blob: 8c454509f299f26296f5e4979c33f13b0bf89893 [file] [log] [blame]
b.liue9582032025-04-17 19:18:16 +08001/* 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
144struct 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
154struct 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
163int ip6_fraglist_init(struct sk_buff *skb, unsigned int hlen, u8 *prevhdr,
164 u8 nexthdr, __be32 frag_id,
165 struct ip6_fraglist_iter *iter);
166void ip6_fraglist_prepare(struct sk_buff *skb, struct ip6_fraglist_iter *iter);
167
168static 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
178struct 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
191void 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);
194struct 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 */
203extern int sysctl_mld_max_msf;
204extern 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
272struct ip6_ra_chain {
273 struct ip6_ra_chain *next;
274 struct sock *sk;
275 int sel;
276 void (*destructor)(struct sock *);
277};
278
279extern struct ip6_ra_chain *ip6_ra_chain;
280extern 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
287struct 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 */
306enum flowlabel_reflect {
307 FLOWLABEL_REFLECT_ESTABLISHED = 1,
308 FLOWLABEL_REFLECT_TCP_RESET = 2,
309 FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES = 4,
310};
311
312struct 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
337struct ipv6_fl_socklist {
338 struct ipv6_fl_socklist __rcu *next;
339 struct ip6_flowlabel *fl;
340 struct rcu_head rcu;
341};
342
343struct 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
352static 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
361static 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
371static 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
387static 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
393struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label);
394
395extern struct static_key_false_deferred ipv6_flowlabel_exclusive;
396static 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
405struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space,
406 struct ip6_flowlabel *fl,
407 struct ipv6_txoptions *fopt);
408void fl6_free_socklist(struct sock *sk);
409int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen);
410int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
411 int flags);
412int ip6_flowlabel_init(void);
413void ip6_flowlabel_cleanup(void);
414bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np);
415
416static inline void fl6_sock_release(struct ip6_flowlabel *fl)
417{
418 if (fl)
419 atomic_dec(&fl->users);
420}
421
422void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info);
423
424void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
425 struct icmp6hdr *thdr, int len);
426
427int ip6_ra_control(struct sock *sk, int sel);
428
429int ipv6_parse_hopopts(struct sk_buff *skb);
430
431struct ipv6_txoptions *ipv6_dup_options(struct sock *sk,
432 struct ipv6_txoptions *opt);
433struct ipv6_txoptions *ipv6_renew_options(struct sock *sk,
434 struct ipv6_txoptions *opt,
435 int newtype,
436 struct ipv6_opt_hdr *newopt);
437struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
438 struct ipv6_txoptions *opt);
439
440bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
441 const struct inet6_skb_parm *opt);
442struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
443 struct ipv6_txoptions *opt);
444
445static 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
458int __ipv6_addr_type(const struct in6_addr *addr);
459static inline int ipv6_addr_type(const struct in6_addr *addr)
460{
461 return __ipv6_addr_type(addr) & 0xffff;
462}
463
464static inline int ipv6_addr_scope(const struct in6_addr *addr)
465{
466 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK;
467}
468
469static inline int __ipv6_addr_src_scope(int type)
470{
471 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16);
472}
473
474static inline int ipv6_addr_src_scope(const struct in6_addr *addr)
475{
476 return __ipv6_addr_src_scope(__ipv6_addr_type(addr));
477}
478
479static 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
486static 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
491static 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
496static inline bool
497ipv6_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
515static 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
529static 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
544static 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
564static 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
572static 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
589static 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
598static 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
613static 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
635static 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
647static 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() */
661static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval)
662{
663 return jhash2((__force const u32 *)a->s6_addr32,
664 ARRAY_SIZE(a->s6_addr32), initval);
665}
666
667static inline bool ipv6_addr_loopback(const struct in6_addr *a)
668{
669#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
670 const __be64 *be = (const __be64 *)a;
671
672 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL;
673#else
674 return (a->s6_addr32[0] | a->s6_addr32[1] |
675 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0;
676#endif
677}
678
679/*
680 * Note that we must __force cast these to unsigned long to make sparse happy,
681 * since all of the endian-annotated types are fixed size regardless of arch.
682 */
683static inline bool ipv6_addr_v4mapped(const struct in6_addr *a)
684{
685 return (
686#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
687 *(unsigned long *)a |
688#else
689 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) |
690#endif
691 (__force unsigned long)(a->s6_addr32[2] ^
692 cpu_to_be32(0x0000ffff))) == 0UL;
693}
694
695static inline u32 ipv6_portaddr_hash(const struct net *net,
696 const struct in6_addr *addr6,
697 unsigned int port)
698{
699 unsigned int hash, mix = net_hash_mix(net);
700
701 if (ipv6_addr_any(addr6))
702 hash = jhash_1word(0, mix);
703 else if (ipv6_addr_v4mapped(addr6))
704 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
705 else
706 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
707
708 return hash ^ port;
709}
710
711/*
712 * Check for a RFC 4843 ORCHID address
713 * (Overlay Routable Cryptographic Hash Identifiers)
714 */
715static inline bool ipv6_addr_orchid(const struct in6_addr *a)
716{
717 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010);
718}
719
720static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr)
721{
722 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000);
723}
724
725static inline void ipv6_addr_set_v4mapped(const __be32 addr,
726 struct in6_addr *v4mapped)
727{
728 ipv6_addr_set(v4mapped,
729 0, 0,
730 htonl(0x0000FFFF),
731 addr);
732}
733
734/*
735 * find the first different bit between two addresses
736 * length of address must be a multiple of 32bits
737 */
738static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen)
739{
740 const __be32 *a1 = token1, *a2 = token2;
741 int i;
742
743 addrlen >>= 2;
744
745 for (i = 0; i < addrlen; i++) {
746 __be32 xb = a1[i] ^ a2[i];
747 if (xb)
748 return i * 32 + 31 - __fls(ntohl(xb));
749 }
750
751 /*
752 * we should *never* get to this point since that
753 * would mean the addrs are equal
754 *
755 * However, we do get to it 8) And exacly, when
756 * addresses are equal 8)
757 *
758 * ip route add 1111::/128 via ...
759 * ip route add 1111::/64 via ...
760 * and we are here.
761 *
762 * Ideally, this function should stop comparison
763 * at prefix length. It does not, but it is still OK,
764 * if returned value is greater than prefix length.
765 * --ANK (980803)
766 */
767 return addrlen << 5;
768}
769
770#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
771static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen)
772{
773 const __be64 *a1 = token1, *a2 = token2;
774 int i;
775
776 addrlen >>= 3;
777
778 for (i = 0; i < addrlen; i++) {
779 __be64 xb = a1[i] ^ a2[i];
780 if (xb)
781 return i * 64 + 63 - __fls(be64_to_cpu(xb));
782 }
783
784 return addrlen << 6;
785}
786#endif
787
788static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen)
789{
790#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
791 if (__builtin_constant_p(addrlen) && !(addrlen & 7))
792 return __ipv6_addr_diff64(token1, token2, addrlen);
793#endif
794 return __ipv6_addr_diff32(token1, token2, addrlen);
795}
796
797static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2)
798{
799 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr));
800}
801
802__be32 ipv6_select_ident(struct net *net,
803 const struct in6_addr *daddr,
804 const struct in6_addr *saddr);
805__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb);
806
807int ip6_dst_hoplimit(struct dst_entry *dst);
808
809static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
810 struct dst_entry *dst)
811{
812 int hlimit;
813
814 if (ipv6_addr_is_multicast(&fl6->daddr))
815 hlimit = np->mcast_hops;
816 else
817 hlimit = np->hop_limit;
818 if (hlimit < 0)
819 hlimit = ip6_dst_hoplimit(dst);
820 return hlimit;
821}
822
823/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store
824 * Equivalent to : flow->v6addrs.src = iph->saddr;
825 * flow->v6addrs.dst = iph->daddr;
826 */
827static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow,
828 const struct ipv6hdr *iph)
829{
830 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) !=
831 offsetof(typeof(flow->addrs), v6addrs.src) +
832 sizeof(flow->addrs.v6addrs.src));
833 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs));
834 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
835}
836
837#if IS_ENABLED(CONFIG_IPV6)
838
839static inline bool ipv6_can_nonlocal_bind(struct net *net,
840 struct inet_sock *inet)
841{
842 return net->ipv6.sysctl.ip_nonlocal_bind ||
843 inet->freebind || inet->transparent;
844}
845
846/* Sysctl settings for net ipv6.auto_flowlabels */
847#define IP6_AUTO_FLOW_LABEL_OFF 0
848#define IP6_AUTO_FLOW_LABEL_OPTOUT 1
849#define IP6_AUTO_FLOW_LABEL_OPTIN 2
850#define IP6_AUTO_FLOW_LABEL_FORCED 3
851
852#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED
853
854#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT
855
856static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
857 __be32 flowlabel, bool autolabel,
858 struct flowi6 *fl6)
859{
860 u32 hash;
861
862 /* @flowlabel may include more than a flow label, eg, the traffic class.
863 * Here we want only the flow label value.
864 */
865 flowlabel &= IPV6_FLOWLABEL_MASK;
866
867 if (flowlabel ||
868 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF ||
869 (!autolabel &&
870 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED))
871 return flowlabel;
872
873 hash = skb_get_hash_flowi6(skb, fl6);
874
875 /* Since this is being sent on the wire obfuscate hash a bit
876 * to minimize possbility that any useful information to an
877 * attacker is leaked. Only lower 20 bits are relevant.
878 */
879 hash = rol32(hash, 16);
880
881 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
882
883 if (net->ipv6.sysctl.flowlabel_state_ranges)
884 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG;
885
886 return flowlabel;
887}
888
889static inline int ip6_default_np_autolabel(struct net *net)
890{
891 switch (net->ipv6.sysctl.auto_flowlabels) {
892 case IP6_AUTO_FLOW_LABEL_OFF:
893 case IP6_AUTO_FLOW_LABEL_OPTIN:
894 default:
895 return 0;
896 case IP6_AUTO_FLOW_LABEL_OPTOUT:
897 case IP6_AUTO_FLOW_LABEL_FORCED:
898 return 1;
899 }
900}
901#else
902static inline void ip6_set_txhash(struct sock *sk) { }
903static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
904 __be32 flowlabel, bool autolabel,
905 struct flowi6 *fl6)
906{
907 return flowlabel;
908}
909static inline int ip6_default_np_autolabel(struct net *net)
910{
911 return 0;
912}
913#endif
914
915#if IS_ENABLED(CONFIG_IPV6)
916static inline int ip6_multipath_hash_policy(const struct net *net)
917{
918 return net->ipv6.sysctl.multipath_hash_policy;
919}
920#else
921static inline int ip6_multipath_hash_policy(const struct net *net)
922{
923 return 0;
924}
925#endif
926
927/*
928 * Header manipulation
929 */
930static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass,
931 __be32 flowlabel)
932{
933 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel;
934}
935
936static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr)
937{
938 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK;
939}
940
941static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr)
942{
943 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK;
944}
945
946static inline u8 ip6_tclass(__be32 flowinfo)
947{
948 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT;
949}
950
951static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel)
952{
953 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel;
954}
955
956static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6)
957{
958 return fl6->flowlabel & IPV6_FLOWLABEL_MASK;
959}
960
961/*
962 * Prototypes exported by ipv6
963 */
964
965/*
966 * rcv function (called from netdevice level)
967 */
968
969int ipv6_rcv(struct sk_buff *skb, struct net_device *dev,
970 struct packet_type *pt, struct net_device *orig_dev);
971void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
972 struct net_device *orig_dev);
973
974int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb);
975
976/*
977 * upper-layer output functions
978 */
979int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
980 __u32 mark, struct ipv6_txoptions *opt, int tclass, u32 priority);
981
982int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr);
983
984int ip6_append_data(struct sock *sk,
985 int getfrag(void *from, char *to, int offset, int len,
986 int odd, struct sk_buff *skb),
987 void *from, int length, int transhdrlen,
988 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
989 struct rt6_info *rt, unsigned int flags);
990
991int ip6_push_pending_frames(struct sock *sk);
992
993void ip6_flush_pending_frames(struct sock *sk);
994
995int ip6_send_skb(struct sk_buff *skb);
996
997struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue,
998 struct inet_cork_full *cork,
999 struct inet6_cork *v6_cork);
1000struct sk_buff *ip6_make_skb(struct sock *sk,
1001 int getfrag(void *from, char *to, int offset,
1002 int len, int odd, struct sk_buff *skb),
1003 void *from, int length, int transhdrlen,
1004 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1005 struct rt6_info *rt, unsigned int flags,
1006 struct inet_cork_full *cork);
1007
1008static inline struct sk_buff *ip6_finish_skb(struct sock *sk)
1009{
1010 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork,
1011 &inet6_sk(sk)->cork);
1012}
1013
1014int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1015 struct flowi6 *fl6);
1016struct dst_entry *ip6_dst_lookup_flow(struct net *net, const struct sock *sk, struct flowi6 *fl6,
1017 const struct in6_addr *final_dst);
1018struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
1019 const struct in6_addr *final_dst,
1020 bool connected);
1021struct dst_entry *ip6_blackhole_route(struct net *net,
1022 struct dst_entry *orig_dst);
1023
1024/*
1025 * skb processing functions
1026 */
1027
1028int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb);
1029int ip6_forward(struct sk_buff *skb);
1030int ip6_input(struct sk_buff *skb);
1031int ip6_mc_input(struct sk_buff *skb);
1032void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
1033 bool have_final);
1034
1035int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1036int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1037
1038/*
1039 * Extension header (options) processing
1040 */
1041
1042void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1043 u8 *proto, struct in6_addr **daddr_p,
1044 struct in6_addr *saddr);
1045void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1046 u8 *proto);
1047
1048int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp,
1049 __be16 *frag_offp);
1050
1051bool ipv6_ext_hdr(u8 nexthdr);
1052
1053enum {
1054 IP6_FH_F_FRAG = (1 << 0),
1055 IP6_FH_F_AUTH = (1 << 1),
1056 IP6_FH_F_SKIP_RH = (1 << 2),
1057};
1058
1059/* find specified header and get offset to it */
1060int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target,
1061 unsigned short *fragoff, int *fragflg);
1062
1063int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type);
1064
1065struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1066 const struct ipv6_txoptions *opt,
1067 struct in6_addr *orig);
1068
1069/*
1070 * socket options (ipv6_sockglue.c)
1071 */
1072
1073int ipv6_setsockopt(struct sock *sk, int level, int optname,
1074 char __user *optval, unsigned int optlen);
1075int ipv6_getsockopt(struct sock *sk, int level, int optname,
1076 char __user *optval, int __user *optlen);
1077int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
1078 char __user *optval, unsigned int optlen);
1079int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1080 char __user *optval, int __user *optlen);
1081
1082int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr,
1083 int addr_len);
1084int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len);
1085int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr,
1086 int addr_len);
1087int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr);
1088void ip6_datagram_release_cb(struct sock *sk);
1089
1090int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len,
1091 int *addr_len);
1092int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
1093 int *addr_len);
1094void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
1095 u32 info, u8 *payload);
1096void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info);
1097void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu);
1098
1099void inet6_cleanup_sock(struct sock *sk);
1100void inet6_sock_destruct(struct sock *sk);
1101int inet6_release(struct socket *sock);
1102int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1103int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
1104 int peer);
1105int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg);
1106
1107int inet6_hash_connect(struct inet_timewait_death_row *death_row,
1108 struct sock *sk);
1109
1110/*
1111 * reassembly.c
1112 */
1113extern const struct proto_ops inet6_stream_ops;
1114extern const struct proto_ops inet6_dgram_ops;
1115extern const struct proto_ops inet6_sockraw_ops;
1116
1117struct group_source_req;
1118struct group_filter;
1119
1120int ip6_mc_source(int add, int omode, struct sock *sk,
1121 struct group_source_req *pgsr);
1122int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf);
1123int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
1124 struct group_filter __user *optval, int __user *optlen);
1125
1126#ifdef CONFIG_PROC_FS
1127int ac6_proc_init(struct net *net);
1128void ac6_proc_exit(struct net *net);
1129int raw6_proc_init(void);
1130void raw6_proc_exit(void);
1131int tcp6_proc_init(struct net *net);
1132void tcp6_proc_exit(struct net *net);
1133int udp6_proc_init(struct net *net);
1134void udp6_proc_exit(struct net *net);
1135int udplite6_proc_init(void);
1136void udplite6_proc_exit(void);
1137int ipv6_misc_proc_init(void);
1138void ipv6_misc_proc_exit(void);
1139int snmp6_register_dev(struct inet6_dev *idev);
1140int snmp6_unregister_dev(struct inet6_dev *idev);
1141
1142#else
1143static inline int ac6_proc_init(struct net *net) { return 0; }
1144static inline void ac6_proc_exit(struct net *net) { }
1145static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; }
1146static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; }
1147#endif
1148
1149#ifdef CONFIG_SYSCTL
1150struct ctl_table *ipv6_icmp_sysctl_init(struct net *net);
1151struct ctl_table *ipv6_route_sysctl_init(struct net *net);
1152int ipv6_sysctl_register(void);
1153void ipv6_sysctl_unregister(void);
1154#endif
1155
1156int ipv6_sock_mc_join(struct sock *sk, int ifindex,
1157 const struct in6_addr *addr);
1158int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
1159 const struct in6_addr *addr, unsigned int mode);
1160int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
1161 const struct in6_addr *addr);
1162#endif /* _NET_IPV6_H */