Linux Audio

Check our new training course

Loading...
v5.4
  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
 27#include <net/inet_sock.h>
 28#include <net/route.h>
 29#include <net/snmp.h>
 30#include <net/flow.h>
 31#include <net/flow_dissector.h>
 32#include <net/netns/hash.h>
 33
 34#define IPV4_MAX_PMTU		65535U		/* RFC 2675, Section 5.1 */
 35#define IPV4_MIN_MTU		68			/* RFC 791 */
 36
 37extern unsigned int sysctl_fib_sync_mem;
 38extern unsigned int sysctl_fib_sync_mem_min;
 39extern unsigned int sysctl_fib_sync_mem_max;
 40
 41struct sock;
 42
 43struct inet_skb_parm {
 44	int			iif;
 45	struct ip_options	opt;		/* Compiled IP options		*/
 46	u16			flags;
 47
 48#define IPSKB_FORWARDED		BIT(0)
 49#define IPSKB_XFRM_TUNNEL_SIZE	BIT(1)
 50#define IPSKB_XFRM_TRANSFORMED	BIT(2)
 51#define IPSKB_FRAG_COMPLETE	BIT(3)
 52#define IPSKB_REROUTED		BIT(4)
 53#define IPSKB_DOREDIRECT	BIT(5)
 54#define IPSKB_FRAG_PMTU		BIT(6)
 55#define IPSKB_L3SLAVE		BIT(7)
 56
 57	u16			frag_max_size;
 58};
 59
 60static inline bool ipv4_l3mdev_skb(u16 flags)
 61{
 62	return !!(flags & IPSKB_L3SLAVE);
 63}
 64
 65static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
 66{
 67	return ip_hdr(skb)->ihl * 4;
 68}
 69
 70struct ipcm_cookie {
 71	struct sockcm_cookie	sockc;
 72	__be32			addr;
 73	int			oif;
 74	struct ip_options_rcu	*opt;
 
 75	__u8			ttl;
 76	__s16			tos;
 77	char			priority;
 78	__u16			gso_size;
 79};
 80
 81static inline void ipcm_init(struct ipcm_cookie *ipcm)
 82{
 83	*ipcm = (struct ipcm_cookie) { .tos = -1 };
 84}
 85
 86static inline void ipcm_init_sk(struct ipcm_cookie *ipcm,
 87				const struct inet_sock *inet)
 88{
 89	ipcm_init(ipcm);
 90
 91	ipcm->sockc.mark = inet->sk.sk_mark;
 92	ipcm->sockc.tsflags = inet->sk.sk_tsflags;
 93	ipcm->oif = inet->sk.sk_bound_dev_if;
 94	ipcm->addr = inet->inet_saddr;
 95}
 96
 97#define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
 98#define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
 99
100/* return enslaved device index if relevant */
101static inline int inet_sdif(struct sk_buff *skb)
102{
103#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
104	if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
105		return IPCB(skb)->iif;
106#endif
107	return 0;
108}
109
110/* Special input handler for packets caught by router alert option.
111   They are selected only by protocol field, and then processed likely
112   local ones; but only if someone wants them! Otherwise, router
113   not running rsvpd will kill RSVP.
114
115   It is user level problem, what it will make with them.
116   I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
117   but receiver should be enough clever f.e. to forward mtrace requests,
118   sent to multicast group to reach destination designated router.
119 */
120
121struct ip_ra_chain {
122	struct ip_ra_chain __rcu *next;
123	struct sock		*sk;
124	union {
125		void			(*destructor)(struct sock *);
126		struct sock		*saved_sk;
127	};
128	struct rcu_head		rcu;
129};
130
131/* IP flags. */
132#define IP_CE		0x8000		/* Flag: "Congestion"		*/
133#define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
134#define IP_MF		0x2000		/* Flag: "More Fragments"	*/
135#define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
136
137#define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
138
139struct msghdr;
140struct net_device;
141struct packet_type;
142struct rtable;
143struct sockaddr;
144
145int igmp_mc_init(void);
146
147/*
148 *	Functions provided by ip.c
149 */
150
151int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
152			  __be32 saddr, __be32 daddr,
153			  struct ip_options_rcu *opt);
154int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
155	   struct net_device *orig_dev);
156void ip_list_rcv(struct list_head *head, struct packet_type *pt,
157		 struct net_device *orig_dev);
158int ip_local_deliver(struct sk_buff *skb);
159void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int proto);
160int ip_mr_input(struct sk_buff *skb);
161int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
162int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
163int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
164		   int (*output)(struct net *, struct sock *, struct sk_buff *));
165
166struct ip_fraglist_iter {
167	struct sk_buff	*frag;
168	struct iphdr	*iph;
169	int		offset;
170	unsigned int	hlen;
171};
172
173void ip_fraglist_init(struct sk_buff *skb, struct iphdr *iph,
174		      unsigned int hlen, struct ip_fraglist_iter *iter);
175void ip_fraglist_prepare(struct sk_buff *skb, struct ip_fraglist_iter *iter);
176
177static inline struct sk_buff *ip_fraglist_next(struct ip_fraglist_iter *iter)
178{
179	struct sk_buff *skb = iter->frag;
180
181	iter->frag = skb->next;
182	skb_mark_not_on_list(skb);
183
184	return skb;
185}
186
187struct ip_frag_state {
188	bool		DF;
189	unsigned int	hlen;
190	unsigned int	ll_rs;
191	unsigned int	mtu;
192	unsigned int	left;
193	int		offset;
194	int		ptr;
195	__be16		not_last_frag;
196};
197
198void ip_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int ll_rs,
199		  unsigned int mtu, bool DF, struct ip_frag_state *state);
200struct sk_buff *ip_frag_next(struct sk_buff *skb,
201			     struct ip_frag_state *state);
202
203void ip_send_check(struct iphdr *ip);
204int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
205int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
206
207int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
208		    __u8 tos);
209void ip_init(void);
210int ip_append_data(struct sock *sk, struct flowi4 *fl4,
211		   int getfrag(void *from, char *to, int offset, int len,
212			       int odd, struct sk_buff *skb),
213		   void *from, int len, int protolen,
214		   struct ipcm_cookie *ipc,
215		   struct rtable **rt,
216		   unsigned int flags);
217int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
218		       struct sk_buff *skb);
219ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
220		       int offset, size_t size, int flags);
221struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
222			      struct sk_buff_head *queue,
223			      struct inet_cork *cork);
224int ip_send_skb(struct net *net, struct sk_buff *skb);
225int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
226void ip_flush_pending_frames(struct sock *sk);
227struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
228			    int getfrag(void *from, char *to, int offset,
229					int len, int odd, struct sk_buff *skb),
230			    void *from, int length, int transhdrlen,
231			    struct ipcm_cookie *ipc, struct rtable **rtp,
232			    struct inet_cork *cork, unsigned int flags);
233
234static inline int ip_queue_xmit(struct sock *sk, struct sk_buff *skb,
235				struct flowi *fl)
236{
237	return __ip_queue_xmit(sk, skb, fl, inet_sk(sk)->tos);
238}
239
240static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
241{
242	return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
243}
244
245static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
246{
247	return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
248}
249
250static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
251{
252	return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
253}
254
255/* datagram.c */
256int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
257int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
258
259void ip4_datagram_release_cb(struct sock *sk);
260
261struct ip_reply_arg {
262	struct kvec iov[1];
263	int	    flags;
264	__wsum 	    csum;
265	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
266				/* -1 if not needed */
267	int	    bound_dev_if;
268	u8  	    tos;
269	kuid_t	    uid;
270};
271
272#define IP_REPLY_ARG_NOSRCCHECK 1
273
274static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
275{
276	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
277}
278
279void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
280			   const struct ip_options *sopt,
281			   __be32 daddr, __be32 saddr,
282			   const struct ip_reply_arg *arg,
283			   unsigned int len, u64 transmit_time);
284
285#define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
286#define __IP_INC_STATS(net, field)	__SNMP_INC_STATS64((net)->mib.ip_statistics, field)
287#define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
288#define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
289#define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
290#define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
291#define NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
292#define __NET_INC_STATS(net, field)	__SNMP_INC_STATS((net)->mib.net_statistics, field)
293#define NET_ADD_STATS(net, field, adnd)	SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
294#define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
295
296u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
297unsigned long snmp_fold_field(void __percpu *mib, int offt);
298#if BITS_PER_LONG==32
299u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
300			 size_t syncp_offset);
301u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
302#else
303static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
304					size_t syncp_offset)
305{
306	return snmp_get_cpu_field(mib, cpu, offct);
307
308}
309
310static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
311{
312	return snmp_fold_field(mib, offt);
313}
314#endif
315
316#define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
317{ \
318	int i, c; \
319	for_each_possible_cpu(c) { \
320		for (i = 0; stats_list[i].name; i++) \
321			buff64[i] += snmp_get_cpu_field64( \
322					mib_statistic, \
323					c, stats_list[i].entry, \
324					offset); \
325	} \
326}
327
328#define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
329{ \
330	int i, c; \
331	for_each_possible_cpu(c) { \
332		for (i = 0; stats_list[i].name; i++) \
333			buff[i] += snmp_get_cpu_field( \
334						mib_statistic, \
335						c, stats_list[i].entry); \
336	} \
337}
338
339void inet_get_local_port_range(struct net *net, int *low, int *high);
340
341#ifdef CONFIG_SYSCTL
342static inline int inet_is_local_reserved_port(struct net *net, int port)
343{
344	if (!net->ipv4.sysctl_local_reserved_ports)
345		return 0;
346	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
347}
348
349static inline bool sysctl_dev_name_is_allowed(const char *name)
350{
351	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
352}
353
354static inline int inet_prot_sock(struct net *net)
355{
356	return net->ipv4.sysctl_ip_prot_sock;
357}
358
359#else
360static inline int inet_is_local_reserved_port(struct net *net, int port)
361{
362	return 0;
363}
364
365static inline int inet_prot_sock(struct net *net)
366{
367	return PROT_SOCK;
368}
369#endif
370
371__be32 inet_current_timestamp(void);
372
373/* From inetpeer.c */
374extern int inet_peer_threshold;
375extern int inet_peer_minttl;
376extern int inet_peer_maxttl;
377
378void ipfrag_init(void);
379
380void ip_static_sysctl_init(void);
381
382#define IP4_REPLY_MARK(net, mark) \
383	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
384
385static inline bool ip_is_fragment(const struct iphdr *iph)
386{
387	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
388}
389
390#ifdef CONFIG_INET
391#include <net/dst.h>
392
393/* The function in 2.2 was invalid, producing wrong result for
394 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
395static inline
396int ip_decrease_ttl(struct iphdr *iph)
397{
398	u32 check = (__force u32)iph->check;
399	check += (__force u32)htons(0x0100);
400	iph->check = (__force __sum16)(check + (check>=0xFFFF));
401	return --iph->ttl;
402}
403
404static inline int ip_mtu_locked(const struct dst_entry *dst)
405{
406	const struct rtable *rt = (const struct rtable *)dst;
407
408	return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
409}
410
411static inline
412int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
413{
414	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
415
416	return  pmtudisc == IP_PMTUDISC_DO ||
417		(pmtudisc == IP_PMTUDISC_WANT &&
418		 !ip_mtu_locked(dst));
419}
420
421static inline bool ip_sk_accept_pmtu(const struct sock *sk)
422{
423	return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
424	       inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
425}
426
427static inline bool ip_sk_use_pmtu(const struct sock *sk)
428{
429	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
430}
431
432static inline bool ip_sk_ignore_df(const struct sock *sk)
433{
434	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
435	       inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
436}
437
438static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
439						    bool forwarding)
440{
441	struct net *net = dev_net(dst->dev);
442
443	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
444	    ip_mtu_locked(dst) ||
445	    !forwarding)
446		return dst_mtu(dst);
447
448	return min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);
449}
450
451static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
452					  const struct sk_buff *skb)
453{
454	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
455		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
456
457		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
458	}
459
460	return min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
461}
462
463struct dst_metrics *ip_fib_metrics_init(struct net *net, struct nlattr *fc_mx,
464					int fc_mx_len,
465					struct netlink_ext_ack *extack);
466static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
467{
468	if (fib_metrics != &dst_default_metrics &&
469	    refcount_dec_and_test(&fib_metrics->refcnt))
470		kfree(fib_metrics);
471}
472
473/* ipv4 and ipv6 both use refcounted metrics if it is not the default */
474static inline
475void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
476{
477	dst_init_metrics(dst, fib_metrics->metrics, true);
478
479	if (fib_metrics != &dst_default_metrics) {
480		dst->_metrics |= DST_METRICS_REFCOUNTED;
481		refcount_inc(&fib_metrics->refcnt);
482	}
483}
484
485static inline
486void ip_dst_metrics_put(struct dst_entry *dst)
487{
488	struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
489
490	if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
491		kfree(p);
492}
493
494u32 ip_idents_reserve(u32 hash, int segs);
495void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
496
497static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
498					struct sock *sk, int segs)
499{
500	struct iphdr *iph = ip_hdr(skb);
501
502	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
503		/* This is only to work around buggy Windows95/2000
504		 * VJ compression implementations.  If the ID field
505		 * does not change, they drop every other packet in
506		 * a TCP stream using header compression.
507		 */
508		if (sk && inet_sk(sk)->inet_daddr) {
509			iph->id = htons(inet_sk(sk)->inet_id);
510			inet_sk(sk)->inet_id += segs;
511		} else {
512			iph->id = 0;
513		}
514	} else {
515		__ip_select_ident(net, iph, segs);
516	}
517}
518
519static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
520				   struct sock *sk)
521{
522	ip_select_ident_segs(net, skb, sk, 1);
523}
524
525static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
526{
527	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
528				  skb->len, proto, 0);
529}
530
531/* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
532 * Equivalent to :	flow->v4addrs.src = iph->saddr;
533 *			flow->v4addrs.dst = iph->daddr;
534 */
535static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
536					    const struct iphdr *iph)
537{
538	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
539		     offsetof(typeof(flow->addrs), v4addrs.src) +
540			      sizeof(flow->addrs.v4addrs.src));
541	memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
542	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
543}
544
545static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
546{
547	const struct iphdr *iph = skb_gro_network_header(skb);
548
549	return csum_tcpudp_nofold(iph->saddr, iph->daddr,
550				  skb_gro_len(skb), proto, 0);
551}
552
553/*
554 *	Map a multicast IP onto multicast MAC for type ethernet.
555 */
556
557static inline void ip_eth_mc_map(__be32 naddr, char *buf)
558{
559	__u32 addr=ntohl(naddr);
560	buf[0]=0x01;
561	buf[1]=0x00;
562	buf[2]=0x5e;
563	buf[5]=addr&0xFF;
564	addr>>=8;
565	buf[4]=addr&0xFF;
566	addr>>=8;
567	buf[3]=addr&0x7F;
568}
569
570/*
571 *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
572 *	Leave P_Key as 0 to be filled in by driver.
573 */
574
575static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
576{
577	__u32 addr;
578	unsigned char scope = broadcast[5] & 0xF;
579
580	buf[0]  = 0;		/* Reserved */
581	buf[1]  = 0xff;		/* Multicast QPN */
582	buf[2]  = 0xff;
583	buf[3]  = 0xff;
584	addr    = ntohl(naddr);
585	buf[4]  = 0xff;
586	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
587	buf[6]  = 0x40;		/* IPv4 signature */
588	buf[7]  = 0x1b;
589	buf[8]  = broadcast[8];		/* P_Key */
590	buf[9]  = broadcast[9];
591	buf[10] = 0;
592	buf[11] = 0;
593	buf[12] = 0;
594	buf[13] = 0;
595	buf[14] = 0;
596	buf[15] = 0;
597	buf[19] = addr & 0xff;
598	addr  >>= 8;
599	buf[18] = addr & 0xff;
600	addr  >>= 8;
601	buf[17] = addr & 0xff;
602	addr  >>= 8;
603	buf[16] = addr & 0x0f;
604}
605
606static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
607{
608	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
609		memcpy(buf, broadcast, 4);
610	else
611		memcpy(buf, &naddr, sizeof(naddr));
612}
613
614#if IS_ENABLED(CONFIG_IPV6)
615#include <linux/ipv6.h>
616#endif
617
618static __inline__ void inet_reset_saddr(struct sock *sk)
619{
620	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
621#if IS_ENABLED(CONFIG_IPV6)
622	if (sk->sk_family == PF_INET6) {
623		struct ipv6_pinfo *np = inet6_sk(sk);
624
625		memset(&np->saddr, 0, sizeof(np->saddr));
626		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
627	}
628#endif
629}
630
631#endif
632
633static inline unsigned int ipv4_addr_hash(__be32 ip)
634{
635	return (__force unsigned int) ip;
636}
637
638static inline u32 ipv4_portaddr_hash(const struct net *net,
639				     __be32 saddr,
640				     unsigned int port)
641{
642	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
643}
644
645bool ip_call_ra_chain(struct sk_buff *skb);
646
647/*
648 *	Functions provided by ip_fragment.c
649 */
650
651enum ip_defrag_users {
652	IP_DEFRAG_LOCAL_DELIVER,
653	IP_DEFRAG_CALL_RA_CHAIN,
654	IP_DEFRAG_CONNTRACK_IN,
655	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
656	IP_DEFRAG_CONNTRACK_OUT,
657	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
658	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
659	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
660	IP_DEFRAG_VS_IN,
661	IP_DEFRAG_VS_OUT,
662	IP_DEFRAG_VS_FWD,
663	IP_DEFRAG_AF_PACKET,
664	IP_DEFRAG_MACVLAN,
665};
666
667/* Return true if the value of 'user' is between 'lower_bond'
668 * and 'upper_bond' inclusively.
669 */
670static inline bool ip_defrag_user_in_between(u32 user,
671					     enum ip_defrag_users lower_bond,
672					     enum ip_defrag_users upper_bond)
673{
674	return user >= lower_bond && user <= upper_bond;
675}
676
677int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
678#ifdef CONFIG_INET
679struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
680#else
681static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
682{
683	return skb;
684}
685#endif
686
687/*
688 *	Functions provided by ip_forward.c
689 */
690
691int ip_forward(struct sk_buff *skb);
692
693/*
694 *	Functions provided by ip_options.c
695 */
696
697void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
698		      __be32 daddr, struct rtable *rt, int is_frag);
699
700int __ip_options_echo(struct net *net, struct ip_options *dopt,
701		      struct sk_buff *skb, const struct ip_options *sopt);
702static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
703				  struct sk_buff *skb)
704{
705	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
706}
707
708void ip_options_fragment(struct sk_buff *skb);
709int __ip_options_compile(struct net *net, struct ip_options *opt,
710			 struct sk_buff *skb, __be32 *info);
711int ip_options_compile(struct net *net, struct ip_options *opt,
712		       struct sk_buff *skb);
713int ip_options_get(struct net *net, struct ip_options_rcu **optp,
714		   unsigned char *data, int optlen);
715int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
716			     unsigned char __user *data, int optlen);
717void ip_options_undo(struct ip_options *opt);
718void ip_forward_options(struct sk_buff *skb);
719int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
720
721/*
722 *	Functions provided by ip_sockglue.c
723 */
724
725void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
726void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
727			 struct sk_buff *skb, int tlen, int offset);
728int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
729		 struct ipcm_cookie *ipc, bool allow_ipv6);
730int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
731		  unsigned int optlen);
732int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
733		  int __user *optlen);
734int compat_ip_setsockopt(struct sock *sk, int level, int optname,
735			 char __user *optval, unsigned int optlen);
736int compat_ip_getsockopt(struct sock *sk, int level, int optname,
737			 char __user *optval, int __user *optlen);
738int ip_ra_control(struct sock *sk, unsigned char on,
739		  void (*destructor)(struct sock *));
740
741int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
742void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
743		   u32 info, u8 *payload);
744void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
745		    u32 info);
746
747static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
748{
749	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
750}
751
752bool icmp_global_allow(void);
753extern int sysctl_icmp_msgs_per_sec;
754extern int sysctl_icmp_msgs_burst;
755
756#ifdef CONFIG_PROC_FS
757int ip_misc_proc_init(void);
758#endif
759
760int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
761				struct netlink_ext_ack *extack);
762
763#endif	/* _IP_H */
v4.17
 
  1/*
  2 * INET		An implementation of the TCP/IP protocol suite for the LINUX
  3 *		operating system.  INET is implemented using the  BSD Socket
  4 *		interface as the means of communication with the user level.
  5 *
  6 *		Definitions for the IP module.
  7 *
  8 * Version:	@(#)ip.h	1.0.2	05/07/93
  9 *
 10 * Authors:	Ross Biro
 11 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 12 *		Alan Cox, <gw4pts@gw4pts.ampr.org>
 13 *
 14 * Changes:
 15 *		Mike McLagan    :       Routing by source
 16 *
 17 *		This program is free software; you can redistribute it and/or
 18 *		modify it under the terms of the GNU General Public License
 19 *		as published by the Free Software Foundation; either version
 20 *		2 of the License, or (at your option) any later version.
 21 */
 22#ifndef _IP_H
 23#define _IP_H
 24
 25#include <linux/types.h>
 26#include <linux/ip.h>
 27#include <linux/in.h>
 28#include <linux/skbuff.h>
 29#include <linux/jhash.h>
 30
 31#include <net/inet_sock.h>
 32#include <net/route.h>
 33#include <net/snmp.h>
 34#include <net/flow.h>
 35#include <net/flow_dissector.h>
 36#include <net/netns/hash.h>
 37
 38#define IPV4_MAX_PMTU		65535U		/* RFC 2675, Section 5.1 */
 39#define IPV4_MIN_MTU		68			/* RFC 791 */
 40
 
 
 
 
 41struct sock;
 42
 43struct inet_skb_parm {
 44	int			iif;
 45	struct ip_options	opt;		/* Compiled IP options		*/
 46	u16			flags;
 47
 48#define IPSKB_FORWARDED		BIT(0)
 49#define IPSKB_XFRM_TUNNEL_SIZE	BIT(1)
 50#define IPSKB_XFRM_TRANSFORMED	BIT(2)
 51#define IPSKB_FRAG_COMPLETE	BIT(3)
 52#define IPSKB_REROUTED		BIT(4)
 53#define IPSKB_DOREDIRECT	BIT(5)
 54#define IPSKB_FRAG_PMTU		BIT(6)
 55#define IPSKB_L3SLAVE		BIT(7)
 56
 57	u16			frag_max_size;
 58};
 59
 60static inline bool ipv4_l3mdev_skb(u16 flags)
 61{
 62	return !!(flags & IPSKB_L3SLAVE);
 63}
 64
 65static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
 66{
 67	return ip_hdr(skb)->ihl * 4;
 68}
 69
 70struct ipcm_cookie {
 71	struct sockcm_cookie	sockc;
 72	__be32			addr;
 73	int			oif;
 74	struct ip_options_rcu	*opt;
 75	__u8			tx_flags;
 76	__u8			ttl;
 77	__s16			tos;
 78	char			priority;
 
 79};
 80
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 81#define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
 82#define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
 83
 84/* return enslaved device index if relevant */
 85static inline int inet_sdif(struct sk_buff *skb)
 86{
 87#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
 88	if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
 89		return IPCB(skb)->iif;
 90#endif
 91	return 0;
 92}
 93
 94/* Special input handler for packets caught by router alert option.
 95   They are selected only by protocol field, and then processed likely
 96   local ones; but only if someone wants them! Otherwise, router
 97   not running rsvpd will kill RSVP.
 98
 99   It is user level problem, what it will make with them.
100   I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
101   but receiver should be enough clever f.e. to forward mtrace requests,
102   sent to multicast group to reach destination designated router.
103 */
104
105struct ip_ra_chain {
106	struct ip_ra_chain __rcu *next;
107	struct sock		*sk;
108	union {
109		void			(*destructor)(struct sock *);
110		struct sock		*saved_sk;
111	};
112	struct rcu_head		rcu;
113};
114
115/* IP flags. */
116#define IP_CE		0x8000		/* Flag: "Congestion"		*/
117#define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
118#define IP_MF		0x2000		/* Flag: "More Fragments"	*/
119#define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
120
121#define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
122
123struct msghdr;
124struct net_device;
125struct packet_type;
126struct rtable;
127struct sockaddr;
128
129int igmp_mc_init(void);
130
131/*
132 *	Functions provided by ip.c
133 */
134
135int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
136			  __be32 saddr, __be32 daddr,
137			  struct ip_options_rcu *opt);
138int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
139	   struct net_device *orig_dev);
 
 
140int ip_local_deliver(struct sk_buff *skb);
 
141int ip_mr_input(struct sk_buff *skb);
142int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
143int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
144int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
145		   int (*output)(struct net *, struct sock *, struct sk_buff *));
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
146void ip_send_check(struct iphdr *ip);
147int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
148int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
149
150int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
 
151void ip_init(void);
152int ip_append_data(struct sock *sk, struct flowi4 *fl4,
153		   int getfrag(void *from, char *to, int offset, int len,
154			       int odd, struct sk_buff *skb),
155		   void *from, int len, int protolen,
156		   struct ipcm_cookie *ipc,
157		   struct rtable **rt,
158		   unsigned int flags);
159int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
160		       struct sk_buff *skb);
161ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
162		       int offset, size_t size, int flags);
163struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
164			      struct sk_buff_head *queue,
165			      struct inet_cork *cork);
166int ip_send_skb(struct net *net, struct sk_buff *skb);
167int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
168void ip_flush_pending_frames(struct sock *sk);
169struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
170			    int getfrag(void *from, char *to, int offset,
171					int len, int odd, struct sk_buff *skb),
172			    void *from, int length, int transhdrlen,
173			    struct ipcm_cookie *ipc, struct rtable **rtp,
174			    unsigned int flags);
 
 
 
 
 
 
175
176static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
177{
178	return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
179}
180
181static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
182{
183	return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
184}
185
186static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
187{
188	return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
189}
190
191/* datagram.c */
192int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
193int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
194
195void ip4_datagram_release_cb(struct sock *sk);
196
197struct ip_reply_arg {
198	struct kvec iov[1];
199	int	    flags;
200	__wsum 	    csum;
201	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
202				/* -1 if not needed */
203	int	    bound_dev_if;
204	u8  	    tos;
205	kuid_t	    uid;
206};
207
208#define IP_REPLY_ARG_NOSRCCHECK 1
209
210static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
211{
212	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
213}
214
215void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
216			   const struct ip_options *sopt,
217			   __be32 daddr, __be32 saddr,
218			   const struct ip_reply_arg *arg,
219			   unsigned int len);
220
221#define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
222#define __IP_INC_STATS(net, field)	__SNMP_INC_STATS64((net)->mib.ip_statistics, field)
223#define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
224#define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
225#define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
226#define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
227#define NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
228#define __NET_INC_STATS(net, field)	__SNMP_INC_STATS((net)->mib.net_statistics, field)
229#define NET_ADD_STATS(net, field, adnd)	SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
230#define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
231
232u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
233unsigned long snmp_fold_field(void __percpu *mib, int offt);
234#if BITS_PER_LONG==32
235u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
236			 size_t syncp_offset);
237u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
238#else
239static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
240					size_t syncp_offset)
241{
242	return snmp_get_cpu_field(mib, cpu, offct);
243
244}
245
246static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
247{
248	return snmp_fold_field(mib, offt);
249}
250#endif
251
252#define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
253{ \
254	int i, c; \
255	for_each_possible_cpu(c) { \
256		for (i = 0; stats_list[i].name; i++) \
257			buff64[i] += snmp_get_cpu_field64( \
258					mib_statistic, \
259					c, stats_list[i].entry, \
260					offset); \
261	} \
262}
263
264#define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
265{ \
266	int i, c; \
267	for_each_possible_cpu(c) { \
268		for (i = 0; stats_list[i].name; i++) \
269			buff[i] += snmp_get_cpu_field( \
270						mib_statistic, \
271						c, stats_list[i].entry); \
272	} \
273}
274
275void inet_get_local_port_range(struct net *net, int *low, int *high);
276
277#ifdef CONFIG_SYSCTL
278static inline int inet_is_local_reserved_port(struct net *net, int port)
279{
280	if (!net->ipv4.sysctl_local_reserved_ports)
281		return 0;
282	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
283}
284
285static inline bool sysctl_dev_name_is_allowed(const char *name)
286{
287	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
288}
289
290static inline int inet_prot_sock(struct net *net)
291{
292	return net->ipv4.sysctl_ip_prot_sock;
293}
294
295#else
296static inline int inet_is_local_reserved_port(struct net *net, int port)
297{
298	return 0;
299}
300
301static inline int inet_prot_sock(struct net *net)
302{
303	return PROT_SOCK;
304}
305#endif
306
307__be32 inet_current_timestamp(void);
308
309/* From inetpeer.c */
310extern int inet_peer_threshold;
311extern int inet_peer_minttl;
312extern int inet_peer_maxttl;
313
314void ipfrag_init(void);
315
316void ip_static_sysctl_init(void);
317
318#define IP4_REPLY_MARK(net, mark) \
319	((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
320
321static inline bool ip_is_fragment(const struct iphdr *iph)
322{
323	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
324}
325
326#ifdef CONFIG_INET
327#include <net/dst.h>
328
329/* The function in 2.2 was invalid, producing wrong result for
330 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
331static inline
332int ip_decrease_ttl(struct iphdr *iph)
333{
334	u32 check = (__force u32)iph->check;
335	check += (__force u32)htons(0x0100);
336	iph->check = (__force __sum16)(check + (check>=0xFFFF));
337	return --iph->ttl;
338}
339
340static inline int ip_mtu_locked(const struct dst_entry *dst)
341{
342	const struct rtable *rt = (const struct rtable *)dst;
343
344	return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
345}
346
347static inline
348int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
349{
350	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
351
352	return  pmtudisc == IP_PMTUDISC_DO ||
353		(pmtudisc == IP_PMTUDISC_WANT &&
354		 !ip_mtu_locked(dst));
355}
356
357static inline bool ip_sk_accept_pmtu(const struct sock *sk)
358{
359	return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
360	       inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
361}
362
363static inline bool ip_sk_use_pmtu(const struct sock *sk)
364{
365	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
366}
367
368static inline bool ip_sk_ignore_df(const struct sock *sk)
369{
370	return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
371	       inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
372}
373
374static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
375						    bool forwarding)
376{
377	struct net *net = dev_net(dst->dev);
378
379	if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
380	    ip_mtu_locked(dst) ||
381	    !forwarding)
382		return dst_mtu(dst);
383
384	return min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);
385}
386
387static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
388					  const struct sk_buff *skb)
389{
390	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
391		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
392
393		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
394	}
395
396	return min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
397}
398
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
399u32 ip_idents_reserve(u32 hash, int segs);
400void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
401
402static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
403					struct sock *sk, int segs)
404{
405	struct iphdr *iph = ip_hdr(skb);
406
407	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
408		/* This is only to work around buggy Windows95/2000
409		 * VJ compression implementations.  If the ID field
410		 * does not change, they drop every other packet in
411		 * a TCP stream using header compression.
412		 */
413		if (sk && inet_sk(sk)->inet_daddr) {
414			iph->id = htons(inet_sk(sk)->inet_id);
415			inet_sk(sk)->inet_id += segs;
416		} else {
417			iph->id = 0;
418		}
419	} else {
420		__ip_select_ident(net, iph, segs);
421	}
422}
423
424static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
425				   struct sock *sk)
426{
427	ip_select_ident_segs(net, skb, sk, 1);
428}
429
430static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
431{
432	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
433				  skb->len, proto, 0);
434}
435
436/* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
437 * Equivalent to :	flow->v4addrs.src = iph->saddr;
438 *			flow->v4addrs.dst = iph->daddr;
439 */
440static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
441					    const struct iphdr *iph)
442{
443	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
444		     offsetof(typeof(flow->addrs), v4addrs.src) +
445			      sizeof(flow->addrs.v4addrs.src));
446	memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
447	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
448}
449
450static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
451{
452	const struct iphdr *iph = skb_gro_network_header(skb);
453
454	return csum_tcpudp_nofold(iph->saddr, iph->daddr,
455				  skb_gro_len(skb), proto, 0);
456}
457
458/*
459 *	Map a multicast IP onto multicast MAC for type ethernet.
460 */
461
462static inline void ip_eth_mc_map(__be32 naddr, char *buf)
463{
464	__u32 addr=ntohl(naddr);
465	buf[0]=0x01;
466	buf[1]=0x00;
467	buf[2]=0x5e;
468	buf[5]=addr&0xFF;
469	addr>>=8;
470	buf[4]=addr&0xFF;
471	addr>>=8;
472	buf[3]=addr&0x7F;
473}
474
475/*
476 *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
477 *	Leave P_Key as 0 to be filled in by driver.
478 */
479
480static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
481{
482	__u32 addr;
483	unsigned char scope = broadcast[5] & 0xF;
484
485	buf[0]  = 0;		/* Reserved */
486	buf[1]  = 0xff;		/* Multicast QPN */
487	buf[2]  = 0xff;
488	buf[3]  = 0xff;
489	addr    = ntohl(naddr);
490	buf[4]  = 0xff;
491	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
492	buf[6]  = 0x40;		/* IPv4 signature */
493	buf[7]  = 0x1b;
494	buf[8]  = broadcast[8];		/* P_Key */
495	buf[9]  = broadcast[9];
496	buf[10] = 0;
497	buf[11] = 0;
498	buf[12] = 0;
499	buf[13] = 0;
500	buf[14] = 0;
501	buf[15] = 0;
502	buf[19] = addr & 0xff;
503	addr  >>= 8;
504	buf[18] = addr & 0xff;
505	addr  >>= 8;
506	buf[17] = addr & 0xff;
507	addr  >>= 8;
508	buf[16] = addr & 0x0f;
509}
510
511static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
512{
513	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
514		memcpy(buf, broadcast, 4);
515	else
516		memcpy(buf, &naddr, sizeof(naddr));
517}
518
519#if IS_ENABLED(CONFIG_IPV6)
520#include <linux/ipv6.h>
521#endif
522
523static __inline__ void inet_reset_saddr(struct sock *sk)
524{
525	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
526#if IS_ENABLED(CONFIG_IPV6)
527	if (sk->sk_family == PF_INET6) {
528		struct ipv6_pinfo *np = inet6_sk(sk);
529
530		memset(&np->saddr, 0, sizeof(np->saddr));
531		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
532	}
533#endif
534}
535
536#endif
537
538static inline unsigned int ipv4_addr_hash(__be32 ip)
539{
540	return (__force unsigned int) ip;
541}
542
543static inline u32 ipv4_portaddr_hash(const struct net *net,
544				     __be32 saddr,
545				     unsigned int port)
546{
547	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
548}
549
550bool ip_call_ra_chain(struct sk_buff *skb);
551
552/*
553 *	Functions provided by ip_fragment.c
554 */
555
556enum ip_defrag_users {
557	IP_DEFRAG_LOCAL_DELIVER,
558	IP_DEFRAG_CALL_RA_CHAIN,
559	IP_DEFRAG_CONNTRACK_IN,
560	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
561	IP_DEFRAG_CONNTRACK_OUT,
562	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
563	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
564	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
565	IP_DEFRAG_VS_IN,
566	IP_DEFRAG_VS_OUT,
567	IP_DEFRAG_VS_FWD,
568	IP_DEFRAG_AF_PACKET,
569	IP_DEFRAG_MACVLAN,
570};
571
572/* Return true if the value of 'user' is between 'lower_bond'
573 * and 'upper_bond' inclusively.
574 */
575static inline bool ip_defrag_user_in_between(u32 user,
576					     enum ip_defrag_users lower_bond,
577					     enum ip_defrag_users upper_bond)
578{
579	return user >= lower_bond && user <= upper_bond;
580}
581
582int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
583#ifdef CONFIG_INET
584struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
585#else
586static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
587{
588	return skb;
589}
590#endif
591
592/*
593 *	Functions provided by ip_forward.c
594 */
595
596int ip_forward(struct sk_buff *skb);
597
598/*
599 *	Functions provided by ip_options.c
600 */
601
602void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
603		      __be32 daddr, struct rtable *rt, int is_frag);
604
605int __ip_options_echo(struct net *net, struct ip_options *dopt,
606		      struct sk_buff *skb, const struct ip_options *sopt);
607static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
608				  struct sk_buff *skb)
609{
610	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
611}
612
613void ip_options_fragment(struct sk_buff *skb);
 
 
614int ip_options_compile(struct net *net, struct ip_options *opt,
615		       struct sk_buff *skb);
616int ip_options_get(struct net *net, struct ip_options_rcu **optp,
617		   unsigned char *data, int optlen);
618int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
619			     unsigned char __user *data, int optlen);
620void ip_options_undo(struct ip_options *opt);
621void ip_forward_options(struct sk_buff *skb);
622int ip_options_rcv_srr(struct sk_buff *skb);
623
624/*
625 *	Functions provided by ip_sockglue.c
626 */
627
628void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
629void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
630			 struct sk_buff *skb, int tlen, int offset);
631int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
632		 struct ipcm_cookie *ipc, bool allow_ipv6);
633int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
634		  unsigned int optlen);
635int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
636		  int __user *optlen);
637int compat_ip_setsockopt(struct sock *sk, int level, int optname,
638			 char __user *optval, unsigned int optlen);
639int compat_ip_getsockopt(struct sock *sk, int level, int optname,
640			 char __user *optval, int __user *optlen);
641int ip_ra_control(struct sock *sk, unsigned char on,
642		  void (*destructor)(struct sock *));
643
644int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
645void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
646		   u32 info, u8 *payload);
647void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
648		    u32 info);
649
650static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
651{
652	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
653}
654
655bool icmp_global_allow(void);
656extern int sysctl_icmp_msgs_per_sec;
657extern int sysctl_icmp_msgs_burst;
658
659#ifdef CONFIG_PROC_FS
660int ip_misc_proc_init(void);
661#endif
 
 
 
662
663#endif	/* _IP_H */