1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * NET Generic infrastructure for INET connection oriented protocols.
5 * Definitions for inet_connection_sock
7 * Authors: Many people, see the TCP sources
9 * From code originally in TCP
11 #ifndef _INET_CONNECTION_SOCK_H
12 #define _INET_CONNECTION_SOCK_H
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>
21 #include <net/inet_sock.h>
22 #include <net/request_sock.h>
24 /* Cancel timers, when they are not required. */
25 #undef INET_CSK_CLEAR_TIMERS
27 struct inet_bind_bucket;
28 struct tcp_congestion_ops;
31 * Pointers to address related TCP functions
32 * (i.e. things that depend on the address family)
34 struct inet_connection_sock_af_ops {
35 int (*queue_xmit)(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
36 void (*send_check)(struct sock *sk, struct sk_buff *skb);
37 int (*rebuild_header)(struct sock *sk);
38 void (*sk_rx_dst_set)(struct sock *sk, const struct sk_buff *skb);
39 int (*conn_request)(struct sock *sk, struct sk_buff *skb);
40 struct sock *(*syn_recv_sock)(const struct sock *sk, struct sk_buff *skb,
41 struct request_sock *req,
42 struct dst_entry *dst,
43 struct request_sock *req_unhash,
46 u16 net_frag_header_len;
48 int (*setsockopt)(struct sock *sk, int level, int optname,
49 sockptr_t optval, unsigned int optlen);
50 int (*getsockopt)(struct sock *sk, int level, int optname,
51 char __user *optval, int __user *optlen);
52 void (*addr2sockaddr)(struct sock *sk, struct sockaddr *);
53 void (*mtu_reduced)(struct sock *sk);
56 /** inet_connection_sock - INET connection oriented sock
58 * @icsk_accept_queue: FIFO of established children
59 * @icsk_bind_hash: Bind node
60 * @icsk_timeout: Timeout
61 * @icsk_retransmit_timer: Resend (no ack)
62 * @icsk_rto: Retransmit timeout
63 * @icsk_pmtu_cookie Last pmtu seen by socket
64 * @icsk_ca_ops Pluggable congestion control hook
65 * @icsk_af_ops Operations which are AF_INET{4,6} specific
66 * @icsk_ulp_ops Pluggable ULP control hook
67 * @icsk_ulp_data ULP private data
68 * @icsk_clean_acked Clean acked data hook
69 * @icsk_listen_portaddr_node hash to the portaddr listener hashtable
70 * @icsk_ca_state: Congestion control state
71 * @icsk_retransmits: Number of unrecovered [RTO] timeouts
72 * @icsk_pending: Scheduled timer event
73 * @icsk_backoff: Backoff
74 * @icsk_syn_retries: Number of allowed SYN (or equivalent) retries
75 * @icsk_probes_out: unanswered 0 window probes
76 * @icsk_ext_hdr_len: Network protocol overhead (IP/IPv6 options)
77 * @icsk_ack: Delayed ACK control data
78 * @icsk_mtup; MTU probing control data
79 * @icsk_probes_tstamp: Probe timestamp (cleared by non-zero window ack)
80 * @icsk_user_timeout: TCP_USER_TIMEOUT value
82 struct inet_connection_sock {
83 /* inet_sock has to be the first member! */
84 struct inet_sock icsk_inet;
85 struct request_sock_queue icsk_accept_queue;
86 struct inet_bind_bucket *icsk_bind_hash;
87 unsigned long icsk_timeout;
88 struct timer_list icsk_retransmit_timer;
89 struct timer_list icsk_delack_timer;
92 __u32 icsk_delack_max;
93 __u32 icsk_pmtu_cookie;
94 const struct tcp_congestion_ops *icsk_ca_ops;
95 const struct inet_connection_sock_af_ops *icsk_af_ops;
96 const struct tcp_ulp_ops *icsk_ulp_ops;
97 void __rcu *icsk_ulp_data;
98 void (*icsk_clean_acked)(struct sock *sk, u32 acked_seq);
99 struct hlist_node icsk_listen_portaddr_node;
100 unsigned int (*icsk_sync_mss)(struct sock *sk, u32 pmtu);
101 __u8 icsk_ca_state:5,
102 icsk_ca_initialized:1,
103 icsk_ca_setsockopt:1,
104 icsk_ca_dst_locked:1;
105 __u8 icsk_retransmits;
108 __u8 icsk_syn_retries;
109 __u8 icsk_probes_out;
110 __u16 icsk_ext_hdr_len;
112 __u8 pending; /* ACK is pending */
113 __u8 quick; /* Scheduled number of quick acks */
114 __u8 pingpong; /* The session is interactive */
115 __u8 retry; /* Number of attempts */
116 __u32 ato; /* Predicted tick of soft clock */
117 unsigned long timeout; /* Currently scheduled timeout */
118 __u32 lrcvtime; /* timestamp of last received data packet */
119 __u16 last_seg_size; /* Size of last incoming segment */
120 __u16 rcv_mss; /* MSS used for delayed ACK decisions */
123 /* Range of MTUs to search */
127 /* Information on the current probe. */
129 /* Is the MTUP feature enabled for this connection? */
134 u32 icsk_probes_tstamp;
135 u32 icsk_user_timeout;
137 u64 icsk_ca_priv[104 / sizeof(u64)];
138 #define ICSK_CA_PRIV_SIZE (13 * sizeof(u64))
141 #define ICSK_TIME_RETRANS 1 /* Retransmit timer */
142 #define ICSK_TIME_DACK 2 /* Delayed ack timer */
143 #define ICSK_TIME_PROBE0 3 /* Zero window probe timer */
144 #define ICSK_TIME_LOSS_PROBE 5 /* Tail loss probe timer */
145 #define ICSK_TIME_REO_TIMEOUT 6 /* Reordering timer */
147 static inline struct inet_connection_sock *inet_csk(const struct sock *sk)
149 return (struct inet_connection_sock *)sk;
152 static inline void *inet_csk_ca(const struct sock *sk)
154 return (void *)inet_csk(sk)->icsk_ca_priv;
157 struct sock *inet_csk_clone_lock(const struct sock *sk,
158 const struct request_sock *req,
159 const gfp_t priority);
161 enum inet_csk_ack_state_t {
165 ICSK_ACK_PUSHED2 = 8,
166 ICSK_ACK_NOW = 16 /* Send the next ACK immediately (once) */
169 void inet_csk_init_xmit_timers(struct sock *sk,
170 void (*retransmit_handler)(struct timer_list *),
171 void (*delack_handler)(struct timer_list *),
172 void (*keepalive_handler)(struct timer_list *));
173 void inet_csk_clear_xmit_timers(struct sock *sk);
175 static inline void inet_csk_schedule_ack(struct sock *sk)
177 inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_SCHED;
180 static inline int inet_csk_ack_scheduled(const struct sock *sk)
182 return inet_csk(sk)->icsk_ack.pending & ICSK_ACK_SCHED;
185 static inline void inet_csk_delack_init(struct sock *sk)
187 memset(&inet_csk(sk)->icsk_ack, 0, sizeof(inet_csk(sk)->icsk_ack));
190 void inet_csk_delete_keepalive_timer(struct sock *sk);
191 void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long timeout);
193 static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what)
195 struct inet_connection_sock *icsk = inet_csk(sk);
197 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0) {
198 icsk->icsk_pending = 0;
199 #ifdef INET_CSK_CLEAR_TIMERS
200 sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
202 } else if (what == ICSK_TIME_DACK) {
203 icsk->icsk_ack.pending = 0;
204 icsk->icsk_ack.retry = 0;
205 #ifdef INET_CSK_CLEAR_TIMERS
206 sk_stop_timer(sk, &icsk->icsk_delack_timer);
209 pr_debug("inet_csk BUG: unknown timer value\n");
214 * Reset the retransmission timer
216 static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what,
218 const unsigned long max_when)
220 struct inet_connection_sock *icsk = inet_csk(sk);
222 if (when > max_when) {
223 pr_debug("reset_xmit_timer: sk=%p %d when=0x%lx, caller=%p\n",
224 sk, what, when, (void *)_THIS_IP_);
228 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 ||
229 what == ICSK_TIME_LOSS_PROBE || what == ICSK_TIME_REO_TIMEOUT) {
230 icsk->icsk_pending = what;
231 icsk->icsk_timeout = jiffies + when;
232 sk_reset_timer(sk, &icsk->icsk_retransmit_timer, icsk->icsk_timeout);
233 } else if (what == ICSK_TIME_DACK) {
234 icsk->icsk_ack.pending |= ICSK_ACK_TIMER;
235 icsk->icsk_ack.timeout = jiffies + when;
236 sk_reset_timer(sk, &icsk->icsk_delack_timer, icsk->icsk_ack.timeout);
238 pr_debug("inet_csk BUG: unknown timer value\n");
242 static inline unsigned long
243 inet_csk_rto_backoff(const struct inet_connection_sock *icsk,
244 unsigned long max_when)
246 u64 when = (u64)icsk->icsk_rto << icsk->icsk_backoff;
248 return (unsigned long)min_t(u64, when, max_when);
251 struct sock *inet_csk_accept(struct sock *sk, int flags, int *err, bool kern);
253 int inet_csk_get_port(struct sock *sk, unsigned short snum);
255 struct dst_entry *inet_csk_route_req(const struct sock *sk, struct flowi4 *fl4,
256 const struct request_sock *req);
257 struct dst_entry *inet_csk_route_child_sock(const struct sock *sk,
259 const struct request_sock *req);
261 struct sock *inet_csk_reqsk_queue_add(struct sock *sk,
262 struct request_sock *req,
264 void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
265 unsigned long timeout);
266 struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child,
267 struct request_sock *req,
270 static inline void inet_csk_reqsk_queue_added(struct sock *sk)
272 reqsk_queue_added(&inet_csk(sk)->icsk_accept_queue);
275 static inline int inet_csk_reqsk_queue_len(const struct sock *sk)
277 return reqsk_queue_len(&inet_csk(sk)->icsk_accept_queue);
280 static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk)
282 return inet_csk_reqsk_queue_len(sk) >= sk->sk_max_ack_backlog;
285 void inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req);
286 void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req);
288 static inline void inet_csk_prepare_for_destroy_sock(struct sock *sk)
290 /* The below has to be done to allow calling inet_csk_destroy_sock */
291 sock_set_flag(sk, SOCK_DEAD);
292 percpu_counter_inc(sk->sk_prot->orphan_count);
295 void inet_csk_destroy_sock(struct sock *sk);
296 void inet_csk_prepare_forced_close(struct sock *sk);
299 * LISTEN is a special case for poll..
301 static inline __poll_t inet_csk_listen_poll(const struct sock *sk)
303 return !reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue) ?
304 (EPOLLIN | EPOLLRDNORM) : 0;
307 int inet_csk_listen_start(struct sock *sk, int backlog);
308 void inet_csk_listen_stop(struct sock *sk);
310 void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr);
312 /* update the fast reuse flag when adding a socket */
313 void inet_csk_update_fastreuse(struct inet_bind_bucket *tb,
316 struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu);
318 #define TCP_PINGPONG_THRESH 3
320 static inline void inet_csk_enter_pingpong_mode(struct sock *sk)
322 inet_csk(sk)->icsk_ack.pingpong = TCP_PINGPONG_THRESH;
325 static inline void inet_csk_exit_pingpong_mode(struct sock *sk)
327 inet_csk(sk)->icsk_ack.pingpong = 0;
330 static inline bool inet_csk_in_pingpong_mode(struct sock *sk)
332 return inet_csk(sk)->icsk_ack.pingpong >= TCP_PINGPONG_THRESH;
335 static inline void inet_csk_inc_pingpong_cnt(struct sock *sk)
337 struct inet_connection_sock *icsk = inet_csk(sk);
339 if (icsk->icsk_ack.pingpong < U8_MAX)
340 icsk->icsk_ack.pingpong++;
343 static inline bool inet_csk_has_ulp(struct sock *sk)
345 return inet_sk(sk)->is_icsk && !!inet_csk(sk)->icsk_ulp_ops;
348 #endif /* _INET_CONNECTION_SOCK_H */