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 <linux/jump_label_ratelimit.h> 17 #include <net/if_inet6.h> 18 #include <net/ndisc.h> 19 #include <net/flow.h> 20 #include <net/flow_dissector.h> 21 #include <net/snmp.h> 22 #include <net/netns/hash.h> 23 24 #define SIN6_LEN_RFC2133 24 25 26 #define IPV6_MAXPLEN 65535 27 28 /* 29 * NextHeader field of IPv6 header 30 */ 31 32 #define NEXTHDR_HOP 0 /* Hop-by-hop option header. */ 33 #define NEXTHDR_TCP 6 /* TCP segment. */ 34 #define NEXTHDR_UDP 17 /* UDP message. */ 35 #define NEXTHDR_IPV6 41 /* IPv6 in IPv6 */ 36 #define NEXTHDR_ROUTING 43 /* Routing header. */ 37 #define NEXTHDR_FRAGMENT 44 /* Fragmentation/reassembly header. */ 38 #define NEXTHDR_GRE 47 /* GRE header. */ 39 #define NEXTHDR_ESP 50 /* Encapsulating security payload. */ 40 #define NEXTHDR_AUTH 51 /* Authentication header. */ 41 #define NEXTHDR_ICMP 58 /* ICMP for IPv6. */ 42 #define NEXTHDR_NONE 59 /* No next header */ 43 #define NEXTHDR_DEST 60 /* Destination options header. */ 44 #define NEXTHDR_SCTP 132 /* SCTP message. */ 45 #define NEXTHDR_MOBILITY 135 /* Mobility header. */ 46 47 #define NEXTHDR_MAX 255 48 49 #define IPV6_DEFAULT_HOPLIMIT 64 50 #define IPV6_DEFAULT_MCASTHOPS 1 51 52 /* Limits on Hop-by-Hop and Destination options. 53 * 54 * Per RFC8200 there is no limit on the maximum number or lengths of options in 55 * Hop-by-Hop or Destination options other then the packet must fit in an MTU. 56 * We allow configurable limits in order to mitigate potential denial of 57 * service attacks. 58 * 59 * There are three limits that may be set: 60 * - Limit the number of options in a Hop-by-Hop or Destination options 61 * extension header 62 * - Limit the byte length of a Hop-by-Hop or Destination options extension 63 * header 64 * - Disallow unknown options 65 * 66 * The limits are expressed in corresponding sysctls: 67 * 68 * ipv6.sysctl.max_dst_opts_cnt 69 * ipv6.sysctl.max_hbh_opts_cnt 70 * ipv6.sysctl.max_dst_opts_len 71 * ipv6.sysctl.max_hbh_opts_len 72 * 73 * max_*_opts_cnt is the number of TLVs that are allowed for Destination 74 * options or Hop-by-Hop options. If the number is less than zero then unknown 75 * TLVs are disallowed and the number of known options that are allowed is the 76 * absolute value. Setting the value to INT_MAX indicates no limit. 77 * 78 * max_*_opts_len is the length limit in bytes of a Destination or 79 * Hop-by-Hop options extension header. Setting the value to INT_MAX 80 * indicates no length limit. 81 * 82 * If a limit is exceeded when processing an extension header the packet is 83 * silently discarded. 84 */ 85 86 /* Default limits for Hop-by-Hop and Destination options */ 87 #define IP6_DEFAULT_MAX_DST_OPTS_CNT 8 88 #define IP6_DEFAULT_MAX_HBH_OPTS_CNT 8 89 #define IP6_DEFAULT_MAX_DST_OPTS_LEN INT_MAX /* No limit */ 90 #define IP6_DEFAULT_MAX_HBH_OPTS_LEN INT_MAX /* No limit */ 91 92 /* 93 * Addr type 94 * 95 * type - unicast | multicast 96 * scope - local | site | global 97 * v4 - compat 98 * v4mapped 99 * any 100 * loopback 101 */ 102 103 #define IPV6_ADDR_ANY 0x0000U 104 105 #define IPV6_ADDR_UNICAST 0x0001U 106 #define IPV6_ADDR_MULTICAST 0x0002U 107 108 #define IPV6_ADDR_LOOPBACK 0x0010U 109 #define IPV6_ADDR_LINKLOCAL 0x0020U 110 #define IPV6_ADDR_SITELOCAL 0x0040U 111 112 #define IPV6_ADDR_COMPATv4 0x0080U 113 114 #define IPV6_ADDR_SCOPE_MASK 0x00f0U 115 116 #define IPV6_ADDR_MAPPED 0x1000U 117 118 /* 119 * Addr scopes 120 */ 121 #define IPV6_ADDR_MC_SCOPE(a) \ 122 ((a)->s6_addr[1] & 0x0f) /* nonstandard */ 123 #define __IPV6_ADDR_SCOPE_INVALID -1 124 #define IPV6_ADDR_SCOPE_NODELOCAL 0x01 125 #define IPV6_ADDR_SCOPE_LINKLOCAL 0x02 126 #define IPV6_ADDR_SCOPE_SITELOCAL 0x05 127 #define IPV6_ADDR_SCOPE_ORGLOCAL 0x08 128 #define IPV6_ADDR_SCOPE_GLOBAL 0x0e 129 130 /* 131 * Addr flags 132 */ 133 #define IPV6_ADDR_MC_FLAG_TRANSIENT(a) \ 134 ((a)->s6_addr[1] & 0x10) 135 #define IPV6_ADDR_MC_FLAG_PREFIX(a) \ 136 ((a)->s6_addr[1] & 0x20) 137 #define IPV6_ADDR_MC_FLAG_RENDEZVOUS(a) \ 138 ((a)->s6_addr[1] & 0x40) 139 140 /* 141 * fragmentation header 142 */ 143 144 struct frag_hdr { 145 __u8 nexthdr; 146 __u8 reserved; 147 __be16 frag_off; 148 __be32 identification; 149 }; 150 151 #define IP6_MF 0x0001 152 #define IP6_OFFSET 0xFFF8 153 154 struct ip6_fraglist_iter { 155 struct ipv6hdr *tmp_hdr; 156 struct sk_buff *frag; 157 int offset; 158 unsigned int hlen; 159 __be32 frag_id; 160 u8 nexthdr; 161 }; 162 163 int ip6_fraglist_init(struct sk_buff *skb, unsigned int hlen, u8 *prevhdr, 164 u8 nexthdr, __be32 frag_id, 165 struct ip6_fraglist_iter *iter); 166 void ip6_fraglist_prepare(struct sk_buff *skb, struct ip6_fraglist_iter *iter); 167 168 static inline struct sk_buff *ip6_fraglist_next(struct ip6_fraglist_iter *iter) 169 { 170 struct sk_buff *skb = iter->frag; 171 172 iter->frag = skb->next; 173 skb_mark_not_on_list(skb); 174 175 return skb; 176 } 177 178 struct ip6_frag_state { 179 u8 *prevhdr; 180 unsigned int hlen; 181 unsigned int mtu; 182 unsigned int left; 183 int offset; 184 int ptr; 185 int hroom; 186 int troom; 187 __be32 frag_id; 188 u8 nexthdr; 189 }; 190 191 void ip6_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int mtu, 192 unsigned short needed_tailroom, int hdr_room, u8 *prevhdr, 193 u8 nexthdr, __be32 frag_id, struct ip6_frag_state *state); 194 struct sk_buff *ip6_frag_next(struct sk_buff *skb, 195 struct ip6_frag_state *state); 196 197 #define IP6_REPLY_MARK(net, mark) \ 198 ((net)->ipv6.sysctl.fwmark_reflect ? (mark) : 0) 199 200 #include <net/sock.h> 201 202 /* sysctls */ 203 extern int sysctl_mld_max_msf; 204 extern int sysctl_mld_qrv; 205 206 #define _DEVINC(net, statname, mod, idev, field) \ 207 ({ \ 208 struct inet6_dev *_idev = (idev); \ 209 if (likely(_idev != NULL)) \ 210 mod##SNMP_INC_STATS64((_idev)->stats.statname, (field));\ 211 mod##SNMP_INC_STATS64((net)->mib.statname##_statistics, (field));\ 212 }) 213 214 /* per device counters are atomic_long_t */ 215 #define _DEVINCATOMIC(net, statname, mod, idev, field) \ 216 ({ \ 217 struct inet6_dev *_idev = (idev); \ 218 if (likely(_idev != NULL)) \ 219 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \ 220 mod##SNMP_INC_STATS((net)->mib.statname##_statistics, (field));\ 221 }) 222 223 /* per device and per net counters are atomic_long_t */ 224 #define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \ 225 ({ \ 226 struct inet6_dev *_idev = (idev); \ 227 if (likely(_idev != NULL)) \ 228 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \ 229 SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\ 230 }) 231 232 #define _DEVADD(net, statname, mod, idev, field, val) \ 233 ({ \ 234 struct inet6_dev *_idev = (idev); \ 235 if (likely(_idev != NULL)) \ 236 mod##SNMP_ADD_STATS((_idev)->stats.statname, (field), (val)); \ 237 mod##SNMP_ADD_STATS((net)->mib.statname##_statistics, (field), (val));\ 238 }) 239 240 #define _DEVUPD(net, statname, mod, idev, field, val) \ 241 ({ \ 242 struct inet6_dev *_idev = (idev); \ 243 if (likely(_idev != NULL)) \ 244 mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \ 245 mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\ 246 }) 247 248 /* MIBs */ 249 250 #define IP6_INC_STATS(net, idev,field) \ 251 _DEVINC(net, ipv6, , idev, field) 252 #define __IP6_INC_STATS(net, idev,field) \ 253 _DEVINC(net, ipv6, __, idev, field) 254 #define IP6_ADD_STATS(net, idev,field,val) \ 255 _DEVADD(net, ipv6, , idev, field, val) 256 #define __IP6_ADD_STATS(net, idev,field,val) \ 257 _DEVADD(net, ipv6, __, idev, field, val) 258 #define IP6_UPD_PO_STATS(net, idev,field,val) \ 259 _DEVUPD(net, ipv6, , idev, field, val) 260 #define __IP6_UPD_PO_STATS(net, idev,field,val) \ 261 _DEVUPD(net, ipv6, __, idev, field, val) 262 #define ICMP6_INC_STATS(net, idev, field) \ 263 _DEVINCATOMIC(net, icmpv6, , idev, field) 264 #define __ICMP6_INC_STATS(net, idev, field) \ 265 _DEVINCATOMIC(net, icmpv6, __, idev, field) 266 267 #define ICMP6MSGOUT_INC_STATS(net, idev, field) \ 268 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256) 269 #define ICMP6MSGIN_INC_STATS(net, idev, field) \ 270 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field) 271 272 struct ip6_ra_chain { 273 struct ip6_ra_chain *next; 274 struct sock *sk; 275 int sel; 276 void (*destructor)(struct sock *); 277 }; 278 279 extern struct ip6_ra_chain *ip6_ra_chain; 280 extern rwlock_t ip6_ra_lock; 281 282 /* 283 This structure is prepared by protocol, when parsing 284 ancillary data and passed to IPv6. 285 */ 286 287 struct ipv6_txoptions { 288 refcount_t refcnt; 289 /* Length of this structure */ 290 int tot_len; 291 292 /* length of extension headers */ 293 294 __u16 opt_flen; /* after fragment hdr */ 295 __u16 opt_nflen; /* before fragment hdr */ 296 297 struct ipv6_opt_hdr *hopopt; 298 struct ipv6_opt_hdr *dst0opt; 299 struct ipv6_rt_hdr *srcrt; /* Routing Header */ 300 struct ipv6_opt_hdr *dst1opt; 301 struct rcu_head rcu; 302 /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */ 303 }; 304 305 /* flowlabel_reflect sysctl values */ 306 enum flowlabel_reflect { 307 FLOWLABEL_REFLECT_ESTABLISHED = 1, 308 FLOWLABEL_REFLECT_TCP_RESET = 2, 309 FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES = 4, 310 }; 311 312 struct ip6_flowlabel { 313 struct ip6_flowlabel __rcu *next; 314 __be32 label; 315 atomic_t users; 316 struct in6_addr dst; 317 struct ipv6_txoptions *opt; 318 unsigned long linger; 319 struct rcu_head rcu; 320 u8 share; 321 union { 322 struct pid *pid; 323 kuid_t uid; 324 } owner; 325 unsigned long lastuse; 326 unsigned long expires; 327 struct net *fl_net; 328 }; 329 330 #define IPV6_FLOWINFO_MASK cpu_to_be32(0x0FFFFFFF) 331 #define IPV6_FLOWLABEL_MASK cpu_to_be32(0x000FFFFF) 332 #define IPV6_FLOWLABEL_STATELESS_FLAG cpu_to_be32(0x00080000) 333 334 #define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 335 #define IPV6_TCLASS_SHIFT 20 336 337 struct ipv6_fl_socklist { 338 struct ipv6_fl_socklist __rcu *next; 339 struct ip6_flowlabel *fl; 340 struct rcu_head rcu; 341 }; 342 343 struct ipcm6_cookie { 344 struct sockcm_cookie sockc; 345 __s16 hlimit; 346 __s16 tclass; 347 __s8 dontfrag; 348 struct ipv6_txoptions *opt; 349 __u16 gso_size; 350 }; 351 352 static inline void ipcm6_init(struct ipcm6_cookie *ipc6) 353 { 354 *ipc6 = (struct ipcm6_cookie) { 355 .hlimit = -1, 356 .tclass = -1, 357 .dontfrag = -1, 358 }; 359 } 360 361 static inline void ipcm6_init_sk(struct ipcm6_cookie *ipc6, 362 const struct ipv6_pinfo *np) 363 { 364 *ipc6 = (struct ipcm6_cookie) { 365 .hlimit = -1, 366 .tclass = np->tclass, 367 .dontfrag = np->dontfrag, 368 }; 369 } 370 371 static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np) 372 { 373 struct ipv6_txoptions *opt; 374 375 rcu_read_lock(); 376 opt = rcu_dereference(np->opt); 377 if (opt) { 378 if (!refcount_inc_not_zero(&opt->refcnt)) 379 opt = NULL; 380 else 381 opt = rcu_pointer_handoff(opt); 382 } 383 rcu_read_unlock(); 384 return opt; 385 } 386 387 static inline void txopt_put(struct ipv6_txoptions *opt) 388 { 389 if (opt && refcount_dec_and_test(&opt->refcnt)) 390 kfree_rcu(opt, rcu); 391 } 392 393 struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label); 394 395 extern struct static_key_false_deferred ipv6_flowlabel_exclusive; 396 static inline struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, 397 __be32 label) 398 { 399 if (static_branch_unlikely(&ipv6_flowlabel_exclusive.key)) 400 return __fl6_sock_lookup(sk, label) ? : ERR_PTR(-ENOENT); 401 402 return NULL; 403 } 404 405 struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space, 406 struct ip6_flowlabel *fl, 407 struct ipv6_txoptions *fopt); 408 void fl6_free_socklist(struct sock *sk); 409 int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen); 410 int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq, 411 int flags); 412 int ip6_flowlabel_init(void); 413 void ip6_flowlabel_cleanup(void); 414 bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np); 415 416 static inline void fl6_sock_release(struct ip6_flowlabel *fl) 417 { 418 if (fl) 419 atomic_dec(&fl->users); 420 } 421 422 void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info); 423 424 void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 425 struct icmp6hdr *thdr, int len); 426 427 int ip6_ra_control(struct sock *sk, int sel); 428 429 int ipv6_parse_hopopts(struct sk_buff *skb); 430 431 struct ipv6_txoptions *ipv6_dup_options(struct sock *sk, 432 struct ipv6_txoptions *opt); 433 struct ipv6_txoptions *ipv6_renew_options(struct sock *sk, 434 struct ipv6_txoptions *opt, 435 int newtype, 436 struct ipv6_opt_hdr *newopt); 437 struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space, 438 struct ipv6_txoptions *opt); 439 440 bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 441 const struct inet6_skb_parm *opt); 442 struct ipv6_txoptions *ipv6_update_options(struct sock *sk, 443 struct ipv6_txoptions *opt); 444 445 static inline bool ipv6_accept_ra(struct inet6_dev *idev) 446 { 447 /* If forwarding is enabled, RA are not accepted unless the special 448 * hybrid mode (accept_ra=2) is enabled. 449 */ 450 return idev->cnf.forwarding ? idev->cnf.accept_ra == 2 : 451 idev->cnf.accept_ra; 452 } 453 454 #define IPV6_FRAG_HIGH_THRESH (4 * 1024*1024) /* 4194304 */ 455 #define IPV6_FRAG_LOW_THRESH (3 * 1024*1024) /* 3145728 */ 456 #define IPV6_FRAG_TIMEOUT (60 * HZ) /* 60 seconds */ 457 458 int __ipv6_addr_type(const struct in6_addr *addr); 459 static inline int ipv6_addr_type(const struct in6_addr *addr) 460 { 461 return __ipv6_addr_type(addr) & 0xffff; 462 } 463 464 static inline int ipv6_addr_scope(const struct in6_addr *addr) 465 { 466 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK; 467 } 468 469 static inline int __ipv6_addr_src_scope(int type) 470 { 471 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16); 472 } 473 474 static inline int ipv6_addr_src_scope(const struct in6_addr *addr) 475 { 476 return __ipv6_addr_src_scope(__ipv6_addr_type(addr)); 477 } 478 479 static inline bool __ipv6_addr_needs_scope_id(int type) 480 { 481 return type & IPV6_ADDR_LINKLOCAL || 482 (type & IPV6_ADDR_MULTICAST && 483 (type & (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL))); 484 } 485 486 static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr, int iface) 487 { 488 return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0; 489 } 490 491 static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2) 492 { 493 return memcmp(a1, a2, sizeof(struct in6_addr)); 494 } 495 496 static inline bool 497 ipv6_masked_addr_cmp(const struct in6_addr *a1, const struct in6_addr *m, 498 const struct in6_addr *a2) 499 { 500 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 501 const unsigned long *ul1 = (const unsigned long *)a1; 502 const unsigned long *ulm = (const unsigned long *)m; 503 const unsigned long *ul2 = (const unsigned long *)a2; 504 505 return !!(((ul1[0] ^ ul2[0]) & ulm[0]) | 506 ((ul1[1] ^ ul2[1]) & ulm[1])); 507 #else 508 return !!(((a1->s6_addr32[0] ^ a2->s6_addr32[0]) & m->s6_addr32[0]) | 509 ((a1->s6_addr32[1] ^ a2->s6_addr32[1]) & m->s6_addr32[1]) | 510 ((a1->s6_addr32[2] ^ a2->s6_addr32[2]) & m->s6_addr32[2]) | 511 ((a1->s6_addr32[3] ^ a2->s6_addr32[3]) & m->s6_addr32[3])); 512 #endif 513 } 514 515 static inline void ipv6_addr_prefix(struct in6_addr *pfx, 516 const struct in6_addr *addr, 517 int plen) 518 { 519 /* caller must guarantee 0 <= plen <= 128 */ 520 int o = plen >> 3, 521 b = plen & 0x7; 522 523 memset(pfx->s6_addr, 0, sizeof(pfx->s6_addr)); 524 memcpy(pfx->s6_addr, addr, o); 525 if (b != 0) 526 pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b); 527 } 528 529 static inline void ipv6_addr_prefix_copy(struct in6_addr *addr, 530 const struct in6_addr *pfx, 531 int plen) 532 { 533 /* caller must guarantee 0 <= plen <= 128 */ 534 int o = plen >> 3, 535 b = plen & 0x7; 536 537 memcpy(addr->s6_addr, pfx, o); 538 if (b != 0) { 539 addr->s6_addr[o] &= ~(0xff00 >> b); 540 addr->s6_addr[o] |= (pfx->s6_addr[o] & (0xff00 >> b)); 541 } 542 } 543 544 static inline void __ipv6_addr_set_half(__be32 *addr, 545 __be32 wh, __be32 wl) 546 { 547 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 548 #if defined(__BIG_ENDIAN) 549 if (__builtin_constant_p(wh) && __builtin_constant_p(wl)) { 550 *(__force u64 *)addr = ((__force u64)(wh) << 32 | (__force u64)(wl)); 551 return; 552 } 553 #elif defined(__LITTLE_ENDIAN) 554 if (__builtin_constant_p(wl) && __builtin_constant_p(wh)) { 555 *(__force u64 *)addr = ((__force u64)(wl) << 32 | (__force u64)(wh)); 556 return; 557 } 558 #endif 559 #endif 560 addr[0] = wh; 561 addr[1] = wl; 562 } 563 564 static inline void ipv6_addr_set(struct in6_addr *addr, 565 __be32 w1, __be32 w2, 566 __be32 w3, __be32 w4) 567 { 568 __ipv6_addr_set_half(&addr->s6_addr32[0], w1, w2); 569 __ipv6_addr_set_half(&addr->s6_addr32[2], w3, w4); 570 } 571 572 static inline bool ipv6_addr_equal(const struct in6_addr *a1, 573 const struct in6_addr *a2) 574 { 575 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 576 const unsigned long *ul1 = (const unsigned long *)a1; 577 const unsigned long *ul2 = (const unsigned long *)a2; 578 579 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL; 580 #else 581 return ((a1->s6_addr32[0] ^ a2->s6_addr32[0]) | 582 (a1->s6_addr32[1] ^ a2->s6_addr32[1]) | 583 (a1->s6_addr32[2] ^ a2->s6_addr32[2]) | 584 (a1->s6_addr32[3] ^ a2->s6_addr32[3])) == 0; 585 #endif 586 } 587 588 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 589 static inline bool __ipv6_prefix_equal64_half(const __be64 *a1, 590 const __be64 *a2, 591 unsigned int len) 592 { 593 if (len && ((*a1 ^ *a2) & cpu_to_be64((~0UL) << (64 - len)))) 594 return false; 595 return true; 596 } 597 598 static inline bool ipv6_prefix_equal(const struct in6_addr *addr1, 599 const struct in6_addr *addr2, 600 unsigned int prefixlen) 601 { 602 const __be64 *a1 = (const __be64 *)addr1; 603 const __be64 *a2 = (const __be64 *)addr2; 604 605 if (prefixlen >= 64) { 606 if (a1[0] ^ a2[0]) 607 return false; 608 return __ipv6_prefix_equal64_half(a1 + 1, a2 + 1, prefixlen - 64); 609 } 610 return __ipv6_prefix_equal64_half(a1, a2, prefixlen); 611 } 612 #else 613 static inline bool ipv6_prefix_equal(const struct in6_addr *addr1, 614 const struct in6_addr *addr2, 615 unsigned int prefixlen) 616 { 617 const __be32 *a1 = addr1->s6_addr32; 618 const __be32 *a2 = addr2->s6_addr32; 619 unsigned int pdw, pbi; 620 621 /* check complete u32 in prefix */ 622 pdw = prefixlen >> 5; 623 if (pdw && memcmp(a1, a2, pdw << 2)) 624 return false; 625 626 /* check incomplete u32 in prefix */ 627 pbi = prefixlen & 0x1f; 628 if (pbi && ((a1[pdw] ^ a2[pdw]) & htonl((0xffffffff) << (32 - pbi)))) 629 return false; 630 631 return true; 632 } 633 #endif 634 635 static inline bool ipv6_addr_any(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; 639 640 return (ul[0] | ul[1]) == 0UL; 641 #else 642 return (a->s6_addr32[0] | a->s6_addr32[1] | 643 a->s6_addr32[2] | a->s6_addr32[3]) == 0; 644 #endif 645 } 646 647 static inline u32 ipv6_addr_hash(const struct in6_addr *a) 648 { 649 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 650 const unsigned long *ul = (const unsigned long *)a; 651 unsigned long x = ul[0] ^ ul[1]; 652 653 return (u32)(x ^ (x >> 32)); 654 #else 655 return (__force u32)(a->s6_addr32[0] ^ a->s6_addr32[1] ^ 656 a->s6_addr32[2] ^ a->s6_addr32[3]); 657 #endif 658 } 659 660 /* more secured version of ipv6_addr_hash() */ 661 static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval) 662 { 663 u32 v = (__force u32)a->s6_addr32[0] ^ (__force u32)a->s6_addr32[1]; 664 665 return jhash_3words(v, 666 (__force u32)a->s6_addr32[2], 667 (__force u32)a->s6_addr32[3], 668 initval); 669 } 670 671 static inline bool ipv6_addr_loopback(const struct in6_addr *a) 672 { 673 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 674 const __be64 *be = (const __be64 *)a; 675 676 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL; 677 #else 678 return (a->s6_addr32[0] | a->s6_addr32[1] | 679 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0; 680 #endif 681 } 682 683 /* 684 * Note that we must __force cast these to unsigned long to make sparse happy, 685 * since all of the endian-annotated types are fixed size regardless of arch. 686 */ 687 static inline bool ipv6_addr_v4mapped(const struct in6_addr *a) 688 { 689 return ( 690 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 691 *(unsigned long *)a | 692 #else 693 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) | 694 #endif 695 (__force unsigned long)(a->s6_addr32[2] ^ 696 cpu_to_be32(0x0000ffff))) == 0UL; 697 } 698 699 static inline u32 ipv6_portaddr_hash(const struct net *net, 700 const struct in6_addr *addr6, 701 unsigned int port) 702 { 703 unsigned int hash, mix = net_hash_mix(net); 704 705 if (ipv6_addr_any(addr6)) 706 hash = jhash_1word(0, mix); 707 else if (ipv6_addr_v4mapped(addr6)) 708 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix); 709 else 710 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix); 711 712 return hash ^ port; 713 } 714 715 /* 716 * Check for a RFC 4843 ORCHID address 717 * (Overlay Routable Cryptographic Hash Identifiers) 718 */ 719 static inline bool ipv6_addr_orchid(const struct in6_addr *a) 720 { 721 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010); 722 } 723 724 static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr) 725 { 726 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000); 727 } 728 729 static inline void ipv6_addr_set_v4mapped(const __be32 addr, 730 struct in6_addr *v4mapped) 731 { 732 ipv6_addr_set(v4mapped, 733 0, 0, 734 htonl(0x0000FFFF), 735 addr); 736 } 737 738 /* 739 * find the first different bit between two addresses 740 * length of address must be a multiple of 32bits 741 */ 742 static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen) 743 { 744 const __be32 *a1 = token1, *a2 = token2; 745 int i; 746 747 addrlen >>= 2; 748 749 for (i = 0; i < addrlen; i++) { 750 __be32 xb = a1[i] ^ a2[i]; 751 if (xb) 752 return i * 32 + 31 - __fls(ntohl(xb)); 753 } 754 755 /* 756 * we should *never* get to this point since that 757 * would mean the addrs are equal 758 * 759 * However, we do get to it 8) And exacly, when 760 * addresses are equal 8) 761 * 762 * ip route add 1111::/128 via ... 763 * ip route add 1111::/64 via ... 764 * and we are here. 765 * 766 * Ideally, this function should stop comparison 767 * at prefix length. It does not, but it is still OK, 768 * if returned value is greater than prefix length. 769 * --ANK (980803) 770 */ 771 return addrlen << 5; 772 } 773 774 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 775 static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen) 776 { 777 const __be64 *a1 = token1, *a2 = token2; 778 int i; 779 780 addrlen >>= 3; 781 782 for (i = 0; i < addrlen; i++) { 783 __be64 xb = a1[i] ^ a2[i]; 784 if (xb) 785 return i * 64 + 63 - __fls(be64_to_cpu(xb)); 786 } 787 788 return addrlen << 6; 789 } 790 #endif 791 792 static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen) 793 { 794 #if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 795 if (__builtin_constant_p(addrlen) && !(addrlen & 7)) 796 return __ipv6_addr_diff64(token1, token2, addrlen); 797 #endif 798 return __ipv6_addr_diff32(token1, token2, addrlen); 799 } 800 801 static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2) 802 { 803 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr)); 804 } 805 806 __be32 ipv6_select_ident(struct net *net, 807 const struct in6_addr *daddr, 808 const struct in6_addr *saddr); 809 __be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb); 810 811 int ip6_dst_hoplimit(struct dst_entry *dst); 812 813 static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6, 814 struct dst_entry *dst) 815 { 816 int hlimit; 817 818 if (ipv6_addr_is_multicast(&fl6->daddr)) 819 hlimit = np->mcast_hops; 820 else 821 hlimit = np->hop_limit; 822 if (hlimit < 0) 823 hlimit = ip6_dst_hoplimit(dst); 824 return hlimit; 825 } 826 827 /* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store 828 * Equivalent to : flow->v6addrs.src = iph->saddr; 829 * flow->v6addrs.dst = iph->daddr; 830 */ 831 static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow, 832 const struct ipv6hdr *iph) 833 { 834 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) != 835 offsetof(typeof(flow->addrs), v6addrs.src) + 836 sizeof(flow->addrs.v6addrs.src)); 837 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs)); 838 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 839 } 840 841 #if IS_ENABLED(CONFIG_IPV6) 842 843 static inline bool ipv6_can_nonlocal_bind(struct net *net, 844 struct inet_sock *inet) 845 { 846 return net->ipv6.sysctl.ip_nonlocal_bind || 847 inet->freebind || inet->transparent; 848 } 849 850 /* Sysctl settings for net ipv6.auto_flowlabels */ 851 #define IP6_AUTO_FLOW_LABEL_OFF 0 852 #define IP6_AUTO_FLOW_LABEL_OPTOUT 1 853 #define IP6_AUTO_FLOW_LABEL_OPTIN 2 854 #define IP6_AUTO_FLOW_LABEL_FORCED 3 855 856 #define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED 857 858 #define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT 859 860 static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb, 861 __be32 flowlabel, bool autolabel, 862 struct flowi6 *fl6) 863 { 864 u32 hash; 865 866 /* @flowlabel may include more than a flow label, eg, the traffic class. 867 * Here we want only the flow label value. 868 */ 869 flowlabel &= IPV6_FLOWLABEL_MASK; 870 871 if (flowlabel || 872 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF || 873 (!autolabel && 874 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED)) 875 return flowlabel; 876 877 hash = skb_get_hash_flowi6(skb, fl6); 878 879 /* Since this is being sent on the wire obfuscate hash a bit 880 * to minimize possbility that any useful information to an 881 * attacker is leaked. Only lower 20 bits are relevant. 882 */ 883 hash = rol32(hash, 16); 884 885 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK; 886 887 if (net->ipv6.sysctl.flowlabel_state_ranges) 888 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG; 889 890 return flowlabel; 891 } 892 893 static inline int ip6_default_np_autolabel(struct net *net) 894 { 895 switch (net->ipv6.sysctl.auto_flowlabels) { 896 case IP6_AUTO_FLOW_LABEL_OFF: 897 case IP6_AUTO_FLOW_LABEL_OPTIN: 898 default: 899 return 0; 900 case IP6_AUTO_FLOW_LABEL_OPTOUT: 901 case IP6_AUTO_FLOW_LABEL_FORCED: 902 return 1; 903 } 904 } 905 #else 906 static inline void ip6_set_txhash(struct sock *sk) { } 907 static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb, 908 __be32 flowlabel, bool autolabel, 909 struct flowi6 *fl6) 910 { 911 return flowlabel; 912 } 913 static inline int ip6_default_np_autolabel(struct net *net) 914 { 915 return 0; 916 } 917 #endif 918 919 #if IS_ENABLED(CONFIG_IPV6) 920 static inline int ip6_multipath_hash_policy(const struct net *net) 921 { 922 return net->ipv6.sysctl.multipath_hash_policy; 923 } 924 #else 925 static inline int ip6_multipath_hash_policy(const struct net *net) 926 { 927 return 0; 928 } 929 #endif 930 931 /* 932 * Header manipulation 933 */ 934 static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass, 935 __be32 flowlabel) 936 { 937 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel; 938 } 939 940 static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr) 941 { 942 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK; 943 } 944 945 static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr) 946 { 947 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK; 948 } 949 950 static inline u8 ip6_tclass(__be32 flowinfo) 951 { 952 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT; 953 } 954 955 static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel) 956 { 957 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel; 958 } 959 960 static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6) 961 { 962 return fl6->flowlabel & IPV6_FLOWLABEL_MASK; 963 } 964 965 /* 966 * Prototypes exported by ipv6 967 */ 968 969 /* 970 * rcv function (called from netdevice level) 971 */ 972 973 int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, 974 struct packet_type *pt, struct net_device *orig_dev); 975 void ipv6_list_rcv(struct list_head *head, struct packet_type *pt, 976 struct net_device *orig_dev); 977 978 int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb); 979 980 /* 981 * upper-layer output functions 982 */ 983 int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6, 984 __u32 mark, struct ipv6_txoptions *opt, int tclass); 985 986 int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr); 987 988 int ip6_append_data(struct sock *sk, 989 int getfrag(void *from, char *to, int offset, int len, 990 int odd, struct sk_buff *skb), 991 void *from, int length, int transhdrlen, 992 struct ipcm6_cookie *ipc6, struct flowi6 *fl6, 993 struct rt6_info *rt, unsigned int flags); 994 995 int ip6_push_pending_frames(struct sock *sk); 996 997 void ip6_flush_pending_frames(struct sock *sk); 998 999 int ip6_send_skb(struct sk_buff *skb); 1000 1001 struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue, 1002 struct inet_cork_full *cork, 1003 struct inet6_cork *v6_cork); 1004 struct sk_buff *ip6_make_skb(struct sock *sk, 1005 int getfrag(void *from, char *to, int offset, 1006 int len, int odd, struct sk_buff *skb), 1007 void *from, int length, int transhdrlen, 1008 struct ipcm6_cookie *ipc6, struct flowi6 *fl6, 1009 struct rt6_info *rt, unsigned int flags, 1010 struct inet_cork_full *cork); 1011 1012 static inline struct sk_buff *ip6_finish_skb(struct sock *sk) 1013 { 1014 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork, 1015 &inet6_sk(sk)->cork); 1016 } 1017 1018 int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst, 1019 struct flowi6 *fl6); 1020 struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6, 1021 const struct in6_addr *final_dst); 1022 struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, 1023 const struct in6_addr *final_dst, 1024 bool connected); 1025 struct dst_entry *ip6_blackhole_route(struct net *net, 1026 struct dst_entry *orig_dst); 1027 1028 /* 1029 * skb processing functions 1030 */ 1031 1032 int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb); 1033 int ip6_forward(struct sk_buff *skb); 1034 int ip6_input(struct sk_buff *skb); 1035 int ip6_mc_input(struct sk_buff *skb); 1036 void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr, 1037 bool have_final); 1038 1039 int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb); 1040 int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb); 1041 1042 /* 1043 * Extension header (options) processing 1044 */ 1045 1046 void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, 1047 u8 *proto, struct in6_addr **daddr_p, 1048 struct in6_addr *saddr); 1049 void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, 1050 u8 *proto); 1051 1052 int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp, 1053 __be16 *frag_offp); 1054 1055 bool ipv6_ext_hdr(u8 nexthdr); 1056 1057 enum { 1058 IP6_FH_F_FRAG = (1 << 0), 1059 IP6_FH_F_AUTH = (1 << 1), 1060 IP6_FH_F_SKIP_RH = (1 << 2), 1061 }; 1062 1063 /* find specified header and get offset to it */ 1064 int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target, 1065 unsigned short *fragoff, int *fragflg); 1066 1067 int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type); 1068 1069 struct in6_addr *fl6_update_dst(struct flowi6 *fl6, 1070 const struct ipv6_txoptions *opt, 1071 struct in6_addr *orig); 1072 1073 /* 1074 * socket options (ipv6_sockglue.c) 1075 */ 1076 1077 int ipv6_setsockopt(struct sock *sk, int level, int optname, 1078 char __user *optval, unsigned int optlen); 1079 int ipv6_getsockopt(struct sock *sk, int level, int optname, 1080 char __user *optval, int __user *optlen); 1081 int compat_ipv6_setsockopt(struct sock *sk, int level, int optname, 1082 char __user *optval, unsigned int optlen); 1083 int compat_ipv6_getsockopt(struct sock *sk, int level, int optname, 1084 char __user *optval, int __user *optlen); 1085 1086 int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, 1087 int addr_len); 1088 int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len); 1089 int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr, 1090 int addr_len); 1091 int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr); 1092 void ip6_datagram_release_cb(struct sock *sk); 1093 1094 int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, 1095 int *addr_len); 1096 int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len, 1097 int *addr_len); 1098 void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port, 1099 u32 info, u8 *payload); 1100 void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info); 1101 void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu); 1102 1103 int inet6_release(struct socket *sock); 1104 int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len); 1105 int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 1106 int peer); 1107 int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg); 1108 1109 int inet6_hash_connect(struct inet_timewait_death_row *death_row, 1110 struct sock *sk); 1111 1112 /* 1113 * reassembly.c 1114 */ 1115 extern const struct proto_ops inet6_stream_ops; 1116 extern const struct proto_ops inet6_dgram_ops; 1117 extern const struct proto_ops inet6_sockraw_ops; 1118 1119 struct group_source_req; 1120 struct group_filter; 1121 1122 int ip6_mc_source(int add, int omode, struct sock *sk, 1123 struct group_source_req *pgsr); 1124 int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf); 1125 int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf, 1126 struct group_filter __user *optval, int __user *optlen); 1127 1128 #ifdef CONFIG_PROC_FS 1129 int ac6_proc_init(struct net *net); 1130 void ac6_proc_exit(struct net *net); 1131 int raw6_proc_init(void); 1132 void raw6_proc_exit(void); 1133 int tcp6_proc_init(struct net *net); 1134 void tcp6_proc_exit(struct net *net); 1135 int udp6_proc_init(struct net *net); 1136 void udp6_proc_exit(struct net *net); 1137 int udplite6_proc_init(void); 1138 void udplite6_proc_exit(void); 1139 int ipv6_misc_proc_init(void); 1140 void ipv6_misc_proc_exit(void); 1141 int snmp6_register_dev(struct inet6_dev *idev); 1142 int snmp6_unregister_dev(struct inet6_dev *idev); 1143 1144 #else 1145 static inline int ac6_proc_init(struct net *net) { return 0; } 1146 static inline void ac6_proc_exit(struct net *net) { } 1147 static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; } 1148 static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; } 1149 #endif 1150 1151 #ifdef CONFIG_SYSCTL 1152 struct ctl_table *ipv6_icmp_sysctl_init(struct net *net); 1153 struct ctl_table *ipv6_route_sysctl_init(struct net *net); 1154 int ipv6_sysctl_register(void); 1155 void ipv6_sysctl_unregister(void); 1156 #endif 1157 1158 int ipv6_sock_mc_join(struct sock *sk, int ifindex, 1159 const struct in6_addr *addr); 1160 int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex, 1161 const struct in6_addr *addr, unsigned int mode); 1162 int ipv6_sock_mc_drop(struct sock *sk, int ifindex, 1163 const struct in6_addr *addr); 1164 #endif /* _NET_IPV6_H */ 1165