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_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 u32 v = (__force u32)a->s6_addr32[0] ^ (__force u32)a->s6_addr32[1];
664
665 return jhash_3words(v,
666 (__force u32)a->s6_addr32[2],
667 (__force u32)a->s6_addr32[3],
668 initval);
669}
670
671static inline bool ipv6_addr_loopback(const struct in6_addr *a)
672{
673#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
674 const __be64 *be = (const __be64 *)a;
675
676 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL;
677#else
678 return (a->s6_addr32[0] | a->s6_addr32[1] |
679 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0;
680#endif
681}
682
683/*
684 * Note that we must __force cast these to unsigned long to make sparse happy,
685 * since all of the endian-annotated types are fixed size regardless of arch.
686 */
687static inline bool ipv6_addr_v4mapped(const struct in6_addr *a)
688{
689 return (
690#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
691 *(unsigned long *)a |
692#else
693 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) |
694#endif
695 (__force unsigned long)(a->s6_addr32[2] ^
696 cpu_to_be32(0x0000ffff))) == 0UL;
697}
698
699static inline u32 ipv6_portaddr_hash(const struct net *net,
700 const struct in6_addr *addr6,
701 unsigned int port)
702{
703 unsigned int hash, mix = net_hash_mix(net);
704
705 if (ipv6_addr_any(addr6))
706 hash = jhash_1word(0, mix);
707 else if (ipv6_addr_v4mapped(addr6))
708 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
709 else
710 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
711
712 return hash ^ port;
713}
714
715/*
716 * Check for a RFC 4843 ORCHID address
717 * (Overlay Routable Cryptographic Hash Identifiers)
718 */
719static inline bool ipv6_addr_orchid(const struct in6_addr *a)
720{
721 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010);
722}
723
724static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr)
725{
726 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000);
727}
728
729static inline void ipv6_addr_set_v4mapped(const __be32 addr,
730 struct in6_addr *v4mapped)
731{
732 ipv6_addr_set(v4mapped,
733 0, 0,
734 htonl(0x0000FFFF),
735 addr);
736}
737
738/*
739 * find the first different bit between two addresses
740 * length of address must be a multiple of 32bits
741 */
742static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen)
743{
744 const __be32 *a1 = token1, *a2 = token2;
745 int i;
746
747 addrlen >>= 2;
748
749 for (i = 0; i < addrlen; i++) {
750 __be32 xb = a1[i] ^ a2[i];
751 if (xb)
752 return i * 32 + 31 - __fls(ntohl(xb));
753 }
754
755 /*
756 * we should *never* get to this point since that
757 * would mean the addrs are equal
758 *
759 * However, we do get to it 8) And exacly, when
760 * addresses are equal 8)
761 *
762 * ip route add 1111::/128 via ...
763 * ip route add 1111::/64 via ...
764 * and we are here.
765 *
766 * Ideally, this function should stop comparison
767 * at prefix length. It does not, but it is still OK,
768 * if returned value is greater than prefix length.
769 * --ANK (980803)
770 */
771 return addrlen << 5;
772}
773
774#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
775static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen)
776{
777 const __be64 *a1 = token1, *a2 = token2;
778 int i;
779
780 addrlen >>= 3;
781
782 for (i = 0; i < addrlen; i++) {
783 __be64 xb = a1[i] ^ a2[i];
784 if (xb)
785 return i * 64 + 63 - __fls(be64_to_cpu(xb));
786 }
787
788 return addrlen << 6;
789}
790#endif
791
792static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen)
793{
794#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
795 if (__builtin_constant_p(addrlen) && !(addrlen & 7))
796 return __ipv6_addr_diff64(token1, token2, addrlen);
797#endif
798 return __ipv6_addr_diff32(token1, token2, addrlen);
799}
800
801static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2)
802{
803 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr));
804}
805
806__be32 ipv6_select_ident(struct net *net,
807 const struct in6_addr *daddr,
808 const struct in6_addr *saddr);
809__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb);
810
811int ip6_dst_hoplimit(struct dst_entry *dst);
812
813static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
814 struct dst_entry *dst)
815{
816 int hlimit;
817
818 if (ipv6_addr_is_multicast(&fl6->daddr))
819 hlimit = np->mcast_hops;
820 else
821 hlimit = np->hop_limit;
822 if (hlimit < 0)
823 hlimit = ip6_dst_hoplimit(dst);
824 return hlimit;
825}
826
827/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store
828 * Equivalent to : flow->v6addrs.src = iph->saddr;
829 * flow->v6addrs.dst = iph->daddr;
830 */
831static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow,
832 const struct ipv6hdr *iph)
833{
834 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) !=
835 offsetof(typeof(flow->addrs), v6addrs.src) +
836 sizeof(flow->addrs.v6addrs.src));
837 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs));
838 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
839}
840
841#if IS_ENABLED(CONFIG_IPV6)
842
843static inline bool ipv6_can_nonlocal_bind(struct net *net,
844 struct inet_sock *inet)
845{
846 return net->ipv6.sysctl.ip_nonlocal_bind ||
847 inet->freebind || inet->transparent;
848}
849
850/* Sysctl settings for net ipv6.auto_flowlabels */
851#define IP6_AUTO_FLOW_LABEL_OFF 0
852#define IP6_AUTO_FLOW_LABEL_OPTOUT 1
853#define IP6_AUTO_FLOW_LABEL_OPTIN 2
854#define IP6_AUTO_FLOW_LABEL_FORCED 3
855
856#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED
857
858#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT
859
860static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
861 __be32 flowlabel, bool autolabel,
862 struct flowi6 *fl6)
863{
864 u32 hash;
865
866 /* @flowlabel may include more than a flow label, eg, the traffic class.
867 * Here we want only the flow label value.
868 */
869 flowlabel &= IPV6_FLOWLABEL_MASK;
870
871 if (flowlabel ||
872 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF ||
873 (!autolabel &&
874 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED))
875 return flowlabel;
876
877 hash = skb_get_hash_flowi6(skb, fl6);
878
879 /* Since this is being sent on the wire obfuscate hash a bit
880 * to minimize possbility that any useful information to an
881 * attacker is leaked. Only lower 20 bits are relevant.
882 */
883 hash = rol32(hash, 16);
884
885 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
886
887 if (net->ipv6.sysctl.flowlabel_state_ranges)
888 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG;
889
890 return flowlabel;
891}
892
893static inline int ip6_default_np_autolabel(struct net *net)
894{
895 switch (net->ipv6.sysctl.auto_flowlabels) {
896 case IP6_AUTO_FLOW_LABEL_OFF:
897 case IP6_AUTO_FLOW_LABEL_OPTIN:
898 default:
899 return 0;
900 case IP6_AUTO_FLOW_LABEL_OPTOUT:
901 case IP6_AUTO_FLOW_LABEL_FORCED:
902 return 1;
903 }
904}
905#else
906static inline void ip6_set_txhash(struct sock *sk) { }
907static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
908 __be32 flowlabel, bool autolabel,
909 struct flowi6 *fl6)
910{
911 return flowlabel;
912}
913static inline int ip6_default_np_autolabel(struct net *net)
914{
915 return 0;
916}
917#endif
918
919#if IS_ENABLED(CONFIG_IPV6)
920static inline int ip6_multipath_hash_policy(const struct net *net)
921{
922 return net->ipv6.sysctl.multipath_hash_policy;
923}
924#else
925static inline int ip6_multipath_hash_policy(const struct net *net)
926{
927 return 0;
928}
929#endif
930
931/*
932 * Header manipulation
933 */
934static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass,
935 __be32 flowlabel)
936{
937 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel;
938}
939
940static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr)
941{
942 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK;
943}
944
945static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr)
946{
947 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK;
948}
949
950static inline u8 ip6_tclass(__be32 flowinfo)
951{
952 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT;
953}
954
955static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel)
956{
957 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel;
958}
959
960static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6)
961{
962 return fl6->flowlabel & IPV6_FLOWLABEL_MASK;
963}
964
965/*
966 * Prototypes exported by ipv6
967 */
968
969/*
970 * rcv function (called from netdevice level)
971 */
972
973int ipv6_rcv(struct sk_buff *skb, struct net_device *dev,
974 struct packet_type *pt, struct net_device *orig_dev);
975void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
976 struct net_device *orig_dev);
977
978int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb);
979
980/*
981 * upper-layer output functions
982 */
983int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
984 __u32 mark, struct ipv6_txoptions *opt, int tclass, u32 priority);
985
986int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr);
987
988int ip6_append_data(struct sock *sk,
989 int getfrag(void *from, char *to, int offset, int len,
990 int odd, struct sk_buff *skb),
991 void *from, int length, int transhdrlen,
992 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
993 struct rt6_info *rt, unsigned int flags);
994
995int ip6_push_pending_frames(struct sock *sk);
996
997void ip6_flush_pending_frames(struct sock *sk);
998
999int ip6_send_skb(struct sk_buff *skb);
1000
1001struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue,
1002 struct inet_cork_full *cork,
1003 struct inet6_cork *v6_cork);
1004struct sk_buff *ip6_make_skb(struct sock *sk,
1005 int getfrag(void *from, char *to, int offset,
1006 int len, int odd, struct sk_buff *skb),
1007 void *from, int length, int transhdrlen,
1008 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1009 struct rt6_info *rt, unsigned int flags,
1010 struct inet_cork_full *cork);
1011
1012static inline struct sk_buff *ip6_finish_skb(struct sock *sk)
1013{
1014 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork,
1015 &inet6_sk(sk)->cork);
1016}
1017
1018int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1019 struct flowi6 *fl6);
1020struct dst_entry *ip6_dst_lookup_flow(const struct sock *sk, struct flowi6 *fl6,
1021 const struct in6_addr *final_dst);
1022struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
1023 const struct in6_addr *final_dst,
1024 bool connected);
1025struct dst_entry *ip6_blackhole_route(struct net *net,
1026 struct dst_entry *orig_dst);
1027
1028/*
1029 * skb processing functions
1030 */
1031
1032int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb);
1033int ip6_forward(struct sk_buff *skb);
1034int ip6_input(struct sk_buff *skb);
1035int ip6_mc_input(struct sk_buff *skb);
1036void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
1037 bool have_final);
1038
1039int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1040int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1041
1042/*
1043 * Extension header (options) processing
1044 */
1045
1046void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1047 u8 *proto, struct in6_addr **daddr_p,
1048 struct in6_addr *saddr);
1049void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1050 u8 *proto);
1051
1052int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp,
1053 __be16 *frag_offp);
1054
1055bool ipv6_ext_hdr(u8 nexthdr);
1056
1057enum {
1058 IP6_FH_F_FRAG = (1 << 0),
1059 IP6_FH_F_AUTH = (1 << 1),
1060 IP6_FH_F_SKIP_RH = (1 << 2),
1061};
1062
1063/* find specified header and get offset to it */
1064int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target,
1065 unsigned short *fragoff, int *fragflg);
1066
1067int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type);
1068
1069struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1070 const struct ipv6_txoptions *opt,
1071 struct in6_addr *orig);
1072
1073/*
1074 * socket options (ipv6_sockglue.c)
1075 */
1076
1077int ipv6_setsockopt(struct sock *sk, int level, int optname,
1078 char __user *optval, unsigned int optlen);
1079int ipv6_getsockopt(struct sock *sk, int level, int optname,
1080 char __user *optval, int __user *optlen);
1081int compat_ipv6_setsockopt(struct sock *sk, int level, int optname,
1082 char __user *optval, unsigned int optlen);
1083int compat_ipv6_getsockopt(struct sock *sk, int level, int optname,
1084 char __user *optval, int __user *optlen);
1085
1086int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr,
1087 int addr_len);
1088int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len);
1089int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr,
1090 int addr_len);
1091int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr);
1092void ip6_datagram_release_cb(struct sock *sk);
1093
1094int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len,
1095 int *addr_len);
1096int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
1097 int *addr_len);
1098void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
1099 u32 info, u8 *payload);
1100void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info);
1101void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu);
1102
1103int inet6_release(struct socket *sock);
1104int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1105int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
1106 int peer);
1107int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg);
1108
1109int inet6_hash_connect(struct inet_timewait_death_row *death_row,
1110 struct sock *sk);
1111
1112/*
1113 * reassembly.c
1114 */
1115extern const struct proto_ops inet6_stream_ops;
1116extern const struct proto_ops inet6_dgram_ops;
1117extern const struct proto_ops inet6_sockraw_ops;
1118
1119struct group_source_req;
1120struct group_filter;
1121
1122int ip6_mc_source(int add, int omode, struct sock *sk,
1123 struct group_source_req *pgsr);
1124int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf);
1125int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
1126 struct group_filter __user *optval, int __user *optlen);
1127
1128#ifdef CONFIG_PROC_FS
1129int ac6_proc_init(struct net *net);
1130void ac6_proc_exit(struct net *net);
1131int raw6_proc_init(void);
1132void raw6_proc_exit(void);
1133int tcp6_proc_init(struct net *net);
1134void tcp6_proc_exit(struct net *net);
1135int udp6_proc_init(struct net *net);
1136void udp6_proc_exit(struct net *net);
1137int udplite6_proc_init(void);
1138void udplite6_proc_exit(void);
1139int ipv6_misc_proc_init(void);
1140void ipv6_misc_proc_exit(void);
1141int snmp6_register_dev(struct inet6_dev *idev);
1142int snmp6_unregister_dev(struct inet6_dev *idev);
1143
1144#else
1145static inline int ac6_proc_init(struct net *net) { return 0; }
1146static inline void ac6_proc_exit(struct net *net) { }
1147static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; }
1148static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; }
1149#endif
1150
1151#ifdef CONFIG_SYSCTL
1152struct ctl_table *ipv6_icmp_sysctl_init(struct net *net);
1153struct ctl_table *ipv6_route_sysctl_init(struct net *net);
1154int ipv6_sysctl_register(void);
1155void ipv6_sysctl_unregister(void);
1156#endif
1157
1158int ipv6_sock_mc_join(struct sock *sk, int ifindex,
1159 const struct in6_addr *addr);
1160int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
1161 const struct in6_addr *addr, unsigned int mode);
1162int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
1163 const struct in6_addr *addr);
1164#endif /* _NET_IPV6_H */
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * Linux INET6 implementation
4 *
5 * Authors:
6 * Pedro Roque <roque@di.fc.ul.pt>
7 */
8
9#ifndef _NET_IPV6_H
10#define _NET_IPV6_H
11
12#include <linux/ipv6.h>
13#include <linux/hardirq.h>
14#include <linux/jhash.h>
15#include <linux/refcount.h>
16#include <linux/jump_label_ratelimit.h>
17#include <net/if_inet6.h>
18#include <net/ndisc.h>
19#include <net/flow.h>
20#include <net/flow_dissector.h>
21#include <net/snmp.h>
22#include <net/netns/hash.h>
23
24#define SIN6_LEN_RFC2133 24
25
26#define IPV6_MAXPLEN 65535
27
28/*
29 * NextHeader field of IPv6 header
30 */
31
32#define NEXTHDR_HOP 0 /* Hop-by-hop option header. */
33#define NEXTHDR_TCP 6 /* TCP segment. */
34#define NEXTHDR_UDP 17 /* UDP message. */
35#define NEXTHDR_IPV6 41 /* IPv6 in IPv6 */
36#define NEXTHDR_ROUTING 43 /* Routing header. */
37#define NEXTHDR_FRAGMENT 44 /* Fragmentation/reassembly header. */
38#define NEXTHDR_GRE 47 /* GRE header. */
39#define NEXTHDR_ESP 50 /* Encapsulating security payload. */
40#define NEXTHDR_AUTH 51 /* Authentication header. */
41#define NEXTHDR_ICMP 58 /* ICMP for IPv6. */
42#define NEXTHDR_NONE 59 /* No next header */
43#define NEXTHDR_DEST 60 /* Destination options header. */
44#define NEXTHDR_SCTP 132 /* SCTP message. */
45#define NEXTHDR_MOBILITY 135 /* Mobility header. */
46
47#define NEXTHDR_MAX 255
48
49#define IPV6_DEFAULT_HOPLIMIT 64
50#define IPV6_DEFAULT_MCASTHOPS 1
51
52/* Limits on Hop-by-Hop and Destination options.
53 *
54 * Per RFC8200 there is no limit on the maximum number or lengths of options in
55 * Hop-by-Hop or Destination options other then the packet must fit in an MTU.
56 * We allow configurable limits in order to mitigate potential denial of
57 * service attacks.
58 *
59 * There are three limits that may be set:
60 * - Limit the number of options in a Hop-by-Hop or Destination options
61 * extension header
62 * - Limit the byte length of a Hop-by-Hop or Destination options extension
63 * header
64 * - Disallow unknown options
65 *
66 * The limits are expressed in corresponding sysctls:
67 *
68 * ipv6.sysctl.max_dst_opts_cnt
69 * ipv6.sysctl.max_hbh_opts_cnt
70 * ipv6.sysctl.max_dst_opts_len
71 * ipv6.sysctl.max_hbh_opts_len
72 *
73 * max_*_opts_cnt is the number of TLVs that are allowed for Destination
74 * options or Hop-by-Hop options. If the number is less than zero then unknown
75 * TLVs are disallowed and the number of known options that are allowed is the
76 * absolute value. Setting the value to INT_MAX indicates no limit.
77 *
78 * max_*_opts_len is the length limit in bytes of a Destination or
79 * Hop-by-Hop options extension header. Setting the value to INT_MAX
80 * indicates no length limit.
81 *
82 * If a limit is exceeded when processing an extension header the packet is
83 * silently discarded.
84 */
85
86/* Default limits for Hop-by-Hop and Destination options */
87#define IP6_DEFAULT_MAX_DST_OPTS_CNT 8
88#define IP6_DEFAULT_MAX_HBH_OPTS_CNT 8
89#define IP6_DEFAULT_MAX_DST_OPTS_LEN INT_MAX /* No limit */
90#define IP6_DEFAULT_MAX_HBH_OPTS_LEN INT_MAX /* No limit */
91
92/*
93 * Addr type
94 *
95 * type - unicast | multicast
96 * scope - local | site | global
97 * v4 - compat
98 * v4mapped
99 * any
100 * loopback
101 */
102
103#define IPV6_ADDR_ANY 0x0000U
104
105#define IPV6_ADDR_UNICAST 0x0001U
106#define IPV6_ADDR_MULTICAST 0x0002U
107
108#define IPV6_ADDR_LOOPBACK 0x0010U
109#define IPV6_ADDR_LINKLOCAL 0x0020U
110#define IPV6_ADDR_SITELOCAL 0x0040U
111
112#define IPV6_ADDR_COMPATv4 0x0080U
113
114#define IPV6_ADDR_SCOPE_MASK 0x00f0U
115
116#define IPV6_ADDR_MAPPED 0x1000U
117
118/*
119 * Addr scopes
120 */
121#define IPV6_ADDR_MC_SCOPE(a) \
122 ((a)->s6_addr[1] & 0x0f) /* nonstandard */
123#define __IPV6_ADDR_SCOPE_INVALID -1
124#define IPV6_ADDR_SCOPE_NODELOCAL 0x01
125#define IPV6_ADDR_SCOPE_LINKLOCAL 0x02
126#define IPV6_ADDR_SCOPE_SITELOCAL 0x05
127#define IPV6_ADDR_SCOPE_ORGLOCAL 0x08
128#define IPV6_ADDR_SCOPE_GLOBAL 0x0e
129
130/*
131 * Addr flags
132 */
133#define IPV6_ADDR_MC_FLAG_TRANSIENT(a) \
134 ((a)->s6_addr[1] & 0x10)
135#define IPV6_ADDR_MC_FLAG_PREFIX(a) \
136 ((a)->s6_addr[1] & 0x20)
137#define IPV6_ADDR_MC_FLAG_RENDEZVOUS(a) \
138 ((a)->s6_addr[1] & 0x40)
139
140/*
141 * fragmentation header
142 */
143
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, sockptr_t 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 u32 v = (__force u32)a->s6_addr32[0] ^ (__force u32)a->s6_addr32[1];
664
665 return jhash_3words(v,
666 (__force u32)a->s6_addr32[2],
667 (__force u32)a->s6_addr32[3],
668 initval);
669}
670
671static inline bool ipv6_addr_loopback(const struct in6_addr *a)
672{
673#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
674 const __be64 *be = (const __be64 *)a;
675
676 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL;
677#else
678 return (a->s6_addr32[0] | a->s6_addr32[1] |
679 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0;
680#endif
681}
682
683/*
684 * Note that we must __force cast these to unsigned long to make sparse happy,
685 * since all of the endian-annotated types are fixed size regardless of arch.
686 */
687static inline bool ipv6_addr_v4mapped(const struct in6_addr *a)
688{
689 return (
690#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
691 *(unsigned long *)a |
692#else
693 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) |
694#endif
695 (__force unsigned long)(a->s6_addr32[2] ^
696 cpu_to_be32(0x0000ffff))) == 0UL;
697}
698
699static inline bool ipv6_addr_v4mapped_loopback(const struct in6_addr *a)
700{
701 return ipv6_addr_v4mapped(a) && ipv4_is_loopback(a->s6_addr32[3]);
702}
703
704static inline u32 ipv6_portaddr_hash(const struct net *net,
705 const struct in6_addr *addr6,
706 unsigned int port)
707{
708 unsigned int hash, mix = net_hash_mix(net);
709
710 if (ipv6_addr_any(addr6))
711 hash = jhash_1word(0, mix);
712 else if (ipv6_addr_v4mapped(addr6))
713 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
714 else
715 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
716
717 return hash ^ port;
718}
719
720/*
721 * Check for a RFC 4843 ORCHID address
722 * (Overlay Routable Cryptographic Hash Identifiers)
723 */
724static inline bool ipv6_addr_orchid(const struct in6_addr *a)
725{
726 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010);
727}
728
729static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr)
730{
731 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000);
732}
733
734static inline void ipv6_addr_set_v4mapped(const __be32 addr,
735 struct in6_addr *v4mapped)
736{
737 ipv6_addr_set(v4mapped,
738 0, 0,
739 htonl(0x0000FFFF),
740 addr);
741}
742
743/*
744 * find the first different bit between two addresses
745 * length of address must be a multiple of 32bits
746 */
747static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen)
748{
749 const __be32 *a1 = token1, *a2 = token2;
750 int i;
751
752 addrlen >>= 2;
753
754 for (i = 0; i < addrlen; i++) {
755 __be32 xb = a1[i] ^ a2[i];
756 if (xb)
757 return i * 32 + 31 - __fls(ntohl(xb));
758 }
759
760 /*
761 * we should *never* get to this point since that
762 * would mean the addrs are equal
763 *
764 * However, we do get to it 8) And exacly, when
765 * addresses are equal 8)
766 *
767 * ip route add 1111::/128 via ...
768 * ip route add 1111::/64 via ...
769 * and we are here.
770 *
771 * Ideally, this function should stop comparison
772 * at prefix length. It does not, but it is still OK,
773 * if returned value is greater than prefix length.
774 * --ANK (980803)
775 */
776 return addrlen << 5;
777}
778
779#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
780static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen)
781{
782 const __be64 *a1 = token1, *a2 = token2;
783 int i;
784
785 addrlen >>= 3;
786
787 for (i = 0; i < addrlen; i++) {
788 __be64 xb = a1[i] ^ a2[i];
789 if (xb)
790 return i * 64 + 63 - __fls(be64_to_cpu(xb));
791 }
792
793 return addrlen << 6;
794}
795#endif
796
797static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen)
798{
799#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
800 if (__builtin_constant_p(addrlen) && !(addrlen & 7))
801 return __ipv6_addr_diff64(token1, token2, addrlen);
802#endif
803 return __ipv6_addr_diff32(token1, token2, addrlen);
804}
805
806static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2)
807{
808 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr));
809}
810
811__be32 ipv6_select_ident(struct net *net,
812 const struct in6_addr *daddr,
813 const struct in6_addr *saddr);
814__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb);
815
816int ip6_dst_hoplimit(struct dst_entry *dst);
817
818static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
819 struct dst_entry *dst)
820{
821 int hlimit;
822
823 if (ipv6_addr_is_multicast(&fl6->daddr))
824 hlimit = np->mcast_hops;
825 else
826 hlimit = np->hop_limit;
827 if (hlimit < 0)
828 hlimit = ip6_dst_hoplimit(dst);
829 return hlimit;
830}
831
832/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store
833 * Equivalent to : flow->v6addrs.src = iph->saddr;
834 * flow->v6addrs.dst = iph->daddr;
835 */
836static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow,
837 const struct ipv6hdr *iph)
838{
839 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) !=
840 offsetof(typeof(flow->addrs), v6addrs.src) +
841 sizeof(flow->addrs.v6addrs.src));
842 memcpy(&flow->addrs.v6addrs, &iph->saddr, sizeof(flow->addrs.v6addrs));
843 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
844}
845
846#if IS_ENABLED(CONFIG_IPV6)
847
848static inline bool ipv6_can_nonlocal_bind(struct net *net,
849 struct inet_sock *inet)
850{
851 return net->ipv6.sysctl.ip_nonlocal_bind ||
852 inet->freebind || inet->transparent;
853}
854
855/* Sysctl settings for net ipv6.auto_flowlabels */
856#define IP6_AUTO_FLOW_LABEL_OFF 0
857#define IP6_AUTO_FLOW_LABEL_OPTOUT 1
858#define IP6_AUTO_FLOW_LABEL_OPTIN 2
859#define IP6_AUTO_FLOW_LABEL_FORCED 3
860
861#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED
862
863#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT
864
865static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
866 __be32 flowlabel, bool autolabel,
867 struct flowi6 *fl6)
868{
869 u32 hash;
870
871 /* @flowlabel may include more than a flow label, eg, the traffic class.
872 * Here we want only the flow label value.
873 */
874 flowlabel &= IPV6_FLOWLABEL_MASK;
875
876 if (flowlabel ||
877 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF ||
878 (!autolabel &&
879 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED))
880 return flowlabel;
881
882 hash = skb_get_hash_flowi6(skb, fl6);
883
884 /* Since this is being sent on the wire obfuscate hash a bit
885 * to minimize possbility that any useful information to an
886 * attacker is leaked. Only lower 20 bits are relevant.
887 */
888 hash = rol32(hash, 16);
889
890 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
891
892 if (net->ipv6.sysctl.flowlabel_state_ranges)
893 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG;
894
895 return flowlabel;
896}
897
898static inline int ip6_default_np_autolabel(struct net *net)
899{
900 switch (net->ipv6.sysctl.auto_flowlabels) {
901 case IP6_AUTO_FLOW_LABEL_OFF:
902 case IP6_AUTO_FLOW_LABEL_OPTIN:
903 default:
904 return 0;
905 case IP6_AUTO_FLOW_LABEL_OPTOUT:
906 case IP6_AUTO_FLOW_LABEL_FORCED:
907 return 1;
908 }
909}
910#else
911static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
912 __be32 flowlabel, bool autolabel,
913 struct flowi6 *fl6)
914{
915 return flowlabel;
916}
917static inline int ip6_default_np_autolabel(struct net *net)
918{
919 return 0;
920}
921#endif
922
923#if IS_ENABLED(CONFIG_IPV6)
924static inline int ip6_multipath_hash_policy(const struct net *net)
925{
926 return net->ipv6.sysctl.multipath_hash_policy;
927}
928#else
929static inline int ip6_multipath_hash_policy(const struct net *net)
930{
931 return 0;
932}
933#endif
934
935/*
936 * Header manipulation
937 */
938static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass,
939 __be32 flowlabel)
940{
941 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel;
942}
943
944static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr)
945{
946 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK;
947}
948
949static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr)
950{
951 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK;
952}
953
954static inline u8 ip6_tclass(__be32 flowinfo)
955{
956 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT;
957}
958
959static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel)
960{
961 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel;
962}
963
964static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6)
965{
966 return fl6->flowlabel & IPV6_FLOWLABEL_MASK;
967}
968
969/*
970 * Prototypes exported by ipv6
971 */
972
973/*
974 * rcv function (called from netdevice level)
975 */
976
977int ipv6_rcv(struct sk_buff *skb, struct net_device *dev,
978 struct packet_type *pt, struct net_device *orig_dev);
979void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
980 struct net_device *orig_dev);
981
982int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb);
983
984/*
985 * upper-layer output functions
986 */
987int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
988 __u32 mark, struct ipv6_txoptions *opt, int tclass, u32 priority);
989
990int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr);
991
992int ip6_append_data(struct sock *sk,
993 int getfrag(void *from, char *to, int offset, int len,
994 int odd, struct sk_buff *skb),
995 void *from, int length, int transhdrlen,
996 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
997 struct rt6_info *rt, unsigned int flags);
998
999int ip6_push_pending_frames(struct sock *sk);
1000
1001void ip6_flush_pending_frames(struct sock *sk);
1002
1003int ip6_send_skb(struct sk_buff *skb);
1004
1005struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue,
1006 struct inet_cork_full *cork,
1007 struct inet6_cork *v6_cork);
1008struct sk_buff *ip6_make_skb(struct sock *sk,
1009 int getfrag(void *from, char *to, int offset,
1010 int len, int odd, struct sk_buff *skb),
1011 void *from, int length, int transhdrlen,
1012 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1013 struct rt6_info *rt, unsigned int flags,
1014 struct inet_cork_full *cork);
1015
1016static inline struct sk_buff *ip6_finish_skb(struct sock *sk)
1017{
1018 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork,
1019 &inet6_sk(sk)->cork);
1020}
1021
1022int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1023 struct flowi6 *fl6);
1024struct dst_entry *ip6_dst_lookup_flow(struct net *net, const struct sock *sk, struct flowi6 *fl6,
1025 const struct in6_addr *final_dst);
1026struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
1027 const struct in6_addr *final_dst,
1028 bool connected);
1029struct dst_entry *ip6_dst_lookup_tunnel(struct sk_buff *skb,
1030 struct net_device *dev,
1031 struct net *net, struct socket *sock,
1032 struct in6_addr *saddr,
1033 const struct ip_tunnel_info *info,
1034 u8 protocol, bool use_cache);
1035struct dst_entry *ip6_blackhole_route(struct net *net,
1036 struct dst_entry *orig_dst);
1037
1038/*
1039 * skb processing functions
1040 */
1041
1042int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb);
1043int ip6_forward(struct sk_buff *skb);
1044int ip6_input(struct sk_buff *skb);
1045int ip6_mc_input(struct sk_buff *skb);
1046void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
1047 bool have_final);
1048
1049int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1050int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1051
1052/*
1053 * Extension header (options) processing
1054 */
1055
1056void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1057 u8 *proto, struct in6_addr **daddr_p,
1058 struct in6_addr *saddr);
1059void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1060 u8 *proto);
1061
1062int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp,
1063 __be16 *frag_offp);
1064
1065bool ipv6_ext_hdr(u8 nexthdr);
1066
1067enum {
1068 IP6_FH_F_FRAG = (1 << 0),
1069 IP6_FH_F_AUTH = (1 << 1),
1070 IP6_FH_F_SKIP_RH = (1 << 2),
1071};
1072
1073/* find specified header and get offset to it */
1074int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target,
1075 unsigned short *fragoff, int *fragflg);
1076
1077int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type);
1078
1079struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1080 const struct ipv6_txoptions *opt,
1081 struct in6_addr *orig);
1082
1083/*
1084 * socket options (ipv6_sockglue.c)
1085 */
1086
1087int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1088 unsigned int optlen);
1089int ipv6_getsockopt(struct sock *sk, int level, int optname,
1090 char __user *optval, int __user *optlen);
1091
1092int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr,
1093 int addr_len);
1094int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len);
1095int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr,
1096 int addr_len);
1097int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr);
1098void ip6_datagram_release_cb(struct sock *sk);
1099
1100int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len,
1101 int *addr_len);
1102int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
1103 int *addr_len);
1104void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
1105 u32 info, u8 *payload);
1106void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info);
1107void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu);
1108
1109int inet6_release(struct socket *sock);
1110int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1111int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
1112 int peer);
1113int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg);
1114int inet6_compat_ioctl(struct socket *sock, unsigned int cmd,
1115 unsigned long arg);
1116
1117int inet6_hash_connect(struct inet_timewait_death_row *death_row,
1118 struct sock *sk);
1119int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size);
1120int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1121 int flags);
1122
1123/*
1124 * reassembly.c
1125 */
1126extern const struct proto_ops inet6_stream_ops;
1127extern const struct proto_ops inet6_dgram_ops;
1128extern const struct proto_ops inet6_sockraw_ops;
1129
1130struct group_source_req;
1131struct group_filter;
1132
1133int ip6_mc_source(int add, int omode, struct sock *sk,
1134 struct group_source_req *pgsr);
1135int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
1136 struct sockaddr_storage *list);
1137int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
1138 struct sockaddr_storage __user *p);
1139
1140#ifdef CONFIG_PROC_FS
1141int ac6_proc_init(struct net *net);
1142void ac6_proc_exit(struct net *net);
1143int raw6_proc_init(void);
1144void raw6_proc_exit(void);
1145int tcp6_proc_init(struct net *net);
1146void tcp6_proc_exit(struct net *net);
1147int udp6_proc_init(struct net *net);
1148void udp6_proc_exit(struct net *net);
1149int udplite6_proc_init(void);
1150void udplite6_proc_exit(void);
1151int ipv6_misc_proc_init(void);
1152void ipv6_misc_proc_exit(void);
1153int snmp6_register_dev(struct inet6_dev *idev);
1154int snmp6_unregister_dev(struct inet6_dev *idev);
1155
1156#else
1157static inline int ac6_proc_init(struct net *net) { return 0; }
1158static inline void ac6_proc_exit(struct net *net) { }
1159static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; }
1160static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; }
1161#endif
1162
1163#ifdef CONFIG_SYSCTL
1164struct ctl_table *ipv6_icmp_sysctl_init(struct net *net);
1165struct ctl_table *ipv6_route_sysctl_init(struct net *net);
1166int ipv6_sysctl_register(void);
1167void ipv6_sysctl_unregister(void);
1168#endif
1169
1170int ipv6_sock_mc_join(struct sock *sk, int ifindex,
1171 const struct in6_addr *addr);
1172int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
1173 const struct in6_addr *addr, unsigned int mode);
1174int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
1175 const struct in6_addr *addr);
1176
1177static inline int ip6_sock_set_v6only(struct sock *sk)
1178{
1179 if (inet_sk(sk)->inet_num)
1180 return -EINVAL;
1181 lock_sock(sk);
1182 sk->sk_ipv6only = true;
1183 release_sock(sk);
1184 return 0;
1185}
1186
1187static inline void ip6_sock_set_recverr(struct sock *sk)
1188{
1189 lock_sock(sk);
1190 inet6_sk(sk)->recverr = true;
1191 release_sock(sk);
1192}
1193
1194static inline int __ip6_sock_set_addr_preferences(struct sock *sk, int val)
1195{
1196 unsigned int pref = 0;
1197 unsigned int prefmask = ~0;
1198
1199 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
1200 switch (val & (IPV6_PREFER_SRC_PUBLIC |
1201 IPV6_PREFER_SRC_TMP |
1202 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
1203 case IPV6_PREFER_SRC_PUBLIC:
1204 pref |= IPV6_PREFER_SRC_PUBLIC;
1205 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1206 IPV6_PREFER_SRC_TMP);
1207 break;
1208 case IPV6_PREFER_SRC_TMP:
1209 pref |= IPV6_PREFER_SRC_TMP;
1210 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1211 IPV6_PREFER_SRC_TMP);
1212 break;
1213 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
1214 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1215 IPV6_PREFER_SRC_TMP);
1216 break;
1217 case 0:
1218 break;
1219 default:
1220 return -EINVAL;
1221 }
1222
1223 /* check HOME/COA conflicts */
1224 switch (val & (IPV6_PREFER_SRC_HOME | IPV6_PREFER_SRC_COA)) {
1225 case IPV6_PREFER_SRC_HOME:
1226 prefmask &= ~IPV6_PREFER_SRC_COA;
1227 break;
1228 case IPV6_PREFER_SRC_COA:
1229 pref |= IPV6_PREFER_SRC_COA;
1230 break;
1231 case 0:
1232 break;
1233 default:
1234 return -EINVAL;
1235 }
1236
1237 /* check CGA/NONCGA conflicts */
1238 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
1239 case IPV6_PREFER_SRC_CGA:
1240 case IPV6_PREFER_SRC_NONCGA:
1241 case 0:
1242 break;
1243 default:
1244 return -EINVAL;
1245 }
1246
1247 inet6_sk(sk)->srcprefs = (inet6_sk(sk)->srcprefs & prefmask) | pref;
1248 return 0;
1249}
1250
1251static inline int ip6_sock_set_addr_preferences(struct sock *sk, bool val)
1252{
1253 int ret;
1254
1255 lock_sock(sk);
1256 ret = __ip6_sock_set_addr_preferences(sk, val);
1257 release_sock(sk);
1258 return ret;
1259}
1260
1261static inline void ip6_sock_set_recvpktinfo(struct sock *sk)
1262{
1263 lock_sock(sk);
1264 inet6_sk(sk)->rxopt.bits.rxinfo = true;
1265 release_sock(sk);
1266}
1267
1268#endif /* _NET_IPV6_H */