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 */
43 struct 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
62 struct ip_options_rcu {
63 struct rcu_head rcu;
64 struct ip_options opt;
65 };
66
67 struct ip_options_data {
68 struct ip_options_rcu opt;
69 char data[40];
70 };
71
72 struct 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
inet_rsk(const struct request_sock * sk)107 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
108 {
109 return (struct inet_request_sock *)sk;
110 }
111
inet_request_mark(const struct sock * sk,struct sk_buff * skb)112 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
113 {
114 if (!sk->sk_mark &&
115 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept))
116 return skb->mark;
117
118 return sk->sk_mark;
119 }
120
inet_request_bound_dev_if(const struct sock * sk,struct sk_buff * skb)121 static inline int inet_request_bound_dev_if(const struct sock *sk,
122 struct sk_buff *skb)
123 {
124 #ifdef CONFIG_NET_L3_MASTER_DEV
125 struct net *net = sock_net(sk);
126
127 if (!sk->sk_bound_dev_if && net->ipv4.sysctl_tcp_l3mdev_accept)
128 return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
129 #endif
130
131 return sk->sk_bound_dev_if;
132 }
133
134 struct inet_cork {
135 unsigned int flags;
136 __be32 addr;
137 struct ip_options *opt;
138 unsigned int fragsize;
139 int length; /* Total length of all frames */
140 struct dst_entry *dst;
141 u8 tx_flags;
142 __u8 ttl;
143 __s16 tos;
144 char priority;
145 __u16 gso_size;
146 u64 transmit_time;
147 };
148
149 struct inet_cork_full {
150 struct inet_cork base;
151 struct flowi fl;
152 };
153
154 struct ip_mc_socklist;
155 struct ipv6_pinfo;
156 struct rtable;
157
158 /** struct inet_sock - representation of INET sockets
159 *
160 * @sk - ancestor class
161 * @pinet6 - pointer to IPv6 control block
162 * @inet_daddr - Foreign IPv4 addr
163 * @inet_rcv_saddr - Bound local IPv4 addr
164 * @inet_dport - Destination port
165 * @inet_num - Local port
166 * @inet_saddr - Sending source
167 * @uc_ttl - Unicast TTL
168 * @inet_sport - Source port
169 * @inet_id - ID counter for DF pkts
170 * @tos - TOS
171 * @mc_ttl - Multicasting TTL
172 * @is_icsk - is this an inet_connection_sock?
173 * @uc_index - Unicast outgoing device index
174 * @mc_index - Multicast device index
175 * @mc_list - Group array
176 * @cork - info to build ip hdr on each ip frag while socket is corked
177 */
178 struct inet_sock {
179 /* sk and pinet6 has to be the first two members of inet_sock */
180 struct sock sk;
181 #if IS_ENABLED(CONFIG_IPV6)
182 struct ipv6_pinfo *pinet6;
183 #endif
184 /* Socket demultiplex comparisons on incoming packets. */
185 #define inet_daddr sk.__sk_common.skc_daddr
186 #define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr
187 #define inet_dport sk.__sk_common.skc_dport
188 #define inet_num sk.__sk_common.skc_num
189
190 __be32 inet_saddr;
191 __s16 uc_ttl;
192 __u16 cmsg_flags;
193 __be16 inet_sport;
194 __u16 inet_id;
195
196 struct ip_options_rcu __rcu *inet_opt;
197 int rx_dst_ifindex;
198 __u8 tos;
199 __u8 min_ttl;
200 __u8 mc_ttl;
201 __u8 pmtudisc;
202 __u8 recverr:1,
203 is_icsk:1,
204 freebind:1,
205 hdrincl:1,
206 mc_loop:1,
207 transparent:1,
208 mc_all:1,
209 nodefrag:1;
210 __u8 bind_address_no_port:1,
211 defer_connect:1; /* Indicates that fastopen_connect is set
212 * and cookie exists so we defer connect
213 * until first data frame is written
214 */
215 __u8 rcv_tos;
216 __u8 convert_csum;
217 int uc_index;
218 int mc_index;
219 __be32 mc_addr;
220 struct ip_mc_socklist __rcu *mc_list;
221 struct inet_cork_full cork;
222 };
223
224 #define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */
225 #define IPCORK_ALLFRAG 2 /* always fragment (for ipv6 for now) */
226
227 /* cmsg flags for inet */
228 #define IP_CMSG_PKTINFO BIT(0)
229 #define IP_CMSG_TTL BIT(1)
230 #define IP_CMSG_TOS BIT(2)
231 #define IP_CMSG_RECVOPTS BIT(3)
232 #define IP_CMSG_RETOPTS BIT(4)
233 #define IP_CMSG_PASSSEC BIT(5)
234 #define IP_CMSG_ORIGDSTADDR BIT(6)
235 #define IP_CMSG_CHECKSUM BIT(7)
236 #define IP_CMSG_RECVFRAGSIZE BIT(8)
237
238 /**
239 * sk_to_full_sk - Access to a full socket
240 * @sk: pointer to a socket
241 *
242 * SYNACK messages might be attached to request sockets.
243 * Some places want to reach the listener in this case.
244 */
sk_to_full_sk(struct sock * sk)245 static inline struct sock *sk_to_full_sk(struct sock *sk)
246 {
247 #ifdef CONFIG_INET
248 if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
249 sk = inet_reqsk(sk)->rsk_listener;
250 #endif
251 return sk;
252 }
253
254 /* sk_to_full_sk() variant with a const argument */
sk_const_to_full_sk(const struct sock * sk)255 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
256 {
257 #ifdef CONFIG_INET
258 if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
259 sk = ((const struct request_sock *)sk)->rsk_listener;
260 #endif
261 return sk;
262 }
263
skb_to_full_sk(const struct sk_buff * skb)264 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
265 {
266 return sk_to_full_sk(skb->sk);
267 }
268
inet_sk(const struct sock * sk)269 static inline struct inet_sock *inet_sk(const struct sock *sk)
270 {
271 return (struct inet_sock *)sk;
272 }
273
__inet_sk_copy_descendant(struct sock * sk_to,const struct sock * sk_from,const int ancestor_size)274 static inline void __inet_sk_copy_descendant(struct sock *sk_to,
275 const struct sock *sk_from,
276 const int ancestor_size)
277 {
278 memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
279 sk_from->sk_prot->obj_size - ancestor_size);
280 }
281 #if !(IS_ENABLED(CONFIG_IPV6))
inet_sk_copy_descendant(struct sock * sk_to,const struct sock * sk_from)282 static inline void inet_sk_copy_descendant(struct sock *sk_to,
283 const struct sock *sk_from)
284 {
285 __inet_sk_copy_descendant(sk_to, sk_from, sizeof(struct inet_sock));
286 }
287 #endif
288
289 int inet_sk_rebuild_header(struct sock *sk);
290
291 /**
292 * inet_sk_state_load - read sk->sk_state for lockless contexts
293 * @sk: socket pointer
294 *
295 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
296 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
297 */
inet_sk_state_load(const struct sock * sk)298 static inline int inet_sk_state_load(const struct sock *sk)
299 {
300 /* state change might impact lockless readers. */
301 return smp_load_acquire(&sk->sk_state);
302 }
303
304 /**
305 * inet_sk_state_store - update sk->sk_state
306 * @sk: socket pointer
307 * @newstate: new state
308 *
309 * Paired with inet_sk_state_load(). Should be used in contexts where
310 * state change might impact lockless readers.
311 */
312 void inet_sk_state_store(struct sock *sk, int newstate);
313
314 void inet_sk_set_state(struct sock *sk, int state);
315
__inet_ehashfn(const __be32 laddr,const __u16 lport,const __be32 faddr,const __be16 fport,u32 initval)316 static inline unsigned int __inet_ehashfn(const __be32 laddr,
317 const __u16 lport,
318 const __be32 faddr,
319 const __be16 fport,
320 u32 initval)
321 {
322 return jhash_3words((__force __u32) laddr,
323 (__force __u32) faddr,
324 ((__u32) lport) << 16 | (__force __u32)fport,
325 initval);
326 }
327
328 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
329 struct sock *sk_listener,
330 bool attach_listener);
331
inet_sk_flowi_flags(const struct sock * sk)332 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
333 {
334 __u8 flags = 0;
335
336 if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl)
337 flags |= FLOWI_FLAG_ANYSRC;
338 return flags;
339 }
340
inet_inc_convert_csum(struct sock * sk)341 static inline void inet_inc_convert_csum(struct sock *sk)
342 {
343 inet_sk(sk)->convert_csum++;
344 }
345
inet_dec_convert_csum(struct sock * sk)346 static inline void inet_dec_convert_csum(struct sock *sk)
347 {
348 if (inet_sk(sk)->convert_csum > 0)
349 inet_sk(sk)->convert_csum--;
350 }
351
inet_get_convert_csum(struct sock * sk)352 static inline bool inet_get_convert_csum(struct sock *sk)
353 {
354 return !!inet_sk(sk)->convert_csum;
355 }
356
357
inet_can_nonlocal_bind(struct net * net,struct inet_sock * inet)358 static inline bool inet_can_nonlocal_bind(struct net *net,
359 struct inet_sock *inet)
360 {
361 return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) ||
362 inet->freebind || inet->transparent;
363 }
364
365 #endif /* _INET_SOCK_H */
366