1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Definitions for inet_sock
8 *
9 * Authors: Many, reorganised here by
10 * Arnaldo Carvalho de Melo <acme@mandriva.com>
11 */
12 #ifndef _INET_SOCK_H
13 #define _INET_SOCK_H
14
15 #include <linux/bitops.h>
16 #include <linux/string.h>
17 #include <linux/types.h>
18 #include <linux/jhash.h>
19 #include <linux/netdevice.h>
20
21 #include <net/flow.h>
22 #include <net/sock.h>
23 #include <net/request_sock.h>
24 #include <net/netns/hash.h>
25 #include <net/tcp_states.h>
26 #include <net/l3mdev.h>
27
28 /** struct ip_options - IP Options
29 *
30 * @faddr - Saved first hop address
31 * @nexthop - Saved nexthop address in LSRR and SSRR
32 * @is_strictroute - Strict source route
33 * @srr_is_hit - Packet destination addr was our one
34 * @is_changed - IP checksum more not valid
35 * @rr_needaddr - Need to record addr of outgoing dev
36 * @ts_needtime - Need to record timestamp
37 * @ts_needaddr - Need to record addr of outgoing dev
38 */
39 struct ip_options {
40 __be32 faddr;
41 __be32 nexthop;
42 unsigned char optlen;
43 unsigned char srr;
44 unsigned char rr;
45 unsigned char ts;
46 unsigned char is_strictroute:1,
47 srr_is_hit:1,
48 is_changed:1,
49 rr_needaddr:1,
50 ts_needtime:1,
51 ts_needaddr:1;
52 unsigned char router_alert;
53 unsigned char cipso;
54 unsigned char __pad2;
55 unsigned char __data[0];
56 };
57
58 struct ip_options_rcu {
59 struct rcu_head rcu;
60 struct ip_options opt;
61 };
62
63 struct ip_options_data {
64 struct ip_options_rcu opt;
65 char data[40];
66 };
67
68 struct inet_request_sock {
69 struct request_sock req;
70 #define ir_loc_addr req.__req_common.skc_rcv_saddr
71 #define ir_rmt_addr req.__req_common.skc_daddr
72 #define ir_num req.__req_common.skc_num
73 #define ir_rmt_port req.__req_common.skc_dport
74 #define ir_v6_rmt_addr req.__req_common.skc_v6_daddr
75 #define ir_v6_loc_addr req.__req_common.skc_v6_rcv_saddr
76 #define ir_iif req.__req_common.skc_bound_dev_if
77 #define ir_cookie req.__req_common.skc_cookie
78 #define ireq_net req.__req_common.skc_net
79 #define ireq_state req.__req_common.skc_state
80 #define ireq_family req.__req_common.skc_family
81
82 u16 snd_wscale : 4,
83 rcv_wscale : 4,
84 tstamp_ok : 1,
85 sack_ok : 1,
86 wscale_ok : 1,
87 ecn_ok : 1,
88 acked : 1,
89 no_srccheck: 1,
90 smc_ok : 1;
91 u32 ir_mark;
92 union {
93 struct ip_options_rcu __rcu *ireq_opt;
94 #if IS_ENABLED(CONFIG_IPV6)
95 struct {
96 struct ipv6_txoptions *ipv6_opt;
97 struct sk_buff *pktopts;
98 };
99 #endif
100 };
101 };
102
inet_rsk(const struct request_sock * sk)103 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
104 {
105 return (struct inet_request_sock *)sk;
106 }
107
inet_request_mark(const struct sock * sk,struct sk_buff * skb)108 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
109 {
110 if (!sk->sk_mark &&
111 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept))
112 return skb->mark;
113
114 return sk->sk_mark;
115 }
116
inet_request_bound_dev_if(const struct sock * sk,struct sk_buff * skb)117 static inline int inet_request_bound_dev_if(const struct sock *sk,
118 struct sk_buff *skb)
119 {
120 #ifdef CONFIG_NET_L3_MASTER_DEV
121 struct net *net = sock_net(sk);
122
123 if (!sk->sk_bound_dev_if && net->ipv4.sysctl_tcp_l3mdev_accept)
124 return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
125 #endif
126
127 return sk->sk_bound_dev_if;
128 }
129
inet_sk_bound_l3mdev(const struct sock * sk)130 static inline int inet_sk_bound_l3mdev(const struct sock *sk)
131 {
132 #ifdef CONFIG_NET_L3_MASTER_DEV
133 struct net *net = sock_net(sk);
134
135 if (!net->ipv4.sysctl_tcp_l3mdev_accept)
136 return l3mdev_master_ifindex_by_index(net,
137 sk->sk_bound_dev_if);
138 #endif
139
140 return 0;
141 }
142
inet_bound_dev_eq(bool l3mdev_accept,int bound_dev_if,int dif,int sdif)143 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
144 int dif, int sdif)
145 {
146 if (!bound_dev_if)
147 return !sdif || l3mdev_accept;
148 return bound_dev_if == dif || bound_dev_if == sdif;
149 }
150
151 struct inet_cork {
152 unsigned int flags;
153 __be32 addr;
154 struct ip_options *opt;
155 unsigned int fragsize;
156 int length; /* Total length of all frames */
157 struct dst_entry *dst;
158 u8 tx_flags;
159 __u8 ttl;
160 __s16 tos;
161 char priority;
162 __u16 gso_size;
163 u64 transmit_time;
164 u32 mark;
165 };
166
167 struct inet_cork_full {
168 struct inet_cork base;
169 struct flowi fl;
170 };
171
172 struct ip_mc_socklist;
173 struct ipv6_pinfo;
174 struct rtable;
175
176 /** struct inet_sock - representation of INET sockets
177 *
178 * @sk - ancestor class
179 * @pinet6 - pointer to IPv6 control block
180 * @inet_daddr - Foreign IPv4 addr
181 * @inet_rcv_saddr - Bound local IPv4 addr
182 * @inet_dport - Destination port
183 * @inet_num - Local port
184 * @inet_saddr - Sending source
185 * @uc_ttl - Unicast TTL
186 * @inet_sport - Source port
187 * @inet_id - ID counter for DF pkts
188 * @tos - TOS
189 * @mc_ttl - Multicasting TTL
190 * @is_icsk - is this an inet_connection_sock?
191 * @uc_index - Unicast outgoing device index
192 * @mc_index - Multicast device index
193 * @mc_list - Group array
194 * @cork - info to build ip hdr on each ip frag while socket is corked
195 */
196 struct inet_sock {
197 /* sk and pinet6 has to be the first two members of inet_sock */
198 struct sock sk;
199 #if IS_ENABLED(CONFIG_IPV6)
200 struct ipv6_pinfo *pinet6;
201 #endif
202 /* Socket demultiplex comparisons on incoming packets. */
203 #define inet_daddr sk.__sk_common.skc_daddr
204 #define inet_rcv_saddr sk.__sk_common.skc_rcv_saddr
205 #define inet_dport sk.__sk_common.skc_dport
206 #define inet_num sk.__sk_common.skc_num
207
208 __be32 inet_saddr;
209 __s16 uc_ttl;
210 __u16 cmsg_flags;
211 __be16 inet_sport;
212 __u16 inet_id;
213
214 struct ip_options_rcu __rcu *inet_opt;
215 int rx_dst_ifindex;
216 __u8 tos;
217 __u8 min_ttl;
218 __u8 mc_ttl;
219 __u8 pmtudisc;
220 __u8 recverr:1,
221 is_icsk:1,
222 freebind:1,
223 hdrincl:1,
224 mc_loop:1,
225 transparent:1,
226 mc_all:1,
227 nodefrag:1;
228 __u8 bind_address_no_port:1,
229 defer_connect:1; /* Indicates that fastopen_connect is set
230 * and cookie exists so we defer connect
231 * until first data frame is written
232 */
233 __u8 rcv_tos;
234 __u8 convert_csum;
235 int uc_index;
236 int mc_index;
237 __be32 mc_addr;
238 struct ip_mc_socklist __rcu *mc_list;
239 struct inet_cork_full cork;
240 };
241
242 #define IPCORK_OPT 1 /* ip-options has been held in ipcork.opt */
243 #define IPCORK_ALLFRAG 2 /* always fragment (for ipv6 for now) */
244
245 /* cmsg flags for inet */
246 #define IP_CMSG_PKTINFO BIT(0)
247 #define IP_CMSG_TTL BIT(1)
248 #define IP_CMSG_TOS BIT(2)
249 #define IP_CMSG_RECVOPTS BIT(3)
250 #define IP_CMSG_RETOPTS BIT(4)
251 #define IP_CMSG_PASSSEC BIT(5)
252 #define IP_CMSG_ORIGDSTADDR BIT(6)
253 #define IP_CMSG_CHECKSUM BIT(7)
254 #define IP_CMSG_RECVFRAGSIZE BIT(8)
255
256 /**
257 * sk_to_full_sk - Access to a full socket
258 * @sk: pointer to a socket
259 *
260 * SYNACK messages might be attached to request sockets.
261 * Some places want to reach the listener in this case.
262 */
sk_to_full_sk(struct sock * sk)263 static inline struct sock *sk_to_full_sk(struct sock *sk)
264 {
265 #ifdef CONFIG_INET
266 if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
267 sk = inet_reqsk(sk)->rsk_listener;
268 #endif
269 return sk;
270 }
271
272 /* sk_to_full_sk() variant with a const argument */
sk_const_to_full_sk(const struct sock * sk)273 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
274 {
275 #ifdef CONFIG_INET
276 if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
277 sk = ((const struct request_sock *)sk)->rsk_listener;
278 #endif
279 return sk;
280 }
281
skb_to_full_sk(const struct sk_buff * skb)282 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
283 {
284 return sk_to_full_sk(skb->sk);
285 }
286
inet_sk(const struct sock * sk)287 static inline struct inet_sock *inet_sk(const struct sock *sk)
288 {
289 return (struct inet_sock *)sk;
290 }
291
__inet_sk_copy_descendant(struct sock * sk_to,const struct sock * sk_from,const int ancestor_size)292 static inline void __inet_sk_copy_descendant(struct sock *sk_to,
293 const struct sock *sk_from,
294 const int ancestor_size)
295 {
296 memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
297 sk_from->sk_prot->obj_size - ancestor_size);
298 }
299 #if !(IS_ENABLED(CONFIG_IPV6))
inet_sk_copy_descendant(struct sock * sk_to,const struct sock * sk_from)300 static inline void inet_sk_copy_descendant(struct sock *sk_to,
301 const struct sock *sk_from)
302 {
303 __inet_sk_copy_descendant(sk_to, sk_from, sizeof(struct inet_sock));
304 }
305 #endif
306
307 int inet_sk_rebuild_header(struct sock *sk);
308
309 /**
310 * inet_sk_state_load - read sk->sk_state for lockless contexts
311 * @sk: socket pointer
312 *
313 * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
314 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
315 */
inet_sk_state_load(const struct sock * sk)316 static inline int inet_sk_state_load(const struct sock *sk)
317 {
318 /* state change might impact lockless readers. */
319 return smp_load_acquire(&sk->sk_state);
320 }
321
322 /**
323 * inet_sk_state_store - update sk->sk_state
324 * @sk: socket pointer
325 * @newstate: new state
326 *
327 * Paired with inet_sk_state_load(). Should be used in contexts where
328 * state change might impact lockless readers.
329 */
330 void inet_sk_state_store(struct sock *sk, int newstate);
331
332 void inet_sk_set_state(struct sock *sk, int state);
333
__inet_ehashfn(const __be32 laddr,const __u16 lport,const __be32 faddr,const __be16 fport,u32 initval)334 static inline unsigned int __inet_ehashfn(const __be32 laddr,
335 const __u16 lport,
336 const __be32 faddr,
337 const __be16 fport,
338 u32 initval)
339 {
340 return jhash_3words((__force __u32) laddr,
341 (__force __u32) faddr,
342 ((__u32) lport) << 16 | (__force __u32)fport,
343 initval);
344 }
345
346 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
347 struct sock *sk_listener,
348 bool attach_listener);
349
inet_sk_flowi_flags(const struct sock * sk)350 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
351 {
352 __u8 flags = 0;
353
354 if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl)
355 flags |= FLOWI_FLAG_ANYSRC;
356 return flags;
357 }
358
inet_inc_convert_csum(struct sock * sk)359 static inline void inet_inc_convert_csum(struct sock *sk)
360 {
361 inet_sk(sk)->convert_csum++;
362 }
363
inet_dec_convert_csum(struct sock * sk)364 static inline void inet_dec_convert_csum(struct sock *sk)
365 {
366 if (inet_sk(sk)->convert_csum > 0)
367 inet_sk(sk)->convert_csum--;
368 }
369
inet_get_convert_csum(struct sock * sk)370 static inline bool inet_get_convert_csum(struct sock *sk)
371 {
372 return !!inet_sk(sk)->convert_csum;
373 }
374
375
inet_can_nonlocal_bind(struct net * net,struct inet_sock * inet)376 static inline bool inet_can_nonlocal_bind(struct net *net,
377 struct inet_sock *inet)
378 {
379 return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) ||
380 inet->freebind || inet->transparent;
381 }
382
383 #endif /* _INET_SOCK_H */
384