tjh.dev
Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
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
24/* Cancel timers, when they are not required. */
25#undef INET_CSK_CLEAR_TIMERS
26
27struct inet_bind_bucket;
28struct inet_bind2_bucket;
29struct tcp_congestion_ops;
30
31/*
32 * Pointers to address related TCP functions
33 * (i.e. things that depend on the address family)
34 */
35struct inet_connection_sock_af_ops {
36 int (*queue_xmit)(struct sock *sk, struct sk_buff *skb, struct flowi *fl);
37 void (*send_check)(struct sock *sk, struct sk_buff *skb);
38 int (*rebuild_header)(struct sock *sk);
39 void (*sk_rx_dst_set)(struct sock *sk, const struct sk_buff *skb);
40 int (*conn_request)(struct sock *sk, struct sk_buff *skb);
41 struct sock *(*syn_recv_sock)(const struct sock *sk, struct sk_buff *skb,
42 struct request_sock *req,
43 struct dst_entry *dst,
44 struct request_sock *req_unhash,
45 bool *own_req);
46 u16 net_header_len;
47 u16 sockaddr_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);
54};
55
56/** inet_connection_sock - INET connection oriented sock
57 *
58 * @icsk_accept_queue: FIFO of established children
59 * @icsk_bind_hash: Bind node
60 * @icsk_bind2_hash: Bind node in the bhash2 table
61 * @icsk_timeout: Timeout
62 * @icsk_retransmit_timer: Resend (no ack)
63 * @icsk_rto: Retransmit timeout
64 * @icsk_pmtu_cookie Last pmtu seen by socket
65 * @icsk_ca_ops Pluggable congestion control hook
66 * @icsk_af_ops Operations which are AF_INET{4,6} specific
67 * @icsk_ulp_ops Pluggable ULP control hook
68 * @icsk_ulp_data ULP private data
69 * @icsk_clean_acked Clean acked data hook
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
81 */
82struct 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 struct inet_bind2_bucket *icsk_bind2_hash;
88 unsigned long icsk_timeout;
89 struct timer_list icsk_retransmit_timer;
90 struct timer_list icsk_delack_timer;
91 __u32 icsk_rto;
92 __u32 icsk_rto_min;
93 __u32 icsk_delack_max;
94 __u32 icsk_pmtu_cookie;
95 const struct tcp_congestion_ops *icsk_ca_ops;
96 const struct inet_connection_sock_af_ops *icsk_af_ops;
97 const struct tcp_ulp_ops *icsk_ulp_ops;
98 void __rcu *icsk_ulp_data;
99 void (*icsk_clean_acked)(struct sock *sk, u32 acked_seq);
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;
106 __u8 icsk_pending;
107 __u8 icsk_backoff;
108 __u8 icsk_syn_retries;
109 __u8 icsk_probes_out;
110 __u16 icsk_ext_hdr_len;
111 struct {
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 #define ATO_BITS 8
117 __u32 ato:ATO_BITS, /* Predicted tick of soft clock */
118 lrcv_flowlabel:20, /* last received ipv6 flowlabel */
119 unused:4;
120 unsigned long timeout; /* Currently scheduled timeout */
121 __u32 lrcvtime; /* timestamp of last received data packet */
122 __u16 last_seg_size; /* Size of last incoming segment */
123 __u16 rcv_mss; /* MSS used for delayed ACK decisions */
124 } icsk_ack;
125 struct {
126 /* Range of MTUs to search */
127 int search_high;
128 int search_low;
129
130 /* Information on the current probe. */
131 u32 probe_size:31,
132 /* Is the MTUP feature enabled for this connection? */
133 enabled:1;
134
135 u32 probe_timestamp;
136 } icsk_mtup;
137 u32 icsk_probes_tstamp;
138 u32 icsk_user_timeout;
139
140 u64 icsk_ca_priv[104 / sizeof(u64)];
141#define ICSK_CA_PRIV_SIZE sizeof_field(struct inet_connection_sock, icsk_ca_priv)
142};
143
144#define ICSK_TIME_RETRANS 1 /* Retransmit timer */
145#define ICSK_TIME_DACK 2 /* Delayed ack timer */
146#define ICSK_TIME_PROBE0 3 /* Zero window probe timer */
147#define ICSK_TIME_LOSS_PROBE 5 /* Tail loss probe timer */
148#define ICSK_TIME_REO_TIMEOUT 6 /* Reordering timer */
149
150#define inet_csk(ptr) container_of_const(ptr, struct inet_connection_sock, icsk_inet.sk)
151
152static inline void *inet_csk_ca(const struct sock *sk)
153{
154 return (void *)inet_csk(sk)->icsk_ca_priv;
155}
156
157struct sock *inet_csk_clone_lock(const struct sock *sk,
158 const struct request_sock *req,
159 const gfp_t priority);
160
161enum inet_csk_ack_state_t {
162 ICSK_ACK_SCHED = 1,
163 ICSK_ACK_TIMER = 2,
164 ICSK_ACK_PUSHED = 4,
165 ICSK_ACK_PUSHED2 = 8,
166 ICSK_ACK_NOW = 16, /* Send the next ACK immediately (once) */
167 ICSK_ACK_NOMEM = 32,
168};
169
170void inet_csk_init_xmit_timers(struct sock *sk,
171 void (*retransmit_handler)(struct timer_list *),
172 void (*delack_handler)(struct timer_list *),
173 void (*keepalive_handler)(struct timer_list *));
174void inet_csk_clear_xmit_timers(struct sock *sk);
175void inet_csk_clear_xmit_timers_sync(struct sock *sk);
176
177static inline void inet_csk_schedule_ack(struct sock *sk)
178{
179 inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_SCHED;
180}
181
182static inline int inet_csk_ack_scheduled(const struct sock *sk)
183{
184 return inet_csk(sk)->icsk_ack.pending & ICSK_ACK_SCHED;
185}
186
187static inline void inet_csk_delack_init(struct sock *sk)
188{
189 memset(&inet_csk(sk)->icsk_ack, 0, sizeof(inet_csk(sk)->icsk_ack));
190}
191
192void inet_csk_delete_keepalive_timer(struct sock *sk);
193void inet_csk_reset_keepalive_timer(struct sock *sk, unsigned long timeout);
194
195static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what)
196{
197 struct inet_connection_sock *icsk = inet_csk(sk);
198
199 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0) {
200 icsk->icsk_pending = 0;
201#ifdef INET_CSK_CLEAR_TIMERS
202 sk_stop_timer(sk, &icsk->icsk_retransmit_timer);
203#endif
204 } else if (what == ICSK_TIME_DACK) {
205 icsk->icsk_ack.pending = 0;
206 icsk->icsk_ack.retry = 0;
207#ifdef INET_CSK_CLEAR_TIMERS
208 sk_stop_timer(sk, &icsk->icsk_delack_timer);
209#endif
210 } else {
211 pr_debug("inet_csk BUG: unknown timer value\n");
212 }
213}
214
215/*
216 * Reset the retransmission timer
217 */
218static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what,
219 unsigned long when,
220 const unsigned long max_when)
221{
222 struct inet_connection_sock *icsk = inet_csk(sk);
223
224 if (when > max_when) {
225 pr_debug("reset_xmit_timer: sk=%p %d when=0x%lx, caller=%p\n",
226 sk, what, when, (void *)_THIS_IP_);
227 when = max_when;
228 }
229
230 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 ||
231 what == ICSK_TIME_LOSS_PROBE || what == ICSK_TIME_REO_TIMEOUT) {
232 icsk->icsk_pending = what;
233 icsk->icsk_timeout = jiffies + when;
234 sk_reset_timer(sk, &icsk->icsk_retransmit_timer, icsk->icsk_timeout);
235 } else if (what == ICSK_TIME_DACK) {
236 icsk->icsk_ack.pending |= ICSK_ACK_TIMER;
237 icsk->icsk_ack.timeout = jiffies + when;
238 sk_reset_timer(sk, &icsk->icsk_delack_timer, icsk->icsk_ack.timeout);
239 } else {
240 pr_debug("inet_csk BUG: unknown timer value\n");
241 }
242}
243
244static inline unsigned long
245inet_csk_rto_backoff(const struct inet_connection_sock *icsk,
246 unsigned long max_when)
247{
248 u64 when = (u64)icsk->icsk_rto << icsk->icsk_backoff;
249
250 return (unsigned long)min_t(u64, when, max_when);
251}
252
253struct sock *inet_csk_accept(struct sock *sk, struct proto_accept_arg *arg);
254
255int inet_csk_get_port(struct sock *sk, unsigned short snum);
256
257struct dst_entry *inet_csk_route_req(const struct sock *sk, struct flowi4 *fl4,
258 const struct request_sock *req);
259struct dst_entry *inet_csk_route_child_sock(const struct sock *sk,
260 struct sock *newsk,
261 const struct request_sock *req);
262
263struct sock *inet_csk_reqsk_queue_add(struct sock *sk,
264 struct request_sock *req,
265 struct sock *child);
266bool inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req,
267 unsigned long timeout);
268struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child,
269 struct request_sock *req,
270 bool own_req);
271
272static inline void inet_csk_reqsk_queue_added(struct sock *sk)
273{
274 reqsk_queue_added(&inet_csk(sk)->icsk_accept_queue);
275}
276
277static inline int inet_csk_reqsk_queue_len(const struct sock *sk)
278{
279 return reqsk_queue_len(&inet_csk(sk)->icsk_accept_queue);
280}
281
282static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk)
283{
284 return inet_csk_reqsk_queue_len(sk) >= READ_ONCE(sk->sk_max_ack_backlog);
285}
286
287bool inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req);
288void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req);
289
290static inline unsigned long
291reqsk_timeout(struct request_sock *req, unsigned long max_timeout)
292{
293 u64 timeout = (u64)req->timeout << req->num_timeout;
294
295 return (unsigned long)min_t(u64, timeout, max_timeout);
296}
297
298static inline void inet_csk_prepare_for_destroy_sock(struct sock *sk)
299{
300 /* The below has to be done to allow calling inet_csk_destroy_sock */
301 sock_set_flag(sk, SOCK_DEAD);
302 this_cpu_inc(*sk->sk_prot->orphan_count);
303}
304
305void inet_csk_destroy_sock(struct sock *sk);
306void inet_csk_prepare_forced_close(struct sock *sk);
307
308/*
309 * LISTEN is a special case for poll..
310 */
311static inline __poll_t inet_csk_listen_poll(const struct sock *sk)
312{
313 return !reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue) ?
314 (EPOLLIN | EPOLLRDNORM) : 0;
315}
316
317int inet_csk_listen_start(struct sock *sk);
318void inet_csk_listen_stop(struct sock *sk);
319
320void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr);
321
322/* update the fast reuse flag when adding a socket */
323void inet_csk_update_fastreuse(struct inet_bind_bucket *tb,
324 struct sock *sk);
325
326struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu);
327
328static inline void inet_csk_enter_pingpong_mode(struct sock *sk)
329{
330 inet_csk(sk)->icsk_ack.pingpong =
331 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_pingpong_thresh);
332}
333
334static inline void inet_csk_exit_pingpong_mode(struct sock *sk)
335{
336 inet_csk(sk)->icsk_ack.pingpong = 0;
337}
338
339static inline bool inet_csk_in_pingpong_mode(struct sock *sk)
340{
341 return inet_csk(sk)->icsk_ack.pingpong >=
342 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_pingpong_thresh);
343}
344
345static inline void inet_csk_inc_pingpong_cnt(struct sock *sk)
346{
347 struct inet_connection_sock *icsk = inet_csk(sk);
348
349 if (icsk->icsk_ack.pingpong < U8_MAX)
350 icsk->icsk_ack.pingpong++;
351}
352
353static inline bool inet_csk_has_ulp(const struct sock *sk)
354{
355 return inet_test_bit(IS_ICSK, sk) && !!inet_csk(sk)->icsk_ulp_ops;
356}
357
358static inline void inet_init_csk_locks(struct sock *sk)
359{
360 struct inet_connection_sock *icsk = inet_csk(sk);
361
362 spin_lock_init(&icsk->icsk_accept_queue.rskq_lock);
363 spin_lock_init(&icsk->icsk_accept_queue.fastopenq.lock);
364}
365
366#endif /* _INET_CONNECTION_SOCK_H */