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