xref: /linux/include/net/ip.h (revision a4ff64edf9edc8f05e2183610dc8306d3279c6ac)
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 #include <linux/static_key.h>
28 
29 #include <net/inet_sock.h>
30 #include <net/route.h>
31 #include <net/snmp.h>
32 #include <net/flow.h>
33 #include <net/flow_dissector.h>
34 #include <net/netns/hash.h>
35 #include <net/lwtunnel.h>
36 
37 #define IPV4_MAX_PMTU		65535U		/* RFC 2675, Section 5.1 */
38 #define IPV4_MIN_MTU		68			/* RFC 791 */
39 
40 extern unsigned int sysctl_fib_sync_mem;
41 extern unsigned int sysctl_fib_sync_mem_min;
42 extern unsigned int sysctl_fib_sync_mem_max;
43 
44 struct sock;
45 
46 struct inet_skb_parm {
47 	int			iif;
48 	struct ip_options	opt;		/* Compiled IP options		*/
49 	u16			flags;
50 
51 #define IPSKB_FORWARDED		BIT(0)
52 #define IPSKB_XFRM_TUNNEL_SIZE	BIT(1)
53 #define IPSKB_XFRM_TRANSFORMED	BIT(2)
54 #define IPSKB_FRAG_COMPLETE	BIT(3)
55 #define IPSKB_REROUTED		BIT(4)
56 #define IPSKB_DOREDIRECT	BIT(5)
57 #define IPSKB_FRAG_PMTU		BIT(6)
58 #define IPSKB_L3SLAVE		BIT(7)
59 #define IPSKB_NOPOLICY		BIT(8)
60 #define IPSKB_MULTIPATH		BIT(9)
61 
62 	u16			frag_max_size;
63 };
64 
65 static inline bool ipv4_l3mdev_skb(u16 flags)
66 {
67 	return !!(flags & IPSKB_L3SLAVE);
68 }
69 
70 static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
71 {
72 	return ip_hdr(skb)->ihl * 4;
73 }
74 
75 struct ipcm_cookie {
76 	struct sockcm_cookie	sockc;
77 	__be32			addr;
78 	int			oif;
79 	struct ip_options_rcu	*opt;
80 	__u8			protocol;
81 	__u8			ttl;
82 	__s16			tos;
83 	char			priority;
84 	__u16			gso_size;
85 };
86 
87 static inline void ipcm_init(struct ipcm_cookie *ipcm)
88 {
89 	*ipcm = (struct ipcm_cookie) { .tos = -1 };
90 }
91 
92 static inline void ipcm_init_sk(struct ipcm_cookie *ipcm,
93 				const struct inet_sock *inet)
94 {
95 	ipcm_init(ipcm);
96 
97 	ipcm->sockc.mark = READ_ONCE(inet->sk.sk_mark);
98 	ipcm->sockc.tsflags = READ_ONCE(inet->sk.sk_tsflags);
99 	ipcm->oif = READ_ONCE(inet->sk.sk_bound_dev_if);
100 	ipcm->addr = inet->inet_saddr;
101 	ipcm->protocol = inet->inet_num;
102 }
103 
104 #define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
105 #define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
106 
107 /* return enslaved device index if relevant */
108 static inline int inet_sdif(const struct sk_buff *skb)
109 {
110 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
111 	if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
112 		return IPCB(skb)->iif;
113 #endif
114 	return 0;
115 }
116 
117 /* Special input handler for packets caught by router alert option.
118    They are selected only by protocol field, and then processed likely
119    local ones; but only if someone wants them! Otherwise, router
120    not running rsvpd will kill RSVP.
121 
122    It is user level problem, what it will make with them.
123    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
124    but receiver should be enough clever f.e. to forward mtrace requests,
125    sent to multicast group to reach destination designated router.
126  */
127 
128 struct ip_ra_chain {
129 	struct ip_ra_chain __rcu *next;
130 	struct sock		*sk;
131 	union {
132 		void			(*destructor)(struct sock *);
133 		struct sock		*saved_sk;
134 	};
135 	struct rcu_head		rcu;
136 };
137 
138 /* IP flags. */
139 #define IP_CE		0x8000		/* Flag: "Congestion"		*/
140 #define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
141 #define IP_MF		0x2000		/* Flag: "More Fragments"	*/
142 #define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
143 
144 #define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
145 
146 struct msghdr;
147 struct net_device;
148 struct packet_type;
149 struct rtable;
150 struct sockaddr;
151 
152 int igmp_mc_init(void);
153 
154 /*
155  *	Functions provided by ip.c
156  */
157 
158 int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
159 			  __be32 saddr, __be32 daddr,
160 			  struct ip_options_rcu *opt, u8 tos);
161 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
162 	   struct net_device *orig_dev);
163 void ip_list_rcv(struct list_head *head, struct packet_type *pt,
164 		 struct net_device *orig_dev);
165 int ip_local_deliver(struct sk_buff *skb);
166 void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int proto);
167 int ip_mr_input(struct sk_buff *skb);
168 int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
169 int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
170 int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
171 		   int (*output)(struct net *, struct sock *, struct sk_buff *));
172 
173 struct ip_fraglist_iter {
174 	struct sk_buff	*frag;
175 	struct iphdr	*iph;
176 	int		offset;
177 	unsigned int	hlen;
178 };
179 
180 void ip_fraglist_init(struct sk_buff *skb, struct iphdr *iph,
181 		      unsigned int hlen, struct ip_fraglist_iter *iter);
182 void ip_fraglist_prepare(struct sk_buff *skb, struct ip_fraglist_iter *iter);
183 
184 static inline struct sk_buff *ip_fraglist_next(struct ip_fraglist_iter *iter)
185 {
186 	struct sk_buff *skb = iter->frag;
187 
188 	iter->frag = skb->next;
189 	skb_mark_not_on_list(skb);
190 
191 	return skb;
192 }
193 
194 struct ip_frag_state {
195 	bool		DF;
196 	unsigned int	hlen;
197 	unsigned int	ll_rs;
198 	unsigned int	mtu;
199 	unsigned int	left;
200 	int		offset;
201 	int		ptr;
202 	__be16		not_last_frag;
203 };
204 
205 void ip_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int ll_rs,
206 		  unsigned int mtu, bool DF, struct ip_frag_state *state);
207 struct sk_buff *ip_frag_next(struct sk_buff *skb,
208 			     struct ip_frag_state *state);
209 
210 void ip_send_check(struct iphdr *ip);
211 int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
212 int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
213 
214 int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
215 		    __u8 tos);
216 void ip_init(void);
217 int ip_append_data(struct sock *sk, struct flowi4 *fl4,
218 		   int getfrag(void *from, char *to, int offset, int len,
219 			       int odd, struct sk_buff *skb),
220 		   void *from, int len, int protolen,
221 		   struct ipcm_cookie *ipc,
222 		   struct rtable **rt,
223 		   unsigned int flags);
224 int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
225 		       struct sk_buff *skb);
226 struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
227 			      struct sk_buff_head *queue,
228 			      struct inet_cork *cork);
229 int ip_send_skb(struct net *net, struct sk_buff *skb);
230 int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
231 void ip_flush_pending_frames(struct sock *sk);
232 struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
233 			    int getfrag(void *from, char *to, int offset,
234 					int len, int odd, struct sk_buff *skb),
235 			    void *from, int length, int transhdrlen,
236 			    struct ipcm_cookie *ipc, struct rtable **rtp,
237 			    struct inet_cork *cork, unsigned int flags);
238 
239 int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
240 
241 static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
242 {
243 	return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
244 }
245 
246 /* Get the route scope that should be used when sending a packet. */
247 static inline u8 ip_sendmsg_scope(const struct inet_sock *inet,
248 				  const struct ipcm_cookie *ipc,
249 				  const struct msghdr *msg)
250 {
251 	if (sock_flag(&inet->sk, SOCK_LOCALROUTE) ||
252 	    msg->msg_flags & MSG_DONTROUTE ||
253 	    (ipc->opt && ipc->opt->opt.is_strictroute))
254 		return RT_SCOPE_LINK;
255 
256 	return RT_SCOPE_UNIVERSE;
257 }
258 
259 static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
260 {
261 	return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(READ_ONCE(inet->tos));
262 }
263 
264 /* datagram.c */
265 int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
266 int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
267 
268 void ip4_datagram_release_cb(struct sock *sk);
269 
270 struct ip_reply_arg {
271 	struct kvec iov[1];
272 	int	    flags;
273 	__wsum 	    csum;
274 	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
275 				/* -1 if not needed */
276 	int	    bound_dev_if;
277 	u8  	    tos;
278 	kuid_t	    uid;
279 };
280 
281 #define IP_REPLY_ARG_NOSRCCHECK 1
282 
283 static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
284 {
285 	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
286 }
287 
288 void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
289 			   const struct ip_options *sopt,
290 			   __be32 daddr, __be32 saddr,
291 			   const struct ip_reply_arg *arg,
292 			   unsigned int len, u64 transmit_time, u32 txhash);
293 
294 #define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
295 #define __IP_INC_STATS(net, field)	__SNMP_INC_STATS64((net)->mib.ip_statistics, field)
296 #define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
297 #define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
298 #define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
299 #define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
300 #define NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
301 #define __NET_INC_STATS(net, field)	__SNMP_INC_STATS((net)->mib.net_statistics, field)
302 #define NET_ADD_STATS(net, field, adnd)	SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
303 #define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
304 
305 static inline u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
306 {
307 	return  *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
308 }
309 
310 unsigned long snmp_fold_field(void __percpu *mib, int offt);
311 #if BITS_PER_LONG==32
312 u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
313 			 size_t syncp_offset);
314 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
315 #else
316 static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
317 					size_t syncp_offset)
318 {
319 	return snmp_get_cpu_field(mib, cpu, offct);
320 
321 }
322 
323 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
324 {
325 	return snmp_fold_field(mib, offt);
326 }
327 #endif
328 
329 #define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
330 { \
331 	int i, c; \
332 	for_each_possible_cpu(c) { \
333 		for (i = 0; stats_list[i].name; i++) \
334 			buff64[i] += snmp_get_cpu_field64( \
335 					mib_statistic, \
336 					c, stats_list[i].entry, \
337 					offset); \
338 	} \
339 }
340 
341 #define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
342 { \
343 	int i, c; \
344 	for_each_possible_cpu(c) { \
345 		for (i = 0; stats_list[i].name; i++) \
346 			buff[i] += snmp_get_cpu_field( \
347 						mib_statistic, \
348 						c, stats_list[i].entry); \
349 	} \
350 }
351 
352 void inet_get_local_port_range(const struct net *net, int *low, int *high);
353 void inet_sk_get_local_port_range(const struct sock *sk, int *low, int *high);
354 
355 #ifdef CONFIG_SYSCTL
356 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
357 {
358 	if (!net->ipv4.sysctl_local_reserved_ports)
359 		return false;
360 	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
361 }
362 
363 static inline bool sysctl_dev_name_is_allowed(const char *name)
364 {
365 	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
366 }
367 
368 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
369 {
370 	return port < READ_ONCE(net->ipv4.sysctl_ip_prot_sock);
371 }
372 
373 #else
374 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
375 {
376 	return false;
377 }
378 
379 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
380 {
381 	return port < PROT_SOCK;
382 }
383 #endif
384 
385 __be32 inet_current_timestamp(void);
386 
387 /* From inetpeer.c */
388 extern int inet_peer_threshold;
389 extern int inet_peer_minttl;
390 extern int inet_peer_maxttl;
391 
392 void ipfrag_init(void);
393 
394 void ip_static_sysctl_init(void);
395 
396 #define IP4_REPLY_MARK(net, mark) \
397 	(READ_ONCE((net)->ipv4.sysctl_fwmark_reflect) ? (mark) : 0)
398 
399 static inline bool ip_is_fragment(const struct iphdr *iph)
400 {
401 	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
402 }
403 
404 #ifdef CONFIG_INET
405 #include <net/dst.h>
406 
407 /* The function in 2.2 was invalid, producing wrong result for
408  * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
409 static inline
410 int ip_decrease_ttl(struct iphdr *iph)
411 {
412 	u32 check = (__force u32)iph->check;
413 	check += (__force u32)htons(0x0100);
414 	iph->check = (__force __sum16)(check + (check>=0xFFFF));
415 	return --iph->ttl;
416 }
417 
418 static inline int ip_mtu_locked(const struct dst_entry *dst)
419 {
420 	const struct rtable *rt = (const struct rtable *)dst;
421 
422 	return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
423 }
424 
425 static inline
426 int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
427 {
428 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
429 
430 	return  pmtudisc == IP_PMTUDISC_DO ||
431 		(pmtudisc == IP_PMTUDISC_WANT &&
432 		 !ip_mtu_locked(dst));
433 }
434 
435 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
436 {
437 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
438 
439 	return pmtudisc != IP_PMTUDISC_INTERFACE &&
440 	       pmtudisc != IP_PMTUDISC_OMIT;
441 }
442 
443 static inline bool ip_sk_use_pmtu(const struct sock *sk)
444 {
445 	return READ_ONCE(inet_sk(sk)->pmtudisc) < IP_PMTUDISC_PROBE;
446 }
447 
448 static inline bool ip_sk_ignore_df(const struct sock *sk)
449 {
450 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
451 
452 	return pmtudisc < IP_PMTUDISC_DO || pmtudisc == IP_PMTUDISC_OMIT;
453 }
454 
455 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
456 						    bool forwarding)
457 {
458 	const struct rtable *rt = container_of(dst, struct rtable, dst);
459 	struct net *net = dev_net(dst->dev);
460 	unsigned int mtu;
461 
462 	if (READ_ONCE(net->ipv4.sysctl_ip_fwd_use_pmtu) ||
463 	    ip_mtu_locked(dst) ||
464 	    !forwarding) {
465 		mtu = rt->rt_pmtu;
466 		if (mtu && time_before(jiffies, rt->dst.expires))
467 			goto out;
468 	}
469 
470 	/* 'forwarding = true' case should always honour route mtu */
471 	mtu = dst_metric_raw(dst, RTAX_MTU);
472 	if (mtu)
473 		goto out;
474 
475 	mtu = READ_ONCE(dst->dev->mtu);
476 
477 	if (unlikely(ip_mtu_locked(dst))) {
478 		if (rt->rt_uses_gateway && mtu > 576)
479 			mtu = 576;
480 	}
481 
482 out:
483 	mtu = min_t(unsigned int, mtu, IP_MAX_MTU);
484 
485 	return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
486 }
487 
488 static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
489 					  const struct sk_buff *skb)
490 {
491 	unsigned int mtu;
492 
493 	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
494 		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
495 
496 		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
497 	}
498 
499 	mtu = min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
500 	return mtu - lwtunnel_headroom(skb_dst(skb)->lwtstate, mtu);
501 }
502 
503 struct dst_metrics *ip_fib_metrics_init(struct net *net, struct nlattr *fc_mx,
504 					int fc_mx_len,
505 					struct netlink_ext_ack *extack);
506 static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
507 {
508 	if (fib_metrics != &dst_default_metrics &&
509 	    refcount_dec_and_test(&fib_metrics->refcnt))
510 		kfree(fib_metrics);
511 }
512 
513 /* ipv4 and ipv6 both use refcounted metrics if it is not the default */
514 static inline
515 void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
516 {
517 	dst_init_metrics(dst, fib_metrics->metrics, true);
518 
519 	if (fib_metrics != &dst_default_metrics) {
520 		dst->_metrics |= DST_METRICS_REFCOUNTED;
521 		refcount_inc(&fib_metrics->refcnt);
522 	}
523 }
524 
525 static inline
526 void ip_dst_metrics_put(struct dst_entry *dst)
527 {
528 	struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
529 
530 	if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
531 		kfree(p);
532 }
533 
534 void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
535 
536 static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
537 					struct sock *sk, int segs)
538 {
539 	struct iphdr *iph = ip_hdr(skb);
540 
541 	/* We had many attacks based on IPID, use the private
542 	 * generator as much as we can.
543 	 */
544 	if (sk && inet_sk(sk)->inet_daddr) {
545 		int val;
546 
547 		/* avoid atomic operations for TCP,
548 		 * as we hold socket lock at this point.
549 		 */
550 		if (sk_is_tcp(sk)) {
551 			sock_owned_by_me(sk);
552 			val = atomic_read(&inet_sk(sk)->inet_id);
553 			atomic_set(&inet_sk(sk)->inet_id, val + segs);
554 		} else {
555 			val = atomic_add_return(segs, &inet_sk(sk)->inet_id);
556 		}
557 		iph->id = htons(val);
558 		return;
559 	}
560 	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
561 		iph->id = 0;
562 	} else {
563 		/* Unfortunately we need the big hammer to get a suitable IPID */
564 		__ip_select_ident(net, iph, segs);
565 	}
566 }
567 
568 static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
569 				   struct sock *sk)
570 {
571 	ip_select_ident_segs(net, skb, sk, 1);
572 }
573 
574 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
575 {
576 	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
577 				  skb->len, proto, 0);
578 }
579 
580 /* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
581  * Equivalent to :	flow->v4addrs.src = iph->saddr;
582  *			flow->v4addrs.dst = iph->daddr;
583  */
584 static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
585 					    const struct iphdr *iph)
586 {
587 	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
588 		     offsetof(typeof(flow->addrs), v4addrs.src) +
589 			      sizeof(flow->addrs.v4addrs.src));
590 	memcpy(&flow->addrs.v4addrs, &iph->addrs, sizeof(flow->addrs.v4addrs));
591 	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
592 }
593 
594 /*
595  *	Map a multicast IP onto multicast MAC for type ethernet.
596  */
597 
598 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
599 {
600 	__u32 addr=ntohl(naddr);
601 	buf[0]=0x01;
602 	buf[1]=0x00;
603 	buf[2]=0x5e;
604 	buf[5]=addr&0xFF;
605 	addr>>=8;
606 	buf[4]=addr&0xFF;
607 	addr>>=8;
608 	buf[3]=addr&0x7F;
609 }
610 
611 /*
612  *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
613  *	Leave P_Key as 0 to be filled in by driver.
614  */
615 
616 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
617 {
618 	__u32 addr;
619 	unsigned char scope = broadcast[5] & 0xF;
620 
621 	buf[0]  = 0;		/* Reserved */
622 	buf[1]  = 0xff;		/* Multicast QPN */
623 	buf[2]  = 0xff;
624 	buf[3]  = 0xff;
625 	addr    = ntohl(naddr);
626 	buf[4]  = 0xff;
627 	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
628 	buf[6]  = 0x40;		/* IPv4 signature */
629 	buf[7]  = 0x1b;
630 	buf[8]  = broadcast[8];		/* P_Key */
631 	buf[9]  = broadcast[9];
632 	buf[10] = 0;
633 	buf[11] = 0;
634 	buf[12] = 0;
635 	buf[13] = 0;
636 	buf[14] = 0;
637 	buf[15] = 0;
638 	buf[19] = addr & 0xff;
639 	addr  >>= 8;
640 	buf[18] = addr & 0xff;
641 	addr  >>= 8;
642 	buf[17] = addr & 0xff;
643 	addr  >>= 8;
644 	buf[16] = addr & 0x0f;
645 }
646 
647 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
648 {
649 	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
650 		memcpy(buf, broadcast, 4);
651 	else
652 		memcpy(buf, &naddr, sizeof(naddr));
653 }
654 
655 #if IS_ENABLED(CONFIG_IPV6)
656 #include <linux/ipv6.h>
657 #endif
658 
659 static __inline__ void inet_reset_saddr(struct sock *sk)
660 {
661 	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
662 #if IS_ENABLED(CONFIG_IPV6)
663 	if (sk->sk_family == PF_INET6) {
664 		struct ipv6_pinfo *np = inet6_sk(sk);
665 
666 		memset(&np->saddr, 0, sizeof(np->saddr));
667 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
668 	}
669 #endif
670 }
671 
672 #endif
673 
674 static inline unsigned int ipv4_addr_hash(__be32 ip)
675 {
676 	return (__force unsigned int) ip;
677 }
678 
679 static inline u32 ipv4_portaddr_hash(const struct net *net,
680 				     __be32 saddr,
681 				     unsigned int port)
682 {
683 	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
684 }
685 
686 bool ip_call_ra_chain(struct sk_buff *skb);
687 
688 /*
689  *	Functions provided by ip_fragment.c
690  */
691 
692 enum ip_defrag_users {
693 	IP_DEFRAG_LOCAL_DELIVER,
694 	IP_DEFRAG_CALL_RA_CHAIN,
695 	IP_DEFRAG_CONNTRACK_IN,
696 	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
697 	IP_DEFRAG_CONNTRACK_OUT,
698 	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
699 	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
700 	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
701 	IP_DEFRAG_VS_IN,
702 	IP_DEFRAG_VS_OUT,
703 	IP_DEFRAG_VS_FWD,
704 	IP_DEFRAG_AF_PACKET,
705 	IP_DEFRAG_MACVLAN,
706 };
707 
708 /* Return true if the value of 'user' is between 'lower_bond'
709  * and 'upper_bond' inclusively.
710  */
711 static inline bool ip_defrag_user_in_between(u32 user,
712 					     enum ip_defrag_users lower_bond,
713 					     enum ip_defrag_users upper_bond)
714 {
715 	return user >= lower_bond && user <= upper_bond;
716 }
717 
718 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
719 #ifdef CONFIG_INET
720 struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
721 #else
722 static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
723 {
724 	return skb;
725 }
726 #endif
727 
728 /*
729  *	Functions provided by ip_forward.c
730  */
731 
732 int ip_forward(struct sk_buff *skb);
733 
734 /*
735  *	Functions provided by ip_options.c
736  */
737 
738 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
739 		      __be32 daddr, struct rtable *rt);
740 
741 int __ip_options_echo(struct net *net, struct ip_options *dopt,
742 		      struct sk_buff *skb, const struct ip_options *sopt);
743 static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
744 				  struct sk_buff *skb)
745 {
746 	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
747 }
748 
749 void ip_options_fragment(struct sk_buff *skb);
750 int __ip_options_compile(struct net *net, struct ip_options *opt,
751 			 struct sk_buff *skb, __be32 *info);
752 int ip_options_compile(struct net *net, struct ip_options *opt,
753 		       struct sk_buff *skb);
754 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
755 		   sockptr_t data, int optlen);
756 void ip_options_undo(struct ip_options *opt);
757 void ip_forward_options(struct sk_buff *skb);
758 int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
759 
760 /*
761  *	Functions provided by ip_sockglue.c
762  */
763 
764 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
765 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
766 			 struct sk_buff *skb, int tlen, int offset);
767 int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
768 		 struct ipcm_cookie *ipc, bool allow_ipv6);
769 DECLARE_STATIC_KEY_FALSE(ip4_min_ttl);
770 int do_ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
771 		     unsigned int optlen);
772 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
773 		  unsigned int optlen);
774 int do_ip_getsockopt(struct sock *sk, int level, int optname,
775 		     sockptr_t optval, sockptr_t optlen);
776 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
777 		  int __user *optlen);
778 int ip_ra_control(struct sock *sk, unsigned char on,
779 		  void (*destructor)(struct sock *));
780 
781 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
782 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
783 		   u32 info, u8 *payload);
784 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
785 		    u32 info);
786 
787 static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
788 {
789 	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
790 }
791 
792 bool icmp_global_allow(void);
793 extern int sysctl_icmp_msgs_per_sec;
794 extern int sysctl_icmp_msgs_burst;
795 
796 #ifdef CONFIG_PROC_FS
797 int ip_misc_proc_init(void);
798 #endif
799 
800 int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
801 				struct netlink_ext_ack *extack);
802 
803 static inline bool inetdev_valid_mtu(unsigned int mtu)
804 {
805 	return likely(mtu >= IPV4_MIN_MTU);
806 }
807 
808 void ip_sock_set_freebind(struct sock *sk);
809 int ip_sock_set_mtu_discover(struct sock *sk, int val);
810 void ip_sock_set_pktinfo(struct sock *sk);
811 void ip_sock_set_recverr(struct sock *sk);
812 void ip_sock_set_tos(struct sock *sk, int val);
813 void  __ip_sock_set_tos(struct sock *sk, int val);
814 
815 #endif	/* _IP_H */
816