Loading...
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, <waltje@uWalt.NL.Mugnet.ORG>
12 * Alan Cox, <gw4pts@gw4pts.ampr.org>
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 struct ip_options opt; /* Compiled IP options */
40 unsigned char flags;
41
42#define IPSKB_FORWARDED BIT(0)
43#define IPSKB_XFRM_TUNNEL_SIZE BIT(1)
44#define IPSKB_XFRM_TRANSFORMED BIT(2)
45#define IPSKB_FRAG_COMPLETE BIT(3)
46#define IPSKB_REROUTED BIT(4)
47#define IPSKB_DOREDIRECT BIT(5)
48#define IPSKB_FRAG_PMTU BIT(6)
49
50 u16 frag_max_size;
51};
52
53static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
54{
55 return ip_hdr(skb)->ihl * 4;
56}
57
58struct ipcm_cookie {
59 __be32 addr;
60 int oif;
61 struct ip_options_rcu *opt;
62 __u8 tx_flags;
63 __u8 ttl;
64 __s16 tos;
65 char priority;
66};
67
68#define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
69#define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
70
71struct ip_ra_chain {
72 struct ip_ra_chain __rcu *next;
73 struct sock *sk;
74 union {
75 void (*destructor)(struct sock *);
76 struct sock *saved_sk;
77 };
78 struct rcu_head rcu;
79};
80
81extern struct ip_ra_chain __rcu *ip_ra_chain;
82
83/* IP flags. */
84#define IP_CE 0x8000 /* Flag: "Congestion" */
85#define IP_DF 0x4000 /* Flag: "Don't Fragment" */
86#define IP_MF 0x2000 /* Flag: "More Fragments" */
87#define IP_OFFSET 0x1FFF /* "Fragment Offset" part */
88
89#define IP_FRAG_TIME (30 * HZ) /* fragment lifetime */
90
91struct msghdr;
92struct net_device;
93struct packet_type;
94struct rtable;
95struct sockaddr;
96
97int igmp_mc_init(void);
98
99/*
100 * Functions provided by ip.c
101 */
102
103int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
104 __be32 saddr, __be32 daddr,
105 struct ip_options_rcu *opt);
106int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
107 struct net_device *orig_dev);
108int ip_local_deliver(struct sk_buff *skb);
109int ip_mr_input(struct sk_buff *skb);
110int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
111int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
112int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
113 int (*output)(struct net *, struct sock *, struct sk_buff *));
114void ip_send_check(struct iphdr *ip);
115int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
116int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
117
118int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
119void ip_init(void);
120int ip_append_data(struct sock *sk, struct flowi4 *fl4,
121 int getfrag(void *from, char *to, int offset, int len,
122 int odd, struct sk_buff *skb),
123 void *from, int len, int protolen,
124 struct ipcm_cookie *ipc,
125 struct rtable **rt,
126 unsigned int flags);
127int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
128 struct sk_buff *skb);
129ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
130 int offset, size_t size, int flags);
131struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
132 struct sk_buff_head *queue,
133 struct inet_cork *cork);
134int ip_send_skb(struct net *net, struct sk_buff *skb);
135int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
136void ip_flush_pending_frames(struct sock *sk);
137struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
138 int getfrag(void *from, char *to, int offset,
139 int len, int odd, struct sk_buff *skb),
140 void *from, int length, int transhdrlen,
141 struct ipcm_cookie *ipc, struct rtable **rtp,
142 unsigned int flags);
143
144static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
145{
146 return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
147}
148
149static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
150{
151 return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
152}
153
154static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
155{
156 return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
157}
158
159/* datagram.c */
160int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
161int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
162
163void ip4_datagram_release_cb(struct sock *sk);
164
165struct ip_reply_arg {
166 struct kvec iov[1];
167 int flags;
168 __wsum csum;
169 int csumoffset; /* u16 offset of csum in iov[0].iov_base */
170 /* -1 if not needed */
171 int bound_dev_if;
172 u8 tos;
173};
174
175#define IP_REPLY_ARG_NOSRCCHECK 1
176
177static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
178{
179 return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
180}
181
182void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
183 const struct ip_options *sopt,
184 __be32 daddr, __be32 saddr,
185 const struct ip_reply_arg *arg,
186 unsigned int len);
187
188#define IP_INC_STATS(net, field) SNMP_INC_STATS64((net)->mib.ip_statistics, field)
189#define IP_INC_STATS_BH(net, field) SNMP_INC_STATS64_BH((net)->mib.ip_statistics, field)
190#define IP_ADD_STATS(net, field, val) SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
191#define IP_ADD_STATS_BH(net, field, val) SNMP_ADD_STATS64_BH((net)->mib.ip_statistics, field, val)
192#define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
193#define IP_UPD_PO_STATS_BH(net, field, val) SNMP_UPD_PO_STATS64_BH((net)->mib.ip_statistics, field, val)
194#define NET_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.net_statistics, field)
195#define NET_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.net_statistics, field)
196#define NET_INC_STATS_USER(net, field) SNMP_INC_STATS_USER((net)->mib.net_statistics, field)
197#define NET_ADD_STATS(net, field, adnd) SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
198#define NET_ADD_STATS_BH(net, field, adnd) SNMP_ADD_STATS_BH((net)->mib.net_statistics, field, adnd)
199#define NET_ADD_STATS_USER(net, field, adnd) SNMP_ADD_STATS_USER((net)->mib.net_statistics, field, adnd)
200
201u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
202unsigned long snmp_fold_field(void __percpu *mib, int offt);
203#if BITS_PER_LONG==32
204u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
205 size_t syncp_offset);
206u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
207#else
208static inline u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
209 size_t syncp_offset)
210{
211 return snmp_get_cpu_field(mib, cpu, offct);
212
213}
214
215static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
216{
217 return snmp_fold_field(mib, offt);
218}
219#endif
220
221void inet_get_local_port_range(struct net *net, int *low, int *high);
222
223#ifdef CONFIG_SYSCTL
224static inline int inet_is_local_reserved_port(struct net *net, int port)
225{
226 if (!net->ipv4.sysctl_local_reserved_ports)
227 return 0;
228 return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
229}
230
231static inline bool sysctl_dev_name_is_allowed(const char *name)
232{
233 return strcmp(name, "default") != 0 && strcmp(name, "all") != 0;
234}
235
236#else
237static inline int inet_is_local_reserved_port(struct net *net, int port)
238{
239 return 0;
240}
241#endif
242
243__be32 inet_current_timestamp(void);
244
245/* From inetpeer.c */
246extern int inet_peer_threshold;
247extern int inet_peer_minttl;
248extern int inet_peer_maxttl;
249
250void ipfrag_init(void);
251
252void ip_static_sysctl_init(void);
253
254#define IP4_REPLY_MARK(net, mark) \
255 ((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
256
257static inline bool ip_is_fragment(const struct iphdr *iph)
258{
259 return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
260}
261
262#ifdef CONFIG_INET
263#include <net/dst.h>
264
265/* The function in 2.2 was invalid, producing wrong result for
266 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
267static inline
268int ip_decrease_ttl(struct iphdr *iph)
269{
270 u32 check = (__force u32)iph->check;
271 check += (__force u32)htons(0x0100);
272 iph->check = (__force __sum16)(check + (check>=0xFFFF));
273 return --iph->ttl;
274}
275
276static inline
277int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
278{
279 u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
280
281 return pmtudisc == IP_PMTUDISC_DO ||
282 (pmtudisc == IP_PMTUDISC_WANT &&
283 !(dst_metric_locked(dst, RTAX_MTU)));
284}
285
286static inline bool ip_sk_accept_pmtu(const struct sock *sk)
287{
288 return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
289 inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
290}
291
292static inline bool ip_sk_use_pmtu(const struct sock *sk)
293{
294 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
295}
296
297static inline bool ip_sk_ignore_df(const struct sock *sk)
298{
299 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
300 inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
301}
302
303static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
304 bool forwarding)
305{
306 struct net *net = dev_net(dst->dev);
307
308 if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
309 dst_metric_locked(dst, RTAX_MTU) ||
310 !forwarding)
311 return dst_mtu(dst);
312
313 return min(dst->dev->mtu, IP_MAX_MTU);
314}
315
316static inline unsigned int ip_skb_dst_mtu(const struct sk_buff *skb)
317{
318 struct sock *sk = skb->sk;
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 net *net, 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 */
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, <waltje@uWalt.NL.Mugnet.ORG>
12 * Alan Cox, <gw4pts@gw4pts.ampr.org>
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 u16 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_L3SLAVE BIT(7)
51
52 u16 frag_max_size;
53};
54
55static inline bool ipv4_l3mdev_skb(u16 flags)
56{
57 return !!(flags & IPSKB_L3SLAVE);
58}
59
60static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
61{
62 return ip_hdr(skb)->ihl * 4;
63}
64
65struct ipcm_cookie {
66 struct sockcm_cookie sockc;
67 __be32 addr;
68 int oif;
69 struct ip_options_rcu *opt;
70 __u8 tx_flags;
71 __u8 ttl;
72 __s16 tos;
73 char priority;
74};
75
76#define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
77#define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
78
79struct ip_ra_chain {
80 struct ip_ra_chain __rcu *next;
81 struct sock *sk;
82 union {
83 void (*destructor)(struct sock *);
84 struct sock *saved_sk;
85 };
86 struct rcu_head rcu;
87};
88
89extern struct ip_ra_chain __rcu *ip_ra_chain;
90
91/* IP flags. */
92#define IP_CE 0x8000 /* Flag: "Congestion" */
93#define IP_DF 0x4000 /* Flag: "Don't Fragment" */
94#define IP_MF 0x2000 /* Flag: "More Fragments" */
95#define IP_OFFSET 0x1FFF /* "Fragment Offset" part */
96
97#define IP_FRAG_TIME (30 * HZ) /* fragment lifetime */
98
99struct msghdr;
100struct net_device;
101struct packet_type;
102struct rtable;
103struct sockaddr;
104
105int igmp_mc_init(void);
106
107/*
108 * Functions provided by ip.c
109 */
110
111int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
112 __be32 saddr, __be32 daddr,
113 struct ip_options_rcu *opt);
114int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
115 struct net_device *orig_dev);
116int ip_local_deliver(struct sk_buff *skb);
117int ip_mr_input(struct sk_buff *skb);
118int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
119int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
120int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
121 int (*output)(struct net *, struct sock *, struct sk_buff *));
122void ip_send_check(struct iphdr *ip);
123int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
124int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
125
126int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
127void ip_init(void);
128int ip_append_data(struct sock *sk, struct flowi4 *fl4,
129 int getfrag(void *from, char *to, int offset, int len,
130 int odd, struct sk_buff *skb),
131 void *from, int len, int protolen,
132 struct ipcm_cookie *ipc,
133 struct rtable **rt,
134 unsigned int flags);
135int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
136 struct sk_buff *skb);
137ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
138 int offset, size_t size, int flags);
139struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
140 struct sk_buff_head *queue,
141 struct inet_cork *cork);
142int ip_send_skb(struct net *net, struct sk_buff *skb);
143int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
144void ip_flush_pending_frames(struct sock *sk);
145struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
146 int getfrag(void *from, char *to, int offset,
147 int len, int odd, struct sk_buff *skb),
148 void *from, int length, int transhdrlen,
149 struct ipcm_cookie *ipc, struct rtable **rtp,
150 unsigned int flags);
151
152static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
153{
154 return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
155}
156
157static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
158{
159 return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
160}
161
162static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
163{
164 return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
165}
166
167/* datagram.c */
168int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
169int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
170
171void ip4_datagram_release_cb(struct sock *sk);
172
173struct ip_reply_arg {
174 struct kvec iov[1];
175 int flags;
176 __wsum csum;
177 int csumoffset; /* u16 offset of csum in iov[0].iov_base */
178 /* -1 if not needed */
179 int bound_dev_if;
180 u8 tos;
181 kuid_t uid;
182};
183
184#define IP_REPLY_ARG_NOSRCCHECK 1
185
186static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
187{
188 return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
189}
190
191void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
192 const struct ip_options *sopt,
193 __be32 daddr, __be32 saddr,
194 const struct ip_reply_arg *arg,
195 unsigned int len);
196
197#define IP_INC_STATS(net, field) SNMP_INC_STATS64((net)->mib.ip_statistics, field)
198#define __IP_INC_STATS(net, field) __SNMP_INC_STATS64((net)->mib.ip_statistics, field)
199#define IP_ADD_STATS(net, field, val) SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
200#define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
201#define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
202#define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
203#define NET_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.net_statistics, field)
204#define __NET_INC_STATS(net, field) __SNMP_INC_STATS((net)->mib.net_statistics, field)
205#define NET_ADD_STATS(net, field, adnd) SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
206#define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
207
208u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
209unsigned long snmp_fold_field(void __percpu *mib, int offt);
210#if BITS_PER_LONG==32
211u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
212 size_t syncp_offset);
213u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
214#else
215static inline u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
216 size_t syncp_offset)
217{
218 return snmp_get_cpu_field(mib, cpu, offct);
219
220}
221
222static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
223{
224 return snmp_fold_field(mib, offt);
225}
226#endif
227
228#define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
229{ \
230 int i, c; \
231 for_each_possible_cpu(c) { \
232 for (i = 0; stats_list[i].name; i++) \
233 buff64[i] += snmp_get_cpu_field64( \
234 mib_statistic, \
235 c, stats_list[i].entry, \
236 offset); \
237 } \
238}
239
240#define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
241{ \
242 int i, c; \
243 for_each_possible_cpu(c) { \
244 for (i = 0; stats_list[i].name; i++) \
245 buff[i] += snmp_get_cpu_field( \
246 mib_statistic, \
247 c, stats_list[i].entry); \
248 } \
249}
250
251void inet_get_local_port_range(struct net *net, int *low, int *high);
252
253#ifdef CONFIG_SYSCTL
254static inline int inet_is_local_reserved_port(struct net *net, int port)
255{
256 if (!net->ipv4.sysctl_local_reserved_ports)
257 return 0;
258 return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
259}
260
261static inline bool sysctl_dev_name_is_allowed(const char *name)
262{
263 return strcmp(name, "default") != 0 && strcmp(name, "all") != 0;
264}
265
266#else
267static inline int inet_is_local_reserved_port(struct net *net, int port)
268{
269 return 0;
270}
271#endif
272
273__be32 inet_current_timestamp(void);
274
275/* From inetpeer.c */
276extern int inet_peer_threshold;
277extern int inet_peer_minttl;
278extern int inet_peer_maxttl;
279
280void ipfrag_init(void);
281
282void ip_static_sysctl_init(void);
283
284#define IP4_REPLY_MARK(net, mark) \
285 ((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
286
287static inline bool ip_is_fragment(const struct iphdr *iph)
288{
289 return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
290}
291
292#ifdef CONFIG_INET
293#include <net/dst.h>
294
295/* The function in 2.2 was invalid, producing wrong result for
296 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
297static inline
298int ip_decrease_ttl(struct iphdr *iph)
299{
300 u32 check = (__force u32)iph->check;
301 check += (__force u32)htons(0x0100);
302 iph->check = (__force __sum16)(check + (check>=0xFFFF));
303 return --iph->ttl;
304}
305
306static inline
307int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
308{
309 u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
310
311 return pmtudisc == IP_PMTUDISC_DO ||
312 (pmtudisc == IP_PMTUDISC_WANT &&
313 !(dst_metric_locked(dst, RTAX_MTU)));
314}
315
316static inline bool ip_sk_accept_pmtu(const struct sock *sk)
317{
318 return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
319 inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
320}
321
322static inline bool ip_sk_use_pmtu(const struct sock *sk)
323{
324 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
325}
326
327static inline bool ip_sk_ignore_df(const struct sock *sk)
328{
329 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
330 inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
331}
332
333static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
334 bool forwarding)
335{
336 struct net *net = dev_net(dst->dev);
337
338 if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
339 dst_metric_locked(dst, RTAX_MTU) ||
340 !forwarding)
341 return dst_mtu(dst);
342
343 return min(dst->dev->mtu, IP_MAX_MTU);
344}
345
346static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
347 const struct sk_buff *skb)
348{
349 if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
350 bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
351
352 return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
353 }
354
355 return min(skb_dst(skb)->dev->mtu, IP_MAX_MTU);
356}
357
358u32 ip_idents_reserve(u32 hash, int segs);
359void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
360
361static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
362 struct sock *sk, int segs)
363{
364 struct iphdr *iph = ip_hdr(skb);
365
366 if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
367 /* This is only to work around buggy Windows95/2000
368 * VJ compression implementations. If the ID field
369 * does not change, they drop every other packet in
370 * a TCP stream using header compression.
371 */
372 if (sk && inet_sk(sk)->inet_daddr) {
373 iph->id = htons(inet_sk(sk)->inet_id);
374 inet_sk(sk)->inet_id += segs;
375 } else {
376 iph->id = 0;
377 }
378 } else {
379 __ip_select_ident(net, iph, segs);
380 }
381}
382
383static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
384 struct sock *sk)
385{
386 ip_select_ident_segs(net, skb, sk, 1);
387}
388
389static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
390{
391 return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
392 skb->len, proto, 0);
393}
394
395/* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
396 * Equivalent to : flow->v4addrs.src = iph->saddr;
397 * flow->v4addrs.dst = iph->daddr;
398 */
399static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
400 const struct iphdr *iph)
401{
402 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
403 offsetof(typeof(flow->addrs), v4addrs.src) +
404 sizeof(flow->addrs.v4addrs.src));
405 memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
406 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
407}
408
409static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
410{
411 const struct iphdr *iph = skb_gro_network_header(skb);
412
413 return csum_tcpudp_nofold(iph->saddr, iph->daddr,
414 skb_gro_len(skb), proto, 0);
415}
416
417/*
418 * Map a multicast IP onto multicast MAC for type ethernet.
419 */
420
421static inline void ip_eth_mc_map(__be32 naddr, char *buf)
422{
423 __u32 addr=ntohl(naddr);
424 buf[0]=0x01;
425 buf[1]=0x00;
426 buf[2]=0x5e;
427 buf[5]=addr&0xFF;
428 addr>>=8;
429 buf[4]=addr&0xFF;
430 addr>>=8;
431 buf[3]=addr&0x7F;
432}
433
434/*
435 * Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
436 * Leave P_Key as 0 to be filled in by driver.
437 */
438
439static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
440{
441 __u32 addr;
442 unsigned char scope = broadcast[5] & 0xF;
443
444 buf[0] = 0; /* Reserved */
445 buf[1] = 0xff; /* Multicast QPN */
446 buf[2] = 0xff;
447 buf[3] = 0xff;
448 addr = ntohl(naddr);
449 buf[4] = 0xff;
450 buf[5] = 0x10 | scope; /* scope from broadcast address */
451 buf[6] = 0x40; /* IPv4 signature */
452 buf[7] = 0x1b;
453 buf[8] = broadcast[8]; /* P_Key */
454 buf[9] = broadcast[9];
455 buf[10] = 0;
456 buf[11] = 0;
457 buf[12] = 0;
458 buf[13] = 0;
459 buf[14] = 0;
460 buf[15] = 0;
461 buf[19] = addr & 0xff;
462 addr >>= 8;
463 buf[18] = addr & 0xff;
464 addr >>= 8;
465 buf[17] = addr & 0xff;
466 addr >>= 8;
467 buf[16] = addr & 0x0f;
468}
469
470static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
471{
472 if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
473 memcpy(buf, broadcast, 4);
474 else
475 memcpy(buf, &naddr, sizeof(naddr));
476}
477
478#if IS_ENABLED(CONFIG_IPV6)
479#include <linux/ipv6.h>
480#endif
481
482static __inline__ void inet_reset_saddr(struct sock *sk)
483{
484 inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
485#if IS_ENABLED(CONFIG_IPV6)
486 if (sk->sk_family == PF_INET6) {
487 struct ipv6_pinfo *np = inet6_sk(sk);
488
489 memset(&np->saddr, 0, sizeof(np->saddr));
490 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
491 }
492#endif
493}
494
495#endif
496
497static inline unsigned int ipv4_addr_hash(__be32 ip)
498{
499 return (__force unsigned int) ip;
500}
501
502bool ip_call_ra_chain(struct sk_buff *skb);
503
504/*
505 * Functions provided by ip_fragment.c
506 */
507
508enum ip_defrag_users {
509 IP_DEFRAG_LOCAL_DELIVER,
510 IP_DEFRAG_CALL_RA_CHAIN,
511 IP_DEFRAG_CONNTRACK_IN,
512 __IP_DEFRAG_CONNTRACK_IN_END = IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
513 IP_DEFRAG_CONNTRACK_OUT,
514 __IP_DEFRAG_CONNTRACK_OUT_END = IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
515 IP_DEFRAG_CONNTRACK_BRIDGE_IN,
516 __IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
517 IP_DEFRAG_VS_IN,
518 IP_DEFRAG_VS_OUT,
519 IP_DEFRAG_VS_FWD,
520 IP_DEFRAG_AF_PACKET,
521 IP_DEFRAG_MACVLAN,
522};
523
524/* Return true if the value of 'user' is between 'lower_bond'
525 * and 'upper_bond' inclusively.
526 */
527static inline bool ip_defrag_user_in_between(u32 user,
528 enum ip_defrag_users lower_bond,
529 enum ip_defrag_users upper_bond)
530{
531 return user >= lower_bond && user <= upper_bond;
532}
533
534int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
535#ifdef CONFIG_INET
536struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
537#else
538static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
539{
540 return skb;
541}
542#endif
543int ip_frag_mem(struct net *net);
544
545/*
546 * Functions provided by ip_forward.c
547 */
548
549int ip_forward(struct sk_buff *skb);
550
551/*
552 * Functions provided by ip_options.c
553 */
554
555void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
556 __be32 daddr, struct rtable *rt, int is_frag);
557
558int __ip_options_echo(struct ip_options *dopt, struct sk_buff *skb,
559 const struct ip_options *sopt);
560static inline int ip_options_echo(struct ip_options *dopt, struct sk_buff *skb)
561{
562 return __ip_options_echo(dopt, skb, &IPCB(skb)->opt);
563}
564
565void ip_options_fragment(struct sk_buff *skb);
566int ip_options_compile(struct net *net, struct ip_options *opt,
567 struct sk_buff *skb);
568int ip_options_get(struct net *net, struct ip_options_rcu **optp,
569 unsigned char *data, int optlen);
570int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
571 unsigned char __user *data, int optlen);
572void ip_options_undo(struct ip_options *opt);
573void ip_forward_options(struct sk_buff *skb);
574int ip_options_rcv_srr(struct sk_buff *skb);
575
576/*
577 * Functions provided by ip_sockglue.c
578 */
579
580void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
581void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
582 struct sk_buff *skb, int tlen, int offset);
583int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
584 struct ipcm_cookie *ipc, bool allow_ipv6);
585int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
586 unsigned int optlen);
587int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
588 int __user *optlen);
589int compat_ip_setsockopt(struct sock *sk, int level, int optname,
590 char __user *optval, unsigned int optlen);
591int compat_ip_getsockopt(struct sock *sk, int level, int optname,
592 char __user *optval, int __user *optlen);
593int ip_ra_control(struct sock *sk, unsigned char on,
594 void (*destructor)(struct sock *));
595
596int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
597void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
598 u32 info, u8 *payload);
599void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
600 u32 info);
601
602static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
603{
604 ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
605}
606
607bool icmp_global_allow(void);
608extern int sysctl_icmp_msgs_per_sec;
609extern int sysctl_icmp_msgs_burst;
610
611#ifdef CONFIG_PROC_FS
612int ip_misc_proc_init(void);
613#endif
614
615#endif /* _IP_H */