• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 #include <linux/android_kabi.h>
21 
22 #include <net/flow.h>
23 #include <net/sock.h>
24 #include <net/request_sock.h>
25 #include <net/netns/hash.h>
26 #include <net/tcp_states.h>
27 #include <net/l3mdev.h>
28 
29 /** struct ip_options - IP Options
30  *
31  * @faddr - Saved first hop address
32  * @nexthop - Saved nexthop address in LSRR and SSRR
33  * @is_strictroute - Strict source route
34  * @srr_is_hit - Packet destination addr was our one
35  * @is_changed - IP checksum more not valid
36  * @rr_needaddr - Need to record addr of outgoing dev
37  * @ts_needtime - Need to record timestamp
38  * @ts_needaddr - Need to record addr of outgoing dev
39  */
40 struct ip_options {
41 	__be32		faddr;
42 	__be32		nexthop;
43 	unsigned char	optlen;
44 	unsigned char	srr;
45 	unsigned char	rr;
46 	unsigned char	ts;
47 	unsigned char	is_strictroute:1,
48 			srr_is_hit:1,
49 			is_changed:1,
50 			rr_needaddr:1,
51 			ts_needtime:1,
52 			ts_needaddr:1;
53 	unsigned char	router_alert;
54 	unsigned char	cipso;
55 	unsigned char	__pad2;
56 	unsigned char	__data[];
57 };
58 
59 struct ip_options_rcu {
60 	struct rcu_head rcu;
61 	struct ip_options opt;
62 };
63 
64 struct ip_options_data {
65 	struct ip_options_rcu	opt;
66 	char			data[40];
67 };
68 
69 struct inet_request_sock {
70 	struct request_sock	req;
71 #define ir_loc_addr		req.__req_common.skc_rcv_saddr
72 #define ir_rmt_addr		req.__req_common.skc_daddr
73 #define ir_num			req.__req_common.skc_num
74 #define ir_rmt_port		req.__req_common.skc_dport
75 #define ir_v6_rmt_addr		req.__req_common.skc_v6_daddr
76 #define ir_v6_loc_addr		req.__req_common.skc_v6_rcv_saddr
77 #define ir_iif			req.__req_common.skc_bound_dev_if
78 #define ir_cookie		req.__req_common.skc_cookie
79 #define ireq_net		req.__req_common.skc_net
80 #define ireq_state		req.__req_common.skc_state
81 #define ireq_family		req.__req_common.skc_family
82 
83 	u16			snd_wscale : 4,
84 				rcv_wscale : 4,
85 				tstamp_ok  : 1,
86 				sack_ok	   : 1,
87 				wscale_ok  : 1,
88 				ecn_ok	   : 1,
89 				acked	   : 1,
90 				no_srccheck: 1,
91 				smc_ok	   : 1;
92 	u32                     ir_mark;
93 	union {
94 		struct ip_options_rcu __rcu	*ireq_opt;
95 #if IS_ENABLED(CONFIG_IPV6)
96 		struct {
97 			struct ipv6_txoptions	*ipv6_opt;
98 			struct sk_buff		*pktopts;
99 		};
100 #endif
101 	};
102 };
103 
inet_rsk(const struct request_sock * sk)104 static inline struct inet_request_sock *inet_rsk(const struct request_sock *sk)
105 {
106 	return (struct inet_request_sock *)sk;
107 }
108 
inet_request_mark(const struct sock * sk,struct sk_buff * skb)109 static inline u32 inet_request_mark(const struct sock *sk, struct sk_buff *skb)
110 {
111 	u32 mark = READ_ONCE(sk->sk_mark);
112 
113 	if (!mark && READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_fwmark_accept))
114 		return skb->mark;
115 
116 	return mark;
117 }
118 
inet_request_bound_dev_if(const struct sock * sk,struct sk_buff * skb)119 static inline int inet_request_bound_dev_if(const struct sock *sk,
120 					    struct sk_buff *skb)
121 {
122 	int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
123 #ifdef CONFIG_NET_L3_MASTER_DEV
124 	struct net *net = sock_net(sk);
125 
126 	if (!bound_dev_if && READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
127 		return l3mdev_master_ifindex_by_index(net, skb->skb_iif);
128 #endif
129 
130 	return bound_dev_if;
131 }
132 
inet_sk_bound_l3mdev(const struct sock * sk)133 static inline int inet_sk_bound_l3mdev(const struct sock *sk)
134 {
135 #ifdef CONFIG_NET_L3_MASTER_DEV
136 	struct net *net = sock_net(sk);
137 
138 	if (!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept))
139 		return l3mdev_master_ifindex_by_index(net,
140 						      sk->sk_bound_dev_if);
141 #endif
142 
143 	return 0;
144 }
145 
inet_bound_dev_eq(bool l3mdev_accept,int bound_dev_if,int dif,int sdif)146 static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
147 				     int dif, int sdif)
148 {
149 	if (!bound_dev_if)
150 		return !sdif || l3mdev_accept;
151 	return bound_dev_if == dif || bound_dev_if == sdif;
152 }
153 
inet_sk_bound_dev_eq(struct net * net,int bound_dev_if,int dif,int sdif)154 static inline bool inet_sk_bound_dev_eq(struct net *net, int bound_dev_if,
155 					int dif, int sdif)
156 {
157 #if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
158 	return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
159 				 bound_dev_if, dif, sdif);
160 #else
161 	return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
162 #endif
163 }
164 
165 struct inet_cork {
166 	unsigned int		flags;
167 	__be32			addr;
168 	struct ip_options	*opt;
169 	unsigned int		fragsize;
170 	int			length; /* Total length of all frames */
171 	struct dst_entry	*dst;
172 	u8			tx_flags;
173 	__u8			ttl;
174 	__s16			tos;
175 	char			priority;
176 	__u16			gso_size;
177 	u64			transmit_time;
178 	u32			mark;
179 };
180 
181 struct inet_cork_full {
182 	struct inet_cork	base;
183 	struct flowi		fl;
184 };
185 
186 struct ip_mc_socklist;
187 struct ipv6_pinfo;
188 struct rtable;
189 
190 /** struct inet_sock - representation of INET sockets
191  *
192  * @sk - ancestor class
193  * @pinet6 - pointer to IPv6 control block
194  * @inet_daddr - Foreign IPv4 addr
195  * @inet_rcv_saddr - Bound local IPv4 addr
196  * @inet_dport - Destination port
197  * @inet_num - Local port
198  * @inet_flags - various atomic flags
199  * @inet_saddr - Sending source
200  * @uc_ttl - Unicast TTL
201  * @inet_sport - Source port
202  * @inet_id - ID counter for DF pkts
203  * @tos - TOS
204  * @mc_ttl - Multicasting TTL
205  * @uc_index - Unicast outgoing device index
206  * @mc_index - Multicast device index
207  * @mc_list - Group array
208  * @cork - info to build ip hdr on each ip frag while socket is corked
209  */
210 struct inet_sock {
211 	/* sk and pinet6 has to be the first two members of inet_sock */
212 	struct sock		sk;
213 #if IS_ENABLED(CONFIG_IPV6)
214 	struct ipv6_pinfo	*pinet6;
215 #endif
216 	/* Socket demultiplex comparisons on incoming packets. */
217 #define inet_daddr		sk.__sk_common.skc_daddr
218 #define inet_rcv_saddr		sk.__sk_common.skc_rcv_saddr
219 #define inet_dport		sk.__sk_common.skc_dport
220 #define inet_num		sk.__sk_common.skc_num
221 
222 	unsigned long		inet_flags;
223 	__be32			inet_saddr;
224 	__s16			uc_ttl;
225 	__be16			inet_sport;
226 	struct ip_options_rcu __rcu	*inet_opt;
227 	atomic_t		inet_id;
228 
229 	__u8			tos;
230 	__u8			min_ttl;
231 	__u8			mc_ttl;
232 	__u8			pmtudisc;
233 	__u8			rcv_tos;
234 	__u8			convert_csum;
235 	int			uc_index;
236 	int			mc_index;
237 	__be32			mc_addr;
238 	struct {
239 		__u16 lo;
240 		__u16 hi;
241 	}			local_port_range;
242 
243 	struct ip_mc_socklist __rcu	*mc_list;
244 	struct inet_cork_full	cork;
245 
246 	ANDROID_KABI_RESERVE(1);
247 	ANDROID_KABI_RESERVE(2);
248 };
249 
250 #define IPCORK_OPT	1	/* ip-options has been held in ipcork.opt */
251 #define IPCORK_ALLFRAG	2	/* always fragment (for ipv6 for now) */
252 
253 enum {
254 	INET_FLAGS_PKTINFO	= 0,
255 	INET_FLAGS_TTL		= 1,
256 	INET_FLAGS_TOS		= 2,
257 	INET_FLAGS_RECVOPTS	= 3,
258 	INET_FLAGS_RETOPTS	= 4,
259 	INET_FLAGS_PASSSEC	= 5,
260 	INET_FLAGS_ORIGDSTADDR	= 6,
261 	INET_FLAGS_CHECKSUM	= 7,
262 	INET_FLAGS_RECVFRAGSIZE	= 8,
263 
264 	INET_FLAGS_RECVERR	= 9,
265 	INET_FLAGS_RECVERR_RFC4884 = 10,
266 	INET_FLAGS_FREEBIND	= 11,
267 	INET_FLAGS_HDRINCL	= 12,
268 	INET_FLAGS_MC_LOOP	= 13,
269 	INET_FLAGS_MC_ALL	= 14,
270 	INET_FLAGS_TRANSPARENT	= 15,
271 	INET_FLAGS_IS_ICSK	= 16,
272 	INET_FLAGS_NODEFRAG	= 17,
273 	INET_FLAGS_BIND_ADDRESS_NO_PORT = 18,
274 	INET_FLAGS_DEFER_CONNECT = 19,
275 };
276 
277 /* cmsg flags for inet */
278 #define IP_CMSG_PKTINFO		BIT(INET_FLAGS_PKTINFO)
279 #define IP_CMSG_TTL		BIT(INET_FLAGS_TTL)
280 #define IP_CMSG_TOS		BIT(INET_FLAGS_TOS)
281 #define IP_CMSG_RECVOPTS	BIT(INET_FLAGS_RECVOPTS)
282 #define IP_CMSG_RETOPTS		BIT(INET_FLAGS_RETOPTS)
283 #define IP_CMSG_PASSSEC		BIT(INET_FLAGS_PASSSEC)
284 #define IP_CMSG_ORIGDSTADDR	BIT(INET_FLAGS_ORIGDSTADDR)
285 #define IP_CMSG_CHECKSUM	BIT(INET_FLAGS_CHECKSUM)
286 #define IP_CMSG_RECVFRAGSIZE	BIT(INET_FLAGS_RECVFRAGSIZE)
287 
288 #define IP_CMSG_ALL	(IP_CMSG_PKTINFO | IP_CMSG_TTL |		\
289 			 IP_CMSG_TOS | IP_CMSG_RECVOPTS |		\
290 			 IP_CMSG_RETOPTS | IP_CMSG_PASSSEC |		\
291 			 IP_CMSG_ORIGDSTADDR | IP_CMSG_CHECKSUM |	\
292 			 IP_CMSG_RECVFRAGSIZE)
293 
inet_cmsg_flags(const struct inet_sock * inet)294 static inline unsigned long inet_cmsg_flags(const struct inet_sock *inet)
295 {
296 	return READ_ONCE(inet->inet_flags) & IP_CMSG_ALL;
297 }
298 
299 #define inet_test_bit(nr, sk)			\
300 	test_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
301 #define inet_set_bit(nr, sk)			\
302 	set_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
303 #define inet_clear_bit(nr, sk)			\
304 	clear_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags)
305 #define inet_assign_bit(nr, sk, val)		\
306 	assign_bit(INET_FLAGS_##nr, &inet_sk(sk)->inet_flags, val)
307 
308 /**
309  * sk_to_full_sk - Access to a full socket
310  * @sk: pointer to a socket
311  *
312  * SYNACK messages might be attached to request sockets.
313  * Some places want to reach the listener in this case.
314  */
sk_to_full_sk(struct sock * sk)315 static inline struct sock *sk_to_full_sk(struct sock *sk)
316 {
317 #ifdef CONFIG_INET
318 	if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
319 		sk = inet_reqsk(sk)->rsk_listener;
320 #endif
321 	return sk;
322 }
323 
324 /* sk_to_full_sk() variant with a const argument */
sk_const_to_full_sk(const struct sock * sk)325 static inline const struct sock *sk_const_to_full_sk(const struct sock *sk)
326 {
327 #ifdef CONFIG_INET
328 	if (sk && sk->sk_state == TCP_NEW_SYN_RECV)
329 		sk = ((const struct request_sock *)sk)->rsk_listener;
330 #endif
331 	return sk;
332 }
333 
skb_to_full_sk(const struct sk_buff * skb)334 static inline struct sock *skb_to_full_sk(const struct sk_buff *skb)
335 {
336 	return sk_to_full_sk(skb->sk);
337 }
338 
339 #define inet_sk(ptr) container_of_const(ptr, struct inet_sock, sk)
340 
__inet_sk_copy_descendant(struct sock * sk_to,const struct sock * sk_from,const int ancestor_size)341 static inline void __inet_sk_copy_descendant(struct sock *sk_to,
342 					     const struct sock *sk_from,
343 					     const int ancestor_size)
344 {
345 	memcpy(inet_sk(sk_to) + 1, inet_sk(sk_from) + 1,
346 	       sk_from->sk_prot->obj_size - ancestor_size);
347 }
348 
349 int inet_sk_rebuild_header(struct sock *sk);
350 
351 /**
352  * inet_sk_state_load - read sk->sk_state for lockless contexts
353  * @sk: socket pointer
354  *
355  * Paired with inet_sk_state_store(). Used in places we don't hold socket lock:
356  * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
357  */
inet_sk_state_load(const struct sock * sk)358 static inline int inet_sk_state_load(const struct sock *sk)
359 {
360 	/* state change might impact lockless readers. */
361 	return smp_load_acquire(&sk->sk_state);
362 }
363 
364 /**
365  * inet_sk_state_store - update sk->sk_state
366  * @sk: socket pointer
367  * @newstate: new state
368  *
369  * Paired with inet_sk_state_load(). Should be used in contexts where
370  * state change might impact lockless readers.
371  */
372 void inet_sk_state_store(struct sock *sk, int newstate);
373 
374 void inet_sk_set_state(struct sock *sk, int state);
375 
__inet_ehashfn(const __be32 laddr,const __u16 lport,const __be32 faddr,const __be16 fport,u32 initval)376 static inline unsigned int __inet_ehashfn(const __be32 laddr,
377 					  const __u16 lport,
378 					  const __be32 faddr,
379 					  const __be16 fport,
380 					  u32 initval)
381 {
382 	return jhash_3words((__force __u32) laddr,
383 			    (__force __u32) faddr,
384 			    ((__u32) lport) << 16 | (__force __u32)fport,
385 			    initval);
386 }
387 
388 struct request_sock *inet_reqsk_alloc(const struct request_sock_ops *ops,
389 				      struct sock *sk_listener,
390 				      bool attach_listener);
391 
inet_sk_flowi_flags(const struct sock * sk)392 static inline __u8 inet_sk_flowi_flags(const struct sock *sk)
393 {
394 	__u8 flags = 0;
395 
396 	if (inet_test_bit(TRANSPARENT, sk) || inet_test_bit(HDRINCL, sk))
397 		flags |= FLOWI_FLAG_ANYSRC;
398 	return flags;
399 }
400 
inet_inc_convert_csum(struct sock * sk)401 static inline void inet_inc_convert_csum(struct sock *sk)
402 {
403 	inet_sk(sk)->convert_csum++;
404 }
405 
inet_dec_convert_csum(struct sock * sk)406 static inline void inet_dec_convert_csum(struct sock *sk)
407 {
408 	if (inet_sk(sk)->convert_csum > 0)
409 		inet_sk(sk)->convert_csum--;
410 }
411 
inet_get_convert_csum(struct sock * sk)412 static inline bool inet_get_convert_csum(struct sock *sk)
413 {
414 	return !!inet_sk(sk)->convert_csum;
415 }
416 
417 
inet_can_nonlocal_bind(struct net * net,struct inet_sock * inet)418 static inline bool inet_can_nonlocal_bind(struct net *net,
419 					  struct inet_sock *inet)
420 {
421 	return READ_ONCE(net->ipv4.sysctl_ip_nonlocal_bind) ||
422 		test_bit(INET_FLAGS_FREEBIND, &inet->inet_flags) ||
423 		test_bit(INET_FLAGS_TRANSPARENT, &inet->inet_flags);
424 }
425 
inet_addr_valid_or_nonlocal(struct net * net,struct inet_sock * inet,__be32 addr,int addr_type)426 static inline bool inet_addr_valid_or_nonlocal(struct net *net,
427 					       struct inet_sock *inet,
428 					       __be32 addr,
429 					       int addr_type)
430 {
431 	return inet_can_nonlocal_bind(net, inet) ||
432 		addr == htonl(INADDR_ANY) ||
433 		addr_type == RTN_LOCAL ||
434 		addr_type == RTN_MULTICAST ||
435 		addr_type == RTN_BROADCAST;
436 }
437 
438 #endif	/* _INET_SOCK_H */
439