xref: /linux/include/net/ip.h (revision 4949009eb8d40a441dcddcd96e101e77d31cf1b2)
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		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 
49 	u16			frag_max_size;
50 };
51 
52 static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
53 {
54 	return ip_hdr(skb)->ihl * 4;
55 }
56 
57 struct ipcm_cookie {
58 	__be32			addr;
59 	int			oif;
60 	struct ip_options_rcu	*opt;
61 	__u8			tx_flags;
62 	__u8			ttl;
63 	__s16			tos;
64 	char			priority;
65 };
66 
67 #define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
68 #define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
69 
70 struct ip_ra_chain {
71 	struct ip_ra_chain __rcu *next;
72 	struct sock		*sk;
73 	union {
74 		void			(*destructor)(struct sock *);
75 		struct sock		*saved_sk;
76 	};
77 	struct rcu_head		rcu;
78 };
79 
80 extern struct ip_ra_chain __rcu *ip_ra_chain;
81 
82 /* IP flags. */
83 #define IP_CE		0x8000		/* Flag: "Congestion"		*/
84 #define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
85 #define IP_MF		0x2000		/* Flag: "More Fragments"	*/
86 #define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
87 
88 #define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
89 
90 struct msghdr;
91 struct net_device;
92 struct packet_type;
93 struct rtable;
94 struct sockaddr;
95 
96 int igmp_mc_init(void);
97 
98 /*
99  *	Functions provided by ip.c
100  */
101 
102 int ip_build_and_send_pkt(struct sk_buff *skb, struct sock *sk,
103 			  __be32 saddr, __be32 daddr,
104 			  struct ip_options_rcu *opt);
105 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
106 	   struct net_device *orig_dev);
107 int ip_local_deliver(struct sk_buff *skb);
108 int ip_mr_input(struct sk_buff *skb);
109 int ip_output(struct sock *sk, struct sk_buff *skb);
110 int ip_mc_output(struct sock *sk, struct sk_buff *skb);
111 int ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *));
112 int ip_do_nat(struct sk_buff *skb);
113 void ip_send_check(struct iphdr *ip);
114 int __ip_local_out(struct sk_buff *skb);
115 int ip_local_out_sk(struct sock *sk, struct sk_buff *skb);
116 static inline int ip_local_out(struct sk_buff *skb)
117 {
118 	return ip_local_out_sk(skb->sk, skb);
119 }
120 
121 int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
122 void ip_init(void);
123 int 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);
130 int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
131 		       struct sk_buff *skb);
132 ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
133 		       int offset, size_t size, int flags);
134 struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
135 			      struct sk_buff_head *queue,
136 			      struct inet_cork *cork);
137 int ip_send_skb(struct net *net, struct sk_buff *skb);
138 int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
139 void ip_flush_pending_frames(struct sock *sk);
140 struct 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 
147 static 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 
152 static 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 
157 static 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 */
163 int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
164 
165 void ip4_datagram_release_cb(struct sock *sk);
166 
167 struct ip_reply_arg {
168 	struct kvec iov[1];
169 	int	    flags;
170 	__wsum 	    csum;
171 	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
172 				/* -1 if not needed */
173 	int	    bound_dev_if;
174 	u8  	    tos;
175 };
176 
177 #define IP_REPLY_ARG_NOSRCCHECK 1
178 
179 static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
180 {
181 	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
182 }
183 
184 void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
185 			   const struct ip_options *sopt,
186 			   __be32 daddr, __be32 saddr,
187 			   const struct ip_reply_arg *arg,
188 			   unsigned int len);
189 
190 #define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
191 #define IP_INC_STATS_BH(net, field)	SNMP_INC_STATS64_BH((net)->mib.ip_statistics, field)
192 #define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
193 #define IP_ADD_STATS_BH(net, field, val) SNMP_ADD_STATS64_BH((net)->mib.ip_statistics, field, val)
194 #define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
195 #define IP_UPD_PO_STATS_BH(net, field, val) SNMP_UPD_PO_STATS64_BH((net)->mib.ip_statistics, field, val)
196 #define NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
197 #define NET_INC_STATS_BH(net, field)	SNMP_INC_STATS_BH((net)->mib.net_statistics, field)
198 #define NET_INC_STATS_USER(net, field) 	SNMP_INC_STATS_USER((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_BH(net, field, adnd) SNMP_ADD_STATS_BH((net)->mib.net_statistics, field, adnd)
201 #define NET_ADD_STATS_USER(net, field, adnd) SNMP_ADD_STATS_USER((net)->mib.net_statistics, field, adnd)
202 
203 unsigned long snmp_fold_field(void __percpu *mib, int offt);
204 #if BITS_PER_LONG==32
205 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
206 #else
207 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
208 {
209 	return snmp_fold_field(mib, offt);
210 }
211 #endif
212 
213 void inet_get_local_port_range(struct net *net, int *low, int *high);
214 
215 #ifdef CONFIG_SYSCTL
216 static inline int inet_is_local_reserved_port(struct net *net, int port)
217 {
218 	if (!net->ipv4.sysctl_local_reserved_ports)
219 		return 0;
220 	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
221 }
222 
223 static inline bool sysctl_dev_name_is_allowed(const char *name)
224 {
225 	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
226 }
227 
228 #else
229 static inline int inet_is_local_reserved_port(struct net *net, int port)
230 {
231 	return 0;
232 }
233 #endif
234 
235 /* From inetpeer.c */
236 extern int inet_peer_threshold;
237 extern int inet_peer_minttl;
238 extern int inet_peer_maxttl;
239 
240 /* From ip_input.c */
241 extern int sysctl_ip_early_demux;
242 
243 /* From ip_output.c */
244 extern int sysctl_ip_dynaddr;
245 
246 void ipfrag_init(void);
247 
248 void ip_static_sysctl_init(void);
249 
250 #define IP4_REPLY_MARK(net, mark) \
251 	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
252 
253 static inline bool ip_is_fragment(const struct iphdr *iph)
254 {
255 	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
256 }
257 
258 #ifdef CONFIG_INET
259 #include <net/dst.h>
260 
261 /* The function in 2.2 was invalid, producing wrong result for
262  * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
263 static inline
264 int ip_decrease_ttl(struct iphdr *iph)
265 {
266 	u32 check = (__force u32)iph->check;
267 	check += (__force u32)htons(0x0100);
268 	iph->check = (__force __sum16)(check + (check>=0xFFFF));
269 	return --iph->ttl;
270 }
271 
272 static inline
273 int ip_dont_fragment(struct sock *sk, struct dst_entry *dst)
274 {
275 	return  inet_sk(sk)->pmtudisc == IP_PMTUDISC_DO ||
276 		(inet_sk(sk)->pmtudisc == IP_PMTUDISC_WANT &&
277 		 !(dst_metric_locked(dst, RTAX_MTU)));
278 }
279 
280 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
281 {
282 	return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
283 	       inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
284 }
285 
286 static inline bool ip_sk_use_pmtu(const struct sock *sk)
287 {
288 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
289 }
290 
291 static inline bool ip_sk_ignore_df(const struct sock *sk)
292 {
293 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
294 	       inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
295 }
296 
297 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
298 						    bool forwarding)
299 {
300 	struct net *net = dev_net(dst->dev);
301 
302 	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
303 	    dst_metric_locked(dst, RTAX_MTU) ||
304 	    !forwarding)
305 		return dst_mtu(dst);
306 
307 	return min(dst->dev->mtu, IP_MAX_MTU);
308 }
309 
310 static inline unsigned int ip_skb_dst_mtu(const struct sk_buff *skb)
311 {
312 	if (!skb->sk || ip_sk_use_pmtu(skb->sk)) {
313 		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
314 		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
315 	} else {
316 		return min(skb_dst(skb)->dev->mtu, IP_MAX_MTU);
317 	}
318 }
319 
320 u32 ip_idents_reserve(u32 hash, int segs);
321 void __ip_select_ident(struct iphdr *iph, int segs);
322 
323 static inline void ip_select_ident_segs(struct sk_buff *skb, struct sock *sk, int segs)
324 {
325 	struct iphdr *iph = ip_hdr(skb);
326 
327 	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
328 		/* This is only to work around buggy Windows95/2000
329 		 * VJ compression implementations.  If the ID field
330 		 * does not change, they drop every other packet in
331 		 * a TCP stream using header compression.
332 		 */
333 		if (sk && inet_sk(sk)->inet_daddr) {
334 			iph->id = htons(inet_sk(sk)->inet_id);
335 			inet_sk(sk)->inet_id += segs;
336 		} else {
337 			iph->id = 0;
338 		}
339 	} else {
340 		__ip_select_ident(iph, segs);
341 	}
342 }
343 
344 static inline void ip_select_ident(struct sk_buff *skb, struct sock *sk)
345 {
346 	ip_select_ident_segs(skb, sk, 1);
347 }
348 
349 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
350 {
351 	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
352 				  skb->len, proto, 0);
353 }
354 
355 static inline void inet_set_txhash(struct sock *sk)
356 {
357 	struct inet_sock *inet = inet_sk(sk);
358 	struct flow_keys keys;
359 
360 	keys.src = inet->inet_saddr;
361 	keys.dst = inet->inet_daddr;
362 	keys.port16[0] = inet->inet_sport;
363 	keys.port16[1] = inet->inet_dport;
364 
365 	sk->sk_txhash = flow_hash_from_keys(&keys);
366 }
367 
368 static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
369 {
370 	const struct iphdr *iph = skb_gro_network_header(skb);
371 
372 	return csum_tcpudp_nofold(iph->saddr, iph->daddr,
373 				  skb_gro_len(skb), proto, 0);
374 }
375 
376 /*
377  *	Map a multicast IP onto multicast MAC for type ethernet.
378  */
379 
380 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
381 {
382 	__u32 addr=ntohl(naddr);
383 	buf[0]=0x01;
384 	buf[1]=0x00;
385 	buf[2]=0x5e;
386 	buf[5]=addr&0xFF;
387 	addr>>=8;
388 	buf[4]=addr&0xFF;
389 	addr>>=8;
390 	buf[3]=addr&0x7F;
391 }
392 
393 /*
394  *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
395  *	Leave P_Key as 0 to be filled in by driver.
396  */
397 
398 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
399 {
400 	__u32 addr;
401 	unsigned char scope = broadcast[5] & 0xF;
402 
403 	buf[0]  = 0;		/* Reserved */
404 	buf[1]  = 0xff;		/* Multicast QPN */
405 	buf[2]  = 0xff;
406 	buf[3]  = 0xff;
407 	addr    = ntohl(naddr);
408 	buf[4]  = 0xff;
409 	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
410 	buf[6]  = 0x40;		/* IPv4 signature */
411 	buf[7]  = 0x1b;
412 	buf[8]  = broadcast[8];		/* P_Key */
413 	buf[9]  = broadcast[9];
414 	buf[10] = 0;
415 	buf[11] = 0;
416 	buf[12] = 0;
417 	buf[13] = 0;
418 	buf[14] = 0;
419 	buf[15] = 0;
420 	buf[19] = addr & 0xff;
421 	addr  >>= 8;
422 	buf[18] = addr & 0xff;
423 	addr  >>= 8;
424 	buf[17] = addr & 0xff;
425 	addr  >>= 8;
426 	buf[16] = addr & 0x0f;
427 }
428 
429 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
430 {
431 	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
432 		memcpy(buf, broadcast, 4);
433 	else
434 		memcpy(buf, &naddr, sizeof(naddr));
435 }
436 
437 #if IS_ENABLED(CONFIG_IPV6)
438 #include <linux/ipv6.h>
439 #endif
440 
441 static __inline__ void inet_reset_saddr(struct sock *sk)
442 {
443 	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
444 #if IS_ENABLED(CONFIG_IPV6)
445 	if (sk->sk_family == PF_INET6) {
446 		struct ipv6_pinfo *np = inet6_sk(sk);
447 
448 		memset(&np->saddr, 0, sizeof(np->saddr));
449 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
450 	}
451 #endif
452 }
453 
454 #endif
455 
456 static inline int sk_mc_loop(struct sock *sk)
457 {
458 	if (!sk)
459 		return 1;
460 	switch (sk->sk_family) {
461 	case AF_INET:
462 		return inet_sk(sk)->mc_loop;
463 #if IS_ENABLED(CONFIG_IPV6)
464 	case AF_INET6:
465 		return inet6_sk(sk)->mc_loop;
466 #endif
467 	}
468 	WARN_ON(1);
469 	return 1;
470 }
471 
472 bool ip_call_ra_chain(struct sk_buff *skb);
473 
474 /*
475  *	Functions provided by ip_fragment.c
476  */
477 
478 enum ip_defrag_users {
479 	IP_DEFRAG_LOCAL_DELIVER,
480 	IP_DEFRAG_CALL_RA_CHAIN,
481 	IP_DEFRAG_CONNTRACK_IN,
482 	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
483 	IP_DEFRAG_CONNTRACK_OUT,
484 	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
485 	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
486 	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
487 	IP_DEFRAG_VS_IN,
488 	IP_DEFRAG_VS_OUT,
489 	IP_DEFRAG_VS_FWD,
490 	IP_DEFRAG_AF_PACKET,
491 	IP_DEFRAG_MACVLAN,
492 };
493 
494 int ip_defrag(struct sk_buff *skb, u32 user);
495 #ifdef CONFIG_INET
496 struct sk_buff *ip_check_defrag(struct sk_buff *skb, u32 user);
497 #else
498 static inline struct sk_buff *ip_check_defrag(struct sk_buff *skb, u32 user)
499 {
500 	return skb;
501 }
502 #endif
503 int ip_frag_mem(struct net *net);
504 
505 /*
506  *	Functions provided by ip_forward.c
507  */
508 
509 int ip_forward(struct sk_buff *skb);
510 
511 /*
512  *	Functions provided by ip_options.c
513  */
514 
515 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
516 		      __be32 daddr, struct rtable *rt, int is_frag);
517 
518 int __ip_options_echo(struct ip_options *dopt, struct sk_buff *skb,
519 		      const struct ip_options *sopt);
520 static inline int ip_options_echo(struct ip_options *dopt, struct sk_buff *skb)
521 {
522 	return __ip_options_echo(dopt, skb, &IPCB(skb)->opt);
523 }
524 
525 void ip_options_fragment(struct sk_buff *skb);
526 int ip_options_compile(struct net *net, struct ip_options *opt,
527 		       struct sk_buff *skb);
528 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
529 		   unsigned char *data, int optlen);
530 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
531 			     unsigned char __user *data, int optlen);
532 void ip_options_undo(struct ip_options *opt);
533 void ip_forward_options(struct sk_buff *skb);
534 int ip_options_rcv_srr(struct sk_buff *skb);
535 
536 /*
537  *	Functions provided by ip_sockglue.c
538  */
539 
540 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
541 void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb);
542 int ip_cmsg_send(struct net *net, struct msghdr *msg,
543 		 struct ipcm_cookie *ipc, bool allow_ipv6);
544 int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
545 		  unsigned int optlen);
546 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
547 		  int __user *optlen);
548 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
549 			 char __user *optval, unsigned int optlen);
550 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
551 			 char __user *optval, int __user *optlen);
552 int ip_ra_control(struct sock *sk, unsigned char on,
553 		  void (*destructor)(struct sock *));
554 
555 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
556 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
557 		   u32 info, u8 *payload);
558 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
559 		    u32 info);
560 
561 bool icmp_global_allow(void);
562 extern int sysctl_icmp_msgs_per_sec;
563 extern int sysctl_icmp_msgs_burst;
564 
565 #ifdef CONFIG_PROC_FS
566 int ip_misc_proc_init(void);
567 #endif
568 
569 #endif	/* _IP_H */
570