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