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#include <net/inet_connection_nata.h>
26#endif
27
28/* Cancel timers, when they are not required. */
29#undef INET_CSK_CLEAR_TIMERS
30
31struct inet_bind_bucket;
32struct tcp_congestion_ops;
33
34/*
35 * Pointers to address related TCP functions
36 * (i.e. things that depend on the address family)
37 */
38struct inet_connection_sock_af_ops {
39	int	    (*queue_xmit)(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
40	void	    (*send_check)(struct sock *sk, struct sk_buff *skb);
41	int	    (*rebuild_header)(struct sock *sk);
42	void	    (*sk_rx_dst_set)(struct sock *sk, const struct sk_buff *skb);
43	int	    (*conn_request)(struct sock *sk, struct sk_buff *skb);
44	struct sock *(*syn_recv_sock)(const struct sock *sk, struct sk_buff *skb,
45				      struct request_sock *req,
46				      struct dst_entry *dst,
47				      struct request_sock *req_unhash,
48				      bool *own_req);
49	u16	    net_header_len;
50	u16	    net_frag_header_len;
51	u16	    sockaddr_len;
52	int	    (*setsockopt)(struct sock *sk, int level, int optname,
53				  sockptr_t optval, unsigned int optlen);
54	int	    (*getsockopt)(struct sock *sk, int level, int optname,
55				  char __user *optval, int __user *optlen);
56	void	    (*addr2sockaddr)(struct sock *sk, struct sockaddr *);
57	void	    (*mtu_reduced)(struct sock *sk);
58};
59
60/** inet_connection_sock - INET connection oriented sock
61 *
62 * @icsk_accept_queue:	   FIFO of established children
63 * @icsk_bind_hash:	   Bind node
64 * @icsk_timeout:	   Timeout
65 * @icsk_retransmit_timer: Resend (no ack)
66 * @icsk_rto:		   Retransmit timeout
67 * @icsk_pmtu_cookie	   Last pmtu seen by socket
68 * @icsk_ca_ops		   Pluggable congestion control hook
69 * @icsk_af_ops		   Operations which are AF_INET{4,6} specific
70 * @icsk_ulp_ops	   Pluggable ULP control hook
71 * @icsk_ulp_data	   ULP private data
72 * @icsk_clean_acked	   Clean acked data hook
73 * @icsk_listen_portaddr_node	hash to the portaddr listener hashtable
74 * @icsk_ca_state:	   Congestion control state
75 * @icsk_retransmits:	   Number of unrecovered [RTO] timeouts
76 * @icsk_pending:	   Scheduled timer event
77 * @icsk_backoff:	   Backoff
78 * @icsk_syn_retries:      Number of allowed SYN (or equivalent) retries
79 * @icsk_probes_out:	   unanswered 0 window probes
80 * @icsk_ext_hdr_len:	   Network protocol overhead (IP/IPv6 options)
81 * @icsk_ack:		   Delayed ACK control data
82 * @icsk_mtup;		   MTU probing control data
83 * @icsk_probes_tstamp:    Probe timestamp (cleared by non-zero window ack)
84 * @icsk_user_timeout:	   TCP_USER_TIMEOUT value
85 */
86struct inet_connection_sock {
87	/* inet_sock has to be the first member! */
88	struct inet_sock	  icsk_inet;
89	struct request_sock_queue icsk_accept_queue;
90	struct inet_bind_bucket	  *icsk_bind_hash;
91	unsigned long		  icsk_timeout;
92 	struct timer_list	  icsk_retransmit_timer;
93 	struct timer_list	  icsk_delack_timer;
94	__u32			  icsk_rto;
95	__u32                     icsk_rto_min;
96	__u32                     icsk_delack_max;
97	__u32			  icsk_pmtu_cookie;
98	const struct tcp_congestion_ops *icsk_ca_ops;
99	const struct inet_connection_sock_af_ops *icsk_af_ops;
100	const struct tcp_ulp_ops  *icsk_ulp_ops;
101	void __rcu		  *icsk_ulp_data;
102	void (*icsk_clean_acked)(struct sock *sk, u32 acked_seq);
103	struct hlist_node         icsk_listen_portaddr_node;
104	unsigned int		  (*icsk_sync_mss)(struct sock *sk, u32 pmtu);
105	__u8			  icsk_ca_state:5,
106				  icsk_ca_initialized:1,
107				  icsk_ca_setsockopt:1,
108				  icsk_ca_dst_locked:1;
109	__u8			  icsk_retransmits;
110	__u8			  icsk_pending;
111	__u8			  icsk_backoff;
112	__u8			  icsk_syn_retries;
113	__u8			  icsk_probes_out;
114	__u16			  icsk_ext_hdr_len;
115#if defined(CONFIG_TCP_NATA_URC) || defined(CONFIG_TCP_NATA_STL)
116	__u8			  nata_retries_enabled:1,
117				  nata_reserved:7;
118	__u8			  nata_data_retries;
119	__u8			  nata_retries_type;
120	__u32			  nata_syn_rto;
121	__u32			  nata_data_rto;
122#endif
123	struct {
124		__u8		  pending;	 /* ACK is pending			   */
125		__u8		  quick;	 /* Scheduled number of quick acks	   */
126		__u8		  pingpong;	 /* The session is interactive		   */
127		__u8		  retry;	 /* Number of attempts			   */
128		__u32		  ato;		 /* Predicted tick of soft clock	   */
129		unsigned long	  timeout;	 /* Currently scheduled timeout		   */
130		__u32		  lrcvtime;	 /* timestamp of last received data packet */
131		__u16		  last_seg_size; /* Size of last incoming segment	   */
132		__u16		  rcv_mss;	 /* MSS used for delayed ACK decisions	   */
133	} icsk_ack;
134	struct {
135		int		  enabled;
136
137		/* Range of MTUs to search */
138		int		  search_high;
139		int		  search_low;
140
141		/* Information on the current probe. */
142		int		  probe_size;
143
144		u32		  probe_timestamp;
145	} icsk_mtup;
146	u32			  icsk_probes_tstamp;
147	u32			  icsk_user_timeout;
148
149	u64			  icsk_ca_priv[104 / sizeof(u64)];
150#define ICSK_CA_PRIV_SIZE      (13 * sizeof(u64))
151};
152
153#define ICSK_TIME_RETRANS	1	/* Retransmit timer */
154#define ICSK_TIME_DACK		2	/* Delayed ack timer */
155#define ICSK_TIME_PROBE0	3	/* Zero window probe timer */
156#define ICSK_TIME_EARLY_RETRANS 4	/* Early retransmit timer */
157#define ICSK_TIME_LOSS_PROBE	5	/* Tail loss probe timer */
158#define ICSK_TIME_REO_TIMEOUT	6	/* Reordering timer */
159
160static inline struct inet_connection_sock *inet_csk(const struct sock *sk)
161{
162	return (struct inet_connection_sock *)sk;
163}
164
165static inline void *inet_csk_ca(const struct sock *sk)
166{
167	return (void *)inet_csk(sk)->icsk_ca_priv;
168}
169
170struct sock *inet_csk_clone_lock(const struct sock *sk,
171				 const struct request_sock *req,
172				 const gfp_t priority);
173
174enum inet_csk_ack_state_t {
175	ICSK_ACK_SCHED	= 1,
176	ICSK_ACK_TIMER  = 2,
177	ICSK_ACK_PUSHED = 4,
178	ICSK_ACK_PUSHED2 = 8,
179	ICSK_ACK_NOW = 16	/* Send the next ACK immediately (once) */
180};
181
182void inet_csk_init_xmit_timers(struct sock *sk,
183			       void (*retransmit_handler)(struct timer_list *),
184			       void (*delack_handler)(struct timer_list *),
185			       void (*keepalive_handler)(struct timer_list *));
186void inet_csk_clear_xmit_timers(struct sock *sk);
187void inet_csk_clear_xmit_timers_sync(struct sock *sk);
188
189static inline void inet_csk_schedule_ack(struct sock *sk)
190{
191	inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_SCHED;
192}
193
194static inline int inet_csk_ack_scheduled(const struct sock *sk)
195{
196	return inet_csk(sk)->icsk_ack.pending & ICSK_ACK_SCHED;
197}
198
199static inline void inet_csk_delack_init(struct sock *sk)
200{
201	memset(&inet_csk(sk)->icsk_ack, 0, sizeof(inet_csk(sk)->icsk_ack));
202}
203
204void inet_csk_delete_keepalive_timer(struct sock *sk);
205void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long timeout);
206
207static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what)
208{
209	struct inet_connection_sock *icsk = inet_csk(sk);
210
211	if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0) {
212		icsk->icsk_pending = 0;
213#ifdef INET_CSK_CLEAR_TIMERS
214		sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
215#endif
216	} else if (what == ICSK_TIME_DACK) {
217		icsk->icsk_ack.pending = 0;
218		icsk->icsk_ack.retry = 0;
219#ifdef INET_CSK_CLEAR_TIMERS
220		sk_stop_timer(sk, &icsk->icsk_delack_timer);
221#endif
222	} else {
223		pr_debug("inet_csk BUG: unknown timer value\n");
224	}
225}
226
227#if defined(CONFIG_TCP_NATA_URC) || defined(CONFIG_TCP_NATA_STL)
228static inline unsigned long get_nata_rto(struct sock *sk,
229					 struct inet_connection_sock *icsk,
230					 unsigned long when, const int what)
231{
232	unsigned long when_nata;
233	unsigned long shift;
234
235	if (!icsk->nata_retries_enabled)
236		return when;
237
238	switch (what) {
239	case ICSK_TIME_RETRANS:
240	case ICSK_TIME_EARLY_RETRANS:
241	case ICSK_TIME_LOSS_PROBE:
242	case ICSK_TIME_REO_TIMEOUT:
243		break;
244	default:
245		return when;
246	}
247
248	if (icsk->nata_retries_type == NATA_STL)
249		return sk->sk_state == TCP_SYN_SENT ?
250			icsk->nata_syn_rto : icsk->nata_data_rto;
251
252	if (!nata_thin_stream_check(sk))
253		return when;
254
255	when_nata = icsk->nata_data_rto;
256	if (icsk->icsk_retransmits > icsk->nata_data_retries) {
257		shift = icsk->icsk_retransmits - icsk->nata_data_retries;
258		if (shift > MAX_SHIFT) {
259			when_nata = NATA_RTO_MAX;
260		} else {
261			when_nata <<= shift;
262		}
263	}
264	return min(when, when_nata);
265}
266#endif
267
268/*
269 *	Reset the retransmission timer
270 */
271static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what,
272					     unsigned long when,
273					     const unsigned long max_when)
274{
275	struct inet_connection_sock *icsk = inet_csk(sk);
276
277#if defined(CONFIG_TCP_NATA_URC) || defined(CONFIG_TCP_NATA_STL)
278	when = get_nata_rto(sk, icsk, when, what);
279#endif
280
281	if (when > max_when) {
282		pr_debug("reset_xmit_timer: sk=%p %d when=0x%lx, caller=%p\n",
283			 sk, what, when, (void *)_THIS_IP_);
284		when = max_when;
285	}
286
287	if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 ||
288	    what == ICSK_TIME_EARLY_RETRANS || what == ICSK_TIME_LOSS_PROBE ||
289	    what == ICSK_TIME_REO_TIMEOUT) {
290		icsk->icsk_pending = what;
291		icsk->icsk_timeout = jiffies + when;
292		sk_reset_timer(sk, &icsk->icsk_retransmit_timer, icsk->icsk_timeout);
293	} else if (what == ICSK_TIME_DACK) {
294		icsk->icsk_ack.pending |= ICSK_ACK_TIMER;
295		icsk->icsk_ack.timeout = jiffies + when;
296		sk_reset_timer(sk, &icsk->icsk_delack_timer, icsk->icsk_ack.timeout);
297	} else {
298		pr_debug("inet_csk BUG: unknown timer value\n");
299	}
300}
301
302static inline unsigned long
303inet_csk_rto_backoff(const struct inet_connection_sock *icsk,
304		     unsigned long max_when)
305{
306        u64 when = (u64)icsk->icsk_rto << icsk->icsk_backoff;
307
308        return (unsigned long)min_t(u64, when, max_when);
309}
310
311struct sock *inet_csk_accept(struct sock *sk, int flags, int *err, bool kern);
312
313int inet_csk_get_port(struct sock *sk, unsigned short snum);
314
315struct dst_entry *inet_csk_route_req(const struct sock *sk, struct flowi4 *fl4,
316				     const struct request_sock *req);
317struct dst_entry *inet_csk_route_child_sock(const struct sock *sk,
318					    struct sock *newsk,
319					    const struct request_sock *req);
320
321struct sock *inet_csk_reqsk_queue_add(struct sock *sk,
322				      struct request_sock *req,
323				      struct sock *child);
324void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
325				   unsigned long timeout);
326struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child,
327					 struct request_sock *req,
328					 bool own_req);
329
330static inline void inet_csk_reqsk_queue_added(struct sock *sk)
331{
332	reqsk_queue_added(&inet_csk(sk)->icsk_accept_queue);
333}
334
335static inline int inet_csk_reqsk_queue_len(const struct sock *sk)
336{
337	return reqsk_queue_len(&inet_csk(sk)->icsk_accept_queue);
338}
339
340static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk)
341{
342	return inet_csk_reqsk_queue_len(sk) >= sk->sk_max_ack_backlog;
343}
344
345bool inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req);
346void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req);
347
348static inline void inet_csk_prepare_for_destroy_sock(struct sock *sk)
349{
350	/* The below has to be done to allow calling inet_csk_destroy_sock */
351	sock_set_flag(sk, SOCK_DEAD);
352	this_cpu_inc(*sk->sk_prot->orphan_count);
353}
354
355void inet_csk_destroy_sock(struct sock *sk);
356void inet_csk_prepare_forced_close(struct sock *sk);
357
358/*
359 * LISTEN is a special case for poll..
360 */
361static inline __poll_t inet_csk_listen_poll(const struct sock *sk)
362{
363	return !reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue) ?
364			(EPOLLIN | EPOLLRDNORM) : 0;
365}
366
367int inet_csk_listen_start(struct sock *sk, int backlog);
368void inet_csk_listen_stop(struct sock *sk);
369
370void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr);
371
372/* update the fast reuse flag when adding a socket */
373void inet_csk_update_fastreuse(struct inet_bind_bucket *tb,
374			       struct sock *sk);
375
376struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu);
377
378#define TCP_PINGPONG_THRESH	1
379
380static inline void inet_csk_enter_pingpong_mode(struct sock *sk)
381{
382	inet_csk(sk)->icsk_ack.pingpong = TCP_PINGPONG_THRESH;
383}
384
385static inline void inet_csk_exit_pingpong_mode(struct sock *sk)
386{
387	inet_csk(sk)->icsk_ack.pingpong = 0;
388}
389
390static inline bool inet_csk_in_pingpong_mode(struct sock *sk)
391{
392	return inet_csk(sk)->icsk_ack.pingpong >= TCP_PINGPONG_THRESH;
393}
394
395static inline bool inet_csk_has_ulp(struct sock *sk)
396{
397	return inet_sk(sk)->is_icsk && !!inet_csk(sk)->icsk_ulp_ops;
398}
399
400static inline void inet_init_csk_locks(struct sock *sk)
401{
402	struct inet_connection_sock *icsk = inet_csk(sk);
403
404	spin_lock_init(&icsk->icsk_accept_queue.rskq_lock);
405	spin_lock_init(&icsk->icsk_accept_queue.fastopenq.lock);
406}
407
408#endif /* _INET_CONNECTION_SOCK_H */
409