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 27 struct inet_bind_bucket; 28 struct inet_bind2_bucket; 29 struct tcp_congestion_ops; 30 31 /* 32 * Pointers to address related TCP functions 33 * (i.e. things that depend on the address family) 34 */ 35 struct 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 */ 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 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_rto_max; 94 __u32 icsk_delack_max; 95 __u32 icsk_pmtu_cookie; 96 const struct tcp_congestion_ops *icsk_ca_ops; 97 const struct inet_connection_sock_af_ops *icsk_af_ops; 98 const struct tcp_ulp_ops *icsk_ulp_ops; 99 void __rcu *icsk_ulp_data; 100 void (*icsk_clean_acked)(struct sock *sk, u32 acked_seq); 101 unsigned int (*icsk_sync_mss)(struct sock *sk, u32 pmtu); 102 __u8 icsk_ca_state:5, 103 icsk_ca_initialized:1, 104 icsk_ca_setsockopt:1, 105 icsk_ca_dst_locked:1; 106 __u8 icsk_retransmits; 107 __u8 icsk_pending; 108 __u8 icsk_backoff; 109 __u8 icsk_syn_retries; 110 __u8 icsk_probes_out; 111 __u16 icsk_ext_hdr_len; 112 struct { 113 __u8 pending; /* ACK is pending */ 114 __u8 quick; /* Scheduled number of quick acks */ 115 __u8 pingpong; /* The session is interactive */ 116 __u8 retry; /* Number of attempts */ 117 #define ATO_BITS 8 118 __u32 ato:ATO_BITS, /* Predicted tick of soft clock */ 119 lrcv_flowlabel:20, /* last received ipv6 flowlabel */ 120 unused:4; 121 unsigned long timeout; /* Currently scheduled timeout */ 122 __u32 lrcvtime; /* timestamp of last received data packet */ 123 __u16 last_seg_size; /* Size of last incoming segment */ 124 __u16 rcv_mss; /* MSS used for delayed ACK decisions */ 125 } icsk_ack; 126 struct { 127 /* Range of MTUs to search */ 128 int search_high; 129 int search_low; 130 131 /* Information on the current probe. */ 132 u32 probe_size:31, 133 /* Is the MTUP feature enabled for this connection? */ 134 enabled:1; 135 136 u32 probe_timestamp; 137 } icsk_mtup; 138 u32 icsk_probes_tstamp; 139 u32 icsk_user_timeout; 140 141 u64 icsk_ca_priv[104 / sizeof(u64)]; 142 #define ICSK_CA_PRIV_SIZE sizeof_field(struct inet_connection_sock, icsk_ca_priv) 143 }; 144 145 #define ICSK_TIME_RETRANS 1 /* Retransmit timer */ 146 #define ICSK_TIME_DACK 2 /* Delayed ack timer */ 147 #define ICSK_TIME_PROBE0 3 /* Zero window probe timer */ 148 #define ICSK_TIME_LOSS_PROBE 5 /* Tail loss probe timer */ 149 #define ICSK_TIME_REO_TIMEOUT 6 /* Reordering timer */ 150 151 #define inet_csk(ptr) container_of_const(ptr, struct inet_connection_sock, icsk_inet.sk) 152 153 static inline void *inet_csk_ca(const struct sock *sk) 154 { 155 return (void *)inet_csk(sk)->icsk_ca_priv; 156 } 157 158 struct sock *inet_csk_clone_lock(const struct sock *sk, 159 const struct request_sock *req, 160 const gfp_t priority); 161 162 enum inet_csk_ack_state_t { 163 ICSK_ACK_SCHED = 1, 164 ICSK_ACK_TIMER = 2, 165 ICSK_ACK_PUSHED = 4, 166 ICSK_ACK_PUSHED2 = 8, 167 ICSK_ACK_NOW = 16, /* Send the next ACK immediately (once) */ 168 ICSK_ACK_NOMEM = 32, 169 }; 170 171 void inet_csk_init_xmit_timers(struct sock *sk, 172 void (*retransmit_handler)(struct timer_list *), 173 void (*delack_handler)(struct timer_list *), 174 void (*keepalive_handler)(struct timer_list *)); 175 void inet_csk_clear_xmit_timers(struct sock *sk); 176 void inet_csk_clear_xmit_timers_sync(struct sock *sk); 177 178 static inline void inet_csk_schedule_ack(struct sock *sk) 179 { 180 inet_csk(sk)->icsk_ack.pending |= ICSK_ACK_SCHED; 181 } 182 183 static inline int inet_csk_ack_scheduled(const struct sock *sk) 184 { 185 return inet_csk(sk)->icsk_ack.pending & ICSK_ACK_SCHED; 186 } 187 188 static inline void inet_csk_delack_init(struct sock *sk) 189 { 190 memset(&inet_csk(sk)->icsk_ack, 0, sizeof(inet_csk(sk)->icsk_ack)); 191 } 192 193 static inline void inet_csk_clear_xmit_timer(struct sock *sk, const int what) 194 { 195 struct inet_connection_sock *icsk = inet_csk(sk); 196 197 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0) { 198 smp_store_release(&icsk->icsk_pending, 0); 199 #ifdef INET_CSK_CLEAR_TIMERS 200 sk_stop_timer(sk, &icsk->icsk_retransmit_timer); 201 #endif 202 } else if (what == ICSK_TIME_DACK) { 203 smp_store_release(&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); 207 #endif 208 } else { 209 pr_debug("inet_csk BUG: unknown timer value\n"); 210 } 211 } 212 213 /* 214 * Reset the retransmission timer 215 */ 216 static inline void inet_csk_reset_xmit_timer(struct sock *sk, const int what, 217 unsigned long when, 218 const unsigned long max_when) 219 { 220 struct inet_connection_sock *icsk = inet_csk(sk); 221 222 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_); 225 when = max_when; 226 } 227 228 if (what == ICSK_TIME_RETRANS || what == ICSK_TIME_PROBE0 || 229 what == ICSK_TIME_LOSS_PROBE || what == ICSK_TIME_REO_TIMEOUT) { 230 smp_store_release(&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 smp_store_release(&icsk->icsk_ack.pending, 235 icsk->icsk_ack.pending | ICSK_ACK_TIMER); 236 icsk->icsk_ack.timeout = jiffies + when; 237 sk_reset_timer(sk, &icsk->icsk_delack_timer, icsk->icsk_ack.timeout); 238 } else { 239 pr_debug("inet_csk BUG: unknown timer value\n"); 240 } 241 } 242 243 static inline unsigned long 244 inet_csk_rto_backoff(const struct inet_connection_sock *icsk, 245 unsigned long max_when) 246 { 247 u64 when = (u64)icsk->icsk_rto << icsk->icsk_backoff; 248 249 return (unsigned long)min_t(u64, when, max_when); 250 } 251 252 struct sock *inet_csk_accept(struct sock *sk, struct proto_accept_arg *arg); 253 254 int inet_csk_get_port(struct sock *sk, unsigned short snum); 255 256 struct dst_entry *inet_csk_route_req(const struct sock *sk, struct flowi4 *fl4, 257 const struct request_sock *req); 258 struct dst_entry *inet_csk_route_child_sock(const struct sock *sk, 259 struct sock *newsk, 260 const struct request_sock *req); 261 262 struct sock *inet_csk_reqsk_queue_add(struct sock *sk, 263 struct request_sock *req, 264 struct sock *child); 265 bool inet_csk_reqsk_queue_hash_add(struct sock *sk, struct request_sock *req, 266 unsigned long timeout); 267 struct sock *inet_csk_complete_hashdance(struct sock *sk, struct sock *child, 268 struct request_sock *req, 269 bool own_req); 270 271 static inline void inet_csk_reqsk_queue_added(struct sock *sk) 272 { 273 reqsk_queue_added(&inet_csk(sk)->icsk_accept_queue); 274 } 275 276 static inline int inet_csk_reqsk_queue_len(const struct sock *sk) 277 { 278 return reqsk_queue_len(&inet_csk(sk)->icsk_accept_queue); 279 } 280 281 static inline int inet_csk_reqsk_queue_is_full(const struct sock *sk) 282 { 283 return inet_csk_reqsk_queue_len(sk) > READ_ONCE(sk->sk_max_ack_backlog); 284 } 285 286 bool inet_csk_reqsk_queue_drop(struct sock *sk, struct request_sock *req); 287 void inet_csk_reqsk_queue_drop_and_put(struct sock *sk, struct request_sock *req); 288 289 static inline unsigned long 290 reqsk_timeout(struct request_sock *req, unsigned long max_timeout) 291 { 292 u64 timeout = (u64)req->timeout << req->num_timeout; 293 294 return (unsigned long)min_t(u64, timeout, max_timeout); 295 } 296 297 static inline void inet_csk_prepare_for_destroy_sock(struct sock *sk) 298 { 299 /* The below has to be done to allow calling inet_csk_destroy_sock */ 300 sock_set_flag(sk, SOCK_DEAD); 301 this_cpu_inc(*sk->sk_prot->orphan_count); 302 } 303 304 void inet_csk_destroy_sock(struct sock *sk); 305 void inet_csk_prepare_forced_close(struct sock *sk); 306 307 /* 308 * LISTEN is a special case for poll.. 309 */ 310 static inline __poll_t inet_csk_listen_poll(const struct sock *sk) 311 { 312 return !reqsk_queue_empty(&inet_csk(sk)->icsk_accept_queue) ? 313 (EPOLLIN | EPOLLRDNORM) : 0; 314 } 315 316 int inet_csk_listen_start(struct sock *sk); 317 void inet_csk_listen_stop(struct sock *sk); 318 319 void inet_csk_addr2sockaddr(struct sock *sk, struct sockaddr *uaddr); 320 321 /* update the fast reuse flag when adding a socket */ 322 void inet_csk_update_fastreuse(struct inet_bind_bucket *tb, 323 struct sock *sk); 324 325 struct dst_entry *inet_csk_update_pmtu(struct sock *sk, u32 mtu); 326 327 static inline void inet_csk_enter_pingpong_mode(struct sock *sk) 328 { 329 inet_csk(sk)->icsk_ack.pingpong = 330 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_pingpong_thresh); 331 } 332 333 static inline void inet_csk_exit_pingpong_mode(struct sock *sk) 334 { 335 inet_csk(sk)->icsk_ack.pingpong = 0; 336 } 337 338 static inline bool inet_csk_in_pingpong_mode(struct sock *sk) 339 { 340 return inet_csk(sk)->icsk_ack.pingpong >= 341 READ_ONCE(sock_net(sk)->ipv4.sysctl_tcp_pingpong_thresh); 342 } 343 344 static inline void inet_csk_inc_pingpong_cnt(struct sock *sk) 345 { 346 struct inet_connection_sock *icsk = inet_csk(sk); 347 348 if (icsk->icsk_ack.pingpong < U8_MAX) 349 icsk->icsk_ack.pingpong++; 350 } 351 352 static inline bool inet_csk_has_ulp(const struct sock *sk) 353 { 354 return inet_test_bit(IS_ICSK, sk) && !!inet_csk(sk)->icsk_ulp_ops; 355 } 356 357 static inline void inet_init_csk_locks(struct sock *sk) 358 { 359 struct inet_connection_sock *icsk = inet_csk(sk); 360 361 spin_lock_init(&icsk->icsk_accept_queue.rskq_lock); 362 spin_lock_init(&icsk->icsk_accept_queue.fastopenq.lock); 363 } 364 365 #endif /* _INET_CONNECTION_SOCK_H */ 366