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/flow.h>
19#include <net/flow_dissector.h>
20#include <net/inet_dscp.h>
21#include <net/snmp.h>
22#include <net/netns/hash.h>
23
24struct ip_tunnel_info;
25
26#define SIN6_LEN_RFC2133 24
27
28#define IPV6_MAXPLEN 65535
29
30/*
31 * NextHeader field of IPv6 header
32 */
33
34#define NEXTHDR_HOP 0 /* Hop-by-hop option header. */
35#define NEXTHDR_IPV4 4 /* IPv4 in IPv6 */
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/*
155 * Jumbo payload option, as described in RFC 2675 2.
156 */
157struct hop_jumbo_hdr {
158 u8 nexthdr;
159 u8 hdrlen;
160 u8 tlv_type; /* IPV6_TLV_JUMBO, 0xC2 */
161 u8 tlv_len; /* 4 */
162 __be32 jumbo_payload_len;
163};
164
165#define IP6_MF 0x0001
166#define IP6_OFFSET 0xFFF8
167
168struct ip6_fraglist_iter {
169 struct ipv6hdr *tmp_hdr;
170 struct sk_buff *frag;
171 int offset;
172 unsigned int hlen;
173 __be32 frag_id;
174 u8 nexthdr;
175};
176
177int ip6_fraglist_init(struct sk_buff *skb, unsigned int hlen, u8 *prevhdr,
178 u8 nexthdr, __be32 frag_id,
179 struct ip6_fraglist_iter *iter);
180void ip6_fraglist_prepare(struct sk_buff *skb, struct ip6_fraglist_iter *iter);
181
182static inline struct sk_buff *ip6_fraglist_next(struct ip6_fraglist_iter *iter)
183{
184 struct sk_buff *skb = iter->frag;
185
186 iter->frag = skb->next;
187 skb_mark_not_on_list(skb);
188
189 return skb;
190}
191
192struct ip6_frag_state {
193 u8 *prevhdr;
194 unsigned int hlen;
195 unsigned int mtu;
196 unsigned int left;
197 int offset;
198 int ptr;
199 int hroom;
200 int troom;
201 __be32 frag_id;
202 u8 nexthdr;
203};
204
205void ip6_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int mtu,
206 unsigned short needed_tailroom, int hdr_room, u8 *prevhdr,
207 u8 nexthdr, __be32 frag_id, struct ip6_frag_state *state);
208struct sk_buff *ip6_frag_next(struct sk_buff *skb,
209 struct ip6_frag_state *state);
210
211#define IP6_REPLY_MARK(net, mark) \
212 ((net)->ipv6.sysctl.fwmark_reflect ? (mark) : 0)
213
214#include <net/sock.h>
215
216/* sysctls */
217extern int sysctl_mld_max_msf;
218extern int sysctl_mld_qrv;
219
220#define _DEVINC(net, statname, mod, idev, field) \
221({ \
222 struct inet6_dev *_idev = (idev); \
223 if (likely(_idev != NULL)) \
224 mod##SNMP_INC_STATS64((_idev)->stats.statname, (field));\
225 mod##SNMP_INC_STATS64((net)->mib.statname##_statistics, (field));\
226})
227
228/* per device counters are atomic_long_t */
229#define _DEVINCATOMIC(net, statname, mod, idev, field) \
230({ \
231 struct inet6_dev *_idev = (idev); \
232 if (likely(_idev != NULL)) \
233 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
234 mod##SNMP_INC_STATS((net)->mib.statname##_statistics, (field));\
235})
236
237/* per device and per net counters are atomic_long_t */
238#define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \
239({ \
240 struct inet6_dev *_idev = (idev); \
241 if (likely(_idev != NULL)) \
242 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
243 SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\
244})
245
246#define _DEVADD(net, statname, mod, idev, field, val) \
247({ \
248 struct inet6_dev *_idev = (idev); \
249 if (likely(_idev != NULL)) \
250 mod##SNMP_ADD_STATS((_idev)->stats.statname, (field), (val)); \
251 mod##SNMP_ADD_STATS((net)->mib.statname##_statistics, (field), (val));\
252})
253
254#define _DEVUPD(net, statname, mod, idev, field, val) \
255({ \
256 struct inet6_dev *_idev = (idev); \
257 if (likely(_idev != NULL)) \
258 mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \
259 mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
260})
261
262/* MIBs */
263
264#define IP6_INC_STATS(net, idev,field) \
265 _DEVINC(net, ipv6, , idev, field)
266#define __IP6_INC_STATS(net, idev,field) \
267 _DEVINC(net, ipv6, __, idev, field)
268#define IP6_ADD_STATS(net, idev,field,val) \
269 _DEVADD(net, ipv6, , idev, field, val)
270#define __IP6_ADD_STATS(net, idev,field,val) \
271 _DEVADD(net, ipv6, __, idev, field, val)
272#define IP6_UPD_PO_STATS(net, idev,field,val) \
273 _DEVUPD(net, ipv6, , idev, field, val)
274#define __IP6_UPD_PO_STATS(net, idev,field,val) \
275 _DEVUPD(net, ipv6, __, idev, field, val)
276#define ICMP6_INC_STATS(net, idev, field) \
277 _DEVINCATOMIC(net, icmpv6, , idev, field)
278#define __ICMP6_INC_STATS(net, idev, field) \
279 _DEVINCATOMIC(net, icmpv6, __, idev, field)
280
281#define ICMP6MSGOUT_INC_STATS(net, idev, field) \
282 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256)
283#define ICMP6MSGIN_INC_STATS(net, idev, field) \
284 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field)
285
286struct ip6_ra_chain {
287 struct ip6_ra_chain *next;
288 struct sock *sk;
289 int sel;
290 void (*destructor)(struct sock *);
291};
292
293extern struct ip6_ra_chain *ip6_ra_chain;
294extern rwlock_t ip6_ra_lock;
295
296/*
297 This structure is prepared by protocol, when parsing
298 ancillary data and passed to IPv6.
299 */
300
301struct ipv6_txoptions {
302 refcount_t refcnt;
303 /* Length of this structure */
304 int tot_len;
305
306 /* length of extension headers */
307
308 __u16 opt_flen; /* after fragment hdr */
309 __u16 opt_nflen; /* before fragment hdr */
310
311 struct ipv6_opt_hdr *hopopt;
312 struct ipv6_opt_hdr *dst0opt;
313 struct ipv6_rt_hdr *srcrt; /* Routing Header */
314 struct ipv6_opt_hdr *dst1opt;
315 struct rcu_head rcu;
316 /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */
317};
318
319/* flowlabel_reflect sysctl values */
320enum flowlabel_reflect {
321 FLOWLABEL_REFLECT_ESTABLISHED = 1,
322 FLOWLABEL_REFLECT_TCP_RESET = 2,
323 FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES = 4,
324};
325
326struct ip6_flowlabel {
327 struct ip6_flowlabel __rcu *next;
328 __be32 label;
329 atomic_t users;
330 struct in6_addr dst;
331 struct ipv6_txoptions *opt;
332 unsigned long linger;
333 struct rcu_head rcu;
334 u8 share;
335 union {
336 struct pid *pid;
337 kuid_t uid;
338 } owner;
339 unsigned long lastuse;
340 unsigned long expires;
341 struct net *fl_net;
342};
343
344#define IPV6_FLOWINFO_MASK cpu_to_be32(0x0FFFFFFF)
345#define IPV6_FLOWLABEL_MASK cpu_to_be32(0x000FFFFF)
346#define IPV6_FLOWLABEL_STATELESS_FLAG cpu_to_be32(0x00080000)
347
348#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
349#define IPV6_TCLASS_SHIFT 20
350
351struct ipv6_fl_socklist {
352 struct ipv6_fl_socklist __rcu *next;
353 struct ip6_flowlabel *fl;
354 struct rcu_head rcu;
355};
356
357struct ipcm6_cookie {
358 struct sockcm_cookie sockc;
359 __s16 hlimit;
360 __s16 tclass;
361 __u16 gso_size;
362 __s8 dontfrag;
363 struct ipv6_txoptions *opt;
364};
365
366static inline void ipcm6_init(struct ipcm6_cookie *ipc6)
367{
368 *ipc6 = (struct ipcm6_cookie) {
369 .hlimit = -1,
370 .tclass = -1,
371 .dontfrag = -1,
372 };
373}
374
375static inline void ipcm6_init_sk(struct ipcm6_cookie *ipc6,
376 const struct sock *sk)
377{
378 *ipc6 = (struct ipcm6_cookie) {
379 .hlimit = -1,
380 .tclass = inet6_sk(sk)->tclass,
381 .dontfrag = inet6_test_bit(DONTFRAG, sk),
382 };
383}
384
385static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np)
386{
387 struct ipv6_txoptions *opt;
388
389 rcu_read_lock();
390 opt = rcu_dereference(np->opt);
391 if (opt) {
392 if (!refcount_inc_not_zero(&opt->refcnt))
393 opt = NULL;
394 else
395 opt = rcu_pointer_handoff(opt);
396 }
397 rcu_read_unlock();
398 return opt;
399}
400
401static inline void txopt_put(struct ipv6_txoptions *opt)
402{
403 if (opt && refcount_dec_and_test(&opt->refcnt))
404 kfree_rcu(opt, rcu);
405}
406
407#if IS_ENABLED(CONFIG_IPV6)
408struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label);
409
410extern struct static_key_false_deferred ipv6_flowlabel_exclusive;
411static inline struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk,
412 __be32 label)
413{
414 if (static_branch_unlikely(&ipv6_flowlabel_exclusive.key) &&
415 READ_ONCE(sock_net(sk)->ipv6.flowlabel_has_excl))
416 return __fl6_sock_lookup(sk, label) ? : ERR_PTR(-ENOENT);
417
418 return NULL;
419}
420#endif
421
422struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space,
423 struct ip6_flowlabel *fl,
424 struct ipv6_txoptions *fopt);
425void fl6_free_socklist(struct sock *sk);
426int ipv6_flowlabel_opt(struct sock *sk, sockptr_t optval, int optlen);
427int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
428 int flags);
429int ip6_flowlabel_init(void);
430void ip6_flowlabel_cleanup(void);
431bool ip6_autoflowlabel(struct net *net, const struct sock *sk);
432
433static inline void fl6_sock_release(struct ip6_flowlabel *fl)
434{
435 if (fl)
436 atomic_dec(&fl->users);
437}
438
439enum skb_drop_reason icmpv6_notify(struct sk_buff *skb, u8 type,
440 u8 code, __be32 info);
441
442void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
443 struct icmp6hdr *thdr, int len);
444
445int ip6_ra_control(struct sock *sk, int sel);
446
447int ipv6_parse_hopopts(struct sk_buff *skb);
448
449struct ipv6_txoptions *ipv6_dup_options(struct sock *sk,
450 struct ipv6_txoptions *opt);
451struct ipv6_txoptions *ipv6_renew_options(struct sock *sk,
452 struct ipv6_txoptions *opt,
453 int newtype,
454 struct ipv6_opt_hdr *newopt);
455struct ipv6_txoptions *__ipv6_fixup_options(struct ipv6_txoptions *opt_space,
456 struct ipv6_txoptions *opt);
457
458static inline struct ipv6_txoptions *
459ipv6_fixup_options(struct ipv6_txoptions *opt_space, struct ipv6_txoptions *opt)
460{
461 if (!opt)
462 return NULL;
463 return __ipv6_fixup_options(opt_space, opt);
464}
465
466bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
467 const struct inet6_skb_parm *opt);
468struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
469 struct ipv6_txoptions *opt);
470
471/* This helper is specialized for BIG TCP needs.
472 * It assumes the hop_jumbo_hdr will immediately follow the IPV6 header.
473 * It assumes headers are already in skb->head.
474 * Returns 0, or IPPROTO_TCP if a BIG TCP packet is there.
475 */
476static inline int ipv6_has_hopopt_jumbo(const struct sk_buff *skb)
477{
478 const struct hop_jumbo_hdr *jhdr;
479 const struct ipv6hdr *nhdr;
480
481 if (likely(skb->len <= GRO_LEGACY_MAX_SIZE))
482 return 0;
483
484 if (skb->protocol != htons(ETH_P_IPV6))
485 return 0;
486
487 if (skb_network_offset(skb) +
488 sizeof(struct ipv6hdr) +
489 sizeof(struct hop_jumbo_hdr) > skb_headlen(skb))
490 return 0;
491
492 nhdr = ipv6_hdr(skb);
493
494 if (nhdr->nexthdr != NEXTHDR_HOP)
495 return 0;
496
497 jhdr = (const struct hop_jumbo_hdr *) (nhdr + 1);
498 if (jhdr->tlv_type != IPV6_TLV_JUMBO || jhdr->hdrlen != 0 ||
499 jhdr->nexthdr != IPPROTO_TCP)
500 return 0;
501 return jhdr->nexthdr;
502}
503
504/* Return 0 if HBH header is successfully removed
505 * Or if HBH removal is unnecessary (packet is not big TCP)
506 * Return error to indicate dropping the packet
507 */
508static inline int ipv6_hopopt_jumbo_remove(struct sk_buff *skb)
509{
510 const int hophdr_len = sizeof(struct hop_jumbo_hdr);
511 int nexthdr = ipv6_has_hopopt_jumbo(skb);
512 struct ipv6hdr *h6;
513
514 if (!nexthdr)
515 return 0;
516
517 if (skb_cow_head(skb, 0))
518 return -1;
519
520 /* Remove the HBH header.
521 * Layout: [Ethernet header][IPv6 header][HBH][L4 Header]
522 */
523 memmove(skb_mac_header(skb) + hophdr_len, skb_mac_header(skb),
524 skb_network_header(skb) - skb_mac_header(skb) +
525 sizeof(struct ipv6hdr));
526
527 __skb_pull(skb, hophdr_len);
528 skb->network_header += hophdr_len;
529 skb->mac_header += hophdr_len;
530
531 h6 = ipv6_hdr(skb);
532 h6->nexthdr = nexthdr;
533
534 return 0;
535}
536
537static inline bool ipv6_accept_ra(const struct inet6_dev *idev)
538{
539 s32 accept_ra = READ_ONCE(idev->cnf.accept_ra);
540
541 /* If forwarding is enabled, RA are not accepted unless the special
542 * hybrid mode (accept_ra=2) is enabled.
543 */
544 return READ_ONCE(idev->cnf.forwarding) ? accept_ra == 2 :
545 accept_ra;
546}
547
548#define IPV6_FRAG_HIGH_THRESH (4 * 1024*1024) /* 4194304 */
549#define IPV6_FRAG_LOW_THRESH (3 * 1024*1024) /* 3145728 */
550#define IPV6_FRAG_TIMEOUT (60 * HZ) /* 60 seconds */
551
552int __ipv6_addr_type(const struct in6_addr *addr);
553static inline int ipv6_addr_type(const struct in6_addr *addr)
554{
555 return __ipv6_addr_type(addr) & 0xffff;
556}
557
558static inline int ipv6_addr_scope(const struct in6_addr *addr)
559{
560 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK;
561}
562
563static inline int __ipv6_addr_src_scope(int type)
564{
565 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16);
566}
567
568static inline int ipv6_addr_src_scope(const struct in6_addr *addr)
569{
570 return __ipv6_addr_src_scope(__ipv6_addr_type(addr));
571}
572
573static inline bool __ipv6_addr_needs_scope_id(int type)
574{
575 return type & IPV6_ADDR_LINKLOCAL ||
576 (type & IPV6_ADDR_MULTICAST &&
577 (type & (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)));
578}
579
580static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr, int iface)
581{
582 return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0;
583}
584
585static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2)
586{
587 return memcmp(a1, a2, sizeof(struct in6_addr));
588}
589
590static inline bool
591ipv6_masked_addr_cmp(const struct in6_addr *a1, const struct in6_addr *m,
592 const struct in6_addr *a2)
593{
594#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
595 const unsigned long *ul1 = (const unsigned long *)a1;
596 const unsigned long *ulm = (const unsigned long *)m;
597 const unsigned long *ul2 = (const unsigned long *)a2;
598
599 return !!(((ul1[0] ^ ul2[0]) & ulm[0]) |
600 ((ul1[1] ^ ul2[1]) & ulm[1]));
601#else
602 return !!(((a1->s6_addr32[0] ^ a2->s6_addr32[0]) & m->s6_addr32[0]) |
603 ((a1->s6_addr32[1] ^ a2->s6_addr32[1]) & m->s6_addr32[1]) |
604 ((a1->s6_addr32[2] ^ a2->s6_addr32[2]) & m->s6_addr32[2]) |
605 ((a1->s6_addr32[3] ^ a2->s6_addr32[3]) & m->s6_addr32[3]));
606#endif
607}
608
609static inline void ipv6_addr_prefix(struct in6_addr *pfx,
610 const struct in6_addr *addr,
611 int plen)
612{
613 /* caller must guarantee 0 <= plen <= 128 */
614 int o = plen >> 3,
615 b = plen & 0x7;
616
617 memset(pfx->s6_addr, 0, sizeof(pfx->s6_addr));
618 memcpy(pfx->s6_addr, addr, o);
619 if (b != 0)
620 pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b);
621}
622
623static inline void ipv6_addr_prefix_copy(struct in6_addr *addr,
624 const struct in6_addr *pfx,
625 int plen)
626{
627 /* caller must guarantee 0 <= plen <= 128 */
628 int o = plen >> 3,
629 b = plen & 0x7;
630
631 memcpy(addr->s6_addr, pfx, o);
632 if (b != 0) {
633 addr->s6_addr[o] &= ~(0xff00 >> b);
634 addr->s6_addr[o] |= (pfx->s6_addr[o] & (0xff00 >> b));
635 }
636}
637
638static inline void __ipv6_addr_set_half(__be32 *addr,
639 __be32 wh, __be32 wl)
640{
641#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
642#if defined(__BIG_ENDIAN)
643 if (__builtin_constant_p(wh) && __builtin_constant_p(wl)) {
644 *(__force u64 *)addr = ((__force u64)(wh) << 32 | (__force u64)(wl));
645 return;
646 }
647#elif defined(__LITTLE_ENDIAN)
648 if (__builtin_constant_p(wl) && __builtin_constant_p(wh)) {
649 *(__force u64 *)addr = ((__force u64)(wl) << 32 | (__force u64)(wh));
650 return;
651 }
652#endif
653#endif
654 addr[0] = wh;
655 addr[1] = wl;
656}
657
658static inline void ipv6_addr_set(struct in6_addr *addr,
659 __be32 w1, __be32 w2,
660 __be32 w3, __be32 w4)
661{
662 __ipv6_addr_set_half(&addr->s6_addr32[0], w1, w2);
663 __ipv6_addr_set_half(&addr->s6_addr32[2], w3, w4);
664}
665
666static inline bool ipv6_addr_equal(const struct in6_addr *a1,
667 const struct in6_addr *a2)
668{
669#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
670 const unsigned long *ul1 = (const unsigned long *)a1;
671 const unsigned long *ul2 = (const unsigned long *)a2;
672
673 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL;
674#else
675 return ((a1->s6_addr32[0] ^ a2->s6_addr32[0]) |
676 (a1->s6_addr32[1] ^ a2->s6_addr32[1]) |
677 (a1->s6_addr32[2] ^ a2->s6_addr32[2]) |
678 (a1->s6_addr32[3] ^ a2->s6_addr32[3])) == 0;
679#endif
680}
681
682#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
683static inline bool __ipv6_prefix_equal64_half(const __be64 *a1,
684 const __be64 *a2,
685 unsigned int len)
686{
687 if (len && ((*a1 ^ *a2) & cpu_to_be64((~0UL) << (64 - len))))
688 return false;
689 return true;
690}
691
692static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
693 const struct in6_addr *addr2,
694 unsigned int prefixlen)
695{
696 const __be64 *a1 = (const __be64 *)addr1;
697 const __be64 *a2 = (const __be64 *)addr2;
698
699 if (prefixlen >= 64) {
700 if (a1[0] ^ a2[0])
701 return false;
702 return __ipv6_prefix_equal64_half(a1 + 1, a2 + 1, prefixlen - 64);
703 }
704 return __ipv6_prefix_equal64_half(a1, a2, prefixlen);
705}
706#else
707static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
708 const struct in6_addr *addr2,
709 unsigned int prefixlen)
710{
711 const __be32 *a1 = addr1->s6_addr32;
712 const __be32 *a2 = addr2->s6_addr32;
713 unsigned int pdw, pbi;
714
715 /* check complete u32 in prefix */
716 pdw = prefixlen >> 5;
717 if (pdw && memcmp(a1, a2, pdw << 2))
718 return false;
719
720 /* check incomplete u32 in prefix */
721 pbi = prefixlen & 0x1f;
722 if (pbi && ((a1[pdw] ^ a2[pdw]) & htonl((0xffffffff) << (32 - pbi))))
723 return false;
724
725 return true;
726}
727#endif
728
729static inline bool ipv6_addr_any(const struct in6_addr *a)
730{
731#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
732 const unsigned long *ul = (const unsigned long *)a;
733
734 return (ul[0] | ul[1]) == 0UL;
735#else
736 return (a->s6_addr32[0] | a->s6_addr32[1] |
737 a->s6_addr32[2] | a->s6_addr32[3]) == 0;
738#endif
739}
740
741static inline u32 ipv6_addr_hash(const struct in6_addr *a)
742{
743#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
744 const unsigned long *ul = (const unsigned long *)a;
745 unsigned long x = ul[0] ^ ul[1];
746
747 return (u32)(x ^ (x >> 32));
748#else
749 return (__force u32)(a->s6_addr32[0] ^ a->s6_addr32[1] ^
750 a->s6_addr32[2] ^ a->s6_addr32[3]);
751#endif
752}
753
754/* more secured version of ipv6_addr_hash() */
755static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval)
756{
757 return jhash2((__force const u32 *)a->s6_addr32,
758 ARRAY_SIZE(a->s6_addr32), initval);
759}
760
761static inline bool ipv6_addr_loopback(const struct in6_addr *a)
762{
763#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
764 const __be64 *be = (const __be64 *)a;
765
766 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL;
767#else
768 return (a->s6_addr32[0] | a->s6_addr32[1] |
769 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0;
770#endif
771}
772
773/*
774 * Note that we must __force cast these to unsigned long to make sparse happy,
775 * since all of the endian-annotated types are fixed size regardless of arch.
776 */
777static inline bool ipv6_addr_v4mapped(const struct in6_addr *a)
778{
779 return (
780#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
781 *(unsigned long *)a |
782#else
783 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) |
784#endif
785 (__force unsigned long)(a->s6_addr32[2] ^
786 cpu_to_be32(0x0000ffff))) == 0UL;
787}
788
789static inline bool ipv6_addr_v4mapped_loopback(const struct in6_addr *a)
790{
791 return ipv6_addr_v4mapped(a) && ipv4_is_loopback(a->s6_addr32[3]);
792}
793
794static inline u32 ipv6_portaddr_hash(const struct net *net,
795 const struct in6_addr *addr6,
796 unsigned int port)
797{
798 unsigned int hash, mix = net_hash_mix(net);
799
800 if (ipv6_addr_any(addr6))
801 hash = jhash_1word(0, mix);
802 else if (ipv6_addr_v4mapped(addr6))
803 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
804 else
805 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
806
807 return hash ^ port;
808}
809
810/*
811 * Check for a RFC 4843 ORCHID address
812 * (Overlay Routable Cryptographic Hash Identifiers)
813 */
814static inline bool ipv6_addr_orchid(const struct in6_addr *a)
815{
816 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010);
817}
818
819static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr)
820{
821 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000);
822}
823
824static inline void ipv6_addr_set_v4mapped(const __be32 addr,
825 struct in6_addr *v4mapped)
826{
827 ipv6_addr_set(v4mapped,
828 0, 0,
829 htonl(0x0000FFFF),
830 addr);
831}
832
833/*
834 * find the first different bit between two addresses
835 * length of address must be a multiple of 32bits
836 */
837static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen)
838{
839 const __be32 *a1 = token1, *a2 = token2;
840 int i;
841
842 addrlen >>= 2;
843
844 for (i = 0; i < addrlen; i++) {
845 __be32 xb = a1[i] ^ a2[i];
846 if (xb)
847 return i * 32 + 31 - __fls(ntohl(xb));
848 }
849
850 /*
851 * we should *never* get to this point since that
852 * would mean the addrs are equal
853 *
854 * However, we do get to it 8) And exactly, when
855 * addresses are equal 8)
856 *
857 * ip route add 1111::/128 via ...
858 * ip route add 1111::/64 via ...
859 * and we are here.
860 *
861 * Ideally, this function should stop comparison
862 * at prefix length. It does not, but it is still OK,
863 * if returned value is greater than prefix length.
864 * --ANK (980803)
865 */
866 return addrlen << 5;
867}
868
869#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
870static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen)
871{
872 const __be64 *a1 = token1, *a2 = token2;
873 int i;
874
875 addrlen >>= 3;
876
877 for (i = 0; i < addrlen; i++) {
878 __be64 xb = a1[i] ^ a2[i];
879 if (xb)
880 return i * 64 + 63 - __fls(be64_to_cpu(xb));
881 }
882
883 return addrlen << 6;
884}
885#endif
886
887static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen)
888{
889#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
890 if (__builtin_constant_p(addrlen) && !(addrlen & 7))
891 return __ipv6_addr_diff64(token1, token2, addrlen);
892#endif
893 return __ipv6_addr_diff32(token1, token2, addrlen);
894}
895
896static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2)
897{
898 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr));
899}
900
901__be32 ipv6_select_ident(struct net *net,
902 const struct in6_addr *daddr,
903 const struct in6_addr *saddr);
904__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb);
905
906int ip6_dst_hoplimit(struct dst_entry *dst);
907
908static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
909 struct dst_entry *dst)
910{
911 int hlimit;
912
913 if (ipv6_addr_is_multicast(&fl6->daddr))
914 hlimit = READ_ONCE(np->mcast_hops);
915 else
916 hlimit = READ_ONCE(np->hop_limit);
917 if (hlimit < 0)
918 hlimit = ip6_dst_hoplimit(dst);
919 return hlimit;
920}
921
922/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store
923 * Equivalent to : flow->v6addrs.src = iph->saddr;
924 * flow->v6addrs.dst = iph->daddr;
925 */
926static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow,
927 const struct ipv6hdr *iph)
928{
929 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) !=
930 offsetof(typeof(flow->addrs), v6addrs.src) +
931 sizeof(flow->addrs.v6addrs.src));
932 memcpy(&flow->addrs.v6addrs, &iph->addrs, sizeof(flow->addrs.v6addrs));
933 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
934}
935
936#if IS_ENABLED(CONFIG_IPV6)
937
938static inline bool ipv6_can_nonlocal_bind(struct net *net,
939 struct inet_sock *inet)
940{
941 return net->ipv6.sysctl.ip_nonlocal_bind ||
942 test_bit(INET_FLAGS_FREEBIND, &inet->inet_flags) ||
943 test_bit(INET_FLAGS_TRANSPARENT, &inet->inet_flags);
944}
945
946/* Sysctl settings for net ipv6.auto_flowlabels */
947#define IP6_AUTO_FLOW_LABEL_OFF 0
948#define IP6_AUTO_FLOW_LABEL_OPTOUT 1
949#define IP6_AUTO_FLOW_LABEL_OPTIN 2
950#define IP6_AUTO_FLOW_LABEL_FORCED 3
951
952#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED
953
954#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT
955
956static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
957 __be32 flowlabel, bool autolabel,
958 struct flowi6 *fl6)
959{
960 u32 hash;
961
962 /* @flowlabel may include more than a flow label, eg, the traffic class.
963 * Here we want only the flow label value.
964 */
965 flowlabel &= IPV6_FLOWLABEL_MASK;
966
967 if (flowlabel ||
968 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF ||
969 (!autolabel &&
970 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED))
971 return flowlabel;
972
973 hash = skb_get_hash_flowi6(skb, fl6);
974
975 /* Since this is being sent on the wire obfuscate hash a bit
976 * to minimize possibility that any useful information to an
977 * attacker is leaked. Only lower 20 bits are relevant.
978 */
979 hash = rol32(hash, 16);
980
981 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
982
983 if (net->ipv6.sysctl.flowlabel_state_ranges)
984 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG;
985
986 return flowlabel;
987}
988
989static inline int ip6_default_np_autolabel(struct net *net)
990{
991 switch (net->ipv6.sysctl.auto_flowlabels) {
992 case IP6_AUTO_FLOW_LABEL_OFF:
993 case IP6_AUTO_FLOW_LABEL_OPTIN:
994 default:
995 return 0;
996 case IP6_AUTO_FLOW_LABEL_OPTOUT:
997 case IP6_AUTO_FLOW_LABEL_FORCED:
998 return 1;
999 }
1000}
1001#else
1002static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
1003 __be32 flowlabel, bool autolabel,
1004 struct flowi6 *fl6)
1005{
1006 return flowlabel;
1007}
1008static inline int ip6_default_np_autolabel(struct net *net)
1009{
1010 return 0;
1011}
1012#endif
1013
1014#if IS_ENABLED(CONFIG_IPV6)
1015static inline int ip6_multipath_hash_policy(const struct net *net)
1016{
1017 return net->ipv6.sysctl.multipath_hash_policy;
1018}
1019static inline u32 ip6_multipath_hash_fields(const struct net *net)
1020{
1021 return net->ipv6.sysctl.multipath_hash_fields;
1022}
1023#else
1024static inline int ip6_multipath_hash_policy(const struct net *net)
1025{
1026 return 0;
1027}
1028static inline u32 ip6_multipath_hash_fields(const struct net *net)
1029{
1030 return 0;
1031}
1032#endif
1033
1034/*
1035 * Header manipulation
1036 */
1037static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass,
1038 __be32 flowlabel)
1039{
1040 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel;
1041}
1042
1043static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr)
1044{
1045 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK;
1046}
1047
1048static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr)
1049{
1050 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK;
1051}
1052
1053static inline u8 ip6_tclass(__be32 flowinfo)
1054{
1055 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT;
1056}
1057
1058static inline dscp_t ip6_dscp(__be32 flowinfo)
1059{
1060 return inet_dsfield_to_dscp(ip6_tclass(flowinfo));
1061}
1062
1063static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel)
1064{
1065 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel;
1066}
1067
1068static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6)
1069{
1070 return fl6->flowlabel & IPV6_FLOWLABEL_MASK;
1071}
1072
1073/*
1074 * Prototypes exported by ipv6
1075 */
1076
1077/*
1078 * rcv function (called from netdevice level)
1079 */
1080
1081int ipv6_rcv(struct sk_buff *skb, struct net_device *dev,
1082 struct packet_type *pt, struct net_device *orig_dev);
1083void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
1084 struct net_device *orig_dev);
1085
1086int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb);
1087
1088/*
1089 * upper-layer output functions
1090 */
1091int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
1092 __u32 mark, struct ipv6_txoptions *opt, int tclass, u32 priority);
1093
1094int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr);
1095
1096int ip6_append_data(struct sock *sk,
1097 int getfrag(void *from, char *to, int offset, int len,
1098 int odd, struct sk_buff *skb),
1099 void *from, size_t length, int transhdrlen,
1100 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1101 struct rt6_info *rt, unsigned int flags);
1102
1103int ip6_push_pending_frames(struct sock *sk);
1104
1105void ip6_flush_pending_frames(struct sock *sk);
1106
1107int ip6_send_skb(struct sk_buff *skb);
1108
1109struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue,
1110 struct inet_cork_full *cork,
1111 struct inet6_cork *v6_cork);
1112struct sk_buff *ip6_make_skb(struct sock *sk,
1113 int getfrag(void *from, char *to, int offset,
1114 int len, int odd, struct sk_buff *skb),
1115 void *from, size_t length, int transhdrlen,
1116 struct ipcm6_cookie *ipc6,
1117 struct rt6_info *rt, unsigned int flags,
1118 struct inet_cork_full *cork);
1119
1120static inline struct sk_buff *ip6_finish_skb(struct sock *sk)
1121{
1122 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork,
1123 &inet6_sk(sk)->cork);
1124}
1125
1126int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1127 struct flowi6 *fl6);
1128struct dst_entry *ip6_dst_lookup_flow(struct net *net, const struct sock *sk, struct flowi6 *fl6,
1129 const struct in6_addr *final_dst);
1130struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
1131 const struct in6_addr *final_dst,
1132 bool connected);
1133struct dst_entry *ip6_blackhole_route(struct net *net,
1134 struct dst_entry *orig_dst);
1135
1136/*
1137 * skb processing functions
1138 */
1139
1140int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb);
1141int ip6_forward(struct sk_buff *skb);
1142int ip6_input(struct sk_buff *skb);
1143int ip6_mc_input(struct sk_buff *skb);
1144void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
1145 bool have_final);
1146
1147int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1148int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1149
1150/*
1151 * Extension header (options) processing
1152 */
1153
1154void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1155 u8 *proto, struct in6_addr **daddr_p,
1156 struct in6_addr *saddr);
1157void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1158 u8 *proto);
1159
1160int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp,
1161 __be16 *frag_offp);
1162
1163bool ipv6_ext_hdr(u8 nexthdr);
1164
1165enum {
1166 IP6_FH_F_FRAG = (1 << 0),
1167 IP6_FH_F_AUTH = (1 << 1),
1168 IP6_FH_F_SKIP_RH = (1 << 2),
1169};
1170
1171/* find specified header and get offset to it */
1172int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target,
1173 unsigned short *fragoff, int *fragflg);
1174
1175int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type);
1176
1177struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1178 const struct ipv6_txoptions *opt,
1179 struct in6_addr *orig);
1180
1181/*
1182 * socket options (ipv6_sockglue.c)
1183 */
1184DECLARE_STATIC_KEY_FALSE(ip6_min_hopcount);
1185
1186int do_ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1187 unsigned int optlen);
1188int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1189 unsigned int optlen);
1190int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1191 sockptr_t optval, sockptr_t optlen);
1192int ipv6_getsockopt(struct sock *sk, int level, int optname,
1193 char __user *optval, int __user *optlen);
1194
1195int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr,
1196 int addr_len);
1197int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len);
1198int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr,
1199 int addr_len);
1200int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr);
1201void ip6_datagram_release_cb(struct sock *sk);
1202
1203int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len,
1204 int *addr_len);
1205int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
1206 int *addr_len);
1207void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
1208 u32 info, u8 *payload);
1209void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info);
1210void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu);
1211
1212void inet6_cleanup_sock(struct sock *sk);
1213void inet6_sock_destruct(struct sock *sk);
1214int inet6_release(struct socket *sock);
1215int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1216int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len);
1217int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
1218 int peer);
1219int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg);
1220int inet6_compat_ioctl(struct socket *sock, unsigned int cmd,
1221 unsigned long arg);
1222
1223int inet6_hash_connect(struct inet_timewait_death_row *death_row,
1224 struct sock *sk);
1225int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size);
1226int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1227 int flags);
1228
1229/*
1230 * reassembly.c
1231 */
1232extern const struct proto_ops inet6_stream_ops;
1233extern const struct proto_ops inet6_dgram_ops;
1234extern const struct proto_ops inet6_sockraw_ops;
1235
1236struct group_source_req;
1237struct group_filter;
1238
1239int ip6_mc_source(int add, int omode, struct sock *sk,
1240 struct group_source_req *pgsr);
1241int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
1242 struct sockaddr_storage *list);
1243int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
1244 sockptr_t optval, size_t ss_offset);
1245
1246#ifdef CONFIG_PROC_FS
1247int ac6_proc_init(struct net *net);
1248void ac6_proc_exit(struct net *net);
1249int raw6_proc_init(void);
1250void raw6_proc_exit(void);
1251int tcp6_proc_init(struct net *net);
1252void tcp6_proc_exit(struct net *net);
1253int udp6_proc_init(struct net *net);
1254void udp6_proc_exit(struct net *net);
1255int udplite6_proc_init(void);
1256void udplite6_proc_exit(void);
1257int ipv6_misc_proc_init(void);
1258void ipv6_misc_proc_exit(void);
1259int snmp6_register_dev(struct inet6_dev *idev);
1260int snmp6_unregister_dev(struct inet6_dev *idev);
1261
1262#else
1263static inline int ac6_proc_init(struct net *net) { return 0; }
1264static inline void ac6_proc_exit(struct net *net) { }
1265static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; }
1266static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; }
1267#endif
1268
1269#ifdef CONFIG_SYSCTL
1270struct ctl_table *ipv6_icmp_sysctl_init(struct net *net);
1271size_t ipv6_icmp_sysctl_table_size(void);
1272struct ctl_table *ipv6_route_sysctl_init(struct net *net);
1273size_t ipv6_route_sysctl_table_size(struct net *net);
1274int ipv6_sysctl_register(void);
1275void ipv6_sysctl_unregister(void);
1276#endif
1277
1278int ipv6_sock_mc_join(struct sock *sk, int ifindex,
1279 const struct in6_addr *addr);
1280int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
1281 const struct in6_addr *addr, unsigned int mode);
1282int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
1283 const struct in6_addr *addr);
1284
1285static inline int ip6_sock_set_v6only(struct sock *sk)
1286{
1287 if (inet_sk(sk)->inet_num)
1288 return -EINVAL;
1289 lock_sock(sk);
1290 sk->sk_ipv6only = true;
1291 release_sock(sk);
1292 return 0;
1293}
1294
1295static inline void ip6_sock_set_recverr(struct sock *sk)
1296{
1297 inet6_set_bit(RECVERR6, sk);
1298}
1299
1300#define IPV6_PREFER_SRC_MASK (IPV6_PREFER_SRC_TMP | IPV6_PREFER_SRC_PUBLIC | \
1301 IPV6_PREFER_SRC_COA)
1302
1303static inline int ip6_sock_set_addr_preferences(struct sock *sk, int val)
1304{
1305 unsigned int prefmask = ~IPV6_PREFER_SRC_MASK;
1306 unsigned int pref = 0;
1307
1308 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
1309 switch (val & (IPV6_PREFER_SRC_PUBLIC |
1310 IPV6_PREFER_SRC_TMP |
1311 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
1312 case IPV6_PREFER_SRC_PUBLIC:
1313 pref |= IPV6_PREFER_SRC_PUBLIC;
1314 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1315 IPV6_PREFER_SRC_TMP);
1316 break;
1317 case IPV6_PREFER_SRC_TMP:
1318 pref |= IPV6_PREFER_SRC_TMP;
1319 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1320 IPV6_PREFER_SRC_TMP);
1321 break;
1322 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
1323 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1324 IPV6_PREFER_SRC_TMP);
1325 break;
1326 case 0:
1327 break;
1328 default:
1329 return -EINVAL;
1330 }
1331
1332 /* check HOME/COA conflicts */
1333 switch (val & (IPV6_PREFER_SRC_HOME | IPV6_PREFER_SRC_COA)) {
1334 case IPV6_PREFER_SRC_HOME:
1335 prefmask &= ~IPV6_PREFER_SRC_COA;
1336 break;
1337 case IPV6_PREFER_SRC_COA:
1338 pref |= IPV6_PREFER_SRC_COA;
1339 break;
1340 case 0:
1341 break;
1342 default:
1343 return -EINVAL;
1344 }
1345
1346 /* check CGA/NONCGA conflicts */
1347 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
1348 case IPV6_PREFER_SRC_CGA:
1349 case IPV6_PREFER_SRC_NONCGA:
1350 case 0:
1351 break;
1352 default:
1353 return -EINVAL;
1354 }
1355
1356 WRITE_ONCE(inet6_sk(sk)->srcprefs,
1357 (READ_ONCE(inet6_sk(sk)->srcprefs) & prefmask) | pref);
1358 return 0;
1359}
1360
1361static inline void ip6_sock_set_recvpktinfo(struct sock *sk)
1362{
1363 lock_sock(sk);
1364 inet6_sk(sk)->rxopt.bits.rxinfo = true;
1365 release_sock(sk);
1366}
1367
1368#define IPV6_ADDR_WORDS 4
1369
1370static inline void ipv6_addr_cpu_to_be32(__be32 *dst, const u32 *src)
1371{
1372 cpu_to_be32_array(dst, src, IPV6_ADDR_WORDS);
1373}
1374
1375static inline void ipv6_addr_be32_to_cpu(u32 *dst, const __be32 *src)
1376{
1377 be32_to_cpu_array(dst, src, IPV6_ADDR_WORDS);
1378}
1379
1380#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/flow.h>
19#include <net/flow_dissector.h>
20#include <net/inet_dscp.h>
21#include <net/snmp.h>
22#include <net/netns/hash.h>
23
24struct ip_tunnel_info;
25
26#define SIN6_LEN_RFC2133 24
27
28#define IPV6_MAXPLEN 65535
29
30/*
31 * NextHeader field of IPv6 header
32 */
33
34#define NEXTHDR_HOP 0 /* Hop-by-hop option header. */
35#define NEXTHDR_IPV4 4 /* IPv4 in IPv6 */
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/*
155 * Jumbo payload option, as described in RFC 2675 2.
156 */
157struct hop_jumbo_hdr {
158 u8 nexthdr;
159 u8 hdrlen;
160 u8 tlv_type; /* IPV6_TLV_JUMBO, 0xC2 */
161 u8 tlv_len; /* 4 */
162 __be32 jumbo_payload_len;
163};
164
165#define IP6_MF 0x0001
166#define IP6_OFFSET 0xFFF8
167
168struct ip6_fraglist_iter {
169 struct ipv6hdr *tmp_hdr;
170 struct sk_buff *frag;
171 int offset;
172 unsigned int hlen;
173 __be32 frag_id;
174 u8 nexthdr;
175};
176
177int ip6_fraglist_init(struct sk_buff *skb, unsigned int hlen, u8 *prevhdr,
178 u8 nexthdr, __be32 frag_id,
179 struct ip6_fraglist_iter *iter);
180void ip6_fraglist_prepare(struct sk_buff *skb, struct ip6_fraglist_iter *iter);
181
182static inline struct sk_buff *ip6_fraglist_next(struct ip6_fraglist_iter *iter)
183{
184 struct sk_buff *skb = iter->frag;
185
186 iter->frag = skb->next;
187 skb_mark_not_on_list(skb);
188
189 return skb;
190}
191
192struct ip6_frag_state {
193 u8 *prevhdr;
194 unsigned int hlen;
195 unsigned int mtu;
196 unsigned int left;
197 int offset;
198 int ptr;
199 int hroom;
200 int troom;
201 __be32 frag_id;
202 u8 nexthdr;
203};
204
205void ip6_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int mtu,
206 unsigned short needed_tailroom, int hdr_room, u8 *prevhdr,
207 u8 nexthdr, __be32 frag_id, struct ip6_frag_state *state);
208struct sk_buff *ip6_frag_next(struct sk_buff *skb,
209 struct ip6_frag_state *state);
210
211#define IP6_REPLY_MARK(net, mark) \
212 ((net)->ipv6.sysctl.fwmark_reflect ? (mark) : 0)
213
214#include <net/sock.h>
215
216/* sysctls */
217extern int sysctl_mld_max_msf;
218extern int sysctl_mld_qrv;
219
220#define _DEVINC(net, statname, mod, idev, field) \
221({ \
222 struct inet6_dev *_idev = (idev); \
223 if (likely(_idev != NULL)) \
224 mod##SNMP_INC_STATS64((_idev)->stats.statname, (field));\
225 mod##SNMP_INC_STATS64((net)->mib.statname##_statistics, (field));\
226})
227
228/* per device counters are atomic_long_t */
229#define _DEVINCATOMIC(net, statname, mod, idev, field) \
230({ \
231 struct inet6_dev *_idev = (idev); \
232 if (likely(_idev != NULL)) \
233 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
234 mod##SNMP_INC_STATS((net)->mib.statname##_statistics, (field));\
235})
236
237/* per device and per net counters are atomic_long_t */
238#define _DEVINC_ATOMIC_ATOMIC(net, statname, idev, field) \
239({ \
240 struct inet6_dev *_idev = (idev); \
241 if (likely(_idev != NULL)) \
242 SNMP_INC_STATS_ATOMIC_LONG((_idev)->stats.statname##dev, (field)); \
243 SNMP_INC_STATS_ATOMIC_LONG((net)->mib.statname##_statistics, (field));\
244})
245
246#define _DEVADD(net, statname, mod, idev, field, val) \
247({ \
248 struct inet6_dev *_idev = (idev); \
249 if (likely(_idev != NULL)) \
250 mod##SNMP_ADD_STATS((_idev)->stats.statname, (field), (val)); \
251 mod##SNMP_ADD_STATS((net)->mib.statname##_statistics, (field), (val));\
252})
253
254#define _DEVUPD(net, statname, mod, idev, field, val) \
255({ \
256 struct inet6_dev *_idev = (idev); \
257 if (likely(_idev != NULL)) \
258 mod##SNMP_UPD_PO_STATS((_idev)->stats.statname, field, (val)); \
259 mod##SNMP_UPD_PO_STATS((net)->mib.statname##_statistics, field, (val));\
260})
261
262/* MIBs */
263
264#define IP6_INC_STATS(net, idev,field) \
265 _DEVINC(net, ipv6, , idev, field)
266#define __IP6_INC_STATS(net, idev,field) \
267 _DEVINC(net, ipv6, __, idev, field)
268#define IP6_ADD_STATS(net, idev,field,val) \
269 _DEVADD(net, ipv6, , idev, field, val)
270#define __IP6_ADD_STATS(net, idev,field,val) \
271 _DEVADD(net, ipv6, __, idev, field, val)
272#define IP6_UPD_PO_STATS(net, idev,field,val) \
273 _DEVUPD(net, ipv6, , idev, field, val)
274#define __IP6_UPD_PO_STATS(net, idev,field,val) \
275 _DEVUPD(net, ipv6, __, idev, field, val)
276#define ICMP6_INC_STATS(net, idev, field) \
277 _DEVINCATOMIC(net, icmpv6, , idev, field)
278#define __ICMP6_INC_STATS(net, idev, field) \
279 _DEVINCATOMIC(net, icmpv6, __, idev, field)
280
281#define ICMP6MSGOUT_INC_STATS(net, idev, field) \
282 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field +256)
283#define ICMP6MSGIN_INC_STATS(net, idev, field) \
284 _DEVINC_ATOMIC_ATOMIC(net, icmpv6msg, idev, field)
285
286struct ip6_ra_chain {
287 struct ip6_ra_chain *next;
288 struct sock *sk;
289 int sel;
290 void (*destructor)(struct sock *);
291};
292
293extern struct ip6_ra_chain *ip6_ra_chain;
294extern rwlock_t ip6_ra_lock;
295
296/*
297 This structure is prepared by protocol, when parsing
298 ancillary data and passed to IPv6.
299 */
300
301struct ipv6_txoptions {
302 refcount_t refcnt;
303 /* Length of this structure */
304 int tot_len;
305
306 /* length of extension headers */
307
308 __u16 opt_flen; /* after fragment hdr */
309 __u16 opt_nflen; /* before fragment hdr */
310
311 struct ipv6_opt_hdr *hopopt;
312 struct ipv6_opt_hdr *dst0opt;
313 struct ipv6_rt_hdr *srcrt; /* Routing Header */
314 struct ipv6_opt_hdr *dst1opt;
315 struct rcu_head rcu;
316 /* Option buffer, as read by IPV6_PKTOPTIONS, starts here. */
317};
318
319/* flowlabel_reflect sysctl values */
320enum flowlabel_reflect {
321 FLOWLABEL_REFLECT_ESTABLISHED = 1,
322 FLOWLABEL_REFLECT_TCP_RESET = 2,
323 FLOWLABEL_REFLECT_ICMPV6_ECHO_REPLIES = 4,
324};
325
326struct ip6_flowlabel {
327 struct ip6_flowlabel __rcu *next;
328 __be32 label;
329 atomic_t users;
330 struct in6_addr dst;
331 struct ipv6_txoptions *opt;
332 unsigned long linger;
333 struct rcu_head rcu;
334 u8 share;
335 union {
336 struct pid *pid;
337 kuid_t uid;
338 } owner;
339 unsigned long lastuse;
340 unsigned long expires;
341 struct net *fl_net;
342};
343
344#define IPV6_FLOWINFO_MASK cpu_to_be32(0x0FFFFFFF)
345#define IPV6_FLOWLABEL_MASK cpu_to_be32(0x000FFFFF)
346#define IPV6_FLOWLABEL_STATELESS_FLAG cpu_to_be32(0x00080000)
347
348#define IPV6_TCLASS_MASK (IPV6_FLOWINFO_MASK & ~IPV6_FLOWLABEL_MASK)
349#define IPV6_TCLASS_SHIFT 20
350
351struct ipv6_fl_socklist {
352 struct ipv6_fl_socklist __rcu *next;
353 struct ip6_flowlabel *fl;
354 struct rcu_head rcu;
355};
356
357struct ipcm6_cookie {
358 struct sockcm_cookie sockc;
359 __s16 hlimit;
360 __s16 tclass;
361 __u16 gso_size;
362 __s8 dontfrag;
363 struct ipv6_txoptions *opt;
364};
365
366static inline void ipcm6_init(struct ipcm6_cookie *ipc6)
367{
368 *ipc6 = (struct ipcm6_cookie) {
369 .hlimit = -1,
370 .tclass = -1,
371 .dontfrag = -1,
372 };
373}
374
375static inline void ipcm6_init_sk(struct ipcm6_cookie *ipc6,
376 const struct sock *sk)
377{
378 *ipc6 = (struct ipcm6_cookie) {
379 .hlimit = -1,
380 .tclass = inet6_sk(sk)->tclass,
381 .dontfrag = inet6_test_bit(DONTFRAG, sk),
382 };
383}
384
385static inline struct ipv6_txoptions *txopt_get(const struct ipv6_pinfo *np)
386{
387 struct ipv6_txoptions *opt;
388
389 rcu_read_lock();
390 opt = rcu_dereference(np->opt);
391 if (opt) {
392 if (!refcount_inc_not_zero(&opt->refcnt))
393 opt = NULL;
394 else
395 opt = rcu_pointer_handoff(opt);
396 }
397 rcu_read_unlock();
398 return opt;
399}
400
401static inline void txopt_put(struct ipv6_txoptions *opt)
402{
403 if (opt && refcount_dec_and_test(&opt->refcnt))
404 kfree_rcu(opt, rcu);
405}
406
407#if IS_ENABLED(CONFIG_IPV6)
408struct ip6_flowlabel *__fl6_sock_lookup(struct sock *sk, __be32 label);
409
410extern struct static_key_false_deferred ipv6_flowlabel_exclusive;
411static inline struct ip6_flowlabel *fl6_sock_lookup(struct sock *sk,
412 __be32 label)
413{
414 if (static_branch_unlikely(&ipv6_flowlabel_exclusive.key) &&
415 READ_ONCE(sock_net(sk)->ipv6.flowlabel_has_excl))
416 return __fl6_sock_lookup(sk, label) ? : ERR_PTR(-ENOENT);
417
418 return NULL;
419}
420#endif
421
422struct ipv6_txoptions *fl6_merge_options(struct ipv6_txoptions *opt_space,
423 struct ip6_flowlabel *fl,
424 struct ipv6_txoptions *fopt);
425void fl6_free_socklist(struct sock *sk);
426int ipv6_flowlabel_opt(struct sock *sk, sockptr_t optval, int optlen);
427int ipv6_flowlabel_opt_get(struct sock *sk, struct in6_flowlabel_req *freq,
428 int flags);
429int ip6_flowlabel_init(void);
430void ip6_flowlabel_cleanup(void);
431bool ip6_autoflowlabel(struct net *net, const struct sock *sk);
432
433static inline void fl6_sock_release(struct ip6_flowlabel *fl)
434{
435 if (fl)
436 atomic_dec(&fl->users);
437}
438
439enum skb_drop_reason icmpv6_notify(struct sk_buff *skb, u8 type,
440 u8 code, __be32 info);
441
442void icmpv6_push_pending_frames(struct sock *sk, struct flowi6 *fl6,
443 struct icmp6hdr *thdr, int len);
444
445int ip6_ra_control(struct sock *sk, int sel);
446
447int ipv6_parse_hopopts(struct sk_buff *skb);
448
449struct ipv6_txoptions *ipv6_dup_options(struct sock *sk,
450 struct ipv6_txoptions *opt);
451struct ipv6_txoptions *ipv6_renew_options(struct sock *sk,
452 struct ipv6_txoptions *opt,
453 int newtype,
454 struct ipv6_opt_hdr *newopt);
455struct ipv6_txoptions *__ipv6_fixup_options(struct ipv6_txoptions *opt_space,
456 struct ipv6_txoptions *opt);
457
458static inline struct ipv6_txoptions *
459ipv6_fixup_options(struct ipv6_txoptions *opt_space, struct ipv6_txoptions *opt)
460{
461 if (!opt)
462 return NULL;
463 return __ipv6_fixup_options(opt_space, opt);
464}
465
466bool ipv6_opt_accepted(const struct sock *sk, const struct sk_buff *skb,
467 const struct inet6_skb_parm *opt);
468struct ipv6_txoptions *ipv6_update_options(struct sock *sk,
469 struct ipv6_txoptions *opt);
470
471/* This helper is specialized for BIG TCP needs.
472 * It assumes the hop_jumbo_hdr will immediately follow the IPV6 header.
473 * It assumes headers are already in skb->head.
474 * Returns 0, or IPPROTO_TCP if a BIG TCP packet is there.
475 */
476static inline int ipv6_has_hopopt_jumbo(const struct sk_buff *skb)
477{
478 const struct hop_jumbo_hdr *jhdr;
479 const struct ipv6hdr *nhdr;
480
481 if (likely(skb->len <= GRO_LEGACY_MAX_SIZE))
482 return 0;
483
484 if (skb->protocol != htons(ETH_P_IPV6))
485 return 0;
486
487 if (skb_network_offset(skb) +
488 sizeof(struct ipv6hdr) +
489 sizeof(struct hop_jumbo_hdr) > skb_headlen(skb))
490 return 0;
491
492 nhdr = ipv6_hdr(skb);
493
494 if (nhdr->nexthdr != NEXTHDR_HOP)
495 return 0;
496
497 jhdr = (const struct hop_jumbo_hdr *) (nhdr + 1);
498 if (jhdr->tlv_type != IPV6_TLV_JUMBO || jhdr->hdrlen != 0 ||
499 jhdr->nexthdr != IPPROTO_TCP)
500 return 0;
501 return jhdr->nexthdr;
502}
503
504/* Return 0 if HBH header is successfully removed
505 * Or if HBH removal is unnecessary (packet is not big TCP)
506 * Return error to indicate dropping the packet
507 */
508static inline int ipv6_hopopt_jumbo_remove(struct sk_buff *skb)
509{
510 const int hophdr_len = sizeof(struct hop_jumbo_hdr);
511 int nexthdr = ipv6_has_hopopt_jumbo(skb);
512 struct ipv6hdr *h6;
513
514 if (!nexthdr)
515 return 0;
516
517 if (skb_cow_head(skb, 0))
518 return -1;
519
520 /* Remove the HBH header.
521 * Layout: [Ethernet header][IPv6 header][HBH][L4 Header]
522 */
523 memmove(skb_mac_header(skb) + hophdr_len, skb_mac_header(skb),
524 skb_network_header(skb) - skb_mac_header(skb) +
525 sizeof(struct ipv6hdr));
526
527 __skb_pull(skb, hophdr_len);
528 skb->network_header += hophdr_len;
529 skb->mac_header += hophdr_len;
530
531 h6 = ipv6_hdr(skb);
532 h6->nexthdr = nexthdr;
533
534 return 0;
535}
536
537static inline bool ipv6_accept_ra(struct inet6_dev *idev)
538{
539 /* If forwarding is enabled, RA are not accepted unless the special
540 * hybrid mode (accept_ra=2) is enabled.
541 */
542 return idev->cnf.forwarding ? idev->cnf.accept_ra == 2 :
543 idev->cnf.accept_ra;
544}
545
546#define IPV6_FRAG_HIGH_THRESH (4 * 1024*1024) /* 4194304 */
547#define IPV6_FRAG_LOW_THRESH (3 * 1024*1024) /* 3145728 */
548#define IPV6_FRAG_TIMEOUT (60 * HZ) /* 60 seconds */
549
550int __ipv6_addr_type(const struct in6_addr *addr);
551static inline int ipv6_addr_type(const struct in6_addr *addr)
552{
553 return __ipv6_addr_type(addr) & 0xffff;
554}
555
556static inline int ipv6_addr_scope(const struct in6_addr *addr)
557{
558 return __ipv6_addr_type(addr) & IPV6_ADDR_SCOPE_MASK;
559}
560
561static inline int __ipv6_addr_src_scope(int type)
562{
563 return (type == IPV6_ADDR_ANY) ? __IPV6_ADDR_SCOPE_INVALID : (type >> 16);
564}
565
566static inline int ipv6_addr_src_scope(const struct in6_addr *addr)
567{
568 return __ipv6_addr_src_scope(__ipv6_addr_type(addr));
569}
570
571static inline bool __ipv6_addr_needs_scope_id(int type)
572{
573 return type & IPV6_ADDR_LINKLOCAL ||
574 (type & IPV6_ADDR_MULTICAST &&
575 (type & (IPV6_ADDR_LOOPBACK|IPV6_ADDR_LINKLOCAL)));
576}
577
578static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr, int iface)
579{
580 return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0;
581}
582
583static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2)
584{
585 return memcmp(a1, a2, sizeof(struct in6_addr));
586}
587
588static inline bool
589ipv6_masked_addr_cmp(const struct in6_addr *a1, const struct in6_addr *m,
590 const struct in6_addr *a2)
591{
592#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
593 const unsigned long *ul1 = (const unsigned long *)a1;
594 const unsigned long *ulm = (const unsigned long *)m;
595 const unsigned long *ul2 = (const unsigned long *)a2;
596
597 return !!(((ul1[0] ^ ul2[0]) & ulm[0]) |
598 ((ul1[1] ^ ul2[1]) & ulm[1]));
599#else
600 return !!(((a1->s6_addr32[0] ^ a2->s6_addr32[0]) & m->s6_addr32[0]) |
601 ((a1->s6_addr32[1] ^ a2->s6_addr32[1]) & m->s6_addr32[1]) |
602 ((a1->s6_addr32[2] ^ a2->s6_addr32[2]) & m->s6_addr32[2]) |
603 ((a1->s6_addr32[3] ^ a2->s6_addr32[3]) & m->s6_addr32[3]));
604#endif
605}
606
607static inline void ipv6_addr_prefix(struct in6_addr *pfx,
608 const struct in6_addr *addr,
609 int plen)
610{
611 /* caller must guarantee 0 <= plen <= 128 */
612 int o = plen >> 3,
613 b = plen & 0x7;
614
615 memset(pfx->s6_addr, 0, sizeof(pfx->s6_addr));
616 memcpy(pfx->s6_addr, addr, o);
617 if (b != 0)
618 pfx->s6_addr[o] = addr->s6_addr[o] & (0xff00 >> b);
619}
620
621static inline void ipv6_addr_prefix_copy(struct in6_addr *addr,
622 const struct in6_addr *pfx,
623 int plen)
624{
625 /* caller must guarantee 0 <= plen <= 128 */
626 int o = plen >> 3,
627 b = plen & 0x7;
628
629 memcpy(addr->s6_addr, pfx, o);
630 if (b != 0) {
631 addr->s6_addr[o] &= ~(0xff00 >> b);
632 addr->s6_addr[o] |= (pfx->s6_addr[o] & (0xff00 >> b));
633 }
634}
635
636static inline void __ipv6_addr_set_half(__be32 *addr,
637 __be32 wh, __be32 wl)
638{
639#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
640#if defined(__BIG_ENDIAN)
641 if (__builtin_constant_p(wh) && __builtin_constant_p(wl)) {
642 *(__force u64 *)addr = ((__force u64)(wh) << 32 | (__force u64)(wl));
643 return;
644 }
645#elif defined(__LITTLE_ENDIAN)
646 if (__builtin_constant_p(wl) && __builtin_constant_p(wh)) {
647 *(__force u64 *)addr = ((__force u64)(wl) << 32 | (__force u64)(wh));
648 return;
649 }
650#endif
651#endif
652 addr[0] = wh;
653 addr[1] = wl;
654}
655
656static inline void ipv6_addr_set(struct in6_addr *addr,
657 __be32 w1, __be32 w2,
658 __be32 w3, __be32 w4)
659{
660 __ipv6_addr_set_half(&addr->s6_addr32[0], w1, w2);
661 __ipv6_addr_set_half(&addr->s6_addr32[2], w3, w4);
662}
663
664static inline bool ipv6_addr_equal(const struct in6_addr *a1,
665 const struct in6_addr *a2)
666{
667#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
668 const unsigned long *ul1 = (const unsigned long *)a1;
669 const unsigned long *ul2 = (const unsigned long *)a2;
670
671 return ((ul1[0] ^ ul2[0]) | (ul1[1] ^ ul2[1])) == 0UL;
672#else
673 return ((a1->s6_addr32[0] ^ a2->s6_addr32[0]) |
674 (a1->s6_addr32[1] ^ a2->s6_addr32[1]) |
675 (a1->s6_addr32[2] ^ a2->s6_addr32[2]) |
676 (a1->s6_addr32[3] ^ a2->s6_addr32[3])) == 0;
677#endif
678}
679
680#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
681static inline bool __ipv6_prefix_equal64_half(const __be64 *a1,
682 const __be64 *a2,
683 unsigned int len)
684{
685 if (len && ((*a1 ^ *a2) & cpu_to_be64((~0UL) << (64 - len))))
686 return false;
687 return true;
688}
689
690static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
691 const struct in6_addr *addr2,
692 unsigned int prefixlen)
693{
694 const __be64 *a1 = (const __be64 *)addr1;
695 const __be64 *a2 = (const __be64 *)addr2;
696
697 if (prefixlen >= 64) {
698 if (a1[0] ^ a2[0])
699 return false;
700 return __ipv6_prefix_equal64_half(a1 + 1, a2 + 1, prefixlen - 64);
701 }
702 return __ipv6_prefix_equal64_half(a1, a2, prefixlen);
703}
704#else
705static inline bool ipv6_prefix_equal(const struct in6_addr *addr1,
706 const struct in6_addr *addr2,
707 unsigned int prefixlen)
708{
709 const __be32 *a1 = addr1->s6_addr32;
710 const __be32 *a2 = addr2->s6_addr32;
711 unsigned int pdw, pbi;
712
713 /* check complete u32 in prefix */
714 pdw = prefixlen >> 5;
715 if (pdw && memcmp(a1, a2, pdw << 2))
716 return false;
717
718 /* check incomplete u32 in prefix */
719 pbi = prefixlen & 0x1f;
720 if (pbi && ((a1[pdw] ^ a2[pdw]) & htonl((0xffffffff) << (32 - pbi))))
721 return false;
722
723 return true;
724}
725#endif
726
727static inline bool ipv6_addr_any(const struct in6_addr *a)
728{
729#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
730 const unsigned long *ul = (const unsigned long *)a;
731
732 return (ul[0] | ul[1]) == 0UL;
733#else
734 return (a->s6_addr32[0] | a->s6_addr32[1] |
735 a->s6_addr32[2] | a->s6_addr32[3]) == 0;
736#endif
737}
738
739static inline u32 ipv6_addr_hash(const struct in6_addr *a)
740{
741#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
742 const unsigned long *ul = (const unsigned long *)a;
743 unsigned long x = ul[0] ^ ul[1];
744
745 return (u32)(x ^ (x >> 32));
746#else
747 return (__force u32)(a->s6_addr32[0] ^ a->s6_addr32[1] ^
748 a->s6_addr32[2] ^ a->s6_addr32[3]);
749#endif
750}
751
752/* more secured version of ipv6_addr_hash() */
753static inline u32 __ipv6_addr_jhash(const struct in6_addr *a, const u32 initval)
754{
755 return jhash2((__force const u32 *)a->s6_addr32,
756 ARRAY_SIZE(a->s6_addr32), initval);
757}
758
759static inline bool ipv6_addr_loopback(const struct in6_addr *a)
760{
761#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
762 const __be64 *be = (const __be64 *)a;
763
764 return (be[0] | (be[1] ^ cpu_to_be64(1))) == 0UL;
765#else
766 return (a->s6_addr32[0] | a->s6_addr32[1] |
767 a->s6_addr32[2] | (a->s6_addr32[3] ^ cpu_to_be32(1))) == 0;
768#endif
769}
770
771/*
772 * Note that we must __force cast these to unsigned long to make sparse happy,
773 * since all of the endian-annotated types are fixed size regardless of arch.
774 */
775static inline bool ipv6_addr_v4mapped(const struct in6_addr *a)
776{
777 return (
778#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
779 *(unsigned long *)a |
780#else
781 (__force unsigned long)(a->s6_addr32[0] | a->s6_addr32[1]) |
782#endif
783 (__force unsigned long)(a->s6_addr32[2] ^
784 cpu_to_be32(0x0000ffff))) == 0UL;
785}
786
787static inline bool ipv6_addr_v4mapped_loopback(const struct in6_addr *a)
788{
789 return ipv6_addr_v4mapped(a) && ipv4_is_loopback(a->s6_addr32[3]);
790}
791
792static inline u32 ipv6_portaddr_hash(const struct net *net,
793 const struct in6_addr *addr6,
794 unsigned int port)
795{
796 unsigned int hash, mix = net_hash_mix(net);
797
798 if (ipv6_addr_any(addr6))
799 hash = jhash_1word(0, mix);
800 else if (ipv6_addr_v4mapped(addr6))
801 hash = jhash_1word((__force u32)addr6->s6_addr32[3], mix);
802 else
803 hash = jhash2((__force u32 *)addr6->s6_addr32, 4, mix);
804
805 return hash ^ port;
806}
807
808/*
809 * Check for a RFC 4843 ORCHID address
810 * (Overlay Routable Cryptographic Hash Identifiers)
811 */
812static inline bool ipv6_addr_orchid(const struct in6_addr *a)
813{
814 return (a->s6_addr32[0] & htonl(0xfffffff0)) == htonl(0x20010010);
815}
816
817static inline bool ipv6_addr_is_multicast(const struct in6_addr *addr)
818{
819 return (addr->s6_addr32[0] & htonl(0xFF000000)) == htonl(0xFF000000);
820}
821
822static inline void ipv6_addr_set_v4mapped(const __be32 addr,
823 struct in6_addr *v4mapped)
824{
825 ipv6_addr_set(v4mapped,
826 0, 0,
827 htonl(0x0000FFFF),
828 addr);
829}
830
831/*
832 * find the first different bit between two addresses
833 * length of address must be a multiple of 32bits
834 */
835static inline int __ipv6_addr_diff32(const void *token1, const void *token2, int addrlen)
836{
837 const __be32 *a1 = token1, *a2 = token2;
838 int i;
839
840 addrlen >>= 2;
841
842 for (i = 0; i < addrlen; i++) {
843 __be32 xb = a1[i] ^ a2[i];
844 if (xb)
845 return i * 32 + 31 - __fls(ntohl(xb));
846 }
847
848 /*
849 * we should *never* get to this point since that
850 * would mean the addrs are equal
851 *
852 * However, we do get to it 8) And exacly, when
853 * addresses are equal 8)
854 *
855 * ip route add 1111::/128 via ...
856 * ip route add 1111::/64 via ...
857 * and we are here.
858 *
859 * Ideally, this function should stop comparison
860 * at prefix length. It does not, but it is still OK,
861 * if returned value is greater than prefix length.
862 * --ANK (980803)
863 */
864 return addrlen << 5;
865}
866
867#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
868static inline int __ipv6_addr_diff64(const void *token1, const void *token2, int addrlen)
869{
870 const __be64 *a1 = token1, *a2 = token2;
871 int i;
872
873 addrlen >>= 3;
874
875 for (i = 0; i < addrlen; i++) {
876 __be64 xb = a1[i] ^ a2[i];
877 if (xb)
878 return i * 64 + 63 - __fls(be64_to_cpu(xb));
879 }
880
881 return addrlen << 6;
882}
883#endif
884
885static inline int __ipv6_addr_diff(const void *token1, const void *token2, int addrlen)
886{
887#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
888 if (__builtin_constant_p(addrlen) && !(addrlen & 7))
889 return __ipv6_addr_diff64(token1, token2, addrlen);
890#endif
891 return __ipv6_addr_diff32(token1, token2, addrlen);
892}
893
894static inline int ipv6_addr_diff(const struct in6_addr *a1, const struct in6_addr *a2)
895{
896 return __ipv6_addr_diff(a1, a2, sizeof(struct in6_addr));
897}
898
899__be32 ipv6_select_ident(struct net *net,
900 const struct in6_addr *daddr,
901 const struct in6_addr *saddr);
902__be32 ipv6_proxy_select_ident(struct net *net, struct sk_buff *skb);
903
904int ip6_dst_hoplimit(struct dst_entry *dst);
905
906static inline int ip6_sk_dst_hoplimit(struct ipv6_pinfo *np, struct flowi6 *fl6,
907 struct dst_entry *dst)
908{
909 int hlimit;
910
911 if (ipv6_addr_is_multicast(&fl6->daddr))
912 hlimit = READ_ONCE(np->mcast_hops);
913 else
914 hlimit = READ_ONCE(np->hop_limit);
915 if (hlimit < 0)
916 hlimit = ip6_dst_hoplimit(dst);
917 return hlimit;
918}
919
920/* copy IPv6 saddr & daddr to flow_keys, possibly using 64bit load/store
921 * Equivalent to : flow->v6addrs.src = iph->saddr;
922 * flow->v6addrs.dst = iph->daddr;
923 */
924static inline void iph_to_flow_copy_v6addrs(struct flow_keys *flow,
925 const struct ipv6hdr *iph)
926{
927 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v6addrs.dst) !=
928 offsetof(typeof(flow->addrs), v6addrs.src) +
929 sizeof(flow->addrs.v6addrs.src));
930 memcpy(&flow->addrs.v6addrs, &iph->addrs, sizeof(flow->addrs.v6addrs));
931 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
932}
933
934#if IS_ENABLED(CONFIG_IPV6)
935
936static inline bool ipv6_can_nonlocal_bind(struct net *net,
937 struct inet_sock *inet)
938{
939 return net->ipv6.sysctl.ip_nonlocal_bind ||
940 test_bit(INET_FLAGS_FREEBIND, &inet->inet_flags) ||
941 test_bit(INET_FLAGS_TRANSPARENT, &inet->inet_flags);
942}
943
944/* Sysctl settings for net ipv6.auto_flowlabels */
945#define IP6_AUTO_FLOW_LABEL_OFF 0
946#define IP6_AUTO_FLOW_LABEL_OPTOUT 1
947#define IP6_AUTO_FLOW_LABEL_OPTIN 2
948#define IP6_AUTO_FLOW_LABEL_FORCED 3
949
950#define IP6_AUTO_FLOW_LABEL_MAX IP6_AUTO_FLOW_LABEL_FORCED
951
952#define IP6_DEFAULT_AUTO_FLOW_LABELS IP6_AUTO_FLOW_LABEL_OPTOUT
953
954static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
955 __be32 flowlabel, bool autolabel,
956 struct flowi6 *fl6)
957{
958 u32 hash;
959
960 /* @flowlabel may include more than a flow label, eg, the traffic class.
961 * Here we want only the flow label value.
962 */
963 flowlabel &= IPV6_FLOWLABEL_MASK;
964
965 if (flowlabel ||
966 net->ipv6.sysctl.auto_flowlabels == IP6_AUTO_FLOW_LABEL_OFF ||
967 (!autolabel &&
968 net->ipv6.sysctl.auto_flowlabels != IP6_AUTO_FLOW_LABEL_FORCED))
969 return flowlabel;
970
971 hash = skb_get_hash_flowi6(skb, fl6);
972
973 /* Since this is being sent on the wire obfuscate hash a bit
974 * to minimize possbility that any useful information to an
975 * attacker is leaked. Only lower 20 bits are relevant.
976 */
977 hash = rol32(hash, 16);
978
979 flowlabel = (__force __be32)hash & IPV6_FLOWLABEL_MASK;
980
981 if (net->ipv6.sysctl.flowlabel_state_ranges)
982 flowlabel |= IPV6_FLOWLABEL_STATELESS_FLAG;
983
984 return flowlabel;
985}
986
987static inline int ip6_default_np_autolabel(struct net *net)
988{
989 switch (net->ipv6.sysctl.auto_flowlabels) {
990 case IP6_AUTO_FLOW_LABEL_OFF:
991 case IP6_AUTO_FLOW_LABEL_OPTIN:
992 default:
993 return 0;
994 case IP6_AUTO_FLOW_LABEL_OPTOUT:
995 case IP6_AUTO_FLOW_LABEL_FORCED:
996 return 1;
997 }
998}
999#else
1000static inline __be32 ip6_make_flowlabel(struct net *net, struct sk_buff *skb,
1001 __be32 flowlabel, bool autolabel,
1002 struct flowi6 *fl6)
1003{
1004 return flowlabel;
1005}
1006static inline int ip6_default_np_autolabel(struct net *net)
1007{
1008 return 0;
1009}
1010#endif
1011
1012#if IS_ENABLED(CONFIG_IPV6)
1013static inline int ip6_multipath_hash_policy(const struct net *net)
1014{
1015 return net->ipv6.sysctl.multipath_hash_policy;
1016}
1017static inline u32 ip6_multipath_hash_fields(const struct net *net)
1018{
1019 return net->ipv6.sysctl.multipath_hash_fields;
1020}
1021#else
1022static inline int ip6_multipath_hash_policy(const struct net *net)
1023{
1024 return 0;
1025}
1026static inline u32 ip6_multipath_hash_fields(const struct net *net)
1027{
1028 return 0;
1029}
1030#endif
1031
1032/*
1033 * Header manipulation
1034 */
1035static inline void ip6_flow_hdr(struct ipv6hdr *hdr, unsigned int tclass,
1036 __be32 flowlabel)
1037{
1038 *(__be32 *)hdr = htonl(0x60000000 | (tclass << 20)) | flowlabel;
1039}
1040
1041static inline __be32 ip6_flowinfo(const struct ipv6hdr *hdr)
1042{
1043 return *(__be32 *)hdr & IPV6_FLOWINFO_MASK;
1044}
1045
1046static inline __be32 ip6_flowlabel(const struct ipv6hdr *hdr)
1047{
1048 return *(__be32 *)hdr & IPV6_FLOWLABEL_MASK;
1049}
1050
1051static inline u8 ip6_tclass(__be32 flowinfo)
1052{
1053 return ntohl(flowinfo & IPV6_TCLASS_MASK) >> IPV6_TCLASS_SHIFT;
1054}
1055
1056static inline dscp_t ip6_dscp(__be32 flowinfo)
1057{
1058 return inet_dsfield_to_dscp(ip6_tclass(flowinfo));
1059}
1060
1061static inline __be32 ip6_make_flowinfo(unsigned int tclass, __be32 flowlabel)
1062{
1063 return htonl(tclass << IPV6_TCLASS_SHIFT) | flowlabel;
1064}
1065
1066static inline __be32 flowi6_get_flowlabel(const struct flowi6 *fl6)
1067{
1068 return fl6->flowlabel & IPV6_FLOWLABEL_MASK;
1069}
1070
1071/*
1072 * Prototypes exported by ipv6
1073 */
1074
1075/*
1076 * rcv function (called from netdevice level)
1077 */
1078
1079int ipv6_rcv(struct sk_buff *skb, struct net_device *dev,
1080 struct packet_type *pt, struct net_device *orig_dev);
1081void ipv6_list_rcv(struct list_head *head, struct packet_type *pt,
1082 struct net_device *orig_dev);
1083
1084int ip6_rcv_finish(struct net *net, struct sock *sk, struct sk_buff *skb);
1085
1086/*
1087 * upper-layer output functions
1088 */
1089int ip6_xmit(const struct sock *sk, struct sk_buff *skb, struct flowi6 *fl6,
1090 __u32 mark, struct ipv6_txoptions *opt, int tclass, u32 priority);
1091
1092int ip6_find_1stfragopt(struct sk_buff *skb, u8 **nexthdr);
1093
1094int ip6_append_data(struct sock *sk,
1095 int getfrag(void *from, char *to, int offset, int len,
1096 int odd, struct sk_buff *skb),
1097 void *from, size_t length, int transhdrlen,
1098 struct ipcm6_cookie *ipc6, struct flowi6 *fl6,
1099 struct rt6_info *rt, unsigned int flags);
1100
1101int ip6_push_pending_frames(struct sock *sk);
1102
1103void ip6_flush_pending_frames(struct sock *sk);
1104
1105int ip6_send_skb(struct sk_buff *skb);
1106
1107struct sk_buff *__ip6_make_skb(struct sock *sk, struct sk_buff_head *queue,
1108 struct inet_cork_full *cork,
1109 struct inet6_cork *v6_cork);
1110struct sk_buff *ip6_make_skb(struct sock *sk,
1111 int getfrag(void *from, char *to, int offset,
1112 int len, int odd, struct sk_buff *skb),
1113 void *from, size_t length, int transhdrlen,
1114 struct ipcm6_cookie *ipc6,
1115 struct rt6_info *rt, unsigned int flags,
1116 struct inet_cork_full *cork);
1117
1118static inline struct sk_buff *ip6_finish_skb(struct sock *sk)
1119{
1120 return __ip6_make_skb(sk, &sk->sk_write_queue, &inet_sk(sk)->cork,
1121 &inet6_sk(sk)->cork);
1122}
1123
1124int ip6_dst_lookup(struct net *net, struct sock *sk, struct dst_entry **dst,
1125 struct flowi6 *fl6);
1126struct dst_entry *ip6_dst_lookup_flow(struct net *net, const struct sock *sk, struct flowi6 *fl6,
1127 const struct in6_addr *final_dst);
1128struct dst_entry *ip6_sk_dst_lookup_flow(struct sock *sk, struct flowi6 *fl6,
1129 const struct in6_addr *final_dst,
1130 bool connected);
1131struct dst_entry *ip6_blackhole_route(struct net *net,
1132 struct dst_entry *orig_dst);
1133
1134/*
1135 * skb processing functions
1136 */
1137
1138int ip6_output(struct net *net, struct sock *sk, struct sk_buff *skb);
1139int ip6_forward(struct sk_buff *skb);
1140int ip6_input(struct sk_buff *skb);
1141int ip6_mc_input(struct sk_buff *skb);
1142void ip6_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int nexthdr,
1143 bool have_final);
1144
1145int __ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1146int ip6_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
1147
1148/*
1149 * Extension header (options) processing
1150 */
1151
1152void ipv6_push_nfrag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1153 u8 *proto, struct in6_addr **daddr_p,
1154 struct in6_addr *saddr);
1155void ipv6_push_frag_opts(struct sk_buff *skb, struct ipv6_txoptions *opt,
1156 u8 *proto);
1157
1158int ipv6_skip_exthdr(const struct sk_buff *, int start, u8 *nexthdrp,
1159 __be16 *frag_offp);
1160
1161bool ipv6_ext_hdr(u8 nexthdr);
1162
1163enum {
1164 IP6_FH_F_FRAG = (1 << 0),
1165 IP6_FH_F_AUTH = (1 << 1),
1166 IP6_FH_F_SKIP_RH = (1 << 2),
1167};
1168
1169/* find specified header and get offset to it */
1170int ipv6_find_hdr(const struct sk_buff *skb, unsigned int *offset, int target,
1171 unsigned short *fragoff, int *fragflg);
1172
1173int ipv6_find_tlv(const struct sk_buff *skb, int offset, int type);
1174
1175struct in6_addr *fl6_update_dst(struct flowi6 *fl6,
1176 const struct ipv6_txoptions *opt,
1177 struct in6_addr *orig);
1178
1179/*
1180 * socket options (ipv6_sockglue.c)
1181 */
1182DECLARE_STATIC_KEY_FALSE(ip6_min_hopcount);
1183
1184int do_ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1185 unsigned int optlen);
1186int ipv6_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
1187 unsigned int optlen);
1188int do_ipv6_getsockopt(struct sock *sk, int level, int optname,
1189 sockptr_t optval, sockptr_t optlen);
1190int ipv6_getsockopt(struct sock *sk, int level, int optname,
1191 char __user *optval, int __user *optlen);
1192
1193int __ip6_datagram_connect(struct sock *sk, struct sockaddr *addr,
1194 int addr_len);
1195int ip6_datagram_connect(struct sock *sk, struct sockaddr *addr, int addr_len);
1196int ip6_datagram_connect_v6_only(struct sock *sk, struct sockaddr *addr,
1197 int addr_len);
1198int ip6_datagram_dst_update(struct sock *sk, bool fix_sk_saddr);
1199void ip6_datagram_release_cb(struct sock *sk);
1200
1201int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len,
1202 int *addr_len);
1203int ipv6_recv_rxpmtu(struct sock *sk, struct msghdr *msg, int len,
1204 int *addr_len);
1205void ipv6_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
1206 u32 info, u8 *payload);
1207void ipv6_local_error(struct sock *sk, int err, struct flowi6 *fl6, u32 info);
1208void ipv6_local_rxpmtu(struct sock *sk, struct flowi6 *fl6, u32 mtu);
1209
1210void inet6_cleanup_sock(struct sock *sk);
1211void inet6_sock_destruct(struct sock *sk);
1212int inet6_release(struct socket *sock);
1213int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len);
1214int inet6_bind_sk(struct sock *sk, struct sockaddr *uaddr, int addr_len);
1215int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
1216 int peer);
1217int inet6_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg);
1218int inet6_compat_ioctl(struct socket *sock, unsigned int cmd,
1219 unsigned long arg);
1220
1221int inet6_hash_connect(struct inet_timewait_death_row *death_row,
1222 struct sock *sk);
1223int inet6_sendmsg(struct socket *sock, struct msghdr *msg, size_t size);
1224int inet6_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1225 int flags);
1226
1227/*
1228 * reassembly.c
1229 */
1230extern const struct proto_ops inet6_stream_ops;
1231extern const struct proto_ops inet6_dgram_ops;
1232extern const struct proto_ops inet6_sockraw_ops;
1233
1234struct group_source_req;
1235struct group_filter;
1236
1237int ip6_mc_source(int add, int omode, struct sock *sk,
1238 struct group_source_req *pgsr);
1239int ip6_mc_msfilter(struct sock *sk, struct group_filter *gsf,
1240 struct sockaddr_storage *list);
1241int ip6_mc_msfget(struct sock *sk, struct group_filter *gsf,
1242 sockptr_t optval, size_t ss_offset);
1243
1244#ifdef CONFIG_PROC_FS
1245int ac6_proc_init(struct net *net);
1246void ac6_proc_exit(struct net *net);
1247int raw6_proc_init(void);
1248void raw6_proc_exit(void);
1249int tcp6_proc_init(struct net *net);
1250void tcp6_proc_exit(struct net *net);
1251int udp6_proc_init(struct net *net);
1252void udp6_proc_exit(struct net *net);
1253int udplite6_proc_init(void);
1254void udplite6_proc_exit(void);
1255int ipv6_misc_proc_init(void);
1256void ipv6_misc_proc_exit(void);
1257int snmp6_register_dev(struct inet6_dev *idev);
1258int snmp6_unregister_dev(struct inet6_dev *idev);
1259
1260#else
1261static inline int ac6_proc_init(struct net *net) { return 0; }
1262static inline void ac6_proc_exit(struct net *net) { }
1263static inline int snmp6_register_dev(struct inet6_dev *idev) { return 0; }
1264static inline int snmp6_unregister_dev(struct inet6_dev *idev) { return 0; }
1265#endif
1266
1267#ifdef CONFIG_SYSCTL
1268struct ctl_table *ipv6_icmp_sysctl_init(struct net *net);
1269size_t ipv6_icmp_sysctl_table_size(void);
1270struct ctl_table *ipv6_route_sysctl_init(struct net *net);
1271size_t ipv6_route_sysctl_table_size(struct net *net);
1272int ipv6_sysctl_register(void);
1273void ipv6_sysctl_unregister(void);
1274#endif
1275
1276int ipv6_sock_mc_join(struct sock *sk, int ifindex,
1277 const struct in6_addr *addr);
1278int ipv6_sock_mc_join_ssm(struct sock *sk, int ifindex,
1279 const struct in6_addr *addr, unsigned int mode);
1280int ipv6_sock_mc_drop(struct sock *sk, int ifindex,
1281 const struct in6_addr *addr);
1282
1283static inline int ip6_sock_set_v6only(struct sock *sk)
1284{
1285 if (inet_sk(sk)->inet_num)
1286 return -EINVAL;
1287 lock_sock(sk);
1288 sk->sk_ipv6only = true;
1289 release_sock(sk);
1290 return 0;
1291}
1292
1293static inline void ip6_sock_set_recverr(struct sock *sk)
1294{
1295 inet6_set_bit(RECVERR6, sk);
1296}
1297
1298#define IPV6_PREFER_SRC_MASK (IPV6_PREFER_SRC_TMP | IPV6_PREFER_SRC_PUBLIC | \
1299 IPV6_PREFER_SRC_COA)
1300
1301static inline int ip6_sock_set_addr_preferences(struct sock *sk, int val)
1302{
1303 unsigned int prefmask = ~IPV6_PREFER_SRC_MASK;
1304 unsigned int pref = 0;
1305
1306 /* check PUBLIC/TMP/PUBTMP_DEFAULT conflicts */
1307 switch (val & (IPV6_PREFER_SRC_PUBLIC |
1308 IPV6_PREFER_SRC_TMP |
1309 IPV6_PREFER_SRC_PUBTMP_DEFAULT)) {
1310 case IPV6_PREFER_SRC_PUBLIC:
1311 pref |= IPV6_PREFER_SRC_PUBLIC;
1312 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1313 IPV6_PREFER_SRC_TMP);
1314 break;
1315 case IPV6_PREFER_SRC_TMP:
1316 pref |= IPV6_PREFER_SRC_TMP;
1317 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1318 IPV6_PREFER_SRC_TMP);
1319 break;
1320 case IPV6_PREFER_SRC_PUBTMP_DEFAULT:
1321 prefmask &= ~(IPV6_PREFER_SRC_PUBLIC |
1322 IPV6_PREFER_SRC_TMP);
1323 break;
1324 case 0:
1325 break;
1326 default:
1327 return -EINVAL;
1328 }
1329
1330 /* check HOME/COA conflicts */
1331 switch (val & (IPV6_PREFER_SRC_HOME | IPV6_PREFER_SRC_COA)) {
1332 case IPV6_PREFER_SRC_HOME:
1333 prefmask &= ~IPV6_PREFER_SRC_COA;
1334 break;
1335 case IPV6_PREFER_SRC_COA:
1336 pref |= IPV6_PREFER_SRC_COA;
1337 break;
1338 case 0:
1339 break;
1340 default:
1341 return -EINVAL;
1342 }
1343
1344 /* check CGA/NONCGA conflicts */
1345 switch (val & (IPV6_PREFER_SRC_CGA|IPV6_PREFER_SRC_NONCGA)) {
1346 case IPV6_PREFER_SRC_CGA:
1347 case IPV6_PREFER_SRC_NONCGA:
1348 case 0:
1349 break;
1350 default:
1351 return -EINVAL;
1352 }
1353
1354 WRITE_ONCE(inet6_sk(sk)->srcprefs,
1355 (READ_ONCE(inet6_sk(sk)->srcprefs) & prefmask) | pref);
1356 return 0;
1357}
1358
1359static inline void ip6_sock_set_recvpktinfo(struct sock *sk)
1360{
1361 lock_sock(sk);
1362 inet6_sk(sk)->rxopt.bits.rxinfo = true;
1363 release_sock(sk);
1364}
1365
1366#endif /* _NET_IPV6_H */