xref: /linux/include/net/ip.h (revision 3fd6c59042dbba50391e30862beac979491145fe)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * INET		An implementation of the TCP/IP protocol suite for the LINUX
4  *		operating system.  INET is implemented using the  BSD Socket
5  *		interface as the means of communication with the user level.
6  *
7  *		Definitions for the IP module.
8  *
9  * Version:	@(#)ip.h	1.0.2	05/07/93
10  *
11  * Authors:	Ross Biro
12  *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
13  *		Alan Cox, <gw4pts@gw4pts.ampr.org>
14  *
15  * Changes:
16  *		Mike McLagan    :       Routing by source
17  */
18 #ifndef _IP_H
19 #define _IP_H
20 
21 #include <linux/types.h>
22 #include <linux/ip.h>
23 #include <linux/in.h>
24 #include <linux/skbuff.h>
25 #include <linux/jhash.h>
26 #include <linux/sockptr.h>
27 #include <linux/static_key.h>
28 
29 #include <net/inet_sock.h>
30 #include <net/route.h>
31 #include <net/snmp.h>
32 #include <net/flow.h>
33 #include <net/flow_dissector.h>
34 #include <net/netns/hash.h>
35 #include <net/lwtunnel.h>
36 #include <net/inet_dscp.h>
37 
38 #define IPV4_MAX_PMTU		65535U		/* RFC 2675, Section 5.1 */
39 #define IPV4_MIN_MTU		68			/* RFC 791 */
40 
41 extern unsigned int sysctl_fib_sync_mem;
42 extern unsigned int sysctl_fib_sync_mem_min;
43 extern unsigned int sysctl_fib_sync_mem_max;
44 
45 struct sock;
46 
47 struct inet_skb_parm {
48 	int			iif;
49 	struct ip_options	opt;		/* Compiled IP options		*/
50 	u16			flags;
51 
52 #define IPSKB_FORWARDED		BIT(0)
53 #define IPSKB_XFRM_TUNNEL_SIZE	BIT(1)
54 #define IPSKB_XFRM_TRANSFORMED	BIT(2)
55 #define IPSKB_FRAG_COMPLETE	BIT(3)
56 #define IPSKB_REROUTED		BIT(4)
57 #define IPSKB_DOREDIRECT	BIT(5)
58 #define IPSKB_FRAG_PMTU		BIT(6)
59 #define IPSKB_L3SLAVE		BIT(7)
60 #define IPSKB_NOPOLICY		BIT(8)
61 #define IPSKB_MULTIPATH		BIT(9)
62 
63 	u16			frag_max_size;
64 };
65 
ipv4_l3mdev_skb(u16 flags)66 static inline bool ipv4_l3mdev_skb(u16 flags)
67 {
68 	return !!(flags & IPSKB_L3SLAVE);
69 }
70 
ip_hdrlen(const struct sk_buff * skb)71 static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
72 {
73 	return ip_hdr(skb)->ihl * 4;
74 }
75 
76 struct ipcm_cookie {
77 	struct sockcm_cookie	sockc;
78 	__be32			addr;
79 	int			oif;
80 	struct ip_options_rcu	*opt;
81 	__u8			protocol;
82 	__u8			ttl;
83 	__s16			tos;
84 	char			priority;
85 	__u16			gso_size;
86 };
87 
ipcm_init(struct ipcm_cookie * ipcm)88 static inline void ipcm_init(struct ipcm_cookie *ipcm)
89 {
90 	*ipcm = (struct ipcm_cookie) { .tos = -1 };
91 }
92 
ipcm_init_sk(struct ipcm_cookie * ipcm,const struct inet_sock * inet)93 static inline void ipcm_init_sk(struct ipcm_cookie *ipcm,
94 				const struct inet_sock *inet)
95 {
96 	ipcm_init(ipcm);
97 
98 	ipcm->sockc.mark = READ_ONCE(inet->sk.sk_mark);
99 	ipcm->sockc.tsflags = READ_ONCE(inet->sk.sk_tsflags);
100 	ipcm->oif = READ_ONCE(inet->sk.sk_bound_dev_if);
101 	ipcm->addr = inet->inet_saddr;
102 	ipcm->protocol = inet->inet_num;
103 }
104 
105 #define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
106 #define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
107 
108 /* return enslaved device index if relevant */
inet_sdif(const struct sk_buff * skb)109 static inline int inet_sdif(const struct sk_buff *skb)
110 {
111 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
112 	if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
113 		return IPCB(skb)->iif;
114 #endif
115 	return 0;
116 }
117 
118 /* Special input handler for packets caught by router alert option.
119    They are selected only by protocol field, and then processed likely
120    local ones; but only if someone wants them! Otherwise, router
121    not running rsvpd will kill RSVP.
122 
123    It is user level problem, what it will make with them.
124    I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
125    but receiver should be enough clever f.e. to forward mtrace requests,
126    sent to multicast group to reach destination designated router.
127  */
128 
129 struct ip_ra_chain {
130 	struct ip_ra_chain __rcu *next;
131 	struct sock		*sk;
132 	union {
133 		void			(*destructor)(struct sock *);
134 		struct sock		*saved_sk;
135 	};
136 	struct rcu_head		rcu;
137 };
138 
139 /* IP flags. */
140 #define IP_CE		0x8000		/* Flag: "Congestion"		*/
141 #define IP_DF		0x4000		/* Flag: "Don't Fragment"	*/
142 #define IP_MF		0x2000		/* Flag: "More Fragments"	*/
143 #define IP_OFFSET	0x1FFF		/* "Fragment Offset" part	*/
144 
145 #define IP_FRAG_TIME	(30 * HZ)		/* fragment lifetime	*/
146 
147 struct msghdr;
148 struct net_device;
149 struct packet_type;
150 struct rtable;
151 struct sockaddr;
152 
153 int igmp_mc_init(void);
154 
155 /*
156  *	Functions provided by ip.c
157  */
158 
159 int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
160 			  __be32 saddr, __be32 daddr,
161 			  struct ip_options_rcu *opt, u8 tos);
162 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
163 	   struct net_device *orig_dev);
164 void ip_list_rcv(struct list_head *head, struct packet_type *pt,
165 		 struct net_device *orig_dev);
166 int ip_local_deliver(struct sk_buff *skb);
167 void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int proto);
168 int ip_mr_input(struct sk_buff *skb);
169 int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
170 int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
171 int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
172 		   int (*output)(struct net *, struct sock *, struct sk_buff *));
173 
174 struct ip_fraglist_iter {
175 	struct sk_buff	*frag;
176 	struct iphdr	*iph;
177 	int		offset;
178 	unsigned int	hlen;
179 };
180 
181 void ip_fraglist_init(struct sk_buff *skb, struct iphdr *iph,
182 		      unsigned int hlen, struct ip_fraglist_iter *iter);
183 void ip_fraglist_prepare(struct sk_buff *skb, struct ip_fraglist_iter *iter);
184 
ip_fraglist_next(struct ip_fraglist_iter * iter)185 static inline struct sk_buff *ip_fraglist_next(struct ip_fraglist_iter *iter)
186 {
187 	struct sk_buff *skb = iter->frag;
188 
189 	iter->frag = skb->next;
190 	skb_mark_not_on_list(skb);
191 
192 	return skb;
193 }
194 
195 struct ip_frag_state {
196 	bool		DF;
197 	unsigned int	hlen;
198 	unsigned int	ll_rs;
199 	unsigned int	mtu;
200 	unsigned int	left;
201 	int		offset;
202 	int		ptr;
203 	__be16		not_last_frag;
204 };
205 
206 void ip_frag_init(struct sk_buff *skb, unsigned int hlen, unsigned int ll_rs,
207 		  unsigned int mtu, bool DF, struct ip_frag_state *state);
208 struct sk_buff *ip_frag_next(struct sk_buff *skb,
209 			     struct ip_frag_state *state);
210 
211 void ip_send_check(struct iphdr *ip);
212 int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
213 int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
214 
215 int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
216 		    __u8 tos);
217 void ip_init(void);
218 int ip_append_data(struct sock *sk, struct flowi4 *fl4,
219 		   int getfrag(void *from, char *to, int offset, int len,
220 			       int odd, struct sk_buff *skb),
221 		   void *from, int len, int protolen,
222 		   struct ipcm_cookie *ipc,
223 		   struct rtable **rt,
224 		   unsigned int flags);
225 int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
226 		       struct sk_buff *skb);
227 struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
228 			      struct sk_buff_head *queue,
229 			      struct inet_cork *cork);
230 int ip_send_skb(struct net *net, struct sk_buff *skb);
231 int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
232 void ip_flush_pending_frames(struct sock *sk);
233 struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
234 			    int getfrag(void *from, char *to, int offset,
235 					int len, int odd, struct sk_buff *skb),
236 			    void *from, int length, int transhdrlen,
237 			    struct ipcm_cookie *ipc, struct rtable **rtp,
238 			    struct inet_cork *cork, unsigned int flags);
239 
240 int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
241 
ip_finish_skb(struct sock * sk,struct flowi4 * fl4)242 static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
243 {
244 	return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
245 }
246 
247 /* Get the route scope that should be used when sending a packet. */
ip_sendmsg_scope(const struct inet_sock * inet,const struct ipcm_cookie * ipc,const struct msghdr * msg)248 static inline u8 ip_sendmsg_scope(const struct inet_sock *inet,
249 				  const struct ipcm_cookie *ipc,
250 				  const struct msghdr *msg)
251 {
252 	if (sock_flag(&inet->sk, SOCK_LOCALROUTE) ||
253 	    msg->msg_flags & MSG_DONTROUTE ||
254 	    (ipc->opt && ipc->opt->opt.is_strictroute))
255 		return RT_SCOPE_LINK;
256 
257 	return RT_SCOPE_UNIVERSE;
258 }
259 
get_rttos(struct ipcm_cookie * ipc,struct inet_sock * inet)260 static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
261 {
262 	u8 dsfield = ipc->tos != -1 ? ipc->tos : READ_ONCE(inet->tos);
263 
264 	return dsfield & INET_DSCP_MASK;
265 }
266 
267 /* datagram.c */
268 int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
269 int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
270 
271 void ip4_datagram_release_cb(struct sock *sk);
272 
273 struct ip_reply_arg {
274 	struct kvec iov[1];
275 	int	    flags;
276 	__wsum 	    csum;
277 	int	    csumoffset; /* u16 offset of csum in iov[0].iov_base */
278 				/* -1 if not needed */
279 	int	    bound_dev_if;
280 	u8  	    tos;
281 	kuid_t	    uid;
282 };
283 
284 #define IP_REPLY_ARG_NOSRCCHECK 1
285 
ip_reply_arg_flowi_flags(const struct ip_reply_arg * arg)286 static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
287 {
288 	return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
289 }
290 
291 void ip_send_unicast_reply(struct sock *sk, const struct sock *orig_sk,
292 			   struct sk_buff *skb,
293 			   const struct ip_options *sopt,
294 			   __be32 daddr, __be32 saddr,
295 			   const struct ip_reply_arg *arg,
296 			   unsigned int len, u64 transmit_time, u32 txhash);
297 
298 #define IP_INC_STATS(net, field)	SNMP_INC_STATS64((net)->mib.ip_statistics, field)
299 #define __IP_INC_STATS(net, field)	__SNMP_INC_STATS64((net)->mib.ip_statistics, field)
300 #define IP_ADD_STATS(net, field, val)	SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
301 #define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
302 #define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
303 #define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
304 #define NET_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.net_statistics, field)
305 #define __NET_INC_STATS(net, field)	__SNMP_INC_STATS((net)->mib.net_statistics, field)
306 #define NET_ADD_STATS(net, field, adnd)	SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
307 #define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
308 
snmp_get_cpu_field(void __percpu * mib,int cpu,int offt)309 static inline u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offt)
310 {
311 	return  *(((unsigned long *)per_cpu_ptr(mib, cpu)) + offt);
312 }
313 
314 unsigned long snmp_fold_field(void __percpu *mib, int offt);
315 #if BITS_PER_LONG==32
316 u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
317 			 size_t syncp_offset);
318 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
319 #else
snmp_get_cpu_field64(void __percpu * mib,int cpu,int offct,size_t syncp_offset)320 static inline u64  snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
321 					size_t syncp_offset)
322 {
323 	return snmp_get_cpu_field(mib, cpu, offct);
324 
325 }
326 
snmp_fold_field64(void __percpu * mib,int offt,size_t syncp_off)327 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
328 {
329 	return snmp_fold_field(mib, offt);
330 }
331 #endif
332 
333 #define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
334 { \
335 	int i, c; \
336 	for_each_possible_cpu(c) { \
337 		for (i = 0; stats_list[i].name; i++) \
338 			buff64[i] += snmp_get_cpu_field64( \
339 					mib_statistic, \
340 					c, stats_list[i].entry, \
341 					offset); \
342 	} \
343 }
344 
345 #define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
346 { \
347 	int i, c; \
348 	for_each_possible_cpu(c) { \
349 		for (i = 0; stats_list[i].name; i++) \
350 			buff[i] += snmp_get_cpu_field( \
351 						mib_statistic, \
352 						c, stats_list[i].entry); \
353 	} \
354 }
355 
inet_get_local_port_range(const struct net * net,int * low,int * high)356 static inline void inet_get_local_port_range(const struct net *net, int *low, int *high)
357 {
358 	u32 range = READ_ONCE(net->ipv4.ip_local_ports.range);
359 
360 	*low = range & 0xffff;
361 	*high = range >> 16;
362 }
363 bool inet_sk_get_local_port_range(const struct sock *sk, int *low, int *high);
364 
365 #ifdef CONFIG_SYSCTL
inet_is_local_reserved_port(struct net * net,unsigned short port)366 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
367 {
368 	if (!net->ipv4.sysctl_local_reserved_ports)
369 		return false;
370 	return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
371 }
372 
sysctl_dev_name_is_allowed(const char * name)373 static inline bool sysctl_dev_name_is_allowed(const char *name)
374 {
375 	return strcmp(name, "default") != 0  && strcmp(name, "all") != 0;
376 }
377 
inet_port_requires_bind_service(struct net * net,unsigned short port)378 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
379 {
380 	return port < READ_ONCE(net->ipv4.sysctl_ip_prot_sock);
381 }
382 
383 #else
inet_is_local_reserved_port(struct net * net,unsigned short port)384 static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
385 {
386 	return false;
387 }
388 
inet_port_requires_bind_service(struct net * net,unsigned short port)389 static inline bool inet_port_requires_bind_service(struct net *net, unsigned short port)
390 {
391 	return port < PROT_SOCK;
392 }
393 #endif
394 
395 __be32 inet_current_timestamp(void);
396 
397 /* From inetpeer.c */
398 extern int inet_peer_threshold;
399 extern int inet_peer_minttl;
400 extern int inet_peer_maxttl;
401 
402 void ipfrag_init(void);
403 
404 void ip_static_sysctl_init(void);
405 
406 #define IP4_REPLY_MARK(net, mark) \
407 	(READ_ONCE((net)->ipv4.sysctl_fwmark_reflect) ? (mark) : 0)
408 
ip_is_fragment(const struct iphdr * iph)409 static inline bool ip_is_fragment(const struct iphdr *iph)
410 {
411 	return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
412 }
413 
414 #ifdef CONFIG_INET
415 #include <net/dst.h>
416 
417 /* The function in 2.2 was invalid, producing wrong result for
418  * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
419 static inline
ip_decrease_ttl(struct iphdr * iph)420 int ip_decrease_ttl(struct iphdr *iph)
421 {
422 	u32 check = (__force u32)iph->check;
423 	check += (__force u32)htons(0x0100);
424 	iph->check = (__force __sum16)(check + (check>=0xFFFF));
425 	return --iph->ttl;
426 }
427 
ip4h_dscp(const struct iphdr * ip4h)428 static inline dscp_t ip4h_dscp(const struct iphdr *ip4h)
429 {
430 	return inet_dsfield_to_dscp(ip4h->tos);
431 }
432 
ip_mtu_locked(const struct dst_entry * dst)433 static inline int ip_mtu_locked(const struct dst_entry *dst)
434 {
435 	const struct rtable *rt = dst_rtable(dst);
436 
437 	return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
438 }
439 
440 static inline
ip_dont_fragment(const struct sock * sk,const struct dst_entry * dst)441 int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
442 {
443 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
444 
445 	return  pmtudisc == IP_PMTUDISC_DO ||
446 		(pmtudisc == IP_PMTUDISC_WANT &&
447 		 !ip_mtu_locked(dst));
448 }
449 
ip_sk_accept_pmtu(const struct sock * sk)450 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
451 {
452 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
453 
454 	return pmtudisc != IP_PMTUDISC_INTERFACE &&
455 	       pmtudisc != IP_PMTUDISC_OMIT;
456 }
457 
ip_sk_use_pmtu(const struct sock * sk)458 static inline bool ip_sk_use_pmtu(const struct sock *sk)
459 {
460 	return READ_ONCE(inet_sk(sk)->pmtudisc) < IP_PMTUDISC_PROBE;
461 }
462 
ip_sk_ignore_df(const struct sock * sk)463 static inline bool ip_sk_ignore_df(const struct sock *sk)
464 {
465 	u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
466 
467 	return pmtudisc < IP_PMTUDISC_DO || pmtudisc == IP_PMTUDISC_OMIT;
468 }
469 
ip_dst_mtu_maybe_forward(const struct dst_entry * dst,bool forwarding)470 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
471 						    bool forwarding)
472 {
473 	const struct rtable *rt = dst_rtable(dst);
474 	struct net *net = dev_net(dst->dev);
475 	unsigned int mtu;
476 
477 	if (READ_ONCE(net->ipv4.sysctl_ip_fwd_use_pmtu) ||
478 	    ip_mtu_locked(dst) ||
479 	    !forwarding) {
480 		mtu = rt->rt_pmtu;
481 		if (mtu && time_before(jiffies, rt->dst.expires))
482 			goto out;
483 	}
484 
485 	/* 'forwarding = true' case should always honour route mtu */
486 	mtu = dst_metric_raw(dst, RTAX_MTU);
487 	if (mtu)
488 		goto out;
489 
490 	mtu = READ_ONCE(dst->dev->mtu);
491 
492 	if (unlikely(ip_mtu_locked(dst))) {
493 		if (rt->rt_uses_gateway && mtu > 576)
494 			mtu = 576;
495 	}
496 
497 out:
498 	mtu = min_t(unsigned int, mtu, IP_MAX_MTU);
499 
500 	return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
501 }
502 
ip_skb_dst_mtu(struct sock * sk,const struct sk_buff * skb)503 static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
504 					  const struct sk_buff *skb)
505 {
506 	unsigned int mtu;
507 
508 	if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
509 		bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
510 
511 		return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
512 	}
513 
514 	mtu = min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
515 	return mtu - lwtunnel_headroom(skb_dst(skb)->lwtstate, mtu);
516 }
517 
518 struct dst_metrics *ip_fib_metrics_init(struct nlattr *fc_mx, int fc_mx_len,
519 					struct netlink_ext_ack *extack);
ip_fib_metrics_put(struct dst_metrics * fib_metrics)520 static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
521 {
522 	if (fib_metrics != &dst_default_metrics &&
523 	    refcount_dec_and_test(&fib_metrics->refcnt))
524 		kfree(fib_metrics);
525 }
526 
527 /* ipv4 and ipv6 both use refcounted metrics if it is not the default */
528 static inline
ip_dst_init_metrics(struct dst_entry * dst,struct dst_metrics * fib_metrics)529 void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
530 {
531 	dst_init_metrics(dst, fib_metrics->metrics, true);
532 
533 	if (fib_metrics != &dst_default_metrics) {
534 		dst->_metrics |= DST_METRICS_REFCOUNTED;
535 		refcount_inc(&fib_metrics->refcnt);
536 	}
537 }
538 
539 static inline
ip_dst_metrics_put(struct dst_entry * dst)540 void ip_dst_metrics_put(struct dst_entry *dst)
541 {
542 	struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
543 
544 	if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
545 		kfree(p);
546 }
547 
548 void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
549 
ip_select_ident_segs(struct net * net,struct sk_buff * skb,struct sock * sk,int segs)550 static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
551 					struct sock *sk, int segs)
552 {
553 	struct iphdr *iph = ip_hdr(skb);
554 
555 	/* We had many attacks based on IPID, use the private
556 	 * generator as much as we can.
557 	 */
558 	if (sk && inet_sk(sk)->inet_daddr) {
559 		int val;
560 
561 		/* avoid atomic operations for TCP,
562 		 * as we hold socket lock at this point.
563 		 */
564 		if (sk_is_tcp(sk)) {
565 			sock_owned_by_me(sk);
566 			val = atomic_read(&inet_sk(sk)->inet_id);
567 			atomic_set(&inet_sk(sk)->inet_id, val + segs);
568 		} else {
569 			val = atomic_add_return(segs, &inet_sk(sk)->inet_id);
570 		}
571 		iph->id = htons(val);
572 		return;
573 	}
574 	if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
575 		iph->id = 0;
576 	} else {
577 		/* Unfortunately we need the big hammer to get a suitable IPID */
578 		__ip_select_ident(net, iph, segs);
579 	}
580 }
581 
ip_select_ident(struct net * net,struct sk_buff * skb,struct sock * sk)582 static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
583 				   struct sock *sk)
584 {
585 	ip_select_ident_segs(net, skb, sk, 1);
586 }
587 
inet_compute_pseudo(struct sk_buff * skb,int proto)588 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
589 {
590 	return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
591 				  skb->len, proto, 0);
592 }
593 
594 /* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
595  * Equivalent to :	flow->v4addrs.src = iph->saddr;
596  *			flow->v4addrs.dst = iph->daddr;
597  */
iph_to_flow_copy_v4addrs(struct flow_keys * flow,const struct iphdr * iph)598 static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
599 					    const struct iphdr *iph)
600 {
601 	BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
602 		     offsetof(typeof(flow->addrs), v4addrs.src) +
603 			      sizeof(flow->addrs.v4addrs.src));
604 	memcpy(&flow->addrs.v4addrs, &iph->addrs, sizeof(flow->addrs.v4addrs));
605 	flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
606 }
607 
608 /*
609  *	Map a multicast IP onto multicast MAC for type ethernet.
610  */
611 
ip_eth_mc_map(__be32 naddr,char * buf)612 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
613 {
614 	__u32 addr=ntohl(naddr);
615 	buf[0]=0x01;
616 	buf[1]=0x00;
617 	buf[2]=0x5e;
618 	buf[5]=addr&0xFF;
619 	addr>>=8;
620 	buf[4]=addr&0xFF;
621 	addr>>=8;
622 	buf[3]=addr&0x7F;
623 }
624 
625 /*
626  *	Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
627  *	Leave P_Key as 0 to be filled in by driver.
628  */
629 
ip_ib_mc_map(__be32 naddr,const unsigned char * broadcast,char * buf)630 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
631 {
632 	__u32 addr;
633 	unsigned char scope = broadcast[5] & 0xF;
634 
635 	buf[0]  = 0;		/* Reserved */
636 	buf[1]  = 0xff;		/* Multicast QPN */
637 	buf[2]  = 0xff;
638 	buf[3]  = 0xff;
639 	addr    = ntohl(naddr);
640 	buf[4]  = 0xff;
641 	buf[5]  = 0x10 | scope;	/* scope from broadcast address */
642 	buf[6]  = 0x40;		/* IPv4 signature */
643 	buf[7]  = 0x1b;
644 	buf[8]  = broadcast[8];		/* P_Key */
645 	buf[9]  = broadcast[9];
646 	buf[10] = 0;
647 	buf[11] = 0;
648 	buf[12] = 0;
649 	buf[13] = 0;
650 	buf[14] = 0;
651 	buf[15] = 0;
652 	buf[19] = addr & 0xff;
653 	addr  >>= 8;
654 	buf[18] = addr & 0xff;
655 	addr  >>= 8;
656 	buf[17] = addr & 0xff;
657 	addr  >>= 8;
658 	buf[16] = addr & 0x0f;
659 }
660 
ip_ipgre_mc_map(__be32 naddr,const unsigned char * broadcast,char * buf)661 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
662 {
663 	if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
664 		memcpy(buf, broadcast, 4);
665 	else
666 		memcpy(buf, &naddr, sizeof(naddr));
667 }
668 
669 #if IS_ENABLED(CONFIG_IPV6)
670 #include <linux/ipv6.h>
671 #endif
672 
inet_reset_saddr(struct sock * sk)673 static __inline__ void inet_reset_saddr(struct sock *sk)
674 {
675 	inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
676 #if IS_ENABLED(CONFIG_IPV6)
677 	if (sk->sk_family == PF_INET6) {
678 		struct ipv6_pinfo *np = inet6_sk(sk);
679 
680 		memset(&np->saddr, 0, sizeof(np->saddr));
681 		memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
682 	}
683 #endif
684 }
685 
686 #endif
687 
ipv4_addr_hash(__be32 ip)688 static inline unsigned int ipv4_addr_hash(__be32 ip)
689 {
690 	return (__force unsigned int) ip;
691 }
692 
__ipv4_addr_hash(const __be32 ip,const u32 initval)693 static inline u32 __ipv4_addr_hash(const __be32 ip, const u32 initval)
694 {
695 	return jhash_1word((__force u32)ip, initval);
696 }
697 
ipv4_portaddr_hash(const struct net * net,__be32 saddr,unsigned int port)698 static inline u32 ipv4_portaddr_hash(const struct net *net,
699 				     __be32 saddr,
700 				     unsigned int port)
701 {
702 	return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
703 }
704 
705 bool ip_call_ra_chain(struct sk_buff *skb);
706 
707 /*
708  *	Functions provided by ip_fragment.c
709  */
710 
711 enum ip_defrag_users {
712 	IP_DEFRAG_LOCAL_DELIVER,
713 	IP_DEFRAG_CALL_RA_CHAIN,
714 	IP_DEFRAG_CONNTRACK_IN,
715 	__IP_DEFRAG_CONNTRACK_IN_END	= IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
716 	IP_DEFRAG_CONNTRACK_OUT,
717 	__IP_DEFRAG_CONNTRACK_OUT_END	= IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
718 	IP_DEFRAG_CONNTRACK_BRIDGE_IN,
719 	__IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
720 	IP_DEFRAG_VS_IN,
721 	IP_DEFRAG_VS_OUT,
722 	IP_DEFRAG_VS_FWD,
723 	IP_DEFRAG_AF_PACKET,
724 	IP_DEFRAG_MACVLAN,
725 };
726 
727 /* Return true if the value of 'user' is between 'lower_bond'
728  * and 'upper_bond' inclusively.
729  */
ip_defrag_user_in_between(u32 user,enum ip_defrag_users lower_bond,enum ip_defrag_users upper_bond)730 static inline bool ip_defrag_user_in_between(u32 user,
731 					     enum ip_defrag_users lower_bond,
732 					     enum ip_defrag_users upper_bond)
733 {
734 	return user >= lower_bond && user <= upper_bond;
735 }
736 
737 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
738 #ifdef CONFIG_INET
739 struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
740 #else
ip_check_defrag(struct net * net,struct sk_buff * skb,u32 user)741 static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
742 {
743 	return skb;
744 }
745 #endif
746 
747 /*
748  *	Functions provided by ip_forward.c
749  */
750 
751 int ip_forward(struct sk_buff *skb);
752 
753 /*
754  *	Functions provided by ip_options.c
755  */
756 
757 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
758 		      __be32 daddr, struct rtable *rt);
759 
760 int __ip_options_echo(struct net *net, struct ip_options *dopt,
761 		      struct sk_buff *skb, const struct ip_options *sopt);
ip_options_echo(struct net * net,struct ip_options * dopt,struct sk_buff * skb)762 static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
763 				  struct sk_buff *skb)
764 {
765 	return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
766 }
767 
768 void ip_options_fragment(struct sk_buff *skb);
769 int __ip_options_compile(struct net *net, struct ip_options *opt,
770 			 struct sk_buff *skb, __be32 *info);
771 int ip_options_compile(struct net *net, struct ip_options *opt,
772 		       struct sk_buff *skb);
773 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
774 		   sockptr_t data, int optlen);
775 void ip_options_undo(struct ip_options *opt);
776 void ip_forward_options(struct sk_buff *skb);
777 int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
778 
779 /*
780  *	Functions provided by ip_sockglue.c
781  */
782 
783 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb, bool drop_dst);
784 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
785 			 struct sk_buff *skb, int tlen, int offset);
786 int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
787 		 struct ipcm_cookie *ipc, bool allow_ipv6);
788 DECLARE_STATIC_KEY_FALSE(ip4_min_ttl);
789 int do_ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
790 		     unsigned int optlen);
791 int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
792 		  unsigned int optlen);
793 int do_ip_getsockopt(struct sock *sk, int level, int optname,
794 		     sockptr_t optval, sockptr_t optlen);
795 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
796 		  int __user *optlen);
797 int ip_ra_control(struct sock *sk, unsigned char on,
798 		  void (*destructor)(struct sock *));
799 
800 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
801 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
802 		   u32 info, u8 *payload);
803 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
804 		    u32 info);
805 
ip_cmsg_recv(struct msghdr * msg,struct sk_buff * skb)806 static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
807 {
808 	ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
809 }
810 
811 bool icmp_global_allow(struct net *net);
812 void icmp_global_consume(struct net *net);
813 
814 #ifdef CONFIG_PROC_FS
815 int ip_misc_proc_init(void);
816 #endif
817 
818 int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
819 				struct netlink_ext_ack *extack);
820 
inetdev_valid_mtu(unsigned int mtu)821 static inline bool inetdev_valid_mtu(unsigned int mtu)
822 {
823 	return likely(mtu >= IPV4_MIN_MTU);
824 }
825 
826 void ip_sock_set_freebind(struct sock *sk);
827 int ip_sock_set_mtu_discover(struct sock *sk, int val);
828 void ip_sock_set_pktinfo(struct sock *sk);
829 void ip_sock_set_recverr(struct sock *sk);
830 void ip_sock_set_tos(struct sock *sk, int val);
831 void  __ip_sock_set_tos(struct sock *sk, int val);
832 
833 #endif	/* _IP_H */
834