tcp.h 71.7 KB
Newer Older
1
/* SPDX-License-Identifier: GPL-2.0-or-later */
L
Linus Torvalds 已提交
2 3 4 5 6 7 8 9 10
/*
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		Definitions for the TCP module.
 *
 * Version:	@(#)tcp.h	1.0.5	05/23/93
 *
11
 * Authors:	Ross Biro
L
Linus Torvalds 已提交
12 13 14 15 16 17 18 19 20
 *		Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
 */
#ifndef _TCP_H
#define _TCP_H

#define FASTRETRANS_DEBUG 1

#include <linux/list.h>
#include <linux/tcp.h>
21
#include <linux/bug.h>
L
Linus Torvalds 已提交
22 23 24
#include <linux/slab.h>
#include <linux/cache.h>
#include <linux/percpu.h>
25
#include <linux/skbuff.h>
26
#include <linux/kref.h>
27
#include <linux/ktime.h>
28
#include <linux/indirect_call_wrapper.h>
29 30

#include <net/inet_connection_sock.h>
31
#include <net/inet_timewait_sock.h>
32
#include <net/inet_hashtables.h>
L
Linus Torvalds 已提交
33
#include <net/checksum.h>
34
#include <net/request_sock.h>
35
#include <net/sock_reuseport.h>
L
Linus Torvalds 已提交
36 37 38
#include <net/sock.h>
#include <net/snmp.h>
#include <net/ip.h>
39
#include <net/tcp_states.h>
40
#include <net/inet_ecn.h>
41
#include <net/dst.h>
42
#include <net/mptcp.h>
43

L
Linus Torvalds 已提交
44
#include <linux/seq_file.h>
45
#include <linux/memcontrol.h>
L
Lawrence Brakmo 已提交
46
#include <linux/bpf-cgroup.h>
47
#include <linux/siphash.h>
L
Lawrence Brakmo 已提交
48

49
extern struct inet_hashinfo tcp_hashinfo;
L
Linus Torvalds 已提交
50

51
extern struct percpu_counter tcp_orphan_count;
52
void tcp_time_wait(struct sock *sk, int state, int timeo);
L
Linus Torvalds 已提交
53

54
#define MAX_TCP_HEADER	L1_CACHE_ALIGN(128 + MAX_HEADER)
A
Adam Langley 已提交
55
#define MAX_TCP_OPTION_SPACE 40
56 57
#define TCP_MIN_SND_MSS		48
#define TCP_MIN_GSO_SIZE	(TCP_MIN_SND_MSS - MAX_TCP_OPTION_SPACE)
L
Linus Torvalds 已提交
58

59
/*
L
Linus Torvalds 已提交
60
 * Never offer a window over 32767 without using window scaling. Some
61
 * poor stacks do signed 16bit maths!
L
Linus Torvalds 已提交
62 63 64 65 66 67
 */
#define MAX_TCP_WINDOW		32767U

/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
#define TCP_MIN_MSS		88U

J
Josh Hunt 已提交
68
/* The initial MTU to use for probing */
69
#define TCP_BASE_MSS		1024
J
John Heffner 已提交
70

71 72 73
/* probing interval, default to 10 minutes as per RFC4821 */
#define TCP_PROBE_INTERVAL	600

74 75 76
/* Specify interval when tcp mtu probing will stop */
#define TCP_PROBE_THRESHOLD	8

L
Linus Torvalds 已提交
77 78 79 80 81 82
/* After receiving this amount of duplicate ACKs fast retransmit starts. */
#define TCP_FASTRETRANS_THRESH 3

/* Maximal number of ACKs sent quickly to accelerate slow-start. */
#define TCP_MAX_QUICKACKS	16U

83 84 85
/* Maximal number of window scale according to RFC1323 */
#define TCP_MAX_WSCALE		14U

L
Linus Torvalds 已提交
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104
/* urg_data states */
#define TCP_URG_VALID	0x0100
#define TCP_URG_NOTYET	0x0200
#define TCP_URG_READ	0x0400

#define TCP_RETR1	3	/*
				 * This is how many retries it does before it
				 * tries to figure out if the gateway is
				 * down. Minimal RFC value is 3; it corresponds
				 * to ~3sec-8min depending on RTO.
				 */

#define TCP_RETR2	15	/*
				 * This should take at least
				 * 90 minutes to time out.
				 * RFC1122 says that the limit is 100 sec.
				 * 15 is ~13-30min depending on RTO.
				 */

105 106 107 108 109 110 111 112
#define TCP_SYN_RETRIES	 6	/* This is how many retries are done
				 * when active opening a connection.
				 * RFC1122 says the minimum retry MUST
				 * be at least 180secs.  Nevertheless
				 * this value is corresponding to
				 * 63secs of retransmission with the
				 * current initial RTO.
				 */
L
Linus Torvalds 已提交
113

114 115 116 117 118 119
#define TCP_SYNACK_RETRIES 5	/* This is how may retries are done
				 * when passive opening a connection.
				 * This is corresponding to 31secs of
				 * retransmission with the current
				 * initial RTO.
				 */
L
Linus Torvalds 已提交
120 121 122 123 124 125 126 127 128

#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
				  * state, about 60 seconds	*/
#define TCP_FIN_TIMEOUT	TCP_TIMEWAIT_LEN
                                 /* BSD style FIN_WAIT2 deadlock breaker.
				  * It used to be 3min, new value is 60sec,
				  * to combine FIN-WAIT-2 timeout with
				  * TIME-WAIT timer.
				  */
129
#define TCP_FIN_TIMEOUT_MAX (120 * HZ) /* max TCP_LINGER2 value (two minutes) */
L
Linus Torvalds 已提交
130 131 132 133 134 135 136 137 138 139 140

#define TCP_DELACK_MAX	((unsigned)(HZ/5))	/* maximal time to delay before sending an ACK */
#if HZ >= 100
#define TCP_DELACK_MIN	((unsigned)(HZ/25))	/* minimal time to delay before sending an ACK */
#define TCP_ATO_MIN	((unsigned)(HZ/25))
#else
#define TCP_DELACK_MIN	4U
#define TCP_ATO_MIN	4U
#endif
#define TCP_RTO_MAX	((unsigned)(120*HZ))
#define TCP_RTO_MIN	((unsigned)(HZ/5))
141
#define TCP_TIMEOUT_MIN	(2U) /* Min timeout for TCP timers in jiffies */
142
#define TCP_TIMEOUT_INIT ((unsigned)(1*HZ))	/* RFC6298 2.1 initial RTO value	*/
143 144 145 146 147 148
#define TCP_TIMEOUT_FALLBACK ((unsigned)(3*HZ))	/* RFC 1122 initial RTO value, now
						 * used as a fallback RTO for the
						 * initial data transmission if no
						 * valid RTT sample has been acquired,
						 * most likely due to retrans in 3WHS.
						 */
L
Linus Torvalds 已提交
149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177

#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
					                 * for local resources.
					                 */
#define TCP_KEEPALIVE_TIME	(120*60*HZ)	/* two hours */
#define TCP_KEEPALIVE_PROBES	9		/* Max of 9 keepalive probes	*/
#define TCP_KEEPALIVE_INTVL	(75*HZ)

#define MAX_TCP_KEEPIDLE	32767
#define MAX_TCP_KEEPINTVL	32767
#define MAX_TCP_KEEPCNT		127
#define MAX_TCP_SYNCNT		127

#define TCP_SYNQ_INTERVAL	(HZ/5)	/* Period of SYNACK timer */

#define TCP_PAWS_24DAYS	(60 * 60 * 24 * 24)
#define TCP_PAWS_MSL	60		/* Per-host timestamps are invalidated
					 * after this time. It should be equal
					 * (or greater than) TCP_TIMEWAIT_LEN
					 * to provide reliability equal to one
					 * provided by timewait state.
					 */
#define TCP_PAWS_WINDOW	1		/* Replay window for per-host
					 * timestamps. It must be less than
					 * minimal timewait lifetime.
					 */
/*
 *	TCP option
 */
178

L
Linus Torvalds 已提交
179 180 181 182 183 184 185
#define TCPOPT_NOP		1	/* Padding */
#define TCPOPT_EOL		0	/* End of options */
#define TCPOPT_MSS		2	/* Segment size negotiating */
#define TCPOPT_WINDOW		3	/* Window scaling */
#define TCPOPT_SACK_PERM        4       /* SACK Permitted */
#define TCPOPT_SACK             5       /* SACK Block */
#define TCPOPT_TIMESTAMP	8	/* Better RTT estimations/PAWS */
186
#define TCPOPT_MD5SIG		19	/* MD5 Signature (RFC2385) */
M
Mat Martineau 已提交
187
#define TCPOPT_MPTCP		30	/* Multipath TCP (RFC6824) */
188
#define TCPOPT_FASTOPEN		34	/* Fast open (RFC7413) */
Y
Yuchung Cheng 已提交
189 190 191 192 193
#define TCPOPT_EXP		254	/* Experimental */
/* Magic number to be after the option value for sharing TCP
 * experimental options. See draft-ietf-tcpm-experimental-options-00.txt
 */
#define TCPOPT_FASTOPEN_MAGIC	0xF989
194
#define TCPOPT_SMC_MAGIC	0xE2D4C3D9
L
Linus Torvalds 已提交
195 196 197 198 199 200 201 202 203

/*
 *     TCP option lengths
 */

#define TCPOLEN_MSS            4
#define TCPOLEN_WINDOW         3
#define TCPOLEN_SACK_PERM      2
#define TCPOLEN_TIMESTAMP      10
204
#define TCPOLEN_MD5SIG         18
205
#define TCPOLEN_FASTOPEN_BASE  2
Y
Yuchung Cheng 已提交
206
#define TCPOLEN_EXP_FASTOPEN_BASE  4
207
#define TCPOLEN_EXP_SMC_BASE   6
L
Linus Torvalds 已提交
208 209 210 211 212 213 214 215

/* But this is what stacks really send out. */
#define TCPOLEN_TSTAMP_ALIGNED		12
#define TCPOLEN_WSCALE_ALIGNED		4
#define TCPOLEN_SACKPERM_ALIGNED	4
#define TCPOLEN_SACK_BASE		2
#define TCPOLEN_SACK_BASE_ALIGNED	4
#define TCPOLEN_SACK_PERBLOCK		8
216
#define TCPOLEN_MD5SIG_ALIGNED		20
A
Adam Langley 已提交
217
#define TCPOLEN_MSS_ALIGNED		4
218
#define TCPOLEN_EXP_SMC_BASE_ALIGNED	8
L
Linus Torvalds 已提交
219 220 221 222

/* Flags in tp->nonagle */
#define TCP_NAGLE_OFF		1	/* Nagle's algo is disabled */
#define TCP_NAGLE_CORK		2	/* Socket is corked	    */
S
Stephen Hemminger 已提交
223
#define TCP_NAGLE_PUSH		4	/* Cork is overridden for already queued data */
L
Linus Torvalds 已提交
224

A
Andreas Petlund 已提交
225 226 227
/* TCP thin-stream limits */
#define TCP_THIN_LINEAR_RETRIES 6       /* After 6 linear retries, do exp. backoff */

228
/* TCP initial congestion window as per rfc6928 */
229 230
#define TCP_INIT_CWND		10

231 232
/* Bit Flags for sysctl_tcp_fastopen */
#define	TFO_CLIENT_ENABLE	1
233
#define	TFO_SERVER_ENABLE	2
234
#define	TFO_CLIENT_NO_COOKIE	4	/* Data in SYN w/o cookie option */
235

236 237 238 239
/* Accept SYN data w/o any cookie option */
#define	TFO_SERVER_COOKIE_NOT_REQD	0x200

/* Force enable TFO on all listeners, i.e., not requiring the
240
 * TCP_FASTOPEN socket option.
241 242 243
 */
#define	TFO_SERVER_WO_SOCKOPT1	0x400

244

L
Linus Torvalds 已提交
245 246
/* sysctl variables for tcp */
extern int sysctl_tcp_max_orphans;
247
extern long sysctl_tcp_mem[3];
248

249
#define TCP_RACK_LOSS_DETECTION  0x1 /* Use RACK to detect losses */
250
#define TCP_RACK_STATIC_REO_WND  0x2 /* Use static RACK reo wnd */
251
#define TCP_RACK_NO_DUPTHRESH    0x4 /* Do not use DUPACK threshold in RACK */
252

E
Eric Dumazet 已提交
253
extern atomic_long_t tcp_memory_allocated;
254
extern struct percpu_counter tcp_sockets_allocated;
255
extern unsigned long tcp_memory_pressure;
L
Linus Torvalds 已提交
256

257 258 259
/* optimized version of sk_under_memory_pressure() for TCP sockets */
static inline bool tcp_under_memory_pressure(const struct sock *sk)
{
260 261
	if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
	    mem_cgroup_under_socket_pressure(sk->sk_memcg))
262
		return true;
263

264
	return READ_ONCE(tcp_memory_pressure);
265
}
L
Linus Torvalds 已提交
266 267 268 269 270
/*
 * The next routines deal with comparing 32 bit unsigned ints
 * and worry about wraparound (automatic with unsigned arithmetic).
 */

E
Eric Dumazet 已提交
271
static inline bool before(__u32 seq1, __u32 seq2)
L
Linus Torvalds 已提交
272
{
273
        return (__s32)(seq1-seq2) < 0;
L
Linus Torvalds 已提交
274
}
275
#define after(seq2, seq1) 	before(seq1, seq2)
L
Linus Torvalds 已提交
276 277

/* is s2<=s1<=s3 ? */
E
Eric Dumazet 已提交
278
static inline bool between(__u32 seq1, __u32 seq2, __u32 seq3)
L
Linus Torvalds 已提交
279 280 281 282
{
	return seq3 - seq2 >= seq1 - seq2;
}

A
Arun Sharma 已提交
283 284 285 286 287 288 289 290
static inline bool tcp_out_of_memory(struct sock *sk)
{
	if (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
	    sk_memory_allocated(sk) > sk_prot_mem_limits(sk, 2))
		return true;
	return false;
}

291 292
void sk_forced_mem_schedule(struct sock *sk, int size);

293
static inline bool tcp_too_many_orphans(struct sock *sk, int shift)
294
{
295 296 297 298 299 300 301 302 303
	struct percpu_counter *ocp = sk->sk_prot->orphan_count;
	int orphans = percpu_counter_read_positive(ocp);

	if (orphans << shift > sysctl_tcp_max_orphans) {
		orphans = percpu_counter_sum_positive(ocp);
		if (orphans << shift > sysctl_tcp_max_orphans)
			return true;
	}
	return false;
304
}
L
Linus Torvalds 已提交
305

306
bool tcp_check_oom(struct sock *sk, int shift);
A
Arun Sharma 已提交
307

308

L
Linus Torvalds 已提交
309 310
extern struct proto tcp_prot;

311
#define TCP_INC_STATS(net, field)	SNMP_INC_STATS((net)->mib.tcp_statistics, field)
E
Eric Dumazet 已提交
312
#define __TCP_INC_STATS(net, field)	__SNMP_INC_STATS((net)->mib.tcp_statistics, field)
313
#define TCP_DEC_STATS(net, field)	SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
314
#define TCP_ADD_STATS(net, field, val)	SNMP_ADD_STATS((net)->mib.tcp_statistics, field, val)
L
Linus Torvalds 已提交
315

316 317
void tcp_tasklet_init(void);

318
int tcp_v4_err(struct sk_buff *skb, u32);
319 320 321

void tcp_shutdown(struct sock *sk, int how);

322
int tcp_v4_early_demux(struct sk_buff *skb);
323 324
int tcp_v4_rcv(struct sk_buff *skb);

P
Paolo Abeni 已提交
325
void tcp_remove_empty_skb(struct sock *sk, struct sk_buff *skb);
326
int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
327
int tcp_sendmsg(struct sock *sk, struct msghdr *msg, size_t size);
328
int tcp_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t size);
329 330
int tcp_sendpage(struct sock *sk, struct page *page, int offset, size_t size,
		 int flags);
331 332
int tcp_sendpage_locked(struct sock *sk, struct page *page, int offset,
			size_t size, int flags);
P
Paolo Abeni 已提交
333 334
struct sk_buff *tcp_build_frag(struct sock *sk, int size_goal, int flags,
			       struct page *page, int offset, size_t *size);
335 336
ssize_t do_tcp_sendpages(struct sock *sk, struct page *page, int offset,
		 size_t size, int flags);
337 338 339
int tcp_send_mss(struct sock *sk, int *size_goal, int flags);
void tcp_push(struct sock *sk, int flags, int mss_now, int nonagle,
	      int size_goal);
340 341 342 343 344
void tcp_release_cb(struct sock *sk);
void tcp_wfree(struct sk_buff *skb);
void tcp_write_timer_handler(struct sock *sk);
void tcp_delack_timer_handler(struct sock *sk);
int tcp_ioctl(struct sock *sk, int cmd, unsigned long arg);
345
int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb);
346
void tcp_rcv_established(struct sock *sk, struct sk_buff *skb);
347 348 349 350 351 352
void tcp_rcv_space_adjust(struct sock *sk);
int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp);
void tcp_twsk_destructor(struct sock *sk);
ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
			struct pipe_inode_info *pipe, size_t len,
			unsigned int flags);
J
Jens Axboe 已提交
353

354
void tcp_enter_quickack_mode(struct sock *sk, unsigned int max_quickacks);
355 356
static inline void tcp_dec_quickack_mode(struct sock *sk,
					 const unsigned int pkts)
L
Linus Torvalds 已提交
357
{
358
	struct inet_connection_sock *icsk = inet_csk(sk);
359

360 361 362
	if (icsk->icsk_ack.quick) {
		if (pkts >= icsk->icsk_ack.quick) {
			icsk->icsk_ack.quick = 0;
363
			/* Leaving quickack mode we deflate ATO. */
364
			icsk->icsk_ack.ato   = TCP_ATO_MIN;
365
		} else
366
			icsk->icsk_ack.quick -= pkts;
L
Linus Torvalds 已提交
367 368 369
	}
}

370 371 372
#define	TCP_ECN_OK		1
#define	TCP_ECN_QUEUE_CWR	2
#define	TCP_ECN_DEMAND_CWR	4
373
#define	TCP_ECN_SEEN		8
374

E
Eric Dumazet 已提交
375
enum tcp_tw_status {
L
Linus Torvalds 已提交
376 377 378 379 380 381 382
	TCP_TW_SUCCESS = 0,
	TCP_TW_RST = 1,
	TCP_TW_ACK = 2,
	TCP_TW_SYN = 3
};


383 384 385 386
enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
					      struct sk_buff *skb,
					      const struct tcphdr *th);
struct sock *tcp_check_req(struct sock *sk, struct sk_buff *skb,
387 388
			   struct request_sock *req, bool fastopen,
			   bool *lost_race);
389 390
int tcp_child_process(struct sock *parent, struct sock *child,
		      struct sk_buff *skb);
391
void tcp_enter_loss(struct sock *sk);
392
void tcp_cwnd_reduction(struct sock *sk, int newly_acked_sacked, int newly_lost, int flag);
393 394 395 396
void tcp_clear_retrans(struct tcp_sock *tp);
void tcp_update_metrics(struct sock *sk);
void tcp_init_metrics(struct sock *sk);
void tcp_metrics_init(void);
397
bool tcp_peer_is_proven(struct request_sock *req, struct dst_entry *dst);
P
Paolo Abeni 已提交
398
void __tcp_close(struct sock *sk, long timeout);
399 400
void tcp_close(struct sock *sk, long timeout);
void tcp_init_sock(struct sock *sk);
401
void tcp_init_transfer(struct sock *sk, int bpf_op, struct sk_buff *skb);
402 403
__poll_t tcp_poll(struct file *file, struct socket *sock,
		      struct poll_table_struct *wait);
404 405
int tcp_getsockopt(struct sock *sk, int level, int optname,
		   char __user *optval, int __user *optlen);
406 407
int tcp_setsockopt(struct sock *sk, int level, int optname, sockptr_t optval,
		   unsigned int optlen);
408
void tcp_set_keepalive(struct sock *sk, int val);
409
void tcp_syn_ack_timeout(const struct request_sock *req);
410 411
int tcp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, int nonblock,
		int flags, int *addr_len);
412
int tcp_set_rcvlowat(struct sock *sk, int val);
413
int tcp_set_window_clamp(struct sock *sk, int val);
414
void tcp_data_ready(struct sock *sk);
415
#ifdef CONFIG_MMU
416 417
int tcp_mmap(struct file *file, struct socket *sock,
	     struct vm_area_struct *vma);
418
#endif
419
void tcp_parse_options(const struct net *net, const struct sk_buff *skb,
420 421 422
		       struct tcp_options_received *opt_rx,
		       int estab, struct tcp_fastopen_cookie *foc);
const u8 *tcp_parse_md5sig_option(const struct tcphdr *th);
423

424 425 426 427 428 429 430 431 432 433
/*
 *	BPF SKB-less helpers
 */
u16 tcp_v4_get_syncookie(struct sock *sk, struct iphdr *iph,
			 struct tcphdr *th, u32 *cookie);
u16 tcp_v6_get_syncookie(struct sock *sk, struct ipv6hdr *iph,
			 struct tcphdr *th, u32 *cookie);
u16 tcp_get_syncookie_mss(struct request_sock_ops *rsk_ops,
			  const struct tcp_request_sock_ops *af_ops,
			  struct sock *sk, struct tcphdr *th);
L
Linus Torvalds 已提交
434 435 436 437
/*
 *	TCP v4 functions exported for the inet6 API
 */

438
void tcp_v4_send_check(struct sock *sk, struct sk_buff *skb);
439
void tcp_v4_mtu_reduced(struct sock *sk);
440
void tcp_req_err(struct sock *sk, u32 seq, bool abort);
441
void tcp_ld_RTO_revert(struct sock *sk, u32 seq);
442
int tcp_v4_conn_request(struct sock *sk, struct sk_buff *skb);
443
struct sock *tcp_create_openreq_child(const struct sock *sk,
444 445
				      struct request_sock *req,
				      struct sk_buff *skb);
446
void tcp_ca_openreq_child(struct sock *sk, const struct dst_entry *dst);
447
struct sock *tcp_v4_syn_recv_sock(const struct sock *sk, struct sk_buff *skb,
448
				  struct request_sock *req,
449 450 451
				  struct dst_entry *dst,
				  struct request_sock *req_unhash,
				  bool *own_req);
452 453 454
int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb);
int tcp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len);
int tcp_connect(struct sock *sk);
455 456 457 458 459
enum tcp_synack_type {
	TCP_SYNACK_NORMAL,
	TCP_SYNACK_FASTOPEN,
	TCP_SYNACK_COOKIE,
};
460
struct sk_buff *tcp_make_synack(const struct sock *sk, struct dst_entry *dst,
461
				struct request_sock *req,
462
				struct tcp_fastopen_cookie *foc,
463 464
				enum tcp_synack_type synack_type,
				struct sk_buff *syn_skb);
465
int tcp_disconnect(struct sock *sk, int flags);
L
Linus Torvalds 已提交
466

P
Pavel Emelyanov 已提交
467
void tcp_finish_connect(struct sock *sk, struct sk_buff *skb);
468
int tcp_send_rcvq(struct sock *sk, struct msghdr *msg, size_t size);
469
void inet_sk_rx_dst_set(struct sock *sk, const struct sk_buff *skb);
L
Linus Torvalds 已提交
470 471

/* From syncookies.c */
472 473
struct sock *tcp_get_cookie_sock(struct sock *sk, struct sk_buff *skb,
				 struct request_sock *req,
474
				 struct dst_entry *dst, u32 tsoff);
475 476
int __cookie_v4_check(const struct iphdr *iph, const struct tcphdr *th,
		      u32 cookie);
C
Cong Wang 已提交
477
struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb);
478 479
struct request_sock *cookie_tcp_reqsk_alloc(const struct request_sock_ops *ops,
					    struct sock *sk, struct sk_buff *skb);
480
#ifdef CONFIG_SYN_COOKIES
481

482
/* Syncookies use a monotonic timer which increments every 60 seconds.
483 484 485
 * This counter is used both as a hash input and partially encoded into
 * the cookie value.  A cookie is only validated further if the delta
 * between the current counter value and the encoded one is less than this,
486
 * i.e. a sent cookie is valid only at most for 2*60 seconds (or less if
487 488
 * the counter advances immediately after a cookie is generated).
 */
489 490 491 492 493 494
#define MAX_SYNCOOKIE_AGE	2
#define TCP_SYNCOOKIE_PERIOD	(60 * HZ)
#define TCP_SYNCOOKIE_VALID	(MAX_SYNCOOKIE_AGE * TCP_SYNCOOKIE_PERIOD)

/* syncookies: remember time of last synqueue overflow
 * But do not dirty this field too often (once per second is enough)
495
 * It is racy as we do not hold a lock, but race is very minor.
496
 */
497
static inline void tcp_synq_overflow(const struct sock *sk)
498
{
499
	unsigned int last_overflow;
500
	unsigned int now = jiffies;
501

502 503 504 505 506 507
	if (sk->sk_reuseport) {
		struct sock_reuseport *reuse;

		reuse = rcu_dereference(sk->sk_reuseport_cb);
		if (likely(reuse)) {
			last_overflow = READ_ONCE(reuse->synq_overflow_ts);
508 509
			if (!time_between32(now, last_overflow,
					    last_overflow + HZ))
510 511 512 513 514
				WRITE_ONCE(reuse->synq_overflow_ts, now);
			return;
		}
	}

515
	last_overflow = READ_ONCE(tcp_sk(sk)->rx_opt.ts_recent_stamp);
516
	if (!time_between32(now, last_overflow, last_overflow + HZ))
517
		WRITE_ONCE(tcp_sk(sk)->rx_opt.ts_recent_stamp, now);
518 519 520 521 522
}

/* syncookies: no recent synqueue overflow on this listening socket? */
static inline bool tcp_synq_no_recent_overflow(const struct sock *sk)
{
523
	unsigned int last_overflow;
524
	unsigned int now = jiffies;
525

526 527 528 529 530 531
	if (sk->sk_reuseport) {
		struct sock_reuseport *reuse;

		reuse = rcu_dereference(sk->sk_reuseport_cb);
		if (likely(reuse)) {
			last_overflow = READ_ONCE(reuse->synq_overflow_ts);
532 533 534
			return !time_between32(now, last_overflow - HZ,
					       last_overflow +
					       TCP_SYNCOOKIE_VALID);
535 536 537
		}
	}

538
	last_overflow = READ_ONCE(tcp_sk(sk)->rx_opt.ts_recent_stamp);
539 540 541 542 543 544 545 546 547 548

	/* If last_overflow <= jiffies <= last_overflow + TCP_SYNCOOKIE_VALID,
	 * then we're under synflood. However, we have to use
	 * 'last_overflow - HZ' as lower bound. That's because a concurrent
	 * tcp_synq_overflow() could update .ts_recent_stamp after we read
	 * jiffies but before we store .ts_recent_stamp into last_overflow,
	 * which could lead to rejecting a valid syncookie.
	 */
	return !time_between32(now, last_overflow - HZ,
			       last_overflow + TCP_SYNCOOKIE_VALID);
549
}
550 551 552

static inline u32 tcp_cookie_time(void)
{
553 554
	u64 val = get_jiffies_64();

555
	do_div(val, TCP_SYNCOOKIE_PERIOD);
556
	return val;
557 558
}

559 560
u32 __cookie_v4_init_sequence(const struct iphdr *iph, const struct tcphdr *th,
			      u16 *mssp);
561
__u32 cookie_v4_init_sequence(const struct sk_buff *skb, __u16 *mss);
562
u64 cookie_init_timestamp(struct request_sock *req, u64 now);
E
Eric Dumazet 已提交
563 564
bool cookie_timestamp_decode(const struct net *net,
			     struct tcp_options_received *opt);
565
bool cookie_ecn_ok(const struct tcp_options_received *opt,
566
		   const struct net *net, const struct dst_entry *dst);
567

568
/* From net/ipv6/syncookies.c */
569 570 571
int __cookie_v6_check(const struct ipv6hdr *iph, const struct tcphdr *th,
		      u32 cookie);
struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
572

573 574
u32 __cookie_v6_init_sequence(const struct ipv6hdr *iph,
			      const struct tcphdr *th, u16 *mssp);
575
__u32 cookie_v6_init_sequence(const struct sk_buff *skb, __u16 *mss);
576
#endif
L
Linus Torvalds 已提交
577 578
/* tcp_output.c */

579 580
void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
			       int nonagle);
581 582
int __tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs);
int tcp_retransmit_skb(struct sock *sk, struct sk_buff *skb, int segs);
583 584 585
void tcp_retransmit_timer(struct sock *sk);
void tcp_xmit_retransmit_queue(struct sock *);
void tcp_simple_retransmit(struct sock *);
586
void tcp_enter_recovery(struct sock *sk, bool ece_ack);
587
int tcp_trim_head(struct sock *, struct sk_buff *, u32);
588 589 590 591 592 593 594
enum tcp_queue {
	TCP_FRAG_IN_WRITE_QUEUE,
	TCP_FRAG_IN_RTX_QUEUE,
};
int tcp_fragment(struct sock *sk, enum tcp_queue tcp_queue,
		 struct sk_buff *skb, u32 len,
		 unsigned int mss_now, gfp_t gfp);
595 596 597

void tcp_send_probe0(struct sock *);
void tcp_send_partial(struct sock *);
598
int tcp_write_wakeup(struct sock *, int mib);
599 600 601 602
void tcp_send_fin(struct sock *sk);
void tcp_send_active_reset(struct sock *sk, gfp_t priority);
int tcp_send_synack(struct sock *);
void tcp_push_one(struct sock *, unsigned int mss_now);
603
void __tcp_send_ack(struct sock *sk, u32 rcv_nxt);
604 605 606
void tcp_send_ack(struct sock *sk);
void tcp_send_delayed_ack(struct sock *sk);
void tcp_send_loss_probe(struct sock *sk);
607
bool tcp_schedule_loss_probe(struct sock *sk, bool advancing_rto);
608 609
void tcp_skb_collapse_tstamp(struct sk_buff *skb,
			     const struct sk_buff *next_skb);
L
Linus Torvalds 已提交
610

611
/* tcp_input.c */
612
void tcp_rearm_rto(struct sock *sk);
Y
Yuchung Cheng 已提交
613
void tcp_synack_rtt_meas(struct sock *sk, struct request_sock *req);
614
void tcp_reset(struct sock *sk, struct sk_buff *skb);
Y
Yuchung Cheng 已提交
615
void tcp_skb_mark_lost_uncond_verify(struct tcp_sock *tp, struct sk_buff *skb);
616
void tcp_fin(struct sock *sk);
617

L
Linus Torvalds 已提交
618
/* tcp_timer.c */
619
void tcp_init_xmit_timers(struct sock *);
620 621
static inline void tcp_clear_xmit_timers(struct sock *sk)
{
622
	if (hrtimer_try_to_cancel(&tcp_sk(sk)->pacing_timer) == 1)
623
		__sock_put(sk);
624

E
Eric Dumazet 已提交
625 626 627
	if (hrtimer_try_to_cancel(&tcp_sk(sk)->compressed_ack_timer) == 1)
		__sock_put(sk);

628 629
	inet_csk_clear_xmit_timers(sk);
}
L
Linus Torvalds 已提交
630

631 632
unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
unsigned int tcp_current_mss(struct sock *sk);
I
Ilpo Järvinen 已提交
633 634 635 636

/* Bound MSS / TSO packet size with the half of the window */
static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
{
637 638 639 640 641 642 643 644 645
	int cutoff;

	/* When peer uses tiny windows, there is no use in packetizing
	 * to sub-MSS pieces for the sake of SWS or making sure there
	 * are enough packets in the pipe for fast recovery.
	 *
	 * On the other hand, for extremely large MSS devices, handling
	 * smaller than MSS windows in this way does make sense.
	 */
646
	if (tp->max_window > TCP_MSS_DEFAULT)
647 648 649 650 651 652
		cutoff = (tp->max_window >> 1);
	else
		cutoff = tp->max_window;

	if (cutoff && pktsize > cutoff)
		return max_t(int, cutoff, 68U - tp->tcp_header_len);
I
Ilpo Järvinen 已提交
653 654 655
	else
		return pktsize;
}
L
Linus Torvalds 已提交
656

657
/* tcp.c */
658
void tcp_get_info(struct sock *, struct tcp_info *);
L
Linus Torvalds 已提交
659 660

/* Read 'sendfile()'-style from a TCP socket */
661 662
int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
		  sk_read_actor_t recv_actor);
L
Linus Torvalds 已提交
663

664
void tcp_initialize_rcv_mss(struct sock *sk);
L
Linus Torvalds 已提交
665

666 667 668
int tcp_mtu_to_mss(struct sock *sk, int pmtu);
int tcp_mss_to_mtu(struct sock *sk, int mss);
void tcp_mtup_init(struct sock *sk);
J
John Heffner 已提交
669

670 671 672 673 674 675 676 677
static inline void tcp_bound_rto(const struct sock *sk)
{
	if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
		inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
}

static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
{
678
	return usecs_to_jiffies((tp->srtt_us >> 3) + tp->rttvar_us);
679 680
}

681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703
static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
{
	tp->pred_flags = htonl((tp->tcp_header_len << 26) |
			       ntohl(TCP_FLAG_ACK) |
			       snd_wnd);
}

static inline void tcp_fast_path_on(struct tcp_sock *tp)
{
	__tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
}

static inline void tcp_fast_path_check(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);

	if (RB_EMPTY_ROOT(&tp->out_of_order_queue) &&
	    tp->rcv_wnd &&
	    atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
	    !tp->urg_data)
		tcp_fast_path_on(tp);
}

704 705 706
/* Compute the actual rto_min value */
static inline u32 tcp_rto_min(struct sock *sk)
{
707
	const struct dst_entry *dst = __sk_dst_get(sk);
708
	u32 rto_min = inet_csk(sk)->icsk_rto_min;
709 710 711 712 713 714

	if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
		rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
	return rto_min;
}

715 716 717 718 719
static inline u32 tcp_rto_min_us(struct sock *sk)
{
	return jiffies_to_usecs(tcp_rto_min(sk));
}

720 721 722 723 724
static inline bool tcp_ca_dst_locked(const struct dst_entry *dst)
{
	return dst_metric_locked(dst, RTAX_CC_ALGO);
}

725 726 727
/* Minimum RTT in usec. ~0 means not available. */
static inline u32 tcp_min_rtt(const struct tcp_sock *tp)
{
728
	return minmax_get(&tp->rtt_min);
729 730
}

L
Linus Torvalds 已提交
731 732 733 734
/* Compute the actual receive window we are currently advertising.
 * Rcv_nxt can be after the window if our peer push more data
 * than the offered window.
 */
S
Stephen Hemminger 已提交
735
static inline u32 tcp_receive_window(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
736 737 738 739 740 741 742 743 744 745 746 747
{
	s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;

	if (win < 0)
		win = 0;
	return (u32) win;
}

/* Choose a new window, without checks for shrinking, and without
 * scaling applied to the result.  The caller does these things
 * if necessary.  This is a "raw" window selection.
 */
748
u32 __tcp_select_window(struct sock *sk);
L
Linus Torvalds 已提交
749

P
Pavel Emelyanov 已提交
750 751
void tcp_send_window_probe(struct sock *sk);

E
Eric Dumazet 已提交
752 753 754 755 756 757
/* TCP uses 32bit jiffies to save some space.
 * Note that this is different from tcp_time_stamp, which
 * historically has been the same until linux-4.13.
 */
#define tcp_jiffies32 ((u32)jiffies)

758 759 760 761 762 763 764 765 766
/*
 * Deliver a 32bit value for TCP timestamp option (RFC 7323)
 * It is no longer tied to jiffies, but to 1 ms clock.
 * Note: double check if you want to use tcp_jiffies32 instead of this.
 */
#define TCP_TS_HZ	1000

static inline u64 tcp_clock_ns(void)
{
767
	return ktime_get_ns();
768 769 770 771 772 773 774 775 776 777 778 779 780
}

static inline u64 tcp_clock_us(void)
{
	return div_u64(tcp_clock_ns(), NSEC_PER_USEC);
}

/* This should only be used in contexts where tp->tcp_mstamp is up to date */
static inline u32 tcp_time_stamp(const struct tcp_sock *tp)
{
	return div_u64(tp->tcp_mstamp, USEC_PER_SEC / TCP_TS_HZ);
}

781 782 783 784 785 786
/* Convert a nsec timestamp into TCP TSval timestamp (ms based currently) */
static inline u32 tcp_ns_to_ts(u64 ns)
{
	return div_u64(ns, NSEC_PER_SEC / TCP_TS_HZ);
}

787 788 789
/* Could use tcp_clock_us() / 1000, but this version uses a single divide */
static inline u32 tcp_time_stamp_raw(void)
{
790
	return tcp_ns_to_ts(tcp_clock_ns());
791 792
}

793
void tcp_mstamp_refresh(struct tcp_sock *tp);
794 795 796 797 798

static inline u32 tcp_stamp_us_delta(u64 t1, u64 t0)
{
	return max_t(s64, t1 - t0, 0);
}
L
Linus Torvalds 已提交
799

E
Eric Dumazet 已提交
800 801
static inline u32 tcp_skb_timestamp(const struct sk_buff *skb)
{
802
	return tcp_ns_to_ts(skb->skb_mstamp_ns);
E
Eric Dumazet 已提交
803 804
}

805 806 807
/* provide the departure time in us unit */
static inline u64 tcp_skb_timestamp_us(const struct sk_buff *skb)
{
808
	return div_u64(skb->skb_mstamp_ns, NSEC_PER_USEC);
809 810
}

E
Eric Dumazet 已提交
811

C
Changli Gao 已提交
812 813 814 815 816 817 818 819 820 821 822
#define tcp_flag_byte(th) (((u_int8_t *)th)[13])

#define TCPHDR_FIN 0x01
#define TCPHDR_SYN 0x02
#define TCPHDR_RST 0x04
#define TCPHDR_PSH 0x08
#define TCPHDR_ACK 0x10
#define TCPHDR_URG 0x20
#define TCPHDR_ECE 0x40
#define TCPHDR_CWR 0x80

823 824
#define TCPHDR_SYN_ECN	(TCPHDR_SYN | TCPHDR_ECE | TCPHDR_CWR)

S
Stephen Hemminger 已提交
825
/* This is what the send packet queuing engine uses to pass
826 827 828 829
 * TCP per-packet control information to the transmission code.
 * We also store the host-order sequence numbers in here too.
 * This is 44 bytes if IPV6 is enabled.
 * If this grows please adjust skbuff.h:skbuff->cb[xxx] size appropriately.
L
Linus Torvalds 已提交
830 831 832 833
 */
struct tcp_skb_cb {
	__u32		seq;		/* Starting sequence number	*/
	__u32		end_seq;	/* SEQ + FIN + SYN + datalen	*/
834 835 836 837
	union {
		/* Note : tcp_tw_isn is used in input path only
		 *	  (isn chosen by tcp_timewait_state_process())
		 *
838 839
		 * 	  tcp_gso_segs/size are used in write queue only,
		 *	  cf tcp_skb_pcount()/tcp_skb_mss()
840 841
		 */
		__u32		tcp_tw_isn;
842 843 844 845
		struct {
			u16	tcp_gso_segs;
			u16	tcp_gso_size;
		};
846
	};
E
Eric Dumazet 已提交
847
	__u8		tcp_flags;	/* TCP header flags. (tcp[13])	*/
848

Y
Yuchung Cheng 已提交
849
	__u8		sacked;		/* State flags for SACK.	*/
L
Linus Torvalds 已提交
850 851 852 853
#define TCPCB_SACKED_ACKED	0x01	/* SKB ACK'd by a SACK block	*/
#define TCPCB_SACKED_RETRANS	0x02	/* SKB retransmitted		*/
#define TCPCB_LOST		0x04	/* SKB is lost			*/
#define TCPCB_TAGBITS		0x07	/* All tag bits			*/
854
#define TCPCB_REPAIRED		0x10	/* SKB repaired (no skb_mstamp_ns)	*/
L
Linus Torvalds 已提交
855
#define TCPCB_EVER_RETRANS	0x80	/* Ever retransmitted frame	*/
856 857
#define TCPCB_RETRANS		(TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS| \
				TCPCB_REPAIRED)
L
Linus Torvalds 已提交
858

859
	__u8		ip_dsfield;	/* IPv4 tos or IPv6 dsfield	*/
860
	__u8		txstamp_ack:1,	/* Record TX timestamp for ack? */
861
			eor:1,		/* Is skb MSG_EOR marked? */
862 863
			has_rxtstamp:1,	/* SKB has a RX timestamp	*/
			unused:5;
L
Linus Torvalds 已提交
864
	__u32		ack_seq;	/* Sequence number ACK'd	*/
865
	union {
866
		struct {
867
			/* There is space for up to 24 bytes */
868 869 870
			__u32 in_flight:30,/* Bytes in flight at transmit */
			      is_app_limited:1, /* cwnd not fully used? */
			      unused:1;
871 872 873
			/* pkts S/ACKed so far upon tx of skb, incl retrans: */
			__u32 delivered;
			/* start of send pipeline phase */
874
			u64 first_tx_mstamp;
875
			/* when we reached the "delivered" count */
876
			u64 delivered_mstamp;
877 878 879
		} tx;   /* only used for outgoing skbs */
		union {
			struct inet_skb_parm	h4;
880
#if IS_ENABLED(CONFIG_IPV6)
881
			struct inet6_skb_parm	h6;
882
#endif
883
		} header;	/* For incoming skbs */
884 885
		struct {
			__u32 flags;
886
			struct sock *sk_redir;
887
			void *data_end;
888
		} bpf;
889
	};
L
Linus Torvalds 已提交
890 891 892 893
};

#define TCP_SKB_CB(__skb)	((struct tcp_skb_cb *)&((__skb)->cb[0]))

894 895 896 897
static inline void bpf_compute_data_end_sk_skb(struct sk_buff *skb)
{
	TCP_SKB_CB(skb)->bpf.data_end = skb->data + skb_headlen(skb);
}
E
Eric Dumazet 已提交
898

899 900 901 902 903 904 905 906 907 908 909 910 911 912 913
static inline bool tcp_skb_bpf_ingress(const struct sk_buff *skb)
{
	return TCP_SKB_CB(skb)->bpf.flags & BPF_F_INGRESS;
}

static inline struct sock *tcp_skb_bpf_redirect_fetch(struct sk_buff *skb)
{
	return TCP_SKB_CB(skb)->bpf.sk_redir;
}

static inline void tcp_skb_bpf_redirect_clear(struct sk_buff *skb)
{
	TCP_SKB_CB(skb)->bpf.sk_redir = NULL;
}

914 915
extern const struct inet_connection_sock_af_ops ipv4_specific;

916
#if IS_ENABLED(CONFIG_IPV6)
E
Eric Dumazet 已提交
917 918 919 920
/* This is the variant of inet6_iif() that must be used by TCP,
 * as TCP moves IP6CB into a different location in skb->cb[]
 */
static inline int tcp_v6_iif(const struct sk_buff *skb)
921 922 923 924 925
{
	return TCP_SKB_CB(skb)->header.h6.iif;
}

static inline int tcp_v6_iif_l3_slave(const struct sk_buff *skb)
E
Eric Dumazet 已提交
926
{
927
	bool l3_slave = ipv6_l3mdev_skb(TCP_SKB_CB(skb)->header.h6.flags);
928 929

	return l3_slave ? skb->skb_iif : TCP_SKB_CB(skb)->header.h6.iif;
E
Eric Dumazet 已提交
930
}
931 932 933 934 935 936 937 938 939 940

/* TCP_SKB_CB reference means this can not be used from early demux */
static inline int tcp_v6_sdif(const struct sk_buff *skb)
{
#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
	if (skb && ipv6_l3mdev_skb(TCP_SKB_CB(skb)->header.h6.flags))
		return TCP_SKB_CB(skb)->header.h6.iif;
#endif
	return 0;
}
941

942 943
extern const struct inet_connection_sock_af_ops ipv6_specific;

944
INDIRECT_CALLABLE_DECLARE(void tcp_v6_send_check(struct sock *sk, struct sk_buff *skb));
945 946
INDIRECT_CALLABLE_DECLARE(int tcp_v6_rcv(struct sk_buff *skb));
INDIRECT_CALLABLE_DECLARE(void tcp_v6_early_demux(struct sk_buff *skb));
947

948
#endif
E
Eric Dumazet 已提交
949

950 951 952 953 954 955 956 957 958 959
/* TCP_SKB_CB reference means this can not be used from early demux */
static inline int tcp_v4_sdif(struct sk_buff *skb)
{
#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
	if (skb && ipv4_l3mdev_skb(TCP_SKB_CB(skb)->header.h4.flags))
		return TCP_SKB_CB(skb)->header.h4.iif;
#endif
	return 0;
}

L
Linus Torvalds 已提交
960 961
/* Due to TSO, an SKB can be composed of multiple actual
 * packets.  To keep these tracked properly, we use this.
962
 */
L
Linus Torvalds 已提交
963
static inline int tcp_skb_pcount(const struct sk_buff *skb)
964
{
965 966
	return TCP_SKB_CB(skb)->tcp_gso_segs;
}
967

968 969 970
static inline void tcp_skb_pcount_set(struct sk_buff *skb, int segs)
{
	TCP_SKB_CB(skb)->tcp_gso_segs = segs;
971 972
}

973
static inline void tcp_skb_pcount_add(struct sk_buff *skb, int segs)
L
Linus Torvalds 已提交
974
{
975
	TCP_SKB_CB(skb)->tcp_gso_segs += segs;
L
Linus Torvalds 已提交
976 977
}

978
/* This is valid iff skb is in write queue and tcp_skb_pcount() > 1. */
L
Linus Torvalds 已提交
979 980
static inline int tcp_skb_mss(const struct sk_buff *skb)
{
981
	return TCP_SKB_CB(skb)->tcp_gso_size;
L
Linus Torvalds 已提交
982 983
}

984 985 986 987 988
static inline bool tcp_skb_can_collapse_to(const struct sk_buff *skb)
{
	return likely(!TCP_SKB_CB(skb)->eor);
}

989 990 991 992 993 994 995
static inline bool tcp_skb_can_collapse(const struct sk_buff *to,
					const struct sk_buff *from)
{
	return likely(tcp_skb_can_collapse_to(to) &&
		      mptcp_skb_can_collapse(to, from));
}

996 997 998 999 1000 1001
/* Events passed to congestion control interface */
enum tcp_ca_event {
	CA_EVENT_TX_START,	/* first transmit when no packets in flight */
	CA_EVENT_CWND_RESTART,	/* congestion window restart */
	CA_EVENT_COMPLETE_CWR,	/* end of congestion recovery */
	CA_EVENT_LOSS,		/* loss timeout */
1002 1003
	CA_EVENT_ECN_NO_CE,	/* ECT set, but not CE marked */
	CA_EVENT_ECN_IS_CE,	/* received CE marked IP packet */
1004 1005
};

1006
/* Information about inbound ACK, passed to cong_ops->in_ack_event() */
1007
enum tcp_ca_ack_event_flags {
1008 1009 1010
	CA_ACK_SLOWPATH		= (1 << 0),	/* In slow path processing */
	CA_ACK_WIN_UPDATE	= (1 << 1),	/* ACK updated window */
	CA_ACK_ECE		= (1 << 2),	/* ECE bit is set on ack */
1011 1012 1013 1014 1015 1016
};

/*
 * Interface for adding new TCP congestion control handlers
 */
#define TCP_CA_NAME_MAX	16
1017 1018 1019
#define TCP_CA_MAX	128
#define TCP_CA_BUF_MAX	(TCP_CA_NAME_MAX*TCP_CA_MAX)

1020 1021
#define TCP_CA_UNSPEC	0

1022
/* Algorithm can be set on socket without CAP_NET_ADMIN privileges */
1023
#define TCP_CONG_NON_RESTRICTED 0x1
1024 1025
/* Requires ECN/ECT set on all packets */
#define TCP_CONG_NEEDS_ECN	0x2
1026
#define TCP_CONG_MASK	(TCP_CONG_NON_RESTRICTED | TCP_CONG_NEEDS_ECN)
1027

1028 1029
union tcp_cc_info;

1030 1031 1032
struct ack_sample {
	u32 pkts_acked;
	s32 rtt_us;
1033
	u32 in_flight;
1034 1035
};

1036 1037 1038 1039 1040 1041 1042 1043 1044
/* A rate sample measures the number of (original/retransmitted) data
 * packets delivered "delivered" over an interval of time "interval_us".
 * The tcp_rate.c code fills in the rate sample, and congestion
 * control modules that define a cong_control function to run at the end
 * of ACK processing can optionally chose to consult this sample when
 * setting cwnd and pacing rate.
 * A sample is invalid if "delivered" or "interval_us" is negative.
 */
struct rate_sample {
1045
	u64  prior_mstamp; /* starting timestamp for interval */
1046 1047 1048
	u32  prior_delivered;	/* tp->delivered at "prior_mstamp" */
	s32  delivered;		/* number of packets delivered over interval */
	long interval_us;	/* time for tp->delivered to incr "delivered" */
1049 1050
	u32 snd_interval_us;	/* snd interval for delivered packets */
	u32 rcv_interval_us;	/* rcv interval for delivered packets */
1051 1052 1053 1054
	long rtt_us;		/* RTT of last (S)ACKed packet (or -1) */
	int  losses;		/* number of packets marked lost upon ACK */
	u32  acked_sacked;	/* number of packets newly (S)ACKed upon ACK */
	u32  prior_in_flight;	/* in flight before this ACK */
1055
	bool is_app_limited;	/* is sample from packet with bubble in pipe? */
1056
	bool is_retrans;	/* is sample from retransmission? */
1057
	bool is_ack_delayed;	/* is this (likely) a delayed ACK? */
1058 1059
};

1060 1061
struct tcp_congestion_ops {
	struct list_head	list;
1062 1063
	u32 key;
	u32 flags;
1064 1065

	/* initialize private data (optional) */
1066
	void (*init)(struct sock *sk);
1067
	/* cleanup private data  (optional) */
1068
	void (*release)(struct sock *sk);
1069 1070

	/* return slow start threshold (required) */
1071
	u32 (*ssthresh)(struct sock *sk);
1072
	/* do new cwnd calculation (required) */
1073
	void (*cong_avoid)(struct sock *sk, u32 ack, u32 acked);
1074
	/* call before changing ca_state (optional) */
1075
	void (*set_state)(struct sock *sk, u8 new_state);
1076
	/* call when cwnd event occurs (optional) */
1077
	void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
1078 1079
	/* call when ack arrives (optional) */
	void (*in_ack_event)(struct sock *sk, u32 flags);
1080
	/* new value of cwnd after loss (required) */
1081
	u32  (*undo_cwnd)(struct sock *sk);
1082
	/* hook for packet ack accounting (optional) */
1083
	void (*pkts_acked)(struct sock *sk, const struct ack_sample *sample);
1084 1085
	/* override sysctl_tcp_min_tso_segs */
	u32 (*min_tso_segs)(struct sock *sk);
1086 1087
	/* returns the multiplier used in tcp_sndbuf_expand (optional) */
	u32 (*sndbuf_expand)(struct sock *sk);
1088 1089 1090 1091
	/* call when packets are delivered to update cwnd and pacing rate,
	 * after all the ca_state processing. (optional)
	 */
	void (*cong_control)(struct sock *sk, const struct rate_sample *rs);
1092
	/* get info for inet_diag (optional) */
1093 1094
	size_t (*get_info)(struct sock *sk, u32 ext, int *attr,
			   union tcp_cc_info *info);
1095 1096 1097 1098 1099

	char 		name[TCP_CA_NAME_MAX];
	struct module 	*owner;
};

1100 1101
int tcp_register_congestion_control(struct tcp_congestion_ops *type);
void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
1102

1103
void tcp_assign_congestion_control(struct sock *sk);
1104 1105
void tcp_init_congestion_control(struct sock *sk);
void tcp_cleanup_congestion_control(struct sock *sk);
1106 1107
int tcp_set_default_congestion_control(struct net *net, const char *name);
void tcp_get_default_congestion_control(struct net *net, char *name);
1108 1109 1110
void tcp_get_available_congestion_control(char *buf, size_t len);
void tcp_get_allowed_congestion_control(char *buf, size_t len);
int tcp_set_allowed_congestion_control(char *allowed);
1111
int tcp_set_congestion_control(struct sock *sk, const char *name, bool load,
1112
			       bool cap_net_admin);
N
Neal Cardwell 已提交
1113 1114
u32 tcp_slow_start(struct tcp_sock *tp, u32 acked);
void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w, u32 acked);
1115

1116
u32 tcp_reno_ssthresh(struct sock *sk);
1117
u32 tcp_reno_undo_cwnd(struct sock *sk);
1118
void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 acked);
1119
extern struct tcp_congestion_ops tcp_reno;
1120

1121
struct tcp_congestion_ops *tcp_ca_find(const char *name);
1122
struct tcp_congestion_ops *tcp_ca_find_key(u32 key);
1123
u32 tcp_ca_get_key_by_name(struct net *net, const char *name, bool *ecn_ca);
1124
#ifdef CONFIG_INET
1125
char *tcp_ca_get_name_by_key(u32 key, char *buffer);
1126 1127 1128 1129 1130 1131
#else
static inline char *tcp_ca_get_name_by_key(u32 key, char *buffer)
{
	return NULL;
}
#endif
1132

1133 1134 1135 1136 1137 1138 1139
static inline bool tcp_ca_needs_ecn(const struct sock *sk)
{
	const struct inet_connection_sock *icsk = inet_csk(sk);

	return icsk->icsk_ca_ops->flags & TCP_CONG_NEEDS_ECN;
}

1140
static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
1141
{
1142 1143 1144 1145 1146
	struct inet_connection_sock *icsk = inet_csk(sk);

	if (icsk->icsk_ca_ops->set_state)
		icsk->icsk_ca_ops->set_state(sk, ca_state);
	icsk->icsk_ca_state = ca_state;
1147 1148
}

1149
static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
1150
{
1151 1152 1153 1154
	const struct inet_connection_sock *icsk = inet_csk(sk);

	if (icsk->icsk_ca_ops->cwnd_event)
		icsk->icsk_ca_ops->cwnd_event(sk, event);
1155 1156
}

1157 1158 1159 1160 1161
/* From tcp_rate.c */
void tcp_rate_skb_sent(struct sock *sk, struct sk_buff *skb);
void tcp_rate_skb_delivered(struct sock *sk, struct sk_buff *skb,
			    struct rate_sample *rs);
void tcp_rate_gen(struct sock *sk, u32 delivered, u32 lost,
1162
		  bool is_sack_reneg, struct rate_sample *rs);
1163
void tcp_rate_check_app_limited(struct sock *sk);
1164

1165 1166 1167 1168 1169 1170 1171 1172 1173
/* These functions determine how the current flow behaves in respect of SACK
 * handling. SACK is negotiated with the peer, and therefore it can vary
 * between different flows.
 *
 * tcp_is_sack - SACK enabled
 * tcp_is_reno - No SACK
 */
static inline int tcp_is_sack(const struct tcp_sock *tp)
{
E
Eric Dumazet 已提交
1174
	return likely(tp->rx_opt.sack_ok);
1175 1176
}

E
Eric Dumazet 已提交
1177
static inline bool tcp_is_reno(const struct tcp_sock *tp)
1178 1179 1180 1181
{
	return !tcp_is_sack(tp);
}

1182 1183 1184 1185 1186
static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
{
	return tp->sacked_out + tp->lost_out;
}

L
Linus Torvalds 已提交
1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200
/* This determines how many packets are "in the network" to the best
 * of our knowledge.  In many cases it is conservative, but where
 * detailed information is available from the receiver (via SACK
 * blocks etc.) we can make more aggressive calculations.
 *
 * Use this for decisions involving congestion control, use just
 * tp->packets_out to determine if the send queue is empty or not.
 *
 * Read this equation as:
 *
 *	"Packets sent once on transmission queue" MINUS
 *	"Packets left network, but not honestly ACKed yet" PLUS
 *	"Packets fast retransmitted"
 */
S
Stephen Hemminger 已提交
1201
static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
L
Linus Torvalds 已提交
1202
{
1203
	return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
L
Linus Torvalds 已提交
1204 1205
}

I
Ilpo Järvinen 已提交
1206 1207
#define TCP_INFINITE_SSTHRESH	0x7fffffff

1208 1209
static inline bool tcp_in_slow_start(const struct tcp_sock *tp)
{
1210
	return tp->snd_cwnd < tp->snd_ssthresh;
1211 1212
}

I
Ilpo Järvinen 已提交
1213 1214 1215 1216 1217
static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
{
	return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
}

1218 1219 1220 1221 1222 1223
static inline bool tcp_in_cwnd_reduction(const struct sock *sk)
{
	return (TCPF_CA_CWR | TCPF_CA_Recovery) &
	       (1 << inet_csk(sk)->icsk_ca_state);
}

L
Linus Torvalds 已提交
1224
/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
1225
 * The exception is cwnd reduction phase, when cwnd is decreasing towards
L
Linus Torvalds 已提交
1226 1227
 * ssthresh.
 */
1228
static inline __u32 tcp_current_ssthresh(const struct sock *sk)
L
Linus Torvalds 已提交
1229
{
1230
	const struct tcp_sock *tp = tcp_sk(sk);
1231

1232
	if (tcp_in_cwnd_reduction(sk))
L
Linus Torvalds 已提交
1233 1234 1235 1236 1237 1238 1239
		return tp->snd_ssthresh;
	else
		return max(tp->snd_ssthresh,
			   ((tp->snd_cwnd >> 1) +
			    (tp->snd_cwnd >> 2)));
}

1240 1241
/* Use define here intentionally to get WARN_ON location shown at the caller */
#define tcp_verify_left_out(tp)	WARN_ON(tcp_left_out(tp) > tp->packets_out)
L
Linus Torvalds 已提交
1242

1243
void tcp_enter_cwr(struct sock *sk);
1244
__u32 tcp_init_cwnd(const struct tcp_sock *tp, const struct dst_entry *dst);
L
Linus Torvalds 已提交
1245

1246 1247 1248 1249 1250 1251 1252 1253
/* The maximum number of MSS of available cwnd for which TSO defers
 * sending if not using sysctl_tcp_tso_win_divisor.
 */
static inline __u32 tcp_max_tso_deferred_mss(const struct tcp_sock *tp)
{
	return 3;
}

1254 1255 1256 1257 1258
/* Returns end sequence number of the receiver's advertised window */
static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
{
	return tp->snd_una + tp->snd_wnd;
}
1259 1260 1261

/* We follow the spirit of RFC2861 to validate cwnd but implement a more
 * flexible approach. The RFC suggests cwnd should not be raised unless
1262 1263 1264
 * it was fully used previously. And that's exactly what we do in
 * congestion avoidance mode. But in slow start we allow cwnd to grow
 * as long as the application has used half the cwnd.
1265 1266 1267 1268 1269 1270 1271 1272
 * Example :
 *    cwnd is 10 (IW10), but application sends 9 frames.
 *    We allow cwnd to reach 18 when all frames are ACKed.
 * This check is safe because it's as aggressive as slow start which already
 * risks 100% overshoot. The advantage is that we discourage application to
 * either send more filler packets or data to artificially blow up the cwnd
 * usage, and allow application-limited process to probe bw more aggressively.
 */
1273
static inline bool tcp_is_cwnd_limited(const struct sock *sk)
1274 1275 1276
{
	const struct tcp_sock *tp = tcp_sk(sk);

1277
	/* If in slow start, ensure cwnd grows to twice what was ACKed. */
1278
	if (tcp_in_slow_start(tp))
1279 1280 1281
		return tp->snd_cwnd < 2 * tp->max_packets_out;

	return tp->is_cwnd_limited;
1282
}
1283

1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294
/* BBR congestion control needs pacing.
 * Same remark for SO_MAX_PACING_RATE.
 * sch_fq packet scheduler is efficiently handling pacing,
 * but is not always installed/used.
 * Return true if TCP stack should pace packets itself.
 */
static inline bool tcp_needs_internal_pacing(const struct sock *sk)
{
	return smp_load_acquire(&sk->sk_pacing_status) == SK_PACING_NEEDED;
}

1295 1296
/* Estimates in how many jiffies next packet for this flow can be sent.
 * Scheduling a retransmit timer too early would be silly.
1297
 */
1298
static inline unsigned long tcp_pacing_delay(const struct sock *sk)
1299
{
1300
	s64 delay = tcp_sk(sk)->tcp_wstamp_ns - tcp_sk(sk)->tcp_clock_cache;
1301

1302
	return delay > 0 ? nsecs_to_jiffies(delay) : 0;
1303 1304 1305 1306 1307
}

static inline void tcp_reset_xmit_timer(struct sock *sk,
					const int what,
					unsigned long when,
1308
					const unsigned long max_when)
1309
{
1310
	inet_csk_reset_xmit_timer(sk, what, when + tcp_pacing_delay(sk),
1311 1312 1313
				  max_when);
}

1314
/* Something is really bad, we could not queue an additional packet,
1315
 * because qdisc is full or receiver sent a 0 window, or we are paced.
1316 1317 1318 1319 1320
 * We do not want to add fuel to the fire, or abort too early,
 * so make sure the timer we arm now is at least 200ms in the future,
 * regardless of current icsk_rto value (as it could be ~2ms)
 */
static inline unsigned long tcp_probe0_base(const struct sock *sk)
L
Linus Torvalds 已提交
1321
{
1322 1323
	return max_t(unsigned long, inet_csk(sk)->icsk_rto, TCP_RTO_MIN);
}
1324

1325 1326 1327 1328
/* Variant of inet_csk_rto_backoff() used for zero window probes */
static inline unsigned long tcp_probe0_when(const struct sock *sk,
					    unsigned long max_when)
{
1329 1330 1331
	u8 backoff = min_t(u8, ilog2(TCP_RTO_MAX / TCP_RTO_MIN) + 1,
			   inet_csk(sk)->icsk_backoff);
	u64 when = (u64)tcp_probe0_base(sk) << backoff;
1332 1333 1334 1335 1336 1337 1338

	return (unsigned long)min_t(u64, when, max_when);
}

static inline void tcp_check_probe_timer(struct sock *sk)
{
	if (!tcp_sk(sk)->packets_out && !inet_csk(sk)->icsk_pending)
1339
		tcp_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
1340
				     tcp_probe0_base(sk), TCP_RTO_MAX);
L
Linus Torvalds 已提交
1341 1342
}

1343
static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
L
Linus Torvalds 已提交
1344 1345 1346 1347
{
	tp->snd_wl1 = seq;
}

1348
static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
L
Linus Torvalds 已提交
1349 1350 1351 1352 1353 1354 1355
{
	tp->snd_wl1 = seq;
}

/*
 * Calculate(/check) TCP checksum
 */
1356 1357
static inline __sum16 tcp_v4_check(int len, __be32 saddr,
				   __be32 daddr, __wsum base)
L
Linus Torvalds 已提交
1358
{
1359
	return csum_tcpudp_magic(saddr, daddr, len, IPPROTO_TCP, base);
L
Linus Torvalds 已提交
1360 1361
}

E
Eric Dumazet 已提交
1362
static inline bool tcp_checksum_complete(struct sk_buff *skb)
L
Linus Torvalds 已提交
1363
{
1364
	return !skb_csum_unnecessary(skb) &&
1365
		__skb_checksum_complete(skb);
L
Linus Torvalds 已提交
1366 1367
}

E
Eric Dumazet 已提交
1368
bool tcp_add_backlog(struct sock *sk, struct sk_buff *skb);
1369
int tcp_filter(struct sock *sk, struct sk_buff *skb);
1370 1371
void tcp_set_state(struct sock *sk, int state);
void tcp_done(struct sock *sk);
1372 1373
int tcp_abort(struct sock *sk, int err);

S
Stephen Hemminger 已提交
1374
static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
L
Linus Torvalds 已提交
1375 1376 1377 1378 1379
{
	rx_opt->dsack = 0;
	rx_opt->num_sacks = 0;
}

1380 1381 1382 1383
void tcp_cwnd_restart(struct sock *sk, s32 delta);

static inline void tcp_slow_start_after_idle_check(struct sock *sk)
{
1384
	const struct tcp_congestion_ops *ca_ops = inet_csk(sk)->icsk_ca_ops;
1385 1386 1387
	struct tcp_sock *tp = tcp_sk(sk);
	s32 delta;

1388
	if (!sock_net(sk)->ipv4.sysctl_tcp_slow_start_after_idle || tp->packets_out ||
1389
	    ca_ops->cong_control)
1390
		return;
1391
	delta = tcp_jiffies32 - tp->lsndtime;
1392 1393 1394
	if (delta > inet_csk(sk)->icsk_rto)
		tcp_cwnd_restart(sk, delta);
}
1395

L
Linus Torvalds 已提交
1396
/* Determine a window scaling and initial window to offer. */
1397 1398
void tcp_select_initial_window(const struct sock *sk, int __space,
			       __u32 mss, __u32 *rcv_wnd,
1399 1400
			       __u32 *window_clamp, int wscale_ok,
			       __u8 *rcv_wscale, __u32 init_rcv_wnd);
L
Linus Torvalds 已提交
1401

1402
static inline int tcp_win_from_space(const struct sock *sk, int space)
L
Linus Torvalds 已提交
1403
{
1404
	int tcp_adv_win_scale = sock_net(sk)->ipv4.sysctl_tcp_adv_win_scale;
1405 1406 1407 1408

	return tcp_adv_win_scale <= 0 ?
		(space>>(-tcp_adv_win_scale)) :
		space - (space>>tcp_adv_win_scale);
L
Linus Torvalds 已提交
1409 1410
}

1411
/* Note: caller must be prepared to deal with negative returns */
L
Linus Torvalds 已提交
1412 1413
static inline int tcp_space(const struct sock *sk)
{
1414
	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf) -
1415
				  READ_ONCE(sk->sk_backlog.len) -
L
Linus Torvalds 已提交
1416
				  atomic_read(&sk->sk_rmem_alloc));
1417
}
L
Linus Torvalds 已提交
1418 1419 1420

static inline int tcp_full_space(const struct sock *sk)
{
1421
	return tcp_win_from_space(sk, READ_ONCE(sk->sk_rcvbuf));
L
Linus Torvalds 已提交
1422 1423
}

1424 1425
void tcp_cleanup_rbuf(struct sock *sk, int copied);

1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
/* We provision sk_rcvbuf around 200% of sk_rcvlowat.
 * If 87.5 % (7/8) of the space has been consumed, we want to override
 * SO_RCVLOWAT constraint, since we are receiving skbs with too small
 * len/truesize ratio.
 */
static inline bool tcp_rmem_pressure(const struct sock *sk)
{
	int rcvbuf = READ_ONCE(sk->sk_rcvbuf);
	int threshold = rcvbuf - (rcvbuf >> 3);

	return atomic_read(&sk->sk_rmem_alloc) > threshold;
}

1439
extern void tcp_openreq_init_rwin(struct request_sock *req,
1440 1441
				  const struct sock *sk_listener,
				  const struct dst_entry *dst);
1442

1443
void tcp_enter_memory_pressure(struct sock *sk);
1444
void tcp_leave_memory_pressure(struct sock *sk);
L
Linus Torvalds 已提交
1445 1446 1447

static inline int keepalive_intvl_when(const struct tcp_sock *tp)
{
1448 1449 1450
	struct net *net = sock_net((struct sock *)tp);

	return tp->keepalive_intvl ? : net->ipv4.sysctl_tcp_keepalive_intvl;
L
Linus Torvalds 已提交
1451 1452 1453 1454
}

static inline int keepalive_time_when(const struct tcp_sock *tp)
{
1455 1456 1457
	struct net *net = sock_net((struct sock *)tp);

	return tp->keepalive_time ? : net->ipv4.sysctl_tcp_keepalive_time;
L
Linus Torvalds 已提交
1458 1459
}

E
Eric Dumazet 已提交
1460 1461
static inline int keepalive_probes(const struct tcp_sock *tp)
{
1462 1463 1464
	struct net *net = sock_net((struct sock *)tp);

	return tp->keepalive_probes ? : net->ipv4.sysctl_tcp_keepalive_probes;
E
Eric Dumazet 已提交
1465 1466
}

1467 1468 1469 1470
static inline u32 keepalive_time_elapsed(const struct tcp_sock *tp)
{
	const struct inet_connection_sock *icsk = &tp->inet_conn;

1471 1472
	return min_t(u32, tcp_jiffies32 - icsk->icsk_ack.lrcvtime,
			  tcp_jiffies32 - tp->rcv_tstamp);
1473 1474
}

1475
static inline int tcp_fin_time(const struct sock *sk)
L
Linus Torvalds 已提交
1476
{
1477
	int fin_timeout = tcp_sk(sk)->linger2 ? : sock_net(sk)->ipv4.sysctl_tcp_fin_timeout;
1478
	const int rto = inet_csk(sk)->icsk_rto;
L
Linus Torvalds 已提交
1479

1480 1481
	if (fin_timeout < (rto << 2) - (rto >> 1))
		fin_timeout = (rto << 2) - (rto >> 1);
L
Linus Torvalds 已提交
1482 1483 1484 1485

	return fin_timeout;
}

E
Eric Dumazet 已提交
1486 1487
static inline bool tcp_paws_check(const struct tcp_options_received *rx_opt,
				  int paws_win)
L
Linus Torvalds 已提交
1488
{
I
Ilpo Järvinen 已提交
1489
	if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
E
Eric Dumazet 已提交
1490
		return true;
1491 1492
	if (unlikely(!time_before32(ktime_get_seconds(),
				    rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)))
E
Eric Dumazet 已提交
1493
		return true;
E
Eric Dumazet 已提交
1494 1495 1496 1497 1498 1499
	/*
	 * Some OSes send SYN and SYNACK messages with tsval=0 tsecr=0,
	 * then following tcp messages have valid values. Ignore 0 value,
	 * or else 'negative' tsval might forbid us to accept their packets.
	 */
	if (!rx_opt->ts_recent)
E
Eric Dumazet 已提交
1500 1501
		return true;
	return false;
I
Ilpo Järvinen 已提交
1502 1503
}

E
Eric Dumazet 已提交
1504 1505
static inline bool tcp_paws_reject(const struct tcp_options_received *rx_opt,
				   int rst)
I
Ilpo Järvinen 已提交
1506 1507
{
	if (tcp_paws_check(rx_opt, 0))
E
Eric Dumazet 已提交
1508
		return false;
L
Linus Torvalds 已提交
1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521

	/* RST segments are not recommended to carry timestamp,
	   and, if they do, it is recommended to ignore PAWS because
	   "their cleanup function should take precedence over timestamps."
	   Certainly, it is mistake. It is necessary to understand the reasons
	   of this constraint to relax it: if peer reboots, clock may go
	   out-of-sync and half-open connections will not be reset.
	   Actually, the problem would be not existing if all
	   the implementations followed draft about maintaining clock
	   via reboots. Linux-2.2 DOES NOT!

	   However, we can relax time bounds for RST segments to MSL.
	 */
1522 1523
	if (rst && !time_before32(ktime_get_seconds(),
				  rx_opt->ts_recent_stamp + TCP_PAWS_MSL))
E
Eric Dumazet 已提交
1524 1525
		return false;
	return true;
L
Linus Torvalds 已提交
1526 1527
}

1528 1529
bool tcp_oow_rate_limited(struct net *net, const struct sk_buff *skb,
			  int mib_idx, u32 *last_oow_ack_time);
1530

P
Pavel Emelyanov 已提交
1531
static inline void tcp_mib_init(struct net *net)
L
Linus Torvalds 已提交
1532 1533
{
	/* See RFC 2012 */
1534 1535 1536 1537
	TCP_ADD_STATS(net, TCP_MIB_RTOALGORITHM, 1);
	TCP_ADD_STATS(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
	TCP_ADD_STATS(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
	TCP_ADD_STATS(net, TCP_MIB_MAXCONN, -1);
L
Linus Torvalds 已提交
1538 1539
}

1540
/* from STCP */
1541
static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
D
David S. Miller 已提交
1542
{
1543
	tp->lost_skb_hint = NULL;
1544 1545 1546 1547 1548
}

static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
{
	tcp_clear_retrans_hints_partial(tp);
1549
	tp->retransmit_skb_hint = NULL;
1550 1551
}

E
Eric Dumazet 已提交
1552 1553 1554 1555 1556 1557 1558
union tcp_md5_addr {
	struct in_addr  a4;
#if IS_ENABLED(CONFIG_IPV6)
	struct in6_addr	a6;
#endif
};

1559 1560
/* - key database */
struct tcp_md5sig_key {
E
Eric Dumazet 已提交
1561
	struct hlist_node	node;
1562
	u8			keylen;
E
Eric Dumazet 已提交
1563
	u8			family; /* AF_INET or AF_INET6 */
1564
	u8			prefixlen;
1565 1566
	union tcp_md5_addr	addr;
	int			l3index; /* set if key added with L3 scope */
E
Eric Dumazet 已提交
1567 1568
	u8			key[TCP_MD5SIG_MAXKEYLEN];
	struct rcu_head		rcu;
1569 1570 1571 1572
};

/* - sock block */
struct tcp_md5sig_info {
E
Eric Dumazet 已提交
1573
	struct hlist_head	head;
1574
	struct rcu_head		rcu;
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590 1591 1592 1593 1594
};

/* - pseudo header */
struct tcp4_pseudohdr {
	__be32		saddr;
	__be32		daddr;
	__u8		pad;
	__u8		protocol;
	__be16		len;
};

struct tcp6_pseudohdr {
	struct in6_addr	saddr;
	struct in6_addr daddr;
	__be32		len;
	__be32		protocol;	/* including padding */
};

union tcp_md5sum_block {
	struct tcp4_pseudohdr ip4;
E
Eric Dumazet 已提交
1595
#if IS_ENABLED(CONFIG_IPV6)
1596 1597 1598 1599 1600 1601
	struct tcp6_pseudohdr ip6;
#endif
};

/* - pool: digest algorithm, hash description and scratch buffer */
struct tcp_md5sig_pool {
H
Herbert Xu 已提交
1602
	struct ahash_request	*md5_req;
1603
	void			*scratch;
1604 1605 1606
};

/* - functions */
1607 1608
int tcp_v4_md5_hash_skb(char *md5_hash, const struct tcp_md5sig_key *key,
			const struct sock *sk, const struct sk_buff *skb);
1609
int tcp_md5_do_add(struct sock *sk, const union tcp_md5_addr *addr,
1610 1611
		   int family, u8 prefixlen, int l3index,
		   const u8 *newkey, u8 newkeylen, gfp_t gfp);
1612
int tcp_md5_do_del(struct sock *sk, const union tcp_md5_addr *addr,
1613
		   int family, u8 prefixlen, int l3index);
1614
struct tcp_md5sig_key *tcp_v4_md5_lookup(const struct sock *sk,
1615
					 const struct sock *addr_sk);
1616

1617
#ifdef CONFIG_TCP_MD5SIG
1618
#include <linux/jump_label.h>
1619
extern struct static_key_false tcp_md5_needed;
1620
struct tcp_md5sig_key *__tcp_md5_do_lookup(const struct sock *sk, int l3index,
1621 1622 1623
					   const union tcp_md5_addr *addr,
					   int family);
static inline struct tcp_md5sig_key *
1624 1625
tcp_md5_do_lookup(const struct sock *sk, int l3index,
		  const union tcp_md5_addr *addr, int family)
1626
{
1627
	if (!static_branch_unlikely(&tcp_md5_needed))
1628
		return NULL;
1629
	return __tcp_md5_do_lookup(sk, l3index, addr, family);
1630 1631
}

E
Eric Dumazet 已提交
1632
#define tcp_twsk_md5_key(twsk)	((twsk)->tw_md5_key)
1633
#else
1634 1635 1636
static inline struct tcp_md5sig_key *
tcp_md5_do_lookup(const struct sock *sk, int l3index,
		  const union tcp_md5_addr *addr, int family)
E
Eric Dumazet 已提交
1637 1638 1639
{
	return NULL;
}
1640 1641 1642
#define tcp_twsk_md5_key(twsk)	NULL
#endif

1643
bool tcp_alloc_md5sig_pool(void);
1644

1645
struct tcp_md5sig_pool *tcp_get_md5sig_pool(void);
1646 1647 1648 1649
static inline void tcp_put_md5sig_pool(void)
{
	local_bh_enable();
}
E
Eric Dumazet 已提交
1650

1651 1652 1653 1654
int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, const struct sk_buff *,
			  unsigned int header_len);
int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
		     const struct tcp_md5sig_key *key);
1655

1656
/* From tcp_fastopen.c */
1657
void tcp_fastopen_cache_get(struct sock *sk, u16 *mss,
1658
			    struct tcp_fastopen_cookie *cookie);
1659
void tcp_fastopen_cache_set(struct sock *sk, u16 mss,
1660 1661
			    struct tcp_fastopen_cookie *cookie, bool syn_lost,
			    u16 try_exp);
1662 1663 1664 1665
struct tcp_fastopen_request {
	/* Fast Open cookie. Size 0 means a cookie request */
	struct tcp_fastopen_cookie	cookie;
	struct msghdr			*data;  /* data in MSG_FASTOPEN */
1666 1667
	size_t				size;
	int				copied;	/* queued in tcp_connect() */
1668
	struct ubuf_info		*uarg;
1669 1670
};
void tcp_free_fastopen_req(struct tcp_sock *tp);
1671
void tcp_fastopen_destroy_cipher(struct sock *sk);
1672
void tcp_fastopen_ctx_destroy(struct net *net);
1673
int tcp_fastopen_reset_cipher(struct net *net, struct sock *sk,
1674
			      void *primary_key, void *backup_key);
1675 1676
int tcp_fastopen_get_cipher(struct net *net, struct inet_connection_sock *icsk,
			    u64 *key);
1677
void tcp_fastopen_add_skb(struct sock *sk, struct sk_buff *skb);
1678 1679
struct sock *tcp_try_fastopen(struct sock *sk, struct sk_buff *skb,
			      struct request_sock *req,
1680 1681
			      struct tcp_fastopen_cookie *foc,
			      const struct dst_entry *dst);
1682
void tcp_fastopen_init_key_once(struct net *net);
1683 1684
bool tcp_fastopen_cookie_check(struct sock *sk, u16 *mss,
			     struct tcp_fastopen_cookie *cookie);
W
Wei Wang 已提交
1685
bool tcp_fastopen_defer_connect(struct sock *sk, int *err);
1686
#define TCP_FASTOPEN_KEY_LENGTH sizeof(siphash_key_t)
1687 1688 1689
#define TCP_FASTOPEN_KEY_MAX 2
#define TCP_FASTOPEN_KEY_BUF_LENGTH \
	(TCP_FASTOPEN_KEY_LENGTH * TCP_FASTOPEN_KEY_MAX)
1690 1691 1692

/* Fastopen key context */
struct tcp_fastopen_context {
1693
	siphash_key_t	key[TCP_FASTOPEN_KEY_MAX];
1694 1695
	int		num;
	struct rcu_head	rcu;
1696 1697
};

1698
extern unsigned int sysctl_tcp_fastopen_blackhole_timeout;
1699
void tcp_fastopen_active_disable(struct sock *sk);
1700 1701
bool tcp_fastopen_active_should_disable(struct sock *sk);
void tcp_fastopen_active_disable_ofo_check(struct sock *sk);
1702
void tcp_fastopen_active_detect_blackhole(struct sock *sk, bool expired);
1703

1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
/* Caller needs to wrap with rcu_read_(un)lock() */
static inline
struct tcp_fastopen_context *tcp_fastopen_get_ctx(const struct sock *sk)
{
	struct tcp_fastopen_context *ctx;

	ctx = rcu_dereference(inet_csk(sk)->icsk_accept_queue.fastopenq.ctx);
	if (!ctx)
		ctx = rcu_dereference(sock_net(sk)->ipv4.tcp_fastopen_ctx);
	return ctx;
}

static inline
bool tcp_fastopen_cookie_match(const struct tcp_fastopen_cookie *foc,
			       const struct tcp_fastopen_cookie *orig)
{
	if (orig->len == TCP_FASTOPEN_COOKIE_SIZE &&
	    orig->len == foc->len &&
	    !memcmp(orig->val, foc->val, foc->len))
		return true;
	return false;
}

static inline
int tcp_fastopen_context_len(const struct tcp_fastopen_context *ctx)
{
1730
	return ctx->num;
1731 1732
}

1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
/* Latencies incurred by various limits for a sender. They are
 * chronograph-like stats that are mutually exclusive.
 */
enum tcp_chrono {
	TCP_CHRONO_UNSPEC,
	TCP_CHRONO_BUSY, /* Actively sending data (non-empty write queue) */
	TCP_CHRONO_RWND_LIMITED, /* Stalled by insufficient receive window */
	TCP_CHRONO_SNDBUF_LIMITED, /* Stalled by insufficient send buffer */
	__TCP_CHRONO_MAX,
};

void tcp_chrono_start(struct sock *sk, const enum tcp_chrono type);
void tcp_chrono_stop(struct sock *sk, const enum tcp_chrono type);

1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
/* This helper is needed, because skb->tcp_tsorted_anchor uses
 * the same memory storage than skb->destructor/_skb_refdst
 */
static inline void tcp_skb_tsorted_anchor_cleanup(struct sk_buff *skb)
{
	skb->destructor = NULL;
	skb->_skb_refdst = 0UL;
}

#define tcp_skb_tsorted_save(skb) {		\
	unsigned long _save = skb->_skb_refdst;	\
	skb->_skb_refdst = 0UL;

#define tcp_skb_tsorted_restore(skb)		\
	skb->_skb_refdst = _save;		\
}

1764
void tcp_write_queue_purge(struct sock *sk);
1765

1766 1767 1768 1769 1770
static inline struct sk_buff *tcp_rtx_queue_head(const struct sock *sk)
{
	return skb_rb_first(&sk->tcp_rtx_queue);
}

1771 1772 1773 1774 1775
static inline struct sk_buff *tcp_rtx_queue_tail(const struct sock *sk)
{
	return skb_rb_last(&sk->tcp_rtx_queue);
}

1776
static inline struct sk_buff *tcp_write_queue_head(const struct sock *sk)
1777
{
1778
	return skb_peek(&sk->sk_write_queue);
1779 1780
}

1781
static inline struct sk_buff *tcp_write_queue_tail(const struct sock *sk)
1782
{
1783
	return skb_peek_tail(&sk->sk_write_queue);
1784 1785
}

1786
#define tcp_for_write_queue_from_safe(skb, tmp, sk)			\
1787
	skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
1788

1789
static inline struct sk_buff *tcp_send_head(const struct sock *sk)
1790
{
1791
	return skb_peek(&sk->sk_write_queue);
1792 1793
}

1794 1795 1796 1797 1798 1799
static inline bool tcp_skb_is_last(const struct sock *sk,
				   const struct sk_buff *skb)
{
	return skb_queue_is_last(&sk->sk_write_queue, skb);
}

1800 1801 1802 1803 1804 1805 1806
/**
 * tcp_write_queue_empty - test if any payload (or FIN) is available in write queue
 * @sk: socket
 *
 * Since the write queue can have a temporary empty skb in it,
 * we must not use "return skb_queue_empty(&sk->sk_write_queue)"
 */
1807
static inline bool tcp_write_queue_empty(const struct sock *sk)
1808
{
1809 1810 1811
	const struct tcp_sock *tp = tcp_sk(sk);

	return tp->write_seq == tp->snd_nxt;
1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
}

static inline bool tcp_rtx_queue_empty(const struct sock *sk)
{
	return RB_EMPTY_ROOT(&sk->tcp_rtx_queue);
}

static inline bool tcp_rtx_and_write_queues_empty(const struct sock *sk)
{
	return tcp_rtx_queue_empty(sk) && tcp_write_queue_empty(sk);
1822 1823 1824 1825
}

static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
{
1826
	__skb_queue_tail(&sk->sk_write_queue, skb);
1827 1828

	/* Queue it, remembering where we must start sending. */
E
Eric Dumazet 已提交
1829
	if (sk->sk_write_queue.next == skb)
1830
		tcp_chrono_start(sk, TCP_CHRONO_BUSY);
1831 1832
}

1833
/* Insert new before skb on the write queue of sk.  */
1834 1835 1836 1837
static inline void tcp_insert_write_queue_before(struct sk_buff *new,
						  struct sk_buff *skb,
						  struct sock *sk)
{
1838
	__skb_queue_before(&sk->sk_write_queue, skb, new);
1839 1840 1841 1842
}

static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
{
E
Eric Dumazet 已提交
1843
	tcp_skb_tsorted_anchor_cleanup(skb);
1844 1845 1846
	__skb_unlink(skb, &sk->sk_write_queue);
}

1847 1848 1849
void tcp_rbtree_insert(struct rb_root *root, struct sk_buff *skb);

static inline void tcp_rtx_queue_unlink(struct sk_buff *skb, struct sock *sk)
1850
{
1851 1852 1853 1854 1855 1856 1857 1858 1859
	tcp_skb_tsorted_anchor_cleanup(skb);
	rb_erase(&skb->rbnode, &sk->tcp_rtx_queue);
}

static inline void tcp_rtx_queue_unlink_and_free(struct sk_buff *skb, struct sock *sk)
{
	list_del(&skb->tcp_tsorted_anchor);
	tcp_rtx_queue_unlink(skb, sk);
	sk_wmem_free_skb(sk, skb);
1860 1861
}

1862 1863 1864 1865 1866 1867 1868 1869 1870
static inline void tcp_push_pending_frames(struct sock *sk)
{
	if (tcp_send_head(sk)) {
		struct tcp_sock *tp = tcp_sk(sk);

		__tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
	}
}

1871 1872 1873
/* Start sequence of the skb just after the highest skb with SACKed
 * bit, valid only if sacked_out > 0 or when the caller has ensured
 * validity by itself.
1874 1875 1876 1877 1878
 */
static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
{
	if (!tp->sacked_out)
		return tp->snd_una;
1879 1880 1881 1882

	if (tp->highest_sack == NULL)
		return tp->snd_nxt;

1883 1884 1885
	return TCP_SKB_CB(tp->highest_sack)->seq;
}

1886 1887
static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
{
E
Eric Dumazet 已提交
1888
	tcp_sk(sk)->highest_sack = skb_rb_next(skb);
1889 1890 1891 1892 1893 1894 1895 1896 1897
}

static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
{
	return tcp_sk(sk)->highest_sack;
}

static inline void tcp_highest_sack_reset(struct sock *sk)
{
E
Eric Dumazet 已提交
1898
	tcp_sk(sk)->highest_sack = tcp_rtx_queue_head(sk);
1899 1900
}

1901 1902
/* Called when old skb is about to be deleted and replaced by new skb */
static inline void tcp_highest_sack_replace(struct sock *sk,
1903 1904 1905
					    struct sk_buff *old,
					    struct sk_buff *new)
{
1906
	if (old == tcp_highest_sack(sk))
1907 1908 1909
		tcp_sk(sk)->highest_sack = new;
}

1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921
/* This helper checks if socket has IP_TRANSPARENT set */
static inline bool inet_sk_transparent(const struct sock *sk)
{
	switch (sk->sk_state) {
	case TCP_TIME_WAIT:
		return inet_twsk(sk)->tw_transparent;
	case TCP_NEW_SYN_RECV:
		return inet_rsk(inet_reqsk(sk))->no_srccheck;
	}
	return inet_sk(sk)->transparent;
}

A
Andreas Petlund 已提交
1922 1923 1924
/* Determines whether this is a thin stream (which may suffer from
 * increased latency). Used to trigger latency-reducing mechanisms.
 */
E
Eric Dumazet 已提交
1925
static inline bool tcp_stream_is_thin(struct tcp_sock *tp)
A
Andreas Petlund 已提交
1926 1927 1928 1929
{
	return tp->packets_out < 4 && !tcp_in_initial_slowstart(tp);
}

L
Linus Torvalds 已提交
1930 1931 1932 1933 1934 1935
/* /proc */
enum tcp_seq_states {
	TCP_SEQ_STATE_LISTENING,
	TCP_SEQ_STATE_ESTABLISHED,
};

1936 1937 1938
void *tcp_seq_start(struct seq_file *seq, loff_t *pos);
void *tcp_seq_next(struct seq_file *seq, void *v, loff_t *pos);
void tcp_seq_stop(struct seq_file *seq, void *v);
1939

L
Linus Torvalds 已提交
1940
struct tcp_seq_afinfo {
1941
	sa_family_t			family;
L
Linus Torvalds 已提交
1942 1943 1944
};

struct tcp_iter_state {
1945
	struct seq_net_private	p;
L
Linus Torvalds 已提交
1946 1947
	enum tcp_seq_states	state;
	struct sock		*syn_wait_sk;
1948
	struct tcp_seq_afinfo	*bpf_seq_afinfo;
1949
	int			bucket, offset, sbucket, num;
1950
	loff_t			last_pos;
L
Linus Torvalds 已提交
1951 1952
};

1953
extern struct request_sock_ops tcp_request_sock_ops;
1954
extern struct request_sock_ops tcp6_request_sock_ops;
1955

1956
void tcp_v4_destroy_sock(struct sock *sk);
1957

E
Eric Dumazet 已提交
1958
struct sk_buff *tcp_gso_segment(struct sk_buff *skb,
1959
				netdev_features_t features);
1960
struct sk_buff *tcp_gro_receive(struct list_head *head, struct sk_buff *skb);
1961 1962 1963 1964
INDIRECT_CALLABLE_DECLARE(int tcp4_gro_complete(struct sk_buff *skb, int thoff));
INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp4_gro_receive(struct list_head *head, struct sk_buff *skb));
INDIRECT_CALLABLE_DECLARE(int tcp6_gro_complete(struct sk_buff *skb, int thoff));
INDIRECT_CALLABLE_DECLARE(struct sk_buff *tcp6_gro_receive(struct list_head *head, struct sk_buff *skb));
1965
int tcp_gro_complete(struct sk_buff *skb);
1966

1967
void __tcp_v4_send_check(struct sk_buff *skb, __be32 saddr, __be32 daddr);
H
Herbert Xu 已提交
1968

1969 1970
static inline u32 tcp_notsent_lowat(const struct tcp_sock *tp)
{
1971 1972
	struct net *net = sock_net((struct sock *)tp);
	return tp->notsent_lowat ?: net->ipv4.sysctl_tcp_notsent_lowat;
1973 1974
}

1975
bool tcp_stream_memory_free(const struct sock *sk, int wake);
1976

1977
#ifdef CONFIG_PROC_FS
1978 1979
int tcp4_proc_init(void);
void tcp4_proc_exit(void);
1980 1981
#endif

1982
int tcp_rtx_synack(const struct sock *sk, struct request_sock *req);
O
Octavian Purdila 已提交
1983 1984 1985
int tcp_conn_request(struct request_sock_ops *rsk_ops,
		     const struct tcp_request_sock_ops *af_ops,
		     struct sock *sk, struct sk_buff *skb);
1986

1987 1988 1989
/* TCP af-specific functions */
struct tcp_sock_af_ops {
#ifdef CONFIG_TCP_MD5SIG
1990
	struct tcp_md5sig_key	*(*md5_lookup) (const struct sock *sk,
1991
						const struct sock *addr_sk);
1992 1993 1994 1995 1996
	int		(*calc_md5_hash)(char *location,
					 const struct tcp_md5sig_key *md5,
					 const struct sock *sk,
					 const struct sk_buff *skb);
	int		(*md5_parse)(struct sock *sk,
1997
				     int optname,
1998
				     sockptr_t optval,
1999
				     int optlen);
2000 2001 2002 2003
#endif
};

struct tcp_request_sock_ops {
2004
	u16 mss_clamp;
2005
#ifdef CONFIG_TCP_MD5SIG
2006
	struct tcp_md5sig_key *(*req_md5_lookup)(const struct sock *sk,
2007
						 const struct sock *addr_sk);
2008 2009 2010 2011
	int		(*calc_md5_hash) (char *location,
					  const struct tcp_md5sig_key *md5,
					  const struct sock *sk,
					  const struct sk_buff *skb);
2012
#endif
2013
#ifdef CONFIG_SYN_COOKIES
2014
	__u32 (*cookie_init_seq)(const struct sk_buff *skb,
2015 2016
				 __u16 *mss);
#endif
2017 2018 2019 2020
	struct dst_entry *(*route_req)(const struct sock *sk,
				       struct sk_buff *skb,
				       struct flowi *fl,
				       struct request_sock *req);
2021
	u32 (*init_seq)(const struct sk_buff *skb);
2022
	u32 (*init_ts_off)(const struct net *net, const struct sk_buff *skb);
2023
	int (*send_synack)(const struct sock *sk, struct dst_entry *dst,
2024
			   struct flowi *fl, struct request_sock *req,
2025
			   struct tcp_fastopen_cookie *foc,
2026 2027
			   enum tcp_synack_type synack_type,
			   struct sk_buff *syn_skb);
2028 2029
};

2030 2031 2032 2033 2034
extern const struct tcp_request_sock_ops tcp_request_sock_ipv4_ops;
#if IS_ENABLED(CONFIG_IPV6)
extern const struct tcp_request_sock_ops tcp_request_sock_ipv6_ops;
#endif

2035 2036
#ifdef CONFIG_SYN_COOKIES
static inline __u32 cookie_init_sequence(const struct tcp_request_sock_ops *ops,
2037
					 const struct sock *sk, struct sk_buff *skb,
2038 2039
					 __u16 *mss)
{
2040
	tcp_synq_overflow(sk);
2041
	__NET_INC_STATS(sock_net(sk), LINUX_MIB_SYNCOOKIESSENT);
2042
	return ops->cookie_init_seq(skb, mss);
2043 2044 2045
}
#else
static inline __u32 cookie_init_sequence(const struct tcp_request_sock_ops *ops,
2046
					 const struct sock *sk, struct sk_buff *skb,
2047 2048 2049 2050 2051 2052
					 __u16 *mss)
{
	return 0;
}
#endif

2053
int tcpv4_offload_init(void);
2054

2055 2056
void tcp_v4_init(void);
void tcp_init(void);
2057

2058
/* tcp_recovery.c */
2059
void tcp_mark_skb_lost(struct sock *sk, struct sk_buff *skb);
2060
void tcp_newreno_mark_lost(struct sock *sk, bool snd_una_advanced);
2061 2062
extern s32 tcp_rack_skb_timeout(struct tcp_sock *tp, struct sk_buff *skb,
				u32 reo_wnd);
2063
extern void tcp_rack_mark_lost(struct sock *sk);
2064
extern void tcp_rack_advance(struct tcp_sock *tp, u8 sacked, u32 end_seq,
2065
			     u64 xmit_time);
2066
extern void tcp_rack_reo_timeout(struct sock *sk);
2067
extern void tcp_rack_update_reo_wnd(struct sock *sk, struct rate_sample *rs);
2068

2069 2070 2071
/* At how many usecs into the future should the RTO fire? */
static inline s64 tcp_rto_delta_us(const struct sock *sk)
{
2072
	const struct sk_buff *skb = tcp_rtx_queue_head(sk);
2073
	u32 rto = inet_csk(sk)->icsk_rto;
2074
	u64 rto_time_stamp_us = tcp_skb_timestamp_us(skb) + jiffies_to_usecs(rto);
2075 2076 2077 2078

	return rto_time_stamp_us - tcp_sk(sk)->tcp_mstamp;
}

2079 2080 2081
/*
 * Save and compile IPv4 options, return a pointer to it
 */
2082 2083
static inline struct ip_options_rcu *tcp_v4_save_options(struct net *net,
							 struct sk_buff *skb)
2084 2085 2086 2087
{
	const struct ip_options *opt = &TCP_SKB_CB(skb)->header.h4.opt;
	struct ip_options_rcu *dopt = NULL;

C
Cong Wang 已提交
2088
	if (opt->optlen) {
2089 2090 2091
		int opt_size = sizeof(*dopt) + opt->optlen;

		dopt = kmalloc(opt_size, GFP_ATOMIC);
2092
		if (dopt && __ip_options_echo(net, &dopt->opt, skb, opt)) {
2093 2094 2095 2096 2097 2098 2099
			kfree(dopt);
			dopt = NULL;
		}
	}
	return dopt;
}

E
Eric Dumazet 已提交
2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114
/* locally generated TCP pure ACKs have skb->truesize == 2
 * (check tcp_send_ack() in net/ipv4/tcp_output.c )
 * This is much faster than dissecting the packet to find out.
 * (Think of GRE encapsulations, IPv4, IPv6, ...)
 */
static inline bool skb_is_tcp_pure_ack(const struct sk_buff *skb)
{
	return skb->truesize == 2;
}

static inline void skb_set_tcp_pure_ack(struct sk_buff *skb)
{
	skb->truesize = 2;
}

2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138
static inline int tcp_inq(struct sock *sk)
{
	struct tcp_sock *tp = tcp_sk(sk);
	int answ;

	if ((1 << sk->sk_state) & (TCPF_SYN_SENT | TCPF_SYN_RECV)) {
		answ = 0;
	} else if (sock_flag(sk, SOCK_URGINLINE) ||
		   !tp->urg_data ||
		   before(tp->urg_seq, tp->copied_seq) ||
		   !before(tp->urg_seq, tp->rcv_nxt)) {

		answ = tp->rcv_nxt - tp->copied_seq;

		/* Subtract 1, if FIN was received */
		if (answ && sock_flag(sk, SOCK_DONE))
			answ--;
	} else {
		answ = tp->urg_seq - tp->copied_seq;
	}

	return answ;
}

2139 2140
int tcp_peek_len(struct socket *sock);

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150
static inline void tcp_segs_in(struct tcp_sock *tp, const struct sk_buff *skb)
{
	u16 segs_in;

	segs_in = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
	tp->segs_in += segs_in;
	if (skb->len > tcp_hdrlen(skb))
		tp->data_segs_in += segs_in;
}

2151 2152 2153 2154 2155 2156 2157 2158 2159 2160
/*
 * TCP listen path runs lockless.
 * We forced "struct sock" to be const qualified to make sure
 * we don't modify one of its field by mistake.
 * Here, we increment sk_drops which is an atomic_t, so we can safely
 * make sock writable again.
 */
static inline void tcp_listendrop(const struct sock *sk)
{
	atomic_inc(&((struct sock *)sk)->sk_drops);
2161
	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
2162 2163
}

2164 2165
enum hrtimer_restart tcp_pace_kick(struct hrtimer *timer);

D
Dave Watson 已提交
2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178
/*
 * Interface for adding Upper Level Protocols over TCP
 */

#define TCP_ULP_NAME_MAX	16
#define TCP_ULP_MAX		128
#define TCP_ULP_BUF_MAX		(TCP_ULP_NAME_MAX*TCP_ULP_MAX)

struct tcp_ulp_ops {
	struct list_head	list;

	/* initialize ulp */
	int (*init)(struct sock *sk);
2179
	/* update ulp */
2180 2181
	void (*update)(struct sock *sk, struct proto *p,
		       void (*write_space)(struct sock *sk));
D
Dave Watson 已提交
2182 2183
	/* cleanup ulp */
	void (*release)(struct sock *sk);
2184 2185 2186
	/* diagnostic */
	int (*get_info)(const struct sock *sk, struct sk_buff *skb);
	size_t (*get_info_size)(const struct sock *sk);
2187 2188 2189
	/* clone ulp */
	void (*clone)(const struct request_sock *req, struct sock *newsk,
		      const gfp_t priority);
D
Dave Watson 已提交
2190 2191 2192 2193 2194 2195 2196 2197 2198

	char		name[TCP_ULP_NAME_MAX];
	struct module	*owner;
};
int tcp_register_ulp(struct tcp_ulp_ops *type);
void tcp_unregister_ulp(struct tcp_ulp_ops *type);
int tcp_set_ulp(struct sock *sk, const char *name);
void tcp_get_available_ulp(char *buf, size_t len);
void tcp_cleanup_ulp(struct sock *sk);
2199 2200
void tcp_update_ulp(struct sock *sk, struct proto *p,
		    void (*write_space)(struct sock *sk));
D
Dave Watson 已提交
2201

2202 2203 2204 2205
#define MODULE_ALIAS_TCP_ULP(name)				\
	__MODULE_INFO(alias, alias_userspace, name);		\
	__MODULE_INFO(alias, alias_tcp_ulp, "tcp-ulp-" name)

2206 2207 2208
struct sk_msg;
struct sk_psock;

2209 2210 2211 2212 2213 2214 2215 2216 2217
#ifdef CONFIG_BPF_STREAM_PARSER
struct proto *tcp_bpf_get_proto(struct sock *sk, struct sk_psock *psock);
void tcp_bpf_clone(const struct sock *sk, struct sock *newsk);
#else
static inline void tcp_bpf_clone(const struct sock *sk, struct sock *newsk)
{
}
#endif /* CONFIG_BPF_STREAM_PARSER */

2218
#ifdef CONFIG_NET_SOCK_MSG
2219 2220 2221
int tcp_bpf_sendmsg_redir(struct sock *sk, struct sk_msg *msg, u32 bytes,
			  int flags);
int __tcp_bpf_recvmsg(struct sock *sk, struct sk_psock *psock,
2222
		      struct msghdr *msg, int len, int flags);
2223
#endif /* CONFIG_NET_SOCK_MSG */
2224

2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240
#ifdef CONFIG_CGROUP_BPF
static inline void bpf_skops_init_skb(struct bpf_sock_ops_kern *skops,
				      struct sk_buff *skb,
				      unsigned int end_offset)
{
	skops->skb = skb;
	skops->skb_data_end = skb->data + end_offset;
}
#else
static inline void bpf_skops_init_skb(struct bpf_sock_ops_kern *skops,
				      struct sk_buff *skb,
				      unsigned int end_offset)
{
}
#endif

L
Lawrence Brakmo 已提交
2241 2242 2243 2244 2245 2246
/* Call BPF_SOCK_OPS program that returns an int. If the return value
 * is < 0, then the BPF op failed (for example if the loaded BPF
 * program does not support the chosen operation or there is no BPF
 * program loaded).
 */
#ifdef CONFIG_BPF
2247
static inline int tcp_call_bpf(struct sock *sk, int op, u32 nargs, u32 *args)
L
Lawrence Brakmo 已提交
2248 2249 2250 2251
{
	struct bpf_sock_ops_kern sock_ops;
	int ret;

2252
	memset(&sock_ops, 0, offsetof(struct bpf_sock_ops_kern, temp));
2253 2254
	if (sk_fullsock(sk)) {
		sock_ops.is_fullsock = 1;
L
Lawrence Brakmo 已提交
2255
		sock_owned_by_me(sk);
2256
	}
L
Lawrence Brakmo 已提交
2257 2258 2259

	sock_ops.sk = sk;
	sock_ops.op = op;
2260 2261
	if (nargs > 0)
		memcpy(sock_ops.args, args, nargs * sizeof(*args));
L
Lawrence Brakmo 已提交
2262 2263 2264 2265 2266 2267 2268 2269

	ret = BPF_CGROUP_RUN_PROG_SOCK_OPS(&sock_ops);
	if (ret == 0)
		ret = sock_ops.reply;
	else
		ret = -1;
	return ret;
}
2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285

static inline int tcp_call_bpf_2arg(struct sock *sk, int op, u32 arg1, u32 arg2)
{
	u32 args[2] = {arg1, arg2};

	return tcp_call_bpf(sk, op, 2, args);
}

static inline int tcp_call_bpf_3arg(struct sock *sk, int op, u32 arg1, u32 arg2,
				    u32 arg3)
{
	u32 args[3] = {arg1, arg2, arg3};

	return tcp_call_bpf(sk, op, 3, args);
}

L
Lawrence Brakmo 已提交
2286
#else
2287
static inline int tcp_call_bpf(struct sock *sk, int op, u32 nargs, u32 *args)
L
Lawrence Brakmo 已提交
2288 2289 2290
{
	return -EPERM;
}
2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302

static inline int tcp_call_bpf_2arg(struct sock *sk, int op, u32 arg1, u32 arg2)
{
	return -EPERM;
}

static inline int tcp_call_bpf_3arg(struct sock *sk, int op, u32 arg1, u32 arg2,
				    u32 arg3)
{
	return -EPERM;
}

L
Lawrence Brakmo 已提交
2303 2304
#endif

2305 2306 2307 2308
static inline u32 tcp_timeout_init(struct sock *sk)
{
	int timeout;

2309
	timeout = tcp_call_bpf(sk, BPF_SOCK_OPS_TIMEOUT_INIT, 0, NULL);
2310 2311 2312 2313 2314 2315

	if (timeout <= 0)
		timeout = TCP_TIMEOUT_INIT;
	return timeout;
}

2316 2317 2318 2319
static inline u32 tcp_rwnd_init_bpf(struct sock *sk)
{
	int rwnd;

2320
	rwnd = tcp_call_bpf(sk, BPF_SOCK_OPS_RWND_INIT, 0, NULL);
2321 2322 2323 2324 2325

	if (rwnd < 0)
		rwnd = 0;
	return rwnd;
}
2326 2327 2328

static inline bool tcp_bpf_ca_needs_ecn(struct sock *sk)
{
2329
	return (tcp_call_bpf(sk, BPF_SOCK_OPS_NEEDS_ECN, 0, NULL) == 1);
2330
}
2331

2332 2333
static inline void tcp_bpf_rtt(struct sock *sk)
{
2334
	if (BPF_SOCK_OPS_TEST_FLAG(tcp_sk(sk), BPF_SOCK_OPS_RTT_CB_FLAG))
2335 2336 2337
		tcp_call_bpf(sk, BPF_SOCK_OPS_RTT_CB, 0, NULL);
}

2338 2339 2340
#if IS_ENABLED(CONFIG_SMC)
extern struct static_key_false tcp_have_smc;
#endif
I
Ilya Lesokhin 已提交
2341 2342 2343 2344 2345

#if IS_ENABLED(CONFIG_TLS_DEVICE)
void clean_acked_data_enable(struct inet_connection_sock *icsk,
			     void (*cad)(struct sock *sk, u32 ack_seq));
void clean_acked_data_disable(struct inet_connection_sock *icsk);
2346
void clean_acked_data_flush(void);
I
Ilya Lesokhin 已提交
2347 2348
#endif

2349 2350 2351 2352 2353 2354 2355 2356
DECLARE_STATIC_KEY_FALSE(tcp_tx_delay_enabled);
static inline void tcp_add_tx_delay(struct sk_buff *skb,
				    const struct tcp_sock *tp)
{
	if (static_branch_unlikely(&tcp_tx_delay_enabled))
		skb->skb_mstamp_ns += (u64)tp->tcp_tx_delay * NSEC_PER_USEC;
}

2357 2358 2359 2360
/* Compute Earliest Departure Time for some control packets
 * like ACK or RST for TIME_WAIT or non ESTABLISHED sockets.
 */
static inline u64 tcp_transmit_time(const struct sock *sk)
2361 2362 2363 2364 2365
{
	if (static_branch_unlikely(&tcp_tx_delay_enabled)) {
		u32 delay = (sk->sk_state == TCP_TIME_WAIT) ?
			tcp_twsk(sk)->tw_tx_delay : tcp_sk(sk)->tcp_tx_delay;

2366
		return tcp_clock_ns() + (u64)delay * NSEC_PER_USEC;
2367
	}
2368
	return 0;
2369 2370
}

L
Linus Torvalds 已提交
2371
#endif	/* _TCP_H */