1 /* SPDX-License-Identifier: GPL-2.0-or-later */ 2 /* 3 * Linux INET6 implementation 4 * 5 * Authors: 6 * Pedro Roque <roque@di.fc.ul.pt> 7 */ 8 9 #ifndef _NET_IPV6_H 10 #define _NET_IPV6_H 11 12 #include <linux/ipv6.h> 13 #include <linux/hardirq.h> 14 #include <linux/jhash.h> 15 #include <linux/refcount.h> 16 #include <net/if_inet6.h> 17 #include <net/ndisc.h> 18 #include <net/flow.h> 19 #include <net/flow_dissector.h> 20 #include <net/snmp.h> 21 #include <net/netns/hash.h> 22 23 #define SIN6_LEN_RFC2133 24 24 25 #define IPV6_MAXPLEN 65535 26 27 /* 28 * NextHeader field of IPv6 header 29 */ 30 31 #define NEXTHDR_HOP 0 /* Hop-by-hop option header. */ 32 #define NEXTHDR_TCP 6 /* TCP segment. */ 33 #define NEXTHDR_UDP 17 /* UDP message. */ 34 #define NEXTHDR_IPV6 41 /* IPv6 in IPv6 */ 35 #define NEXTHDR_ROUTING 43 /* Routing header. */ 36 #define NEXTHDR_FRAGMENT 44 /* Fragmentation/reassembly header. */ 37 #define NEXTHDR_GRE 47 /* GRE header. */ 38 #define NEXTHDR_ESP 50 /* Encapsulating security payload. */ 39 #define NEXTHDR_AUTH 51 /* Authentication header. */ 40 #define NEXTHDR_ICMP 58 /* ICMP for IPv6. */ 41 #define NEXTHDR_NONE 59 /* No next header */ 42 #define NEXTHDR_DEST 60 /* Destination options header. */ 43 #define NEXTHDR_SCTP 132 /* SCTP message. */ 44 #define NEXTHDR_MOBILITY 135 /* Mobility header. */ 45 46 #define NEXTHDR_MAX 255 47 48 #define IPV6_DEFAULT_HOPLIMIT 64 49 #define IPV6_DEFAULT_MCASTHOPS 1 50 51 /* Limits on Hop-by-Hop and Destination options. 52 * 53 * Per RFC8200 there is no limit on the maximum number or lengths of options in 54 * Hop-by-Hop or Destination options other then the packet must fit in an MTU. 55 * We allow configurable limits in order to mitigate potential denial of 56 * service attacks. 57 * 58 * There are three limits that may be set: 59 * - Limit the number of options in a Hop-by-Hop or Destination options 60 * extension header 61 * - Limit the byte length of a Hop-by-Hop or Destination options extension 62 * header 63 * - Disallow unknown options 64 * 65 * The limits are expressed in corresponding sysctls: 66 * 67 * ipv6.sysctl.max_dst_opts_cnt 68 * ipv6.sysctl.max_hbh_opts_cnt 69 * ipv6.sysctl.max_dst_opts_len 70 * ipv6.sysctl.max_hbh_opts_len 71 * 72 * max_*_opts_cnt is the number of TLVs that are allowed for Destination 73 * options or Hop-by-Hop options. If the number is less than zero then unknown 74 * TLVs are disallowed and the number of known options that are allowed is the 75 * absolute value. Setting the value to INT_MAX indicates no limit. 76 * 77 * max_*_opts_len is the length limit in bytes of a Destination or 78 * Hop-by-Hop options extension header. Setting the value to INT_MAX 79 * indicates no length limit. 80 * 81 * If a limit is exceeded when processing an extension header the packet is 82 * silently discarded. 83 */ 84 85 /* Default limits for Hop-by-Hop and Destination options */ 86 #define IP6_DEFAULT_MAX_DST_OPTS_CNT 8 87 #define IP6_DEFAULT_MAX_HBH_OPTS_CNT 8 88 #define IP6_DEFAULT_MAX_DST_OPTS_LEN INT_MAX /* No limit */ 89 #define IP6_DEFAULT_MAX_HBH_OPTS_LEN INT_MAX /* No limit */ 90 91 /* 92 * Addr type 93 * 94 * type - unicast | multicast 95 * scope - local | site | global 96 * v4 - compat 97 * v4mapped 98 * any 99 * loopback 100 */ 101 102 #define IPV6_ADDR_ANY 0x0000U 103 104 #define IPV6_ADDR_UNICAST 0x0001U 105 #define IPV6_ADDR_MULTICAST 0x0002U 106 107 #define IPV6_ADDR_LOOPBACK 0x0010U 108 #define IPV6_ADDR_LINKLOCAL 0x0020U 109 #define IPV6_ADDR_SITELOCAL 0x0040U 110 111 #define IPV6_ADDR_COMPATv4 0x0080U 112 113 #define IPV6_ADDR_SCOPE_MASK 0x00f0U 114 115 #define IPV6_ADDR_MAPPED 0x1000U 116 117 /* 118 * Addr scopes 119 */ 120 #define IPV6_ADDR_MC_SCOPE(a) \ 121 ((a)->s6_addr[1] & 0x0f) /* nonstandard */ 122 #define __IPV6_ADDR_SCOPE_INVALID -1 123 #define IPV6_ADDR_SCOPE_NODELOCAL 0x01 124 #define IPV6_ADDR_SCOPE_LINKLOCAL 0x02 125 #define IPV6_ADDR_SCOPE_SITELOCAL 0x05 126 #define IPV6_ADDR_SCOPE_ORGLOCAL 0x08 127 #define IPV6_ADDR_SCOPE_GLOBAL 0x0e 128 129 /* 130 * Addr flags 131 */ 132 #define IPV6_ADDR_MC_FLAG_TRANSIENT(a) \ 133 ((a)->s6_addr[1] & 0x10) 134 #define IPV6_ADDR_MC_FLAG_PREFIX(a) \ 135 ((a)->s6_addr[1] & 0x20) 136 #define IPV6_ADDR_MC_FLAG_RENDEZVOUS(a) \ 137 ((a)->s6_addr[1] & 0x40) 138 139 /* 140 * fragmentation header 141 */ 142 143 struct frag_hdr { 144 __u8 nexthdr; 145 __u8 reserved; 146 __be16 frag_off; 147 __be32 identification; 148 }; 149 150 #define IP6_MF 0x0001 151 #define IP6_OFFSET 0xFFF8 152 153 struct ip6_fraglist_iter { 154 struct ipv6hdr *tmp_hdr; 155 struct sk_buff *frag; 156 int offset; 157 unsigned int hlen; 158 __be32 frag_id; 159 u8 nexthdr; 160 }; 161 162 int ip6_fraglist_init(struct sk_buff *skb, unsigned int hlen, u8 *prevhdr, 163 u8 nexthdr, __be32 frag_id, 164 struct ip6_fraglist_iter *iter); 165 void ip6_fraglist_prepare(struct sk_buff *skb, struct ip6_fraglist_iter *iter); 166 167 static inline struct sk_buff *ip6_fraglist_next(struct ip6_fraglist_iter *iter) 168 { 169 struct sk_buff *skb = iter->frag; 170 171 iter->frag = skb->next; 172 skb_mark_not_on_list(skb); 173 174 return skb; 175 } 176 177 struct ip6_frag_state { 178 u8 *prevhdr; 179 unsigned int hlen; 180 unsigned int mtu; 181 unsigned int left; 182 int offset; 183 int ptr; 184 int hroom; 185 int troom; 186 __be32 frag_id; 187 u8 nexthdr; 188 }; 189 190 void ip6_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int mtu, 191 unsigned short needed_tailroom, int hdr_room, u8 *prevhdr, 192 u8 nexthdr, __be32 frag_id, struct ip6_frag_state *state); 193 struct sk_buff *ip6_frag_next(struct sk_buff *skb, 194 struct ip6_frag_state *state); 195 196 #define IP6_REPLY_MARK(net, mark) \ 197 ((net)->ipv6.sysctl.fwmark_reflect ? (mark) : 0) 198 199 #include <net/sock.h> 200 201 /* sysctls */ 202 extern int sysctl_mld_max_msf; 203 extern int sysctl_mld_qrv; 204 205 #define _DEVINC(net, statname, mod, idev, field) \ 206 ({ \ 207 struct inet6_dev *_idev = (idev); \ 208 if (likely(_idev != NULL)) \ 209 mod##SNMP_INC_STATS64((_idev)->stats.statname, (field));\ 210 mod##SNMP_INC_STATS64((net)->mib.statname##_statistics, (field));\ 211 }) 212 213 /* per device counters are atomic_long_t */ 214 #define _DEVINCATOMIC(net, statname, mod, idev, field) \ 215 ({ \ 216 struct inet6_dev *_idev = (idev); \ 217 if (likely(_idev != NULL)) \ 218 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \ 219 mod##SNMP_INC_STATS((net)->mib.statname##_statistics, (field));\ 220 }) 221 222 /* per device and per net counters are atomic_long_t */ 223 #define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \ 224 ({ \ 225 struct inet6_dev *_idev = (idev); \ 226 if (likely(_idev != NULL)) \ 227 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \ 228 SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\ 229 }) 230 231 #define _DEVADD(net, statname, mod, idev, field, val) \ 232 ({ \ 233 struct inet6_dev *_idev = (idev); \ 234 if (likely(_idev != NULL)) \ 235 mod##SNMP_ADD_STATS((_idev)->stats.statname, (field), (val)); \ 236 mod##SNMP_ADD_STATS((net)->mib.statname##_statistics, (field), (val));\ 237 }) 238 239 #define _DEVUPD(net, statname, mod, idev, field, val) \ 240 ({ \ 241 struct inet6_dev *_idev = (idev); \ 242 if (likely(_idev != NULL)) \ 243 mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \ 244 mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\ 245 }) 246 247 /* MIBs */ 248 249 #define IP6_INC_STATS(net, idev,field) \ 250 _DEVINC(net, ipv6, , idev, field) 251 #define __IP6_INC_STATS(net, idev,field) \ 252 _DEVINC(net, ipv6, __, idev, field) 253 #define IP6_ADD_STATS(net, idev,field,val) \ 254 _DEVADD(net, ipv6, , idev, field, val) 255 #define __IP6_ADD_STATS(net, idev,field,val) \ 256 _DEVADD(net, ipv6, __, idev, field, val) 257 #define IP6_UPD_PO_STATS(net, idev,field,val) \ 258 _DEVUPD(net, ipv6, , idev, field, val) 259 #define __IP6_UPD_PO_STATS(net, idev,field,val) \ 260 _DEVUPD(net, ipv6, __, idev, field, val) 261 #define ICMP6_INC_STATS(net, idev, field) \ 262 _DEVINCATOMIC(net, icmpv6, , idev, field) 263 #define __ICMP6_INC_STATS(net, idev, field) \ 264 _DEVINCATOMIC(net, icmpv6, __, idev, field) 265 266 #define ICMP6MSGOUT_INC_STATS(net, idev, field) \ 267 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256) 268 #define ICMP6MSGIN_INC_STATS(net, idev, field) \ 269 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field) 270 271 struct ip6_ra_chain { 272 struct ip6_ra_chain *next; 273 struct sock *sk; 274 int sel; 275 void (*destructor)(struct sock *); 276 }; 277 278 extern struct ip6_ra_chain *ip6_ra_chain; 279 extern rwlock_t ip6_ra_lock; 280 281 /* 282 This structure is prepared by protocol, when parsing 283 ancillary data and passed to IPv6. 284 */ 285 286 struct ipv6_txoptions { 287 refcount_t refcnt; 288 /* Length of this structure */ 289 int tot_len; 290 291 /* length of extension headers */ 292 293 __u16 opt_flen; /* after fragment hdr */ 294 __u16 opt_nflen; /* before fragment hdr */ 295 296 struct ipv6_opt_hdr *hopopt; 297 struct ipv6_opt_hdr *dst0opt; 298 struct ipv6_rt_hdr *srcrt; /* Routing Header */ 299 struct ipv6_opt_hdr *dst1opt; 300 struct rcu_head rcu; 301 /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */ 302 }; 303 304 /* flowlabel_reflect sysctl values */ 305 enum flowlabel_reflect { 306 FLOWLABEL_REFLECT_ESTABLISHED = 1, 307 FLOWLABEL_REFLECT_TCP_RESET = 2, 308 FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES = 4, 309 }; 310 311 struct ip6_flowlabel { 312 struct ip6_flowlabel __rcu *next; 313 __be32 label; 314 atomic_t users; 315 struct in6_addr dst; 316 struct ipv6_txoptions *opt; 317 unsigned long linger; 318 struct rcu_head rcu; 319 u8 share; 320 union { 321 struct pid *pid; 322 kuid_t uid; 323 } owner; 324 unsigned long lastuse; 325 unsigned long expires; 326 struct net *fl_net; 327 }; 328 329 #define IPV6_FLOWINFO_MASK cpu_to_be32(0x0FFFFFFF) 330 #define IPV6_FLOWLABEL_MASK cpu_to_be32(0x000FFFFF) 331 #define IPV6_FLOWLABEL_STATELESS_FLAG cpu_to_be32(0x00080000) 332 333 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 334 #define IPV6_TCLASS_SHIFT 20 335 336 struct ipv6_fl_socklist { 337 struct ipv6_fl_socklist __rcu *next; 338 struct ip6_flowlabel *fl; 339 struct rcu_head rcu; 340 }; 341 342 struct ipcm6_cookie { 343 struct sockcm_cookie sockc; 344 __s16 hlimit; 345 __s16 tclass; 346 __s8 dontfrag; 347 struct ipv6_txoptions *opt; 348 __u16 gso_size; 349 }; 350 351 static inline void ipcm6_init(struct ipcm6_cookie *ipc6) 352 { 353 *ipc6 = (struct ipcm6_cookie) { 354 .hlimit = -1, 355 .tclass = -1, 356 .dontfrag = -1, 357 }; 358 } 359 360 static inline void ipcm6_init_sk(struct ipcm6_cookie *ipc6, 361 const struct ipv6_pinfo *np) 362 { 363 *ipc6 = (struct ipcm6_cookie) { 364 .hlimit = -1, 365 .tclass = np->tclass, 366 .dontfrag = np->dontfrag, 367 }; 368 } 369 370 static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np) 371 { 372 struct ipv6_txoptions *opt; 373 374 rcu_read_lock(); 375 opt = rcu_dereference(np->opt); 376 if (opt) { 377 if (!refcount_inc_not_zero(&opt->refcnt)) 378 opt = NULL; 379 else 380 opt = rcu_pointer_handoff(opt); 381 } 382 rcu_read_unlock(); 383 return opt; 384 } 385 386 static inline void txopt_put(struct ipv6_txoptions *opt) 387 { 388 if (opt && refcount_dec_and_test(&opt->refcnt)) 389 kfree_rcu(opt, rcu); 390 } 391 392 struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, __be32 label); 393 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space, 394 struct ip6_flowlabel *fl, 395 struct ipv6_txoptions *fopt); 396 void fl6_free_socklist(struct sock *sk); 397 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen); 398 int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq, 399 int flags); 400 int ip6_flowlabel_init(void); 401 void ip6_flowlabel_cleanup(void); 402 bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np); 403 404 static inline void fl6_sock_release(struct ip6_flowlabel *fl) 405 { 406 if (fl) 407 atomic_dec(&fl->users); 408 } 409 410 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info); 411 412 void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 413 struct icmp6hdr *thdr, int len); 414 415 int ip6_ra_control(struct sock *sk, int sel); 416 417 int ipv6_parse_hopopts(struct sk_buff *skb); 418 419 struct ipv6_txoptions *ipv6_dup_options(struct sock *sk, 420 struct ipv6_txoptions *opt); 421 struct ipv6_txoptions *ipv6_renew_options(struct sock *sk, 422 struct ipv6_txoptions *opt, 423 int newtype, 424 struct ipv6_opt_hdr *newopt); 425 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space, 426 struct ipv6_txoptions *opt); 427 428 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 429 const struct inet6_skb_parm *opt); 430 struct ipv6_txoptions *ipv6_update_options(struct sock *sk, 431 struct ipv6_txoptions *opt); 432 433 static inline bool ipv6_accept_ra(struct inet6_dev *idev) 434 { 435 /* If forwarding is enabled, RA are not accepted unless the special 436 * hybrid mode (accept_ra=2) is enabled. 437 */ 438 return idev->cnf.forwarding ? idev->cnf.accept_ra == 2 : 439 idev->cnf.accept_ra; 440 } 441 442 #define IPV6_FRAG_HIGH_THRESH (4 * 1024*1024) /* 4194304 */ 443 #define IPV6_FRAG_LOW_THRESH (3 * 1024*1024) /* 3145728 */ 444 #define IPV6_FRAG_TIMEOUT (60 * HZ) /* 60 seconds */ 445 446 int __ipv6_addr_type(const struct in6_addr *addr); 447 static inline int ipv6_addr_type(const struct in6_addr *addr) 448 { 449 return __ipv6_addr_type(addr) & 0xffff; 450 } 451 452 static inline int ipv6_addr_scope(const struct in6_addr *addr) 453 { 454 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK; 455 } 456 457 static inline int __ipv6_addr_src_scope(int type) 458 { 459 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16); 460 } 461 462 static inline int ipv6_addr_src_scope(const struct in6_addr *addr) 463 { 464 return __ipv6_addr_src_scope(__ipv6_addr_type(addr)); 465 } 466 467 static inline bool __ipv6_addr_needs_scope_id(int type) 468 { 469 return type & IPV6_ADDR_LINKLOCAL || 470 (type & IPV6_ADDR_MULTICAST && 471 (type & (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL))); 472 } 473 474 static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr, int iface) 475 { 476 return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0; 477 } 478 479 static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2) 480 { 481 return memcmp(a1, a2, sizeof(struct in6_addr)); 482 } 483 484 static inline bool 485 ipv6_masked_addr_cmp(const struct in6_addr *a1, const struct in6_addr *m, 486 const struct in6_addr *a2) 487 { 488 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 489 const unsigned long *ul1 = (const unsigned long *)a1; 490 const unsigned long *ulm = (const unsigned long *)m; 491 const unsigned long *ul2 = (const unsigned long *)a2; 492 493 return !!(((ul1[0] ^ ul2[0]) & ulm[0]) | 494 ((ul1[1] ^ ul2[1]) & ulm[1])); 495 #else 496 return !!(((a1->s6_addr32[0] ^ a2->s6_addr32[0]) & m->s6_addr32[0]) | 497 ((a1->s6_addr32[1] ^ a2->s6_addr32[1]) & m->s6_addr32[1]) | 498 ((a1->s6_addr32[2] ^ a2->s6_addr32[2]) & m->s6_addr32[2]) | 499 ((a1->s6_addr32[3] ^ a2->s6_addr32[3]) & m->s6_addr32[3])); 500 #endif 501 } 502 503 static inline void ipv6_addr_prefix(struct in6_addr *pfx, 504 const struct in6_addr *addr, 505 int plen) 506 { 507 /* caller must guarantee 0 <= plen <= 128 */ 508 int o = plen >> 3, 509 b = plen & 0x7; 510 511 memset(pfx->s6_addr, 0, sizeof(pfx->s6_addr)); 512 memcpy(pfx->s6_addr, addr, o); 513 if (b != 0) 514 pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b); 515 } 516 517 static inline void ipv6_addr_prefix_copy(struct in6_addr *addr, 518 const struct in6_addr *pfx, 519 int plen) 520 { 521 /* caller must guarantee 0 <= plen <= 128 */ 522 int o = plen >> 3, 523 b = plen & 0x7; 524 525 memcpy(addr->s6_addr, pfx, o); 526 if (b != 0) { 527 addr->s6_addr[o] &= ~(0xff00 >> b); 528 addr->s6_addr[o] |= (pfx->s6_addr[o] & (0xff00 >> b)); 529 } 530 } 531 532 static inline void __ipv6_addr_set_half(__be32 *addr, 533 __be32 wh, __be32 wl) 534 { 535 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 536 #if defined(__BIG_ENDIAN) 537 if (__builtin_constant_p(wh) && __builtin_constant_p(wl)) { 538 *(__force u64 *)addr = ((__force u64)(wh) << 32 | (__force u64)(wl)); 539 return; 540 } 541 #elif defined(__LITTLE_ENDIAN) 542 if (__builtin_constant_p(wl) && __builtin_constant_p(wh)) { 543 *(__force u64 *)addr = ((__force u64)(wl) << 32 | (__force u64)(wh)); 544 return; 545 } 546 #endif 547 #endif 548 addr[0] = wh; 549 addr[1] = wl; 550 } 551 552 static inline void ipv6_addr_set(struct in6_addr *addr, 553 __be32 w1, __be32 w2, 554 __be32 w3, __be32 w4) 555 { 556 __ipv6_addr_set_half(&addr->s6_addr32[0], w1, w2); 557 __ipv6_addr_set_half(&addr->s6_addr32[2], w3, w4); 558 } 559 560 static inline bool ipv6_addr_equal(const struct in6_addr *a1, 561 const struct in6_addr *a2) 562 { 563 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 564 const unsigned long *ul1 = (const unsigned long *)a1; 565 const unsigned long *ul2 = (const unsigned long *)a2; 566 567 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL; 568 #else 569 return ((a1->s6_addr32[0] ^ a2->s6_addr32[0]) | 570 (a1->s6_addr32[1] ^ a2->s6_addr32[1]) | 571 (a1->s6_addr32[2] ^ a2->s6_addr32[2]) | 572 (a1->s6_addr32[3] ^ a2->s6_addr32[3])) == 0; 573 #endif 574 } 575 576 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 577 static inline bool __ipv6_prefix_equal64_half(const __be64 *a1, 578 const __be64 *a2, 579 unsigned int len) 580 { 581 if (len && ((*a1 ^ *a2) & cpu_to_be64((~0UL) << (64 - len)))) 582 return false; 583 return true; 584 } 585 586 static inline bool ipv6_prefix_equal(const struct in6_addr *addr1, 587 const struct in6_addr *addr2, 588 unsigned int prefixlen) 589 { 590 const __be64 *a1 = (const __be64 *)addr1; 591 const __be64 *a2 = (const __be64 *)addr2; 592 593 if (prefixlen >= 64) { 594 if (a1[0] ^ a2[0]) 595 return false; 596 return __ipv6_prefix_equal64_half(a1 + 1, a2 + 1, prefixlen - 64); 597 } 598 return __ipv6_prefix_equal64_half(a1, a2, prefixlen); 599 } 600 #else 601 static inline bool ipv6_prefix_equal(const struct in6_addr *addr1, 602 const struct in6_addr *addr2, 603 unsigned int prefixlen) 604 { 605 const __be32 *a1 = addr1->s6_addr32; 606 const __be32 *a2 = addr2->s6_addr32; 607 unsigned int pdw, pbi; 608 609 /* check complete u32 in prefix */ 610 pdw = prefixlen >> 5; 611 if (pdw && memcmp(a1, a2, pdw << 2)) 612 return false; 613 614 /* check incomplete u32 in prefix */ 615 pbi = prefixlen & 0x1f; 616 if (pbi && ((a1[pdw] ^ a2[pdw]) & htonl((0xffffffff) << (32 - pbi)))) 617 return false; 618 619 return true; 620 } 621 #endif 622 623 static inline bool ipv6_addr_any(const struct in6_addr *a) 624 { 625 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 626 const unsigned long *ul = (const unsigned long *)a; 627 628 return (ul[0] | ul[1]) == 0UL; 629 #else 630 return (a->s6_addr32[0] | a->s6_addr32[1] | 631 a->s6_addr32[2] | a->s6_addr32[3]) == 0; 632 #endif 633 } 634 635 static inline u32 ipv6_addr_hash(const struct in6_addr *a) 636 { 637 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 638 const unsigned long *ul = (const unsigned long *)a; 639 unsigned long x = ul[0] ^ ul[1]; 640 641 return (u32)(x ^ (x >> 32)); 642 #else 643 return (__force u32)(a->s6_addr32[0] ^ a->s6_addr32[1] ^ 644 a->s6_addr32[2] ^ a->s6_addr32[3]); 645 #endif 646 } 647 648 /* more secured version of ipv6_addr_hash() */ 649 static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval) 650 { 651 u32 v = (__force u32)a->s6_addr32[0] ^ (__force u32)a->s6_addr32[1]; 652 653 return jhash_3words(v, 654 (__force u32)a->s6_addr32[2], 655 (__force u32)a->s6_addr32[3], 656 initval); 657 } 658 659 static inline bool ipv6_addr_loopback(const struct in6_addr *a) 660 { 661 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 662 const __be64 *be = (const __be64 *)a; 663 664 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL; 665 #else 666 return (a->s6_addr32[0] | a->s6_addr32[1] | 667 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0; 668 #endif 669 } 670 671 /* 672 * Note that we must __force cast these to unsigned long to make sparse happy, 673 * since all of the endian-annotated types are fixed size regardless of arch. 674 */ 675 static inline bool ipv6_addr_v4mapped(const struct in6_addr *a) 676 { 677 return ( 678 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 679 *(unsigned long *)a | 680 #else 681 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) | 682 #endif 683 (__force unsigned long)(a->s6_addr32[2] ^ 684 cpu_to_be32(0x0000ffff))) == 0UL; 685 } 686 687 static inline u32 ipv6_portaddr_hash(const struct net *net, 688 const struct in6_addr *addr6, 689 unsigned int port) 690 { 691 unsigned int hash, mix = net_hash_mix(net); 692 693 if (ipv6_addr_any(addr6)) 694 hash = jhash_1word(0, mix); 695 else if (ipv6_addr_v4mapped(addr6)) 696 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix); 697 else 698 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix); 699 700 return hash ^ port; 701 } 702 703 /* 704 * Check for a RFC 4843 ORCHID address 705 * (Overlay Routable Cryptographic Hash Identifiers) 706 */ 707 static inline bool ipv6_addr_orchid(const struct in6_addr *a) 708 { 709 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010); 710 } 711 712 static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr) 713 { 714 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000); 715 } 716 717 static inline void ipv6_addr_set_v4mapped(const __be32 addr, 718 struct in6_addr *v4mapped) 719 { 720 ipv6_addr_set(v4mapped, 721 0, 0, 722 htonl(0x0000FFFF), 723 addr); 724 } 725 726 /* 727 * find the first different bit between two addresses 728 * length of address must be a multiple of 32bits 729 */ 730 static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen) 731 { 732 const __be32 *a1 = token1, *a2 = token2; 733 int i; 734 735 addrlen >>= 2; 736 737 for (i = 0; i < addrlen; i++) { 738 __be32 xb = a1[i] ^ a2[i]; 739 if (xb) 740 return i * 32 + 31 - __fls(ntohl(xb)); 741 } 742 743 /* 744 * we should *never* get to this point since that 745 * would mean the addrs are equal 746 * 747 * However, we do get to it 8) And exacly, when 748 * addresses are equal 8) 749 * 750 * ip route add 1111::/128 via ... 751 * ip route add 1111::/64 via ... 752 * and we are here. 753 * 754 * Ideally, this function should stop comparison 755 * at prefix length. It does not, but it is still OK, 756 * if returned value is greater than prefix length. 757 * --ANK (980803) 758 */ 759 return addrlen << 5; 760 } 761 762 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 763 static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen) 764 { 765 const __be64 *a1 = token1, *a2 = token2; 766 int i; 767 768 addrlen >>= 3; 769 770 for (i = 0; i < addrlen; i++) { 771 __be64 xb = a1[i] ^ a2[i]; 772 if (xb) 773 return i * 64 + 63 - __fls(be64_to_cpu(xb)); 774 } 775 776 return addrlen << 6; 777 } 778 #endif 779 780 static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen) 781 { 782 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 783 if (__builtin_constant_p(addrlen) && !(addrlen & 7)) 784 return __ipv6_addr_diff64(token1, token2, addrlen); 785 #endif 786 return __ipv6_addr_diff32(token1, token2, addrlen); 787 } 788 789 static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2) 790 { 791 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr)); 792 } 793 794 __be32 ipv6_select_ident(struct net *net, 795 const struct in6_addr *daddr, 796 const struct in6_addr *saddr); 797 __be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb); 798 799 int ip6_dst_hoplimit(struct dst_entry *dst); 800 801 static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6, 802 struct dst_entry *dst) 803 { 804 int hlimit; 805 806 if (ipv6_addr_is_multicast(&fl6->daddr)) 807 hlimit = np->mcast_hops; 808 else 809 hlimit = np->hop_limit; 810 if (hlimit < 0) 811 hlimit = ip6_dst_hoplimit(dst); 812 return hlimit; 813 } 814 815 /* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store 816 * Equivalent to : flow->v6addrs.src = iph->saddr; 817 * flow->v6addrs.dst = iph->daddr; 818 */ 819 static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow, 820 const struct ipv6hdr *iph) 821 { 822 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) != 823 offsetof(typeof(flow->addrs), v6addrs.src) + 824 sizeof(flow->addrs.v6addrs.src)); 825 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs)); 826 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 827 } 828 829 #if IS_ENABLED(CONFIG_IPV6) 830 831 static inline bool ipv6_can_nonlocal_bind(struct net *net, 832 struct inet_sock *inet) 833 { 834 return net->ipv6.sysctl.ip_nonlocal_bind || 835 inet->freebind || inet->transparent; 836 } 837 838 /* Sysctl settings for net ipv6.auto_flowlabels */ 839 #define IP6_AUTO_FLOW_LABEL_OFF 0 840 #define IP6_AUTO_FLOW_LABEL_OPTOUT 1 841 #define IP6_AUTO_FLOW_LABEL_OPTIN 2 842 #define IP6_AUTO_FLOW_LABEL_FORCED 3 843 844 #define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED 845 846 #define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT 847 848 static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb, 849 __be32 flowlabel, bool autolabel, 850 struct flowi6 *fl6) 851 { 852 u32 hash; 853 854 /* @flowlabel may include more than a flow label, eg, the traffic class. 855 * Here we want only the flow label value. 856 */ 857 flowlabel &= IPV6_FLOWLABEL_MASK; 858 859 if (flowlabel || 860 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF || 861 (!autolabel && 862 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED)) 863 return flowlabel; 864 865 hash = skb_get_hash_flowi6(skb, fl6); 866 867 /* Since this is being sent on the wire obfuscate hash a bit 868 * to minimize possbility that any useful information to an 869 * attacker is leaked. Only lower 20 bits are relevant. 870 */ 871 hash = rol32(hash, 16); 872 873 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK; 874 875 if (net->ipv6.sysctl.flowlabel_state_ranges) 876 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG; 877 878 return flowlabel; 879 } 880 881 static inline int ip6_default_np_autolabel(struct net *net) 882 { 883 switch (net->ipv6.sysctl.auto_flowlabels) { 884 case IP6_AUTO_FLOW_LABEL_OFF: 885 case IP6_AUTO_FLOW_LABEL_OPTIN: 886 default: 887 return 0; 888 case IP6_AUTO_FLOW_LABEL_OPTOUT: 889 case IP6_AUTO_FLOW_LABEL_FORCED: 890 return 1; 891 } 892 } 893 #else 894 static inline void ip6_set_txhash(struct sock *sk) { } 895 static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb, 896 __be32 flowlabel, bool autolabel, 897 struct flowi6 *fl6) 898 { 899 return flowlabel; 900 } 901 static inline int ip6_default_np_autolabel(struct net *net) 902 { 903 return 0; 904 } 905 #endif 906 907 #if IS_ENABLED(CONFIG_IPV6) 908 static inline int ip6_multipath_hash_policy(const struct net *net) 909 { 910 return net->ipv6.sysctl.multipath_hash_policy; 911 } 912 #else 913 static inline int ip6_multipath_hash_policy(const struct net *net) 914 { 915 return 0; 916 } 917 #endif 918 919 /* 920 * Header manipulation 921 */ 922 static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass, 923 __be32 flowlabel) 924 { 925 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel; 926 } 927 928 static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr) 929 { 930 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK; 931 } 932 933 static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr) 934 { 935 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK; 936 } 937 938 static inline u8 ip6_tclass(__be32 flowinfo) 939 { 940 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT; 941 } 942 943 static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel) 944 { 945 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel; 946 } 947 948 static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6) 949 { 950 return fl6->flowlabel & IPV6_FLOWLABEL_MASK; 951 } 952 953 /* 954 * Prototypes exported by ipv6 955 */ 956 957 /* 958 * rcv function (called from netdevice level) 959 */ 960 961 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, 962 struct packet_type *pt, struct net_device *orig_dev); 963 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt, 964 struct net_device *orig_dev); 965 966 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb); 967 968 /* 969 * upper-layer output functions 970 */ 971 int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6, 972 __u32 mark, struct ipv6_txoptions *opt, int tclass); 973 974 int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr); 975 976 int ip6_append_data(struct sock *sk, 977 int getfrag(void *from, char *to, int offset, int len, 978 int odd, struct sk_buff *skb), 979 void *from, int length, int transhdrlen, 980 struct ipcm6_cookie *ipc6, struct flowi6 *fl6, 981 struct rt6_info *rt, unsigned int flags); 982 983 int ip6_push_pending_frames(struct sock *sk); 984 985 void ip6_flush_pending_frames(struct sock *sk); 986 987 int ip6_send_skb(struct sk_buff *skb); 988 989 struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue, 990 struct inet_cork_full *cork, 991 struct inet6_cork *v6_cork); 992 struct sk_buff *ip6_make_skb(struct sock *sk, 993 int getfrag(void *from, char *to, int offset, 994 int len, int odd, struct sk_buff *skb), 995 void *from, int length, int transhdrlen, 996 struct ipcm6_cookie *ipc6, struct flowi6 *fl6, 997 struct rt6_info *rt, unsigned int flags, 998 struct inet_cork_full *cork); 999 1000 static inline struct sk_buff *ip6_finish_skb(struct sock *sk) 1001 { 1002 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork, 1003 &inet6_sk(sk)->cork); 1004 } 1005 1006 int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst, 1007 struct flowi6 *fl6); 1008 struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6, 1009 const struct in6_addr *final_dst); 1010 struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, 1011 const struct in6_addr *final_dst, 1012 bool connected); 1013 struct dst_entry *ip6_blackhole_route(struct net *net, 1014 struct dst_entry *orig_dst); 1015 1016 /* 1017 * skb processing functions 1018 */ 1019 1020 int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb); 1021 int ip6_forward(struct sk_buff *skb); 1022 int ip6_input(struct sk_buff *skb); 1023 int ip6_mc_input(struct sk_buff *skb); 1024 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr, 1025 bool have_final); 1026 1027 int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb); 1028 int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb); 1029 1030 /* 1031 * Extension header (options) processing 1032 */ 1033 1034 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, 1035 u8 *proto, struct in6_addr **daddr_p, 1036 struct in6_addr *saddr); 1037 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, 1038 u8 *proto); 1039 1040 int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp, 1041 __be16 *frag_offp); 1042 1043 bool ipv6_ext_hdr(u8 nexthdr); 1044 1045 enum { 1046 IP6_FH_F_FRAG = (1 << 0), 1047 IP6_FH_F_AUTH = (1 << 1), 1048 IP6_FH_F_SKIP_RH = (1 << 2), 1049 }; 1050 1051 /* find specified header and get offset to it */ 1052 int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target, 1053 unsigned short *fragoff, int *fragflg); 1054 1055 int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type); 1056 1057 struct in6_addr *fl6_update_dst(struct flowi6 *fl6, 1058 const struct ipv6_txoptions *opt, 1059 struct in6_addr *orig); 1060 1061 /* 1062 * socket options (ipv6_sockglue.c) 1063 */ 1064 1065 int ipv6_setsockopt(struct sock *sk, int level, int optname, 1066 char __user *optval, unsigned int optlen); 1067 int ipv6_getsockopt(struct sock *sk, int level, int optname, 1068 char __user *optval, int __user *optlen); 1069 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname, 1070 char __user *optval, unsigned int optlen); 1071 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname, 1072 char __user *optval, int __user *optlen); 1073 1074 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, 1075 int addr_len); 1076 int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len); 1077 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr, 1078 int addr_len); 1079 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr); 1080 void ip6_datagram_release_cb(struct sock *sk); 1081 1082 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, 1083 int *addr_len); 1084 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len, 1085 int *addr_len); 1086 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port, 1087 u32 info, u8 *payload); 1088 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info); 1089 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu); 1090 1091 int inet6_release(struct socket *sock); 1092 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len); 1093 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 1094 int peer); 1095 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg); 1096 1097 int inet6_hash_connect(struct inet_timewait_death_row *death_row, 1098 struct sock *sk); 1099 1100 /* 1101 * reassembly.c 1102 */ 1103 extern const struct proto_ops inet6_stream_ops; 1104 extern const struct proto_ops inet6_dgram_ops; 1105 extern const struct proto_ops inet6_sockraw_ops; 1106 1107 struct group_source_req; 1108 struct group_filter; 1109 1110 int ip6_mc_source(int add, int omode, struct sock *sk, 1111 struct group_source_req *pgsr); 1112 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf); 1113 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf, 1114 struct group_filter __user *optval, int __user *optlen); 1115 1116 #ifdef CONFIG_PROC_FS 1117 int ac6_proc_init(struct net *net); 1118 void ac6_proc_exit(struct net *net); 1119 int raw6_proc_init(void); 1120 void raw6_proc_exit(void); 1121 int tcp6_proc_init(struct net *net); 1122 void tcp6_proc_exit(struct net *net); 1123 int udp6_proc_init(struct net *net); 1124 void udp6_proc_exit(struct net *net); 1125 int udplite6_proc_init(void); 1126 void udplite6_proc_exit(void); 1127 int ipv6_misc_proc_init(void); 1128 void ipv6_misc_proc_exit(void); 1129 int snmp6_register_dev(struct inet6_dev *idev); 1130 int snmp6_unregister_dev(struct inet6_dev *idev); 1131 1132 #else 1133 static inline int ac6_proc_init(struct net *net) { return 0; } 1134 static inline void ac6_proc_exit(struct net *net) { } 1135 static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; } 1136 static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; } 1137 #endif 1138 1139 #ifdef CONFIG_SYSCTL 1140 struct ctl_table *ipv6_icmp_sysctl_init(struct net *net); 1141 struct ctl_table *ipv6_route_sysctl_init(struct net *net); 1142 int ipv6_sysctl_register(void); 1143 void ipv6_sysctl_unregister(void); 1144 #endif 1145 1146 int ipv6_sock_mc_join(struct sock *sk, int ifindex, 1147 const struct in6_addr *addr); 1148 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex, 1149 const struct in6_addr *addr, unsigned int mode); 1150 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, 1151 const struct in6_addr *addr); 1152 #endif /* _NET_IPV6_H */ 1153