1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * IPv6 Address Label subsystem 4 * for the IPv6 "Default" Source Address Selection 5 * 6 * Copyright (C)2007 USAGI/WIDE Project 7 */ 8 /* 9 * Author: 10 * YOSHIFUJI Hideaki @ USAGI/WIDE Project <yoshfuji@linux-ipv6.org> 11 */ 12 13 #include <linux/kernel.h> 14 #include <linux/list.h> 15 #include <linux/rcupdate.h> 16 #include <linux/in6.h> 17 #include <linux/slab.h> 18 #include <net/addrconf.h> 19 #include <linux/if_addrlabel.h> 20 #include <linux/netlink.h> 21 #include <linux/rtnetlink.h> 22 #include <linux/refcount.h> 23 24 #if 0 25 #define ADDRLABEL(x...) printk(x) 26 #else 27 #define ADDRLABEL(x...) do { ; } while (0) 28 #endif 29 30 /* 31 * Policy Table 32 */ 33 struct ip6addrlbl_entry { 34 possible_net_t lbl_net; 35 struct in6_addr prefix; 36 int prefixlen; 37 int ifindex; 38 int addrtype; 39 u32 label; 40 struct hlist_node list; 41 refcount_t refcnt; 42 struct rcu_head rcu; 43 }; 44 45 static struct ip6addrlbl_table 46 { 47 struct hlist_head head; 48 spinlock_t lock; 49 u32 seq; 50 } ip6addrlbl_table; 51 52 static inline 53 struct net *ip6addrlbl_net(const struct ip6addrlbl_entry *lbl) 54 { 55 return read_pnet(&lbl->lbl_net); 56 } 57 58 /* 59 * Default policy table (RFC6724 + extensions) 60 * 61 * prefix addr_type label 62 * ------------------------------------------------------------------------- 63 * ::1/128 LOOPBACK 0 64 * ::/0 N/A 1 65 * 2002::/16 N/A 2 66 * ::/96 COMPATv4 3 67 * ::ffff:0:0/96 V4MAPPED 4 68 * fc00::/7 N/A 5 ULA (RFC 4193) 69 * 2001::/32 N/A 6 Teredo (RFC 4380) 70 * 2001:10::/28 N/A 7 ORCHID (RFC 4843) 71 * fec0::/10 N/A 11 Site-local 72 * (deprecated by RFC3879) 73 * 3ffe::/16 N/A 12 6bone 74 * 75 * Note: 0xffffffff is used if we do not have any policies. 76 * Note: Labels for ULA and 6to4 are different from labels listed in RFC6724. 77 */ 78 79 #define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL 80 81 static const __net_initconst struct ip6addrlbl_init_table 82 { 83 const struct in6_addr *prefix; 84 int prefixlen; 85 u32 label; 86 } ip6addrlbl_init_table[] = { 87 { /* ::/0 */ 88 .prefix = &in6addr_any, 89 .label = 1, 90 }, { /* fc00::/7 */ 91 .prefix = &(struct in6_addr){ { { 0xfc } } } , 92 .prefixlen = 7, 93 .label = 5, 94 }, { /* fec0::/10 */ 95 .prefix = &(struct in6_addr){ { { 0xfe, 0xc0 } } }, 96 .prefixlen = 10, 97 .label = 11, 98 }, { /* 2002::/16 */ 99 .prefix = &(struct in6_addr){ { { 0x20, 0x02 } } }, 100 .prefixlen = 16, 101 .label = 2, 102 }, { /* 3ffe::/16 */ 103 .prefix = &(struct in6_addr){ { { 0x3f, 0xfe } } }, 104 .prefixlen = 16, 105 .label = 12, 106 }, { /* 2001::/32 */ 107 .prefix = &(struct in6_addr){ { { 0x20, 0x01 } } }, 108 .prefixlen = 32, 109 .label = 6, 110 }, { /* 2001:10::/28 */ 111 .prefix = &(struct in6_addr){ { { 0x20, 0x01, 0x00, 0x10 } } }, 112 .prefixlen = 28, 113 .label = 7, 114 }, { /* ::ffff:0:0 */ 115 .prefix = &(struct in6_addr){ { { [10] = 0xff, [11] = 0xff } } }, 116 .prefixlen = 96, 117 .label = 4, 118 }, { /* ::/96 */ 119 .prefix = &in6addr_any, 120 .prefixlen = 96, 121 .label = 3, 122 }, { /* ::1/128 */ 123 .prefix = &in6addr_loopback, 124 .prefixlen = 128, 125 .label = 0, 126 } 127 }; 128 129 /* Object management */ 130 static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p) 131 { 132 kfree(p); 133 } 134 135 static void ip6addrlbl_free_rcu(struct rcu_head *h) 136 { 137 ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu)); 138 } 139 140 static bool ip6addrlbl_hold(struct ip6addrlbl_entry *p) 141 { 142 return refcount_inc_not_zero(&p->refcnt); 143 } 144 145 static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p) 146 { 147 if (refcount_dec_and_test(&p->refcnt)) 148 call_rcu(&p->rcu, ip6addrlbl_free_rcu); 149 } 150 151 /* Find label */ 152 static bool __ip6addrlbl_match(struct net *net, 153 const struct ip6addrlbl_entry *p, 154 const struct in6_addr *addr, 155 int addrtype, int ifindex) 156 { 157 if (!net_eq(ip6addrlbl_net(p), net)) 158 return false; 159 if (p->ifindex && p->ifindex != ifindex) 160 return false; 161 if (p->addrtype && p->addrtype != addrtype) 162 return false; 163 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen)) 164 return false; 165 return true; 166 } 167 168 static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net, 169 const struct in6_addr *addr, 170 int type, int ifindex) 171 { 172 struct ip6addrlbl_entry *p; 173 hlist_for_each_entry_rcu(p, &ip6addrlbl_table.head, list) { 174 if (__ip6addrlbl_match(net, p, addr, type, ifindex)) 175 return p; 176 } 177 return NULL; 178 } 179 180 u32 ipv6_addr_label(struct net *net, 181 const struct in6_addr *addr, int type, int ifindex) 182 { 183 u32 label; 184 struct ip6addrlbl_entry *p; 185 186 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK; 187 188 rcu_read_lock(); 189 p = __ipv6_addr_label(net, addr, type, ifindex); 190 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT; 191 rcu_read_unlock(); 192 193 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n", 194 __func__, addr, type, ifindex, label); 195 196 return label; 197 } 198 199 /* allocate one entry */ 200 static struct ip6addrlbl_entry *ip6addrlbl_alloc(struct net *net, 201 const struct in6_addr *prefix, 202 int prefixlen, int ifindex, 203 u32 label) 204 { 205 struct ip6addrlbl_entry *newp; 206 int addrtype; 207 208 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n", 209 __func__, prefix, prefixlen, ifindex, (unsigned int)label); 210 211 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK); 212 213 switch (addrtype) { 214 case IPV6_ADDR_MAPPED: 215 if (prefixlen > 96) 216 return ERR_PTR(-EINVAL); 217 if (prefixlen < 96) 218 addrtype = 0; 219 break; 220 case IPV6_ADDR_COMPATv4: 221 if (prefixlen != 96) 222 addrtype = 0; 223 break; 224 case IPV6_ADDR_LOOPBACK: 225 if (prefixlen != 128) 226 addrtype = 0; 227 break; 228 } 229 230 newp = kmalloc(sizeof(*newp), GFP_KERNEL); 231 if (!newp) 232 return ERR_PTR(-ENOMEM); 233 234 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen); 235 newp->prefixlen = prefixlen; 236 newp->ifindex = ifindex; 237 newp->addrtype = addrtype; 238 newp->label = label; 239 INIT_HLIST_NODE(&newp->list); 240 write_pnet(&newp->lbl_net, net); 241 refcount_set(&newp->refcnt, 1); 242 return newp; 243 } 244 245 /* add a label */ 246 static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace) 247 { 248 struct hlist_node *n; 249 struct ip6addrlbl_entry *last = NULL, *p = NULL; 250 int ret = 0; 251 252 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", __func__, newp, 253 replace); 254 255 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) { 256 if (p->prefixlen == newp->prefixlen && 257 net_eq(ip6addrlbl_net(p), ip6addrlbl_net(newp)) && 258 p->ifindex == newp->ifindex && 259 ipv6_addr_equal(&p->prefix, &newp->prefix)) { 260 if (!replace) { 261 ret = -EEXIST; 262 goto out; 263 } 264 hlist_replace_rcu(&p->list, &newp->list); 265 ip6addrlbl_put(p); 266 goto out; 267 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) || 268 (p->prefixlen < newp->prefixlen)) { 269 hlist_add_before_rcu(&newp->list, &p->list); 270 goto out; 271 } 272 last = p; 273 } 274 if (last) 275 hlist_add_behind_rcu(&newp->list, &last->list); 276 else 277 hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head); 278 out: 279 if (!ret) 280 ip6addrlbl_table.seq++; 281 return ret; 282 } 283 284 /* add a label */ 285 static int ip6addrlbl_add(struct net *net, 286 const struct in6_addr *prefix, int prefixlen, 287 int ifindex, u32 label, int replace) 288 { 289 struct ip6addrlbl_entry *newp; 290 int ret = 0; 291 292 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n", 293 __func__, prefix, prefixlen, ifindex, (unsigned int)label, 294 replace); 295 296 newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label); 297 if (IS_ERR(newp)) 298 return PTR_ERR(newp); 299 spin_lock(&ip6addrlbl_table.lock); 300 ret = __ip6addrlbl_add(newp, replace); 301 spin_unlock(&ip6addrlbl_table.lock); 302 if (ret) 303 ip6addrlbl_free(newp); 304 return ret; 305 } 306 307 /* remove a label */ 308 static int __ip6addrlbl_del(struct net *net, 309 const struct in6_addr *prefix, int prefixlen, 310 int ifindex) 311 { 312 struct ip6addrlbl_entry *p = NULL; 313 struct hlist_node *n; 314 int ret = -ESRCH; 315 316 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n", 317 __func__, prefix, prefixlen, ifindex); 318 319 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) { 320 if (p->prefixlen == prefixlen && 321 net_eq(ip6addrlbl_net(p), net) && 322 p->ifindex == ifindex && 323 ipv6_addr_equal(&p->prefix, prefix)) { 324 hlist_del_rcu(&p->list); 325 ip6addrlbl_put(p); 326 ret = 0; 327 break; 328 } 329 } 330 return ret; 331 } 332 333 static int ip6addrlbl_del(struct net *net, 334 const struct in6_addr *prefix, int prefixlen, 335 int ifindex) 336 { 337 struct in6_addr prefix_buf; 338 int ret; 339 340 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n", 341 __func__, prefix, prefixlen, ifindex); 342 343 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen); 344 spin_lock(&ip6addrlbl_table.lock); 345 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex); 346 spin_unlock(&ip6addrlbl_table.lock); 347 return ret; 348 } 349 350 /* add default label */ 351 static int __net_init ip6addrlbl_net_init(struct net *net) 352 { 353 int err = 0; 354 int i; 355 356 ADDRLABEL(KERN_DEBUG "%s\n", __func__); 357 358 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) { 359 int ret = ip6addrlbl_add(net, 360 ip6addrlbl_init_table[i].prefix, 361 ip6addrlbl_init_table[i].prefixlen, 362 0, 363 ip6addrlbl_init_table[i].label, 0); 364 /* XXX: should we free all rules when we catch an error? */ 365 if (ret && (!err || err != -ENOMEM)) 366 err = ret; 367 } 368 return err; 369 } 370 371 static void __net_exit ip6addrlbl_net_exit(struct net *net) 372 { 373 struct ip6addrlbl_entry *p = NULL; 374 struct hlist_node *n; 375 376 /* Remove all labels belonging to the exiting net */ 377 spin_lock(&ip6addrlbl_table.lock); 378 hlist_for_each_entry_safe(p, n, &ip6addrlbl_table.head, list) { 379 if (net_eq(ip6addrlbl_net(p), net)) { 380 hlist_del_rcu(&p->list); 381 ip6addrlbl_put(p); 382 } 383 } 384 spin_unlock(&ip6addrlbl_table.lock); 385 } 386 387 static struct pernet_operations ipv6_addr_label_ops = { 388 .init = ip6addrlbl_net_init, 389 .exit = ip6addrlbl_net_exit, 390 }; 391 392 int __init ipv6_addr_label_init(void) 393 { 394 spin_lock_init(&ip6addrlbl_table.lock); 395 396 return register_pernet_subsys(&ipv6_addr_label_ops); 397 } 398 399 void ipv6_addr_label_cleanup(void) 400 { 401 unregister_pernet_subsys(&ipv6_addr_label_ops); 402 } 403 404 static const struct nla_policy ifal_policy[IFAL_MAX+1] = { 405 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), }, 406 [IFAL_LABEL] = { .len = sizeof(u32), }, 407 }; 408 409 static bool addrlbl_ifindex_exists(struct net *net, int ifindex) 410 { 411 412 struct net_device *dev; 413 414 rcu_read_lock(); 415 dev = dev_get_by_index_rcu(net, ifindex); 416 rcu_read_unlock(); 417 418 return dev != NULL; 419 } 420 421 static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh, 422 struct netlink_ext_ack *extack) 423 { 424 struct net *net = sock_net(skb->sk); 425 struct ifaddrlblmsg *ifal; 426 struct nlattr *tb[IFAL_MAX+1]; 427 struct in6_addr *pfx; 428 u32 label; 429 int err = 0; 430 431 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy, 432 extack); 433 if (err < 0) 434 return err; 435 436 ifal = nlmsg_data(nlh); 437 438 if (ifal->ifal_family != AF_INET6 || 439 ifal->ifal_prefixlen > 128) 440 return -EINVAL; 441 442 if (!tb[IFAL_ADDRESS]) 443 return -EINVAL; 444 pfx = nla_data(tb[IFAL_ADDRESS]); 445 446 if (!tb[IFAL_LABEL]) 447 return -EINVAL; 448 label = nla_get_u32(tb[IFAL_LABEL]); 449 if (label == IPV6_ADDR_LABEL_DEFAULT) 450 return -EINVAL; 451 452 switch (nlh->nlmsg_type) { 453 case RTM_NEWADDRLABEL: 454 if (ifal->ifal_index && 455 !addrlbl_ifindex_exists(net, ifal->ifal_index)) 456 return -EINVAL; 457 458 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen, 459 ifal->ifal_index, label, 460 nlh->nlmsg_flags & NLM_F_REPLACE); 461 break; 462 case RTM_DELADDRLABEL: 463 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen, 464 ifal->ifal_index); 465 break; 466 default: 467 err = -EOPNOTSUPP; 468 } 469 return err; 470 } 471 472 static void ip6addrlbl_putmsg(struct nlmsghdr *nlh, 473 int prefixlen, int ifindex, u32 lseq) 474 { 475 struct ifaddrlblmsg *ifal = nlmsg_data(nlh); 476 ifal->ifal_family = AF_INET6; 477 ifal->ifal_prefixlen = prefixlen; 478 ifal->ifal_flags = 0; 479 ifal->ifal_index = ifindex; 480 ifal->ifal_seq = lseq; 481 }; 482 483 static int ip6addrlbl_fill(struct sk_buff *skb, 484 struct ip6addrlbl_entry *p, 485 u32 lseq, 486 u32 portid, u32 seq, int event, 487 unsigned int flags) 488 { 489 struct nlmsghdr *nlh = nlmsg_put(skb, portid, seq, event, 490 sizeof(struct ifaddrlblmsg), flags); 491 if (!nlh) 492 return -EMSGSIZE; 493 494 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq); 495 496 if (nla_put_in6_addr(skb, IFAL_ADDRESS, &p->prefix) < 0 || 497 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) { 498 nlmsg_cancel(skb, nlh); 499 return -EMSGSIZE; 500 } 501 502 nlmsg_end(skb, nlh); 503 return 0; 504 } 505 506 static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb) 507 { 508 struct net *net = sock_net(skb->sk); 509 struct ip6addrlbl_entry *p; 510 int idx = 0, s_idx = cb->args[0]; 511 int err; 512 513 rcu_read_lock(); 514 hlist_for_each_entry_rcu(p, &ip6addrlbl_table.head, list) { 515 if (idx >= s_idx && 516 net_eq(ip6addrlbl_net(p), net)) { 517 err = ip6addrlbl_fill(skb, p, 518 ip6addrlbl_table.seq, 519 NETLINK_CB(cb->skb).portid, 520 cb->nlh->nlmsg_seq, 521 RTM_NEWADDRLABEL, 522 NLM_F_MULTI); 523 if (err < 0) 524 break; 525 } 526 idx++; 527 } 528 rcu_read_unlock(); 529 cb->args[0] = idx; 530 return skb->len; 531 } 532 533 static inline int ip6addrlbl_msgsize(void) 534 { 535 return NLMSG_ALIGN(sizeof(struct ifaddrlblmsg)) 536 + nla_total_size(16) /* IFAL_ADDRESS */ 537 + nla_total_size(4); /* IFAL_LABEL */ 538 } 539 540 static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr *nlh, 541 struct netlink_ext_ack *extack) 542 { 543 struct net *net = sock_net(in_skb->sk); 544 struct ifaddrlblmsg *ifal; 545 struct nlattr *tb[IFAL_MAX+1]; 546 struct in6_addr *addr; 547 u32 lseq; 548 int err = 0; 549 struct ip6addrlbl_entry *p; 550 struct sk_buff *skb; 551 552 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy, 553 extack); 554 if (err < 0) 555 return err; 556 557 ifal = nlmsg_data(nlh); 558 559 if (ifal->ifal_family != AF_INET6 || 560 ifal->ifal_prefixlen != 128) 561 return -EINVAL; 562 563 if (ifal->ifal_index && 564 !addrlbl_ifindex_exists(net, ifal->ifal_index)) 565 return -EINVAL; 566 567 if (!tb[IFAL_ADDRESS]) 568 return -EINVAL; 569 addr = nla_data(tb[IFAL_ADDRESS]); 570 571 rcu_read_lock(); 572 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index); 573 if (p && !ip6addrlbl_hold(p)) 574 p = NULL; 575 lseq = ip6addrlbl_table.seq; 576 rcu_read_unlock(); 577 578 if (!p) { 579 err = -ESRCH; 580 goto out; 581 } 582 583 skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL); 584 if (!skb) { 585 ip6addrlbl_put(p); 586 return -ENOBUFS; 587 } 588 589 err = ip6addrlbl_fill(skb, p, lseq, 590 NETLINK_CB(in_skb).portid, nlh->nlmsg_seq, 591 RTM_NEWADDRLABEL, 0); 592 593 ip6addrlbl_put(p); 594 595 if (err < 0) { 596 WARN_ON(err == -EMSGSIZE); 597 kfree_skb(skb); 598 goto out; 599 } 600 601 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid); 602 out: 603 return err; 604 } 605 606 void __init ipv6_addr_label_rtnl_register(void) 607 { 608 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel, 609 NULL, RTNL_FLAG_DOIT_UNLOCKED); 610 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel, 611 NULL, RTNL_FLAG_DOIT_UNLOCKED); 612 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get, 613 ip6addrlbl_dump, RTNL_FLAG_DOIT_UNLOCKED); 614 } 615 616