xref: /openbmc/linux/include/net/inet_sock.h (revision b732539e)
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 
107 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
108 {
109 	return (struct inet_request_sock *)sk;
110 }
111 
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 
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 static inline struct ip_options_rcu *ireq_opt_deref(const struct inet_request_sock *ireq)
134 {
135 	return rcu_dereference_check(ireq->ireq_opt,
136 				     refcount_read(&ireq->req.rsk_refcnt) > 0);
137 }
138 
139 struct inet_cork {
140 	unsigned int		flags;
141 	__be32			addr;
142 	struct ip_options	*opt;
143 	unsigned int		fragsize;
144 	int			length; /* Total length of all frames */
145 	struct dst_entry	*dst;
146 	u8			tx_flags;
147 	__u8			ttl;
148 	__s16			tos;
149 	char			priority;
150 };
151 
152 struct inet_cork_full {
153 	struct inet_cork	base;
154 	struct flowi		fl;
155 };
156 
157 struct ip_mc_socklist;
158 struct ipv6_pinfo;
159 struct rtable;
160 
161 /** struct inet_sock - representation of INET sockets
162  *
163  * @sk - ancestor class
164  * @pinet6 - pointer to IPv6 control block
165  * @inet_daddr - Foreign IPv4 addr
166  * @inet_rcv_saddr - Bound local IPv4 addr
167  * @inet_dport - Destination port
168  * @inet_num - Local port
169  * @inet_saddr - Sending source
170  * @uc_ttl - Unicast TTL
171  * @inet_sport - Source port
172  * @inet_id - ID counter for DF pkts
173  * @tos - TOS
174  * @mc_ttl - Multicasting TTL
175  * @is_icsk - is this an inet_connection_sock?
176  * @uc_index - Unicast outgoing device index
177  * @mc_index - Multicast device index
178  * @mc_list - Group array
179  * @cork - info to build ip hdr on each ip frag while socket is corked
180  */
181 struct inet_sock {
182 	/* sk and pinet6 has to be the first two members of inet_sock */
183 	struct sock		sk;
184 #if IS_ENABLED(CONFIG_IPV6)
185 	struct ipv6_pinfo	*pinet6;
186 #endif
187 	/* Socket demultiplex comparisons on incoming packets. */
188 #define inet_daddr		sk.__sk_common.skc_daddr
189 #define inet_rcv_saddr		sk.__sk_common.skc_rcv_saddr
190 #define inet_dport		sk.__sk_common.skc_dport
191 #define inet_num		sk.__sk_common.skc_num
192 
193 	__be32			inet_saddr;
194 	__s16			uc_ttl;
195 	__u16			cmsg_flags;
196 	__be16			inet_sport;
197 	__u16			inet_id;
198 
199 	struct ip_options_rcu __rcu	*inet_opt;
200 	int			rx_dst_ifindex;
201 	__u8			tos;
202 	__u8			min_ttl;
203 	__u8			mc_ttl;
204 	__u8			pmtudisc;
205 	__u8			recverr:1,
206 				is_icsk:1,
207 				freebind:1,
208 				hdrincl:1,
209 				mc_loop:1,
210 				transparent:1,
211 				mc_all:1,
212 				nodefrag:1;
213 	__u8			bind_address_no_port:1,
214 				defer_connect:1; /* Indicates that fastopen_connect is set
215 						  * and cookie exists so we defer connect
216 						  * until first data frame is written
217 						  */
218 	__u8			rcv_tos;
219 	__u8			convert_csum;
220 	int			uc_index;
221 	int			mc_index;
222 	__be32			mc_addr;
223 	struct ip_mc_socklist __rcu	*mc_list;
224 	struct inet_cork_full	cork;
225 };
226 
227 #define IPCORK_OPT	1	/* ip-options has been held in ipcork.opt */
228 #define IPCORK_ALLFRAG	2	/* always fragment (for ipv6 for now) */
229 
230 /* cmsg flags for inet */
231 #define IP_CMSG_PKTINFO		BIT(0)
232 #define IP_CMSG_TTL		BIT(1)
233 #define IP_CMSG_TOS		BIT(2)
234 #define IP_CMSG_RECVOPTS	BIT(3)
235 #define IP_CMSG_RETOPTS		BIT(4)
236 #define IP_CMSG_PASSSEC		BIT(5)
237 #define IP_CMSG_ORIGDSTADDR	BIT(6)
238 #define IP_CMSG_CHECKSUM	BIT(7)
239 #define IP_CMSG_RECVFRAGSIZE	BIT(8)
240 
241 /**
242  * sk_to_full_sk - Access to a full socket
243  * @sk: pointer to a socket
244  *
245  * SYNACK messages might be attached to request sockets.
246  * Some places want to reach the listener in this case.
247  */
248 static inline struct sock *sk_to_full_sk(struct sock *sk)
249 {
250 #ifdef CONFIG_INET
251 	if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
252 		sk = inet_reqsk(sk)->rsk_listener;
253 #endif
254 	return sk;
255 }
256 
257 /* sk_to_full_sk() variant with a const argument */
258 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
259 {
260 #ifdef CONFIG_INET
261 	if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
262 		sk = ((const struct request_sock *)sk)->rsk_listener;
263 #endif
264 	return sk;
265 }
266 
267 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
268 {
269 	return sk_to_full_sk(skb->sk);
270 }
271 
272 static inline struct inet_sock *inet_sk(const struct sock *sk)
273 {
274 	return (struct inet_sock *)sk;
275 }
276 
277 static inline void __inet_sk_copy_descendant(struct sock *sk_to,
278 					     const struct sock *sk_from,
279 					     const int ancestor_size)
280 {
281 	memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
282 	       sk_from->sk_prot->obj_size - ancestor_size);
283 }
284 #if !(IS_ENABLED(CONFIG_IPV6))
285 static inline void inet_sk_copy_descendant(struct sock *sk_to,
286 					   const struct sock *sk_from)
287 {
288 	__inet_sk_copy_descendant(sk_to, sk_from, sizeof(struct inet_sock));
289 }
290 #endif
291 
292 int inet_sk_rebuild_header(struct sock *sk);
293 
294 /**
295  * inet_sk_state_load - read sk->sk_state for lockless contexts
296  * @sk: socket pointer
297  *
298  * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
299  * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
300  */
301 static inline int inet_sk_state_load(const struct sock *sk)
302 {
303 	/* state change might impact lockless readers. */
304 	return smp_load_acquire(&sk->sk_state);
305 }
306 
307 /**
308  * inet_sk_state_store - update sk->sk_state
309  * @sk: socket pointer
310  * @newstate: new state
311  *
312  * Paired with inet_sk_state_load(). Should be used in contexts where
313  * state change might impact lockless readers.
314  */
315 void inet_sk_state_store(struct sock *sk, int newstate);
316 
317 void inet_sk_set_state(struct sock *sk, int state);
318 
319 static inline unsigned int __inet_ehashfn(const __be32 laddr,
320 					  const __u16 lport,
321 					  const __be32 faddr,
322 					  const __be16 fport,
323 					  u32 initval)
324 {
325 	return jhash_3words((__force __u32) laddr,
326 			    (__force __u32) faddr,
327 			    ((__u32) lport) << 16 | (__force __u32)fport,
328 			    initval);
329 }
330 
331 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
332 				      struct sock *sk_listener,
333 				      bool attach_listener);
334 
335 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
336 {
337 	__u8 flags = 0;
338 
339 	if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl)
340 		flags |= FLOWI_FLAG_ANYSRC;
341 	return flags;
342 }
343 
344 static inline void inet_inc_convert_csum(struct sock *sk)
345 {
346 	inet_sk(sk)->convert_csum++;
347 }
348 
349 static inline void inet_dec_convert_csum(struct sock *sk)
350 {
351 	if (inet_sk(sk)->convert_csum > 0)
352 		inet_sk(sk)->convert_csum--;
353 }
354 
355 static inline bool inet_get_convert_csum(struct sock *sk)
356 {
357 	return !!inet_sk(sk)->convert_csum;
358 }
359 
360 #endif	/* _INET_SOCK_H */
361