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 #include <linux/android_kabi.h>
21
22 #include <net/inet_sock.h>
23 #include <net/request_sock.h>
24
25 /* Cancel timers, when they are not required. */
26 #undef INET_CSK_CLEAR_TIMERS
27
28 struct inet_bind_bucket;
29 struct inet_bind2_bucket;
30 struct tcp_congestion_ops;
31
32 /*
33 * Pointers to address related TCP functions
34 * (i.e. things that depend on the address family)
35 */
36 struct inet_connection_sock_af_ops {
37 int (*queue_xmit)(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
38 void (*send_check)(struct sock *sk, struct sk_buff *skb);
39 int (*rebuild_header)(struct sock *sk);
40 void (*sk_rx_dst_set)(struct sock *sk, const struct sk_buff *skb);
41 int (*conn_request)(struct sock *sk, struct sk_buff *skb);
42 struct sock *(*syn_recv_sock)(const struct sock *sk, struct sk_buff *skb,
43 struct request_sock *req,
44 struct dst_entry *dst,
45 struct request_sock *req_unhash,
46 bool *own_req);
47 u16 net_header_len;
48 u16 sockaddr_len;
49 int (*setsockopt)(struct sock *sk, int level, int optname,
50 sockptr_t optval, unsigned int optlen);
51 int (*getsockopt)(struct sock *sk, int level, int optname,
52 char __user *optval, int __user *optlen);
53 void (*addr2sockaddr)(struct sock *sk, struct sockaddr *);
54 void (*mtu_reduced)(struct sock *sk);
55
56 ANDROID_KABI_RESERVE(1);
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_bind2_hash: Bind node in the bhash2 table
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_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 struct inet_bind2_bucket *icsk_bind2_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 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 struct {
115 __u8 pending; /* ACK is pending */
116 __u8 quick; /* Scheduled number of quick acks */
117 __u8 pingpong; /* The session is interactive */
118 __u8 retry; /* Number of attempts */
119 #define ATO_BITS 8
120 __u32 ato:ATO_BITS, /* Predicted tick of soft clock */
121 lrcv_flowlabel:20, /* last received ipv6 flowlabel */
122 unused:4;
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 /* Range of MTUs to search */
130 int search_high;
131 int search_low;
132
133 /* Information on the current probe. */
134 u32 probe_size:31,
135 /* Is the MTUP feature enabled for this connection? */
136 enabled:1;
137
138 u32 probe_timestamp;
139 } icsk_mtup;
140 u32 icsk_probes_tstamp;
141 u32 icsk_user_timeout;
142
143 ANDROID_KABI_RESERVE(1);
144
145 u64 icsk_ca_priv[104 / sizeof(u64)];
146 #define ICSK_CA_PRIV_SIZE sizeof_field(struct inet_connection_sock, icsk_ca_priv)
147 };
148
149 #define ICSK_TIME_RETRANS 1 /* Retransmit timer */
150 #define ICSK_TIME_DACK 2 /* Delayed ack timer */
151 #define ICSK_TIME_PROBE0 3 /* Zero window probe timer */
152 #define ICSK_TIME_LOSS_PROBE 5 /* Tail loss probe timer */
153 #define ICSK_TIME_REO_TIMEOUT 6 /* Reordering timer */
154
155 #define inet_csk(ptr) container_of_const(ptr, struct inet_connection_sock, icsk_inet.sk)
156
inet_csk_ca(const struct sock * sk)157 static inline void *inet_csk_ca(const struct sock *sk)
158 {
159 return (void *)inet_csk(sk)->icsk_ca_priv;
160 }
161
162 struct sock *inet_csk_clone_lock(const struct sock *sk,
163 const struct request_sock *req,
164 const gfp_t priority);
165
166 enum inet_csk_ack_state_t {
167 ICSK_ACK_SCHED = 1,
168 ICSK_ACK_TIMER = 2,
169 ICSK_ACK_PUSHED = 4,
170 ICSK_ACK_PUSHED2 = 8,
171 ICSK_ACK_NOW = 16, /* Send the next ACK immediately (once) */
172 ICSK_ACK_NOMEM = 32,
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 void inet_csk_clear_xmit_timers_sync(struct sock *sk);
181
inet_csk_schedule_ack(struct sock * sk)182 static inline void inet_csk_schedule_ack(struct sock *sk)
183 {
184 inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_SCHED;
185 }
186
inet_csk_ack_scheduled(const struct sock * sk)187 static inline int inet_csk_ack_scheduled(const struct sock *sk)
188 {
189 return inet_csk(sk)->icsk_ack.pending & ICSK_ACK_SCHED;
190 }
191
inet_csk_delack_init(struct sock * sk)192 static inline void inet_csk_delack_init(struct sock *sk)
193 {
194 memset(&inet_csk(sk)->icsk_ack, 0, sizeof(inet_csk(sk)->icsk_ack));
195 }
196
197 void inet_csk_delete_keepalive_timer(struct sock *sk);
198 void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long timeout);
199
inet_csk_clear_xmit_timer(struct sock * sk,const int what)200 static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what)
201 {
202 struct inet_connection_sock *icsk = inet_csk(sk);
203
204 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0) {
205 icsk->icsk_pending = 0;
206 #ifdef INET_CSK_CLEAR_TIMERS
207 sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
208 #endif
209 } else if (what == ICSK_TIME_DACK) {
210 icsk->icsk_ack.pending = 0;
211 icsk->icsk_ack.retry = 0;
212 #ifdef INET_CSK_CLEAR_TIMERS
213 sk_stop_timer(sk, &icsk->icsk_delack_timer);
214 #endif
215 } else {
216 pr_debug("inet_csk BUG: unknown timer value\n");
217 }
218 }
219
220 /*
221 * Reset the retransmission timer
222 */
inet_csk_reset_xmit_timer(struct sock * sk,const int what,unsigned long when,const unsigned long max_when)223 static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what,
224 unsigned long when,
225 const unsigned long max_when)
226 {
227 struct inet_connection_sock *icsk = inet_csk(sk);
228
229 if (when > max_when) {
230 pr_debug("reset_xmit_timer: sk=%p %d when=0x%lx, caller=%p\n",
231 sk, what, when, (void *)_THIS_IP_);
232 when = max_when;
233 }
234
235 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 ||
236 what == ICSK_TIME_LOSS_PROBE || what == ICSK_TIME_REO_TIMEOUT) {
237 icsk->icsk_pending = what;
238 icsk->icsk_timeout = jiffies + when;
239 sk_reset_timer(sk, &icsk->icsk_retransmit_timer, icsk->icsk_timeout);
240 } else if (what == ICSK_TIME_DACK) {
241 icsk->icsk_ack.pending |= ICSK_ACK_TIMER;
242 icsk->icsk_ack.timeout = jiffies + when;
243 sk_reset_timer(sk, &icsk->icsk_delack_timer, icsk->icsk_ack.timeout);
244 } else {
245 pr_debug("inet_csk BUG: unknown timer value\n");
246 }
247 }
248
249 static inline unsigned long
inet_csk_rto_backoff(const struct inet_connection_sock * icsk,unsigned long max_when)250 inet_csk_rto_backoff(const struct inet_connection_sock *icsk,
251 unsigned long max_when)
252 {
253 u64 when = (u64)icsk->icsk_rto << icsk->icsk_backoff;
254
255 return (unsigned long)min_t(u64, when, max_when);
256 }
257
258 struct sock *inet_csk_accept(struct sock *sk, struct proto_accept_arg *arg);
259
260 int inet_csk_get_port(struct sock *sk, unsigned short snum);
261
262 struct dst_entry *inet_csk_route_req(const struct sock *sk, struct flowi4 *fl4,
263 const struct request_sock *req);
264 struct dst_entry *inet_csk_route_child_sock(const struct sock *sk,
265 struct sock *newsk,
266 const struct request_sock *req);
267
268 struct sock *inet_csk_reqsk_queue_add(struct sock *sk,
269 struct request_sock *req,
270 struct sock *child);
271 bool inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
272 unsigned long timeout);
273 struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child,
274 struct request_sock *req,
275 bool own_req);
276
inet_csk_reqsk_queue_added(struct sock * sk)277 static inline void inet_csk_reqsk_queue_added(struct sock *sk)
278 {
279 reqsk_queue_added(&inet_csk(sk)->icsk_accept_queue);
280 }
281
inet_csk_reqsk_queue_len(const struct sock * sk)282 static inline int inet_csk_reqsk_queue_len(const struct sock *sk)
283 {
284 return reqsk_queue_len(&inet_csk(sk)->icsk_accept_queue);
285 }
286
inet_csk_reqsk_queue_is_full(const struct sock * sk)287 static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk)
288 {
289 return inet_csk_reqsk_queue_len(sk) > READ_ONCE(sk->sk_max_ack_backlog);
290 }
291
292 bool inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req);
293 void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req);
294
295 static inline unsigned long
reqsk_timeout(struct request_sock * req,unsigned long max_timeout)296 reqsk_timeout(struct request_sock *req, unsigned long max_timeout)
297 {
298 u64 timeout = (u64)req->timeout << req->num_timeout;
299
300 return (unsigned long)min_t(u64, timeout, max_timeout);
301 }
302
inet_csk_prepare_for_destroy_sock(struct sock * sk)303 static inline void inet_csk_prepare_for_destroy_sock(struct sock *sk)
304 {
305 /* The below has to be done to allow calling inet_csk_destroy_sock */
306 sock_set_flag(sk, SOCK_DEAD);
307 this_cpu_inc(*sk->sk_prot->orphan_count);
308 }
309
310 void inet_csk_destroy_sock(struct sock *sk);
311 void inet_csk_prepare_forced_close(struct sock *sk);
312
313 /*
314 * LISTEN is a special case for poll..
315 */
inet_csk_listen_poll(const struct sock * sk)316 static inline __poll_t inet_csk_listen_poll(const struct sock *sk)
317 {
318 return !reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue) ?
319 (EPOLLIN | EPOLLRDNORM) : 0;
320 }
321
322 int inet_csk_listen_start(struct sock *sk);
323 void inet_csk_listen_stop(struct sock *sk);
324
325 void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr);
326
327 /* update the fast reuse flag when adding a socket */
328 void inet_csk_update_fastreuse(struct inet_bind_bucket *tb,
329 struct sock *sk);
330
331 struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu);
332
inet_csk_enter_pingpong_mode(struct sock * sk)333 static inline void inet_csk_enter_pingpong_mode(struct sock *sk)
334 {
335 inet_csk(sk)->icsk_ack.pingpong =
336 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_pingpong_thresh);
337 }
338
inet_csk_exit_pingpong_mode(struct sock * sk)339 static inline void inet_csk_exit_pingpong_mode(struct sock *sk)
340 {
341 inet_csk(sk)->icsk_ack.pingpong = 0;
342 }
343
inet_csk_in_pingpong_mode(struct sock * sk)344 static inline bool inet_csk_in_pingpong_mode(struct sock *sk)
345 {
346 return inet_csk(sk)->icsk_ack.pingpong >=
347 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_pingpong_thresh);
348 }
349
inet_csk_inc_pingpong_cnt(struct sock * sk)350 static inline void inet_csk_inc_pingpong_cnt(struct sock *sk)
351 {
352 struct inet_connection_sock *icsk = inet_csk(sk);
353
354 if (icsk->icsk_ack.pingpong < U8_MAX)
355 icsk->icsk_ack.pingpong++;
356 }
357
inet_csk_has_ulp(const struct sock * sk)358 static inline bool inet_csk_has_ulp(const struct sock *sk)
359 {
360 return inet_test_bit(IS_ICSK, sk) && !!inet_csk(sk)->icsk_ulp_ops;
361 }
362
inet_init_csk_locks(struct sock * sk)363 static inline void inet_init_csk_locks(struct sock *sk)
364 {
365 struct inet_connection_sock *icsk = inet_csk(sk);
366
367 spin_lock_init(&icsk->icsk_accept_queue.rskq_lock);
368 spin_lock_init(&icsk->icsk_accept_queue.fastopenq.lock);
369 }
370
371 #endif /* _INET_CONNECTION_SOCK_H */
372