Loading...
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
41extern unsigned int sysctl_fib_sync_mem;
42extern unsigned int sysctl_fib_sync_mem_min;
43extern unsigned int sysctl_fib_sync_mem_max;
44
45struct sock;
46
47struct 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
66static inline bool ipv4_l3mdev_skb(u16 flags)
67{
68 return !!(flags & IPSKB_L3SLAVE);
69}
70
71static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
72{
73 return ip_hdr(skb)->ihl * 4;
74}
75
76struct 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
88static inline void ipcm_init(struct ipcm_cookie *ipcm)
89{
90 *ipcm = (struct ipcm_cookie) { .tos = -1 };
91}
92
93static 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 */
109static 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
129struct 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
147struct msghdr;
148struct net_device;
149struct packet_type;
150struct rtable;
151struct sockaddr;
152
153int igmp_mc_init(void);
154
155/*
156 * Functions provided by ip.c
157 */
158
159int 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);
162int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
163 struct net_device *orig_dev);
164void ip_list_rcv(struct list_head *head, struct packet_type *pt,
165 struct net_device *orig_dev);
166int ip_local_deliver(struct sk_buff *skb);
167void ip_protocol_deliver_rcu(struct net *net, struct sk_buff *skb, int proto);
168int ip_mr_input(struct sk_buff *skb);
169int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
170int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
171int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
172 int (*output)(struct net *, struct sock *, struct sk_buff *));
173
174struct ip_fraglist_iter {
175 struct sk_buff *frag;
176 struct iphdr *iph;
177 int offset;
178 unsigned int hlen;
179};
180
181void ip_fraglist_init(struct sk_buff *skb, struct iphdr *iph,
182 unsigned int hlen, struct ip_fraglist_iter *iter);
183void ip_fraglist_prepare(struct sk_buff *skb, struct ip_fraglist_iter *iter);
184
185static 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
195struct 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
206void 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);
208struct sk_buff *ip_frag_next(struct sk_buff *skb,
209 struct ip_frag_state *state);
210
211void ip_send_check(struct iphdr *ip);
212int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
213int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
214
215int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
216 __u8 tos);
217void ip_init(void);
218int 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);
225int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
226 struct sk_buff *skb);
227struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
228 struct sk_buff_head *queue,
229 struct inet_cork *cork);
230int ip_send_skb(struct net *net, struct sk_buff *skb);
231int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
232void ip_flush_pending_frames(struct sock *sk);
233struct 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
240int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
241
242static 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. */
248static 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
260static 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 */
268int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
269int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
270
271void ip4_datagram_release_cb(struct sock *sk);
272
273struct 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
286static 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
291void 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
309static 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
314unsigned long snmp_fold_field(void __percpu *mib, int offt);
315#if BITS_PER_LONG==32
316u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
317 size_t syncp_offset);
318u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
319#else
320static 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
327static 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
356static 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}
363bool inet_sk_get_local_port_range(const struct sock *sk, int *low, int *high);
364
365#ifdef CONFIG_SYSCTL
366static 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
373static inline bool sysctl_dev_name_is_allowed(const char *name)
374{
375 return strcmp(name, "default") != 0 && strcmp(name, "all") != 0;
376}
377
378static 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
384static inline bool inet_is_local_reserved_port(struct net *net, unsigned short port)
385{
386 return false;
387}
388
389static 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 */
398extern int inet_peer_threshold;
399extern int inet_peer_minttl;
400extern int inet_peer_maxttl;
401
402void ipfrag_init(void);
403
404void ip_static_sysctl_init(void);
405
406#define IP4_REPLY_MARK(net, mark) \
407 (READ_ONCE((net)->ipv4.sysctl_fwmark_reflect) ? (mark) : 0)
408
409static 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) */
419static inline
420int 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
428static inline dscp_t ip4h_dscp(const struct iphdr *ip4h)
429{
430 return inet_dsfield_to_dscp(ip4h->tos);
431}
432
433static 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
440static inline
441int 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
450static 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
458static inline bool ip_sk_use_pmtu(const struct sock *sk)
459{
460 return READ_ONCE(inet_sk(sk)->pmtudisc) < IP_PMTUDISC_PROBE;
461}
462
463static 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
470static 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 unsigned int mtu, res;
475 struct net *net;
476
477 rcu_read_lock();
478
479 net = dev_net_rcu(dst->dev);
480 if (READ_ONCE(net->ipv4.sysctl_ip_fwd_use_pmtu) ||
481 ip_mtu_locked(dst) ||
482 !forwarding) {
483 mtu = rt->rt_pmtu;
484 if (mtu && time_before(jiffies, rt->dst.expires))
485 goto out;
486 }
487
488 /* 'forwarding = true' case should always honour route mtu */
489 mtu = dst_metric_raw(dst, RTAX_MTU);
490 if (mtu)
491 goto out;
492
493 mtu = READ_ONCE(dst->dev->mtu);
494
495 if (unlikely(ip_mtu_locked(dst))) {
496 if (rt->rt_uses_gateway && mtu > 576)
497 mtu = 576;
498 }
499
500out:
501 mtu = min_t(unsigned int, mtu, IP_MAX_MTU);
502
503 res = mtu - lwtunnel_headroom(dst->lwtstate, mtu);
504
505 rcu_read_unlock();
506
507 return res;
508}
509
510static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
511 const struct sk_buff *skb)
512{
513 unsigned int mtu;
514
515 if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
516 bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
517
518 return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
519 }
520
521 mtu = min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
522 return mtu - lwtunnel_headroom(skb_dst(skb)->lwtstate, mtu);
523}
524
525struct dst_metrics *ip_fib_metrics_init(struct nlattr *fc_mx, int fc_mx_len,
526 struct netlink_ext_ack *extack);
527static inline void ip_fib_metrics_put(struct dst_metrics *fib_metrics)
528{
529 if (fib_metrics != &dst_default_metrics &&
530 refcount_dec_and_test(&fib_metrics->refcnt))
531 kfree(fib_metrics);
532}
533
534/* ipv4 and ipv6 both use refcounted metrics if it is not the default */
535static inline
536void ip_dst_init_metrics(struct dst_entry *dst, struct dst_metrics *fib_metrics)
537{
538 dst_init_metrics(dst, fib_metrics->metrics, true);
539
540 if (fib_metrics != &dst_default_metrics) {
541 dst->_metrics |= DST_METRICS_REFCOUNTED;
542 refcount_inc(&fib_metrics->refcnt);
543 }
544}
545
546static inline
547void ip_dst_metrics_put(struct dst_entry *dst)
548{
549 struct dst_metrics *p = (struct dst_metrics *)DST_METRICS_PTR(dst);
550
551 if (p != &dst_default_metrics && refcount_dec_and_test(&p->refcnt))
552 kfree(p);
553}
554
555void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
556
557static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
558 struct sock *sk, int segs)
559{
560 struct iphdr *iph = ip_hdr(skb);
561
562 /* We had many attacks based on IPID, use the private
563 * generator as much as we can.
564 */
565 if (sk && inet_sk(sk)->inet_daddr) {
566 int val;
567
568 /* avoid atomic operations for TCP,
569 * as we hold socket lock at this point.
570 */
571 if (sk_is_tcp(sk)) {
572 sock_owned_by_me(sk);
573 val = atomic_read(&inet_sk(sk)->inet_id);
574 atomic_set(&inet_sk(sk)->inet_id, val + segs);
575 } else {
576 val = atomic_add_return(segs, &inet_sk(sk)->inet_id);
577 }
578 iph->id = htons(val);
579 return;
580 }
581 if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
582 iph->id = 0;
583 } else {
584 /* Unfortunately we need the big hammer to get a suitable IPID */
585 __ip_select_ident(net, iph, segs);
586 }
587}
588
589static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
590 struct sock *sk)
591{
592 ip_select_ident_segs(net, skb, sk, 1);
593}
594
595static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
596{
597 return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
598 skb->len, proto, 0);
599}
600
601/* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
602 * Equivalent to : flow->v4addrs.src = iph->saddr;
603 * flow->v4addrs.dst = iph->daddr;
604 */
605static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
606 const struct iphdr *iph)
607{
608 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
609 offsetof(typeof(flow->addrs), v4addrs.src) +
610 sizeof(flow->addrs.v4addrs.src));
611 memcpy(&flow->addrs.v4addrs, &iph->addrs, sizeof(flow->addrs.v4addrs));
612 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
613}
614
615/*
616 * Map a multicast IP onto multicast MAC for type ethernet.
617 */
618
619static inline void ip_eth_mc_map(__be32 naddr, char *buf)
620{
621 __u32 addr=ntohl(naddr);
622 buf[0]=0x01;
623 buf[1]=0x00;
624 buf[2]=0x5e;
625 buf[5]=addr&0xFF;
626 addr>>=8;
627 buf[4]=addr&0xFF;
628 addr>>=8;
629 buf[3]=addr&0x7F;
630}
631
632/*
633 * Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
634 * Leave P_Key as 0 to be filled in by driver.
635 */
636
637static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
638{
639 __u32 addr;
640 unsigned char scope = broadcast[5] & 0xF;
641
642 buf[0] = 0; /* Reserved */
643 buf[1] = 0xff; /* Multicast QPN */
644 buf[2] = 0xff;
645 buf[3] = 0xff;
646 addr = ntohl(naddr);
647 buf[4] = 0xff;
648 buf[5] = 0x10 | scope; /* scope from broadcast address */
649 buf[6] = 0x40; /* IPv4 signature */
650 buf[7] = 0x1b;
651 buf[8] = broadcast[8]; /* P_Key */
652 buf[9] = broadcast[9];
653 buf[10] = 0;
654 buf[11] = 0;
655 buf[12] = 0;
656 buf[13] = 0;
657 buf[14] = 0;
658 buf[15] = 0;
659 buf[19] = addr & 0xff;
660 addr >>= 8;
661 buf[18] = addr & 0xff;
662 addr >>= 8;
663 buf[17] = addr & 0xff;
664 addr >>= 8;
665 buf[16] = addr & 0x0f;
666}
667
668static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
669{
670 if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
671 memcpy(buf, broadcast, 4);
672 else
673 memcpy(buf, &naddr, sizeof(naddr));
674}
675
676#if IS_ENABLED(CONFIG_IPV6)
677#include <linux/ipv6.h>
678#endif
679
680static __inline__ void inet_reset_saddr(struct sock *sk)
681{
682 inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
683#if IS_ENABLED(CONFIG_IPV6)
684 if (sk->sk_family == PF_INET6) {
685 struct ipv6_pinfo *np = inet6_sk(sk);
686
687 memset(&np->saddr, 0, sizeof(np->saddr));
688 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
689 }
690#endif
691}
692
693#endif
694
695static inline unsigned int ipv4_addr_hash(__be32 ip)
696{
697 return (__force unsigned int) ip;
698}
699
700static inline u32 __ipv4_addr_hash(const __be32 ip, const u32 initval)
701{
702 return jhash_1word((__force u32)ip, initval);
703}
704
705static inline u32 ipv4_portaddr_hash(const struct net *net,
706 __be32 saddr,
707 unsigned int port)
708{
709 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
710}
711
712bool ip_call_ra_chain(struct sk_buff *skb);
713
714/*
715 * Functions provided by ip_fragment.c
716 */
717
718enum ip_defrag_users {
719 IP_DEFRAG_LOCAL_DELIVER,
720 IP_DEFRAG_CALL_RA_CHAIN,
721 IP_DEFRAG_CONNTRACK_IN,
722 __IP_DEFRAG_CONNTRACK_IN_END = IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
723 IP_DEFRAG_CONNTRACK_OUT,
724 __IP_DEFRAG_CONNTRACK_OUT_END = IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
725 IP_DEFRAG_CONNTRACK_BRIDGE_IN,
726 __IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
727 IP_DEFRAG_VS_IN,
728 IP_DEFRAG_VS_OUT,
729 IP_DEFRAG_VS_FWD,
730 IP_DEFRAG_AF_PACKET,
731 IP_DEFRAG_MACVLAN,
732};
733
734/* Return true if the value of 'user' is between 'lower_bond'
735 * and 'upper_bond' inclusively.
736 */
737static inline bool ip_defrag_user_in_between(u32 user,
738 enum ip_defrag_users lower_bond,
739 enum ip_defrag_users upper_bond)
740{
741 return user >= lower_bond && user <= upper_bond;
742}
743
744int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
745#ifdef CONFIG_INET
746struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
747#else
748static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
749{
750 return skb;
751}
752#endif
753
754/*
755 * Functions provided by ip_forward.c
756 */
757
758int ip_forward(struct sk_buff *skb);
759
760/*
761 * Functions provided by ip_options.c
762 */
763
764void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
765 __be32 daddr, struct rtable *rt);
766
767int __ip_options_echo(struct net *net, struct ip_options *dopt,
768 struct sk_buff *skb, const struct ip_options *sopt);
769static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
770 struct sk_buff *skb)
771{
772 return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
773}
774
775void ip_options_fragment(struct sk_buff *skb);
776int __ip_options_compile(struct net *net, struct ip_options *opt,
777 struct sk_buff *skb, __be32 *info);
778int ip_options_compile(struct net *net, struct ip_options *opt,
779 struct sk_buff *skb);
780int ip_options_get(struct net *net, struct ip_options_rcu **optp,
781 sockptr_t data, int optlen);
782void ip_options_undo(struct ip_options *opt);
783void ip_forward_options(struct sk_buff *skb);
784int ip_options_rcv_srr(struct sk_buff *skb, struct net_device *dev);
785
786/*
787 * Functions provided by ip_sockglue.c
788 */
789
790void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb, bool drop_dst);
791void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
792 struct sk_buff *skb, int tlen, int offset);
793int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
794 struct ipcm_cookie *ipc, bool allow_ipv6);
795DECLARE_STATIC_KEY_FALSE(ip4_min_ttl);
796int do_ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
797 unsigned int optlen);
798int ip_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
799 unsigned int optlen);
800int do_ip_getsockopt(struct sock *sk, int level, int optname,
801 sockptr_t optval, sockptr_t optlen);
802int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
803 int __user *optlen);
804int ip_ra_control(struct sock *sk, unsigned char on,
805 void (*destructor)(struct sock *));
806
807int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
808void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
809 u32 info, u8 *payload);
810void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
811 u32 info);
812
813static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
814{
815 ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
816}
817
818bool icmp_global_allow(struct net *net);
819void icmp_global_consume(struct net *net);
820
821#ifdef CONFIG_PROC_FS
822int ip_misc_proc_init(void);
823#endif
824
825int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto, u8 family,
826 struct netlink_ext_ack *extack);
827
828static inline bool inetdev_valid_mtu(unsigned int mtu)
829{
830 return likely(mtu >= IPV4_MIN_MTU);
831}
832
833void ip_sock_set_freebind(struct sock *sk);
834int ip_sock_set_mtu_discover(struct sock *sk, int val);
835void ip_sock_set_pktinfo(struct sock *sk);
836void ip_sock_set_recverr(struct sock *sk);
837void ip_sock_set_tos(struct sock *sk, int val);
838void __ip_sock_set_tos(struct sock *sk, int val);
839
840#endif /* _IP_H */
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 */