xref: /linux/include/net/ip.h (revision 0ad53fe3ae82443c74ff8cfd7bd13377cc1134a3)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Definitions for the IP module.
8  *
9  * Version:	@(#)ip.h	1.0.2	05/07/93
10  *
11  * Authors:	Ross Biro
12  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13  *		Alan Cox, <gw4pts@gw4pts.ampr.org>
14  *
15  * Changes:
16  *		Mike McLagan    :       Routing by source
17  */
18 #ifndef _IP_H
19 #define _IP_H
20 
21 #include <linux/types.h>
22 #include <linux/ip.h>
23 #include <linux/in.h>
24 #include <linux/skbuff.h>
25 #include <linux/jhash.h>
26 #include <linux/sockptr.h>
27 
28 #include <net/inet_sock.h>
29 #include <net/route.h>
30 #include <net/snmp.h>
31 #include <net/flow.h>
32 #include <net/flow_dissector.h>
33 #include <net/netns/hash.h>
34 #include <net/lwtunnel.h>
35 
36 #define IPV4_MAX_PMTU		65535U		/* RFC 2675, Section 5.1 */
37 #define IPV4_MIN_MTU		68			/* RFC 791 */
38 
39 extern unsigned int sysctl_fib_sync_mem;
40 extern unsigned int sysctl_fib_sync_mem_min;
41 extern unsigned int sysctl_fib_sync_mem_max;
42 
43 struct sock;
44 
45 struct inet_skb_parm {
46 	int			iif;
47 	struct ip_options	opt;		/* Compiled IP options		*/
48 	u16			flags;
49 
50 #define IPSKB_FORWARDED		BIT(0)
51 #define IPSKB_XFRM_TUNNEL_SIZE	BIT(1)
52 #define IPSKB_XFRM_TRANSFORMED	BIT(2)
53 #define IPSKB_FRAG_COMPLETE	BIT(3)
54 #define IPSKB_REROUTED		BIT(4)
55 #define IPSKB_DOREDIRECT	BIT(5)
56 #define IPSKB_FRAG_PMTU		BIT(6)
57 #define IPSKB_L3SLAVE		BIT(7)
58 
59 	u16			frag_max_size;
60 };
61 
62 static inline bool ipv4_l3mdev_skb(u16 flags)
63 {
64 	return !!(flags & IPSKB_L3SLAVE);
65 }
66 
67 static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
68 {
69 	return ip_hdr(skb)->ihl * 4;
70 }
71 
72 struct ipcm_cookie {
73 	struct sockcm_cookie	sockc;
74 	__be32			addr;
75 	int			oif;
76 	struct ip_options_rcu	*opt;
77 	__u8			ttl;
78 	__s16			tos;
79 	char			priority;
80 	__u16			gso_size;
81 };
82 
83 static inline void ipcm_init(struct ipcm_cookie *ipcm)
84 {
85 	*ipcm = (struct ipcm_cookie) { .tos = -1 };
86 }
87 
88 static inline void ipcm_init_sk(struct ipcm_cookie *ipcm,
89 				const struct inet_sock *inet)
90 {
91 	ipcm_init(ipcm);
92 
93 	ipcm->sockc.mark = inet->sk.sk_mark;
94 	ipcm->sockc.tsflags = inet->sk.sk_tsflags;
95 	ipcm->oif = inet->sk.sk_bound_dev_if;
96 	ipcm->addr = inet->inet_saddr;
97 }
98 
99 #define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
100 #define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
101 
102 /* return enslaved device index if relevant */
103 static inline int inet_sdif(const struct sk_buff *skb)
104 {
105 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
106 	if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
107 		return IPCB(skb)->iif;
108 #endif
109 	return 0;
110 }
111 
112 /* Special input handler for packets caught by router alert option.
113    They are selected only by protocol field, and then processed likely
114    local ones; but only if someone wants them! Otherwise, router
115    not running rsvpd will kill RSVP.
116 
117    It is user level problem, what it will make with them.
118    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
119    but receiver should be enough clever f.e. to forward mtrace requests,
120    sent to multicast group to reach destination designated router.
121  */
122 
123 struct ip_ra_chain {
124 	struct ip_ra_chain __rcu *next;
125 	struct sock		*sk;
126 	union {
127 		void			(*destructor)(struct sock *);
128 		struct sock		*saved_sk;
129 	};
130 	struct rcu_head		rcu;
131 };
132 
133 /* IP flags. */
134 #define IP_CE		0x8000		/* Flag: "Congestion"		*/
135 #define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
136 #define IP_MF		0x2000		/* Flag: "More Fragments"	*/
137 #define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
138 
139 #define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
140 
141 struct msghdr;
142 struct net_device;
143 struct packet_type;
144 struct rtable;
145 struct sockaddr;
146 
147 int igmp_mc_init(void);
148 
149 /*
150  *	Functions provided by ip.c
151  */
152 
153 int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
154 			  __be32 saddr, __be32 daddr,
155 			  struct ip_options_rcu *opt, u8 tos);
156 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
157 	   struct net_device *orig_dev);
158 void ip_list_rcv(struct list_head *head, struct packet_type *pt,
159 		 struct net_device *orig_dev);
160 int ip_local_deliver(struct sk_buff *skb);
161 void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int proto);
162 int ip_mr_input(struct sk_buff *skb);
163 int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
164 int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
165 int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
166 		   int (*output)(struct net *, struct sock *, struct sk_buff *));
167 
168 struct ip_fraglist_iter {
169 	struct sk_buff	*frag;
170 	struct iphdr	*iph;
171 	int		offset;
172 	unsigned int	hlen;
173 };
174 
175 void ip_fraglist_init(struct sk_buff *skb, struct iphdr *iph,
176 		      unsigned int hlen, struct ip_fraglist_iter *iter);
177 void ip_fraglist_prepare(struct sk_buff *skb, struct ip_fraglist_iter *iter);
178 
179 static inline struct sk_buff *ip_fraglist_next(struct ip_fraglist_iter *iter)
180 {
181 	struct sk_buff *skb = iter->frag;
182 
183 	iter->frag = skb->next;
184 	skb_mark_not_on_list(skb);
185 
186 	return skb;
187 }
188 
189 struct ip_frag_state {
190 	bool		DF;
191 	unsigned int	hlen;
192 	unsigned int	ll_rs;
193 	unsigned int	mtu;
194 	unsigned int	left;
195 	int		offset;
196 	int		ptr;
197 	__be16		not_last_frag;
198 };
199 
200 void ip_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int ll_rs,
201 		  unsigned int mtu, bool DF, struct ip_frag_state *state);
202 struct sk_buff *ip_frag_next(struct sk_buff *skb,
203 			     struct ip_frag_state *state);
204 
205 void ip_send_check(struct iphdr *ip);
206 int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
207 int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
208 
209 int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
210 		    __u8 tos);
211 void ip_init(void);
212 int ip_append_data(struct sock *sk, struct flowi4 *fl4,
213 		   int getfrag(void *from, char *to, int offset, int len,
214 			       int odd, struct sk_buff *skb),
215 		   void *from, int len, int protolen,
216 		   struct ipcm_cookie *ipc,
217 		   struct rtable **rt,
218 		   unsigned int flags);
219 int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
220 		       struct sk_buff *skb);
221 ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
222 		       int offset, size_t size, int flags);
223 struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
224 			      struct sk_buff_head *queue,
225 			      struct inet_cork *cork);
226 int ip_send_skb(struct net *net, struct sk_buff *skb);
227 int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
228 void ip_flush_pending_frames(struct sock *sk);
229 struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
230 			    int getfrag(void *from, char *to, int offset,
231 					int len, int odd, struct sk_buff *skb),
232 			    void *from, int length, int transhdrlen,
233 			    struct ipcm_cookie *ipc, struct rtable **rtp,
234 			    struct inet_cork *cork, unsigned int flags);
235 
236 int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
237 
238 static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
239 {
240 	return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
241 }
242 
243 static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
244 {
245 	return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
246 }
247 
248 static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
249 {
250 	return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
251 }
252 
253 /* datagram.c */
254 int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
255 int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
256 
257 void ip4_datagram_release_cb(struct sock *sk);
258 
259 struct ip_reply_arg {
260 	struct kvec iov[1];
261 	int	    flags;
262 	__wsum 	    csum;
263 	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
264 				/* -1 if not needed */
265 	int	    bound_dev_if;
266 	u8  	    tos;
267 	kuid_t	    uid;
268 };
269 
270 #define IP_REPLY_ARG_NOSRCCHECK 1
271 
272 static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
273 {
274 	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
275 }
276 
277 void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
278 			   const struct ip_options *sopt,
279 			   __be32 daddr, __be32 saddr,
280 			   const struct ip_reply_arg *arg,
281 			   unsigned int len, u64 transmit_time);
282 
283 #define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
284 #define __IP_INC_STATS(net, field)	__SNMP_INC_STATS64((net)->mib.ip_statistics, field)
285 #define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
286 #define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
287 #define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
288 #define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
289 #define NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
290 #define __NET_INC_STATS(net, field)	__SNMP_INC_STATS((net)->mib.net_statistics, field)
291 #define NET_ADD_STATS(net, field, adnd)	SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
292 #define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
293 
294 u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
295 unsigned long snmp_fold_field(void __percpu *mib, int offt);
296 #if BITS_PER_LONG==32
297 u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
298 			 size_t syncp_offset);
299 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
300 #else
301 static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
302 					size_t syncp_offset)
303 {
304 	return snmp_get_cpu_field(mib, cpu, offct);
305 
306 }
307 
308 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
309 {
310 	return snmp_fold_field(mib, offt);
311 }
312 #endif
313 
314 #define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
315 { \
316 	int i, c; \
317 	for_each_possible_cpu(c) { \
318 		for (i = 0; stats_list[i].name; i++) \
319 			buff64[i] += snmp_get_cpu_field64( \
320 					mib_statistic, \
321 					c, stats_list[i].entry, \
322 					offset); \
323 	} \
324 }
325 
326 #define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
327 { \
328 	int i, c; \
329 	for_each_possible_cpu(c) { \
330 		for (i = 0; stats_list[i].name; i++) \
331 			buff[i] += snmp_get_cpu_field( \
332 						mib_statistic, \
333 						c, stats_list[i].entry); \
334 	} \
335 }
336 
337 void inet_get_local_port_range(struct net *net, int *low, int *high);
338 
339 #ifdef CONFIG_SYSCTL
340 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
341 {
342 	if (!net->ipv4.sysctl_local_reserved_ports)
343 		return false;
344 	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
345 }
346 
347 static inline bool sysctl_dev_name_is_allowed(const char *name)
348 {
349 	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
350 }
351 
352 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
353 {
354 	return port < net->ipv4.sysctl_ip_prot_sock;
355 }
356 
357 #else
358 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
359 {
360 	return false;
361 }
362 
363 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
364 {
365 	return port < PROT_SOCK;
366 }
367 #endif
368 
369 __be32 inet_current_timestamp(void);
370 
371 /* From inetpeer.c */
372 extern int inet_peer_threshold;
373 extern int inet_peer_minttl;
374 extern int inet_peer_maxttl;
375 
376 void ipfrag_init(void);
377 
378 void ip_static_sysctl_init(void);
379 
380 #define IP4_REPLY_MARK(net, mark) \
381 	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
382 
383 static inline bool ip_is_fragment(const struct iphdr *iph)
384 {
385 	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
386 }
387 
388 #ifdef CONFIG_INET
389 #include <net/dst.h>
390 
391 /* The function in 2.2 was invalid, producing wrong result for
392  * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
393 static inline
394 int ip_decrease_ttl(struct iphdr *iph)
395 {
396 	u32 check = (__force u32)iph->check;
397 	check += (__force u32)htons(0x0100);
398 	iph->check = (__force __sum16)(check + (check>=0xFFFF));
399 	return --iph->ttl;
400 }
401 
402 static inline int ip_mtu_locked(const struct dst_entry *dst)
403 {
404 	const struct rtable *rt = (const struct rtable *)dst;
405 
406 	return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
407 }
408 
409 static inline
410 int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
411 {
412 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
413 
414 	return  pmtudisc == IP_PMTUDISC_DO ||
415 		(pmtudisc == IP_PMTUDISC_WANT &&
416 		 !ip_mtu_locked(dst));
417 }
418 
419 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
420 {
421 	return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
422 	       inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
423 }
424 
425 static inline bool ip_sk_use_pmtu(const struct sock *sk)
426 {
427 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
428 }
429 
430 static inline bool ip_sk_ignore_df(const struct sock *sk)
431 {
432 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
433 	       inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
434 }
435 
436 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
437 						    bool forwarding)
438 {
439 	const struct rtable *rt = container_of(dst, struct rtable, dst);
440 	struct net *net = dev_net(dst->dev);
441 	unsigned int mtu;
442 
443 	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
444 	    ip_mtu_locked(dst) ||
445 	    !forwarding) {
446 		mtu = rt->rt_pmtu;
447 		if (mtu && time_before(jiffies, rt->dst.expires))
448 			goto out;
449 	}
450 
451 	/* 'forwarding = true' case should always honour route mtu */
452 	mtu = dst_metric_raw(dst, RTAX_MTU);
453 	if (mtu)
454 		goto out;
455 
456 	mtu = READ_ONCE(dst->dev->mtu);
457 
458 	if (unlikely(ip_mtu_locked(dst))) {
459 		if (rt->rt_uses_gateway && mtu > 576)
460 			mtu = 576;
461 	}
462 
463 out:
464 	mtu = min_t(unsigned int, mtu, IP_MAX_MTU);
465 
466 	return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
467 }
468 
469 static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
470 					  const struct sk_buff *skb)
471 {
472 	unsigned int mtu;
473 
474 	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
475 		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
476 
477 		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
478 	}
479 
480 	mtu = min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
481 	return mtu - lwtunnel_headroom(skb_dst(skb)->lwtstate, mtu);
482 }
483 
484 struct dst_metrics *ip_fib_metrics_init(struct net *net, struct nlattr *fc_mx,
485 					int fc_mx_len,
486 					struct netlink_ext_ack *extack);
487 static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
488 {
489 	if (fib_metrics != &dst_default_metrics &&
490 	    refcount_dec_and_test(&fib_metrics->refcnt))
491 		kfree(fib_metrics);
492 }
493 
494 /* ipv4 and ipv6 both use refcounted metrics if it is not the default */
495 static inline
496 void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
497 {
498 	dst_init_metrics(dst, fib_metrics->metrics, true);
499 
500 	if (fib_metrics != &dst_default_metrics) {
501 		dst->_metrics |= DST_METRICS_REFCOUNTED;
502 		refcount_inc(&fib_metrics->refcnt);
503 	}
504 }
505 
506 static inline
507 void ip_dst_metrics_put(struct dst_entry *dst)
508 {
509 	struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
510 
511 	if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
512 		kfree(p);
513 }
514 
515 u32 ip_idents_reserve(u32 hash, int segs);
516 void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
517 
518 static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
519 					struct sock *sk, int segs)
520 {
521 	struct iphdr *iph = ip_hdr(skb);
522 
523 	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
524 		/* This is only to work around buggy Windows95/2000
525 		 * VJ compression implementations.  If the ID field
526 		 * does not change, they drop every other packet in
527 		 * a TCP stream using header compression.
528 		 */
529 		if (sk && inet_sk(sk)->inet_daddr) {
530 			iph->id = htons(inet_sk(sk)->inet_id);
531 			inet_sk(sk)->inet_id += segs;
532 		} else {
533 			iph->id = 0;
534 		}
535 	} else {
536 		__ip_select_ident(net, iph, segs);
537 	}
538 }
539 
540 static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
541 				   struct sock *sk)
542 {
543 	ip_select_ident_segs(net, skb, sk, 1);
544 }
545 
546 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
547 {
548 	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
549 				  skb->len, proto, 0);
550 }
551 
552 /* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
553  * Equivalent to :	flow->v4addrs.src = iph->saddr;
554  *			flow->v4addrs.dst = iph->daddr;
555  */
556 static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
557 					    const struct iphdr *iph)
558 {
559 	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
560 		     offsetof(typeof(flow->addrs), v4addrs.src) +
561 			      sizeof(flow->addrs.v4addrs.src));
562 	memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
563 	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
564 }
565 
566 static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
567 {
568 	const struct iphdr *iph = skb_gro_network_header(skb);
569 
570 	return csum_tcpudp_nofold(iph->saddr, iph->daddr,
571 				  skb_gro_len(skb), proto, 0);
572 }
573 
574 /*
575  *	Map a multicast IP onto multicast MAC for type ethernet.
576  */
577 
578 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
579 {
580 	__u32 addr=ntohl(naddr);
581 	buf[0]=0x01;
582 	buf[1]=0x00;
583 	buf[2]=0x5e;
584 	buf[5]=addr&0xFF;
585 	addr>>=8;
586 	buf[4]=addr&0xFF;
587 	addr>>=8;
588 	buf[3]=addr&0x7F;
589 }
590 
591 /*
592  *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
593  *	Leave P_Key as 0 to be filled in by driver.
594  */
595 
596 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
597 {
598 	__u32 addr;
599 	unsigned char scope = broadcast[5] & 0xF;
600 
601 	buf[0]  = 0;		/* Reserved */
602 	buf[1]  = 0xff;		/* Multicast QPN */
603 	buf[2]  = 0xff;
604 	buf[3]  = 0xff;
605 	addr    = ntohl(naddr);
606 	buf[4]  = 0xff;
607 	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
608 	buf[6]  = 0x40;		/* IPv4 signature */
609 	buf[7]  = 0x1b;
610 	buf[8]  = broadcast[8];		/* P_Key */
611 	buf[9]  = broadcast[9];
612 	buf[10] = 0;
613 	buf[11] = 0;
614 	buf[12] = 0;
615 	buf[13] = 0;
616 	buf[14] = 0;
617 	buf[15] = 0;
618 	buf[19] = addr & 0xff;
619 	addr  >>= 8;
620 	buf[18] = addr & 0xff;
621 	addr  >>= 8;
622 	buf[17] = addr & 0xff;
623 	addr  >>= 8;
624 	buf[16] = addr & 0x0f;
625 }
626 
627 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
628 {
629 	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
630 		memcpy(buf, broadcast, 4);
631 	else
632 		memcpy(buf, &naddr, sizeof(naddr));
633 }
634 
635 #if IS_ENABLED(CONFIG_IPV6)
636 #include <linux/ipv6.h>
637 #endif
638 
639 static __inline__ void inet_reset_saddr(struct sock *sk)
640 {
641 	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
642 #if IS_ENABLED(CONFIG_IPV6)
643 	if (sk->sk_family == PF_INET6) {
644 		struct ipv6_pinfo *np = inet6_sk(sk);
645 
646 		memset(&np->saddr, 0, sizeof(np->saddr));
647 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
648 	}
649 #endif
650 }
651 
652 #endif
653 
654 static inline unsigned int ipv4_addr_hash(__be32 ip)
655 {
656 	return (__force unsigned int) ip;
657 }
658 
659 static inline u32 ipv4_portaddr_hash(const struct net *net,
660 				     __be32 saddr,
661 				     unsigned int port)
662 {
663 	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
664 }
665 
666 bool ip_call_ra_chain(struct sk_buff *skb);
667 
668 /*
669  *	Functions provided by ip_fragment.c
670  */
671 
672 enum ip_defrag_users {
673 	IP_DEFRAG_LOCAL_DELIVER,
674 	IP_DEFRAG_CALL_RA_CHAIN,
675 	IP_DEFRAG_CONNTRACK_IN,
676 	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
677 	IP_DEFRAG_CONNTRACK_OUT,
678 	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
679 	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
680 	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
681 	IP_DEFRAG_VS_IN,
682 	IP_DEFRAG_VS_OUT,
683 	IP_DEFRAG_VS_FWD,
684 	IP_DEFRAG_AF_PACKET,
685 	IP_DEFRAG_MACVLAN,
686 };
687 
688 /* Return true if the value of 'user' is between 'lower_bond'
689  * and 'upper_bond' inclusively.
690  */
691 static inline bool ip_defrag_user_in_between(u32 user,
692 					     enum ip_defrag_users lower_bond,
693 					     enum ip_defrag_users upper_bond)
694 {
695 	return user >= lower_bond && user <= upper_bond;
696 }
697 
698 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
699 #ifdef CONFIG_INET
700 struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
701 #else
702 static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
703 {
704 	return skb;
705 }
706 #endif
707 
708 /*
709  *	Functions provided by ip_forward.c
710  */
711 
712 int ip_forward(struct sk_buff *skb);
713 
714 /*
715  *	Functions provided by ip_options.c
716  */
717 
718 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
719 		      __be32 daddr, struct rtable *rt, int is_frag);
720 
721 int __ip_options_echo(struct net *net, struct ip_options *dopt,
722 		      struct sk_buff *skb, const struct ip_options *sopt);
723 static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
724 				  struct sk_buff *skb)
725 {
726 	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
727 }
728 
729 void ip_options_fragment(struct sk_buff *skb);
730 int __ip_options_compile(struct net *net, struct ip_options *opt,
731 			 struct sk_buff *skb, __be32 *info);
732 int ip_options_compile(struct net *net, struct ip_options *opt,
733 		       struct sk_buff *skb);
734 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
735 		   sockptr_t data, int optlen);
736 void ip_options_undo(struct ip_options *opt);
737 void ip_forward_options(struct sk_buff *skb);
738 int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
739 
740 /*
741  *	Functions provided by ip_sockglue.c
742  */
743 
744 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
745 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
746 			 struct sk_buff *skb, int tlen, int offset);
747 int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
748 		 struct ipcm_cookie *ipc, bool allow_ipv6);
749 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
750 		  unsigned int optlen);
751 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
752 		  int __user *optlen);
753 int ip_ra_control(struct sock *sk, unsigned char on,
754 		  void (*destructor)(struct sock *));
755 
756 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
757 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
758 		   u32 info, u8 *payload);
759 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
760 		    u32 info);
761 
762 static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
763 {
764 	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
765 }
766 
767 bool icmp_global_allow(void);
768 extern int sysctl_icmp_msgs_per_sec;
769 extern int sysctl_icmp_msgs_burst;
770 
771 #ifdef CONFIG_PROC_FS
772 int ip_misc_proc_init(void);
773 #endif
774 
775 int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
776 				struct netlink_ext_ack *extack);
777 
778 static inline bool inetdev_valid_mtu(unsigned int mtu)
779 {
780 	return likely(mtu >= IPV4_MIN_MTU);
781 }
782 
783 void ip_sock_set_freebind(struct sock *sk);
784 int ip_sock_set_mtu_discover(struct sock *sk, int val);
785 void ip_sock_set_pktinfo(struct sock *sk);
786 void ip_sock_set_recverr(struct sock *sk);
787 void ip_sock_set_tos(struct sock *sk, int val);
788 
789 #endif	/* _IP_H */
790