1/* 2 * IPv6 Address Label subsystem 3 * for the IPv6 "Default" Source Address Selection 4 * 5 * Copyright (C)2007 USAGI/WIDE Project 6 */ 7/* 8 * Author: 9 * YOSHIFUJI Hideaki @ USAGI/WIDE Project <yoshfuji@linux-ipv6.org> 10 */ 11 12#include <linux/kernel.h> 13#include <linux/list.h> 14#include <linux/rcupdate.h> 15#include <linux/in6.h> 16#include <linux/slab.h> 17#include <net/addrconf.h> 18#include <linux/if_addrlabel.h> 19#include <linux/netlink.h> 20#include <linux/rtnetlink.h> 21 22#define ADDRLABEL(x...) do { ; } while(0) 23 24/* 25 * Policy Table 26 */ 27struct ip6addrlbl_entry 28{ 29#ifdef CONFIG_NET_NS 30 struct net *lbl_net; 31#endif 32 struct in6_addr prefix; 33 int prefixlen; 34 int ifindex; 35 int addrtype; 36 u32 label; 37 struct hlist_node list; 38 atomic_t refcnt; 39 struct rcu_head rcu; 40}; 41 42static struct ip6addrlbl_table 43{ 44 struct hlist_head head; 45 spinlock_t lock; 46 u32 seq; 47} ip6addrlbl_table; 48 49static inline 50struct net *ip6addrlbl_net(const struct ip6addrlbl_entry *lbl) 51{ 52 return read_pnet(&lbl->lbl_net); 53} 54 55/* 56 * Default policy table (RFC3484 + extensions) 57 * 58 * prefix addr_type label 59 * ------------------------------------------------------------------------- 60 * ::1/128 LOOPBACK 0 61 * ::/0 N/A 1 62 * 2002::/16 N/A 2 63 * ::/96 COMPATv4 3 64 * ::ffff:0:0/96 V4MAPPED 4 65 * fc00::/7 N/A 5 ULA (RFC 4193) 66 * 2001::/32 N/A 6 Teredo (RFC 4380) 67 * 2001:10::/28 N/A 7 ORCHID (RFC 4843) 68 * 69 * Note: 0xffffffff is used if we do not have any policies. 70 */ 71 72#define IPV6_ADDR_LABEL_DEFAULT 0xffffffffUL 73 74static const __net_initdata struct ip6addrlbl_init_table 75{ 76 const struct in6_addr *prefix; 77 int prefixlen; 78 u32 label; 79} ip6addrlbl_init_table[] = { 80 { /* ::/0 */ 81 .prefix = &in6addr_any, 82 .label = 1, 83 },{ /* fc00::/7 */ 84 .prefix = &(struct in6_addr){{{ 0xfc }}}, 85 .prefixlen = 7, 86 .label = 5, 87 },{ /* 2002::/16 */ 88 .prefix = &(struct in6_addr){{{ 0x20, 0x02 }}}, 89 .prefixlen = 16, 90 .label = 2, 91 },{ /* 2001::/32 */ 92 .prefix = &(struct in6_addr){{{ 0x20, 0x01 }}}, 93 .prefixlen = 32, 94 .label = 6, 95 },{ /* 2001:10::/28 */ 96 .prefix = &(struct in6_addr){{{ 0x20, 0x01, 0x00, 0x10 }}}, 97 .prefixlen = 28, 98 .label = 7, 99 },{ /* ::ffff:0:0 */ 100 .prefix = &(struct in6_addr){{{ [10] = 0xff, [11] = 0xff }}}, 101 .prefixlen = 96, 102 .label = 4, 103 },{ /* ::/96 */ 104 .prefix = &in6addr_any, 105 .prefixlen = 96, 106 .label = 3, 107 },{ /* ::1/128 */ 108 .prefix = &in6addr_loopback, 109 .prefixlen = 128, 110 .label = 0, 111 } 112}; 113 114/* Object management */ 115static inline void ip6addrlbl_free(struct ip6addrlbl_entry *p) 116{ 117#ifdef CONFIG_NET_NS 118 release_net(p->lbl_net); 119#endif 120 kfree(p); 121} 122 123static void ip6addrlbl_free_rcu(struct rcu_head *h) 124{ 125 ip6addrlbl_free(container_of(h, struct ip6addrlbl_entry, rcu)); 126} 127 128static inline int ip6addrlbl_hold(struct ip6addrlbl_entry *p) 129{ 130 return atomic_inc_not_zero(&p->refcnt); 131} 132 133static inline void ip6addrlbl_put(struct ip6addrlbl_entry *p) 134{ 135 if (atomic_dec_and_test(&p->refcnt)) 136 call_rcu(&p->rcu, ip6addrlbl_free_rcu); 137} 138 139/* Find label */ 140static int __ip6addrlbl_match(struct net *net, 141 struct ip6addrlbl_entry *p, 142 const struct in6_addr *addr, 143 int addrtype, int ifindex) 144{ 145 if (!net_eq(ip6addrlbl_net(p), net)) 146 return 0; 147 if (p->ifindex && p->ifindex != ifindex) 148 return 0; 149 if (p->addrtype && p->addrtype != addrtype) 150 return 0; 151 if (!ipv6_prefix_equal(addr, &p->prefix, p->prefixlen)) 152 return 0; 153 return 1; 154} 155 156static struct ip6addrlbl_entry *__ipv6_addr_label(struct net *net, 157 const struct in6_addr *addr, 158 int type, int ifindex) 159{ 160 struct hlist_node *pos; 161 struct ip6addrlbl_entry *p; 162 hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) { 163 if (__ip6addrlbl_match(net, p, addr, type, ifindex)) 164 return p; 165 } 166 return NULL; 167} 168 169u32 ipv6_addr_label(struct net *net, 170 const struct in6_addr *addr, int type, int ifindex) 171{ 172 u32 label; 173 struct ip6addrlbl_entry *p; 174 175 type &= IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK; 176 177 rcu_read_lock(); 178 p = __ipv6_addr_label(net, addr, type, ifindex); 179 label = p ? p->label : IPV6_ADDR_LABEL_DEFAULT; 180 rcu_read_unlock(); 181 182 ADDRLABEL(KERN_DEBUG "%s(addr=%pI6, type=%d, ifindex=%d) => %08x\n", 183 __func__, addr, type, ifindex, label); 184 185 return label; 186} 187 188/* allocate one entry */ 189static struct ip6addrlbl_entry *ip6addrlbl_alloc(struct net *net, 190 const struct in6_addr *prefix, 191 int prefixlen, int ifindex, 192 u32 label) 193{ 194 struct ip6addrlbl_entry *newp; 195 int addrtype; 196 197 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u)\n", 198 __func__, prefix, prefixlen, ifindex, (unsigned int)label); 199 200 addrtype = ipv6_addr_type(prefix) & (IPV6_ADDR_MAPPED | IPV6_ADDR_COMPATv4 | IPV6_ADDR_LOOPBACK); 201 202 switch (addrtype) { 203 case IPV6_ADDR_MAPPED: 204 if (prefixlen > 96) 205 return ERR_PTR(-EINVAL); 206 if (prefixlen < 96) 207 addrtype = 0; 208 break; 209 case IPV6_ADDR_COMPATv4: 210 if (prefixlen != 96) 211 addrtype = 0; 212 break; 213 case IPV6_ADDR_LOOPBACK: 214 if (prefixlen != 128) 215 addrtype = 0; 216 break; 217 } 218 219 newp = kmalloc(sizeof(*newp), GFP_KERNEL); 220 if (!newp) 221 return ERR_PTR(-ENOMEM); 222 223 ipv6_addr_prefix(&newp->prefix, prefix, prefixlen); 224 newp->prefixlen = prefixlen; 225 newp->ifindex = ifindex; 226 newp->addrtype = addrtype; 227 newp->label = label; 228 INIT_HLIST_NODE(&newp->list); 229#ifdef CONFIG_NET_NS 230 newp->lbl_net = hold_net(net); 231#endif 232 atomic_set(&newp->refcnt, 1); 233 return newp; 234} 235 236/* add a label */ 237static int __ip6addrlbl_add(struct ip6addrlbl_entry *newp, int replace) 238{ 239 int ret = 0; 240 241 ADDRLABEL(KERN_DEBUG "%s(newp=%p, replace=%d)\n", 242 __func__, 243 newp, replace); 244 245 if (hlist_empty(&ip6addrlbl_table.head)) { 246 hlist_add_head_rcu(&newp->list, &ip6addrlbl_table.head); 247 } else { 248 struct hlist_node *pos, *n; 249 struct ip6addrlbl_entry *p = NULL; 250 hlist_for_each_entry_safe(p, pos, n, 251 &ip6addrlbl_table.head, list) { 252 if (p->prefixlen == newp->prefixlen && 253 net_eq(ip6addrlbl_net(p), ip6addrlbl_net(newp)) && 254 p->ifindex == newp->ifindex && 255 ipv6_addr_equal(&p->prefix, &newp->prefix)) { 256 if (!replace) { 257 ret = -EEXIST; 258 goto out; 259 } 260 hlist_replace_rcu(&p->list, &newp->list); 261 ip6addrlbl_put(p); 262 goto out; 263 } else if ((p->prefixlen == newp->prefixlen && !p->ifindex) || 264 (p->prefixlen < newp->prefixlen)) { 265 hlist_add_before_rcu(&newp->list, &p->list); 266 goto out; 267 } 268 } 269 hlist_add_after_rcu(&p->list, &newp->list); 270 } 271out: 272 if (!ret) 273 ip6addrlbl_table.seq++; 274 return ret; 275} 276 277/* add a label */ 278static int ip6addrlbl_add(struct net *net, 279 const struct in6_addr *prefix, int prefixlen, 280 int ifindex, u32 label, int replace) 281{ 282 struct ip6addrlbl_entry *newp; 283 int ret = 0; 284 285 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d, label=%u, replace=%d)\n", 286 __func__, prefix, prefixlen, ifindex, (unsigned int)label, 287 replace); 288 289 newp = ip6addrlbl_alloc(net, prefix, prefixlen, ifindex, label); 290 if (IS_ERR(newp)) 291 return PTR_ERR(newp); 292 spin_lock(&ip6addrlbl_table.lock); 293 ret = __ip6addrlbl_add(newp, replace); 294 spin_unlock(&ip6addrlbl_table.lock); 295 if (ret) 296 ip6addrlbl_free(newp); 297 return ret; 298} 299 300/* remove a label */ 301static int __ip6addrlbl_del(struct net *net, 302 const struct in6_addr *prefix, int prefixlen, 303 int ifindex) 304{ 305 struct ip6addrlbl_entry *p = NULL; 306 struct hlist_node *pos, *n; 307 int ret = -ESRCH; 308 309 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n", 310 __func__, prefix, prefixlen, ifindex); 311 312 hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) { 313 if (p->prefixlen == prefixlen && 314 net_eq(ip6addrlbl_net(p), net) && 315 p->ifindex == ifindex && 316 ipv6_addr_equal(&p->prefix, prefix)) { 317 hlist_del_rcu(&p->list); 318 ip6addrlbl_put(p); 319 ret = 0; 320 break; 321 } 322 } 323 return ret; 324} 325 326static int ip6addrlbl_del(struct net *net, 327 const struct in6_addr *prefix, int prefixlen, 328 int ifindex) 329{ 330 struct in6_addr prefix_buf; 331 int ret; 332 333 ADDRLABEL(KERN_DEBUG "%s(prefix=%pI6, prefixlen=%d, ifindex=%d)\n", 334 __func__, prefix, prefixlen, ifindex); 335 336 ipv6_addr_prefix(&prefix_buf, prefix, prefixlen); 337 spin_lock(&ip6addrlbl_table.lock); 338 ret = __ip6addrlbl_del(net, &prefix_buf, prefixlen, ifindex); 339 spin_unlock(&ip6addrlbl_table.lock); 340 return ret; 341} 342 343/* add default label */ 344static int __net_init ip6addrlbl_net_init(struct net *net) 345{ 346 int err = 0; 347 int i; 348 349 ADDRLABEL(KERN_DEBUG "%s()\n", __func__); 350 351 for (i = 0; i < ARRAY_SIZE(ip6addrlbl_init_table); i++) { 352 int ret = ip6addrlbl_add(net, 353 ip6addrlbl_init_table[i].prefix, 354 ip6addrlbl_init_table[i].prefixlen, 355 0, 356 ip6addrlbl_init_table[i].label, 0); 357 if (ret && (!err || err != -ENOMEM)) 358 err = ret; 359 } 360 return err; 361} 362 363static void __net_exit ip6addrlbl_net_exit(struct net *net) 364{ 365 struct ip6addrlbl_entry *p = NULL; 366 struct hlist_node *pos, *n; 367 368 /* Remove all labels belonging to the exiting net */ 369 spin_lock(&ip6addrlbl_table.lock); 370 hlist_for_each_entry_safe(p, pos, n, &ip6addrlbl_table.head, list) { 371 if (net_eq(ip6addrlbl_net(p), net)) { 372 hlist_del_rcu(&p->list); 373 ip6addrlbl_put(p); 374 } 375 } 376 spin_unlock(&ip6addrlbl_table.lock); 377} 378 379static struct pernet_operations ipv6_addr_label_ops = { 380 .init = ip6addrlbl_net_init, 381 .exit = ip6addrlbl_net_exit, 382}; 383 384int __init ipv6_addr_label_init(void) 385{ 386 spin_lock_init(&ip6addrlbl_table.lock); 387 388 return register_pernet_subsys(&ipv6_addr_label_ops); 389} 390 391void ipv6_addr_label_cleanup(void) 392{ 393 unregister_pernet_subsys(&ipv6_addr_label_ops); 394} 395 396static const struct nla_policy ifal_policy[IFAL_MAX+1] = { 397 [IFAL_ADDRESS] = { .len = sizeof(struct in6_addr), }, 398 [IFAL_LABEL] = { .len = sizeof(u32), }, 399}; 400 401static int ip6addrlbl_newdel(struct sk_buff *skb, struct nlmsghdr *nlh, 402 void *arg) 403{ 404 struct net *net = sock_net(skb->sk); 405 struct ifaddrlblmsg *ifal; 406 struct nlattr *tb[IFAL_MAX+1]; 407 struct in6_addr *pfx; 408 u32 label; 409 int err = 0; 410 411 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy); 412 if (err < 0) 413 return err; 414 415 ifal = nlmsg_data(nlh); 416 417 if (ifal->ifal_family != AF_INET6 || 418 ifal->ifal_prefixlen > 128) 419 return -EINVAL; 420 421 if (!tb[IFAL_ADDRESS]) 422 return -EINVAL; 423 424 pfx = nla_data(tb[IFAL_ADDRESS]); 425 if (!pfx) 426 return -EINVAL; 427 428 if (!tb[IFAL_LABEL]) 429 return -EINVAL; 430 label = nla_get_u32(tb[IFAL_LABEL]); 431 if (label == IPV6_ADDR_LABEL_DEFAULT) 432 return -EINVAL; 433 434 switch(nlh->nlmsg_type) { 435 case RTM_NEWADDRLABEL: 436 if (ifal->ifal_index && 437 !__dev_get_by_index(net, ifal->ifal_index)) 438 return -EINVAL; 439 440 err = ip6addrlbl_add(net, pfx, ifal->ifal_prefixlen, 441 ifal->ifal_index, label, 442 nlh->nlmsg_flags & NLM_F_REPLACE); 443 break; 444 case RTM_DELADDRLABEL: 445 err = ip6addrlbl_del(net, pfx, ifal->ifal_prefixlen, 446 ifal->ifal_index); 447 break; 448 default: 449 err = -EOPNOTSUPP; 450 } 451 return err; 452} 453 454static inline void ip6addrlbl_putmsg(struct nlmsghdr *nlh, 455 int prefixlen, int ifindex, u32 lseq) 456{ 457 struct ifaddrlblmsg *ifal = nlmsg_data(nlh); 458 ifal->ifal_family = AF_INET6; 459 ifal->ifal_prefixlen = prefixlen; 460 ifal->ifal_flags = 0; 461 ifal->ifal_index = ifindex; 462 ifal->ifal_seq = lseq; 463}; 464 465static int ip6addrlbl_fill(struct sk_buff *skb, 466 struct ip6addrlbl_entry *p, 467 u32 lseq, 468 u32 pid, u32 seq, int event, 469 unsigned int flags) 470{ 471 struct nlmsghdr *nlh = nlmsg_put(skb, pid, seq, event, 472 sizeof(struct ifaddrlblmsg), flags); 473 if (!nlh) 474 return -EMSGSIZE; 475 476 ip6addrlbl_putmsg(nlh, p->prefixlen, p->ifindex, lseq); 477 478 if (nla_put(skb, IFAL_ADDRESS, 16, &p->prefix) < 0 || 479 nla_put_u32(skb, IFAL_LABEL, p->label) < 0) { 480 nlmsg_cancel(skb, nlh); 481 return -EMSGSIZE; 482 } 483 484 return nlmsg_end(skb, nlh); 485} 486 487static int ip6addrlbl_dump(struct sk_buff *skb, struct netlink_callback *cb) 488{ 489 struct net *net = sock_net(skb->sk); 490 struct ip6addrlbl_entry *p; 491 struct hlist_node *pos; 492 int idx = 0, s_idx = cb->args[0]; 493 int err; 494 495 rcu_read_lock(); 496 hlist_for_each_entry_rcu(p, pos, &ip6addrlbl_table.head, list) { 497 if (idx >= s_idx && 498 net_eq(ip6addrlbl_net(p), net)) { 499 if ((err = ip6addrlbl_fill(skb, p, 500 ip6addrlbl_table.seq, 501 NETLINK_CB(cb->skb).pid, 502 cb->nlh->nlmsg_seq, 503 RTM_NEWADDRLABEL, 504 NLM_F_MULTI)) <= 0) 505 break; 506 } 507 idx++; 508 } 509 rcu_read_unlock(); 510 cb->args[0] = idx; 511 return skb->len; 512} 513 514static inline int ip6addrlbl_msgsize(void) 515{ 516 return (NLMSG_ALIGN(sizeof(struct ifaddrlblmsg)) 517 + nla_total_size(16) /* IFAL_ADDRESS */ 518 + nla_total_size(4) /* IFAL_LABEL */ 519 ); 520} 521 522static int ip6addrlbl_get(struct sk_buff *in_skb, struct nlmsghdr* nlh, 523 void *arg) 524{ 525 struct net *net = sock_net(in_skb->sk); 526 struct ifaddrlblmsg *ifal; 527 struct nlattr *tb[IFAL_MAX+1]; 528 struct in6_addr *addr; 529 u32 lseq; 530 int err = 0; 531 struct ip6addrlbl_entry *p; 532 struct sk_buff *skb; 533 534 err = nlmsg_parse(nlh, sizeof(*ifal), tb, IFAL_MAX, ifal_policy); 535 if (err < 0) 536 return err; 537 538 ifal = nlmsg_data(nlh); 539 540 if (ifal->ifal_family != AF_INET6 || 541 ifal->ifal_prefixlen != 128) 542 return -EINVAL; 543 544 if (ifal->ifal_index && 545 !__dev_get_by_index(net, ifal->ifal_index)) 546 return -EINVAL; 547 548 if (!tb[IFAL_ADDRESS]) 549 return -EINVAL; 550 551 addr = nla_data(tb[IFAL_ADDRESS]); 552 if (!addr) 553 return -EINVAL; 554 555 rcu_read_lock(); 556 p = __ipv6_addr_label(net, addr, ipv6_addr_type(addr), ifal->ifal_index); 557 if (p && ip6addrlbl_hold(p)) 558 p = NULL; 559 lseq = ip6addrlbl_table.seq; 560 rcu_read_unlock(); 561 562 if (!p) { 563 err = -ESRCH; 564 goto out; 565 } 566 567 if (!(skb = nlmsg_new(ip6addrlbl_msgsize(), GFP_KERNEL))) { 568 ip6addrlbl_put(p); 569 return -ENOBUFS; 570 } 571 572 err = ip6addrlbl_fill(skb, p, lseq, 573 NETLINK_CB(in_skb).pid, nlh->nlmsg_seq, 574 RTM_NEWADDRLABEL, 0); 575 576 ip6addrlbl_put(p); 577 578 if (err < 0) { 579 WARN_ON(err == -EMSGSIZE); 580 kfree_skb(skb); 581 goto out; 582 } 583 584 err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).pid); 585out: 586 return err; 587} 588 589void __init ipv6_addr_label_rtnl_register(void) 590{ 591 __rtnl_register(PF_INET6, RTM_NEWADDRLABEL, ip6addrlbl_newdel, NULL); 592 __rtnl_register(PF_INET6, RTM_DELADDRLABEL, ip6addrlbl_newdel, NULL); 593 __rtnl_register(PF_INET6, RTM_GETADDRLABEL, ip6addrlbl_get, ip6addrlbl_dump); 594} 595