at v5.2 1102 lines 32 kB view raw
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 143struct 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#define IP6_REPLY_MARK(net, mark) \ 154 ((net)->ipv6.sysctl.fwmark_reflect ? (mark) : 0) 155 156#include <net/sock.h> 157 158/* sysctls */ 159extern int sysctl_mld_max_msf; 160extern int sysctl_mld_qrv; 161 162#define _DEVINC(net, statname, mod, idev, field) \ 163({ \ 164 struct inet6_dev *_idev = (idev); \ 165 if (likely(_idev != NULL)) \ 166 mod##SNMP_INC_STATS64((_idev)->stats.statname, (field));\ 167 mod##SNMP_INC_STATS64((net)->mib.statname##_statistics, (field));\ 168}) 169 170/* per device counters are atomic_long_t */ 171#define _DEVINCATOMIC(net, statname, mod, idev, field) \ 172({ \ 173 struct inet6_dev *_idev = (idev); \ 174 if (likely(_idev != NULL)) \ 175 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \ 176 mod##SNMP_INC_STATS((net)->mib.statname##_statistics, (field));\ 177}) 178 179/* per device and per net counters are atomic_long_t */ 180#define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \ 181({ \ 182 struct inet6_dev *_idev = (idev); \ 183 if (likely(_idev != NULL)) \ 184 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \ 185 SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\ 186}) 187 188#define _DEVADD(net, statname, mod, idev, field, val) \ 189({ \ 190 struct inet6_dev *_idev = (idev); \ 191 if (likely(_idev != NULL)) \ 192 mod##SNMP_ADD_STATS((_idev)->stats.statname, (field), (val)); \ 193 mod##SNMP_ADD_STATS((net)->mib.statname##_statistics, (field), (val));\ 194}) 195 196#define _DEVUPD(net, statname, mod, idev, field, val) \ 197({ \ 198 struct inet6_dev *_idev = (idev); \ 199 if (likely(_idev != NULL)) \ 200 mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \ 201 mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\ 202}) 203 204/* MIBs */ 205 206#define IP6_INC_STATS(net, idev,field) \ 207 _DEVINC(net, ipv6, , idev, field) 208#define __IP6_INC_STATS(net, idev,field) \ 209 _DEVINC(net, ipv6, __, idev, field) 210#define IP6_ADD_STATS(net, idev,field,val) \ 211 _DEVADD(net, ipv6, , idev, field, val) 212#define __IP6_ADD_STATS(net, idev,field,val) \ 213 _DEVADD(net, ipv6, __, idev, field, val) 214#define IP6_UPD_PO_STATS(net, idev,field,val) \ 215 _DEVUPD(net, ipv6, , idev, field, val) 216#define __IP6_UPD_PO_STATS(net, idev,field,val) \ 217 _DEVUPD(net, ipv6, __, idev, field, val) 218#define ICMP6_INC_STATS(net, idev, field) \ 219 _DEVINCATOMIC(net, icmpv6, , idev, field) 220#define __ICMP6_INC_STATS(net, idev, field) \ 221 _DEVINCATOMIC(net, icmpv6, __, idev, field) 222 223#define ICMP6MSGOUT_INC_STATS(net, idev, field) \ 224 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256) 225#define ICMP6MSGIN_INC_STATS(net, idev, field) \ 226 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field) 227 228struct ip6_ra_chain { 229 struct ip6_ra_chain *next; 230 struct sock *sk; 231 int sel; 232 void (*destructor)(struct sock *); 233}; 234 235extern struct ip6_ra_chain *ip6_ra_chain; 236extern rwlock_t ip6_ra_lock; 237 238/* 239 This structure is prepared by protocol, when parsing 240 ancillary data and passed to IPv6. 241 */ 242 243struct ipv6_txoptions { 244 refcount_t refcnt; 245 /* Length of this structure */ 246 int tot_len; 247 248 /* length of extension headers */ 249 250 __u16 opt_flen; /* after fragment hdr */ 251 __u16 opt_nflen; /* before fragment hdr */ 252 253 struct ipv6_opt_hdr *hopopt; 254 struct ipv6_opt_hdr *dst0opt; 255 struct ipv6_rt_hdr *srcrt; /* Routing Header */ 256 struct ipv6_opt_hdr *dst1opt; 257 struct rcu_head rcu; 258 /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */ 259}; 260 261struct ip6_flowlabel { 262 struct ip6_flowlabel __rcu *next; 263 __be32 label; 264 atomic_t users; 265 struct in6_addr dst; 266 struct ipv6_txoptions *opt; 267 unsigned long linger; 268 struct rcu_head rcu; 269 u8 share; 270 union { 271 struct pid *pid; 272 kuid_t uid; 273 } owner; 274 unsigned long lastuse; 275 unsigned long expires; 276 struct net *fl_net; 277}; 278 279#define IPV6_FLOWINFO_MASK cpu_to_be32(0x0FFFFFFF) 280#define IPV6_FLOWLABEL_MASK cpu_to_be32(0x000FFFFF) 281#define IPV6_FLOWLABEL_STATELESS_FLAG cpu_to_be32(0x00080000) 282 283#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK) 284#define IPV6_TCLASS_SHIFT 20 285 286struct ipv6_fl_socklist { 287 struct ipv6_fl_socklist __rcu *next; 288 struct ip6_flowlabel *fl; 289 struct rcu_head rcu; 290}; 291 292struct ipcm6_cookie { 293 struct sockcm_cookie sockc; 294 __s16 hlimit; 295 __s16 tclass; 296 __s8 dontfrag; 297 struct ipv6_txoptions *opt; 298 __u16 gso_size; 299}; 300 301static inline void ipcm6_init(struct ipcm6_cookie *ipc6) 302{ 303 *ipc6 = (struct ipcm6_cookie) { 304 .hlimit = -1, 305 .tclass = -1, 306 .dontfrag = -1, 307 }; 308} 309 310static inline void ipcm6_init_sk(struct ipcm6_cookie *ipc6, 311 const struct ipv6_pinfo *np) 312{ 313 *ipc6 = (struct ipcm6_cookie) { 314 .hlimit = -1, 315 .tclass = np->tclass, 316 .dontfrag = np->dontfrag, 317 }; 318} 319 320static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np) 321{ 322 struct ipv6_txoptions *opt; 323 324 rcu_read_lock(); 325 opt = rcu_dereference(np->opt); 326 if (opt) { 327 if (!refcount_inc_not_zero(&opt->refcnt)) 328 opt = NULL; 329 else 330 opt = rcu_pointer_handoff(opt); 331 } 332 rcu_read_unlock(); 333 return opt; 334} 335 336static inline void txopt_put(struct ipv6_txoptions *opt) 337{ 338 if (opt && refcount_dec_and_test(&opt->refcnt)) 339 kfree_rcu(opt, rcu); 340} 341 342struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, __be32 label); 343struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space, 344 struct ip6_flowlabel *fl, 345 struct ipv6_txoptions *fopt); 346void fl6_free_socklist(struct sock *sk); 347int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen); 348int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq, 349 int flags); 350int ip6_flowlabel_init(void); 351void ip6_flowlabel_cleanup(void); 352bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np); 353 354static inline void fl6_sock_release(struct ip6_flowlabel *fl) 355{ 356 if (fl) 357 atomic_dec(&fl->users); 358} 359 360void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info); 361 362void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6, 363 struct icmp6hdr *thdr, int len); 364 365int ip6_ra_control(struct sock *sk, int sel); 366 367int ipv6_parse_hopopts(struct sk_buff *skb); 368 369struct ipv6_txoptions *ipv6_dup_options(struct sock *sk, 370 struct ipv6_txoptions *opt); 371struct ipv6_txoptions *ipv6_renew_options(struct sock *sk, 372 struct ipv6_txoptions *opt, 373 int newtype, 374 struct ipv6_opt_hdr *newopt); 375struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space, 376 struct ipv6_txoptions *opt); 377 378bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb, 379 const struct inet6_skb_parm *opt); 380struct ipv6_txoptions *ipv6_update_options(struct sock *sk, 381 struct ipv6_txoptions *opt); 382 383static inline bool ipv6_accept_ra(struct inet6_dev *idev) 384{ 385 /* If forwarding is enabled, RA are not accepted unless the special 386 * hybrid mode (accept_ra=2) is enabled. 387 */ 388 return idev->cnf.forwarding ? idev->cnf.accept_ra == 2 : 389 idev->cnf.accept_ra; 390} 391 392#define IPV6_FRAG_HIGH_THRESH (4 * 1024*1024) /* 4194304 */ 393#define IPV6_FRAG_LOW_THRESH (3 * 1024*1024) /* 3145728 */ 394#define IPV6_FRAG_TIMEOUT (60 * HZ) /* 60 seconds */ 395 396int __ipv6_addr_type(const struct in6_addr *addr); 397static inline int ipv6_addr_type(const struct in6_addr *addr) 398{ 399 return __ipv6_addr_type(addr) & 0xffff; 400} 401 402static inline int ipv6_addr_scope(const struct in6_addr *addr) 403{ 404 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK; 405} 406 407static inline int __ipv6_addr_src_scope(int type) 408{ 409 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16); 410} 411 412static inline int ipv6_addr_src_scope(const struct in6_addr *addr) 413{ 414 return __ipv6_addr_src_scope(__ipv6_addr_type(addr)); 415} 416 417static inline bool __ipv6_addr_needs_scope_id(int type) 418{ 419 return type & IPV6_ADDR_LINKLOCAL || 420 (type & IPV6_ADDR_MULTICAST && 421 (type & (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL))); 422} 423 424static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr, int iface) 425{ 426 return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0; 427} 428 429static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2) 430{ 431 return memcmp(a1, a2, sizeof(struct in6_addr)); 432} 433 434static inline bool 435ipv6_masked_addr_cmp(const struct in6_addr *a1, const struct in6_addr *m, 436 const struct in6_addr *a2) 437{ 438#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 439 const unsigned long *ul1 = (const unsigned long *)a1; 440 const unsigned long *ulm = (const unsigned long *)m; 441 const unsigned long *ul2 = (const unsigned long *)a2; 442 443 return !!(((ul1[0] ^ ul2[0]) & ulm[0]) | 444 ((ul1[1] ^ ul2[1]) & ulm[1])); 445#else 446 return !!(((a1->s6_addr32[0] ^ a2->s6_addr32[0]) & m->s6_addr32[0]) | 447 ((a1->s6_addr32[1] ^ a2->s6_addr32[1]) & m->s6_addr32[1]) | 448 ((a1->s6_addr32[2] ^ a2->s6_addr32[2]) & m->s6_addr32[2]) | 449 ((a1->s6_addr32[3] ^ a2->s6_addr32[3]) & m->s6_addr32[3])); 450#endif 451} 452 453static inline void ipv6_addr_prefix(struct in6_addr *pfx, 454 const struct in6_addr *addr, 455 int plen) 456{ 457 /* caller must guarantee 0 <= plen <= 128 */ 458 int o = plen >> 3, 459 b = plen & 0x7; 460 461 memset(pfx->s6_addr, 0, sizeof(pfx->s6_addr)); 462 memcpy(pfx->s6_addr, addr, o); 463 if (b != 0) 464 pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b); 465} 466 467static inline void ipv6_addr_prefix_copy(struct in6_addr *addr, 468 const struct in6_addr *pfx, 469 int plen) 470{ 471 /* caller must guarantee 0 <= plen <= 128 */ 472 int o = plen >> 3, 473 b = plen & 0x7; 474 475 memcpy(addr->s6_addr, pfx, o); 476 if (b != 0) { 477 addr->s6_addr[o] &= ~(0xff00 >> b); 478 addr->s6_addr[o] |= (pfx->s6_addr[o] & (0xff00 >> b)); 479 } 480} 481 482static inline void __ipv6_addr_set_half(__be32 *addr, 483 __be32 wh, __be32 wl) 484{ 485#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 486#if defined(__BIG_ENDIAN) 487 if (__builtin_constant_p(wh) && __builtin_constant_p(wl)) { 488 *(__force u64 *)addr = ((__force u64)(wh) << 32 | (__force u64)(wl)); 489 return; 490 } 491#elif defined(__LITTLE_ENDIAN) 492 if (__builtin_constant_p(wl) && __builtin_constant_p(wh)) { 493 *(__force u64 *)addr = ((__force u64)(wl) << 32 | (__force u64)(wh)); 494 return; 495 } 496#endif 497#endif 498 addr[0] = wh; 499 addr[1] = wl; 500} 501 502static inline void ipv6_addr_set(struct in6_addr *addr, 503 __be32 w1, __be32 w2, 504 __be32 w3, __be32 w4) 505{ 506 __ipv6_addr_set_half(&addr->s6_addr32[0], w1, w2); 507 __ipv6_addr_set_half(&addr->s6_addr32[2], w3, w4); 508} 509 510static inline bool ipv6_addr_equal(const struct in6_addr *a1, 511 const struct in6_addr *a2) 512{ 513#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 514 const unsigned long *ul1 = (const unsigned long *)a1; 515 const unsigned long *ul2 = (const unsigned long *)a2; 516 517 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL; 518#else 519 return ((a1->s6_addr32[0] ^ a2->s6_addr32[0]) | 520 (a1->s6_addr32[1] ^ a2->s6_addr32[1]) | 521 (a1->s6_addr32[2] ^ a2->s6_addr32[2]) | 522 (a1->s6_addr32[3] ^ a2->s6_addr32[3])) == 0; 523#endif 524} 525 526#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 527static inline bool __ipv6_prefix_equal64_half(const __be64 *a1, 528 const __be64 *a2, 529 unsigned int len) 530{ 531 if (len && ((*a1 ^ *a2) & cpu_to_be64((~0UL) << (64 - len)))) 532 return false; 533 return true; 534} 535 536static inline bool ipv6_prefix_equal(const struct in6_addr *addr1, 537 const struct in6_addr *addr2, 538 unsigned int prefixlen) 539{ 540 const __be64 *a1 = (const __be64 *)addr1; 541 const __be64 *a2 = (const __be64 *)addr2; 542 543 if (prefixlen >= 64) { 544 if (a1[0] ^ a2[0]) 545 return false; 546 return __ipv6_prefix_equal64_half(a1 + 1, a2 + 1, prefixlen - 64); 547 } 548 return __ipv6_prefix_equal64_half(a1, a2, prefixlen); 549} 550#else 551static inline bool ipv6_prefix_equal(const struct in6_addr *addr1, 552 const struct in6_addr *addr2, 553 unsigned int prefixlen) 554{ 555 const __be32 *a1 = addr1->s6_addr32; 556 const __be32 *a2 = addr2->s6_addr32; 557 unsigned int pdw, pbi; 558 559 /* check complete u32 in prefix */ 560 pdw = prefixlen >> 5; 561 if (pdw && memcmp(a1, a2, pdw << 2)) 562 return false; 563 564 /* check incomplete u32 in prefix */ 565 pbi = prefixlen & 0x1f; 566 if (pbi && ((a1[pdw] ^ a2[pdw]) & htonl((0xffffffff) << (32 - pbi)))) 567 return false; 568 569 return true; 570} 571#endif 572 573static inline bool ipv6_addr_any(const struct in6_addr *a) 574{ 575#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 576 const unsigned long *ul = (const unsigned long *)a; 577 578 return (ul[0] | ul[1]) == 0UL; 579#else 580 return (a->s6_addr32[0] | a->s6_addr32[1] | 581 a->s6_addr32[2] | a->s6_addr32[3]) == 0; 582#endif 583} 584 585static inline u32 ipv6_addr_hash(const struct in6_addr *a) 586{ 587#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 588 const unsigned long *ul = (const unsigned long *)a; 589 unsigned long x = ul[0] ^ ul[1]; 590 591 return (u32)(x ^ (x >> 32)); 592#else 593 return (__force u32)(a->s6_addr32[0] ^ a->s6_addr32[1] ^ 594 a->s6_addr32[2] ^ a->s6_addr32[3]); 595#endif 596} 597 598/* more secured version of ipv6_addr_hash() */ 599static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval) 600{ 601 u32 v = (__force u32)a->s6_addr32[0] ^ (__force u32)a->s6_addr32[1]; 602 603 return jhash_3words(v, 604 (__force u32)a->s6_addr32[2], 605 (__force u32)a->s6_addr32[3], 606 initval); 607} 608 609static inline bool ipv6_addr_loopback(const struct in6_addr *a) 610{ 611#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 612 const __be64 *be = (const __be64 *)a; 613 614 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL; 615#else 616 return (a->s6_addr32[0] | a->s6_addr32[1] | 617 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0; 618#endif 619} 620 621/* 622 * Note that we must __force cast these to unsigned long to make sparse happy, 623 * since all of the endian-annotated types are fixed size regardless of arch. 624 */ 625static inline bool ipv6_addr_v4mapped(const struct in6_addr *a) 626{ 627 return ( 628#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 629 *(unsigned long *)a | 630#else 631 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) | 632#endif 633 (__force unsigned long)(a->s6_addr32[2] ^ 634 cpu_to_be32(0x0000ffff))) == 0UL; 635} 636 637static inline u32 ipv6_portaddr_hash(const struct net *net, 638 const struct in6_addr *addr6, 639 unsigned int port) 640{ 641 unsigned int hash, mix = net_hash_mix(net); 642 643 if (ipv6_addr_any(addr6)) 644 hash = jhash_1word(0, mix); 645 else if (ipv6_addr_v4mapped(addr6)) 646 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix); 647 else 648 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix); 649 650 return hash ^ port; 651} 652 653/* 654 * Check for a RFC 4843 ORCHID address 655 * (Overlay Routable Cryptographic Hash Identifiers) 656 */ 657static inline bool ipv6_addr_orchid(const struct in6_addr *a) 658{ 659 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010); 660} 661 662static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr) 663{ 664 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000); 665} 666 667static inline void ipv6_addr_set_v4mapped(const __be32 addr, 668 struct in6_addr *v4mapped) 669{ 670 ipv6_addr_set(v4mapped, 671 0, 0, 672 htonl(0x0000FFFF), 673 addr); 674} 675 676/* 677 * find the first different bit between two addresses 678 * length of address must be a multiple of 32bits 679 */ 680static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen) 681{ 682 const __be32 *a1 = token1, *a2 = token2; 683 int i; 684 685 addrlen >>= 2; 686 687 for (i = 0; i < addrlen; i++) { 688 __be32 xb = a1[i] ^ a2[i]; 689 if (xb) 690 return i * 32 + 31 - __fls(ntohl(xb)); 691 } 692 693 /* 694 * we should *never* get to this point since that 695 * would mean the addrs are equal 696 * 697 * However, we do get to it 8) And exacly, when 698 * addresses are equal 8) 699 * 700 * ip route add 1111::/128 via ... 701 * ip route add 1111::/64 via ... 702 * and we are here. 703 * 704 * Ideally, this function should stop comparison 705 * at prefix length. It does not, but it is still OK, 706 * if returned value is greater than prefix length. 707 * --ANK (980803) 708 */ 709 return addrlen << 5; 710} 711 712#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 713static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen) 714{ 715 const __be64 *a1 = token1, *a2 = token2; 716 int i; 717 718 addrlen >>= 3; 719 720 for (i = 0; i < addrlen; i++) { 721 __be64 xb = a1[i] ^ a2[i]; 722 if (xb) 723 return i * 64 + 63 - __fls(be64_to_cpu(xb)); 724 } 725 726 return addrlen << 6; 727} 728#endif 729 730static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen) 731{ 732#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64 733 if (__builtin_constant_p(addrlen) && !(addrlen & 7)) 734 return __ipv6_addr_diff64(token1, token2, addrlen); 735#endif 736 return __ipv6_addr_diff32(token1, token2, addrlen); 737} 738 739static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2) 740{ 741 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr)); 742} 743 744__be32 ipv6_select_ident(struct net *net, 745 const struct in6_addr *daddr, 746 const struct in6_addr *saddr); 747__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb); 748 749int ip6_dst_hoplimit(struct dst_entry *dst); 750 751static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6, 752 struct dst_entry *dst) 753{ 754 int hlimit; 755 756 if (ipv6_addr_is_multicast(&fl6->daddr)) 757 hlimit = np->mcast_hops; 758 else 759 hlimit = np->hop_limit; 760 if (hlimit < 0) 761 hlimit = ip6_dst_hoplimit(dst); 762 return hlimit; 763} 764 765/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store 766 * Equivalent to : flow->v6addrs.src = iph->saddr; 767 * flow->v6addrs.dst = iph->daddr; 768 */ 769static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow, 770 const struct ipv6hdr *iph) 771{ 772 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) != 773 offsetof(typeof(flow->addrs), v6addrs.src) + 774 sizeof(flow->addrs.v6addrs.src)); 775 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs)); 776 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS; 777} 778 779#if IS_ENABLED(CONFIG_IPV6) 780 781static inline bool ipv6_can_nonlocal_bind(struct net *net, 782 struct inet_sock *inet) 783{ 784 return net->ipv6.sysctl.ip_nonlocal_bind || 785 inet->freebind || inet->transparent; 786} 787 788/* Sysctl settings for net ipv6.auto_flowlabels */ 789#define IP6_AUTO_FLOW_LABEL_OFF 0 790#define IP6_AUTO_FLOW_LABEL_OPTOUT 1 791#define IP6_AUTO_FLOW_LABEL_OPTIN 2 792#define IP6_AUTO_FLOW_LABEL_FORCED 3 793 794#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED 795 796#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT 797 798static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb, 799 __be32 flowlabel, bool autolabel, 800 struct flowi6 *fl6) 801{ 802 u32 hash; 803 804 /* @flowlabel may include more than a flow label, eg, the traffic class. 805 * Here we want only the flow label value. 806 */ 807 flowlabel &= IPV6_FLOWLABEL_MASK; 808 809 if (flowlabel || 810 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF || 811 (!autolabel && 812 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED)) 813 return flowlabel; 814 815 hash = skb_get_hash_flowi6(skb, fl6); 816 817 /* Since this is being sent on the wire obfuscate hash a bit 818 * to minimize possbility that any useful information to an 819 * attacker is leaked. Only lower 20 bits are relevant. 820 */ 821 hash = rol32(hash, 16); 822 823 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK; 824 825 if (net->ipv6.sysctl.flowlabel_state_ranges) 826 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG; 827 828 return flowlabel; 829} 830 831static inline int ip6_default_np_autolabel(struct net *net) 832{ 833 switch (net->ipv6.sysctl.auto_flowlabels) { 834 case IP6_AUTO_FLOW_LABEL_OFF: 835 case IP6_AUTO_FLOW_LABEL_OPTIN: 836 default: 837 return 0; 838 case IP6_AUTO_FLOW_LABEL_OPTOUT: 839 case IP6_AUTO_FLOW_LABEL_FORCED: 840 return 1; 841 } 842} 843#else 844static inline void ip6_set_txhash(struct sock *sk) { } 845static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb, 846 __be32 flowlabel, bool autolabel, 847 struct flowi6 *fl6) 848{ 849 return flowlabel; 850} 851static inline int ip6_default_np_autolabel(struct net *net) 852{ 853 return 0; 854} 855#endif 856 857#if IS_ENABLED(CONFIG_IPV6) 858static inline int ip6_multipath_hash_policy(const struct net *net) 859{ 860 return net->ipv6.sysctl.multipath_hash_policy; 861} 862#else 863static inline int ip6_multipath_hash_policy(const struct net *net) 864{ 865 return 0; 866} 867#endif 868 869/* 870 * Header manipulation 871 */ 872static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass, 873 __be32 flowlabel) 874{ 875 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel; 876} 877 878static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr) 879{ 880 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK; 881} 882 883static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr) 884{ 885 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK; 886} 887 888static inline u8 ip6_tclass(__be32 flowinfo) 889{ 890 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT; 891} 892 893static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel) 894{ 895 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel; 896} 897 898static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6) 899{ 900 return fl6->flowlabel & IPV6_FLOWLABEL_MASK; 901} 902 903/* 904 * Prototypes exported by ipv6 905 */ 906 907/* 908 * rcv function (called from netdevice level) 909 */ 910 911int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, 912 struct packet_type *pt, struct net_device *orig_dev); 913void ipv6_list_rcv(struct list_head *head, struct packet_type *pt, 914 struct net_device *orig_dev); 915 916int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb); 917 918/* 919 * upper-layer output functions 920 */ 921int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6, 922 __u32 mark, struct ipv6_txoptions *opt, int tclass); 923 924int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr); 925 926int ip6_append_data(struct sock *sk, 927 int getfrag(void *from, char *to, int offset, int len, 928 int odd, struct sk_buff *skb), 929 void *from, int length, int transhdrlen, 930 struct ipcm6_cookie *ipc6, struct flowi6 *fl6, 931 struct rt6_info *rt, unsigned int flags); 932 933int ip6_push_pending_frames(struct sock *sk); 934 935void ip6_flush_pending_frames(struct sock *sk); 936 937int ip6_send_skb(struct sk_buff *skb); 938 939struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue, 940 struct inet_cork_full *cork, 941 struct inet6_cork *v6_cork); 942struct sk_buff *ip6_make_skb(struct sock *sk, 943 int getfrag(void *from, char *to, int offset, 944 int len, int odd, struct sk_buff *skb), 945 void *from, int length, int transhdrlen, 946 struct ipcm6_cookie *ipc6, struct flowi6 *fl6, 947 struct rt6_info *rt, unsigned int flags, 948 struct inet_cork_full *cork); 949 950static inline struct sk_buff *ip6_finish_skb(struct sock *sk) 951{ 952 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork, 953 &inet6_sk(sk)->cork); 954} 955 956int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst, 957 struct flowi6 *fl6); 958struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6, 959 const struct in6_addr *final_dst); 960struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6, 961 const struct in6_addr *final_dst, 962 bool connected); 963struct dst_entry *ip6_blackhole_route(struct net *net, 964 struct dst_entry *orig_dst); 965 966/* 967 * skb processing functions 968 */ 969 970int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb); 971int ip6_forward(struct sk_buff *skb); 972int ip6_input(struct sk_buff *skb); 973int ip6_mc_input(struct sk_buff *skb); 974void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr, 975 bool have_final); 976 977int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb); 978int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb); 979 980/* 981 * Extension header (options) processing 982 */ 983 984void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, 985 u8 *proto, struct in6_addr **daddr_p, 986 struct in6_addr *saddr); 987void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt, 988 u8 *proto); 989 990int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp, 991 __be16 *frag_offp); 992 993bool ipv6_ext_hdr(u8 nexthdr); 994 995enum { 996 IP6_FH_F_FRAG = (1 << 0), 997 IP6_FH_F_AUTH = (1 << 1), 998 IP6_FH_F_SKIP_RH = (1 << 2), 999}; 1000 1001/* find specified header and get offset to it */ 1002int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target, 1003 unsigned short *fragoff, int *fragflg); 1004 1005int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type); 1006 1007struct in6_addr *fl6_update_dst(struct flowi6 *fl6, 1008 const struct ipv6_txoptions *opt, 1009 struct in6_addr *orig); 1010 1011/* 1012 * socket options (ipv6_sockglue.c) 1013 */ 1014 1015int ipv6_setsockopt(struct sock *sk, int level, int optname, 1016 char __user *optval, unsigned int optlen); 1017int ipv6_getsockopt(struct sock *sk, int level, int optname, 1018 char __user *optval, int __user *optlen); 1019int compat_ipv6_setsockopt(struct sock *sk, int level, int optname, 1020 char __user *optval, unsigned int optlen); 1021int compat_ipv6_getsockopt(struct sock *sk, int level, int optname, 1022 char __user *optval, int __user *optlen); 1023 1024int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, 1025 int addr_len); 1026int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len); 1027int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr, 1028 int addr_len); 1029int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr); 1030void ip6_datagram_release_cb(struct sock *sk); 1031 1032int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len, 1033 int *addr_len); 1034int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len, 1035 int *addr_len); 1036void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port, 1037 u32 info, u8 *payload); 1038void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info); 1039void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu); 1040 1041int inet6_release(struct socket *sock); 1042int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len); 1043int inet6_getname(struct socket *sock, struct sockaddr *uaddr, 1044 int peer); 1045int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg); 1046 1047int inet6_hash_connect(struct inet_timewait_death_row *death_row, 1048 struct sock *sk); 1049 1050/* 1051 * reassembly.c 1052 */ 1053extern const struct proto_ops inet6_stream_ops; 1054extern const struct proto_ops inet6_dgram_ops; 1055extern const struct proto_ops inet6_sockraw_ops; 1056 1057struct group_source_req; 1058struct group_filter; 1059 1060int ip6_mc_source(int add, int omode, struct sock *sk, 1061 struct group_source_req *pgsr); 1062int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf); 1063int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf, 1064 struct group_filter __user *optval, int __user *optlen); 1065 1066#ifdef CONFIG_PROC_FS 1067int ac6_proc_init(struct net *net); 1068void ac6_proc_exit(struct net *net); 1069int raw6_proc_init(void); 1070void raw6_proc_exit(void); 1071int tcp6_proc_init(struct net *net); 1072void tcp6_proc_exit(struct net *net); 1073int udp6_proc_init(struct net *net); 1074void udp6_proc_exit(struct net *net); 1075int udplite6_proc_init(void); 1076void udplite6_proc_exit(void); 1077int ipv6_misc_proc_init(void); 1078void ipv6_misc_proc_exit(void); 1079int snmp6_register_dev(struct inet6_dev *idev); 1080int snmp6_unregister_dev(struct inet6_dev *idev); 1081 1082#else 1083static inline int ac6_proc_init(struct net *net) { return 0; } 1084static inline void ac6_proc_exit(struct net *net) { } 1085static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; } 1086static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; } 1087#endif 1088 1089#ifdef CONFIG_SYSCTL 1090struct ctl_table *ipv6_icmp_sysctl_init(struct net *net); 1091struct ctl_table *ipv6_route_sysctl_init(struct net *net); 1092int ipv6_sysctl_register(void); 1093void ipv6_sysctl_unregister(void); 1094#endif 1095 1096int ipv6_sock_mc_join(struct sock *sk, int ifindex, 1097 const struct in6_addr *addr); 1098int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex, 1099 const struct in6_addr *addr, unsigned int mode); 1100int ipv6_sock_mc_drop(struct sock *sk, int ifindex, 1101 const struct in6_addr *addr); 1102#endif /* _NET_IPV6_H */