1 | /* |
2 | * NET Generic infrastructure for INET connection oriented protocols. |
3 | * |
4 | * Definitions for inet_connection_sock |
5 | * |
6 | * Authors: Many people, see the TCP sources |
7 | * |
8 | * From code originally in TCP |
9 | * |
10 | * This program is free software; you can redistribute it and/or |
11 | * modify it under the terms of the GNU General Public License |
12 | * as published by the Free Software Foundation; either version |
13 | * 2 of the License, or (at your option) any later version. |
14 | */ |
15 | #ifndef _INET_CONNECTION_SOCK_H |
16 | #define _INET_CONNECTION_SOCK_H |
17 | |
18 | #include <linux/compiler.h> |
19 | #include <linux/string.h> |
20 | #include <linux/timer.h> |
21 | #include <linux/poll.h> |
22 | #include <linux/kernel.h> |
23 | |
24 | #include <net/inet_sock.h> |
25 | #include <net/request_sock.h> |
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 (*)(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 ; |
49 | u16 ; |
50 | u16 sockaddr_len; |
51 | int (*setsockopt)(struct sock *sk, int level, int optname, |
52 | char __user *optval, unsigned int optlen); |
53 | int (*getsockopt)(struct sock *sk, int level, int optname, |
54 | char __user *optval, int __user *optlen); |
55 | #ifdef CONFIG_COMPAT |
56 | int (*compat_setsockopt)(struct sock *sk, |
57 | int level, int optname, |
58 | char __user *optval, unsigned int optlen); |
59 | int (*compat_getsockopt)(struct sock *sk, |
60 | int level, int optname, |
61 | char __user *optval, int __user *optlen); |
62 | #endif |
63 | void (*addr2sockaddr)(struct sock *sk, struct sockaddr *); |
64 | void (*mtu_reduced)(struct sock *sk); |
65 | }; |
66 | |
67 | /** inet_connection_sock - INET connection oriented sock |
68 | * |
69 | * @icsk_accept_queue: FIFO of established children |
70 | * @icsk_bind_hash: Bind node |
71 | * @icsk_timeout: Timeout |
72 | * @icsk_retransmit_timer: Resend (no ack) |
73 | * @icsk_rto: Retransmit timeout |
74 | * @icsk_pmtu_cookie Last pmtu seen by socket |
75 | * @icsk_ca_ops Pluggable congestion control hook |
76 | * @icsk_af_ops Operations which are AF_INET{4,6} specific |
77 | * @icsk_ulp_ops Pluggable ULP control hook |
78 | * @icsk_ulp_data ULP private data |
79 | * @icsk_clean_acked Clean acked data hook |
80 | * @icsk_listen_portaddr_node hash to the portaddr listener hashtable |
81 | * @icsk_ca_state: Congestion control state |
82 | * @icsk_retransmits: Number of unrecovered [RTO] timeouts |
83 | * @icsk_pending: Scheduled timer event |
84 | * @icsk_backoff: Backoff |
85 | * @icsk_syn_retries: Number of allowed SYN (or equivalent) retries |
86 | * @icsk_probes_out: unanswered 0 window probes |
87 | * @icsk_ext_hdr_len: Network protocol overhead (IP/IPv6 options) |
88 | * @icsk_ack: Delayed ACK control data |
89 | * @icsk_mtup; MTU probing control data |
90 | */ |
91 | struct inet_connection_sock { |
92 | /* inet_sock has to be the first member! */ |
93 | struct inet_sock icsk_inet; |
94 | struct request_sock_queue icsk_accept_queue; |
95 | struct inet_bind_bucket *icsk_bind_hash; |
96 | unsigned long icsk_timeout; |
97 | struct timer_list icsk_retransmit_timer; |
98 | struct timer_list icsk_delack_timer; |
99 | __u32 icsk_rto; |
100 | __u32 icsk_pmtu_cookie; |
101 | const struct tcp_congestion_ops *icsk_ca_ops; |
102 | const struct inet_connection_sock_af_ops *icsk_af_ops; |
103 | const struct tcp_ulp_ops *icsk_ulp_ops; |
104 | void *icsk_ulp_data; |
105 | void (*icsk_clean_acked)(struct sock *sk, u32 acked_seq); |
106 | struct hlist_node icsk_listen_portaddr_node; |
107 | unsigned int (*icsk_sync_mss)(struct sock *sk, u32 pmtu); |
108 | __u8 icsk_ca_state:6, |
109 | icsk_ca_setsockopt:1, |
110 | icsk_ca_dst_locked:1; |
111 | __u8 icsk_retransmits; |
112 | __u8 icsk_pending; |
113 | __u8 icsk_backoff; |
114 | __u8 icsk_syn_retries; |
115 | __u8 icsk_probes_out; |
116 | __u16 icsk_ext_hdr_len; |
117 | struct { |
118 | __u8 pending; /* ACK is pending */ |
119 | __u8 quick; /* Scheduled number of quick acks */ |
120 | __u8 pingpong; /* The session is interactive */ |
121 | __u8 blocked; /* Delayed ACK was blocked by socket lock */ |
122 | __u32 ato; /* Predicted tick of soft clock */ |
123 | unsigned long timeout; /* Currently scheduled timeout */ |
124 | __u32 lrcvtime; /* timestamp of last received data packet */ |
125 | __u16 last_seg_size; /* Size of last incoming segment */ |
126 | __u16 rcv_mss; /* MSS used for delayed ACK decisions */ |
127 | } icsk_ack; |
128 | struct { |
129 | int enabled; |
130 | |
131 | /* Range of MTUs to search */ |
132 | int search_high; |
133 | int search_low; |
134 | |
135 | /* Information on the current probe. */ |
136 | int probe_size; |
137 | |
138 | u32 probe_timestamp; |
139 | } icsk_mtup; |
140 | u32 icsk_user_timeout; |
141 | |
142 | u64 icsk_ca_priv[104 / sizeof(u64)]; |
143 | #define ICSK_CA_PRIV_SIZE (13 * sizeof(u64)) |
144 | }; |
145 | |
146 | #define ICSK_TIME_RETRANS 1 /* Retransmit timer */ |
147 | #define ICSK_TIME_DACK 2 /* Delayed ack timer */ |
148 | #define ICSK_TIME_PROBE0 3 /* Zero window probe timer */ |
149 | #define ICSK_TIME_EARLY_RETRANS 4 /* Early retransmit timer */ |
150 | #define ICSK_TIME_LOSS_PROBE 5 /* Tail loss probe timer */ |
151 | #define ICSK_TIME_REO_TIMEOUT 6 /* Reordering timer */ |
152 | |
153 | static inline struct inet_connection_sock *inet_csk(const struct sock *sk) |
154 | { |
155 | return (struct inet_connection_sock *)sk; |
156 | } |
157 | |
158 | static inline void *inet_csk_ca(const struct sock *sk) |
159 | { |
160 | return (void *)inet_csk(sk)->icsk_ca_priv; |
161 | } |
162 | |
163 | struct sock *inet_csk_clone_lock(const struct sock *sk, |
164 | const struct request_sock *req, |
165 | const gfp_t priority); |
166 | |
167 | enum inet_csk_ack_state_t { |
168 | ICSK_ACK_SCHED = 1, |
169 | ICSK_ACK_TIMER = 2, |
170 | ICSK_ACK_PUSHED = 4, |
171 | ICSK_ACK_PUSHED2 = 8, |
172 | ICSK_ACK_NOW = 16 /* Send the next ACK immediately (once) */ |
173 | }; |
174 | |
175 | void inet_csk_init_xmit_timers(struct sock *sk, |
176 | void (*retransmit_handler)(struct timer_list *), |
177 | void (*delack_handler)(struct timer_list *), |
178 | void (*keepalive_handler)(struct timer_list *)); |
179 | void inet_csk_clear_xmit_timers(struct sock *sk); |
180 | |
181 | static inline void inet_csk_schedule_ack(struct sock *sk) |
182 | { |
183 | inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_SCHED; |
184 | } |
185 | |
186 | static inline int inet_csk_ack_scheduled(const struct sock *sk) |
187 | { |
188 | return inet_csk(sk)->icsk_ack.pending & ICSK_ACK_SCHED; |
189 | } |
190 | |
191 | static inline void inet_csk_delack_init(struct sock *sk) |
192 | { |
193 | memset(&inet_csk(sk)->icsk_ack, 0, sizeof(inet_csk(sk)->icsk_ack)); |
194 | } |
195 | |
196 | void inet_csk_delete_keepalive_timer(struct sock *sk); |
197 | void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long timeout); |
198 | |
199 | static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what) |
200 | { |
201 | struct inet_connection_sock *icsk = inet_csk(sk); |
202 | |
203 | if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0) { |
204 | icsk->icsk_pending = 0; |
205 | #ifdef INET_CSK_CLEAR_TIMERS |
206 | sk_stop_timer(sk, &icsk->icsk_retransmit_timer); |
207 | #endif |
208 | } else if (what == ICSK_TIME_DACK) { |
209 | icsk->icsk_ack.blocked = icsk->icsk_ack.pending = 0; |
210 | #ifdef INET_CSK_CLEAR_TIMERS |
211 | sk_stop_timer(sk, &icsk->icsk_delack_timer); |
212 | #endif |
213 | } else { |
214 | pr_debug("inet_csk BUG: unknown timer value\n" ); |
215 | } |
216 | } |
217 | |
218 | /* |
219 | * Reset the retransmission timer |
220 | */ |
221 | static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what, |
222 | unsigned long when, |
223 | const unsigned long max_when) |
224 | { |
225 | struct inet_connection_sock *icsk = inet_csk(sk); |
226 | |
227 | if (when > max_when) { |
228 | pr_debug("reset_xmit_timer: sk=%p %d when=0x%lx, caller=%p\n" , |
229 | sk, what, when, (void *)_THIS_IP_); |
230 | when = max_when; |
231 | } |
232 | |
233 | if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 || |
234 | what == ICSK_TIME_EARLY_RETRANS || what == ICSK_TIME_LOSS_PROBE || |
235 | what == ICSK_TIME_REO_TIMEOUT) { |
236 | icsk->icsk_pending = what; |
237 | icsk->icsk_timeout = jiffies + when; |
238 | sk_reset_timer(sk, &icsk->icsk_retransmit_timer, icsk->icsk_timeout); |
239 | } else if (what == ICSK_TIME_DACK) { |
240 | icsk->icsk_ack.pending |= ICSK_ACK_TIMER; |
241 | icsk->icsk_ack.timeout = jiffies + when; |
242 | sk_reset_timer(sk, &icsk->icsk_delack_timer, icsk->icsk_ack.timeout); |
243 | } else { |
244 | pr_debug("inet_csk BUG: unknown timer value\n" ); |
245 | } |
246 | } |
247 | |
248 | static inline unsigned long |
249 | inet_csk_rto_backoff(const struct inet_connection_sock *icsk, |
250 | unsigned long max_when) |
251 | { |
252 | u64 when = (u64)icsk->icsk_rto << icsk->icsk_backoff; |
253 | |
254 | return (unsigned long)min_t(u64, when, max_when); |
255 | } |
256 | |
257 | struct sock *inet_csk_accept(struct sock *sk, int flags, int *err, bool kern); |
258 | |
259 | int inet_csk_get_port(struct sock *sk, unsigned short snum); |
260 | |
261 | struct dst_entry *inet_csk_route_req(const struct sock *sk, struct flowi4 *fl4, |
262 | const struct request_sock *req); |
263 | struct dst_entry *inet_csk_route_child_sock(const struct sock *sk, |
264 | struct sock *newsk, |
265 | const struct request_sock *req); |
266 | |
267 | struct sock *inet_csk_reqsk_queue_add(struct sock *sk, |
268 | struct request_sock *req, |
269 | struct sock *child); |
270 | void inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req, |
271 | unsigned long timeout); |
272 | struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child, |
273 | struct request_sock *req, |
274 | bool own_req); |
275 | |
276 | static inline void inet_csk_reqsk_queue_added(struct sock *sk) |
277 | { |
278 | reqsk_queue_added(&inet_csk(sk)->icsk_accept_queue); |
279 | } |
280 | |
281 | static inline int inet_csk_reqsk_queue_len(const struct sock *sk) |
282 | { |
283 | return reqsk_queue_len(&inet_csk(sk)->icsk_accept_queue); |
284 | } |
285 | |
286 | static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk) |
287 | { |
288 | return inet_csk_reqsk_queue_len(sk) >= sk->sk_max_ack_backlog; |
289 | } |
290 | |
291 | void inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req); |
292 | void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req); |
293 | |
294 | void inet_csk_destroy_sock(struct sock *sk); |
295 | void inet_csk_prepare_forced_close(struct sock *sk); |
296 | |
297 | /* |
298 | * LISTEN is a special case for poll.. |
299 | */ |
300 | static inline __poll_t inet_csk_listen_poll(const struct sock *sk) |
301 | { |
302 | return !reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue) ? |
303 | (EPOLLIN | EPOLLRDNORM) : 0; |
304 | } |
305 | |
306 | int inet_csk_listen_start(struct sock *sk, int backlog); |
307 | void inet_csk_listen_stop(struct sock *sk); |
308 | |
309 | void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr); |
310 | |
311 | int inet_csk_compat_getsockopt(struct sock *sk, int level, int optname, |
312 | char __user *optval, int __user *optlen); |
313 | int inet_csk_compat_setsockopt(struct sock *sk, int level, int optname, |
314 | char __user *optval, unsigned int optlen); |
315 | |
316 | struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu); |
317 | |
318 | #define TCP_PINGPONG_THRESH 3 |
319 | |
320 | static inline void inet_csk_enter_pingpong_mode(struct sock *sk) |
321 | { |
322 | inet_csk(sk)->icsk_ack.pingpong = TCP_PINGPONG_THRESH; |
323 | } |
324 | |
325 | static inline void inet_csk_exit_pingpong_mode(struct sock *sk) |
326 | { |
327 | inet_csk(sk)->icsk_ack.pingpong = 0; |
328 | } |
329 | |
330 | static inline bool inet_csk_in_pingpong_mode(struct sock *sk) |
331 | { |
332 | return inet_csk(sk)->icsk_ack.pingpong >= TCP_PINGPONG_THRESH; |
333 | } |
334 | |
335 | static inline void inet_csk_inc_pingpong_cnt(struct sock *sk) |
336 | { |
337 | struct inet_connection_sock *icsk = inet_csk(sk); |
338 | |
339 | if (icsk->icsk_ack.pingpong < U8_MAX) |
340 | icsk->icsk_ack.pingpong++; |
341 | } |
342 | #endif /* _INET_CONNECTION_SOCK_H */ |
343 | |