xref: /openbmc/linux/include/net/inet_sock.h (revision dd21bfa4)
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[];
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 
103 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
104 {
105 	return (struct inet_request_sock *)sk;
106 }
107 
108 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
109 {
110 	if (!sk->sk_mark && sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept)
111 		return skb->mark;
112 
113 	return sk->sk_mark;
114 }
115 
116 static inline int inet_request_bound_dev_if(const struct sock *sk,
117 					    struct sk_buff *skb)
118 {
119 #ifdef CONFIG_NET_L3_MASTER_DEV
120 	struct net *net = sock_net(sk);
121 
122 	if (!sk->sk_bound_dev_if && net->ipv4.sysctl_tcp_l3mdev_accept)
123 		return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
124 #endif
125 
126 	return sk->sk_bound_dev_if;
127 }
128 
129 static inline int inet_sk_bound_l3mdev(const struct sock *sk)
130 {
131 #ifdef CONFIG_NET_L3_MASTER_DEV
132 	struct net *net = sock_net(sk);
133 
134 	if (!net->ipv4.sysctl_tcp_l3mdev_accept)
135 		return l3mdev_master_ifindex_by_index(net,
136 						      sk->sk_bound_dev_if);
137 #endif
138 
139 	return 0;
140 }
141 
142 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
143 				     int dif, int sdif)
144 {
145 	if (!bound_dev_if)
146 		return !sdif || l3mdev_accept;
147 	return bound_dev_if == dif || bound_dev_if == sdif;
148 }
149 
150 struct inet_cork {
151 	unsigned int		flags;
152 	__be32			addr;
153 	struct ip_options	*opt;
154 	unsigned int		fragsize;
155 	int			length; /* Total length of all frames */
156 	struct dst_entry	*dst;
157 	u8			tx_flags;
158 	__u8			ttl;
159 	__s16			tos;
160 	char			priority;
161 	__u16			gso_size;
162 	u64			transmit_time;
163 	u32			mark;
164 };
165 
166 struct inet_cork_full {
167 	struct inet_cork	base;
168 	struct flowi		fl;
169 };
170 
171 struct ip_mc_socklist;
172 struct ipv6_pinfo;
173 struct rtable;
174 
175 /** struct inet_sock - representation of INET sockets
176  *
177  * @sk - ancestor class
178  * @pinet6 - pointer to IPv6 control block
179  * @inet_daddr - Foreign IPv4 addr
180  * @inet_rcv_saddr - Bound local IPv4 addr
181  * @inet_dport - Destination port
182  * @inet_num - Local port
183  * @inet_saddr - Sending source
184  * @uc_ttl - Unicast TTL
185  * @inet_sport - Source port
186  * @inet_id - ID counter for DF pkts
187  * @tos - TOS
188  * @mc_ttl - Multicasting TTL
189  * @is_icsk - is this an inet_connection_sock?
190  * @uc_index - Unicast outgoing device index
191  * @mc_index - Multicast device index
192  * @mc_list - Group array
193  * @cork - info to build ip hdr on each ip frag while socket is corked
194  */
195 struct inet_sock {
196 	/* sk and pinet6 has to be the first two members of inet_sock */
197 	struct sock		sk;
198 #if IS_ENABLED(CONFIG_IPV6)
199 	struct ipv6_pinfo	*pinet6;
200 #endif
201 	/* Socket demultiplex comparisons on incoming packets. */
202 #define inet_daddr		sk.__sk_common.skc_daddr
203 #define inet_rcv_saddr		sk.__sk_common.skc_rcv_saddr
204 #define inet_dport		sk.__sk_common.skc_dport
205 #define inet_num		sk.__sk_common.skc_num
206 
207 	__be32			inet_saddr;
208 	__s16			uc_ttl;
209 	__u16			cmsg_flags;
210 	struct ip_options_rcu __rcu	*inet_opt;
211 	__be16			inet_sport;
212 	__u16			inet_id;
213 
214 	__u8			tos;
215 	__u8			min_ttl;
216 	__u8			mc_ttl;
217 	__u8			pmtudisc;
218 	__u8			recverr:1,
219 				is_icsk:1,
220 				freebind:1,
221 				hdrincl:1,
222 				mc_loop:1,
223 				transparent:1,
224 				mc_all:1,
225 				nodefrag:1;
226 	__u8			bind_address_no_port:1,
227 				recverr_rfc4884:1,
228 				defer_connect:1; /* Indicates that fastopen_connect is set
229 						  * and cookie exists so we defer connect
230 						  * until first data frame is written
231 						  */
232 	__u8			rcv_tos;
233 	__u8			convert_csum;
234 	int			uc_index;
235 	int			mc_index;
236 	__be32			mc_addr;
237 	struct ip_mc_socklist __rcu	*mc_list;
238 	struct inet_cork_full	cork;
239 };
240 
241 #define IPCORK_OPT	1	/* ip-options has been held in ipcork.opt */
242 #define IPCORK_ALLFRAG	2	/* always fragment (for ipv6 for now) */
243 
244 /* cmsg flags for inet */
245 #define IP_CMSG_PKTINFO		BIT(0)
246 #define IP_CMSG_TTL		BIT(1)
247 #define IP_CMSG_TOS		BIT(2)
248 #define IP_CMSG_RECVOPTS	BIT(3)
249 #define IP_CMSG_RETOPTS		BIT(4)
250 #define IP_CMSG_PASSSEC		BIT(5)
251 #define IP_CMSG_ORIGDSTADDR	BIT(6)
252 #define IP_CMSG_CHECKSUM	BIT(7)
253 #define IP_CMSG_RECVFRAGSIZE	BIT(8)
254 
255 /**
256  * sk_to_full_sk - Access to a full socket
257  * @sk: pointer to a socket
258  *
259  * SYNACK messages might be attached to request sockets.
260  * Some places want to reach the listener in this case.
261  */
262 static inline struct sock *sk_to_full_sk(struct sock *sk)
263 {
264 #ifdef CONFIG_INET
265 	if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
266 		sk = inet_reqsk(sk)->rsk_listener;
267 #endif
268 	return sk;
269 }
270 
271 /* sk_to_full_sk() variant with a const argument */
272 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
273 {
274 #ifdef CONFIG_INET
275 	if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
276 		sk = ((const struct request_sock *)sk)->rsk_listener;
277 #endif
278 	return sk;
279 }
280 
281 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
282 {
283 	return sk_to_full_sk(skb->sk);
284 }
285 
286 static inline struct inet_sock *inet_sk(const struct sock *sk)
287 {
288 	return (struct inet_sock *)sk;
289 }
290 
291 static inline void __inet_sk_copy_descendant(struct sock *sk_to,
292 					     const struct sock *sk_from,
293 					     const int ancestor_size)
294 {
295 	memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
296 	       sk_from->sk_prot->obj_size - ancestor_size);
297 }
298 
299 int inet_sk_rebuild_header(struct sock *sk);
300 
301 /**
302  * inet_sk_state_load - read sk->sk_state for lockless contexts
303  * @sk: socket pointer
304  *
305  * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
306  * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
307  */
308 static inline int inet_sk_state_load(const struct sock *sk)
309 {
310 	/* state change might impact lockless readers. */
311 	return smp_load_acquire(&sk->sk_state);
312 }
313 
314 /**
315  * inet_sk_state_store - update sk->sk_state
316  * @sk: socket pointer
317  * @newstate: new state
318  *
319  * Paired with inet_sk_state_load(). Should be used in contexts where
320  * state change might impact lockless readers.
321  */
322 void inet_sk_state_store(struct sock *sk, int newstate);
323 
324 void inet_sk_set_state(struct sock *sk, int state);
325 
326 static inline unsigned int __inet_ehashfn(const __be32 laddr,
327 					  const __u16 lport,
328 					  const __be32 faddr,
329 					  const __be16 fport,
330 					  u32 initval)
331 {
332 	return jhash_3words((__force __u32) laddr,
333 			    (__force __u32) faddr,
334 			    ((__u32) lport) << 16 | (__force __u32)fport,
335 			    initval);
336 }
337 
338 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
339 				      struct sock *sk_listener,
340 				      bool attach_listener);
341 
342 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
343 {
344 	__u8 flags = 0;
345 
346 	if (inet_sk(sk)->transparent || inet_sk(sk)->hdrincl)
347 		flags |= FLOWI_FLAG_ANYSRC;
348 	return flags;
349 }
350 
351 static inline void inet_inc_convert_csum(struct sock *sk)
352 {
353 	inet_sk(sk)->convert_csum++;
354 }
355 
356 static inline void inet_dec_convert_csum(struct sock *sk)
357 {
358 	if (inet_sk(sk)->convert_csum > 0)
359 		inet_sk(sk)->convert_csum--;
360 }
361 
362 static inline bool inet_get_convert_csum(struct sock *sk)
363 {
364 	return !!inet_sk(sk)->convert_csum;
365 }
366 
367 
368 static inline bool inet_can_nonlocal_bind(struct net *net,
369 					  struct inet_sock *inet)
370 {
371 	return net->ipv4.sysctl_ip_nonlocal_bind ||
372 		inet->freebind || inet->transparent;
373 }
374 
375 static inline bool inet_addr_valid_or_nonlocal(struct net *net,
376 					       struct inet_sock *inet,
377 					       __be32 addr,
378 					       int addr_type)
379 {
380 	return inet_can_nonlocal_bind(net, inet) ||
381 		addr == htonl(INADDR_ANY) ||
382 		addr_type == RTN_LOCAL ||
383 		addr_type == RTN_MULTICAST ||
384 		addr_type == RTN_BROADCAST;
385 }
386 
387 #endif	/* _INET_SOCK_H */
388