projects/linux·blob v4.8 · refs

585 lines · 16.8 KiB

1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the IP module.
7 *
8 * Version: @(#)ip.h 1.0.2 05/07/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <[email protected]>
12 * Alan Cox, <[email protected]>
13 *
14 * Changes:
15 * Mike McLagan : Routing by source
16 *
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
21 */
22#ifndef _IP_H
23#define _IP_H
24
25#include <linux/types.h>
26#include <linux/ip.h>
27#include <linux/in.h>
28#include <linux/skbuff.h>
29
30#include <net/inet_sock.h>
31#include <net/route.h>
32#include <net/snmp.h>
33#include <net/flow.h>
34#include <net/flow_dissector.h>
35
36struct sock;
37
38struct inet_skb_parm {
39 int iif;
40 struct ip_options opt; /* Compiled IP options */
41 unsigned char flags;
42
43#define IPSKB_FORWARDED BIT(0)
44#define IPSKB_XFRM_TUNNEL_SIZE BIT(1)
45#define IPSKB_XFRM_TRANSFORMED BIT(2)
46#define IPSKB_FRAG_COMPLETE BIT(3)
47#define IPSKB_REROUTED BIT(4)
48#define IPSKB_DOREDIRECT BIT(5)
49#define IPSKB_FRAG_PMTU BIT(6)
50#define IPSKB_FRAG_SEGS BIT(7)
51
52 u16 frag_max_size;
53};
54
55static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
56{
57 return ip_hdr(skb)->ihl * 4;
58}
59
60struct ipcm_cookie {
61 struct sockcm_cookie sockc;
62 __be32 addr;
63 int oif;
64 struct ip_options_rcu *opt;
65 __u8 tx_flags;
66 __u8 ttl;
67 __s16 tos;
68 char priority;
69};
70
71#define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
72#define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
73
74struct ip_ra_chain {
75 struct ip_ra_chain __rcu *next;
76 struct sock *sk;
77 union {
78 void (*destructor)(struct sock *);
79 struct sock *saved_sk;
80 };
81 struct rcu_head rcu;
82};
83
84extern struct ip_ra_chain __rcu *ip_ra_chain;
85
86/* IP flags. */
87#define IP_CE 0x8000 /* Flag: "Congestion" */
88#define IP_DF 0x4000 /* Flag: "Don't Fragment" */
89#define IP_MF 0x2000 /* Flag: "More Fragments" */
90#define IP_OFFSET 0x1FFF /* "Fragment Offset" part */
91
92#define IP_FRAG_TIME (30 * HZ) /* fragment lifetime */
93
94struct msghdr;
95struct net_device;
96struct packet_type;
97struct rtable;
98struct sockaddr;
99
100int igmp_mc_init(void);
101
102/*
103 * Functions provided by ip.c
104 */
105
106int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
107 __be32 saddr, __be32 daddr,
108 struct ip_options_rcu *opt);
109int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
110 struct net_device *orig_dev);
111int ip_local_deliver(struct sk_buff *skb);
112int ip_mr_input(struct sk_buff *skb);
113int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
114int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
115int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
116 int (*output)(struct net *, struct sock *, struct sk_buff *));
117void ip_send_check(struct iphdr *ip);
118int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
119int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
120
121int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
122void ip_init(void);
123int ip_append_data(struct sock *sk, struct flowi4 *fl4,
124 int getfrag(void *from, char *to, int offset, int len,
125 int odd, struct sk_buff *skb),
126 void *from, int len, int protolen,
127 struct ipcm_cookie *ipc,
128 struct rtable **rt,
129 unsigned int flags);
130int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
131 struct sk_buff *skb);
132ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
133 int offset, size_t size, int flags);
134struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
135 struct sk_buff_head *queue,
136 struct inet_cork *cork);
137int ip_send_skb(struct net *net, struct sk_buff *skb);
138int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
139void ip_flush_pending_frames(struct sock *sk);
140struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
141 int getfrag(void *from, char *to, int offset,
142 int len, int odd, struct sk_buff *skb),
143 void *from, int length, int transhdrlen,
144 struct ipcm_cookie *ipc, struct rtable **rtp,
145 unsigned int flags);
146
147static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
148{
149 return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
150}
151
152static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
153{
154 return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
155}
156
157static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
158{
159 return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
160}
161
162/* datagram.c */
163int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
164int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
165
166void ip4_datagram_release_cb(struct sock *sk);
167
168struct ip_reply_arg {
169 struct kvec iov[1];
170 int flags;
171 __wsum csum;
172 int csumoffset; /* u16 offset of csum in iov[0].iov_base */
173 /* -1 if not needed */
174 int bound_dev_if;
175 u8 tos;
176};
177
178#define IP_REPLY_ARG_NOSRCCHECK 1
179
180static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
181{
182 return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
183}
184
185void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
186 const struct ip_options *sopt,
187 __be32 daddr, __be32 saddr,
188 const struct ip_reply_arg *arg,
189 unsigned int len);
190
191#define IP_INC_STATS(net, field) SNMP_INC_STATS64((net)->mib.ip_statistics, field)
192#define __IP_INC_STATS(net, field) __SNMP_INC_STATS64((net)->mib.ip_statistics, field)
193#define IP_ADD_STATS(net, field, val) SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
194#define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
195#define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
196#define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
197#define NET_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.net_statistics, field)
198#define __NET_INC_STATS(net, field) __SNMP_INC_STATS((net)->mib.net_statistics, field)
199#define NET_ADD_STATS(net, field, adnd) SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
200#define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
201
202u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
203unsigned long snmp_fold_field(void __percpu *mib, int offt);
204#if BITS_PER_LONG==32
205u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
206 size_t syncp_offset);
207u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
208#else
209static inline u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
210 size_t syncp_offset)
211{
212 return snmp_get_cpu_field(mib, cpu, offct);
213
214}
215
216static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
217{
218 return snmp_fold_field(mib, offt);
219}
220#endif
221
222void inet_get_local_port_range(struct net *net, int *low, int *high);
223
224#ifdef CONFIG_SYSCTL
225static inline int inet_is_local_reserved_port(struct net *net, int port)
226{
227 if (!net->ipv4.sysctl_local_reserved_ports)
228 return 0;
229 return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
230}
231
232static inline bool sysctl_dev_name_is_allowed(const char *name)
233{
234 return strcmp(name, "default") != 0 && strcmp(name, "all") != 0;
235}
236
237#else
238static inline int inet_is_local_reserved_port(struct net *net, int port)
239{
240 return 0;
241}
242#endif
243
244__be32 inet_current_timestamp(void);
245
246/* From inetpeer.c */
247extern int inet_peer_threshold;
248extern int inet_peer_minttl;
249extern int inet_peer_maxttl;
250
251void ipfrag_init(void);
252
253void ip_static_sysctl_init(void);
254
255#define IP4_REPLY_MARK(net, mark) \
256 ((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
257
258static inline bool ip_is_fragment(const struct iphdr *iph)
259{
260 return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
261}
262
263#ifdef CONFIG_INET
264#include <net/dst.h>
265
266/* The function in 2.2 was invalid, producing wrong result for
267 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
268static inline
269int ip_decrease_ttl(struct iphdr *iph)
270{
271 u32 check = (__force u32)iph->check;
272 check += (__force u32)htons(0x0100);
273 iph->check = (__force __sum16)(check + (check>=0xFFFF));
274 return --iph->ttl;
275}
276
277static inline
278int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
279{
280 u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
281
282 return pmtudisc == IP_PMTUDISC_DO ||
283 (pmtudisc == IP_PMTUDISC_WANT &&
284 !(dst_metric_locked(dst, RTAX_MTU)));
285}
286
287static inline bool ip_sk_accept_pmtu(const struct sock *sk)
288{
289 return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
290 inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
291}
292
293static inline bool ip_sk_use_pmtu(const struct sock *sk)
294{
295 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
296}
297
298static inline bool ip_sk_ignore_df(const struct sock *sk)
299{
300 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
301 inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
302}
303
304static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
305 bool forwarding)
306{
307 struct net *net = dev_net(dst->dev);
308
309 if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
310 dst_metric_locked(dst, RTAX_MTU) ||
311 !forwarding)
312 return dst_mtu(dst);
313
314 return min(dst->dev->mtu, IP_MAX_MTU);
315}
316
317static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
318 const struct sk_buff *skb)
319{
320 if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
321 bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
322
323 return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
324 }
325
326 return min(skb_dst(skb)->dev->mtu, IP_MAX_MTU);
327}
328
329u32 ip_idents_reserve(u32 hash, int segs);
330void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
331
332static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
333 struct sock *sk, int segs)
334{
335 struct iphdr *iph = ip_hdr(skb);
336
337 if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
338 /* This is only to work around buggy Windows95/2000
339 * VJ compression implementations. If the ID field
340 * does not change, they drop every other packet in
341 * a TCP stream using header compression.
342 */
343 if (sk && inet_sk(sk)->inet_daddr) {
344 iph->id = htons(inet_sk(sk)->inet_id);
345 inet_sk(sk)->inet_id += segs;
346 } else {
347 iph->id = 0;
348 }
349 } else {
350 __ip_select_ident(net, iph, segs);
351 }
352}
353
354static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
355 struct sock *sk)
356{
357 ip_select_ident_segs(net, skb, sk, 1);
358}
359
360static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
361{
362 return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
363 skb->len, proto, 0);
364}
365
366/* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
367 * Equivalent to : flow->v4addrs.src = iph->saddr;
368 * flow->v4addrs.dst = iph->daddr;
369 */
370static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
371 const struct iphdr *iph)
372{
373 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
374 offsetof(typeof(flow->addrs), v4addrs.src) +
375 sizeof(flow->addrs.v4addrs.src));
376 memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
377 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
378}
379
380static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
381{
382 const struct iphdr *iph = skb_gro_network_header(skb);
383
384 return csum_tcpudp_nofold(iph->saddr, iph->daddr,
385 skb_gro_len(skb), proto, 0);
386}
387
388/*
389 * Map a multicast IP onto multicast MAC for type ethernet.
390 */
391
392static inline void ip_eth_mc_map(__be32 naddr, char *buf)
393{
394 __u32 addr=ntohl(naddr);
395 buf[0]=0x01;
396 buf[1]=0x00;
397 buf[2]=0x5e;
398 buf[5]=addr&0xFF;
399 addr>>=8;
400 buf[4]=addr&0xFF;
401 addr>>=8;
402 buf[3]=addr&0x7F;
403}
404
405/*
406 * Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
407 * Leave P_Key as 0 to be filled in by driver.
408 */
409
410static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
411{
412 __u32 addr;
413 unsigned char scope = broadcast[5] & 0xF;
414
415 buf[0] = 0; /* Reserved */
416 buf[1] = 0xff; /* Multicast QPN */
417 buf[2] = 0xff;
418 buf[3] = 0xff;
419 addr = ntohl(naddr);
420 buf[4] = 0xff;
421 buf[5] = 0x10 | scope; /* scope from broadcast address */
422 buf[6] = 0x40; /* IPv4 signature */
423 buf[7] = 0x1b;
424 buf[8] = broadcast[8]; /* P_Key */
425 buf[9] = broadcast[9];
426 buf[10] = 0;
427 buf[11] = 0;
428 buf[12] = 0;
429 buf[13] = 0;
430 buf[14] = 0;
431 buf[15] = 0;
432 buf[19] = addr & 0xff;
433 addr >>= 8;
434 buf[18] = addr & 0xff;
435 addr >>= 8;
436 buf[17] = addr & 0xff;
437 addr >>= 8;
438 buf[16] = addr & 0x0f;
439}
440
441static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
442{
443 if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
444 memcpy(buf, broadcast, 4);
445 else
446 memcpy(buf, &naddr, sizeof(naddr));
447}
448
449#if IS_ENABLED(CONFIG_IPV6)
450#include <linux/ipv6.h>
451#endif
452
453static __inline__ void inet_reset_saddr(struct sock *sk)
454{
455 inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
456#if IS_ENABLED(CONFIG_IPV6)
457 if (sk->sk_family == PF_INET6) {
458 struct ipv6_pinfo *np = inet6_sk(sk);
459
460 memset(&np->saddr, 0, sizeof(np->saddr));
461 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
462 }
463#endif
464}
465
466#endif
467
468static inline unsigned int ipv4_addr_hash(__be32 ip)
469{
470 return (__force unsigned int) ip;
471}
472
473bool ip_call_ra_chain(struct sk_buff *skb);
474
475/*
476 * Functions provided by ip_fragment.c
477 */
478
479enum ip_defrag_users {
480 IP_DEFRAG_LOCAL_DELIVER,
481 IP_DEFRAG_CALL_RA_CHAIN,
482 IP_DEFRAG_CONNTRACK_IN,
483 __IP_DEFRAG_CONNTRACK_IN_END = IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
484 IP_DEFRAG_CONNTRACK_OUT,
485 __IP_DEFRAG_CONNTRACK_OUT_END = IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
486 IP_DEFRAG_CONNTRACK_BRIDGE_IN,
487 __IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
488 IP_DEFRAG_VS_IN,
489 IP_DEFRAG_VS_OUT,
490 IP_DEFRAG_VS_FWD,
491 IP_DEFRAG_AF_PACKET,
492 IP_DEFRAG_MACVLAN,
493};
494
495/* Return true if the value of 'user' is between 'lower_bond'
496 * and 'upper_bond' inclusively.
497 */
498static inline bool ip_defrag_user_in_between(u32 user,
499 enum ip_defrag_users lower_bond,
500 enum ip_defrag_users upper_bond)
501{
502 return user >= lower_bond && user <= upper_bond;
503}
504
505int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
506#ifdef CONFIG_INET
507struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
508#else
509static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
510{
511 return skb;
512}
513#endif
514int ip_frag_mem(struct net *net);
515
516/*
517 * Functions provided by ip_forward.c
518 */
519
520int ip_forward(struct sk_buff *skb);
521
522/*
523 * Functions provided by ip_options.c
524 */
525
526void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
527 __be32 daddr, struct rtable *rt, int is_frag);
528
529int __ip_options_echo(struct ip_options *dopt, struct sk_buff *skb,
530 const struct ip_options *sopt);
531static inline int ip_options_echo(struct ip_options *dopt, struct sk_buff *skb)
532{
533 return __ip_options_echo(dopt, skb, &IPCB(skb)->opt);
534}
535
536void ip_options_fragment(struct sk_buff *skb);
537int ip_options_compile(struct net *net, struct ip_options *opt,
538 struct sk_buff *skb);
539int ip_options_get(struct net *net, struct ip_options_rcu **optp,
540 unsigned char *data, int optlen);
541int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
542 unsigned char __user *data, int optlen);
543void ip_options_undo(struct ip_options *opt);
544void ip_forward_options(struct sk_buff *skb);
545int ip_options_rcv_srr(struct sk_buff *skb);
546
547/*
548 * Functions provided by ip_sockglue.c
549 */
550
551void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
552void ip_cmsg_recv_offset(struct msghdr *msg, struct sk_buff *skb, int offset);
553int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
554 struct ipcm_cookie *ipc, bool allow_ipv6);
555int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
556 unsigned int optlen);
557int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
558 int __user *optlen);
559int compat_ip_setsockopt(struct sock *sk, int level, int optname,
560 char __user *optval, unsigned int optlen);
561int compat_ip_getsockopt(struct sock *sk, int level, int optname,
562 char __user *optval, int __user *optlen);
563int ip_ra_control(struct sock *sk, unsigned char on,
564 void (*destructor)(struct sock *));
565
566int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
567void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
568 u32 info, u8 *payload);
569void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
570 u32 info);
571
572static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
573{
574 ip_cmsg_recv_offset(msg, skb, 0);
575}
576
577bool icmp_global_allow(void);
578extern int sysctl_icmp_msgs_per_sec;
579extern int sysctl_icmp_msgs_burst;
580
581#ifdef CONFIG_PROC_FS
582int ip_misc_proc_init(void);
583#endif
584
585#endif /* _IP_H */