1/*
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.
5 *
6 * Definitions for inet_sock
7 *
8 * Authors: Many, reorganised here by
9 * Arnaldo Carvalho de Melo <acme@mandriva.com>
10 *
11 * This program is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU General Public License
13 * as published by the Free Software Foundation; either version
14 * 2 of the License, or (at your option) any later version.
15 */
16#ifndef _INET_SOCK_H
17#define _INET_SOCK_H
18
19#include <linux/bitops.h>
20#include <linux/string.h>
21#include <linux/types.h>
22#include <linux/jhash.h>
23#include <linux/netdevice.h>
24
25#include <net/flow.h>
26#include <net/sock.h>
27#include <net/request_sock.h>
28#include <net/netns/hash.h>
29#include <net/tcp_states.h>
30#include <net/l3mdev.h>
31
32/** struct ip_options - IP Options
33 *
34 * @faddr - Saved first hop address
35 * @nexthop - Saved nexthop address in LSRR and SSRR
36 * @is_strictroute - Strict source route
37 * @srr_is_hit - Packet destination addr was our one
38 * @is_changed - IP checksum more not valid
39 * @rr_needaddr - Need to record addr of outgoing dev
40 * @ts_needtime - Need to record timestamp
41 * @ts_needaddr - Need to record addr of outgoing dev
42 */
43struct ip_options {
44 __be32 faddr;
45 __be32 nexthop;
46 unsigned char optlen;
47 unsigned char srr;
48 unsigned char rr;
49 unsigned char ts;
50 unsigned char is_strictroute:1,
51 srr_is_hit:1,
52 is_changed:1,
53 rr_needaddr:1,
54 ts_needtime:1,
55 ts_needaddr:1;
56 unsigned char router_alert;
57 unsigned char cipso;
58 unsigned char __pad2;
59 unsigned char __data[0];
60};
61
62struct ip_options_rcu {
63 struct rcu_head rcu;
64 struct ip_options opt;
65};
66
67struct ip_options_data {
68 struct ip_options_rcu opt;
69 char data[40];
70};
71
72struct inet_request_sock {
73 struct request_sock req;
74#define ir_loc_addr req.__req_common.skc_rcv_saddr
75#define ir_rmt_addr req.__req_common.skc_daddr
76#define ir_num req.__req_common.skc_num
77#define ir_rmt_port req.__req_common.skc_dport
78#define ir_v6_rmt_addr req.__req_common.skc_v6_daddr
79#define ir_v6_loc_addr req.__req_common.skc_v6_rcv_saddr
80#define ir_iif req.__req_common.skc_bound_dev_if
81#define ir_cookie req.__req_common.skc_cookie
82#define ireq_net req.__req_common.skc_net
83#define ireq_state req.__req_common.skc_state
84#define ireq_family req.__req_common.skc_family
85
86 u16 snd_wscale : 4,
87 rcv_wscale : 4,
88 tstamp_ok : 1,
89 sack_ok : 1,
90 wscale_ok : 1,
91 ecn_ok : 1,
92 acked : 1,
93 no_srccheck: 1,
94 smc_ok : 1;
95 u32 ir_mark;
96 union {
97 struct ip_options_rcu __rcu *ireq_opt;
98#if IS_ENABLED(CONFIG_IPV6)
99 struct {
100 struct ipv6_txoptions *ipv6_opt;
101 struct sk_buff *pktopts;
102 };
103#endif
104 };
105};
106
107static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
108{
109 return (struct inet_request_sock *)sk;
110}
111
112static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
113{
114 if (!sk->sk_mark && sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept)
115 return skb->mark;
116
117 return sk->sk_mark;
118}
119
120static inline int inet_request_bound_dev_if(const struct sock *sk,
121 struct sk_buff *skb)
122{
123#ifdef CONFIG_NET_L3_MASTER_DEV
124 struct net *net = sock_net(sk);
125
126 if (!sk->sk_bound_dev_if && net->ipv4.sysctl_tcp_l3mdev_accept)
127 return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
128#endif
129
130 return sk->sk_bound_dev_if;
131}
132
133static inline int inet_sk_bound_l3mdev(const struct sock *sk)
134{
135#ifdef CONFIG_NET_L3_MASTER_DEV
136 struct net *net = sock_net(sk);
137
138 if (!net->ipv4.sysctl_tcp_l3mdev_accept)
139 return l3mdev_master_ifindex_by_index(net,
140 sk->sk_bound_dev_if);
141#endif
142
143 return 0;
144}
145
146static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
147 int dif, int sdif)
148{
149 if (!bound_dev_if)
150 return !sdif || l3mdev_accept;
151 return bound_dev_if == dif || bound_dev_if == sdif;
152}
153
154struct inet_cork {
155 unsigned int flags;
156 __be32 addr;
157 struct ip_options *opt;
158 unsigned int fragsize;
159 int length; /* Total length of all frames */
160 struct dst_entry *dst;
161 u8 tx_flags;
162 __u8 ttl;
163 __s16 tos;
164 char priority;
165 __u16 gso_size;
166 u64 transmit_time;
167};
168
169struct inet_cork_full {
170 struct inet_cork base;
171 struct flowi fl;
172};
173
174struct ip_mc_socklist;
175struct ipv6_pinfo;
176struct rtable;
177
178/** struct inet_sock - representation of INET sockets
179 *
180 * @sk - ancestor class
181 * @pinet6 - pointer to IPv6 control block
182 * @inet_daddr - Foreign IPv4 addr
183 * @inet_rcv_saddr - Bound local IPv4 addr
184 * @inet_dport - Destination port
185 * @inet_num - Local port
186 * @inet_saddr - Sending source
187 * @uc_ttl - Unicast TTL
188 * @inet_sport - Source port
189 * @inet_id - ID counter for DF pkts
190 * @tos - TOS
191 * @mc_ttl - Multicasting TTL
192 * @is_icsk - is this an inet_connection_sock?
193 * @uc_index - Unicast outgoing device index
194 * @mc_index - Multicast device index
195 * @mc_list - Group array
196 * @cork - info to build ip hdr on each ip frag while socket is corked
197 */
198struct inet_sock {
199 /* sk and pinet6 has to be the first two members of inet_sock */
200 struct sock sk;
201#if IS_ENABLED(CONFIG_IPV6)
202 struct ipv6_pinfo *pinet6;
203#endif
204 /* Socket demultiplex comparisons on incoming packets. */
205#define inet_daddr sk.__sk_common.skc_daddr
206#define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr
207#define inet_dport sk.__sk_common.skc_dport
208#define inet_num sk.__sk_common.skc_num
209
210 __be32 inet_saddr;
211 __s16 uc_ttl;
212 __u16 cmsg_flags;
213 __be16 inet_sport;
214 __u16 inet_id;
215
216 struct ip_options_rcu __rcu *inet_opt;
217 int rx_dst_ifindex;
218 __u8 tos;
219 __u8 min_ttl;
220 __u8 mc_ttl;
221 __u8 pmtudisc;
222 __u8 recverr:1,
223 is_icsk:1,
224 freebind:1,
225 hdrincl:1,
226 mc_loop:1,
227 transparent:1,
228 mc_all:1,
229 nodefrag:1;
230 __u8 bind_address_no_port:1,
231 defer_connect:1; /* Indicates that fastopen_connect is set
232 * and cookie exists so we defer connect
233 * until first data frame is written
234 */
235 __u8 rcv_tos;
236 __u8 convert_csum;
237 int uc_index;
238 int mc_index;
239 __be32 mc_addr;
240 struct ip_mc_socklist __rcu *mc_list;
241 struct inet_cork_full cork;
242};
243
244#define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */
245#define IPCORK_ALLFRAG 2 /* always fragment (for ipv6 for now) */
246
247/* cmsg flags for inet */
248#define IP_CMSG_PKTINFO BIT(0)
249#define IP_CMSG_TTL BIT(1)
250#define IP_CMSG_TOS BIT(2)
251#define IP_CMSG_RECVOPTS BIT(3)
252#define IP_CMSG_RETOPTS BIT(4)
253#define IP_CMSG_PASSSEC BIT(5)
254#define IP_CMSG_ORIGDSTADDR BIT(6)
255#define IP_CMSG_CHECKSUM BIT(7)
256#define IP_CMSG_RECVFRAGSIZE BIT(8)
257
258/**
259 * sk_to_full_sk - Access to a full socket
260 * @sk: pointer to a socket
261 *
262 * SYNACK messages might be attached to request sockets.
263 * Some places want to reach the listener in this case.
264 */
265static inline struct sock *sk_to_full_sk(struct sock *sk)
266{
267#ifdef CONFIG_INET
268 if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
269 sk = inet_reqsk(sk)->rsk_listener;
270#endif
271 return sk;
272}
273
274/* sk_to_full_sk() variant with a const argument */
275static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
276{
277#ifdef CONFIG_INET
278 if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
279 sk = ((const struct request_sock *)sk)->rsk_listener;
280#endif
281 return sk;
282}
283
284static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
285{
286 return sk_to_full_sk(skb->sk);
287}
288
289static inline struct inet_sock *inet_sk(const struct sock *sk)
290{
291 return (struct inet_sock *)sk;
292}
293
294static inline void __inet_sk_copy_descendant(struct sock *sk_to,
295 const struct sock *sk_from,
296 const int ancestor_size)
297{
298 memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
299 sk_from->sk_prot->obj_size - ancestor_size);
300}
301#if !(IS_ENABLED(CONFIG_IPV6))
302static inline void inet_sk_copy_descendant(struct sock *sk_to,
303 const struct sock *sk_from)
304{
305 __inet_sk_copy_descendant(sk_to, sk_from, sizeof(struct inet_sock));
306}
307#endif
308
309int inet_sk_rebuild_header(struct sock *sk);
310
311/**
312 * inet_sk_state_load - read sk->sk_state for lockless contexts
313 * @sk: socket pointer
314 *
315 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
316 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
317 */
318static inline int inet_sk_state_load(const struct sock *sk)
319{
320 /* state change might impact lockless readers. */
321 return smp_load_acquire(&sk->sk_state);
322}
323
324/**
325 * inet_sk_state_store - update sk->sk_state
326 * @sk: socket pointer
327 * @newstate: new state
328 *
329 * Paired with inet_sk_state_load(). Should be used in contexts where
330 * state change might impact lockless readers.
331 */
332void inet_sk_state_store(struct sock *sk, int newstate);
333
334void inet_sk_set_state(struct sock *sk, int state);
335
336static inline unsigned int __inet_ehashfn(const __be32 laddr,
337 const __u16 lport,
338 const __be32 faddr,
339 const __be16 fport,
340 u32 initval)
341{
342 return jhash_3words((__force __u32) laddr,
343 (__force __u32) faddr,
344 ((__u32) lport) << 16 | (__force __u32)fport,
345 initval);
346}
347
348struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
349 struct sock *sk_listener,
350 bool attach_listener);
351
352static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
353{
354 __u8 flags = 0;
355
356 if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl)
357 flags |= FLOWI_FLAG_ANYSRC;
358 return flags;
359}
360
361static inline void inet_inc_convert_csum(struct sock *sk)
362{
363 inet_sk(sk)->convert_csum++;
364}
365
366static inline void inet_dec_convert_csum(struct sock *sk)
367{
368 if (inet_sk(sk)->convert_csum > 0)
369 inet_sk(sk)->convert_csum--;
370}
371
372static inline bool inet_get_convert_csum(struct sock *sk)
373{
374 return !!inet_sk(sk)->convert_csum;
375}
376
377
378static inline bool inet_can_nonlocal_bind(struct net *net,
379 struct inet_sock *inet)
380{
381 return net->ipv4.sysctl_ip_nonlocal_bind ||
382 inet->freebind || inet->transparent;
383}
384
385#endif /* _INET_SOCK_H */
386