Loading...
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_IPV4 4 /* IPv4 in IPv6 */
34#define NEXTHDR_TCP 6 /* TCP segment. */
35#define NEXTHDR_UDP 17 /* UDP message. */
36#define NEXTHDR_IPV6 41 /* IPv6 in IPv6 */
37#define NEXTHDR_ROUTING 43 /* Routing header. */
38#define NEXTHDR_FRAGMENT 44 /* Fragmentation/reassembly header. */
39#define NEXTHDR_GRE 47 /* GRE header. */
40#define NEXTHDR_ESP 50 /* Encapsulating security payload. */
41#define NEXTHDR_AUTH 51 /* Authentication header. */
42#define NEXTHDR_ICMP 58 /* ICMP for IPv6. */
43#define NEXTHDR_NONE 59 /* No next header */
44#define NEXTHDR_DEST 60 /* Destination options header. */
45#define NEXTHDR_SCTP 132 /* SCTP message. */
46#define NEXTHDR_MOBILITY 135 /* Mobility header. */
47
48#define NEXTHDR_MAX 255
49
50#define IPV6_DEFAULT_HOPLIMIT 64
51#define IPV6_DEFAULT_MCASTHOPS 1
52
53/* Limits on Hop-by-Hop and Destination options.
54 *
55 * Per RFC8200 there is no limit on the maximum number or lengths of options in
56 * Hop-by-Hop or Destination options other then the packet must fit in an MTU.
57 * We allow configurable limits in order to mitigate potential denial of
58 * service attacks.
59 *
60 * There are three limits that may be set:
61 * - Limit the number of options in a Hop-by-Hop or Destination options
62 * extension header
63 * - Limit the byte length of a Hop-by-Hop or Destination options extension
64 * header
65 * - Disallow unknown options
66 *
67 * The limits are expressed in corresponding sysctls:
68 *
69 * ipv6.sysctl.max_dst_opts_cnt
70 * ipv6.sysctl.max_hbh_opts_cnt
71 * ipv6.sysctl.max_dst_opts_len
72 * ipv6.sysctl.max_hbh_opts_len
73 *
74 * max_*_opts_cnt is the number of TLVs that are allowed for Destination
75 * options or Hop-by-Hop options. If the number is less than zero then unknown
76 * TLVs are disallowed and the number of known options that are allowed is the
77 * absolute value. Setting the value to INT_MAX indicates no limit.
78 *
79 * max_*_opts_len is the length limit in bytes of a Destination or
80 * Hop-by-Hop options extension header. Setting the value to INT_MAX
81 * indicates no length limit.
82 *
83 * If a limit is exceeded when processing an extension header the packet is
84 * silently discarded.
85 */
86
87/* Default limits for Hop-by-Hop and Destination options */
88#define IP6_DEFAULT_MAX_DST_OPTS_CNT 8
89#define IP6_DEFAULT_MAX_HBH_OPTS_CNT 8
90#define IP6_DEFAULT_MAX_DST_OPTS_LEN INT_MAX /* No limit */
91#define IP6_DEFAULT_MAX_HBH_OPTS_LEN INT_MAX /* No limit */
92
93/*
94 * Addr type
95 *
96 * type - unicast | multicast
97 * scope - local | site | global
98 * v4 - compat
99 * v4mapped
100 * any
101 * loopback
102 */
103
104#define IPV6_ADDR_ANY 0x0000U
105
106#define IPV6_ADDR_UNICAST 0x0001U
107#define IPV6_ADDR_MULTICAST 0x0002U
108
109#define IPV6_ADDR_LOOPBACK 0x0010U
110#define IPV6_ADDR_LINKLOCAL 0x0020U
111#define IPV6_ADDR_SITELOCAL 0x0040U
112
113#define IPV6_ADDR_COMPATv4 0x0080U
114
115#define IPV6_ADDR_SCOPE_MASK 0x00f0U
116
117#define IPV6_ADDR_MAPPED 0x1000U
118
119/*
120 * Addr scopes
121 */
122#define IPV6_ADDR_MC_SCOPE(a) \
123 ((a)->s6_addr[1] & 0x0f) /* nonstandard */
124#define __IPV6_ADDR_SCOPE_INVALID -1
125#define IPV6_ADDR_SCOPE_NODELOCAL 0x01
126#define IPV6_ADDR_SCOPE_LINKLOCAL 0x02
127#define IPV6_ADDR_SCOPE_SITELOCAL 0x05
128#define IPV6_ADDR_SCOPE_ORGLOCAL 0x08
129#define IPV6_ADDR_SCOPE_GLOBAL 0x0e
130
131/*
132 * Addr flags
133 */
134#define IPV6_ADDR_MC_FLAG_TRANSIENT(a) \
135 ((a)->s6_addr[1] & 0x10)
136#define IPV6_ADDR_MC_FLAG_PREFIX(a) \
137 ((a)->s6_addr[1] & 0x20)
138#define IPV6_ADDR_MC_FLAG_RENDEZVOUS(a) \
139 ((a)->s6_addr[1] & 0x40)
140
141/*
142 * fragmentation header
143 */
144
145struct frag_hdr {
146 __u8 nexthdr;
147 __u8 reserved;
148 __be16 frag_off;
149 __be32 identification;
150};
151
152#define IP6_MF 0x0001
153#define IP6_OFFSET 0xFFF8
154
155struct ip6_fraglist_iter {
156 struct ipv6hdr *tmp_hdr;
157 struct sk_buff *frag;
158 int offset;
159 unsigned int hlen;
160 __be32 frag_id;
161 u8 nexthdr;
162};
163
164int ip6_fraglist_init(struct sk_buff *skb, unsigned int hlen, u8 *prevhdr,
165 u8 nexthdr, __be32 frag_id,
166 struct ip6_fraglist_iter *iter);
167void ip6_fraglist_prepare(struct sk_buff *skb, struct ip6_fraglist_iter *iter);
168
169static inline struct sk_buff *ip6_fraglist_next(struct ip6_fraglist_iter *iter)
170{
171 struct sk_buff *skb = iter->frag;
172
173 iter->frag = skb->next;
174 skb_mark_not_on_list(skb);
175
176 return skb;
177}
178
179struct ip6_frag_state {
180 u8 *prevhdr;
181 unsigned int hlen;
182 unsigned int mtu;
183 unsigned int left;
184 int offset;
185 int ptr;
186 int hroom;
187 int troom;
188 __be32 frag_id;
189 u8 nexthdr;
190};
191
192void ip6_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int mtu,
193 unsigned short needed_tailroom, int hdr_room, u8 *prevhdr,
194 u8 nexthdr, __be32 frag_id, struct ip6_frag_state *state);
195struct sk_buff *ip6_frag_next(struct sk_buff *skb,
196 struct ip6_frag_state *state);
197
198#define IP6_REPLY_MARK(net, mark) \
199 ((net)->ipv6.sysctl.fwmark_reflect ? (mark) : 0)
200
201#include <net/sock.h>
202
203/* sysctls */
204extern int sysctl_mld_max_msf;
205extern int sysctl_mld_qrv;
206
207#define _DEVINC(net, statname, mod, idev, field) \
208({ \
209 struct inet6_dev *_idev = (idev); \
210 if (likely(_idev != NULL)) \
211 mod##SNMP_INC_STATS64((_idev)->stats.statname, (field));\
212 mod##SNMP_INC_STATS64((net)->mib.statname##_statistics, (field));\
213})
214
215/* per device counters are atomic_long_t */
216#define _DEVINCATOMIC(net, statname, mod, idev, field) \
217({ \
218 struct inet6_dev *_idev = (idev); \
219 if (likely(_idev != NULL)) \
220 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
221 mod##SNMP_INC_STATS((net)->mib.statname##_statistics, (field));\
222})
223
224/* per device and per net counters are atomic_long_t */
225#define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \
226({ \
227 struct inet6_dev *_idev = (idev); \
228 if (likely(_idev != NULL)) \
229 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
230 SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\
231})
232
233#define _DEVADD(net, statname, mod, idev, field, val) \
234({ \
235 struct inet6_dev *_idev = (idev); \
236 if (likely(_idev != NULL)) \
237 mod##SNMP_ADD_STATS((_idev)->stats.statname, (field), (val)); \
238 mod##SNMP_ADD_STATS((net)->mib.statname##_statistics, (field), (val));\
239})
240
241#define _DEVUPD(net, statname, mod, idev, field, val) \
242({ \
243 struct inet6_dev *_idev = (idev); \
244 if (likely(_idev != NULL)) \
245 mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \
246 mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
247})
248
249/* MIBs */
250
251#define IP6_INC_STATS(net, idev,field) \
252 _DEVINC(net, ipv6, , idev, field)
253#define __IP6_INC_STATS(net, idev,field) \
254 _DEVINC(net, ipv6, __, idev, field)
255#define IP6_ADD_STATS(net, idev,field,val) \
256 _DEVADD(net, ipv6, , idev, field, val)
257#define __IP6_ADD_STATS(net, idev,field,val) \
258 _DEVADD(net, ipv6, __, idev, field, val)
259#define IP6_UPD_PO_STATS(net, idev,field,val) \
260 _DEVUPD(net, ipv6, , idev, field, val)
261#define __IP6_UPD_PO_STATS(net, idev,field,val) \
262 _DEVUPD(net, ipv6, __, idev, field, val)
263#define ICMP6_INC_STATS(net, idev, field) \
264 _DEVINCATOMIC(net, icmpv6, , idev, field)
265#define __ICMP6_INC_STATS(net, idev, field) \
266 _DEVINCATOMIC(net, icmpv6, __, idev, field)
267
268#define ICMP6MSGOUT_INC_STATS(net, idev, field) \
269 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256)
270#define ICMP6MSGIN_INC_STATS(net, idev, field) \
271 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field)
272
273struct ip6_ra_chain {
274 struct ip6_ra_chain *next;
275 struct sock *sk;
276 int sel;
277 void (*destructor)(struct sock *);
278};
279
280extern struct ip6_ra_chain *ip6_ra_chain;
281extern rwlock_t ip6_ra_lock;
282
283/*
284 This structure is prepared by protocol, when parsing
285 ancillary data and passed to IPv6.
286 */
287
288struct ipv6_txoptions {
289 refcount_t refcnt;
290 /* Length of this structure */
291 int tot_len;
292
293 /* length of extension headers */
294
295 __u16 opt_flen; /* after fragment hdr */
296 __u16 opt_nflen; /* before fragment hdr */
297
298 struct ipv6_opt_hdr *hopopt;
299 struct ipv6_opt_hdr *dst0opt;
300 struct ipv6_rt_hdr *srcrt; /* Routing Header */
301 struct ipv6_opt_hdr *dst1opt;
302 struct rcu_head rcu;
303 /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */
304};
305
306/* flowlabel_reflect sysctl values */
307enum flowlabel_reflect {
308 FLOWLABEL_REFLECT_ESTABLISHED = 1,
309 FLOWLABEL_REFLECT_TCP_RESET = 2,
310 FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES = 4,
311};
312
313struct ip6_flowlabel {
314 struct ip6_flowlabel __rcu *next;
315 __be32 label;
316 atomic_t users;
317 struct in6_addr dst;
318 struct ipv6_txoptions *opt;
319 unsigned long linger;
320 struct rcu_head rcu;
321 u8 share;
322 union {
323 struct pid *pid;
324 kuid_t uid;
325 } owner;
326 unsigned long lastuse;
327 unsigned long expires;
328 struct net *fl_net;
329};
330
331#define IPV6_FLOWINFO_MASK cpu_to_be32(0x0FFFFFFF)
332#define IPV6_FLOWLABEL_MASK cpu_to_be32(0x000FFFFF)
333#define IPV6_FLOWLABEL_STATELESS_FLAG cpu_to_be32(0x00080000)
334
335#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
336#define IPV6_TCLASS_SHIFT 20
337
338struct ipv6_fl_socklist {
339 struct ipv6_fl_socklist __rcu *next;
340 struct ip6_flowlabel *fl;
341 struct rcu_head rcu;
342};
343
344struct ipcm6_cookie {
345 struct sockcm_cookie sockc;
346 __s16 hlimit;
347 __s16 tclass;
348 __s8 dontfrag;
349 struct ipv6_txoptions *opt;
350 __u16 gso_size;
351};
352
353static inline void ipcm6_init(struct ipcm6_cookie *ipc6)
354{
355 *ipc6 = (struct ipcm6_cookie) {
356 .hlimit = -1,
357 .tclass = -1,
358 .dontfrag = -1,
359 };
360}
361
362static inline void ipcm6_init_sk(struct ipcm6_cookie *ipc6,
363 const struct ipv6_pinfo *np)
364{
365 *ipc6 = (struct ipcm6_cookie) {
366 .hlimit = -1,
367 .tclass = np->tclass,
368 .dontfrag = np->dontfrag,
369 };
370}
371
372static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np)
373{
374 struct ipv6_txoptions *opt;
375
376 rcu_read_lock();
377 opt = rcu_dereference(np->opt);
378 if (opt) {
379 if (!refcount_inc_not_zero(&opt->refcnt))
380 opt = NULL;
381 else
382 opt = rcu_pointer_handoff(opt);
383 }
384 rcu_read_unlock();
385 return opt;
386}
387
388static inline void txopt_put(struct ipv6_txoptions *opt)
389{
390 if (opt && refcount_dec_and_test(&opt->refcnt))
391 kfree_rcu(opt, rcu);
392}
393
394struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label);
395
396extern struct static_key_false_deferred ipv6_flowlabel_exclusive;
397static inline struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk,
398 __be32 label)
399{
400 if (static_branch_unlikely(&ipv6_flowlabel_exclusive.key))
401 return __fl6_sock_lookup(sk, label) ? : ERR_PTR(-ENOENT);
402
403 return NULL;
404}
405
406struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space,
407 struct ip6_flowlabel *fl,
408 struct ipv6_txoptions *fopt);
409void fl6_free_socklist(struct sock *sk);
410int ipv6_flowlabel_opt(struct sock *sk, sockptr_t optval, int optlen);
411int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
412 int flags);
413int ip6_flowlabel_init(void);
414void ip6_flowlabel_cleanup(void);
415bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np);
416
417static inline void fl6_sock_release(struct ip6_flowlabel *fl)
418{
419 if (fl)
420 atomic_dec(&fl->users);
421}
422
423void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info);
424
425void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
426 struct icmp6hdr *thdr, int len);
427
428int ip6_ra_control(struct sock *sk, int sel);
429
430int ipv6_parse_hopopts(struct sk_buff *skb);
431
432struct ipv6_txoptions *ipv6_dup_options(struct sock *sk,
433 struct ipv6_txoptions *opt);
434struct ipv6_txoptions *ipv6_renew_options(struct sock *sk,
435 struct ipv6_txoptions *opt,
436 int newtype,
437 struct ipv6_opt_hdr *newopt);
438struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
439 struct ipv6_txoptions *opt);
440
441bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
442 const struct inet6_skb_parm *opt);
443struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
444 struct ipv6_txoptions *opt);
445
446static inline bool ipv6_accept_ra(struct inet6_dev *idev)
447{
448 /* If forwarding is enabled, RA are not accepted unless the special
449 * hybrid mode (accept_ra=2) is enabled.
450 */
451 return idev->cnf.forwarding ? idev->cnf.accept_ra == 2 :
452 idev->cnf.accept_ra;
453}
454
455#define IPV6_FRAG_HIGH_THRESH (4 * 1024*1024) /* 4194304 */
456#define IPV6_FRAG_LOW_THRESH (3 * 1024*1024) /* 3145728 */
457#define IPV6_FRAG_TIMEOUT (60 * HZ) /* 60 seconds */
458
459int __ipv6_addr_type(const struct in6_addr *addr);
460static inline int ipv6_addr_type(const struct in6_addr *addr)
461{
462 return __ipv6_addr_type(addr) & 0xffff;
463}
464
465static inline int ipv6_addr_scope(const struct in6_addr *addr)
466{
467 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK;
468}
469
470static inline int __ipv6_addr_src_scope(int type)
471{
472 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16);
473}
474
475static inline int ipv6_addr_src_scope(const struct in6_addr *addr)
476{
477 return __ipv6_addr_src_scope(__ipv6_addr_type(addr));
478}
479
480static inline bool __ipv6_addr_needs_scope_id(int type)
481{
482 return type & IPV6_ADDR_LINKLOCAL ||
483 (type & IPV6_ADDR_MULTICAST &&
484 (type & (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)));
485}
486
487static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr, int iface)
488{
489 return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0;
490}
491
492static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2)
493{
494 return memcmp(a1, a2, sizeof(struct in6_addr));
495}
496
497static inline bool
498ipv6_masked_addr_cmp(const struct in6_addr *a1, const struct in6_addr *m,
499 const struct in6_addr *a2)
500{
501#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
502 const unsigned long *ul1 = (const unsigned long *)a1;
503 const unsigned long *ulm = (const unsigned long *)m;
504 const unsigned long *ul2 = (const unsigned long *)a2;
505
506 return !!(((ul1[0] ^ ul2[0]) & ulm[0]) |
507 ((ul1[1] ^ ul2[1]) & ulm[1]));
508#else
509 return !!(((a1->s6_addr32[0] ^ a2->s6_addr32[0]) & m->s6_addr32[0]) |
510 ((a1->s6_addr32[1] ^ a2->s6_addr32[1]) & m->s6_addr32[1]) |
511 ((a1->s6_addr32[2] ^ a2->s6_addr32[2]) & m->s6_addr32[2]) |
512 ((a1->s6_addr32[3] ^ a2->s6_addr32[3]) & m->s6_addr32[3]));
513#endif
514}
515
516static inline void ipv6_addr_prefix(struct in6_addr *pfx,
517 const struct in6_addr *addr,
518 int plen)
519{
520 /* caller must guarantee 0 <= plen <= 128 */
521 int o = plen >> 3,
522 b = plen & 0x7;
523
524 memset(pfx->s6_addr, 0, sizeof(pfx->s6_addr));
525 memcpy(pfx->s6_addr, addr, o);
526 if (b != 0)
527 pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b);
528}
529
530static inline void ipv6_addr_prefix_copy(struct in6_addr *addr,
531 const struct in6_addr *pfx,
532 int plen)
533{
534 /* caller must guarantee 0 <= plen <= 128 */
535 int o = plen >> 3,
536 b = plen & 0x7;
537
538 memcpy(addr->s6_addr, pfx, o);
539 if (b != 0) {
540 addr->s6_addr[o] &= ~(0xff00 >> b);
541 addr->s6_addr[o] |= (pfx->s6_addr[o] & (0xff00 >> b));
542 }
543}
544
545static inline void __ipv6_addr_set_half(__be32 *addr,
546 __be32 wh, __be32 wl)
547{
548#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
549#if defined(__BIG_ENDIAN)
550 if (__builtin_constant_p(wh) && __builtin_constant_p(wl)) {
551 *(__force u64 *)addr = ((__force u64)(wh) << 32 | (__force u64)(wl));
552 return;
553 }
554#elif defined(__LITTLE_ENDIAN)
555 if (__builtin_constant_p(wl) && __builtin_constant_p(wh)) {
556 *(__force u64 *)addr = ((__force u64)(wl) << 32 | (__force u64)(wh));
557 return;
558 }
559#endif
560#endif
561 addr[0] = wh;
562 addr[1] = wl;
563}
564
565static inline void ipv6_addr_set(struct in6_addr *addr,
566 __be32 w1, __be32 w2,
567 __be32 w3, __be32 w4)
568{
569 __ipv6_addr_set_half(&addr->s6_addr32[0], w1, w2);
570 __ipv6_addr_set_half(&addr->s6_addr32[2], w3, w4);
571}
572
573static inline bool ipv6_addr_equal(const struct in6_addr *a1,
574 const struct in6_addr *a2)
575{
576#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
577 const unsigned long *ul1 = (const unsigned long *)a1;
578 const unsigned long *ul2 = (const unsigned long *)a2;
579
580 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL;
581#else
582 return ((a1->s6_addr32[0] ^ a2->s6_addr32[0]) |
583 (a1->s6_addr32[1] ^ a2->s6_addr32[1]) |
584 (a1->s6_addr32[2] ^ a2->s6_addr32[2]) |
585 (a1->s6_addr32[3] ^ a2->s6_addr32[3])) == 0;
586#endif
587}
588
589#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
590static inline bool __ipv6_prefix_equal64_half(const __be64 *a1,
591 const __be64 *a2,
592 unsigned int len)
593{
594 if (len && ((*a1 ^ *a2) & cpu_to_be64((~0UL) << (64 - len))))
595 return false;
596 return true;
597}
598
599static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
600 const struct in6_addr *addr2,
601 unsigned int prefixlen)
602{
603 const __be64 *a1 = (const __be64 *)addr1;
604 const __be64 *a2 = (const __be64 *)addr2;
605
606 if (prefixlen >= 64) {
607 if (a1[0] ^ a2[0])
608 return false;
609 return __ipv6_prefix_equal64_half(a1 + 1, a2 + 1, prefixlen - 64);
610 }
611 return __ipv6_prefix_equal64_half(a1, a2, prefixlen);
612}
613#else
614static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
615 const struct in6_addr *addr2,
616 unsigned int prefixlen)
617{
618 const __be32 *a1 = addr1->s6_addr32;
619 const __be32 *a2 = addr2->s6_addr32;
620 unsigned int pdw, pbi;
621
622 /* check complete u32 in prefix */
623 pdw = prefixlen >> 5;
624 if (pdw && memcmp(a1, a2, pdw << 2))
625 return false;
626
627 /* check incomplete u32 in prefix */
628 pbi = prefixlen & 0x1f;
629 if (pbi && ((a1[pdw] ^ a2[pdw]) & htonl((0xffffffff) << (32 - pbi))))
630 return false;
631
632 return true;
633}
634#endif
635
636static inline bool ipv6_addr_any(const struct in6_addr *a)
637{
638#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
639 const unsigned long *ul = (const unsigned long *)a;
640
641 return (ul[0] | ul[1]) == 0UL;
642#else
643 return (a->s6_addr32[0] | a->s6_addr32[1] |
644 a->s6_addr32[2] | a->s6_addr32[3]) == 0;
645#endif
646}
647
648static inline u32 ipv6_addr_hash(const struct in6_addr *a)
649{
650#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
651 const unsigned long *ul = (const unsigned long *)a;
652 unsigned long x = ul[0] ^ ul[1];
653
654 return (u32)(x ^ (x >> 32));
655#else
656 return (__force u32)(a->s6_addr32[0] ^ a->s6_addr32[1] ^
657 a->s6_addr32[2] ^ a->s6_addr32[3]);
658#endif
659}
660
661/* more secured version of ipv6_addr_hash() */
662static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval)
663{
664 u32 v = (__force u32)a->s6_addr32[0] ^ (__force u32)a->s6_addr32[1];
665
666 return jhash_3words(v,
667 (__force u32)a->s6_addr32[2],
668 (__force u32)a->s6_addr32[3],
669 initval);
670}
671
672static inline bool ipv6_addr_loopback(const struct in6_addr *a)
673{
674#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
675 const __be64 *be = (const __be64 *)a;
676
677 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL;
678#else
679 return (a->s6_addr32[0] | a->s6_addr32[1] |
680 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0;
681#endif
682}
683
684/*
685 * Note that we must __force cast these to unsigned long to make sparse happy,
686 * since all of the endian-annotated types are fixed size regardless of arch.
687 */
688static inline bool ipv6_addr_v4mapped(const struct in6_addr *a)
689{
690 return (
691#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
692 *(unsigned long *)a |
693#else
694 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) |
695#endif
696 (__force unsigned long)(a->s6_addr32[2] ^
697 cpu_to_be32(0x0000ffff))) == 0UL;
698}
699
700static inline bool ipv6_addr_v4mapped_loopback(const struct in6_addr *a)
701{
702 return ipv6_addr_v4mapped(a) && ipv4_is_loopback(a->s6_addr32[3]);
703}
704
705static inline u32 ipv6_portaddr_hash(const struct net *net,
706 const struct in6_addr *addr6,
707 unsigned int port)
708{
709 unsigned int hash, mix = net_hash_mix(net);
710
711 if (ipv6_addr_any(addr6))
712 hash = jhash_1word(0, mix);
713 else if (ipv6_addr_v4mapped(addr6))
714 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
715 else
716 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
717
718 return hash ^ port;
719}
720
721/*
722 * Check for a RFC 4843 ORCHID address
723 * (Overlay Routable Cryptographic Hash Identifiers)
724 */
725static inline bool ipv6_addr_orchid(const struct in6_addr *a)
726{
727 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010);
728}
729
730static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr)
731{
732 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000);
733}
734
735static inline void ipv6_addr_set_v4mapped(const __be32 addr,
736 struct in6_addr *v4mapped)
737{
738 ipv6_addr_set(v4mapped,
739 0, 0,
740 htonl(0x0000FFFF),
741 addr);
742}
743
744/*
745 * find the first different bit between two addresses
746 * length of address must be a multiple of 32bits
747 */
748static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen)
749{
750 const __be32 *a1 = token1, *a2 = token2;
751 int i;
752
753 addrlen >>= 2;
754
755 for (i = 0; i < addrlen; i++) {
756 __be32 xb = a1[i] ^ a2[i];
757 if (xb)
758 return i * 32 + 31 - __fls(ntohl(xb));
759 }
760
761 /*
762 * we should *never* get to this point since that
763 * would mean the addrs are equal
764 *
765 * However, we do get to it 8) And exacly, when
766 * addresses are equal 8)
767 *
768 * ip route add 1111::/128 via ...
769 * ip route add 1111::/64 via ...
770 * and we are here.
771 *
772 * Ideally, this function should stop comparison
773 * at prefix length. It does not, but it is still OK,
774 * if returned value is greater than prefix length.
775 * --ANK (980803)
776 */
777 return addrlen << 5;
778}
779
780#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
781static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen)
782{
783 const __be64 *a1 = token1, *a2 = token2;
784 int i;
785
786 addrlen >>= 3;
787
788 for (i = 0; i < addrlen; i++) {
789 __be64 xb = a1[i] ^ a2[i];
790 if (xb)
791 return i * 64 + 63 - __fls(be64_to_cpu(xb));
792 }
793
794 return addrlen << 6;
795}
796#endif
797
798static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen)
799{
800#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
801 if (__builtin_constant_p(addrlen) && !(addrlen & 7))
802 return __ipv6_addr_diff64(token1, token2, addrlen);
803#endif
804 return __ipv6_addr_diff32(token1, token2, addrlen);
805}
806
807static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2)
808{
809 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr));
810}
811
812__be32 ipv6_select_ident(struct net *net,
813 const struct in6_addr *daddr,
814 const struct in6_addr *saddr);
815__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb);
816
817int ip6_dst_hoplimit(struct dst_entry *dst);
818
819static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
820 struct dst_entry *dst)
821{
822 int hlimit;
823
824 if (ipv6_addr_is_multicast(&fl6->daddr))
825 hlimit = np->mcast_hops;
826 else
827 hlimit = np->hop_limit;
828 if (hlimit < 0)
829 hlimit = ip6_dst_hoplimit(dst);
830 return hlimit;
831}
832
833/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store
834 * Equivalent to : flow->v6addrs.src = iph->saddr;
835 * flow->v6addrs.dst = iph->daddr;
836 */
837static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow,
838 const struct ipv6hdr *iph)
839{
840 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) !=
841 offsetof(typeof(flow->addrs), v6addrs.src) +
842 sizeof(flow->addrs.v6addrs.src));
843 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs));
844 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
845}
846
847#if IS_ENABLED(CONFIG_IPV6)
848
849static inline bool ipv6_can_nonlocal_bind(struct net *net,
850 struct inet_sock *inet)
851{
852 return net->ipv6.sysctl.ip_nonlocal_bind ||
853 inet->freebind || inet->transparent;
854}
855
856/* Sysctl settings for net ipv6.auto_flowlabels */
857#define IP6_AUTO_FLOW_LABEL_OFF 0
858#define IP6_AUTO_FLOW_LABEL_OPTOUT 1
859#define IP6_AUTO_FLOW_LABEL_OPTIN 2
860#define IP6_AUTO_FLOW_LABEL_FORCED 3
861
862#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED
863
864#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT
865
866static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
867 __be32 flowlabel, bool autolabel,
868 struct flowi6 *fl6)
869{
870 u32 hash;
871
872 /* @flowlabel may include more than a flow label, eg, the traffic class.
873 * Here we want only the flow label value.
874 */
875 flowlabel &= IPV6_FLOWLABEL_MASK;
876
877 if (flowlabel ||
878 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF ||
879 (!autolabel &&
880 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED))
881 return flowlabel;
882
883 hash = skb_get_hash_flowi6(skb, fl6);
884
885 /* Since this is being sent on the wire obfuscate hash a bit
886 * to minimize possbility that any useful information to an
887 * attacker is leaked. Only lower 20 bits are relevant.
888 */
889 hash = rol32(hash, 16);
890
891 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
892
893 if (net->ipv6.sysctl.flowlabel_state_ranges)
894 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG;
895
896 return flowlabel;
897}
898
899static inline int ip6_default_np_autolabel(struct net *net)
900{
901 switch (net->ipv6.sysctl.auto_flowlabels) {
902 case IP6_AUTO_FLOW_LABEL_OFF:
903 case IP6_AUTO_FLOW_LABEL_OPTIN:
904 default:
905 return 0;
906 case IP6_AUTO_FLOW_LABEL_OPTOUT:
907 case IP6_AUTO_FLOW_LABEL_FORCED:
908 return 1;
909 }
910}
911#else
912static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
913 __be32 flowlabel, bool autolabel,
914 struct flowi6 *fl6)
915{
916 return flowlabel;
917}
918static inline int ip6_default_np_autolabel(struct net *net)
919{
920 return 0;
921}
922#endif
923
924#if IS_ENABLED(CONFIG_IPV6)
925static inline int ip6_multipath_hash_policy(const struct net *net)
926{
927 return net->ipv6.sysctl.multipath_hash_policy;
928}
929static inline u32 ip6_multipath_hash_fields(const struct net *net)
930{
931 return net->ipv6.sysctl.multipath_hash_fields;
932}
933#else
934static inline int ip6_multipath_hash_policy(const struct net *net)
935{
936 return 0;
937}
938static inline u32 ip6_multipath_hash_fields(const struct net *net)
939{
940 return 0;
941}
942#endif
943
944/*
945 * Header manipulation
946 */
947static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass,
948 __be32 flowlabel)
949{
950 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel;
951}
952
953static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr)
954{
955 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK;
956}
957
958static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr)
959{
960 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK;
961}
962
963static inline u8 ip6_tclass(__be32 flowinfo)
964{
965 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT;
966}
967
968static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel)
969{
970 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel;
971}
972
973static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6)
974{
975 return fl6->flowlabel & IPV6_FLOWLABEL_MASK;
976}
977
978/*
979 * Prototypes exported by ipv6
980 */
981
982/*
983 * rcv function (called from netdevice level)
984 */
985
986int ipv6_rcv(struct sk_buff *skb, struct net_device *dev,
987 struct packet_type *pt, struct net_device *orig_dev);
988void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
989 struct net_device *orig_dev);
990
991int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb);
992
993/*
994 * upper-layer output functions
995 */
996int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
997 __u32 mark, struct ipv6_txoptions *opt, int tclass, u32 priority);
998
999int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr);
1000
1001int ip6_append_data(struct sock *sk,
1002 int getfrag(void *from, char *to, int offset, int len,
1003 int odd, struct sk_buff *skb),
1004 void *from, int length, int transhdrlen,
1005 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1006 struct rt6_info *rt, unsigned int flags);
1007
1008int ip6_push_pending_frames(struct sock *sk);
1009
1010void ip6_flush_pending_frames(struct sock *sk);
1011
1012int ip6_send_skb(struct sk_buff *skb);
1013
1014struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue,
1015 struct inet_cork_full *cork,
1016 struct inet6_cork *v6_cork);
1017struct sk_buff *ip6_make_skb(struct sock *sk,
1018 int getfrag(void *from, char *to, int offset,
1019 int len, int odd, struct sk_buff *skb),
1020 void *from, int length, int transhdrlen,
1021 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1022 struct rt6_info *rt, unsigned int flags,
1023 struct inet_cork_full *cork);
1024
1025static inline struct sk_buff *ip6_finish_skb(struct sock *sk)
1026{
1027 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork,
1028 &inet6_sk(sk)->cork);
1029}
1030
1031int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1032 struct flowi6 *fl6);
1033struct dst_entry *ip6_dst_lookup_flow(struct net *net, const struct sock *sk, struct flowi6 *fl6,
1034 const struct in6_addr *final_dst);
1035struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
1036 const struct in6_addr *final_dst,
1037 bool connected);
1038struct dst_entry *ip6_dst_lookup_tunnel(struct sk_buff *skb,
1039 struct net_device *dev,
1040 struct net *net, struct socket *sock,
1041 struct in6_addr *saddr,
1042 const struct ip_tunnel_info *info,
1043 u8 protocol, bool use_cache);
1044struct dst_entry *ip6_blackhole_route(struct net *net,
1045 struct dst_entry *orig_dst);
1046
1047/*
1048 * skb processing functions
1049 */
1050
1051int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb);
1052int ip6_forward(struct sk_buff *skb);
1053int ip6_input(struct sk_buff *skb);
1054int ip6_mc_input(struct sk_buff *skb);
1055void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
1056 bool have_final);
1057
1058int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1059int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1060
1061/*
1062 * Extension header (options) processing
1063 */
1064
1065void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1066 u8 *proto, struct in6_addr **daddr_p,
1067 struct in6_addr *saddr);
1068void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1069 u8 *proto);
1070
1071int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp,
1072 __be16 *frag_offp);
1073
1074bool ipv6_ext_hdr(u8 nexthdr);
1075
1076enum {
1077 IP6_FH_F_FRAG = (1 << 0),
1078 IP6_FH_F_AUTH = (1 << 1),
1079 IP6_FH_F_SKIP_RH = (1 << 2),
1080};
1081
1082/* find specified header and get offset to it */
1083int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target,
1084 unsigned short *fragoff, int *fragflg);
1085
1086int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type);
1087
1088struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1089 const struct ipv6_txoptions *opt,
1090 struct in6_addr *orig);
1091
1092/*
1093 * socket options (ipv6_sockglue.c)
1094 */
1095
1096int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1097 unsigned int optlen);
1098int ipv6_getsockopt(struct sock *sk, int level, int optname,
1099 char __user *optval, int __user *optlen);
1100
1101int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr,
1102 int addr_len);
1103int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len);
1104int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr,
1105 int addr_len);
1106int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr);
1107void ip6_datagram_release_cb(struct sock *sk);
1108
1109int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len,
1110 int *addr_len);
1111int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
1112 int *addr_len);
1113void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
1114 u32 info, u8 *payload);
1115void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info);
1116void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu);
1117
1118int inet6_release(struct socket *sock);
1119int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1120int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
1121 int peer);
1122int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg);
1123int inet6_compat_ioctl(struct socket *sock, unsigned int cmd,
1124 unsigned long arg);
1125
1126int inet6_hash_connect(struct inet_timewait_death_row *death_row,
1127 struct sock *sk);
1128int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size);
1129int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1130 int flags);
1131
1132/*
1133 * reassembly.c
1134 */
1135extern const struct proto_ops inet6_stream_ops;
1136extern const struct proto_ops inet6_dgram_ops;
1137extern const struct proto_ops inet6_sockraw_ops;
1138
1139struct group_source_req;
1140struct group_filter;
1141
1142int ip6_mc_source(int add, int omode, struct sock *sk,
1143 struct group_source_req *pgsr);
1144int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
1145 struct sockaddr_storage *list);
1146int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
1147 struct sockaddr_storage __user *p);
1148
1149#ifdef CONFIG_PROC_FS
1150int ac6_proc_init(struct net *net);
1151void ac6_proc_exit(struct net *net);
1152int raw6_proc_init(void);
1153void raw6_proc_exit(void);
1154int tcp6_proc_init(struct net *net);
1155void tcp6_proc_exit(struct net *net);
1156int udp6_proc_init(struct net *net);
1157void udp6_proc_exit(struct net *net);
1158int udplite6_proc_init(void);
1159void udplite6_proc_exit(void);
1160int ipv6_misc_proc_init(void);
1161void ipv6_misc_proc_exit(void);
1162int snmp6_register_dev(struct inet6_dev *idev);
1163int snmp6_unregister_dev(struct inet6_dev *idev);
1164
1165#else
1166static inline int ac6_proc_init(struct net *net) { return 0; }
1167static inline void ac6_proc_exit(struct net *net) { }
1168static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; }
1169static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; }
1170#endif
1171
1172#ifdef CONFIG_SYSCTL
1173struct ctl_table *ipv6_icmp_sysctl_init(struct net *net);
1174struct ctl_table *ipv6_route_sysctl_init(struct net *net);
1175int ipv6_sysctl_register(void);
1176void ipv6_sysctl_unregister(void);
1177#endif
1178
1179int ipv6_sock_mc_join(struct sock *sk, int ifindex,
1180 const struct in6_addr *addr);
1181int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
1182 const struct in6_addr *addr, unsigned int mode);
1183int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
1184 const struct in6_addr *addr);
1185
1186static inline int ip6_sock_set_v6only(struct sock *sk)
1187{
1188 if (inet_sk(sk)->inet_num)
1189 return -EINVAL;
1190 lock_sock(sk);
1191 sk->sk_ipv6only = true;
1192 release_sock(sk);
1193 return 0;
1194}
1195
1196static inline void ip6_sock_set_recverr(struct sock *sk)
1197{
1198 lock_sock(sk);
1199 inet6_sk(sk)->recverr = true;
1200 release_sock(sk);
1201}
1202
1203static inline int __ip6_sock_set_addr_preferences(struct sock *sk, int val)
1204{
1205 unsigned int pref = 0;
1206 unsigned int prefmask = ~0;
1207
1208 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
1209 switch (val & (IPV6_PREFER_SRC_PUBLIC |
1210 IPV6_PREFER_SRC_TMP |
1211 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
1212 case IPV6_PREFER_SRC_PUBLIC:
1213 pref |= IPV6_PREFER_SRC_PUBLIC;
1214 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1215 IPV6_PREFER_SRC_TMP);
1216 break;
1217 case IPV6_PREFER_SRC_TMP:
1218 pref |= IPV6_PREFER_SRC_TMP;
1219 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1220 IPV6_PREFER_SRC_TMP);
1221 break;
1222 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
1223 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1224 IPV6_PREFER_SRC_TMP);
1225 break;
1226 case 0:
1227 break;
1228 default:
1229 return -EINVAL;
1230 }
1231
1232 /* check HOME/COA conflicts */
1233 switch (val & (IPV6_PREFER_SRC_HOME | IPV6_PREFER_SRC_COA)) {
1234 case IPV6_PREFER_SRC_HOME:
1235 prefmask &= ~IPV6_PREFER_SRC_COA;
1236 break;
1237 case IPV6_PREFER_SRC_COA:
1238 pref |= IPV6_PREFER_SRC_COA;
1239 break;
1240 case 0:
1241 break;
1242 default:
1243 return -EINVAL;
1244 }
1245
1246 /* check CGA/NONCGA conflicts */
1247 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
1248 case IPV6_PREFER_SRC_CGA:
1249 case IPV6_PREFER_SRC_NONCGA:
1250 case 0:
1251 break;
1252 default:
1253 return -EINVAL;
1254 }
1255
1256 inet6_sk(sk)->srcprefs = (inet6_sk(sk)->srcprefs & prefmask) | pref;
1257 return 0;
1258}
1259
1260static inline int ip6_sock_set_addr_preferences(struct sock *sk, bool val)
1261{
1262 int ret;
1263
1264 lock_sock(sk);
1265 ret = __ip6_sock_set_addr_preferences(sk, val);
1266 release_sock(sk);
1267 return ret;
1268}
1269
1270static inline void ip6_sock_set_recvpktinfo(struct sock *sk)
1271{
1272 lock_sock(sk);
1273 inet6_sk(sk)->rxopt.bits.rxinfo = true;
1274 release_sock(sk);
1275}
1276
1277#endif /* _NET_IPV6_H */
1/*
2 * Linux INET6 implementation
3 *
4 * Authors:
5 * Pedro Roque <roque@di.fc.ul.pt>
6 *
7 * This program is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License
9 * as published by the Free Software Foundation; either version
10 * 2 of the License, or (at your option) any later version.
11 */
12
13#ifndef _NET_IPV6_H
14#define _NET_IPV6_H
15
16#include <linux/ipv6.h>
17#include <linux/hardirq.h>
18#include <linux/jhash.h>
19#include <linux/refcount.h>
20#include <net/if_inet6.h>
21#include <net/ndisc.h>
22#include <net/flow.h>
23#include <net/flow_dissector.h>
24#include <net/snmp.h>
25#include <net/netns/hash.h>
26
27#define SIN6_LEN_RFC2133 24
28
29#define IPV6_MAXPLEN 65535
30
31/*
32 * NextHeader field of IPv6 header
33 */
34
35#define NEXTHDR_HOP 0 /* Hop-by-hop option header. */
36#define NEXTHDR_TCP 6 /* TCP segment. */
37#define NEXTHDR_UDP 17 /* UDP message. */
38#define NEXTHDR_IPV6 41 /* IPv6 in IPv6 */
39#define NEXTHDR_ROUTING 43 /* Routing header. */
40#define NEXTHDR_FRAGMENT 44 /* Fragmentation/reassembly header. */
41#define NEXTHDR_GRE 47 /* GRE header. */
42#define NEXTHDR_ESP 50 /* Encapsulating security payload. */
43#define NEXTHDR_AUTH 51 /* Authentication header. */
44#define NEXTHDR_ICMP 58 /* ICMP for IPv6. */
45#define NEXTHDR_NONE 59 /* No next header */
46#define NEXTHDR_DEST 60 /* Destination options header. */
47#define NEXTHDR_SCTP 132 /* SCTP message. */
48#define NEXTHDR_MOBILITY 135 /* Mobility header. */
49
50#define NEXTHDR_MAX 255
51
52#define IPV6_DEFAULT_HOPLIMIT 64
53#define IPV6_DEFAULT_MCASTHOPS 1
54
55/* Limits on Hop-by-Hop and Destination options.
56 *
57 * Per RFC8200 there is no limit on the maximum number or lengths of options in
58 * Hop-by-Hop or Destination options other then the packet must fit in an MTU.
59 * We allow configurable limits in order to mitigate potential denial of
60 * service attacks.
61 *
62 * There are three limits that may be set:
63 * - Limit the number of options in a Hop-by-Hop or Destination options
64 * extension header
65 * - Limit the byte length of a Hop-by-Hop or Destination options extension
66 * header
67 * - Disallow unknown options
68 *
69 * The limits are expressed in corresponding sysctls:
70 *
71 * ipv6.sysctl.max_dst_opts_cnt
72 * ipv6.sysctl.max_hbh_opts_cnt
73 * ipv6.sysctl.max_dst_opts_len
74 * ipv6.sysctl.max_hbh_opts_len
75 *
76 * max_*_opts_cnt is the number of TLVs that are allowed for Destination
77 * options or Hop-by-Hop options. If the number is less than zero then unknown
78 * TLVs are disallowed and the number of known options that are allowed is the
79 * absolute value. Setting the value to INT_MAX indicates no limit.
80 *
81 * max_*_opts_len is the length limit in bytes of a Destination or
82 * Hop-by-Hop options extension header. Setting the value to INT_MAX
83 * indicates no length limit.
84 *
85 * If a limit is exceeded when processing an extension header the packet is
86 * silently discarded.
87 */
88
89/* Default limits for Hop-by-Hop and Destination options */
90#define IP6_DEFAULT_MAX_DST_OPTS_CNT 8
91#define IP6_DEFAULT_MAX_HBH_OPTS_CNT 8
92#define IP6_DEFAULT_MAX_DST_OPTS_LEN INT_MAX /* No limit */
93#define IP6_DEFAULT_MAX_HBH_OPTS_LEN INT_MAX /* No limit */
94
95/*
96 * Addr type
97 *
98 * type - unicast | multicast
99 * scope - local | site | global
100 * v4 - compat
101 * v4mapped
102 * any
103 * loopback
104 */
105
106#define IPV6_ADDR_ANY 0x0000U
107
108#define IPV6_ADDR_UNICAST 0x0001U
109#define IPV6_ADDR_MULTICAST 0x0002U
110
111#define IPV6_ADDR_LOOPBACK 0x0010U
112#define IPV6_ADDR_LINKLOCAL 0x0020U
113#define IPV6_ADDR_SITELOCAL 0x0040U
114
115#define IPV6_ADDR_COMPATv4 0x0080U
116
117#define IPV6_ADDR_SCOPE_MASK 0x00f0U
118
119#define IPV6_ADDR_MAPPED 0x1000U
120
121/*
122 * Addr scopes
123 */
124#define IPV6_ADDR_MC_SCOPE(a) \
125 ((a)->s6_addr[1] & 0x0f) /* nonstandard */
126#define __IPV6_ADDR_SCOPE_INVALID -1
127#define IPV6_ADDR_SCOPE_NODELOCAL 0x01
128#define IPV6_ADDR_SCOPE_LINKLOCAL 0x02
129#define IPV6_ADDR_SCOPE_SITELOCAL 0x05
130#define IPV6_ADDR_SCOPE_ORGLOCAL 0x08
131#define IPV6_ADDR_SCOPE_GLOBAL 0x0e
132
133/*
134 * Addr flags
135 */
136#define IPV6_ADDR_MC_FLAG_TRANSIENT(a) \
137 ((a)->s6_addr[1] & 0x10)
138#define IPV6_ADDR_MC_FLAG_PREFIX(a) \
139 ((a)->s6_addr[1] & 0x20)
140#define IPV6_ADDR_MC_FLAG_RENDEZVOUS(a) \
141 ((a)->s6_addr[1] & 0x40)
142
143/*
144 * fragmentation header
145 */
146
147struct frag_hdr {
148 __u8 nexthdr;
149 __u8 reserved;
150 __be16 frag_off;
151 __be32 identification;
152};
153
154#define IP6_MF 0x0001
155#define IP6_OFFSET 0xFFF8
156
157#define IP6_REPLY_MARK(net, mark) \
158 ((net)->ipv6.sysctl.fwmark_reflect ? (mark) : 0)
159
160#include <net/sock.h>
161
162/* sysctls */
163extern int sysctl_mld_max_msf;
164extern int sysctl_mld_qrv;
165
166#define _DEVINC(net, statname, mod, idev, field) \
167({ \
168 struct inet6_dev *_idev = (idev); \
169 if (likely(_idev != NULL)) \
170 mod##SNMP_INC_STATS64((_idev)->stats.statname, (field));\
171 mod##SNMP_INC_STATS64((net)->mib.statname##_statistics, (field));\
172})
173
174/* per device counters are atomic_long_t */
175#define _DEVINCATOMIC(net, statname, mod, idev, field) \
176({ \
177 struct inet6_dev *_idev = (idev); \
178 if (likely(_idev != NULL)) \
179 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
180 mod##SNMP_INC_STATS((net)->mib.statname##_statistics, (field));\
181})
182
183/* per device and per net counters are atomic_long_t */
184#define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \
185({ \
186 struct inet6_dev *_idev = (idev); \
187 if (likely(_idev != NULL)) \
188 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
189 SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\
190})
191
192#define _DEVADD(net, statname, mod, idev, field, val) \
193({ \
194 struct inet6_dev *_idev = (idev); \
195 if (likely(_idev != NULL)) \
196 mod##SNMP_ADD_STATS((_idev)->stats.statname, (field), (val)); \
197 mod##SNMP_ADD_STATS((net)->mib.statname##_statistics, (field), (val));\
198})
199
200#define _DEVUPD(net, statname, mod, idev, field, val) \
201({ \
202 struct inet6_dev *_idev = (idev); \
203 if (likely(_idev != NULL)) \
204 mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \
205 mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
206})
207
208/* MIBs */
209
210#define IP6_INC_STATS(net, idev,field) \
211 _DEVINC(net, ipv6, , idev, field)
212#define __IP6_INC_STATS(net, idev,field) \
213 _DEVINC(net, ipv6, __, idev, field)
214#define IP6_ADD_STATS(net, idev,field,val) \
215 _DEVADD(net, ipv6, , idev, field, val)
216#define __IP6_ADD_STATS(net, idev,field,val) \
217 _DEVADD(net, ipv6, __, idev, field, val)
218#define IP6_UPD_PO_STATS(net, idev,field,val) \
219 _DEVUPD(net, ipv6, , idev, field, val)
220#define __IP6_UPD_PO_STATS(net, idev,field,val) \
221 _DEVUPD(net, ipv6, __, idev, field, val)
222#define ICMP6_INC_STATS(net, idev, field) \
223 _DEVINCATOMIC(net, icmpv6, , idev, field)
224#define __ICMP6_INC_STATS(net, idev, field) \
225 _DEVINCATOMIC(net, icmpv6, __, idev, field)
226
227#define ICMP6MSGOUT_INC_STATS(net, idev, field) \
228 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256)
229#define ICMP6MSGIN_INC_STATS(net, idev, field) \
230 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field)
231
232struct ip6_ra_chain {
233 struct ip6_ra_chain *next;
234 struct sock *sk;
235 int sel;
236 void (*destructor)(struct sock *);
237};
238
239extern struct ip6_ra_chain *ip6_ra_chain;
240extern rwlock_t ip6_ra_lock;
241
242/*
243 This structure is prepared by protocol, when parsing
244 ancillary data and passed to IPv6.
245 */
246
247struct ipv6_txoptions {
248 refcount_t refcnt;
249 /* Length of this structure */
250 int tot_len;
251
252 /* length of extension headers */
253
254 __u16 opt_flen; /* after fragment hdr */
255 __u16 opt_nflen; /* before fragment hdr */
256
257 struct ipv6_opt_hdr *hopopt;
258 struct ipv6_opt_hdr *dst0opt;
259 struct ipv6_rt_hdr *srcrt; /* Routing Header */
260 struct ipv6_opt_hdr *dst1opt;
261 struct rcu_head rcu;
262 /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */
263};
264
265struct ip6_flowlabel {
266 struct ip6_flowlabel __rcu *next;
267 __be32 label;
268 atomic_t users;
269 struct in6_addr dst;
270 struct ipv6_txoptions *opt;
271 unsigned long linger;
272 struct rcu_head rcu;
273 u8 share;
274 union {
275 struct pid *pid;
276 kuid_t uid;
277 } owner;
278 unsigned long lastuse;
279 unsigned long expires;
280 struct net *fl_net;
281};
282
283#define IPV6_FLOWINFO_MASK cpu_to_be32(0x0FFFFFFF)
284#define IPV6_FLOWLABEL_MASK cpu_to_be32(0x000FFFFF)
285#define IPV6_FLOWLABEL_STATELESS_FLAG cpu_to_be32(0x00080000)
286
287#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
288#define IPV6_TCLASS_SHIFT 20
289
290struct ipv6_fl_socklist {
291 struct ipv6_fl_socklist __rcu *next;
292 struct ip6_flowlabel *fl;
293 struct rcu_head rcu;
294};
295
296struct ipcm6_cookie {
297 __s16 hlimit;
298 __s16 tclass;
299 __s8 dontfrag;
300 struct ipv6_txoptions *opt;
301};
302
303static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np)
304{
305 struct ipv6_txoptions *opt;
306
307 rcu_read_lock();
308 opt = rcu_dereference(np->opt);
309 if (opt) {
310 if (!refcount_inc_not_zero(&opt->refcnt))
311 opt = NULL;
312 else
313 opt = rcu_pointer_handoff(opt);
314 }
315 rcu_read_unlock();
316 return opt;
317}
318
319static inline void txopt_put(struct ipv6_txoptions *opt)
320{
321 if (opt && refcount_dec_and_test(&opt->refcnt))
322 kfree_rcu(opt, rcu);
323}
324
325struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk, __be32 label);
326struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space,
327 struct ip6_flowlabel *fl,
328 struct ipv6_txoptions *fopt);
329void fl6_free_socklist(struct sock *sk);
330int ipv6_flowlabel_opt(struct sock *sk, char __user *optval, int optlen);
331int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
332 int flags);
333int ip6_flowlabel_init(void);
334void ip6_flowlabel_cleanup(void);
335bool ip6_autoflowlabel(struct net *net, const struct ipv6_pinfo *np);
336
337static inline void fl6_sock_release(struct ip6_flowlabel *fl)
338{
339 if (fl)
340 atomic_dec(&fl->users);
341}
342
343void icmpv6_notify(struct sk_buff *skb, u8 type, u8 code, __be32 info);
344
345void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
346 struct icmp6hdr *thdr, int len);
347
348int ip6_ra_control(struct sock *sk, int sel);
349
350int ipv6_parse_hopopts(struct sk_buff *skb);
351
352struct ipv6_txoptions *ipv6_dup_options(struct sock *sk,
353 struct ipv6_txoptions *opt);
354struct ipv6_txoptions *ipv6_renew_options(struct sock *sk,
355 struct ipv6_txoptions *opt,
356 int newtype,
357 struct ipv6_opt_hdr __user *newopt,
358 int newoptlen);
359struct ipv6_txoptions *
360ipv6_renew_options_kern(struct sock *sk,
361 struct ipv6_txoptions *opt,
362 int newtype,
363 struct ipv6_opt_hdr *newopt,
364 int newoptlen);
365struct ipv6_txoptions *ipv6_fixup_options(struct ipv6_txoptions *opt_space,
366 struct ipv6_txoptions *opt);
367
368bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
369 const struct inet6_skb_parm *opt);
370struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
371 struct ipv6_txoptions *opt);
372
373static inline bool ipv6_accept_ra(struct inet6_dev *idev)
374{
375 /* If forwarding is enabled, RA are not accepted unless the special
376 * hybrid mode (accept_ra=2) is enabled.
377 */
378 return idev->cnf.forwarding ? idev->cnf.accept_ra == 2 :
379 idev->cnf.accept_ra;
380}
381
382#define IPV6_FRAG_HIGH_THRESH (4 * 1024*1024) /* 4194304 */
383#define IPV6_FRAG_LOW_THRESH (3 * 1024*1024) /* 3145728 */
384#define IPV6_FRAG_TIMEOUT (60 * HZ) /* 60 seconds */
385
386int __ipv6_addr_type(const struct in6_addr *addr);
387static inline int ipv6_addr_type(const struct in6_addr *addr)
388{
389 return __ipv6_addr_type(addr) & 0xffff;
390}
391
392static inline int ipv6_addr_scope(const struct in6_addr *addr)
393{
394 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK;
395}
396
397static inline int __ipv6_addr_src_scope(int type)
398{
399 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16);
400}
401
402static inline int ipv6_addr_src_scope(const struct in6_addr *addr)
403{
404 return __ipv6_addr_src_scope(__ipv6_addr_type(addr));
405}
406
407static inline bool __ipv6_addr_needs_scope_id(int type)
408{
409 return type & IPV6_ADDR_LINKLOCAL ||
410 (type & IPV6_ADDR_MULTICAST &&
411 (type & (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)));
412}
413
414static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr, int iface)
415{
416 return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0;
417}
418
419static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2)
420{
421 return memcmp(a1, a2, sizeof(struct in6_addr));
422}
423
424static inline bool
425ipv6_masked_addr_cmp(const struct in6_addr *a1, const struct in6_addr *m,
426 const struct in6_addr *a2)
427{
428#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
429 const unsigned long *ul1 = (const unsigned long *)a1;
430 const unsigned long *ulm = (const unsigned long *)m;
431 const unsigned long *ul2 = (const unsigned long *)a2;
432
433 return !!(((ul1[0] ^ ul2[0]) & ulm[0]) |
434 ((ul1[1] ^ ul2[1]) & ulm[1]));
435#else
436 return !!(((a1->s6_addr32[0] ^ a2->s6_addr32[0]) & m->s6_addr32[0]) |
437 ((a1->s6_addr32[1] ^ a2->s6_addr32[1]) & m->s6_addr32[1]) |
438 ((a1->s6_addr32[2] ^ a2->s6_addr32[2]) & m->s6_addr32[2]) |
439 ((a1->s6_addr32[3] ^ a2->s6_addr32[3]) & m->s6_addr32[3]));
440#endif
441}
442
443static inline void ipv6_addr_prefix(struct in6_addr *pfx,
444 const struct in6_addr *addr,
445 int plen)
446{
447 /* caller must guarantee 0 <= plen <= 128 */
448 int o = plen >> 3,
449 b = plen & 0x7;
450
451 memset(pfx->s6_addr, 0, sizeof(pfx->s6_addr));
452 memcpy(pfx->s6_addr, addr, o);
453 if (b != 0)
454 pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b);
455}
456
457static inline void ipv6_addr_prefix_copy(struct in6_addr *addr,
458 const struct in6_addr *pfx,
459 int plen)
460{
461 /* caller must guarantee 0 <= plen <= 128 */
462 int o = plen >> 3,
463 b = plen & 0x7;
464
465 memcpy(addr->s6_addr, pfx, o);
466 if (b != 0) {
467 addr->s6_addr[o] &= ~(0xff00 >> b);
468 addr->s6_addr[o] |= (pfx->s6_addr[o] & (0xff00 >> b));
469 }
470}
471
472static inline void __ipv6_addr_set_half(__be32 *addr,
473 __be32 wh, __be32 wl)
474{
475#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
476#if defined(__BIG_ENDIAN)
477 if (__builtin_constant_p(wh) && __builtin_constant_p(wl)) {
478 *(__force u64 *)addr = ((__force u64)(wh) << 32 | (__force u64)(wl));
479 return;
480 }
481#elif defined(__LITTLE_ENDIAN)
482 if (__builtin_constant_p(wl) && __builtin_constant_p(wh)) {
483 *(__force u64 *)addr = ((__force u64)(wl) << 32 | (__force u64)(wh));
484 return;
485 }
486#endif
487#endif
488 addr[0] = wh;
489 addr[1] = wl;
490}
491
492static inline void ipv6_addr_set(struct in6_addr *addr,
493 __be32 w1, __be32 w2,
494 __be32 w3, __be32 w4)
495{
496 __ipv6_addr_set_half(&addr->s6_addr32[0], w1, w2);
497 __ipv6_addr_set_half(&addr->s6_addr32[2], w3, w4);
498}
499
500static inline bool ipv6_addr_equal(const struct in6_addr *a1,
501 const struct in6_addr *a2)
502{
503#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
504 const unsigned long *ul1 = (const unsigned long *)a1;
505 const unsigned long *ul2 = (const unsigned long *)a2;
506
507 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL;
508#else
509 return ((a1->s6_addr32[0] ^ a2->s6_addr32[0]) |
510 (a1->s6_addr32[1] ^ a2->s6_addr32[1]) |
511 (a1->s6_addr32[2] ^ a2->s6_addr32[2]) |
512 (a1->s6_addr32[3] ^ a2->s6_addr32[3])) == 0;
513#endif
514}
515
516#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
517static inline bool __ipv6_prefix_equal64_half(const __be64 *a1,
518 const __be64 *a2,
519 unsigned int len)
520{
521 if (len && ((*a1 ^ *a2) & cpu_to_be64((~0UL) << (64 - len))))
522 return false;
523 return true;
524}
525
526static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
527 const struct in6_addr *addr2,
528 unsigned int prefixlen)
529{
530 const __be64 *a1 = (const __be64 *)addr1;
531 const __be64 *a2 = (const __be64 *)addr2;
532
533 if (prefixlen >= 64) {
534 if (a1[0] ^ a2[0])
535 return false;
536 return __ipv6_prefix_equal64_half(a1 + 1, a2 + 1, prefixlen - 64);
537 }
538 return __ipv6_prefix_equal64_half(a1, a2, prefixlen);
539}
540#else
541static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
542 const struct in6_addr *addr2,
543 unsigned int prefixlen)
544{
545 const __be32 *a1 = addr1->s6_addr32;
546 const __be32 *a2 = addr2->s6_addr32;
547 unsigned int pdw, pbi;
548
549 /* check complete u32 in prefix */
550 pdw = prefixlen >> 5;
551 if (pdw && memcmp(a1, a2, pdw << 2))
552 return false;
553
554 /* check incomplete u32 in prefix */
555 pbi = prefixlen & 0x1f;
556 if (pbi && ((a1[pdw] ^ a2[pdw]) & htonl((0xffffffff) << (32 - pbi))))
557 return false;
558
559 return true;
560}
561#endif
562
563struct inet_frag_queue;
564
565enum ip6_defrag_users {
566 IP6_DEFRAG_LOCAL_DELIVER,
567 IP6_DEFRAG_CONNTRACK_IN,
568 __IP6_DEFRAG_CONNTRACK_IN = IP6_DEFRAG_CONNTRACK_IN + USHRT_MAX,
569 IP6_DEFRAG_CONNTRACK_OUT,
570 __IP6_DEFRAG_CONNTRACK_OUT = IP6_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
571 IP6_DEFRAG_CONNTRACK_BRIDGE_IN,
572 __IP6_DEFRAG_CONNTRACK_BRIDGE_IN = IP6_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
573};
574
575void ip6_frag_init(struct inet_frag_queue *q, const void *a);
576extern const struct rhashtable_params ip6_rhash_params;
577
578/*
579 * Equivalent of ipv4 struct ip
580 */
581struct frag_queue {
582 struct inet_frag_queue q;
583
584 int iif;
585 __u16 nhoffset;
586 u8 ecn;
587};
588
589void ip6_expire_frag_queue(struct net *net, struct frag_queue *fq);
590
591static inline bool ipv6_addr_any(const struct in6_addr *a)
592{
593#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
594 const unsigned long *ul = (const unsigned long *)a;
595
596 return (ul[0] | ul[1]) == 0UL;
597#else
598 return (a->s6_addr32[0] | a->s6_addr32[1] |
599 a->s6_addr32[2] | a->s6_addr32[3]) == 0;
600#endif
601}
602
603static inline u32 ipv6_addr_hash(const struct in6_addr *a)
604{
605#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
606 const unsigned long *ul = (const unsigned long *)a;
607 unsigned long x = ul[0] ^ ul[1];
608
609 return (u32)(x ^ (x >> 32));
610#else
611 return (__force u32)(a->s6_addr32[0] ^ a->s6_addr32[1] ^
612 a->s6_addr32[2] ^ a->s6_addr32[3]);
613#endif
614}
615
616/* more secured version of ipv6_addr_hash() */
617static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval)
618{
619 u32 v = (__force u32)a->s6_addr32[0] ^ (__force u32)a->s6_addr32[1];
620
621 return jhash_3words(v,
622 (__force u32)a->s6_addr32[2],
623 (__force u32)a->s6_addr32[3],
624 initval);
625}
626
627static inline bool ipv6_addr_loopback(const struct in6_addr *a)
628{
629#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
630 const __be64 *be = (const __be64 *)a;
631
632 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL;
633#else
634 return (a->s6_addr32[0] | a->s6_addr32[1] |
635 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0;
636#endif
637}
638
639/*
640 * Note that we must __force cast these to unsigned long to make sparse happy,
641 * since all of the endian-annotated types are fixed size regardless of arch.
642 */
643static inline bool ipv6_addr_v4mapped(const struct in6_addr *a)
644{
645 return (
646#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
647 *(unsigned long *)a |
648#else
649 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) |
650#endif
651 (__force unsigned long)(a->s6_addr32[2] ^
652 cpu_to_be32(0x0000ffff))) == 0UL;
653}
654
655static inline u32 ipv6_portaddr_hash(const struct net *net,
656 const struct in6_addr *addr6,
657 unsigned int port)
658{
659 unsigned int hash, mix = net_hash_mix(net);
660
661 if (ipv6_addr_any(addr6))
662 hash = jhash_1word(0, mix);
663 else if (ipv6_addr_v4mapped(addr6))
664 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
665 else
666 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
667
668 return hash ^ port;
669}
670
671/*
672 * Check for a RFC 4843 ORCHID address
673 * (Overlay Routable Cryptographic Hash Identifiers)
674 */
675static inline bool ipv6_addr_orchid(const struct in6_addr *a)
676{
677 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010);
678}
679
680static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr)
681{
682 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000);
683}
684
685static inline void ipv6_addr_set_v4mapped(const __be32 addr,
686 struct in6_addr *v4mapped)
687{
688 ipv6_addr_set(v4mapped,
689 0, 0,
690 htonl(0x0000FFFF),
691 addr);
692}
693
694/*
695 * find the first different bit between two addresses
696 * length of address must be a multiple of 32bits
697 */
698static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen)
699{
700 const __be32 *a1 = token1, *a2 = token2;
701 int i;
702
703 addrlen >>= 2;
704
705 for (i = 0; i < addrlen; i++) {
706 __be32 xb = a1[i] ^ a2[i];
707 if (xb)
708 return i * 32 + 31 - __fls(ntohl(xb));
709 }
710
711 /*
712 * we should *never* get to this point since that
713 * would mean the addrs are equal
714 *
715 * However, we do get to it 8) And exacly, when
716 * addresses are equal 8)
717 *
718 * ip route add 1111::/128 via ...
719 * ip route add 1111::/64 via ...
720 * and we are here.
721 *
722 * Ideally, this function should stop comparison
723 * at prefix length. It does not, but it is still OK,
724 * if returned value is greater than prefix length.
725 * --ANK (980803)
726 */
727 return addrlen << 5;
728}
729
730#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
731static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen)
732{
733 const __be64 *a1 = token1, *a2 = token2;
734 int i;
735
736 addrlen >>= 3;
737
738 for (i = 0; i < addrlen; i++) {
739 __be64 xb = a1[i] ^ a2[i];
740 if (xb)
741 return i * 64 + 63 - __fls(be64_to_cpu(xb));
742 }
743
744 return addrlen << 6;
745}
746#endif
747
748static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen)
749{
750#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
751 if (__builtin_constant_p(addrlen) && !(addrlen & 7))
752 return __ipv6_addr_diff64(token1, token2, addrlen);
753#endif
754 return __ipv6_addr_diff32(token1, token2, addrlen);
755}
756
757static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2)
758{
759 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr));
760}
761
762__be32 ipv6_select_ident(struct net *net,
763 const struct in6_addr *daddr,
764 const struct in6_addr *saddr);
765__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb);
766
767int ip6_dst_hoplimit(struct dst_entry *dst);
768
769static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
770 struct dst_entry *dst)
771{
772 int hlimit;
773
774 if (ipv6_addr_is_multicast(&fl6->daddr))
775 hlimit = np->mcast_hops;
776 else
777 hlimit = np->hop_limit;
778 if (hlimit < 0)
779 hlimit = ip6_dst_hoplimit(dst);
780 return hlimit;
781}
782
783/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store
784 * Equivalent to : flow->v6addrs.src = iph->saddr;
785 * flow->v6addrs.dst = iph->daddr;
786 */
787static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow,
788 const struct ipv6hdr *iph)
789{
790 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) !=
791 offsetof(typeof(flow->addrs), v6addrs.src) +
792 sizeof(flow->addrs.v6addrs.src));
793 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs));
794 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
795}
796
797#if IS_ENABLED(CONFIG_IPV6)
798
799/* Sysctl settings for net ipv6.auto_flowlabels */
800#define IP6_AUTO_FLOW_LABEL_OFF 0
801#define IP6_AUTO_FLOW_LABEL_OPTOUT 1
802#define IP6_AUTO_FLOW_LABEL_OPTIN 2
803#define IP6_AUTO_FLOW_LABEL_FORCED 3
804
805#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED
806
807#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT
808
809static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
810 __be32 flowlabel, bool autolabel,
811 struct flowi6 *fl6)
812{
813 u32 hash;
814
815 /* @flowlabel may include more than a flow label, eg, the traffic class.
816 * Here we want only the flow label value.
817 */
818 flowlabel &= IPV6_FLOWLABEL_MASK;
819
820 if (flowlabel ||
821 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF ||
822 (!autolabel &&
823 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED))
824 return flowlabel;
825
826 hash = skb_get_hash_flowi6(skb, fl6);
827
828 /* Since this is being sent on the wire obfuscate hash a bit
829 * to minimize possbility that any useful information to an
830 * attacker is leaked. Only lower 20 bits are relevant.
831 */
832 rol32(hash, 16);
833
834 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
835
836 if (net->ipv6.sysctl.flowlabel_state_ranges)
837 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG;
838
839 return flowlabel;
840}
841
842static inline int ip6_default_np_autolabel(struct net *net)
843{
844 switch (net->ipv6.sysctl.auto_flowlabels) {
845 case IP6_AUTO_FLOW_LABEL_OFF:
846 case IP6_AUTO_FLOW_LABEL_OPTIN:
847 default:
848 return 0;
849 case IP6_AUTO_FLOW_LABEL_OPTOUT:
850 case IP6_AUTO_FLOW_LABEL_FORCED:
851 return 1;
852 }
853}
854#else
855static inline void ip6_set_txhash(struct sock *sk) { }
856static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
857 __be32 flowlabel, bool autolabel,
858 struct flowi6 *fl6)
859{
860 return flowlabel;
861}
862static inline int ip6_default_np_autolabel(struct net *net)
863{
864 return 0;
865}
866#endif
867
868#if IS_ENABLED(CONFIG_IPV6)
869static inline int ip6_multipath_hash_policy(const struct net *net)
870{
871 return net->ipv6.sysctl.multipath_hash_policy;
872}
873#else
874static inline int ip6_multipath_hash_policy(const struct net *net)
875{
876 return 0;
877}
878#endif
879
880/*
881 * Header manipulation
882 */
883static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass,
884 __be32 flowlabel)
885{
886 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel;
887}
888
889static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr)
890{
891 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK;
892}
893
894static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr)
895{
896 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK;
897}
898
899static inline u8 ip6_tclass(__be32 flowinfo)
900{
901 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT;
902}
903
904static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel)
905{
906 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel;
907}
908
909/*
910 * Prototypes exported by ipv6
911 */
912
913/*
914 * rcv function (called from netdevice level)
915 */
916
917int ipv6_rcv(struct sk_buff *skb, struct net_device *dev,
918 struct packet_type *pt, struct net_device *orig_dev);
919
920int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb);
921
922/*
923 * upper-layer output functions
924 */
925int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
926 __u32 mark, struct ipv6_txoptions *opt, int tclass);
927
928int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr);
929
930int ip6_append_data(struct sock *sk,
931 int getfrag(void *from, char *to, int offset, int len,
932 int odd, struct sk_buff *skb),
933 void *from, int length, int transhdrlen,
934 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
935 struct rt6_info *rt, unsigned int flags,
936 const struct sockcm_cookie *sockc);
937
938int ip6_push_pending_frames(struct sock *sk);
939
940void ip6_flush_pending_frames(struct sock *sk);
941
942int ip6_send_skb(struct sk_buff *skb);
943
944struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue,
945 struct inet_cork_full *cork,
946 struct inet6_cork *v6_cork);
947struct sk_buff *ip6_make_skb(struct sock *sk,
948 int getfrag(void *from, char *to, int offset,
949 int len, int odd, struct sk_buff *skb),
950 void *from, int length, int transhdrlen,
951 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
952 struct rt6_info *rt, unsigned int flags,
953 const struct sockcm_cookie *sockc);
954
955static inline struct sk_buff *ip6_finish_skb(struct sock *sk)
956{
957 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork,
958 &inet6_sk(sk)->cork);
959}
960
961unsigned int ip6_dst_mtu_forward(const struct dst_entry *dst);
962
963int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
964 struct flowi6 *fl6);
965struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
966 const struct in6_addr *final_dst);
967struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
968 const struct in6_addr *final_dst,
969 bool connected);
970struct dst_entry *ip6_blackhole_route(struct net *net,
971 struct dst_entry *orig_dst);
972
973/*
974 * skb processing functions
975 */
976
977int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb);
978int ip6_forward(struct sk_buff *skb);
979int ip6_input(struct sk_buff *skb);
980int ip6_mc_input(struct sk_buff *skb);
981
982int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
983int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
984
985/*
986 * Extension header (options) processing
987 */
988
989void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
990 u8 *proto, struct in6_addr **daddr_p,
991 struct in6_addr *saddr);
992void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
993 u8 *proto);
994
995int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp,
996 __be16 *frag_offp);
997
998bool ipv6_ext_hdr(u8 nexthdr);
999
1000enum {
1001 IP6_FH_F_FRAG = (1 << 0),
1002 IP6_FH_F_AUTH = (1 << 1),
1003 IP6_FH_F_SKIP_RH = (1 << 2),
1004};
1005
1006/* find specified header and get offset to it */
1007int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target,
1008 unsigned short *fragoff, int *fragflg);
1009
1010int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type);
1011
1012struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1013 const struct ipv6_txoptions *opt,
1014 struct in6_addr *orig);
1015
1016/*
1017 * socket options (ipv6_sockglue.c)
1018 */
1019
1020int ipv6_setsockopt(struct sock *sk, int level, int optname,
1021 char __user *optval, unsigned int optlen);
1022int ipv6_getsockopt(struct sock *sk, int level, int optname,
1023 char __user *optval, int __user *optlen);
1024int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
1025 char __user *optval, unsigned int optlen);
1026int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1027 char __user *optval, int __user *optlen);
1028
1029int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr,
1030 int addr_len);
1031int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len);
1032int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr,
1033 int addr_len);
1034int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr);
1035void ip6_datagram_release_cb(struct sock *sk);
1036
1037int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len,
1038 int *addr_len);
1039int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
1040 int *addr_len);
1041void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
1042 u32 info, u8 *payload);
1043void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info);
1044void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu);
1045
1046int inet6_release(struct socket *sock);
1047int __inet6_bind(struct sock *sock, struct sockaddr *uaddr, int addr_len,
1048 bool force_bind_address_no_port, bool with_lock);
1049int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1050int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
1051 int peer);
1052int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg);
1053
1054int inet6_hash_connect(struct inet_timewait_death_row *death_row,
1055 struct sock *sk);
1056
1057/*
1058 * reassembly.c
1059 */
1060extern const struct proto_ops inet6_stream_ops;
1061extern const struct proto_ops inet6_dgram_ops;
1062extern const struct proto_ops inet6_sockraw_ops;
1063
1064struct group_source_req;
1065struct group_filter;
1066
1067int ip6_mc_source(int add, int omode, struct sock *sk,
1068 struct group_source_req *pgsr);
1069int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf);
1070int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
1071 struct group_filter __user *optval, int __user *optlen);
1072
1073#ifdef CONFIG_PROC_FS
1074int ac6_proc_init(struct net *net);
1075void ac6_proc_exit(struct net *net);
1076int raw6_proc_init(void);
1077void raw6_proc_exit(void);
1078int tcp6_proc_init(struct net *net);
1079void tcp6_proc_exit(struct net *net);
1080int udp6_proc_init(struct net *net);
1081void udp6_proc_exit(struct net *net);
1082int udplite6_proc_init(void);
1083void udplite6_proc_exit(void);
1084int ipv6_misc_proc_init(void);
1085void ipv6_misc_proc_exit(void);
1086int snmp6_register_dev(struct inet6_dev *idev);
1087int snmp6_unregister_dev(struct inet6_dev *idev);
1088
1089#else
1090static inline int ac6_proc_init(struct net *net) { return 0; }
1091static inline void ac6_proc_exit(struct net *net) { }
1092static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; }
1093static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; }
1094#endif
1095
1096#ifdef CONFIG_SYSCTL
1097extern struct ctl_table ipv6_route_table_template[];
1098
1099struct ctl_table *ipv6_icmp_sysctl_init(struct net *net);
1100struct ctl_table *ipv6_route_sysctl_init(struct net *net);
1101int ipv6_sysctl_register(void);
1102void ipv6_sysctl_unregister(void);
1103#endif
1104
1105int ipv6_sock_mc_join(struct sock *sk, int ifindex,
1106 const struct in6_addr *addr);
1107int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
1108 const struct in6_addr *addr);
1109#endif /* _NET_IPV6_H */