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 static inline void ipcm_init(struct ipcm_cookie *ipcm)
83 *ipcm = (struct ipcm_cookie) { .tos = -1 };
86 static inline void ipcm_init_sk(struct ipcm_cookie *ipcm,
87 const struct inet_sock *inet)
91 ipcm->sockc.tsflags = inet->sk.sk_tsflags;
92 ipcm->oif = inet->sk.sk_bound_dev_if;
93 ipcm->addr = inet->inet_saddr;
96 #define IPCB(skb) ((struct inet_skb_parm*)((skb)->cb))
97 #define PKTINFO_SKB_CB(skb) ((struct in_pktinfo *)((skb)->cb))
99 /* return enslaved device index if relevant */
100 static inline int inet_sdif(struct sk_buff *skb)
102 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
103 if (skb && ipv4_l3mdev_skb(IPCB(skb)->flags))
104 return IPCB(skb)->iif;
109 /* Special input handler for packets caught by router alert option.
110 They are selected only by protocol field, and then processed likely
111 local ones; but only if someone wants them! Otherwise, router
112 not running rsvpd will kill RSVP.
114 It is user level problem, what it will make with them.
115 I have no idea, how it will masquearde or NAT them (it is joke, joke :-)),
116 but receiver should be enough clever f.e. to forward mtrace requests,
117 sent to multicast group to reach destination designated router.
121 struct ip_ra_chain __rcu *next;
124 void (*destructor)(struct sock *);
125 struct sock *saved_sk;
131 #define IP_CE 0x8000 /* Flag: "Congestion" */
132 #define IP_DF 0x4000 /* Flag: "Don't Fragment" */
133 #define IP_MF 0x2000 /* Flag: "More Fragments" */
134 #define IP_OFFSET 0x1FFF /* "Fragment Offset" part */
136 #define IP_FRAG_TIME (30 * HZ) /* fragment lifetime */
144 int igmp_mc_init(void);
147 * Functions provided by ip.c
150 int ip_build_and_send_pkt(struct sk_buff *skb, const struct sock *sk,
151 __be32 saddr, __be32 daddr,
152 struct ip_options_rcu *opt);
153 int ip_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt,
154 struct net_device *orig_dev);
155 void ip_list_rcv(struct list_head *head, struct packet_type *pt,
156 struct net_device *orig_dev);
157 int ip_local_deliver(struct sk_buff *skb);
158 int ip_mr_input(struct sk_buff *skb);
159 int ip_output(struct net *net, struct sock *sk, struct sk_buff *skb);
160 int ip_mc_output(struct net *net, struct sock *sk, struct sk_buff *skb);
161 int ip_do_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
162 int (*output)(struct net *, struct sock *, struct sk_buff *));
163 void ip_send_check(struct iphdr *ip);
164 int __ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
165 int ip_local_out(struct net *net, struct sock *sk, struct sk_buff *skb);
167 int __ip_queue_xmit(struct sock *sk, struct sk_buff *skb, struct flowi *fl,
170 int ip_append_data(struct sock *sk, struct flowi4 *fl4,
171 int getfrag(void *from, char *to, int offset, int len,
172 int odd, struct sk_buff *skb),
173 void *from, int len, int protolen,
174 struct ipcm_cookie *ipc,
177 int ip_generic_getfrag(void *from, char *to, int offset, int len, int odd,
178 struct sk_buff *skb);
179 ssize_t ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
180 int offset, size_t size, int flags);
181 struct sk_buff *__ip_make_skb(struct sock *sk, struct flowi4 *fl4,
182 struct sk_buff_head *queue,
183 struct inet_cork *cork);
184 int ip_send_skb(struct net *net, struct sk_buff *skb);
185 int ip_push_pending_frames(struct sock *sk, struct flowi4 *fl4);
186 void ip_flush_pending_frames(struct sock *sk);
187 struct sk_buff *ip_make_skb(struct sock *sk, struct flowi4 *fl4,
188 int getfrag(void *from, char *to, int offset,
189 int len, int odd, struct sk_buff *skb),
190 void *from, int length, int transhdrlen,
191 struct ipcm_cookie *ipc, struct rtable **rtp,
192 struct inet_cork *cork, unsigned int flags);
194 static inline int ip_queue_xmit(struct sock *sk, struct sk_buff *skb,
197 return __ip_queue_xmit(sk, skb, fl, inet_sk(sk)->tos);
200 static inline struct sk_buff *ip_finish_skb(struct sock *sk, struct flowi4 *fl4)
202 return __ip_make_skb(sk, fl4, &sk->sk_write_queue, &inet_sk(sk)->cork.base);
205 static inline __u8 get_rttos(struct ipcm_cookie* ipc, struct inet_sock *inet)
207 return (ipc->tos != -1) ? RT_TOS(ipc->tos) : RT_TOS(inet->tos);
210 static inline __u8 get_rtconn_flags(struct ipcm_cookie* ipc, struct sock* sk)
212 return (ipc->tos != -1) ? RT_CONN_FLAGS_TOS(sk, ipc->tos) : RT_CONN_FLAGS(sk);
216 int __ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
217 int ip4_datagram_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
219 void ip4_datagram_release_cb(struct sock *sk);
221 struct ip_reply_arg {
225 int csumoffset; /* u16 offset of csum in iov[0].iov_base */
226 /* -1 if not needed */
232 #define IP_REPLY_ARG_NOSRCCHECK 1
234 static inline __u8 ip_reply_arg_flowi_flags(const struct ip_reply_arg *arg)
236 return (arg->flags & IP_REPLY_ARG_NOSRCCHECK) ? FLOWI_FLAG_ANYSRC : 0;
239 void ip_send_unicast_reply(struct sock *sk, struct sk_buff *skb,
240 const struct ip_options *sopt,
241 __be32 daddr, __be32 saddr,
242 const struct ip_reply_arg *arg,
245 #define IP_INC_STATS(net, field) SNMP_INC_STATS64((net)->mib.ip_statistics, field)
246 #define __IP_INC_STATS(net, field) __SNMP_INC_STATS64((net)->mib.ip_statistics, field)
247 #define IP_ADD_STATS(net, field, val) SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
248 #define __IP_ADD_STATS(net, field, val) __SNMP_ADD_STATS64((net)->mib.ip_statistics, field, val)
249 #define IP_UPD_PO_STATS(net, field, val) SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
250 #define __IP_UPD_PO_STATS(net, field, val) __SNMP_UPD_PO_STATS64((net)->mib.ip_statistics, field, val)
251 #define NET_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.net_statistics, field)
252 #define __NET_INC_STATS(net, field) __SNMP_INC_STATS((net)->mib.net_statistics, field)
253 #define NET_ADD_STATS(net, field, adnd) SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
254 #define __NET_ADD_STATS(net, field, adnd) __SNMP_ADD_STATS((net)->mib.net_statistics, field, adnd)
256 u64 snmp_get_cpu_field(void __percpu *mib, int cpu, int offct);
257 unsigned long snmp_fold_field(void __percpu *mib, int offt);
258 #if BITS_PER_LONG==32
259 u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
260 size_t syncp_offset);
261 u64 snmp_fold_field64(void __percpu *mib, int offt, size_t sync_off);
263 static inline u64 snmp_get_cpu_field64(void __percpu *mib, int cpu, int offct,
266 return snmp_get_cpu_field(mib, cpu, offct);
270 static inline u64 snmp_fold_field64(void __percpu *mib, int offt, size_t syncp_off)
272 return snmp_fold_field(mib, offt);
276 #define snmp_get_cpu_field64_batch(buff64, stats_list, mib_statistic, offset) \
279 for_each_possible_cpu(c) { \
280 for (i = 0; stats_list[i].name; i++) \
281 buff64[i] += snmp_get_cpu_field64( \
283 c, stats_list[i].entry, \
288 #define snmp_get_cpu_field_batch(buff, stats_list, mib_statistic) \
291 for_each_possible_cpu(c) { \
292 for (i = 0; stats_list[i].name; i++) \
293 buff[i] += snmp_get_cpu_field( \
295 c, stats_list[i].entry); \
299 void inet_get_local_port_range(struct net *net, int *low, int *high);
302 static inline int inet_is_local_reserved_port(struct net *net, int port)
304 if (!net->ipv4.sysctl_local_reserved_ports)
306 return test_bit(port, net->ipv4.sysctl_local_reserved_ports);
309 static inline bool sysctl_dev_name_is_allowed(const char *name)
311 return strcmp(name, "default") != 0 && strcmp(name, "all") != 0;
314 static inline int inet_prot_sock(struct net *net)
316 return net->ipv4.sysctl_ip_prot_sock;
320 static inline int inet_is_local_reserved_port(struct net *net, int port)
325 static inline int inet_prot_sock(struct net *net)
331 __be32 inet_current_timestamp(void);
333 /* From inetpeer.c */
334 extern int inet_peer_threshold;
335 extern int inet_peer_minttl;
336 extern int inet_peer_maxttl;
338 void ipfrag_init(void);
340 void ip_static_sysctl_init(void);
342 #define IP4_REPLY_MARK(net, mark) \
343 ((net)->ipv4.sysctl_fwmark_reflect ? (mark) : 0)
345 static inline bool ip_is_fragment(const struct iphdr *iph)
347 return (iph->frag_off & htons(IP_MF | IP_OFFSET)) != 0;
353 /* The function in 2.2 was invalid, producing wrong result for
354 * check=0xFEFF. It was noticed by Arthur Skawina _year_ ago. --ANK(000625) */
356 int ip_decrease_ttl(struct iphdr *iph)
358 u32 check = (__force u32)iph->check;
359 check += (__force u32)htons(0x0100);
360 iph->check = (__force __sum16)(check + (check>=0xFFFF));
364 static inline int ip_mtu_locked(const struct dst_entry *dst)
366 const struct rtable *rt = (const struct rtable *)dst;
368 return rt->rt_mtu_locked || dst_metric_locked(dst, RTAX_MTU);
372 int ip_dont_fragment(const struct sock *sk, const struct dst_entry *dst)
374 u8 pmtudisc = READ_ONCE(inet_sk(sk)->pmtudisc);
376 return pmtudisc == IP_PMTUDISC_DO ||
377 (pmtudisc == IP_PMTUDISC_WANT &&
378 !ip_mtu_locked(dst));
381 static inline bool ip_sk_accept_pmtu(const struct sock *sk)
383 return inet_sk(sk)->pmtudisc != IP_PMTUDISC_INTERFACE &&
384 inet_sk(sk)->pmtudisc != IP_PMTUDISC_OMIT;
387 static inline bool ip_sk_use_pmtu(const struct sock *sk)
389 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_PROBE;
392 static inline bool ip_sk_ignore_df(const struct sock *sk)
394 return inet_sk(sk)->pmtudisc < IP_PMTUDISC_DO ||
395 inet_sk(sk)->pmtudisc == IP_PMTUDISC_OMIT;
398 static inline unsigned int ip_dst_mtu_maybe_forward(const struct dst_entry *dst,
401 struct net *net = dev_net(dst->dev);
403 if (net->ipv4.sysctl_ip_fwd_use_pmtu ||
404 ip_mtu_locked(dst) ||
408 return min(READ_ONCE(dst->dev->mtu), IP_MAX_MTU);
411 static inline unsigned int ip_skb_dst_mtu(struct sock *sk,
412 const struct sk_buff *skb)
414 if (!sk || !sk_fullsock(sk) || ip_sk_use_pmtu(sk)) {
415 bool forwarding = IPCB(skb)->flags & IPSKB_FORWARDED;
417 return ip_dst_mtu_maybe_forward(skb_dst(skb), forwarding);
420 return min(READ_ONCE(skb_dst(skb)->dev->mtu), IP_MAX_MTU);
423 int ip_metrics_convert(struct net *net, struct nlattr *fc_mx, int fc_mx_len,
426 u32 ip_idents_reserve(u32 hash, int segs);
427 void __ip_select_ident(struct net *net, struct iphdr *iph, int segs);
429 static inline void ip_select_ident_segs(struct net *net, struct sk_buff *skb,
430 struct sock *sk, int segs)
432 struct iphdr *iph = ip_hdr(skb);
434 if ((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) {
435 /* This is only to work around buggy Windows95/2000
436 * VJ compression implementations. If the ID field
437 * does not change, they drop every other packet in
438 * a TCP stream using header compression.
440 if (sk && inet_sk(sk)->inet_daddr) {
441 iph->id = htons(inet_sk(sk)->inet_id);
442 inet_sk(sk)->inet_id += segs;
447 __ip_select_ident(net, iph, segs);
451 static inline void ip_select_ident(struct net *net, struct sk_buff *skb,
454 ip_select_ident_segs(net, skb, sk, 1);
457 static inline __wsum inet_compute_pseudo(struct sk_buff *skb, int proto)
459 return csum_tcpudp_nofold(ip_hdr(skb)->saddr, ip_hdr(skb)->daddr,
463 /* copy IPv4 saddr & daddr to flow_keys, possibly using 64bit load/store
464 * Equivalent to : flow->v4addrs.src = iph->saddr;
465 * flow->v4addrs.dst = iph->daddr;
467 static inline void iph_to_flow_copy_v4addrs(struct flow_keys *flow,
468 const struct iphdr *iph)
470 BUILD_BUG_ON(offsetof(typeof(flow->addrs), v4addrs.dst) !=
471 offsetof(typeof(flow->addrs), v4addrs.src) +
472 sizeof(flow->addrs.v4addrs.src));
473 memcpy(&flow->addrs.v4addrs, &iph->saddr, sizeof(flow->addrs.v4addrs));
474 flow->control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
477 static inline __wsum inet_gro_compute_pseudo(struct sk_buff *skb, int proto)
479 const struct iphdr *iph = skb_gro_network_header(skb);
481 return csum_tcpudp_nofold(iph->saddr, iph->daddr,
482 skb_gro_len(skb), proto, 0);
486 * Map a multicast IP onto multicast MAC for type ethernet.
489 static inline void ip_eth_mc_map(__be32 naddr, char *buf)
491 __u32 addr=ntohl(naddr);
503 * Map a multicast IP onto multicast MAC for type IP-over-InfiniBand.
504 * Leave P_Key as 0 to be filled in by driver.
507 static inline void ip_ib_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
510 unsigned char scope = broadcast[5] & 0xF;
512 buf[0] = 0; /* Reserved */
513 buf[1] = 0xff; /* Multicast QPN */
518 buf[5] = 0x10 | scope; /* scope from broadcast address */
519 buf[6] = 0x40; /* IPv4 signature */
521 buf[8] = broadcast[8]; /* P_Key */
522 buf[9] = broadcast[9];
529 buf[19] = addr & 0xff;
531 buf[18] = addr & 0xff;
533 buf[17] = addr & 0xff;
535 buf[16] = addr & 0x0f;
538 static inline void ip_ipgre_mc_map(__be32 naddr, const unsigned char *broadcast, char *buf)
540 if ((broadcast[0] | broadcast[1] | broadcast[2] | broadcast[3]) != 0)
541 memcpy(buf, broadcast, 4);
543 memcpy(buf, &naddr, sizeof(naddr));
546 #if IS_ENABLED(CONFIG_IPV6)
547 #include <linux/ipv6.h>
550 static __inline__ void inet_reset_saddr(struct sock *sk)
552 inet_sk(sk)->inet_rcv_saddr = inet_sk(sk)->inet_saddr = 0;
553 #if IS_ENABLED(CONFIG_IPV6)
554 if (sk->sk_family == PF_INET6) {
555 struct ipv6_pinfo *np = inet6_sk(sk);
557 memset(&np->saddr, 0, sizeof(np->saddr));
558 memset(&sk->sk_v6_rcv_saddr, 0, sizeof(sk->sk_v6_rcv_saddr));
565 static inline unsigned int ipv4_addr_hash(__be32 ip)
567 return (__force unsigned int) ip;
570 static inline u32 ipv4_portaddr_hash(const struct net *net,
574 return jhash_1word((__force u32)saddr, net_hash_mix(net)) ^ port;
577 bool ip_call_ra_chain(struct sk_buff *skb);
580 * Functions provided by ip_fragment.c
583 enum ip_defrag_users {
584 IP_DEFRAG_LOCAL_DELIVER,
585 IP_DEFRAG_CALL_RA_CHAIN,
586 IP_DEFRAG_CONNTRACK_IN,
587 __IP_DEFRAG_CONNTRACK_IN_END = IP_DEFRAG_CONNTRACK_IN + USHRT_MAX,
588 IP_DEFRAG_CONNTRACK_OUT,
589 __IP_DEFRAG_CONNTRACK_OUT_END = IP_DEFRAG_CONNTRACK_OUT + USHRT_MAX,
590 IP_DEFRAG_CONNTRACK_BRIDGE_IN,
591 __IP_DEFRAG_CONNTRACK_BRIDGE_IN = IP_DEFRAG_CONNTRACK_BRIDGE_IN + USHRT_MAX,
599 /* Return true if the value of 'user' is between 'lower_bond'
600 * and 'upper_bond' inclusively.
602 static inline bool ip_defrag_user_in_between(u32 user,
603 enum ip_defrag_users lower_bond,
604 enum ip_defrag_users upper_bond)
606 return user >= lower_bond && user <= upper_bond;
609 int ip_defrag(struct net *net, struct sk_buff *skb, u32 user);
611 struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user);
613 static inline struct sk_buff *ip_check_defrag(struct net *net, struct sk_buff *skb, u32 user)
620 * Functions provided by ip_forward.c
623 int ip_forward(struct sk_buff *skb);
626 * Functions provided by ip_options.c
629 void ip_options_build(struct sk_buff *skb, struct ip_options *opt,
630 __be32 daddr, struct rtable *rt, int is_frag);
632 int __ip_options_echo(struct net *net, struct ip_options *dopt,
633 struct sk_buff *skb, const struct ip_options *sopt);
634 static inline int ip_options_echo(struct net *net, struct ip_options *dopt,
637 return __ip_options_echo(net, dopt, skb, &IPCB(skb)->opt);
640 void ip_options_fragment(struct sk_buff *skb);
641 int ip_options_compile(struct net *net, struct ip_options *opt,
642 struct sk_buff *skb);
643 int ip_options_get(struct net *net, struct ip_options_rcu **optp,
644 unsigned char *data, int optlen);
645 int ip_options_get_from_user(struct net *net, struct ip_options_rcu **optp,
646 unsigned char __user *data, int optlen);
647 void ip_options_undo(struct ip_options *opt);
648 void ip_forward_options(struct sk_buff *skb);
649 int ip_options_rcv_srr(struct sk_buff *skb);
652 * Functions provided by ip_sockglue.c
655 void ipv4_pktinfo_prepare(const struct sock *sk, struct sk_buff *skb);
656 void ip_cmsg_recv_offset(struct msghdr *msg, struct sock *sk,
657 struct sk_buff *skb, int tlen, int offset);
658 int ip_cmsg_send(struct sock *sk, struct msghdr *msg,
659 struct ipcm_cookie *ipc, bool allow_ipv6);
660 int ip_setsockopt(struct sock *sk, int level, int optname, char __user *optval,
661 unsigned int optlen);
662 int ip_getsockopt(struct sock *sk, int level, int optname, char __user *optval,
664 int compat_ip_setsockopt(struct sock *sk, int level, int optname,
665 char __user *optval, unsigned int optlen);
666 int compat_ip_getsockopt(struct sock *sk, int level, int optname,
667 char __user *optval, int __user *optlen);
668 int ip_ra_control(struct sock *sk, unsigned char on,
669 void (*destructor)(struct sock *));
671 int ip_recv_error(struct sock *sk, struct msghdr *msg, int len, int *addr_len);
672 void ip_icmp_error(struct sock *sk, struct sk_buff *skb, int err, __be16 port,
673 u32 info, u8 *payload);
674 void ip_local_error(struct sock *sk, int err, __be32 daddr, __be16 dport,
677 static inline void ip_cmsg_recv(struct msghdr *msg, struct sk_buff *skb)
679 ip_cmsg_recv_offset(msg, skb->sk, skb, 0, 0);
682 bool icmp_global_allow(void);
683 extern int sysctl_icmp_msgs_per_sec;
684 extern int sysctl_icmp_msgs_burst;
686 #ifdef CONFIG_PROC_FS
687 int ip_misc_proc_init(void);
690 int rtm_getroute_parse_ip_proto(struct nlattr *attr, u8 *ip_proto,
691 struct netlink_ext_ack *extack);