Linux Audio

Check our new training course

Loading...
v5.14.15
  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
 39extern unsigned int sysctl_fib_sync_mem;
 40extern unsigned int sysctl_fib_sync_mem_min;
 41extern unsigned int sysctl_fib_sync_mem_max;
 42
 43struct sock;
 44
 45struct 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
 62static inline bool ipv4_l3mdev_skb(u16 flags)
 63{
 64	return !!(flags & IPSKB_L3SLAVE);
 65}
 66
 67static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
 68{
 69	return ip_hdr(skb)->ihl * 4;
 70}
 71
 72struct 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
 83static inline void ipcm_init(struct ipcm_cookie *ipcm)
 84{
 85	*ipcm = (struct ipcm_cookie) { .tos = -1 };
 86}
 87
 88static 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 */
103static 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
123struct 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
141struct msghdr;
142struct net_device;
143struct packet_type;
144struct rtable;
145struct sockaddr;
146
147int igmp_mc_init(void);
148
149/*
150 *	Functions provided by ip.c
151 */
152
153int 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);
156int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
157	   struct net_device *orig_dev);
158void ip_list_rcv(struct list_head *head, struct packet_type *pt,
159		 struct net_device *orig_dev);
160int ip_local_deliver(struct sk_buff *skb);
161void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int proto);
162int ip_mr_input(struct sk_buff *skb);
163int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
164int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
165int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
166		   int (*output)(struct net *, struct sock *, struct sk_buff *));
167
168struct ip_fraglist_iter {
169	struct sk_buff	*frag;
170	struct iphdr	*iph;
171	int		offset;
172	unsigned int	hlen;
173};
174
175void ip_fraglist_init(struct sk_buff *skb, struct iphdr *iph,
176		      unsigned int hlen, struct ip_fraglist_iter *iter);
177void ip_fraglist_prepare(struct sk_buff *skb, struct ip_fraglist_iter *iter);
178
179static 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
189struct 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
200void 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);
202struct sk_buff *ip_frag_next(struct sk_buff *skb,
203			     struct ip_frag_state *state);
204
205void ip_send_check(struct iphdr *ip);
206int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
207int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
208
209int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
210		    __u8 tos);
211void ip_init(void);
212int 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);
219int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
220		       struct sk_buff *skb);
221ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
222		       int offset, size_t size, int flags);
223struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
224			      struct sk_buff_head *queue,
225			      struct inet_cork *cork);
226int ip_send_skb(struct net *net, struct sk_buff *skb);
227int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
228void ip_flush_pending_frames(struct sock *sk);
229struct 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
236int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
237
238static 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
243static 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
248static 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 */
254int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
255int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
256
257void ip4_datagram_release_cb(struct sock *sk);
258
259struct 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
272static 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
277void 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
294u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
295unsigned long snmp_fold_field(void __percpu *mib, int offt);
296#if BITS_PER_LONG==32
297u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
298			 size_t syncp_offset);
299u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
300#else
301static 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
308static 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
337void inet_get_local_port_range(struct net *net, int *low, int *high);
338
339#ifdef CONFIG_SYSCTL
340static 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
347static inline bool sysctl_dev_name_is_allowed(const char *name)
348{
349	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
350}
351
352static 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
358static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
359{
360	return false;
361}
362
363static 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 */
372extern int inet_peer_threshold;
373extern int inet_peer_minttl;
374extern int inet_peer_maxttl;
375
376void ipfrag_init(void);
377
378void ip_static_sysctl_init(void);
379
380#define IP4_REPLY_MARK(net, mark) \
381	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
382
383static 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) */
393static inline
394int 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
402static 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
409static inline
410int 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
419static 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
425static inline bool ip_sk_use_pmtu(const struct sock *sk)
426{
427	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
428}
429
430static 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
436static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
437						    bool forwarding)
438{
439	struct net *net = dev_net(dst->dev);
440	unsigned int mtu;
441
442	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
443	    ip_mtu_locked(dst) ||
444	    !forwarding)
445		return dst_mtu(dst);
446
447	/* 'forwarding = true' case should always honour route mtu */
448	mtu = dst_metric_raw(dst, RTAX_MTU);
449	if (!mtu)
450		mtu = min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);
451
452	return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
453}
454
455static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
456					  const struct sk_buff *skb)
457{
458	unsigned int mtu;
459
460	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
461		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
462
463		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
464	}
465
466	mtu = min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
467	return mtu - lwtunnel_headroom(skb_dst(skb)->lwtstate, mtu);
468}
469
470struct dst_metrics *ip_fib_metrics_init(struct net *net, struct nlattr *fc_mx,
471					int fc_mx_len,
472					struct netlink_ext_ack *extack);
473static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
474{
475	if (fib_metrics != &dst_default_metrics &&
476	    refcount_dec_and_test(&fib_metrics->refcnt))
477		kfree(fib_metrics);
478}
479
480/* ipv4 and ipv6 both use refcounted metrics if it is not the default */
481static inline
482void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
483{
484	dst_init_metrics(dst, fib_metrics->metrics, true);
485
486	if (fib_metrics != &dst_default_metrics) {
487		dst->_metrics |= DST_METRICS_REFCOUNTED;
488		refcount_inc(&fib_metrics->refcnt);
489	}
490}
491
492static inline
493void ip_dst_metrics_put(struct dst_entry *dst)
494{
495	struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
496
497	if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
498		kfree(p);
499}
500
501u32 ip_idents_reserve(u32 hash, int segs);
502void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
503
504static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
505					struct sock *sk, int segs)
506{
507	struct iphdr *iph = ip_hdr(skb);
508
509	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
510		/* This is only to work around buggy Windows95/2000
511		 * VJ compression implementations.  If the ID field
512		 * does not change, they drop every other packet in
513		 * a TCP stream using header compression.
514		 */
515		if (sk && inet_sk(sk)->inet_daddr) {
516			iph->id = htons(inet_sk(sk)->inet_id);
517			inet_sk(sk)->inet_id += segs;
518		} else {
519			iph->id = 0;
520		}
521	} else {
522		__ip_select_ident(net, iph, segs);
523	}
524}
525
526static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
527				   struct sock *sk)
528{
529	ip_select_ident_segs(net, skb, sk, 1);
530}
531
532static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
533{
534	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
535				  skb->len, proto, 0);
536}
537
538/* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
539 * Equivalent to :	flow->v4addrs.src = iph->saddr;
540 *			flow->v4addrs.dst = iph->daddr;
541 */
542static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
543					    const struct iphdr *iph)
544{
545	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
546		     offsetof(typeof(flow->addrs), v4addrs.src) +
547			      sizeof(flow->addrs.v4addrs.src));
548	memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
549	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
550}
551
552static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
553{
554	const struct iphdr *iph = skb_gro_network_header(skb);
555
556	return csum_tcpudp_nofold(iph->saddr, iph->daddr,
557				  skb_gro_len(skb), proto, 0);
558}
559
560/*
561 *	Map a multicast IP onto multicast MAC for type ethernet.
562 */
563
564static inline void ip_eth_mc_map(__be32 naddr, char *buf)
565{
566	__u32 addr=ntohl(naddr);
567	buf[0]=0x01;
568	buf[1]=0x00;
569	buf[2]=0x5e;
570	buf[5]=addr&0xFF;
571	addr>>=8;
572	buf[4]=addr&0xFF;
573	addr>>=8;
574	buf[3]=addr&0x7F;
575}
576
577/*
578 *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
579 *	Leave P_Key as 0 to be filled in by driver.
580 */
581
582static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
583{
584	__u32 addr;
585	unsigned char scope = broadcast[5] & 0xF;
586
587	buf[0]  = 0;		/* Reserved */
588	buf[1]  = 0xff;		/* Multicast QPN */
589	buf[2]  = 0xff;
590	buf[3]  = 0xff;
591	addr    = ntohl(naddr);
592	buf[4]  = 0xff;
593	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
594	buf[6]  = 0x40;		/* IPv4 signature */
595	buf[7]  = 0x1b;
596	buf[8]  = broadcast[8];		/* P_Key */
597	buf[9]  = broadcast[9];
598	buf[10] = 0;
599	buf[11] = 0;
600	buf[12] = 0;
601	buf[13] = 0;
602	buf[14] = 0;
603	buf[15] = 0;
604	buf[19] = addr & 0xff;
605	addr  >>= 8;
606	buf[18] = addr & 0xff;
607	addr  >>= 8;
608	buf[17] = addr & 0xff;
609	addr  >>= 8;
610	buf[16] = addr & 0x0f;
611}
612
613static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
614{
615	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
616		memcpy(buf, broadcast, 4);
617	else
618		memcpy(buf, &naddr, sizeof(naddr));
619}
620
621#if IS_ENABLED(CONFIG_IPV6)
622#include <linux/ipv6.h>
623#endif
624
625static __inline__ void inet_reset_saddr(struct sock *sk)
626{
627	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
628#if IS_ENABLED(CONFIG_IPV6)
629	if (sk->sk_family == PF_INET6) {
630		struct ipv6_pinfo *np = inet6_sk(sk);
631
632		memset(&np->saddr, 0, sizeof(np->saddr));
633		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
634	}
635#endif
636}
637
638#endif
639
640static inline unsigned int ipv4_addr_hash(__be32 ip)
641{
642	return (__force unsigned int) ip;
643}
644
645static inline u32 ipv4_portaddr_hash(const struct net *net,
646				     __be32 saddr,
647				     unsigned int port)
648{
649	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
650}
651
652bool ip_call_ra_chain(struct sk_buff *skb);
653
654/*
655 *	Functions provided by ip_fragment.c
656 */
657
658enum ip_defrag_users {
659	IP_DEFRAG_LOCAL_DELIVER,
660	IP_DEFRAG_CALL_RA_CHAIN,
661	IP_DEFRAG_CONNTRACK_IN,
662	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
663	IP_DEFRAG_CONNTRACK_OUT,
664	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
665	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
666	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
667	IP_DEFRAG_VS_IN,
668	IP_DEFRAG_VS_OUT,
669	IP_DEFRAG_VS_FWD,
670	IP_DEFRAG_AF_PACKET,
671	IP_DEFRAG_MACVLAN,
672};
673
674/* Return true if the value of 'user' is between 'lower_bond'
675 * and 'upper_bond' inclusively.
676 */
677static inline bool ip_defrag_user_in_between(u32 user,
678					     enum ip_defrag_users lower_bond,
679					     enum ip_defrag_users upper_bond)
680{
681	return user >= lower_bond && user <= upper_bond;
682}
683
684int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
685#ifdef CONFIG_INET
686struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
687#else
688static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
689{
690	return skb;
691}
692#endif
693
694/*
695 *	Functions provided by ip_forward.c
696 */
697
698int ip_forward(struct sk_buff *skb);
699
700/*
701 *	Functions provided by ip_options.c
702 */
703
704void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
705		      __be32 daddr, struct rtable *rt, int is_frag);
706
707int __ip_options_echo(struct net *net, struct ip_options *dopt,
708		      struct sk_buff *skb, const struct ip_options *sopt);
709static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
710				  struct sk_buff *skb)
711{
712	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
713}
714
715void ip_options_fragment(struct sk_buff *skb);
716int __ip_options_compile(struct net *net, struct ip_options *opt,
717			 struct sk_buff *skb, __be32 *info);
718int ip_options_compile(struct net *net, struct ip_options *opt,
719		       struct sk_buff *skb);
720int ip_options_get(struct net *net, struct ip_options_rcu **optp,
721		   sockptr_t data, int optlen);
722void ip_options_undo(struct ip_options *opt);
723void ip_forward_options(struct sk_buff *skb);
724int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
725
726/*
727 *	Functions provided by ip_sockglue.c
728 */
729
730void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
731void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
732			 struct sk_buff *skb, int tlen, int offset);
733int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
734		 struct ipcm_cookie *ipc, bool allow_ipv6);
735int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
736		  unsigned int optlen);
737int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
738		  int __user *optlen);
739int ip_ra_control(struct sock *sk, unsigned char on,
740		  void (*destructor)(struct sock *));
741
742int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
743void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
744		   u32 info, u8 *payload);
745void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
746		    u32 info);
747
748static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
749{
750	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
751}
752
753bool icmp_global_allow(void);
754extern int sysctl_icmp_msgs_per_sec;
755extern int sysctl_icmp_msgs_burst;
756
757#ifdef CONFIG_PROC_FS
758int ip_misc_proc_init(void);
759#endif
760
761int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
762				struct netlink_ext_ack *extack);
763
764static inline bool inetdev_valid_mtu(unsigned int mtu)
765{
766	return likely(mtu >= IPV4_MIN_MTU);
767}
768
769void ip_sock_set_freebind(struct sock *sk);
770int ip_sock_set_mtu_discover(struct sock *sk, int val);
771void ip_sock_set_pktinfo(struct sock *sk);
772void ip_sock_set_recverr(struct sock *sk);
773void ip_sock_set_tos(struct sock *sk, int val);
774
775#endif	/* _IP_H */
v5.9
  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
 35#define IPV4_MAX_PMTU		65535U		/* RFC 2675, Section 5.1 */
 36#define IPV4_MIN_MTU		68			/* RFC 791 */
 37
 38extern unsigned int sysctl_fib_sync_mem;
 39extern unsigned int sysctl_fib_sync_mem_min;
 40extern unsigned int sysctl_fib_sync_mem_max;
 41
 42struct sock;
 43
 44struct inet_skb_parm {
 45	int			iif;
 46	struct ip_options	opt;		/* Compiled IP options		*/
 47	u16			flags;
 48
 49#define IPSKB_FORWARDED		BIT(0)
 50#define IPSKB_XFRM_TUNNEL_SIZE	BIT(1)
 51#define IPSKB_XFRM_TRANSFORMED	BIT(2)
 52#define IPSKB_FRAG_COMPLETE	BIT(3)
 53#define IPSKB_REROUTED		BIT(4)
 54#define IPSKB_DOREDIRECT	BIT(5)
 55#define IPSKB_FRAG_PMTU		BIT(6)
 56#define IPSKB_L3SLAVE		BIT(7)
 57
 58	u16			frag_max_size;
 59};
 60
 61static inline bool ipv4_l3mdev_skb(u16 flags)
 62{
 63	return !!(flags & IPSKB_L3SLAVE);
 64}
 65
 66static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
 67{
 68	return ip_hdr(skb)->ihl * 4;
 69}
 70
 71struct ipcm_cookie {
 72	struct sockcm_cookie	sockc;
 73	__be32			addr;
 74	int			oif;
 75	struct ip_options_rcu	*opt;
 76	__u8			ttl;
 77	__s16			tos;
 78	char			priority;
 79	__u16			gso_size;
 80};
 81
 82static inline void ipcm_init(struct ipcm_cookie *ipcm)
 83{
 84	*ipcm = (struct ipcm_cookie) { .tos = -1 };
 85}
 86
 87static inline void ipcm_init_sk(struct ipcm_cookie *ipcm,
 88				const struct inet_sock *inet)
 89{
 90	ipcm_init(ipcm);
 91
 92	ipcm->sockc.mark = inet->sk.sk_mark;
 93	ipcm->sockc.tsflags = inet->sk.sk_tsflags;
 94	ipcm->oif = inet->sk.sk_bound_dev_if;
 95	ipcm->addr = inet->inet_saddr;
 96}
 97
 98#define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
 99#define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
100
101/* return enslaved device index if relevant */
102static inline int inet_sdif(struct sk_buff *skb)
103{
104#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
105	if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
106		return IPCB(skb)->iif;
107#endif
108	return 0;
109}
110
111/* Special input handler for packets caught by router alert option.
112   They are selected only by protocol field, and then processed likely
113   local ones; but only if someone wants them! Otherwise, router
114   not running rsvpd will kill RSVP.
115
116   It is user level problem, what it will make with them.
117   I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
118   but receiver should be enough clever f.e. to forward mtrace requests,
119   sent to multicast group to reach destination designated router.
120 */
121
122struct ip_ra_chain {
123	struct ip_ra_chain __rcu *next;
124	struct sock		*sk;
125	union {
126		void			(*destructor)(struct sock *);
127		struct sock		*saved_sk;
128	};
129	struct rcu_head		rcu;
130};
131
132/* IP flags. */
133#define IP_CE		0x8000		/* Flag: "Congestion"		*/
134#define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
135#define IP_MF		0x2000		/* Flag: "More Fragments"	*/
136#define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
137
138#define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
139
140struct msghdr;
141struct net_device;
142struct packet_type;
143struct rtable;
144struct sockaddr;
145
146int igmp_mc_init(void);
147
148/*
149 *	Functions provided by ip.c
150 */
151
152int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
153			  __be32 saddr, __be32 daddr,
154			  struct ip_options_rcu *opt);
155int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
156	   struct net_device *orig_dev);
157void ip_list_rcv(struct list_head *head, struct packet_type *pt,
158		 struct net_device *orig_dev);
159int ip_local_deliver(struct sk_buff *skb);
160void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int proto);
161int ip_mr_input(struct sk_buff *skb);
162int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
163int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
164int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
165		   int (*output)(struct net *, struct sock *, struct sk_buff *));
166
167struct ip_fraglist_iter {
168	struct sk_buff	*frag;
169	struct iphdr	*iph;
170	int		offset;
171	unsigned int	hlen;
172};
173
174void ip_fraglist_init(struct sk_buff *skb, struct iphdr *iph,
175		      unsigned int hlen, struct ip_fraglist_iter *iter);
176void ip_fraglist_prepare(struct sk_buff *skb, struct ip_fraglist_iter *iter);
177
178static inline struct sk_buff *ip_fraglist_next(struct ip_fraglist_iter *iter)
179{
180	struct sk_buff *skb = iter->frag;
181
182	iter->frag = skb->next;
183	skb_mark_not_on_list(skb);
184
185	return skb;
186}
187
188struct ip_frag_state {
189	bool		DF;
190	unsigned int	hlen;
191	unsigned int	ll_rs;
192	unsigned int	mtu;
193	unsigned int	left;
194	int		offset;
195	int		ptr;
196	__be16		not_last_frag;
197};
198
199void ip_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int ll_rs,
200		  unsigned int mtu, bool DF, struct ip_frag_state *state);
201struct sk_buff *ip_frag_next(struct sk_buff *skb,
202			     struct ip_frag_state *state);
203
204void ip_send_check(struct iphdr *ip);
205int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
206int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
207
208int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
209		    __u8 tos);
210void ip_init(void);
211int ip_append_data(struct sock *sk, struct flowi4 *fl4,
212		   int getfrag(void *from, char *to, int offset, int len,
213			       int odd, struct sk_buff *skb),
214		   void *from, int len, int protolen,
215		   struct ipcm_cookie *ipc,
216		   struct rtable **rt,
217		   unsigned int flags);
218int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
219		       struct sk_buff *skb);
220ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
221		       int offset, size_t size, int flags);
222struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
223			      struct sk_buff_head *queue,
224			      struct inet_cork *cork);
225int ip_send_skb(struct net *net, struct sk_buff *skb);
226int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
227void ip_flush_pending_frames(struct sock *sk);
228struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
229			    int getfrag(void *from, char *to, int offset,
230					int len, int odd, struct sk_buff *skb),
231			    void *from, int length, int transhdrlen,
232			    struct ipcm_cookie *ipc, struct rtable **rtp,
233			    struct inet_cork *cork, unsigned int flags);
234
235int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
236
237static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
238{
239	return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
240}
241
242static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
243{
244	return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
245}
246
247static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
248{
249	return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
250}
251
252/* datagram.c */
253int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
254int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
255
256void ip4_datagram_release_cb(struct sock *sk);
257
258struct ip_reply_arg {
259	struct kvec iov[1];
260	int	    flags;
261	__wsum 	    csum;
262	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
263				/* -1 if not needed */
264	int	    bound_dev_if;
265	u8  	    tos;
266	kuid_t	    uid;
267};
268
269#define IP_REPLY_ARG_NOSRCCHECK 1
270
271static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
272{
273	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
274}
275
276void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
277			   const struct ip_options *sopt,
278			   __be32 daddr, __be32 saddr,
279			   const struct ip_reply_arg *arg,
280			   unsigned int len, u64 transmit_time);
281
282#define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
283#define __IP_INC_STATS(net, field)	__SNMP_INC_STATS64((net)->mib.ip_statistics, field)
284#define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
285#define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
286#define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_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 NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
289#define __NET_INC_STATS(net, field)	__SNMP_INC_STATS((net)->mib.net_statistics, field)
290#define NET_ADD_STATS(net, field, adnd)	SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
291#define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
292
293u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
294unsigned long snmp_fold_field(void __percpu *mib, int offt);
295#if BITS_PER_LONG==32
296u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
297			 size_t syncp_offset);
298u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
299#else
300static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
301					size_t syncp_offset)
302{
303	return snmp_get_cpu_field(mib, cpu, offct);
304
305}
306
307static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
308{
309	return snmp_fold_field(mib, offt);
310}
311#endif
312
313#define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
314{ \
315	int i, c; \
316	for_each_possible_cpu(c) { \
317		for (i = 0; stats_list[i].name; i++) \
318			buff64[i] += snmp_get_cpu_field64( \
319					mib_statistic, \
320					c, stats_list[i].entry, \
321					offset); \
322	} \
323}
324
325#define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
326{ \
327	int i, c; \
328	for_each_possible_cpu(c) { \
329		for (i = 0; stats_list[i].name; i++) \
330			buff[i] += snmp_get_cpu_field( \
331						mib_statistic, \
332						c, stats_list[i].entry); \
333	} \
334}
335
336void inet_get_local_port_range(struct net *net, int *low, int *high);
337
338#ifdef CONFIG_SYSCTL
339static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
340{
341	if (!net->ipv4.sysctl_local_reserved_ports)
342		return false;
343	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
344}
345
346static inline bool sysctl_dev_name_is_allowed(const char *name)
347{
348	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
349}
350
351static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
352{
353	return port < net->ipv4.sysctl_ip_prot_sock;
354}
355
356#else
357static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
358{
359	return false;
360}
361
362static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
363{
364	return port < PROT_SOCK;
365}
366#endif
367
368__be32 inet_current_timestamp(void);
369
370/* From inetpeer.c */
371extern int inet_peer_threshold;
372extern int inet_peer_minttl;
373extern int inet_peer_maxttl;
374
375void ipfrag_init(void);
376
377void ip_static_sysctl_init(void);
378
379#define IP4_REPLY_MARK(net, mark) \
380	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
381
382static inline bool ip_is_fragment(const struct iphdr *iph)
383{
384	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
385}
386
387#ifdef CONFIG_INET
388#include <net/dst.h>
389
390/* The function in 2.2 was invalid, producing wrong result for
391 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
392static inline
393int ip_decrease_ttl(struct iphdr *iph)
394{
395	u32 check = (__force u32)iph->check;
396	check += (__force u32)htons(0x0100);
397	iph->check = (__force __sum16)(check + (check>=0xFFFF));
398	return --iph->ttl;
399}
400
401static inline int ip_mtu_locked(const struct dst_entry *dst)
402{
403	const struct rtable *rt = (const struct rtable *)dst;
404
405	return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
406}
407
408static inline
409int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
410{
411	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
412
413	return  pmtudisc == IP_PMTUDISC_DO ||
414		(pmtudisc == IP_PMTUDISC_WANT &&
415		 !ip_mtu_locked(dst));
416}
417
418static inline bool ip_sk_accept_pmtu(const struct sock *sk)
419{
420	return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
421	       inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
422}
423
424static inline bool ip_sk_use_pmtu(const struct sock *sk)
425{
426	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
427}
428
429static inline bool ip_sk_ignore_df(const struct sock *sk)
430{
431	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
432	       inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
433}
434
435static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
436						    bool forwarding)
437{
438	struct net *net = dev_net(dst->dev);
439	unsigned int mtu;
440
441	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
442	    ip_mtu_locked(dst) ||
443	    !forwarding)
444		return dst_mtu(dst);
445
446	/* 'forwarding = true' case should always honour route mtu */
447	mtu = dst_metric_raw(dst, RTAX_MTU);
448	if (mtu)
449		return mtu;
450
451	return min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);
452}
453
454static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
455					  const struct sk_buff *skb)
456{
 
 
457	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
458		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
459
460		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
461	}
462
463	return min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
 
464}
465
466struct dst_metrics *ip_fib_metrics_init(struct net *net, struct nlattr *fc_mx,
467					int fc_mx_len,
468					struct netlink_ext_ack *extack);
469static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
470{
471	if (fib_metrics != &dst_default_metrics &&
472	    refcount_dec_and_test(&fib_metrics->refcnt))
473		kfree(fib_metrics);
474}
475
476/* ipv4 and ipv6 both use refcounted metrics if it is not the default */
477static inline
478void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
479{
480	dst_init_metrics(dst, fib_metrics->metrics, true);
481
482	if (fib_metrics != &dst_default_metrics) {
483		dst->_metrics |= DST_METRICS_REFCOUNTED;
484		refcount_inc(&fib_metrics->refcnt);
485	}
486}
487
488static inline
489void ip_dst_metrics_put(struct dst_entry *dst)
490{
491	struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
492
493	if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
494		kfree(p);
495}
496
497u32 ip_idents_reserve(u32 hash, int segs);
498void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
499
500static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
501					struct sock *sk, int segs)
502{
503	struct iphdr *iph = ip_hdr(skb);
504
505	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
506		/* This is only to work around buggy Windows95/2000
507		 * VJ compression implementations.  If the ID field
508		 * does not change, they drop every other packet in
509		 * a TCP stream using header compression.
510		 */
511		if (sk && inet_sk(sk)->inet_daddr) {
512			iph->id = htons(inet_sk(sk)->inet_id);
513			inet_sk(sk)->inet_id += segs;
514		} else {
515			iph->id = 0;
516		}
517	} else {
518		__ip_select_ident(net, iph, segs);
519	}
520}
521
522static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
523				   struct sock *sk)
524{
525	ip_select_ident_segs(net, skb, sk, 1);
526}
527
528static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
529{
530	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
531				  skb->len, proto, 0);
532}
533
534/* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
535 * Equivalent to :	flow->v4addrs.src = iph->saddr;
536 *			flow->v4addrs.dst = iph->daddr;
537 */
538static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
539					    const struct iphdr *iph)
540{
541	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
542		     offsetof(typeof(flow->addrs), v4addrs.src) +
543			      sizeof(flow->addrs.v4addrs.src));
544	memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
545	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
546}
547
548static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
549{
550	const struct iphdr *iph = skb_gro_network_header(skb);
551
552	return csum_tcpudp_nofold(iph->saddr, iph->daddr,
553				  skb_gro_len(skb), proto, 0);
554}
555
556/*
557 *	Map a multicast IP onto multicast MAC for type ethernet.
558 */
559
560static inline void ip_eth_mc_map(__be32 naddr, char *buf)
561{
562	__u32 addr=ntohl(naddr);
563	buf[0]=0x01;
564	buf[1]=0x00;
565	buf[2]=0x5e;
566	buf[5]=addr&0xFF;
567	addr>>=8;
568	buf[4]=addr&0xFF;
569	addr>>=8;
570	buf[3]=addr&0x7F;
571}
572
573/*
574 *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
575 *	Leave P_Key as 0 to be filled in by driver.
576 */
577
578static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
579{
580	__u32 addr;
581	unsigned char scope = broadcast[5] & 0xF;
582
583	buf[0]  = 0;		/* Reserved */
584	buf[1]  = 0xff;		/* Multicast QPN */
585	buf[2]  = 0xff;
586	buf[3]  = 0xff;
587	addr    = ntohl(naddr);
588	buf[4]  = 0xff;
589	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
590	buf[6]  = 0x40;		/* IPv4 signature */
591	buf[7]  = 0x1b;
592	buf[8]  = broadcast[8];		/* P_Key */
593	buf[9]  = broadcast[9];
594	buf[10] = 0;
595	buf[11] = 0;
596	buf[12] = 0;
597	buf[13] = 0;
598	buf[14] = 0;
599	buf[15] = 0;
600	buf[19] = addr & 0xff;
601	addr  >>= 8;
602	buf[18] = addr & 0xff;
603	addr  >>= 8;
604	buf[17] = addr & 0xff;
605	addr  >>= 8;
606	buf[16] = addr & 0x0f;
607}
608
609static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
610{
611	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
612		memcpy(buf, broadcast, 4);
613	else
614		memcpy(buf, &naddr, sizeof(naddr));
615}
616
617#if IS_ENABLED(CONFIG_IPV6)
618#include <linux/ipv6.h>
619#endif
620
621static __inline__ void inet_reset_saddr(struct sock *sk)
622{
623	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
624#if IS_ENABLED(CONFIG_IPV6)
625	if (sk->sk_family == PF_INET6) {
626		struct ipv6_pinfo *np = inet6_sk(sk);
627
628		memset(&np->saddr, 0, sizeof(np->saddr));
629		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
630	}
631#endif
632}
633
634#endif
635
636static inline unsigned int ipv4_addr_hash(__be32 ip)
637{
638	return (__force unsigned int) ip;
639}
640
641static inline u32 ipv4_portaddr_hash(const struct net *net,
642				     __be32 saddr,
643				     unsigned int port)
644{
645	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
646}
647
648bool ip_call_ra_chain(struct sk_buff *skb);
649
650/*
651 *	Functions provided by ip_fragment.c
652 */
653
654enum ip_defrag_users {
655	IP_DEFRAG_LOCAL_DELIVER,
656	IP_DEFRAG_CALL_RA_CHAIN,
657	IP_DEFRAG_CONNTRACK_IN,
658	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
659	IP_DEFRAG_CONNTRACK_OUT,
660	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
661	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
662	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
663	IP_DEFRAG_VS_IN,
664	IP_DEFRAG_VS_OUT,
665	IP_DEFRAG_VS_FWD,
666	IP_DEFRAG_AF_PACKET,
667	IP_DEFRAG_MACVLAN,
668};
669
670/* Return true if the value of 'user' is between 'lower_bond'
671 * and 'upper_bond' inclusively.
672 */
673static inline bool ip_defrag_user_in_between(u32 user,
674					     enum ip_defrag_users lower_bond,
675					     enum ip_defrag_users upper_bond)
676{
677	return user >= lower_bond && user <= upper_bond;
678}
679
680int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
681#ifdef CONFIG_INET
682struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
683#else
684static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
685{
686	return skb;
687}
688#endif
689
690/*
691 *	Functions provided by ip_forward.c
692 */
693
694int ip_forward(struct sk_buff *skb);
695
696/*
697 *	Functions provided by ip_options.c
698 */
699
700void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
701		      __be32 daddr, struct rtable *rt, int is_frag);
702
703int __ip_options_echo(struct net *net, struct ip_options *dopt,
704		      struct sk_buff *skb, const struct ip_options *sopt);
705static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
706				  struct sk_buff *skb)
707{
708	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
709}
710
711void ip_options_fragment(struct sk_buff *skb);
712int __ip_options_compile(struct net *net, struct ip_options *opt,
713			 struct sk_buff *skb, __be32 *info);
714int ip_options_compile(struct net *net, struct ip_options *opt,
715		       struct sk_buff *skb);
716int ip_options_get(struct net *net, struct ip_options_rcu **optp,
717		   sockptr_t data, int optlen);
718void ip_options_undo(struct ip_options *opt);
719void ip_forward_options(struct sk_buff *skb);
720int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
721
722/*
723 *	Functions provided by ip_sockglue.c
724 */
725
726void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
727void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
728			 struct sk_buff *skb, int tlen, int offset);
729int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
730		 struct ipcm_cookie *ipc, bool allow_ipv6);
731int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
732		  unsigned int optlen);
733int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
734		  int __user *optlen);
735int ip_ra_control(struct sock *sk, unsigned char on,
736		  void (*destructor)(struct sock *));
737
738int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
739void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
740		   u32 info, u8 *payload);
741void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
742		    u32 info);
743
744static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
745{
746	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
747}
748
749bool icmp_global_allow(void);
750extern int sysctl_icmp_msgs_per_sec;
751extern int sysctl_icmp_msgs_burst;
752
753#ifdef CONFIG_PROC_FS
754int ip_misc_proc_init(void);
755#endif
756
757int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
758				struct netlink_ext_ack *extack);
759
760static inline bool inetdev_valid_mtu(unsigned int mtu)
761{
762	return likely(mtu >= IPV4_MIN_MTU);
763}
764
765void ip_sock_set_freebind(struct sock *sk);
766int ip_sock_set_mtu_discover(struct sock *sk, int val);
767void ip_sock_set_pktinfo(struct sock *sk);
768void ip_sock_set_recverr(struct sock *sk);
769void ip_sock_set_tos(struct sock *sk, int val);
770
771#endif	/* _IP_H */