s_ip_idx, 1872 struct inet_fill_args *fillargs) 1873 { 1874 struct in_ifaddr *ifa; 1875 int ip_idx = 0; 1876 int err; 1877 1878 in_dev_for_each_ifa_rcu(ifa, in_dev) { 1879 if (ip_idx < *s_ip_idx) { 1880 ip_idx++; 1881 continue; 1882 } 1883 err = inet_fill_ifaddr(skb, ifa, fillargs); 1884 if (err < 0) 1885 goto done; 1886 1887 nl_dump_check_consistent(cb, nlmsg_hdr(skb)); 1888 ip_idx++; 1889 } 1890 err = 0; 1891 ip_idx = 0; 1892 done: 1893 *s_ip_idx = ip_idx; 1894 1895 return err; 1896 } 1897 1898 static int in_dev_dump_addr(struct in_device *in_dev, struct sk_buff *skb, 1899 struct netlink_callback *cb, int *s_ip_idx, 1900 struct inet_fill_args *fillargs) 1901 { 1902 switch (fillargs->event) { 1903 case RTM_NEWADDR: 1904 return in_dev_dump_ifaddr(in_dev, skb, cb, s_ip_idx, fillargs); 1905 case RTM_GETMULTICAST: 1906 return in_dev_dump_ifmcaddr(in_dev, skb, cb, s_ip_idx, 1907 fillargs); 1908 default: 1909 return -EINVAL; 1910 } 1911 } 1912 1913 /* Combine dev_addr_genid and dev_base_seq to detect changes. 1914 */ 1915 static u32 inet_base_seq(const struct net *net) 1916 { 1917 u32 res = atomic_read(&net->ipv4.dev_addr_genid) + 1918 READ_ONCE(net->dev_base_seq); 1919 1920 /* Must not return 0 (see nl_dump_check_consistent()). 1921 * Chose a value far away from 0. 1922 */ 1923 if (!res) 1924 res = 0x80000000; 1925 return res; 1926 } 1927 1928 static int inet_dump_addr(struct sk_buff *skb, struct netlink_callback *cb, 1929 int event) 1930 { 1931 const struct nlmsghdr *nlh = cb->nlh; 1932 struct inet_fill_args fillargs = { 1933 .portid = NETLINK_CB(cb->skb).portid, 1934 .seq = nlh->nlmsg_seq, 1935 .event = event, 1936 .flags = NLM_F_MULTI, 1937 .netnsid = -1, 1938 }; 1939 struct net *net = sock_net(skb->sk); 1940 struct net *tgt_net = net; 1941 struct { 1942 unsigned long ifindex; 1943 int ip_idx; 1944 } *ctx = (void *)cb->ctx; 1945 struct in_device *in_dev; 1946 struct net_device *dev; 1947 int err = 0; 1948 1949 rcu_read_lock(); 1950 if (cb->strict_check) { 1951 err = inet_valid_dump_ifaddr_req(nlh, &fillargs, &tgt_net, 1952 skb->sk, cb); 1953 if (err < 0) 1954 goto done; 1955 1956 if (fillargs.ifindex) { 1957 dev = dev_get_by_index_rcu(tgt_net, fillargs.ifindex); 1958 if (!dev) { 1959 err = -ENODEV; 1960 goto done; 1961 } 1962 in_dev = __in_dev_get_rcu(dev); 1963 if (!in_dev) 1964 goto done; 1965 err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx, 1966 &fillargs); 1967 goto done; 1968 } 1969 } 1970 1971 cb->seq = inet_base_seq(tgt_net); 1972 1973 for_each_netdev_dump(tgt_net, dev, ctx->ifindex) { 1974 in_dev = __in_dev_get_rcu(dev); 1975 if (!in_dev) 1976 continue; 1977 err = in_dev_dump_addr(in_dev, skb, cb, &ctx->ip_idx, 1978 &fillargs); 1979 if (err < 0) 1980 goto done; 1981 } 1982 done: 1983 if (fillargs.netnsid >= 0) 1984 put_net(tgt_net); 1985 rcu_read_unlock(); 1986 return err; 1987 } 1988 1989 static int inet_dump_ifaddr(struct sk_buff *skb, struct netlink_callback *cb) 1990 { 1991 return inet_dump_addr(skb, cb, RTM_NEWADDR); 1992 } 1993 1994 static int inet_dump_ifmcaddr(struct sk_buff *skb, struct netlink_callback *cb) 1995 { 1996 return inet_dump_addr(skb, cb, RTM_GETMULTICAST); 1997 } 1998 1999 static void rtmsg_ifa(int event, struct in_ifaddr *ifa, struct nlmsghdr *nlh, 2000 u32 portid) 2001 { 2002 struct inet_fill_args fillargs = { 2003 .portid = portid, 2004 .seq = nlh ? nlh->nlmsg_seq : 0, 2005 .event = event, 2006 .flags = 0, 2007 .netnsid = -1, 2008 }; 2009 struct sk_buff *skb; 2010 int err = -ENOBUFS; 2011 struct net *net; 2012 2013 net = dev_net(ifa->ifa_dev->dev); 2014 skb = nlmsg_new(inet_nlmsg_size(), GFP_KERNEL); 2015 if (!skb) 2016 goto errout; 2017 2018 err = inet_fill_ifaddr(skb, ifa, &fillargs); 2019 if (err < 0) { 2020 /* -EMSGSIZE implies BUG in inet_nlmsg_size() */ 2021 WARN_ON(err == -EMSGSIZE); 2022 kfree_skb(skb); 2023 goto errout; 2024 } 2025 rtnl_notify(skb, net, portid, RTNLGRP_IPV4_IFADDR, nlh, GFP_KERNEL); 2026 return; 2027 errout: 2028 rtnl_set_sk_err(net, RTNLGRP_IPV4_IFADDR, err); 2029 } 2030 2031 static size_t inet_get_link_af_size(const struct net_device *dev, 2032 u32 ext_filter_mask) 2033 { 2034 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr); 2035 2036 if (!in_dev) 2037 return 0; 2038 2039 return nla_total_size(IPV4_DEVCONF_MAX * 4); /* IFLA_INET_CONF */ 2040 } 2041 2042 static int inet_fill_link_af(struct sk_buff *skb, const struct net_device *dev, 2043 u32 ext_filter_mask) 2044 { 2045 struct in_device *in_dev = rcu_dereference_rtnl(dev->ip_ptr); 2046 struct nlattr *nla; 2047 int i; 2048 2049 if (!in_dev) 2050 return -ENODATA; 2051 2052 nla = nla_reserve(skb, IFLA_INET_CONF, IPV4_DEVCONF_MAX * 4); 2053 if (!nla) 2054 return -EMSGSIZE; 2055 2056 for (i = 0; i < IPV4_DEVCONF_MAX; i++) 2057 ((u32 *) nla_data(nla))[i] = READ_ONCE(in_dev->cnf.data[i]); 2058 2059 return 0; 2060 } 2061 2062 static const struct nla_policy inet_af_policy[IFLA_INET_MAX+1] = { 2063 [IFLA_INET_CONF] = { .type = NLA_NESTED }, 2064 }; 2065 2066 static int inet_validate_link_af(const struct net_device *dev, 2067 const struct nlattr *nla, 2068 struct netlink_ext_ack *extack) 2069 { 2070 struct nlattr *a, *tb[IFLA_INET_MAX+1]; 2071 int err, rem; 2072 2073 if (dev && !__in_dev_get_rtnl(dev)) 2074 return -EAFNOSUPPORT; 2075 2076 err = nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, 2077 inet_af_policy, extack); 2078 if (err < 0) 2079 return err; 2080 2081 if (tb[IFLA_INET_CONF]) { 2082 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) { 2083 int cfgid = nla_type(a); 2084 2085 if (nla_len(a) < 4) 2086 return -EINVAL; 2087 2088 if (cfgid <= 0 || cfgid > IPV4_DEVCONF_MAX) 2089 return -EINVAL; 2090 } 2091 } 2092 2093 return 0; 2094 } 2095 2096 static int inet_set_link_af(struct net_device *dev, const struct nlattr *nla, 2097 struct netlink_ext_ack *extack) 2098 { 2099 struct in_device *in_dev = __in_dev_get_rtnl(dev); 2100 struct nlattr *a, *tb[IFLA_INET_MAX+1]; 2101 int rem; 2102 2103 if (!in_dev) 2104 return -EAFNOSUPPORT; 2105 2106 if (nla_parse_nested_deprecated(tb, IFLA_INET_MAX, nla, NULL, NULL) < 0) 2107 return -EINVAL; 2108 2109 if (tb[IFLA_INET_CONF]) { 2110 nla_for_each_nested(a, tb[IFLA_INET_CONF], rem) 2111 ipv4_devconf_set(in_dev, nla_type(a), nla_get_u32(a)); 2112 } 2113 2114 return 0; 2115 } 2116 2117 static int inet_netconf_msgsize_devconf(int type) 2118 { 2119 int size = NLMSG_ALIGN(sizeof(struct netconfmsg)) 2120 + nla_total_size(4); /* NETCONFA_IFINDEX */ 2121 bool all = false; 2122 2123 if (type == NETCONFA_ALL) 2124 all = true; 2125 2126 if (all || type == NETCONFA_FORWARDING) 2127 size += nla_total_size(4); 2128 if (all || type == NETCONFA_RP_FILTER) 2129 size += nla_total_size(4); 2130 if (all || type == NETCONFA_MC_FORWARDING) 2131 size += nla_total_size(4); 2132 if (all || type == NETCONFA_BC_FORWARDING) 2133 size += nla_total_size(4); 2134 if (all || type == NETCONFA_PROXY_NEIGH) 2135 size += nla_total_size(4); 2136 if (all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) 2137 size += nla_total_size(4); 2138 2139 return size; 2140 } 2141 2142 static int inet_netconf_fill_devconf(struct sk_buff *skb, int ifindex, 2143 const struct ipv4_devconf *devconf, 2144 u32 portid, u32 seq, int event, 2145 unsigned int flags, int type) 2146 { 2147 struct nlmsghdr *nlh; 2148 struct netconfmsg *ncm; 2149 bool all = false; 2150 2151 nlh = nlmsg_put(skb, portid, seq, event, sizeof(struct netconfmsg), 2152 flags); 2153 if (!nlh) 2154 return -EMSGSIZE; 2155 2156 if (type == NETCONFA_ALL) 2157 all = true; 2158 2159 ncm = nlmsg_data(nlh); 2160 ncm->ncm_family = AF_INET; 2161 2162 if (nla_put_s32(skb, NETCONFA_IFINDEX, ifindex) < 0) 2163 goto nla_put_failure; 2164 2165 if (!devconf) 2166 goto out; 2167 2168 if ((all || type == NETCONFA_FORWARDING) && 2169 nla_put_s32(skb, NETCONFA_FORWARDING, 2170 IPV4_DEVCONF_RO(*devconf, FORWARDING)) < 0) 2171 goto nla_put_failure; 2172 if ((all || type == NETCONFA_RP_FILTER) && 2173 nla_put_s32(skb, NETCONFA_RP_FILTER, 2174 IPV4_DEVCONF_RO(*devconf, RP_FILTER)) < 0) 2175 goto nla_put_failure; 2176 if ((all || type == NETCONFA_MC_FORWARDING) && 2177 nla_put_s32(skb, NETCONFA_MC_FORWARDING, 2178 IPV4_DEVCONF_RO(*devconf, MC_FORWARDING)) < 0) 2179 goto nla_put_failure; 2180 if ((all || type == NETCONFA_BC_FORWARDING) && 2181 nla_put_s32(skb, NETCONFA_BC_FORWARDING, 2182 IPV4_DEVCONF_RO(*devconf, BC_FORWARDING)) < 0) 2183 goto nla_put_failure; 2184 if ((all || type == NETCONFA_PROXY_NEIGH) && 2185 nla_put_s32(skb, NETCONFA_PROXY_NEIGH, 2186 IPV4_DEVCONF_RO(*devconf, PROXY_ARP)) < 0) 2187 goto nla_put_failure; 2188 if ((all || type == NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN) && 2189 nla_put_s32(skb, NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 2190 IPV4_DEVCONF_RO(*devconf, 2191 IGNORE_ROUTES_WITH_LINKDOWN)) < 0) 2192 goto nla_put_failure; 2193 2194 out: 2195 nlmsg_end(skb, nlh); 2196 return 0; 2197 2198 nla_put_failure: 2199 nlmsg_cancel(skb, nlh); 2200 return -EMSGSIZE; 2201 } 2202 2203 void inet_netconf_notify_devconf(struct net *net, int event, int type, 2204 int ifindex, struct ipv4_devconf *devconf) 2205 { 2206 struct sk_buff *skb; 2207 int err = -ENOBUFS; 2208 2209 skb = nlmsg_new(inet_netconf_msgsize_devconf(type), GFP_KERNEL); 2210 if (!skb) 2211 goto errout; 2212 2213 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 0, 0, 2214 event, 0, type); 2215 if (err < 0) { 2216 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */ 2217 WARN_ON(err == -EMSGSIZE); 2218 kfree_skb(skb); 2219 goto errout; 2220 } 2221 rtnl_notify(skb, net, 0, RTNLGRP_IPV4_NETCONF, NULL, GFP_KERNEL); 2222 return; 2223 errout: 2224 rtnl_set_sk_err(net, RTNLGRP_IPV4_NETCONF, err); 2225 } 2226 2227 static const struct nla_policy devconf_ipv4_policy[NETCONFA_MAX+1] = { 2228 [NETCONFA_IFINDEX] = { .len = sizeof(int) }, 2229 [NETCONFA_FORWARDING] = { .len = sizeof(int) }, 2230 [NETCONFA_RP_FILTER] = { .len = sizeof(int) }, 2231 [NETCONFA_PROXY_NEIGH] = { .len = sizeof(int) }, 2232 [NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN] = { .len = sizeof(int) }, 2233 }; 2234 2235 static int inet_netconf_valid_get_req(struct sk_buff *skb, 2236 const struct nlmsghdr *nlh, 2237 struct nlattr **tb, 2238 struct netlink_ext_ack *extack) 2239 { 2240 int i, err; 2241 2242 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(struct netconfmsg))) { 2243 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf get request"); 2244 return -EINVAL; 2245 } 2246 2247 if (!netlink_strict_get_check(skb)) 2248 return nlmsg_parse_deprecated(nlh, sizeof(struct netconfmsg), 2249 tb, NETCONFA_MAX, 2250 devconf_ipv4_policy, extack); 2251 2252 err = nlmsg_parse_deprecated_strict(nlh, sizeof(struct netconfmsg), 2253 tb, NETCONFA_MAX, 2254 devconf_ipv4_policy, extack); 2255 if (err) 2256 return err; 2257 2258 for (i = 0; i <= NETCONFA_MAX; i++) { 2259 if (!tb[i]) 2260 continue; 2261 2262 switch (i) { 2263 case NETCONFA_IFINDEX: 2264 break; 2265 default: 2266 NL_SET_ERR_MSG(extack, "ipv4: Unsupported attribute in netconf get request"); 2267 return -EINVAL; 2268 } 2269 } 2270 2271 return 0; 2272 } 2273 2274 static int inet_netconf_get_devconf(struct sk_buff *in_skb, 2275 struct nlmsghdr *nlh, 2276 struct netlink_ext_ack *extack) 2277 { 2278 struct net *net = sock_net(in_skb->sk); 2279 struct nlattr *tb[NETCONFA_MAX + 1]; 2280 const struct ipv4_devconf *devconf; 2281 struct in_device *in_dev = NULL; 2282 struct net_device *dev = NULL; 2283 struct sk_buff *skb; 2284 int ifindex; 2285 int err; 2286 2287 err = inet_netconf_valid_get_req(in_skb, nlh, tb, extack); 2288 if (err) 2289 return err; 2290 2291 if (!tb[NETCONFA_IFINDEX]) 2292 return -EINVAL; 2293 2294 ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]); 2295 switch (ifindex) { 2296 case NETCONFA_IFINDEX_ALL: 2297 devconf = net->ipv4.devconf_all; 2298 break; 2299 case NETCONFA_IFINDEX_DEFAULT: 2300 devconf = net->ipv4.devconf_dflt; 2301 break; 2302 default: 2303 err = -ENODEV; 2304 dev = dev_get_by_index(net, ifindex); 2305 if (dev) 2306 in_dev = in_dev_get(dev); 2307 if (!in_dev) 2308 goto errout; 2309 devconf = &in_dev->cnf; 2310 break; 2311 } 2312 2313 err = -ENOBUFS; 2314 skb = nlmsg_new(inet_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL); 2315 if (!skb) 2316 goto errout; 2317 2318 err = inet_netconf_fill_devconf(skb, ifindex, devconf, 2319 NETLINK_CB(in_skb).portid, 2320 nlh->nlmsg_seq, RTM_NEWNETCONF, 0, 2321 NETCONFA_ALL); 2322 if (err < 0) { 2323 /* -EMSGSIZE implies BUG in inet_netconf_msgsize_devconf() */ 2324 WARN_ON(err == -EMSGSIZE); 2325 kfree_skb(skb); 2326 goto errout; 2327 } 2328 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 2329 errout: 2330 if (in_dev) 2331 in_dev_put(in_dev); 2332 dev_put(dev); 2333 return err; 2334 } 2335 2336 static int inet_netconf_dump_devconf(struct sk_buff *skb, 2337 struct netlink_callback *cb) 2338 { 2339 const struct nlmsghdr *nlh = cb->nlh; 2340 struct net *net = sock_net(skb->sk); 2341 struct { 2342 unsigned long ifindex; 2343 unsigned int all_default; 2344 } *ctx = (void *)cb->ctx; 2345 const struct in_device *in_dev; 2346 struct net_device *dev; 2347 int err = 0; 2348 2349 if (cb->strict_check) { 2350 struct netlink_ext_ack *extack = cb->extack; 2351 struct netconfmsg *ncm; 2352 2353 if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*ncm))) { 2354 NL_SET_ERR_MSG(extack, "ipv4: Invalid header for netconf dump request"); 2355 return -EINVAL; 2356 } 2357 2358 if (nlmsg_attrlen(nlh, sizeof(*ncm))) { 2359 NL_SET_ERR_MSG(extack, "ipv4: Invalid data after header in netconf dump request"); 2360 return -EINVAL; 2361 } 2362 } 2363 2364 rcu_read_lock(); 2365 for_each_netdev_dump(net, dev, ctx->ifindex) { 2366 in_dev = __in_dev_get_rcu(dev); 2367 if (!in_dev) 2368 continue; 2369 err = inet_netconf_fill_devconf(skb, dev->ifindex, 2370 &in_dev->cnf, 2371 NETLINK_CB(cb->skb).portid, 2372 nlh->nlmsg_seq, 2373 RTM_NEWNETCONF, NLM_F_MULTI, 2374 NETCONFA_ALL); 2375 if (err < 0) 2376 goto done; 2377 } 2378 if (ctx->all_default == 0) { 2379 err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_ALL, 2380 net->ipv4.devconf_all, 2381 NETLINK_CB(cb->skb).portid, 2382 nlh->nlmsg_seq, 2383 RTM_NEWNETCONF, NLM_F_MULTI, 2384 NETCONFA_ALL); 2385 if (err < 0) 2386 goto done; 2387 ctx->all_default++; 2388 } 2389 if (ctx->all_default == 1) { 2390 err = inet_netconf_fill_devconf(skb, NETCONFA_IFINDEX_DEFAULT, 2391 net->ipv4.devconf_dflt, 2392 NETLINK_CB(cb->skb).portid, 2393 nlh->nlmsg_seq, 2394 RTM_NEWNETCONF, NLM_F_MULTI, 2395 NETCONFA_ALL); 2396 if (err < 0) 2397 goto done; 2398 ctx->all_default++; 2399 } 2400 done: 2401 rcu_read_unlock(); 2402 return err; 2403 } 2404 2405 #ifdef CONFIG_SYSCTL 2406 2407 static void devinet_copy_dflt_conf(struct net *net, int i) 2408 { 2409 struct net_device *dev; 2410 2411 rcu_read_lock(); 2412 for_each_netdev_rcu(net, dev) { 2413 struct in_device *in_dev; 2414 2415 in_dev = __in_dev_get_rcu(dev); 2416 if (in_dev && !test_bit(i, in_dev->cnf.state)) 2417 in_dev->cnf.data[i] = net->ipv4.devconf_dflt->data[i]; 2418 } 2419 rcu_read_unlock(); 2420 } 2421 2422 /* called with RTNL locked */ 2423 static void inet_forward_change(struct net *net) 2424 { 2425 struct net_device *dev; 2426 int on = IPV4_DEVCONF_ALL(net, FORWARDING); 2427 2428 IPV4_DEVCONF_ALL(net, ACCEPT_REDIRECTS) = !on; 2429 IPV4_DEVCONF_DFLT(net, FORWARDING) = on; 2430 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2431 NETCONFA_FORWARDING, 2432 NETCONFA_IFINDEX_ALL, 2433 net->ipv4.devconf_all); 2434 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2435 NETCONFA_FORWARDING, 2436 NETCONFA_IFINDEX_DEFAULT, 2437 net->ipv4.devconf_dflt); 2438 2439 for_each_netdev(net, dev) { 2440 struct in_device *in_dev; 2441 2442 if (on) 2443 dev_disable_lro(dev); 2444 2445 in_dev = __in_dev_get_rtnl_net(dev); 2446 if (in_dev) { 2447 IN_DEV_CONF_SET(in_dev, FORWARDING, on); 2448 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2449 NETCONFA_FORWARDING, 2450 dev->ifindex, &in_dev->cnf); 2451 } 2452 } 2453 } 2454 2455 static int devinet_conf_ifindex(struct net *net, struct ipv4_devconf *cnf) 2456 { 2457 if (cnf == net->ipv4.devconf_dflt) 2458 return NETCONFA_IFINDEX_DEFAULT; 2459 else if (cnf == net->ipv4.devconf_all) 2460 return NETCONFA_IFINDEX_ALL; 2461 else { 2462 struct in_device *idev 2463 = container_of(cnf, struct in_device, cnf); 2464 return idev->dev->ifindex; 2465 } 2466 } 2467 2468 static int devinet_conf_proc(const struct ctl_table *ctl, int write, 2469 void *buffer, size_t *lenp, loff_t *ppos) 2470 { 2471 int old_value = *(int *)ctl->data; 2472 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 2473 int new_value = *(int *)ctl->data; 2474 2475 if (write) { 2476 struct ipv4_devconf *cnf = ctl->extra1; 2477 struct net *net = ctl->extra2; 2478 int i = (int *)ctl->data - cnf->data; 2479 int ifindex; 2480 2481 set_bit(i, cnf->state); 2482 2483 if (cnf == net->ipv4.devconf_dflt) 2484 devinet_copy_dflt_conf(net, i); 2485 if (i == IPV4_DEVCONF_ACCEPT_LOCAL - 1 || 2486 i == IPV4_DEVCONF_ROUTE_LOCALNET - 1) 2487 if ((new_value == 0) && (old_value != 0)) 2488 rt_cache_flush(net); 2489 2490 if (i == IPV4_DEVCONF_BC_FORWARDING - 1 && 2491 new_value != old_value) 2492 rt_cache_flush(net); 2493 2494 if (i == IPV4_DEVCONF_RP_FILTER - 1 && 2495 new_value != old_value) { 2496 ifindex = devinet_conf_ifindex(net, cnf); 2497 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2498 NETCONFA_RP_FILTER, 2499 ifindex, cnf); 2500 } 2501 if (i == IPV4_DEVCONF_PROXY_ARP - 1 && 2502 new_value != old_value) { 2503 ifindex = devinet_conf_ifindex(net, cnf); 2504 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2505 NETCONFA_PROXY_NEIGH, 2506 ifindex, cnf); 2507 } 2508 if (i == IPV4_DEVCONF_IGNORE_ROUTES_WITH_LINKDOWN - 1 && 2509 new_value != old_value) { 2510 ifindex = devinet_conf_ifindex(net, cnf); 2511 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2512 NETCONFA_IGNORE_ROUTES_WITH_LINKDOWN, 2513 ifindex, cnf); 2514 } 2515 } 2516 2517 return ret; 2518 } 2519 2520 static int devinet_sysctl_forward(const struct ctl_table *ctl, int write, 2521 void *buffer, size_t *lenp, loff_t *ppos) 2522 { 2523 int *valp = ctl->data; 2524 int val = *valp; 2525 loff_t pos = *ppos; 2526 struct net *net = ctl->extra2; 2527 int ret; 2528 2529 if (write && !ns_capable(net->user_ns, CAP_NET_ADMIN)) 2530 return -EPERM; 2531 2532 ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 2533 2534 if (write && *valp != val) { 2535 if (valp != &IPV4_DEVCONF_DFLT(net, FORWARDING)) { 2536 if (!rtnl_net_trylock(net)) { 2537 /* Restore the original values before restarting */ 2538 *valp = val; 2539 *ppos = pos; 2540 return restart_syscall(); 2541 } 2542 if (valp == &IPV4_DEVCONF_ALL(net, FORWARDING)) { 2543 inet_forward_change(net); 2544 } else { 2545 struct ipv4_devconf *cnf = ctl->extra1; 2546 struct in_device *idev = 2547 container_of(cnf, struct in_device, cnf); 2548 if (*valp) 2549 dev_disable_lro(idev->dev); 2550 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2551 NETCONFA_FORWARDING, 2552 idev->dev->ifindex, 2553 cnf); 2554 } 2555 rtnl_net_unlock(net); 2556 rt_cache_flush(net); 2557 } else 2558 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, 2559 NETCONFA_FORWARDING, 2560 NETCONFA_IFINDEX_DEFAULT, 2561 net->ipv4.devconf_dflt); 2562 } 2563 2564 return ret; 2565 } 2566 2567 static int ipv4_doint_and_flush(const struct ctl_table *ctl, int write, 2568 void *buffer, size_t *lenp, loff_t *ppos) 2569 { 2570 int *valp = ctl->data; 2571 int val = *valp; 2572 int ret = proc_dointvec(ctl, write, buffer, lenp, ppos); 2573 struct net *net = ctl->extra2; 2574 2575 if (write && *valp != val) 2576 rt_cache_flush(net); 2577 2578 return ret; 2579 } 2580 2581 #define DEVINET_SYSCTL_ENTRY(attr, name, mval, proc) \ 2582 { \ 2583 .procname = name, \ 2584 .data = ipv4_devconf.data + \ 2585 IPV4_DEVCONF_ ## attr - 1, \ 2586 .maxlen = sizeof(int), \ 2587 .mode = mval, \ 2588 .proc_handler = proc, \ 2589 .extra1 = &ipv4_devconf, \ 2590 } 2591 2592 #define DEVINET_SYSCTL_RW_ENTRY(attr, name) \ 2593 DEVINET_SYSCTL_ENTRY(attr, name, 0644, devinet_conf_proc) 2594 2595 #define DEVINET_SYSCTL_RO_ENTRY(attr, name) \ 2596 DEVINET_SYSCTL_ENTRY(attr, name, 0444, devinet_conf_proc) 2597 2598 #define DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, proc) \ 2599 DEVINET_SYSCTL_ENTRY(attr, name, 0644, proc) 2600 2601 #define DEVINET_SYSCTL_FLUSHING_ENTRY(attr, name) \ 2602 DEVINET_SYSCTL_COMPLEX_ENTRY(attr, name, ipv4_doint_and_flush) 2603 2604 static struct devinet_sysctl_table { 2605 struct ctl_table_header *sysctl_header; 2606 struct ctl_table devinet_vars[IPV4_DEVCONF_MAX]; 2607 } devinet_sysctl = { 2608 .devinet_vars = { 2609 DEVINET_SYSCTL_COMPLEX_ENTRY(FORWARDING, "forwarding", 2610 devinet_sysctl_forward), 2611 DEVINET_SYSCTL_RO_ENTRY(MC_FORWARDING, "mc_forwarding"), 2612 DEVINET_SYSCTL_RW_ENTRY(BC_FORWARDING, "bc_forwarding"), 2613 2614 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_REDIRECTS, "accept_redirects"), 2615 DEVINET_SYSCTL_RW_ENTRY(SECURE_REDIRECTS, "secure_redirects"), 2616 DEVINET_SYSCTL_RW_ENTRY(SHARED_MEDIA, "shared_media"), 2617 DEVINET_SYSCTL_RW_ENTRY(RP_FILTER, "rp_filter"), 2618 DEVINET_SYSCTL_RW_ENTRY(SEND_REDIRECTS, "send_redirects"), 2619 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_SOURCE_ROUTE, 2620 "accept_source_route"), 2621 DEVINET_SYSCTL_RW_ENTRY(ACCEPT_LOCAL, "accept_local"), 2622 DEVINET_SYSCTL_RW_ENTRY(SRC_VMARK, "src_valid_mark"), 2623 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP, "proxy_arp"), 2624 DEVINET_SYSCTL_RW_ENTRY(MEDIUM_ID, "medium_id"), 2625 DEVINET_SYSCTL_RW_ENTRY(BOOTP_RELAY, "bootp_relay"), 2626 DEVINET_SYSCTL_RW_ENTRY(LOG_MARTIANS, "log_martians"), 2627 DEVINET_SYSCTL_RW_ENTRY(TAG, "tag"), 2628 DEVINET_SYSCTL_RW_ENTRY(ARPFILTER, "arp_filter"), 2629 DEVINET_SYSCTL_RW_ENTRY(ARP_ANNOUNCE, "arp_announce"), 2630 DEVINET_SYSCTL_RW_ENTRY(ARP_IGNORE, "arp_ignore"), 2631 DEVINET_SYSCTL_RW_ENTRY(ARP_ACCEPT, "arp_accept"), 2632 DEVINET_SYSCTL_RW_ENTRY(ARP_NOTIFY, "arp_notify"), 2633 DEVINET_SYSCTL_RW_ENTRY(ARP_EVICT_NOCARRIER, 2634 "arp_evict_nocarrier"), 2635 DEVINET_SYSCTL_RW_ENTRY(PROXY_ARP_PVLAN, "proxy_arp_pvlan"), 2636 DEVINET_SYSCTL_RW_ENTRY(FORCE_IGMP_VERSION, 2637 "force_igmp_version"), 2638 DEVINET_SYSCTL_RW_ENTRY(IGMPV2_UNSOLICITED_REPORT_INTERVAL, 2639 "igmpv2_unsolicited_report_interval"), 2640 DEVINET_SYSCTL_RW_ENTRY(IGMPV3_UNSOLICITED_REPORT_INTERVAL, 2641 "igmpv3_unsolicited_report_interval"), 2642 DEVINET_SYSCTL_RW_ENTRY(IGNORE_ROUTES_WITH_LINKDOWN, 2643 "ignore_routes_with_linkdown"), 2644 DEVINET_SYSCTL_RW_ENTRY(DROP_GRATUITOUS_ARP, 2645 "drop_gratuitous_arp"), 2646 2647 DEVINET_SYSCTL_FLUSHING_ENTRY(NOXFRM, "disable_xfrm"), 2648 DEVINET_SYSCTL_FLUSHING_ENTRY(NOPOLICY, "disable_policy"), 2649 DEVINET_SYSCTL_FLUSHING_ENTRY(PROMOTE_SECONDARIES, 2650 "promote_secondaries"), 2651 DEVINET_SYSCTL_FLUSHING_ENTRY(ROUTE_LOCALNET, 2652 "route_localnet"), 2653 DEVINET_SYSCTL_FLUSHING_ENTRY(DROP_UNICAST_IN_L2_MULTICAST, 2654 "drop_unicast_in_l2_multicast"), 2655 }, 2656 }; 2657 2658 static int __devinet_sysctl_register(struct net *net, char *dev_name, 2659 int ifindex, struct ipv4_devconf *p) 2660 { 2661 int i; 2662 struct devinet_sysctl_table *t; 2663 char path[sizeof("net/ipv4/conf/") + IFNAMSIZ]; 2664 2665 t = kmemdup(&devinet_sysctl, sizeof(*t), GFP_KERNEL_ACCOUNT); 2666 if (!t) 2667 goto out; 2668 2669 for (i = 0; i < ARRAY_SIZE(t->devinet_vars); i++) { 2670 t->devinet_vars[i].data += (char *)p - (char *)&ipv4_devconf; 2671 t->devinet_vars[i].extra1 = p; 2672 t->devinet_vars[i].extra2 = net; 2673 } 2674 2675 snprintf(path, sizeof(path), "net/ipv4/conf/%s", dev_name); 2676 2677 t->sysctl_header = register_net_sysctl(net, path, t->devinet_vars); 2678 if (!t->sysctl_header) 2679 goto free; 2680 2681 p->sysctl = t; 2682 2683 inet_netconf_notify_devconf(net, RTM_NEWNETCONF, NETCONFA_ALL, 2684 ifindex, p); 2685 return 0; 2686 2687 free: 2688 kfree(t); 2689 out: 2690 return -ENOMEM; 2691 } 2692 2693 static void __devinet_sysctl_unregister(struct net *net, 2694 struct ipv4_devconf *cnf, int ifindex) 2695 { 2696 struct devinet_sysctl_table *t = cnf->sysctl; 2697 2698 if (t) { 2699 cnf->sysctl = NULL; 2700 unregister_net_sysctl_table(t->sysctl_header); 2701 kfree(t); 2702 } 2703 2704 inet_netconf_notify_devconf(net, RTM_DELNETCONF, 0, ifindex, NULL); 2705 } 2706 2707 static int devinet_sysctl_register(struct in_device *idev) 2708 { 2709 int err; 2710 2711 if (!sysctl_dev_name_is_allowed(idev->dev->name)) 2712 return -EINVAL; 2713 2714 err = neigh_sysctl_register(idev->dev, idev->arp_parms, NULL); 2715 if (err) 2716 return err; 2717 err = __devinet_sysctl_register(dev_net(idev->dev), idev->dev->name, 2718 idev->dev->ifindex, &idev->cnf); 2719 if (err) 2720 neigh_sysctl_unregister(idev->arp_parms); 2721 return err; 2722 } 2723 2724 static void devinet_sysctl_unregister(struct in_device *idev) 2725 { 2726 struct net *net = dev_net(idev->dev); 2727 2728 __devinet_sysctl_unregister(net, &idev->cnf, idev->dev->ifindex); 2729 neigh_sysctl_unregister(idev->arp_parms); 2730 } 2731 2732 static struct ctl_table ctl_forward_entry[] = { 2733 { 2734 .procname = "ip_forward", 2735 .data = &ipv4_devconf.data[ 2736 IPV4_DEVCONF_FORWARDING - 1], 2737 .maxlen = sizeof(int), 2738 .mode = 0644, 2739 .proc_handler = devinet_sysctl_forward, 2740 .extra1 = &ipv4_devconf, 2741 .extra2 = &init_net, 2742 }, 2743 }; 2744 #endif 2745 2746 static __net_init int devinet_init_net(struct net *net) 2747 { 2748 #ifdef CONFIG_SYSCTL 2749 struct ctl_table_header *forw_hdr; 2750 struct ctl_table *tbl; 2751 #endif 2752 struct ipv4_devconf *all, *dflt; 2753 int err; 2754 int i; 2755 2756 err = -ENOMEM; 2757 net->ipv4.inet_addr_lst = kmalloc_array(IN4_ADDR_HSIZE, 2758 sizeof(struct hlist_head), 2759 GFP_KERNEL); 2760 if (!net->ipv4.inet_addr_lst) 2761 goto err_alloc_hash; 2762 2763 all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL); 2764 if (!all) 2765 goto err_alloc_all; 2766 2767 dflt = kmemdup(&ipv4_devconf_dflt, sizeof(ipv4_devconf_dflt), GFP_KERNEL); 2768 if (!dflt) 2769 goto err_alloc_dflt; 2770 2771 #ifdef CONFIG_SYSCTL 2772 tbl = kmemdup(ctl_forward_entry, sizeof(ctl_forward_entry), GFP_KERNEL); 2773 if (!tbl) 2774 goto err_alloc_ctl; 2775 2776 tbl[0].data = &all->data[IPV4_DEVCONF_FORWARDING - 1]; 2777 tbl[0].extra1 = all; 2778 tbl[0].extra2 = net; 2779 #endif 2780 2781 if (!net_eq(net, &init_net)) { 2782 switch (net_inherit_devconf()) { 2783 case 3: 2784 /* copy from the current netns */ 2785 memcpy(all, current->nsproxy->net_ns->ipv4.devconf_all, 2786 sizeof(ipv4_devconf)); 2787 memcpy(dflt, 2788 current->nsproxy->net_ns->ipv4.devconf_dflt, 2789 sizeof(ipv4_devconf_dflt)); 2790 break; 2791 case 0: 2792 case 1: 2793 /* copy from init_net */ 2794 memcpy(all, init_net.ipv4.devconf_all, 2795 sizeof(ipv4_devconf)); 2796 memcpy(dflt, init_net.ipv4.devconf_dflt, 2797 sizeof(ipv4_devconf_dflt)); 2798 break; 2799 case 2: 2800 /* use compiled values */ 2801 break; 2802 } 2803 } 2804 2805 #ifdef CONFIG_SYSCTL 2806 err = __devinet_sysctl_register(net, "all", NETCONFA_IFINDEX_ALL, all); 2807 if (err < 0) 2808 goto err_reg_all; 2809 2810 err = __devinet_sysctl_register(net, "default", 2811 NETCONFA_IFINDEX_DEFAULT, dflt); 2812 if (err < 0) 2813 goto err_reg_dflt; 2814 2815 err = -ENOMEM; 2816 forw_hdr = register_net_sysctl_sz(net, "net/ipv4", tbl, 2817 ARRAY_SIZE(ctl_forward_entry)); 2818 if (!forw_hdr) 2819 goto err_reg_ctl; 2820 net->ipv4.forw_hdr = forw_hdr; 2821 #endif 2822 2823 for (i = 0; i < IN4_ADDR_HSIZE; i++) 2824 INIT_HLIST_HEAD(&net->ipv4.inet_addr_lst[i]); 2825 2826 INIT_DEFERRABLE_WORK(&net->ipv4.addr_chk_work, check_lifetime); 2827 2828 net->ipv4.devconf_all = all; 2829 net->ipv4.devconf_dflt = dflt; 2830 return 0; 2831 2832 #ifdef CONFIG_SYSCTL 2833 err_reg_ctl: 2834 __devinet_sysctl_unregister(net, dflt, NETCONFA_IFINDEX_DEFAULT); 2835 err_reg_dflt: 2836 __devinet_sysctl_unregister(net, all, NETCONFA_IFINDEX_ALL); 2837 err_reg_all: 2838 kfree(tbl); 2839 err_alloc_ctl: 2840 #endif 2841 kfree(dflt); 2842 err_alloc_dflt: 2843 kfree(all); 2844 err_alloc_all: 2845 kfree(net->ipv4.inet_addr_lst); 2846 err_alloc_hash: 2847 return err; 2848 } 2849 2850 static __net_exit void devinet_exit_net(struct net *net) 2851 { 2852 #ifdef CONFIG_SYSCTL 2853 const struct ctl_table *tbl; 2854 #endif 2855 2856 cancel_delayed_work_sync(&net->ipv4.addr_chk_work); 2857 2858 #ifdef CONFIG_SYSCTL 2859 tbl = net->ipv4.forw_hdr->ctl_table_arg; 2860 unregister_net_sysctl_table(net->ipv4.forw_hdr); 2861 __devinet_sysctl_unregister(net, net->ipv4.devconf_dflt, 2862 NETCONFA_IFINDEX_DEFAULT); 2863 __devinet_sysctl_unregister(net, net->ipv4.devconf_all, 2864 NETCONFA_IFINDEX_ALL); 2865 kfree(tbl); 2866 #endif 2867 kfree(net->ipv4.devconf_dflt); 2868 kfree(net->ipv4.devconf_all); 2869 kfree(net->ipv4.inet_addr_lst); 2870 } 2871 2872 static __net_initdata struct pernet_operations devinet_ops = { 2873 .init = devinet_init_net, 2874 .exit = devinet_exit_net, 2875 }; 2876 2877 static struct rtnl_af_ops inet_af_ops __read_mostly = { 2878 .family = AF_INET, 2879 .fill_link_af = inet_fill_link_af, 2880 .get_link_af_size = inet_get_link_af_size, 2881 .validate_link_af = inet_validate_link_af, 2882 .set_link_af = inet_set_link_af, 2883 }; 2884 2885 static const struct rtnl_msg_handler devinet_rtnl_msg_handlers[] __initconst = { 2886 {.protocol = PF_INET, .msgtype = RTM_NEWADDR, .doit = inet_rtm_newaddr, 2887 .flags = RTNL_FLAG_DOIT_PERNET}, 2888 {.protocol = PF_INET, .msgtype = RTM_DELADDR, .doit = inet_rtm_deladdr, 2889 .flags = RTNL_FLAG_DOIT_PERNET}, 2890 {.protocol = PF_INET, .msgtype = RTM_GETADDR, .dumpit = inet_dump_ifaddr, 2891 .flags = RTNL_FLAG_DUMP_UNLOCKED | RTNL_FLAG_DUMP_SPLIT_NLM_DONE}, 2892 {.protocol = PF_INET, .msgtype = RTM_GETNETCONF, 2893 .doit = inet_netconf_get_devconf, .dumpit = inet_netconf_dump_devconf, 2894 .flags = RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED}, 2895 {.owner = THIS_MODULE, .protocol = PF_INET, .msgtype = RTM_GETMULTICAST, 2896 .dumpit = inet_dump_ifmcaddr, .flags = RTNL_FLAG_DUMP_UNLOCKED}, 2897 }; 2898 2899 void __init devinet_init(void) 2900 { 2901 register_pernet_subsys(&devinet_ops); 2902 register_netdevice_notifier(&ip_netdev_notifier); 2903 2904 if (rtnl_af_register(&inet_af_ops)) 2905 panic("Unable to register inet_af_ops\n"); 2906 2907 rtnl_register_many(devinet_rtnl_msg_handlers); 2908 } 2909