xref: /linux/include/net/ip.h (revision f3956ebb3bf06ab2266ad5ee2214aed46405810c)
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 static inline u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
295 {
296 	return  *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
297 }
298 
299 unsigned long snmp_fold_field(void __percpu *mib, int offt);
300 #if BITS_PER_LONG==32
301 u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
302 			 size_t syncp_offset);
303 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
304 #else
305 static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
306 					size_t syncp_offset)
307 {
308 	return snmp_get_cpu_field(mib, cpu, offct);
309 
310 }
311 
312 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
313 {
314 	return snmp_fold_field(mib, offt);
315 }
316 #endif
317 
318 #define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
319 { \
320 	int i, c; \
321 	for_each_possible_cpu(c) { \
322 		for (i = 0; stats_list[i].name; i++) \
323 			buff64[i] += snmp_get_cpu_field64( \
324 					mib_statistic, \
325 					c, stats_list[i].entry, \
326 					offset); \
327 	} \
328 }
329 
330 #define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
331 { \
332 	int i, c; \
333 	for_each_possible_cpu(c) { \
334 		for (i = 0; stats_list[i].name; i++) \
335 			buff[i] += snmp_get_cpu_field( \
336 						mib_statistic, \
337 						c, stats_list[i].entry); \
338 	} \
339 }
340 
341 void inet_get_local_port_range(struct net *net, int *low, int *high);
342 
343 #ifdef CONFIG_SYSCTL
344 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
345 {
346 	if (!net->ipv4.sysctl_local_reserved_ports)
347 		return false;
348 	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
349 }
350 
351 static inline bool sysctl_dev_name_is_allowed(const char *name)
352 {
353 	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
354 }
355 
356 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
357 {
358 	return port < net->ipv4.sysctl_ip_prot_sock;
359 }
360 
361 #else
362 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
363 {
364 	return false;
365 }
366 
367 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
368 {
369 	return port < PROT_SOCK;
370 }
371 #endif
372 
373 __be32 inet_current_timestamp(void);
374 
375 /* From inetpeer.c */
376 extern int inet_peer_threshold;
377 extern int inet_peer_minttl;
378 extern int inet_peer_maxttl;
379 
380 void ipfrag_init(void);
381 
382 void ip_static_sysctl_init(void);
383 
384 #define IP4_REPLY_MARK(net, mark) \
385 	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
386 
387 static inline bool ip_is_fragment(const struct iphdr *iph)
388 {
389 	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
390 }
391 
392 #ifdef CONFIG_INET
393 #include <net/dst.h>
394 
395 /* The function in 2.2 was invalid, producing wrong result for
396  * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
397 static inline
398 int ip_decrease_ttl(struct iphdr *iph)
399 {
400 	u32 check = (__force u32)iph->check;
401 	check += (__force u32)htons(0x0100);
402 	iph->check = (__force __sum16)(check + (check>=0xFFFF));
403 	return --iph->ttl;
404 }
405 
406 static inline int ip_mtu_locked(const struct dst_entry *dst)
407 {
408 	const struct rtable *rt = (const struct rtable *)dst;
409 
410 	return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
411 }
412 
413 static inline
414 int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
415 {
416 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
417 
418 	return  pmtudisc == IP_PMTUDISC_DO ||
419 		(pmtudisc == IP_PMTUDISC_WANT &&
420 		 !ip_mtu_locked(dst));
421 }
422 
423 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
424 {
425 	return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
426 	       inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
427 }
428 
429 static inline bool ip_sk_use_pmtu(const struct sock *sk)
430 {
431 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
432 }
433 
434 static inline bool ip_sk_ignore_df(const struct sock *sk)
435 {
436 	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
437 	       inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
438 }
439 
440 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
441 						    bool forwarding)
442 {
443 	const struct rtable *rt = container_of(dst, struct rtable, dst);
444 	struct net *net = dev_net(dst->dev);
445 	unsigned int mtu;
446 
447 	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
448 	    ip_mtu_locked(dst) ||
449 	    !forwarding) {
450 		mtu = rt->rt_pmtu;
451 		if (mtu && time_before(jiffies, rt->dst.expires))
452 			goto out;
453 	}
454 
455 	/* 'forwarding = true' case should always honour route mtu */
456 	mtu = dst_metric_raw(dst, RTAX_MTU);
457 	if (mtu)
458 		goto out;
459 
460 	mtu = READ_ONCE(dst->dev->mtu);
461 
462 	if (unlikely(ip_mtu_locked(dst))) {
463 		if (rt->rt_uses_gateway && mtu > 576)
464 			mtu = 576;
465 	}
466 
467 out:
468 	mtu = min_t(unsigned int, mtu, IP_MAX_MTU);
469 
470 	return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
471 }
472 
473 static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
474 					  const struct sk_buff *skb)
475 {
476 	unsigned int mtu;
477 
478 	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
479 		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
480 
481 		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
482 	}
483 
484 	mtu = min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
485 	return mtu - lwtunnel_headroom(skb_dst(skb)->lwtstate, mtu);
486 }
487 
488 struct dst_metrics *ip_fib_metrics_init(struct net *net, struct nlattr *fc_mx,
489 					int fc_mx_len,
490 					struct netlink_ext_ack *extack);
491 static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
492 {
493 	if (fib_metrics != &dst_default_metrics &&
494 	    refcount_dec_and_test(&fib_metrics->refcnt))
495 		kfree(fib_metrics);
496 }
497 
498 /* ipv4 and ipv6 both use refcounted metrics if it is not the default */
499 static inline
500 void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
501 {
502 	dst_init_metrics(dst, fib_metrics->metrics, true);
503 
504 	if (fib_metrics != &dst_default_metrics) {
505 		dst->_metrics |= DST_METRICS_REFCOUNTED;
506 		refcount_inc(&fib_metrics->refcnt);
507 	}
508 }
509 
510 static inline
511 void ip_dst_metrics_put(struct dst_entry *dst)
512 {
513 	struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
514 
515 	if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
516 		kfree(p);
517 }
518 
519 u32 ip_idents_reserve(u32 hash, int segs);
520 void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
521 
522 static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
523 					struct sock *sk, int segs)
524 {
525 	struct iphdr *iph = ip_hdr(skb);
526 
527 	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
528 		/* This is only to work around buggy Windows95/2000
529 		 * VJ compression implementations.  If the ID field
530 		 * does not change, they drop every other packet in
531 		 * a TCP stream using header compression.
532 		 */
533 		if (sk && inet_sk(sk)->inet_daddr) {
534 			iph->id = htons(inet_sk(sk)->inet_id);
535 			inet_sk(sk)->inet_id += segs;
536 		} else {
537 			iph->id = 0;
538 		}
539 	} else {
540 		__ip_select_ident(net, iph, segs);
541 	}
542 }
543 
544 static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
545 				   struct sock *sk)
546 {
547 	ip_select_ident_segs(net, skb, sk, 1);
548 }
549 
550 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
551 {
552 	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
553 				  skb->len, proto, 0);
554 }
555 
556 /* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
557  * Equivalent to :	flow->v4addrs.src = iph->saddr;
558  *			flow->v4addrs.dst = iph->daddr;
559  */
560 static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
561 					    const struct iphdr *iph)
562 {
563 	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
564 		     offsetof(typeof(flow->addrs), v4addrs.src) +
565 			      sizeof(flow->addrs.v4addrs.src));
566 	memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
567 	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
568 }
569 
570 static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
571 {
572 	const struct iphdr *iph = skb_gro_network_header(skb);
573 
574 	return csum_tcpudp_nofold(iph->saddr, iph->daddr,
575 				  skb_gro_len(skb), proto, 0);
576 }
577 
578 /*
579  *	Map a multicast IP onto multicast MAC for type ethernet.
580  */
581 
582 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
583 {
584 	__u32 addr=ntohl(naddr);
585 	buf[0]=0x01;
586 	buf[1]=0x00;
587 	buf[2]=0x5e;
588 	buf[5]=addr&0xFF;
589 	addr>>=8;
590 	buf[4]=addr&0xFF;
591 	addr>>=8;
592 	buf[3]=addr&0x7F;
593 }
594 
595 /*
596  *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
597  *	Leave P_Key as 0 to be filled in by driver.
598  */
599 
600 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
601 {
602 	__u32 addr;
603 	unsigned char scope = broadcast[5] & 0xF;
604 
605 	buf[0]  = 0;		/* Reserved */
606 	buf[1]  = 0xff;		/* Multicast QPN */
607 	buf[2]  = 0xff;
608 	buf[3]  = 0xff;
609 	addr    = ntohl(naddr);
610 	buf[4]  = 0xff;
611 	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
612 	buf[6]  = 0x40;		/* IPv4 signature */
613 	buf[7]  = 0x1b;
614 	buf[8]  = broadcast[8];		/* P_Key */
615 	buf[9]  = broadcast[9];
616 	buf[10] = 0;
617 	buf[11] = 0;
618 	buf[12] = 0;
619 	buf[13] = 0;
620 	buf[14] = 0;
621 	buf[15] = 0;
622 	buf[19] = addr & 0xff;
623 	addr  >>= 8;
624 	buf[18] = addr & 0xff;
625 	addr  >>= 8;
626 	buf[17] = addr & 0xff;
627 	addr  >>= 8;
628 	buf[16] = addr & 0x0f;
629 }
630 
631 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
632 {
633 	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
634 		memcpy(buf, broadcast, 4);
635 	else
636 		memcpy(buf, &naddr, sizeof(naddr));
637 }
638 
639 #if IS_ENABLED(CONFIG_IPV6)
640 #include <linux/ipv6.h>
641 #endif
642 
643 static __inline__ void inet_reset_saddr(struct sock *sk)
644 {
645 	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
646 #if IS_ENABLED(CONFIG_IPV6)
647 	if (sk->sk_family == PF_INET6) {
648 		struct ipv6_pinfo *np = inet6_sk(sk);
649 
650 		memset(&np->saddr, 0, sizeof(np->saddr));
651 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
652 	}
653 #endif
654 }
655 
656 #endif
657 
658 static inline unsigned int ipv4_addr_hash(__be32 ip)
659 {
660 	return (__force unsigned int) ip;
661 }
662 
663 static inline u32 ipv4_portaddr_hash(const struct net *net,
664 				     __be32 saddr,
665 				     unsigned int port)
666 {
667 	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
668 }
669 
670 bool ip_call_ra_chain(struct sk_buff *skb);
671 
672 /*
673  *	Functions provided by ip_fragment.c
674  */
675 
676 enum ip_defrag_users {
677 	IP_DEFRAG_LOCAL_DELIVER,
678 	IP_DEFRAG_CALL_RA_CHAIN,
679 	IP_DEFRAG_CONNTRACK_IN,
680 	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
681 	IP_DEFRAG_CONNTRACK_OUT,
682 	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
683 	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
684 	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
685 	IP_DEFRAG_VS_IN,
686 	IP_DEFRAG_VS_OUT,
687 	IP_DEFRAG_VS_FWD,
688 	IP_DEFRAG_AF_PACKET,
689 	IP_DEFRAG_MACVLAN,
690 };
691 
692 /* Return true if the value of 'user' is between 'lower_bond'
693  * and 'upper_bond' inclusively.
694  */
695 static inline bool ip_defrag_user_in_between(u32 user,
696 					     enum ip_defrag_users lower_bond,
697 					     enum ip_defrag_users upper_bond)
698 {
699 	return user >= lower_bond && user <= upper_bond;
700 }
701 
702 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
703 #ifdef CONFIG_INET
704 struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
705 #else
706 static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
707 {
708 	return skb;
709 }
710 #endif
711 
712 /*
713  *	Functions provided by ip_forward.c
714  */
715 
716 int ip_forward(struct sk_buff *skb);
717 
718 /*
719  *	Functions provided by ip_options.c
720  */
721 
722 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
723 		      __be32 daddr, struct rtable *rt, int is_frag);
724 
725 int __ip_options_echo(struct net *net, struct ip_options *dopt,
726 		      struct sk_buff *skb, const struct ip_options *sopt);
727 static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
728 				  struct sk_buff *skb)
729 {
730 	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
731 }
732 
733 void ip_options_fragment(struct sk_buff *skb);
734 int __ip_options_compile(struct net *net, struct ip_options *opt,
735 			 struct sk_buff *skb, __be32 *info);
736 int ip_options_compile(struct net *net, struct ip_options *opt,
737 		       struct sk_buff *skb);
738 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
739 		   sockptr_t data, int optlen);
740 void ip_options_undo(struct ip_options *opt);
741 void ip_forward_options(struct sk_buff *skb);
742 int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
743 
744 /*
745  *	Functions provided by ip_sockglue.c
746  */
747 
748 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
749 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
750 			 struct sk_buff *skb, int tlen, int offset);
751 int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
752 		 struct ipcm_cookie *ipc, bool allow_ipv6);
753 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
754 		  unsigned int optlen);
755 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
756 		  int __user *optlen);
757 int ip_ra_control(struct sock *sk, unsigned char on,
758 		  void (*destructor)(struct sock *));
759 
760 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
761 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
762 		   u32 info, u8 *payload);
763 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
764 		    u32 info);
765 
766 static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
767 {
768 	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
769 }
770 
771 bool icmp_global_allow(void);
772 extern int sysctl_icmp_msgs_per_sec;
773 extern int sysctl_icmp_msgs_burst;
774 
775 #ifdef CONFIG_PROC_FS
776 int ip_misc_proc_init(void);
777 #endif
778 
779 int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
780 				struct netlink_ext_ack *extack);
781 
782 static inline bool inetdev_valid_mtu(unsigned int mtu)
783 {
784 	return likely(mtu >= IPV4_MIN_MTU);
785 }
786 
787 void ip_sock_set_freebind(struct sock *sk);
788 int ip_sock_set_mtu_discover(struct sock *sk, int val);
789 void ip_sock_set_pktinfo(struct sock *sk);
790 void ip_sock_set_recverr(struct sock *sk);
791 void ip_sock_set_tos(struct sock *sk, int val);
792 
793 #endif	/* _IP_H */
794