xref: /linux/include/net/ip.h (revision 3932b9ca55b0be314a36d3e84faff3e823c081f5)
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_keys.h>
35 
36 struct sock;
37 
38 struct inet_skb_parm {
39 	struct ip_options	opt;		/* Compiled IP options		*/
40 	unsigned char		flags;
41 
42 #define IPSKB_FORWARDED		1
43 #define IPSKB_XFRM_TUNNEL_SIZE	2
44 #define IPSKB_XFRM_TRANSFORMED	4
45 #define IPSKB_FRAG_COMPLETE	8
46 #define IPSKB_REROUTED		16
47 
48 	u16			frag_max_size;
49 };
50 
51 static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
52 {
53 	return ip_hdr(skb)->ihl * 4;
54 }
55 
56 struct ipcm_cookie {
57 	__be32			addr;
58 	int			oif;
59 	struct ip_options_rcu	*opt;
60 	__u8			tx_flags;
61 	__u8			ttl;
62 	__s16			tos;
63 	char			priority;
64 };
65 
66 #define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
67 #define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
68 
69 struct ip_ra_chain {
70 	struct ip_ra_chain __rcu *next;
71 	struct sock		*sk;
72 	union {
73 		void			(*destructor)(struct sock *);
74 		struct sock		*saved_sk;
75 	};
76 	struct rcu_head		rcu;
77 };
78 
79 extern struct ip_ra_chain __rcu *ip_ra_chain;
80 
81 /* IP flags. */
82 #define IP_CE		0x8000		/* Flag: "Congestion"		*/
83 #define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
84 #define IP_MF		0x2000		/* Flag: "More Fragments"	*/
85 #define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
86 
87 #define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
88 
89 struct msghdr;
90 struct net_device;
91 struct packet_type;
92 struct rtable;
93 struct sockaddr;
94 
95 int igmp_mc_init(void);
96 
97 /*
98  *	Functions provided by ip.c
99  */
100 
101 int ip_build_and_send_pkt(struct sk_buff *skb, struct sock *sk,
102 			  __be32 saddr, __be32 daddr,
103 			  struct ip_options_rcu *opt);
104 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
105 	   struct net_device *orig_dev);
106 int ip_local_deliver(struct sk_buff *skb);
107 int ip_mr_input(struct sk_buff *skb);
108 int ip_output(struct sock *sk, struct sk_buff *skb);
109 int ip_mc_output(struct sock *sk, struct sk_buff *skb);
110 int ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *));
111 int ip_do_nat(struct sk_buff *skb);
112 void ip_send_check(struct iphdr *ip);
113 int __ip_local_out(struct sk_buff *skb);
114 int ip_local_out_sk(struct sock *sk, struct sk_buff *skb);
115 static inline int ip_local_out(struct sk_buff *skb)
116 {
117 	return ip_local_out_sk(skb->sk, skb);
118 }
119 
120 int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
121 void ip_init(void);
122 int ip_append_data(struct sock *sk, struct flowi4 *fl4,
123 		   int getfrag(void *from, char *to, int offset, int len,
124 			       int odd, struct sk_buff *skb),
125 		   void *from, int len, int protolen,
126 		   struct ipcm_cookie *ipc,
127 		   struct rtable **rt,
128 		   unsigned int flags);
129 int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
130 		       struct sk_buff *skb);
131 ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
132 		       int offset, size_t size, int flags);
133 struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
134 			      struct sk_buff_head *queue,
135 			      struct inet_cork *cork);
136 int ip_send_skb(struct net *net, struct sk_buff *skb);
137 int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
138 void ip_flush_pending_frames(struct sock *sk);
139 struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
140 			    int getfrag(void *from, char *to, int offset,
141 					int len, int odd, struct sk_buff *skb),
142 			    void *from, int length, int transhdrlen,
143 			    struct ipcm_cookie *ipc, struct rtable **rtp,
144 			    unsigned int flags);
145 
146 static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
147 {
148 	return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
149 }
150 
151 static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
152 {
153 	return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
154 }
155 
156 static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
157 {
158 	return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
159 }
160 
161 /* datagram.c */
162 int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
163 
164 void ip4_datagram_release_cb(struct sock *sk);
165 
166 struct ip_reply_arg {
167 	struct kvec iov[1];
168 	int	    flags;
169 	__wsum 	    csum;
170 	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
171 				/* -1 if not needed */
172 	int	    bound_dev_if;
173 	u8  	    tos;
174 };
175 
176 #define IP_REPLY_ARG_NOSRCCHECK 1
177 
178 static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
179 {
180 	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
181 }
182 
183 void ip_send_unicast_reply(struct net *net, struct sk_buff *skb, __be32 daddr,
184 			   __be32 saddr, const struct ip_reply_arg *arg,
185 			   unsigned int len);
186 
187 #define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
188 #define IP_INC_STATS_BH(net, field)	SNMP_INC_STATS64_BH((net)->mib.ip_statistics, field)
189 #define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
190 #define IP_ADD_STATS_BH(net, field, val) SNMP_ADD_STATS64_BH((net)->mib.ip_statistics, field, val)
191 #define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
192 #define IP_UPD_PO_STATS_BH(net, field, val) SNMP_UPD_PO_STATS64_BH((net)->mib.ip_statistics, field, val)
193 #define NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
194 #define NET_INC_STATS_BH(net, field)	SNMP_INC_STATS_BH((net)->mib.net_statistics, field)
195 #define NET_INC_STATS_USER(net, field) 	SNMP_INC_STATS_USER((net)->mib.net_statistics, field)
196 #define NET_ADD_STATS(net, field, adnd)	SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
197 #define NET_ADD_STATS_BH(net, field, adnd) SNMP_ADD_STATS_BH((net)->mib.net_statistics, field, adnd)
198 #define NET_ADD_STATS_USER(net, field, adnd) SNMP_ADD_STATS_USER((net)->mib.net_statistics, field, adnd)
199 
200 unsigned long snmp_fold_field(void __percpu *mib, int offt);
201 #if BITS_PER_LONG==32
202 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
203 #else
204 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
205 {
206 	return snmp_fold_field(mib, offt);
207 }
208 #endif
209 
210 void inet_get_local_port_range(struct net *net, int *low, int *high);
211 
212 #ifdef CONFIG_SYSCTL
213 static inline int inet_is_local_reserved_port(struct net *net, int port)
214 {
215 	if (!net->ipv4.sysctl_local_reserved_ports)
216 		return 0;
217 	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
218 }
219 
220 static inline bool sysctl_dev_name_is_allowed(const char *name)
221 {
222 	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
223 }
224 
225 #else
226 static inline int inet_is_local_reserved_port(struct net *net, int port)
227 {
228 	return 0;
229 }
230 #endif
231 
232 extern int sysctl_ip_nonlocal_bind;
233 
234 /* From inetpeer.c */
235 extern int inet_peer_threshold;
236 extern int inet_peer_minttl;
237 extern int inet_peer_maxttl;
238 
239 /* From ip_input.c */
240 extern int sysctl_ip_early_demux;
241 
242 /* From ip_output.c */
243 extern int sysctl_ip_dynaddr;
244 
245 void ipfrag_init(void);
246 
247 void ip_static_sysctl_init(void);
248 
249 #define IP4_REPLY_MARK(net, mark) \
250 	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
251 
252 static inline bool ip_is_fragment(const struct iphdr *iph)
253 {
254 	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
255 }
256 
257 #ifdef CONFIG_INET
258 #include <net/dst.h>
259 
260 /* The function in 2.2 was invalid, producing wrong result for
261  * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
262 static inline
263 int ip_decrease_ttl(struct iphdr *iph)
264 {
265 	u32 check = (__force u32)iph->check;
266 	check += (__force u32)htons(0x0100);
267 	iph->check = (__force __sum16)(check + (check>=0xFFFF));
268 	return --iph->ttl;
269 }
270 
271 static inline
272 int ip_dont_fragment(struct sock *sk, struct dst_entry *dst)
273 {
274 	return  inet_sk(sk)->pmtudisc == IP_PMTUDISC_DO ||
275 		(inet_sk(sk)->pmtudisc == IP_PMTUDISC_WANT &&
276 		 !(dst_metric_locked(dst, RTAX_MTU)));
277 }
278 
279 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
280 {
281 	return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
282 	       inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
283 }
284 
285 static inline bool ip_sk_use_pmtu(const struct sock *sk)
286 {
287 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
288 }
289 
290 static inline bool ip_sk_ignore_df(const struct sock *sk)
291 {
292 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
293 	       inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
294 }
295 
296 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
297 						    bool forwarding)
298 {
299 	struct net *net = dev_net(dst->dev);
300 
301 	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
302 	    dst_metric_locked(dst, RTAX_MTU) ||
303 	    !forwarding)
304 		return dst_mtu(dst);
305 
306 	return min(dst->dev->mtu, IP_MAX_MTU);
307 }
308 
309 static inline unsigned int ip_skb_dst_mtu(const struct sk_buff *skb)
310 {
311 	if (!skb->sk || ip_sk_use_pmtu(skb->sk)) {
312 		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
313 		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
314 	} else {
315 		return min(skb_dst(skb)->dev->mtu, IP_MAX_MTU);
316 	}
317 }
318 
319 u32 ip_idents_reserve(u32 hash, int segs);
320 void __ip_select_ident(struct iphdr *iph, int segs);
321 
322 static inline void ip_select_ident_segs(struct sk_buff *skb, struct sock *sk, int segs)
323 {
324 	struct iphdr *iph = ip_hdr(skb);
325 
326 	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
327 		/* This is only to work around buggy Windows95/2000
328 		 * VJ compression implementations.  If the ID field
329 		 * does not change, they drop every other packet in
330 		 * a TCP stream using header compression.
331 		 */
332 		if (sk && inet_sk(sk)->inet_daddr) {
333 			iph->id = htons(inet_sk(sk)->inet_id);
334 			inet_sk(sk)->inet_id += segs;
335 		} else {
336 			iph->id = 0;
337 		}
338 	} else {
339 		__ip_select_ident(iph, segs);
340 	}
341 }
342 
343 static inline void ip_select_ident(struct sk_buff *skb, struct sock *sk)
344 {
345 	ip_select_ident_segs(skb, sk, 1);
346 }
347 
348 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
349 {
350 	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
351 				  skb->len, proto, 0);
352 }
353 
354 static inline void inet_set_txhash(struct sock *sk)
355 {
356 	struct inet_sock *inet = inet_sk(sk);
357 	struct flow_keys keys;
358 
359 	keys.src = inet->inet_saddr;
360 	keys.dst = inet->inet_daddr;
361 	keys.port16[0] = inet->inet_sport;
362 	keys.port16[1] = inet->inet_dport;
363 
364 	sk->sk_txhash = flow_hash_from_keys(&keys);
365 }
366 
367 /*
368  *	Map a multicast IP onto multicast MAC for type ethernet.
369  */
370 
371 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
372 {
373 	__u32 addr=ntohl(naddr);
374 	buf[0]=0x01;
375 	buf[1]=0x00;
376 	buf[2]=0x5e;
377 	buf[5]=addr&0xFF;
378 	addr>>=8;
379 	buf[4]=addr&0xFF;
380 	addr>>=8;
381 	buf[3]=addr&0x7F;
382 }
383 
384 /*
385  *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
386  *	Leave P_Key as 0 to be filled in by driver.
387  */
388 
389 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
390 {
391 	__u32 addr;
392 	unsigned char scope = broadcast[5] & 0xF;
393 
394 	buf[0]  = 0;		/* Reserved */
395 	buf[1]  = 0xff;		/* Multicast QPN */
396 	buf[2]  = 0xff;
397 	buf[3]  = 0xff;
398 	addr    = ntohl(naddr);
399 	buf[4]  = 0xff;
400 	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
401 	buf[6]  = 0x40;		/* IPv4 signature */
402 	buf[7]  = 0x1b;
403 	buf[8]  = broadcast[8];		/* P_Key */
404 	buf[9]  = broadcast[9];
405 	buf[10] = 0;
406 	buf[11] = 0;
407 	buf[12] = 0;
408 	buf[13] = 0;
409 	buf[14] = 0;
410 	buf[15] = 0;
411 	buf[19] = addr & 0xff;
412 	addr  >>= 8;
413 	buf[18] = addr & 0xff;
414 	addr  >>= 8;
415 	buf[17] = addr & 0xff;
416 	addr  >>= 8;
417 	buf[16] = addr & 0x0f;
418 }
419 
420 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
421 {
422 	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
423 		memcpy(buf, broadcast, 4);
424 	else
425 		memcpy(buf, &naddr, sizeof(naddr));
426 }
427 
428 #if IS_ENABLED(CONFIG_IPV6)
429 #include <linux/ipv6.h>
430 #endif
431 
432 static __inline__ void inet_reset_saddr(struct sock *sk)
433 {
434 	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
435 #if IS_ENABLED(CONFIG_IPV6)
436 	if (sk->sk_family == PF_INET6) {
437 		struct ipv6_pinfo *np = inet6_sk(sk);
438 
439 		memset(&np->saddr, 0, sizeof(np->saddr));
440 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
441 	}
442 #endif
443 }
444 
445 #endif
446 
447 static inline int sk_mc_loop(struct sock *sk)
448 {
449 	if (!sk)
450 		return 1;
451 	switch (sk->sk_family) {
452 	case AF_INET:
453 		return inet_sk(sk)->mc_loop;
454 #if IS_ENABLED(CONFIG_IPV6)
455 	case AF_INET6:
456 		return inet6_sk(sk)->mc_loop;
457 #endif
458 	}
459 	WARN_ON(1);
460 	return 1;
461 }
462 
463 bool ip_call_ra_chain(struct sk_buff *skb);
464 
465 /*
466  *	Functions provided by ip_fragment.c
467  */
468 
469 enum ip_defrag_users {
470 	IP_DEFRAG_LOCAL_DELIVER,
471 	IP_DEFRAG_CALL_RA_CHAIN,
472 	IP_DEFRAG_CONNTRACK_IN,
473 	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
474 	IP_DEFRAG_CONNTRACK_OUT,
475 	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
476 	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
477 	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
478 	IP_DEFRAG_VS_IN,
479 	IP_DEFRAG_VS_OUT,
480 	IP_DEFRAG_VS_FWD,
481 	IP_DEFRAG_AF_PACKET,
482 	IP_DEFRAG_MACVLAN,
483 };
484 
485 int ip_defrag(struct sk_buff *skb, u32 user);
486 #ifdef CONFIG_INET
487 struct sk_buff *ip_check_defrag(struct sk_buff *skb, u32 user);
488 #else
489 static inline struct sk_buff *ip_check_defrag(struct sk_buff *skb, u32 user)
490 {
491 	return skb;
492 }
493 #endif
494 int ip_frag_mem(struct net *net);
495 
496 /*
497  *	Functions provided by ip_forward.c
498  */
499 
500 int ip_forward(struct sk_buff *skb);
501 
502 /*
503  *	Functions provided by ip_options.c
504  */
505 
506 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
507 		      __be32 daddr, struct rtable *rt, int is_frag);
508 int ip_options_echo(struct ip_options *dopt, struct sk_buff *skb);
509 void ip_options_fragment(struct sk_buff *skb);
510 int ip_options_compile(struct net *net, struct ip_options *opt,
511 		       struct sk_buff *skb);
512 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
513 		   unsigned char *data, int optlen);
514 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
515 			     unsigned char __user *data, int optlen);
516 void ip_options_undo(struct ip_options *opt);
517 void ip_forward_options(struct sk_buff *skb);
518 int ip_options_rcv_srr(struct sk_buff *skb);
519 
520 /*
521  *	Functions provided by ip_sockglue.c
522  */
523 
524 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
525 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb);
526 int ip_cmsg_send(struct net *net, struct msghdr *msg,
527 		 struct ipcm_cookie *ipc, bool allow_ipv6);
528 int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
529 		  unsigned int optlen);
530 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
531 		  int __user *optlen);
532 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
533 			 char __user *optval, unsigned int optlen);
534 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
535 			 char __user *optval, int __user *optlen);
536 int ip_ra_control(struct sock *sk, unsigned char on,
537 		  void (*destructor)(struct sock *));
538 
539 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
540 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
541 		   u32 info, u8 *payload);
542 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
543 		    u32 info);
544 
545 #ifdef CONFIG_PROC_FS
546 int ip_misc_proc_init(void);
547 #endif
548 
549 #endif	/* _IP_H */
550