• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * NET		Generic infrastructure for INET connection oriented protocols.
4  *
5  *		Definitions for inet_connection_sock
6  *
7  * Authors:	Many people, see the TCP sources
8  *
9  * 		From code originally in TCP
10  */
11 #ifndef _INET_CONNECTION_SOCK_H
12 #define _INET_CONNECTION_SOCK_H
13 
14 #include <linux/compiler.h>
15 #include <linux/string.h>
16 #include <linux/timer.h>
17 #include <linux/poll.h>
18 #include <linux/kernel.h>
19 #include <linux/sockptr.h>
20 
21 #include <net/inet_sock.h>
22 #include <net/request_sock.h>
23 #if defined(CONFIG_TCP_NATA_URC) || defined(CONFIG_TCP_NATA_STL)
24 #include <net/nata.h>
25 #endif
26 
27 /* Cancel timers, when they are not required. */
28 #undef INET_CSK_CLEAR_TIMERS
29 
30 struct inet_bind_bucket;
31 struct tcp_congestion_ops;
32 
33 /*
34  * Pointers to address related TCP functions
35  * (i.e. things that depend on the address family)
36  */
37 struct inet_connection_sock_af_ops {
38 	int	    (*queue_xmit)(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
39 	void	    (*send_check)(struct sock *sk, struct sk_buff *skb);
40 	int	    (*rebuild_header)(struct sock *sk);
41 	void	    (*sk_rx_dst_set)(struct sock *sk, const struct sk_buff *skb);
42 	int	    (*conn_request)(struct sock *sk, struct sk_buff *skb);
43 	struct sock *(*syn_recv_sock)(const struct sock *sk, struct sk_buff *skb,
44 				      struct request_sock *req,
45 				      struct dst_entry *dst,
46 				      struct request_sock *req_unhash,
47 				      bool *own_req);
48 	u16	    net_header_len;
49 	u16	    net_frag_header_len;
50 	u16	    sockaddr_len;
51 	int	    (*setsockopt)(struct sock *sk, int level, int optname,
52 				  sockptr_t optval, unsigned int optlen);
53 	int	    (*getsockopt)(struct sock *sk, int level, int optname,
54 				  char __user *optval, int __user *optlen);
55 	void	    (*addr2sockaddr)(struct sock *sk, struct sockaddr *);
56 	void	    (*mtu_reduced)(struct sock *sk);
57 };
58 
59 /** inet_connection_sock - INET connection oriented sock
60  *
61  * @icsk_accept_queue:	   FIFO of established children
62  * @icsk_bind_hash:	   Bind node
63  * @icsk_timeout:	   Timeout
64  * @icsk_retransmit_timer: Resend (no ack)
65  * @icsk_rto:		   Retransmit timeout
66  * @icsk_pmtu_cookie	   Last pmtu seen by socket
67  * @icsk_ca_ops		   Pluggable congestion control hook
68  * @icsk_af_ops		   Operations which are AF_INET{4,6} specific
69  * @icsk_ulp_ops	   Pluggable ULP control hook
70  * @icsk_ulp_data	   ULP private data
71  * @icsk_clean_acked	   Clean acked data hook
72  * @icsk_listen_portaddr_node	hash to the portaddr listener hashtable
73  * @icsk_ca_state:	   Congestion control state
74  * @icsk_retransmits:	   Number of unrecovered [RTO] timeouts
75  * @icsk_pending:	   Scheduled timer event
76  * @icsk_backoff:	   Backoff
77  * @icsk_syn_retries:      Number of allowed SYN (or equivalent) retries
78  * @icsk_probes_out:	   unanswered 0 window probes
79  * @icsk_ext_hdr_len:	   Network protocol overhead (IP/IPv6 options)
80  * @icsk_ack:		   Delayed ACK control data
81  * @icsk_mtup;		   MTU probing control data
82  * @icsk_probes_tstamp:    Probe timestamp (cleared by non-zero window ack)
83  * @icsk_user_timeout:	   TCP_USER_TIMEOUT value
84  */
85 struct inet_connection_sock {
86 	/* inet_sock has to be the first member! */
87 	struct inet_sock	  icsk_inet;
88 	struct request_sock_queue icsk_accept_queue;
89 	struct inet_bind_bucket	  *icsk_bind_hash;
90 	unsigned long		  icsk_timeout;
91  	struct timer_list	  icsk_retransmit_timer;
92  	struct timer_list	  icsk_delack_timer;
93 	__u32			  icsk_rto;
94 	__u32                     icsk_rto_min;
95 	__u32                     icsk_delack_max;
96 	__u32			  icsk_pmtu_cookie;
97 	const struct tcp_congestion_ops *icsk_ca_ops;
98 	const struct inet_connection_sock_af_ops *icsk_af_ops;
99 	const struct tcp_ulp_ops  *icsk_ulp_ops;
100 	void __rcu		  *icsk_ulp_data;
101 	void (*icsk_clean_acked)(struct sock *sk, u32 acked_seq);
102 	struct hlist_node         icsk_listen_portaddr_node;
103 	unsigned int		  (*icsk_sync_mss)(struct sock *sk, u32 pmtu);
104 	__u8			  icsk_ca_state:5,
105 				  icsk_ca_initialized:1,
106 				  icsk_ca_setsockopt:1,
107 				  icsk_ca_dst_locked:1;
108 	__u8			  icsk_retransmits;
109 	__u8			  icsk_pending;
110 	__u8			  icsk_backoff;
111 	__u8			  icsk_syn_retries;
112 	__u8			  icsk_probes_out;
113 	__u16			  icsk_ext_hdr_len;
114 #if defined(CONFIG_TCP_NATA_URC) || defined(CONFIG_TCP_NATA_STL)
115 	__u8			  nata_retries_enabled:1,
116 				  nata_reserved:7;
117 	__u8			  nata_data_retries;
118 	__u8			  nata_retries_type;
119 	__u32			  nata_syn_rto;
120 	__u32			  nata_data_rto;
121 #endif
122 	struct {
123 		__u8		  pending;	 /* ACK is pending			   */
124 		__u8		  quick;	 /* Scheduled number of quick acks	   */
125 		__u8		  pingpong;	 /* The session is interactive		   */
126 		__u8		  retry;	 /* Number of attempts			   */
127 		__u32		  ato;		 /* Predicted tick of soft clock	   */
128 		unsigned long	  timeout;	 /* Currently scheduled timeout		   */
129 		__u32		  lrcvtime;	 /* timestamp of last received data packet */
130 		__u16		  last_seg_size; /* Size of last incoming segment	   */
131 		__u16		  rcv_mss;	 /* MSS used for delayed ACK decisions	   */
132 	} icsk_ack;
133 	struct {
134 		int		  enabled;
135 
136 		/* Range of MTUs to search */
137 		int		  search_high;
138 		int		  search_low;
139 
140 		/* Information on the current probe. */
141 		int		  probe_size;
142 
143 		u32		  probe_timestamp;
144 	} icsk_mtup;
145 	u32			  icsk_probes_tstamp;
146 	u32			  icsk_user_timeout;
147 
148 	u64			  icsk_ca_priv[104 / sizeof(u64)];
149 #define ICSK_CA_PRIV_SIZE      (13 * sizeof(u64))
150 };
151 
152 #define ICSK_TIME_RETRANS	1	/* Retransmit timer */
153 #define ICSK_TIME_DACK		2	/* Delayed ack timer */
154 #define ICSK_TIME_PROBE0	3	/* Zero window probe timer */
155 #define ICSK_TIME_EARLY_RETRANS 4	/* Early retransmit timer */
156 #define ICSK_TIME_LOSS_PROBE	5	/* Tail loss probe timer */
157 #define ICSK_TIME_REO_TIMEOUT	6	/* Reordering timer */
158 
inet_csk(const struct sock * sk)159 static inline struct inet_connection_sock *inet_csk(const struct sock *sk)
160 {
161 	return (struct inet_connection_sock *)sk;
162 }
163 
inet_csk_ca(const struct sock * sk)164 static inline void *inet_csk_ca(const struct sock *sk)
165 {
166 	return (void *)inet_csk(sk)->icsk_ca_priv;
167 }
168 
169 struct sock *inet_csk_clone_lock(const struct sock *sk,
170 				 const struct request_sock *req,
171 				 const gfp_t priority);
172 
173 enum inet_csk_ack_state_t {
174 	ICSK_ACK_SCHED	= 1,
175 	ICSK_ACK_TIMER  = 2,
176 	ICSK_ACK_PUSHED = 4,
177 	ICSK_ACK_PUSHED2 = 8,
178 	ICSK_ACK_NOW = 16	/* Send the next ACK immediately (once) */
179 };
180 
181 void inet_csk_init_xmit_timers(struct sock *sk,
182 			       void (*retransmit_handler)(struct timer_list *),
183 			       void (*delack_handler)(struct timer_list *),
184 			       void (*keepalive_handler)(struct timer_list *));
185 void inet_csk_clear_xmit_timers(struct sock *sk);
186 void inet_csk_clear_xmit_timers_sync(struct sock *sk);
187 
inet_csk_schedule_ack(struct sock * sk)188 static inline void inet_csk_schedule_ack(struct sock *sk)
189 {
190 	inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_SCHED;
191 }
192 
inet_csk_ack_scheduled(const struct sock * sk)193 static inline int inet_csk_ack_scheduled(const struct sock *sk)
194 {
195 	return inet_csk(sk)->icsk_ack.pending & ICSK_ACK_SCHED;
196 }
197 
inet_csk_delack_init(struct sock * sk)198 static inline void inet_csk_delack_init(struct sock *sk)
199 {
200 	memset(&inet_csk(sk)->icsk_ack, 0, sizeof(inet_csk(sk)->icsk_ack));
201 }
202 
203 void inet_csk_delete_keepalive_timer(struct sock *sk);
204 void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long timeout);
205 
inet_csk_clear_xmit_timer(struct sock * sk,const int what)206 static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what)
207 {
208 	struct inet_connection_sock *icsk = inet_csk(sk);
209 
210 	if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0) {
211 		icsk->icsk_pending = 0;
212 #ifdef INET_CSK_CLEAR_TIMERS
213 		sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
214 #endif
215 	} else if (what == ICSK_TIME_DACK) {
216 		icsk->icsk_ack.pending = 0;
217 		icsk->icsk_ack.retry = 0;
218 #ifdef INET_CSK_CLEAR_TIMERS
219 		sk_stop_timer(sk, &icsk->icsk_delack_timer);
220 #endif
221 	} else {
222 		pr_debug("inet_csk BUG: unknown timer value\n");
223 	}
224 }
225 
226 #if defined(CONFIG_TCP_NATA_URC) || defined(CONFIG_TCP_NATA_STL)
get_nata_rto(struct sock * sk,struct inet_connection_sock * icsk,unsigned long when,const int what)227 static inline unsigned long get_nata_rto(struct sock *sk,
228 					 struct inet_connection_sock *icsk,
229 					 unsigned long when, const int what)
230 {
231 	unsigned long when_nata;
232 	unsigned long shift;
233 
234 	if (!icsk->nata_retries_enabled)
235 		return when;
236 
237 	switch (what) {
238 	case ICSK_TIME_RETRANS:
239 	case ICSK_TIME_EARLY_RETRANS:
240 	case ICSK_TIME_LOSS_PROBE:
241 	case ICSK_TIME_REO_TIMEOUT:
242 		break;
243 	default:
244 		return when;
245 	}
246 
247 	if (icsk->nata_retries_type == NATA_STL)
248 		return sk->sk_state == TCP_SYN_SENT ?
249 			icsk->nata_syn_rto : icsk->nata_data_rto;
250 
251 	when_nata = icsk->nata_data_rto;
252 	if (icsk->icsk_retransmits > icsk->nata_data_retries) {
253 		shift = icsk->icsk_retransmits - icsk->nata_data_retries;
254 		if (shift > MAX_SHIFT) {
255 			when_nata = NATA_RTO_MAX;
256 		} else {
257 			when_nata <<= shift;
258 		}
259 	}
260 	return min(when, when_nata);
261 }
262 #endif
263 
264 /*
265  *	Reset the retransmission timer
266  */
inet_csk_reset_xmit_timer(struct sock * sk,const int what,unsigned long when,const unsigned long max_when)267 static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what,
268 					     unsigned long when,
269 					     const unsigned long max_when)
270 {
271 	struct inet_connection_sock *icsk = inet_csk(sk);
272 
273 #if defined(CONFIG_TCP_NATA_URC) || defined(CONFIG_TCP_NATA_STL)
274 	when = get_nata_rto(sk, icsk, when, what);
275 #endif
276 
277 	if (when > max_when) {
278 		pr_debug("reset_xmit_timer: sk=%p %d when=0x%lx, caller=%p\n",
279 			 sk, what, when, (void *)_THIS_IP_);
280 		when = max_when;
281 	}
282 
283 	if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 ||
284 	    what == ICSK_TIME_EARLY_RETRANS || what == ICSK_TIME_LOSS_PROBE ||
285 	    what == ICSK_TIME_REO_TIMEOUT) {
286 		icsk->icsk_pending = what;
287 		icsk->icsk_timeout = jiffies + when;
288 		sk_reset_timer(sk, &icsk->icsk_retransmit_timer, icsk->icsk_timeout);
289 	} else if (what == ICSK_TIME_DACK) {
290 		icsk->icsk_ack.pending |= ICSK_ACK_TIMER;
291 		icsk->icsk_ack.timeout = jiffies + when;
292 		sk_reset_timer(sk, &icsk->icsk_delack_timer, icsk->icsk_ack.timeout);
293 	} else {
294 		pr_debug("inet_csk BUG: unknown timer value\n");
295 	}
296 }
297 
298 static inline unsigned long
inet_csk_rto_backoff(const struct inet_connection_sock * icsk,unsigned long max_when)299 inet_csk_rto_backoff(const struct inet_connection_sock *icsk,
300 		     unsigned long max_when)
301 {
302         u64 when = (u64)icsk->icsk_rto << icsk->icsk_backoff;
303 
304         return (unsigned long)min_t(u64, when, max_when);
305 }
306 
307 struct sock *inet_csk_accept(struct sock *sk, int flags, int *err, bool kern);
308 
309 int inet_csk_get_port(struct sock *sk, unsigned short snum);
310 
311 struct dst_entry *inet_csk_route_req(const struct sock *sk, struct flowi4 *fl4,
312 				     const struct request_sock *req);
313 struct dst_entry *inet_csk_route_child_sock(const struct sock *sk,
314 					    struct sock *newsk,
315 					    const struct request_sock *req);
316 
317 struct sock *inet_csk_reqsk_queue_add(struct sock *sk,
318 				      struct request_sock *req,
319 				      struct sock *child);
320 void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
321 				   unsigned long timeout);
322 struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child,
323 					 struct request_sock *req,
324 					 bool own_req);
325 
inet_csk_reqsk_queue_added(struct sock * sk)326 static inline void inet_csk_reqsk_queue_added(struct sock *sk)
327 {
328 	reqsk_queue_added(&inet_csk(sk)->icsk_accept_queue);
329 }
330 
inet_csk_reqsk_queue_len(const struct sock * sk)331 static inline int inet_csk_reqsk_queue_len(const struct sock *sk)
332 {
333 	return reqsk_queue_len(&inet_csk(sk)->icsk_accept_queue);
334 }
335 
inet_csk_reqsk_queue_is_full(const struct sock * sk)336 static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk)
337 {
338 	return inet_csk_reqsk_queue_len(sk) >= sk->sk_max_ack_backlog;
339 }
340 
341 bool inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req);
342 void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req);
343 
inet_csk_prepare_for_destroy_sock(struct sock * sk)344 static inline void inet_csk_prepare_for_destroy_sock(struct sock *sk)
345 {
346 	/* The below has to be done to allow calling inet_csk_destroy_sock */
347 	sock_set_flag(sk, SOCK_DEAD);
348 	this_cpu_inc(*sk->sk_prot->orphan_count);
349 }
350 
351 void inet_csk_destroy_sock(struct sock *sk);
352 void inet_csk_prepare_forced_close(struct sock *sk);
353 
354 /*
355  * LISTEN is a special case for poll..
356  */
inet_csk_listen_poll(const struct sock * sk)357 static inline __poll_t inet_csk_listen_poll(const struct sock *sk)
358 {
359 	return !reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue) ?
360 			(EPOLLIN | EPOLLRDNORM) : 0;
361 }
362 
363 int inet_csk_listen_start(struct sock *sk, int backlog);
364 void inet_csk_listen_stop(struct sock *sk);
365 
366 void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr);
367 
368 /* update the fast reuse flag when adding a socket */
369 void inet_csk_update_fastreuse(struct inet_bind_bucket *tb,
370 			       struct sock *sk);
371 
372 struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu);
373 
374 #define TCP_PINGPONG_THRESH	1
375 
inet_csk_enter_pingpong_mode(struct sock * sk)376 static inline void inet_csk_enter_pingpong_mode(struct sock *sk)
377 {
378 	inet_csk(sk)->icsk_ack.pingpong = TCP_PINGPONG_THRESH;
379 }
380 
inet_csk_exit_pingpong_mode(struct sock * sk)381 static inline void inet_csk_exit_pingpong_mode(struct sock *sk)
382 {
383 	inet_csk(sk)->icsk_ack.pingpong = 0;
384 }
385 
inet_csk_in_pingpong_mode(struct sock * sk)386 static inline bool inet_csk_in_pingpong_mode(struct sock *sk)
387 {
388 	return inet_csk(sk)->icsk_ack.pingpong >= TCP_PINGPONG_THRESH;
389 }
390 
inet_csk_has_ulp(struct sock * sk)391 static inline bool inet_csk_has_ulp(struct sock *sk)
392 {
393 	return inet_sk(sk)->is_icsk && !!inet_csk(sk)->icsk_ulp_ops;
394 }
395 
inet_init_csk_locks(struct sock * sk)396 static inline void inet_init_csk_locks(struct sock *sk)
397 {
398 	struct inet_connection_sock *icsk = inet_csk(sk);
399 
400 	spin_lock_init(&icsk->icsk_accept_queue.rskq_lock);
401 	spin_lock_init(&icsk->icsk_accept_queue.fastopenq.lock);
402 }
403 
404 #endif /* _INET_CONNECTION_SOCK_H */
405