2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Definitions for the IP module.
8 * Version: @(#)ip.h 1.0.2 05/07/93
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Alan Cox, <gw4pts@gw4pts.ampr.org>
15 * Mike McLagan : Routing by source
17 * This program is free software; you can redistribute it and/or
18 * modify it under the terms of the GNU General Public License
19 * as published by the Free Software Foundation; either version
20 * 2 of the License, or (at your option) any later version.
25 #include <linux/types.h>
28 #include <linux/skbuff.h>
29 #include <linux/jhash.h>
31 #include <net/inet_sock.h>
32 #include <net/route.h>
35 #include <net/flow_dissector.h>
36 #include <net/netns/hash.h>
38 #define IPV4_MAX_PMTU 65535U /* RFC 2675, Section 5.1 */
39 #define IPV4_MIN_MTU 68 /* RFC 791 */
43 struct inet_skb_parm {
45 struct ip_options opt; /* Compiled IP options */
48 #define IPSKB_FORWARDED BIT(0)
49 #define IPSKB_XFRM_TUNNEL_SIZE BIT(1)
50 #define IPSKB_XFRM_TRANSFORMED BIT(2)
51 #define IPSKB_FRAG_COMPLETE BIT(3)
52 #define IPSKB_REROUTED BIT(4)
53 #define IPSKB_DOREDIRECT BIT(5)
54 #define IPSKB_FRAG_PMTU BIT(6)
55 #define IPSKB_L3SLAVE BIT(7)
60 static inline bool ipv4_l3mdev_skb(u16 flags)
62 return !!(flags & IPSKB_L3SLAVE);
65 static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
67 return ip_hdr(skb)->ihl * 4;
71 struct sockcm_cookie sockc;
74 struct ip_options_rcu *opt;
81 #define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
82 #define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
84 /* return enslaved device index if relevant */
85 static inline int inet_sdif(struct sk_buff *skb)
87 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
88 if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
89 return IPCB(skb)->iif;
94 /* Special input handler for packets caught by router alert option.
95 They are selected only by protocol field, and then processed likely
96 local ones; but only if someone wants them! Otherwise, router
97 not running rsvpd will kill RSVP.
99 It is user level problem, what it will make with them.
100 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
101 but receiver should be enough clever f.e. to forward mtrace requests,
102 sent to multicast group to reach destination designated router.
106 struct ip_ra_chain __rcu *next;
109 void (*destructor)(struct sock *);
110 struct sock *saved_sk;
116 #define IP_CE 0x8000 /* Flag: "Congestion" */
117 #define IP_DF 0x4000 /* Flag: "Don't Fragment" */
118 #define IP_MF 0x2000 /* Flag: "More Fragments" */
119 #define IP_OFFSET 0x1FFF /* "Fragment Offset" part */
121 #define IP_FRAG_TIME (30 * HZ) /* fragment lifetime */
129 int igmp_mc_init(void);
132 * Functions provided by ip.c
135 int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
136 __be32 saddr, __be32 daddr,
137 struct ip_options_rcu *opt);
138 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
139 struct net_device *orig_dev);
140 int ip_local_deliver(struct sk_buff *skb);
141 int ip_mr_input(struct sk_buff *skb);
142 int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
143 int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
144 int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
145 int (*output)(struct net *, struct sock *, struct sk_buff *));
146 void ip_send_check(struct iphdr *ip);
147 int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
148 int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
150 int ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
152 int ip_append_data(struct sock *sk, struct flowi4 *fl4,
153 int getfrag(void *from, char *to, int offset, int len,
154 int odd, struct sk_buff *skb),
155 void *from, int len, int protolen,
156 struct ipcm_cookie *ipc,
159 int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
160 struct sk_buff *skb);
161 ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
162 int offset, size_t size, int flags);
163 struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
164 struct sk_buff_head *queue,
165 struct inet_cork *cork);
166 int ip_send_skb(struct net *net, struct sk_buff *skb);
167 int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
168 void ip_flush_pending_frames(struct sock *sk);
169 struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
170 int getfrag(void *from, char *to, int offset,
171 int len, int odd, struct sk_buff *skb),
172 void *from, int length, int transhdrlen,
173 struct ipcm_cookie *ipc, struct rtable **rtp,
176 static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
178 return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
181 static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
183 return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
186 static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
188 return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
192 int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
193 int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
195 void ip4_datagram_release_cb(struct sock *sk);
197 struct ip_reply_arg {
201 int csumoffset; /* u16 offset of csum in iov[0].iov_base */
202 /* -1 if not needed */
208 #define IP_REPLY_ARG_NOSRCCHECK 1
210 static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
212 return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
215 void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
216 const struct ip_options *sopt,
217 __be32 daddr, __be32 saddr,
218 const struct ip_reply_arg *arg,
221 #define IP_INC_STATS(net, field) SNMP_INC_STATS64((net)->mib.ip_statistics, field)
222 #define __IP_INC_STATS(net, field) __SNMP_INC_STATS64((net)->mib.ip_statistics, field)
223 #define IP_ADD_STATS(net, field, val) SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
224 #define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
225 #define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
226 #define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
227 #define NET_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.net_statistics, field)
228 #define __NET_INC_STATS(net, field) __SNMP_INC_STATS((net)->mib.net_statistics, field)
229 #define NET_ADD_STATS(net, field, adnd) SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
230 #define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
232 u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
233 unsigned long snmp_fold_field(void __percpu *mib, int offt);
234 #if BITS_PER_LONG==32
235 u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
236 size_t syncp_offset);
237 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
239 static inline u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
242 return snmp_get_cpu_field(mib, cpu, offct);
246 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
248 return snmp_fold_field(mib, offt);
252 #define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
255 for_each_possible_cpu(c) { \
256 for (i = 0; stats_list[i].name; i++) \
257 buff64[i] += snmp_get_cpu_field64( \
259 c, stats_list[i].entry, \
264 #define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
267 for_each_possible_cpu(c) { \
268 for (i = 0; stats_list[i].name; i++) \
269 buff[i] += snmp_get_cpu_field( \
271 c, stats_list[i].entry); \
275 void inet_get_local_port_range(struct net *net, int *low, int *high);
278 static inline int inet_is_local_reserved_port(struct net *net, int port)
280 if (!net->ipv4.sysctl_local_reserved_ports)
282 return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
285 static inline bool sysctl_dev_name_is_allowed(const char *name)
287 return strcmp(name, "default") != 0 && strcmp(name, "all") != 0;
290 static inline int inet_prot_sock(struct net *net)
292 return net->ipv4.sysctl_ip_prot_sock;
296 static inline int inet_is_local_reserved_port(struct net *net, int port)
301 static inline int inet_prot_sock(struct net *net)
307 __be32 inet_current_timestamp(void);
309 /* From inetpeer.c */
310 extern int inet_peer_threshold;
311 extern int inet_peer_minttl;
312 extern int inet_peer_maxttl;
314 void ipfrag_init(void);
316 void ip_static_sysctl_init(void);
318 #define IP4_REPLY_MARK(net, mark) \
319 ((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
321 static inline bool ip_is_fragment(const struct iphdr *iph)
323 return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
329 /* The function in 2.2 was invalid, producing wrong result for
330 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
332 int ip_decrease_ttl(struct iphdr *iph)
334 u32 check = (__force u32)iph->check;
335 check += (__force u32)htons(0x0100);
336 iph->check = (__force __sum16)(check + (check>=0xFFFF));
340 static inline int ip_mtu_locked(const struct dst_entry *dst)
342 const struct rtable *rt = (const struct rtable *)dst;
344 return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
348 int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
350 u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
352 return pmtudisc == IP_PMTUDISC_DO ||
353 (pmtudisc == IP_PMTUDISC_WANT &&
354 !ip_mtu_locked(dst));
357 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
359 return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
360 inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
363 static inline bool ip_sk_use_pmtu(const struct sock *sk)
365 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
368 static inline bool ip_sk_ignore_df(const struct sock *sk)
370 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
371 inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
374 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
377 struct net *net = dev_net(dst->dev);
379 if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
380 ip_mtu_locked(dst) ||
384 return min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);
387 static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
388 const struct sk_buff *skb)
390 if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
391 bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
393 return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
396 return min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
399 u32 ip_idents_reserve(u32 hash, int segs);
400 void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
402 static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
403 struct sock *sk, int segs)
405 struct iphdr *iph = ip_hdr(skb);
407 if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
408 /* This is only to work around buggy Windows95/2000
409 * VJ compression implementations. If the ID field
410 * does not change, they drop every other packet in
411 * a TCP stream using header compression.
413 if (sk && inet_sk(sk)->inet_daddr) {
414 iph->id = htons(inet_sk(sk)->inet_id);
415 inet_sk(sk)->inet_id += segs;
420 __ip_select_ident(net, iph, segs);
424 static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
427 ip_select_ident_segs(net, skb, sk, 1);
430 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
432 return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
436 /* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
437 * Equivalent to : flow->v4addrs.src = iph->saddr;
438 * flow->v4addrs.dst = iph->daddr;
440 static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
441 const struct iphdr *iph)
443 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
444 offsetof(typeof(flow->addrs), v4addrs.src) +
445 sizeof(flow->addrs.v4addrs.src));
446 memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
447 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
450 static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
452 const struct iphdr *iph = skb_gro_network_header(skb);
454 return csum_tcpudp_nofold(iph->saddr, iph->daddr,
455 skb_gro_len(skb), proto, 0);
459 * Map a multicast IP onto multicast MAC for type ethernet.
462 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
464 __u32 addr=ntohl(naddr);
476 * Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
477 * Leave P_Key as 0 to be filled in by driver.
480 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
483 unsigned char scope = broadcast[5] & 0xF;
485 buf[0] = 0; /* Reserved */
486 buf[1] = 0xff; /* Multicast QPN */
491 buf[5] = 0x10 | scope; /* scope from broadcast address */
492 buf[6] = 0x40; /* IPv4 signature */
494 buf[8] = broadcast[8]; /* P_Key */
495 buf[9] = broadcast[9];
502 buf[19] = addr & 0xff;
504 buf[18] = addr & 0xff;
506 buf[17] = addr & 0xff;
508 buf[16] = addr & 0x0f;
511 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
513 if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
514 memcpy(buf, broadcast, 4);
516 memcpy(buf, &naddr, sizeof(naddr));
519 #if IS_ENABLED(CONFIG_IPV6)
520 #include <linux/ipv6.h>
523 static __inline__ void inet_reset_saddr(struct sock *sk)
525 inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
526 #if IS_ENABLED(CONFIG_IPV6)
527 if (sk->sk_family == PF_INET6) {
528 struct ipv6_pinfo *np = inet6_sk(sk);
530 memset(&np->saddr, 0, sizeof(np->saddr));
531 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
538 static inline unsigned int ipv4_addr_hash(__be32 ip)
540 return (__force unsigned int) ip;
543 static inline u32 ipv4_portaddr_hash(const struct net *net,
547 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
550 bool ip_call_ra_chain(struct sk_buff *skb);
553 * Functions provided by ip_fragment.c
556 enum ip_defrag_users {
557 IP_DEFRAG_LOCAL_DELIVER,
558 IP_DEFRAG_CALL_RA_CHAIN,
559 IP_DEFRAG_CONNTRACK_IN,
560 __IP_DEFRAG_CONNTRACK_IN_END = IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
561 IP_DEFRAG_CONNTRACK_OUT,
562 __IP_DEFRAG_CONNTRACK_OUT_END = IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
563 IP_DEFRAG_CONNTRACK_BRIDGE_IN,
564 __IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
572 /* Return true if the value of 'user' is between 'lower_bond'
573 * and 'upper_bond' inclusively.
575 static inline bool ip_defrag_user_in_between(u32 user,
576 enum ip_defrag_users lower_bond,
577 enum ip_defrag_users upper_bond)
579 return user >= lower_bond && user <= upper_bond;
582 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
584 struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
586 static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
593 * Functions provided by ip_forward.c
596 int ip_forward(struct sk_buff *skb);
599 * Functions provided by ip_options.c
602 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
603 __be32 daddr, struct rtable *rt, int is_frag);
605 int __ip_options_echo(struct net *net, struct ip_options *dopt,
606 struct sk_buff *skb, const struct ip_options *sopt);
607 static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
610 return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
613 void ip_options_fragment(struct sk_buff *skb);
614 int ip_options_compile(struct net *net, struct ip_options *opt,
615 struct sk_buff *skb);
616 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
617 unsigned char *data, int optlen);
618 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
619 unsigned char __user *data, int optlen);
620 void ip_options_undo(struct ip_options *opt);
621 void ip_forward_options(struct sk_buff *skb);
622 int ip_options_rcv_srr(struct sk_buff *skb);
625 * Functions provided by ip_sockglue.c
628 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
629 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
630 struct sk_buff *skb, int tlen, int offset);
631 int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
632 struct ipcm_cookie *ipc, bool allow_ipv6);
633 int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
634 unsigned int optlen);
635 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
637 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
638 char __user *optval, unsigned int optlen);
639 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
640 char __user *optval, int __user *optlen);
641 int ip_ra_control(struct sock *sk, unsigned char on,
642 void (*destructor)(struct sock *));
644 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
645 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
646 u32 info, u8 *payload);
647 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
650 static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
652 ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
655 bool icmp_global_allow(void);
656 extern int sysctl_icmp_msgs_per_sec;
657 extern int sysctl_icmp_msgs_burst;
659 #ifdef CONFIG_PROC_FS
660 int ip_misc_proc_init(void);