projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
netfilter: nf_conntrack: add support for "conntrack zones"
[deliverable/linux.git] / include / net / tcp.h
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the TCP module.
7 *
8 * Version: @(#)tcp.h 1.0.5 05/23/93
9 *
10 * Authors: Ross Biro
11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 *
13 * This program is free software; you can redistribute it and/or
14 * modify it under the terms of the GNU General Public License
15 * as published by the Free Software Foundation; either version
16 * 2 of the License, or (at your option) any later version.
17 */
18 #ifndef _TCP_H
19 #define _TCP_H
20
21 #define TCP_DEBUG 1
22 #define FASTRETRANS_DEBUG 1
23
24 #include <linux/list.h>
25 #include <linux/tcp.h>
26 #include <linux/slab.h>
27 #include <linux/cache.h>
28 #include <linux/percpu.h>
29 #include <linux/skbuff.h>
30 #include <linux/dmaengine.h>
31 #include <linux/crypto.h>
32 #include <linux/cryptohash.h>
33 #include <linux/kref.h>
34
35 #include <net/inet_connection_sock.h>
36 #include <net/inet_timewait_sock.h>
37 #include <net/inet_hashtables.h>
38 #include <net/checksum.h>
39 #include <net/request_sock.h>
40 #include <net/sock.h>
41 #include <net/snmp.h>
42 #include <net/ip.h>
43 #include <net/tcp_states.h>
44 #include <net/inet_ecn.h>
45 #include <net/dst.h>
46
47 #include <linux/seq_file.h>
48
49 extern struct inet_hashinfo tcp_hashinfo;
50
51 extern struct percpu_counter tcp_orphan_count;
52 extern void tcp_time_wait(struct sock *sk, int state, int timeo);
53
54 #define MAX_TCP_HEADER (128 + MAX_HEADER)
55 #define MAX_TCP_OPTION_SPACE 40
56
57 /*
58 * Never offer a window over 32767 without using window scaling. Some
59 * poor stacks do signed 16bit maths!
60 */
61 #define MAX_TCP_WINDOW 32767U
62
63 /* Minimal accepted MSS. It is (60+60+8) - (20+20). */
64 #define TCP_MIN_MSS 88U
65
66 /* The least MTU to use for probing */
67 #define TCP_BASE_MSS 512
68
69 /* After receiving this amount of duplicate ACKs fast retransmit starts. */
70 #define TCP_FASTRETRANS_THRESH 3
71
72 /* Maximal reordering. */
73 #define TCP_MAX_REORDERING 127
74
75 /* Maximal number of ACKs sent quickly to accelerate slow-start. */
76 #define TCP_MAX_QUICKACKS 16U
77
78 /* urg_data states */
79 #define TCP_URG_VALID 0x0100
80 #define TCP_URG_NOTYET 0x0200
81 #define TCP_URG_READ 0x0400
82
83 #define TCP_RETR1 3 /*
84 * This is how many retries it does before it
85 * tries to figure out if the gateway is
86 * down. Minimal RFC value is 3; it corresponds
87 * to ~3sec-8min depending on RTO.
88 */
89
90 #define TCP_RETR2 15 /*
91 * This should take at least
92 * 90 minutes to time out.
93 * RFC1122 says that the limit is 100 sec.
94 * 15 is ~13-30min depending on RTO.
95 */
96
97 #define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
98 * connection: ~180sec is RFC minimum */
99
100 #define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
101 * connection: ~180sec is RFC minimum */
102
103
104 #define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
105 * socket. 7 is ~50sec-16min.
106 */
107
108
109 #define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
110 * state, about 60 seconds */
111 #define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
112 /* BSD style FIN_WAIT2 deadlock breaker.
113 * It used to be 3min, new value is 60sec,
114 * to combine FIN-WAIT-2 timeout with
115 * TIME-WAIT timer.
116 */
117
118 #define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
119 #if HZ >= 100
120 #define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
121 #define TCP_ATO_MIN ((unsigned)(HZ/25))
122 #else
123 #define TCP_DELACK_MIN 4U
124 #define TCP_ATO_MIN 4U
125 #endif
126 #define TCP_RTO_MAX ((unsigned)(120*HZ))
127 #define TCP_RTO_MIN ((unsigned)(HZ/5))
128 #define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
129
130 #define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
131 * for local resources.
132 */
133
134 #define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
135 #define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
136 #define TCP_KEEPALIVE_INTVL (75*HZ)
137
138 #define MAX_TCP_KEEPIDLE 32767
139 #define MAX_TCP_KEEPINTVL 32767
140 #define MAX_TCP_KEEPCNT 127
141 #define MAX_TCP_SYNCNT 127
142
143 #define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
144
145 #define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
146 #define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
147 * after this time. It should be equal
148 * (or greater than) TCP_TIMEWAIT_LEN
149 * to provide reliability equal to one
150 * provided by timewait state.
151 */
152 #define TCP_PAWS_WINDOW 1 /* Replay window for per-host
153 * timestamps. It must be less than
154 * minimal timewait lifetime.
155 */
156 /*
157 * TCP option
158 */
159
160 #define TCPOPT_NOP 1 /* Padding */
161 #define TCPOPT_EOL 0 /* End of options */
162 #define TCPOPT_MSS 2 /* Segment size negotiating */
163 #define TCPOPT_WINDOW 3 /* Window scaling */
164 #define TCPOPT_SACK_PERM 4 /* SACK Permitted */
165 #define TCPOPT_SACK 5 /* SACK Block */
166 #define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
167 #define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
168 #define TCPOPT_COOKIE 253 /* Cookie extension (experimental) */
169
170 /*
171 * TCP option lengths
172 */
173
174 #define TCPOLEN_MSS 4
175 #define TCPOLEN_WINDOW 3
176 #define TCPOLEN_SACK_PERM 2
177 #define TCPOLEN_TIMESTAMP 10
178 #define TCPOLEN_MD5SIG 18
179 #define TCPOLEN_COOKIE_BASE 2 /* Cookie-less header extension */
180 #define TCPOLEN_COOKIE_PAIR 3 /* Cookie pair header extension */
181 #define TCPOLEN_COOKIE_MIN (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MIN)
182 #define TCPOLEN_COOKIE_MAX (TCPOLEN_COOKIE_BASE+TCP_COOKIE_MAX)
183
184 /* But this is what stacks really send out. */
185 #define TCPOLEN_TSTAMP_ALIGNED 12
186 #define TCPOLEN_WSCALE_ALIGNED 4
187 #define TCPOLEN_SACKPERM_ALIGNED 4
188 #define TCPOLEN_SACK_BASE 2
189 #define TCPOLEN_SACK_BASE_ALIGNED 4
190 #define TCPOLEN_SACK_PERBLOCK 8
191 #define TCPOLEN_MD5SIG_ALIGNED 20
192 #define TCPOLEN_MSS_ALIGNED 4
193
194 /* Flags in tp->nonagle */
195 #define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
196 #define TCP_NAGLE_CORK 2 /* Socket is corked */
197 #define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
198
199 extern struct inet_timewait_death_row tcp_death_row;
200
201 /* sysctl variables for tcp */
202 extern int sysctl_tcp_timestamps;
203 extern int sysctl_tcp_window_scaling;
204 extern int sysctl_tcp_sack;
205 extern int sysctl_tcp_fin_timeout;
206 extern int sysctl_tcp_keepalive_time;
207 extern int sysctl_tcp_keepalive_probes;
208 extern int sysctl_tcp_keepalive_intvl;
209 extern int sysctl_tcp_syn_retries;
210 extern int sysctl_tcp_synack_retries;
211 extern int sysctl_tcp_retries1;
212 extern int sysctl_tcp_retries2;
213 extern int sysctl_tcp_orphan_retries;
214 extern int sysctl_tcp_syncookies;
215 extern int sysctl_tcp_retrans_collapse;
216 extern int sysctl_tcp_stdurg;
217 extern int sysctl_tcp_rfc1337;
218 extern int sysctl_tcp_abort_on_overflow;
219 extern int sysctl_tcp_max_orphans;
220 extern int sysctl_tcp_fack;
221 extern int sysctl_tcp_reordering;
222 extern int sysctl_tcp_ecn;
223 extern int sysctl_tcp_dsack;
224 extern int sysctl_tcp_mem[3];
225 extern int sysctl_tcp_wmem[3];
226 extern int sysctl_tcp_rmem[3];
227 extern int sysctl_tcp_app_win;
228 extern int sysctl_tcp_adv_win_scale;
229 extern int sysctl_tcp_tw_reuse;
230 extern int sysctl_tcp_frto;
231 extern int sysctl_tcp_frto_response;
232 extern int sysctl_tcp_low_latency;
233 extern int sysctl_tcp_dma_copybreak;
234 extern int sysctl_tcp_nometrics_save;
235 extern int sysctl_tcp_moderate_rcvbuf;
236 extern int sysctl_tcp_tso_win_divisor;
237 extern int sysctl_tcp_abc;
238 extern int sysctl_tcp_mtu_probing;
239 extern int sysctl_tcp_base_mss;
240 extern int sysctl_tcp_workaround_signed_windows;
241 extern int sysctl_tcp_slow_start_after_idle;
242 extern int sysctl_tcp_max_ssthresh;
243 extern int sysctl_tcp_cookie_size;
244
245 extern atomic_t tcp_memory_allocated;
246 extern struct percpu_counter tcp_sockets_allocated;
247 extern int tcp_memory_pressure;
248
249 /*
250 * The next routines deal with comparing 32 bit unsigned ints
251 * and worry about wraparound (automatic with unsigned arithmetic).
252 */
253
254 static inline int before(__u32 seq1, __u32 seq2)
255 {
256 return (__s32)(seq1-seq2) < 0;
257 }
258 #define after(seq2, seq1) before(seq1, seq2)
259
260 /* is s2<=s1<=s3 ? */
261 static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
262 {
263 return seq3 - seq2 >= seq1 - seq2;
264 }
265
266 static inline int tcp_too_many_orphans(struct sock *sk, int num)
267 {
268 return (num > sysctl_tcp_max_orphans) ||
269 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
270 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
271 }
272
273 /* syncookies: remember time of last synqueue overflow */
274 static inline void tcp_synq_overflow(struct sock *sk)
275 {
276 tcp_sk(sk)->rx_opt.ts_recent_stamp = jiffies;
277 }
278
279 /* syncookies: no recent synqueue overflow on this listening socket? */
280 static inline int tcp_synq_no_recent_overflow(const struct sock *sk)
281 {
282 unsigned long last_overflow = tcp_sk(sk)->rx_opt.ts_recent_stamp;
283 return time_after(jiffies, last_overflow + TCP_TIMEOUT_INIT);
284 }
285
286 extern struct proto tcp_prot;
287
288 #define TCP_INC_STATS(net, field) SNMP_INC_STATS((net)->mib.tcp_statistics, field)
289 #define TCP_INC_STATS_BH(net, field) SNMP_INC_STATS_BH((net)->mib.tcp_statistics, field)
290 #define TCP_DEC_STATS(net, field) SNMP_DEC_STATS((net)->mib.tcp_statistics, field)
291 #define TCP_ADD_STATS_USER(net, field, val) SNMP_ADD_STATS_USER((net)->mib.tcp_statistics, field, val)
292
293 extern void tcp_v4_err(struct sk_buff *skb, u32);
294
295 extern void tcp_shutdown (struct sock *sk, int how);
296
297 extern int tcp_v4_rcv(struct sk_buff *skb);
298
299 extern int tcp_v4_remember_stamp(struct sock *sk);
300
301 extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
302
303 extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
304 struct msghdr *msg, size_t size);
305 extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
306
307 extern int tcp_ioctl(struct sock *sk,
308 int cmd,
309 unsigned long arg);
310
311 extern int tcp_rcv_state_process(struct sock *sk,
312 struct sk_buff *skb,
313 struct tcphdr *th,
314 unsigned len);
315
316 extern int tcp_rcv_established(struct sock *sk,
317 struct sk_buff *skb,
318 struct tcphdr *th,
319 unsigned len);
320
321 extern void tcp_rcv_space_adjust(struct sock *sk);
322
323 extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
324
325 extern int tcp_twsk_unique(struct sock *sk,
326 struct sock *sktw, void *twp);
327
328 extern void tcp_twsk_destructor(struct sock *sk);
329
330 extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
331 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
332
333 static inline void tcp_dec_quickack_mode(struct sock *sk,
334 const unsigned int pkts)
335 {
336 struct inet_connection_sock *icsk = inet_csk(sk);
337
338 if (icsk->icsk_ack.quick) {
339 if (pkts >= icsk->icsk_ack.quick) {
340 icsk->icsk_ack.quick = 0;
341 /* Leaving quickack mode we deflate ATO. */
342 icsk->icsk_ack.ato = TCP_ATO_MIN;
343 } else
344 icsk->icsk_ack.quick -= pkts;
345 }
346 }
347
348 extern void tcp_enter_quickack_mode(struct sock *sk);
349
350 #define TCP_ECN_OK 1
351 #define TCP_ECN_QUEUE_CWR 2
352 #define TCP_ECN_DEMAND_CWR 4
353
354 static __inline__ void
355 TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
356 {
357 if (sysctl_tcp_ecn && th->ece && th->cwr)
358 inet_rsk(req)->ecn_ok = 1;
359 }
360
361 enum tcp_tw_status {
362 TCP_TW_SUCCESS = 0,
363 TCP_TW_RST = 1,
364 TCP_TW_ACK = 2,
365 TCP_TW_SYN = 3
366 };
367
368
369 extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
370 struct sk_buff *skb,
371 const struct tcphdr *th);
372
373 extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
374 struct request_sock *req,
375 struct request_sock **prev);
376 extern int tcp_child_process(struct sock *parent,
377 struct sock *child,
378 struct sk_buff *skb);
379 extern int tcp_use_frto(struct sock *sk);
380 extern void tcp_enter_frto(struct sock *sk);
381 extern void tcp_enter_loss(struct sock *sk, int how);
382 extern void tcp_clear_retrans(struct tcp_sock *tp);
383 extern void tcp_update_metrics(struct sock *sk);
384
385 extern void tcp_close(struct sock *sk,
386 long timeout);
387 extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
388
389 extern int tcp_getsockopt(struct sock *sk, int level,
390 int optname,
391 char __user *optval,
392 int __user *optlen);
393 extern int tcp_setsockopt(struct sock *sk, int level,
394 int optname, char __user *optval,
395 unsigned int optlen);
396 extern int compat_tcp_getsockopt(struct sock *sk,
397 int level, int optname,
398 char __user *optval, int __user *optlen);
399 extern int compat_tcp_setsockopt(struct sock *sk,
400 int level, int optname,
401 char __user *optval, unsigned int optlen);
402 extern void tcp_set_keepalive(struct sock *sk, int val);
403 extern void tcp_syn_ack_timeout(struct sock *sk,
404 struct request_sock *req);
405 extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
406 struct msghdr *msg,
407 size_t len, int nonblock,
408 int flags, int *addr_len);
409
410 extern void tcp_parse_options(struct sk_buff *skb,
411 struct tcp_options_received *opt_rx,
412 u8 **hvpp,
413 int estab);
414
415 extern u8 *tcp_parse_md5sig_option(struct tcphdr *th);
416
417 /*
418 * TCP v4 functions exported for the inet6 API
419 */
420
421 extern void tcp_v4_send_check(struct sock *sk, int len,
422 struct sk_buff *skb);
423
424 extern int tcp_v4_conn_request(struct sock *sk,
425 struct sk_buff *skb);
426
427 extern struct sock * tcp_create_openreq_child(struct sock *sk,
428 struct request_sock *req,
429 struct sk_buff *skb);
430
431 extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
432 struct sk_buff *skb,
433 struct request_sock *req,
434 struct dst_entry *dst);
435
436 extern int tcp_v4_do_rcv(struct sock *sk,
437 struct sk_buff *skb);
438
439 extern int tcp_v4_connect(struct sock *sk,
440 struct sockaddr *uaddr,
441 int addr_len);
442
443 extern int tcp_connect(struct sock *sk);
444
445 extern struct sk_buff * tcp_make_synack(struct sock *sk,
446 struct dst_entry *dst,
447 struct request_sock *req,
448 struct request_values *rvp);
449
450 extern int tcp_disconnect(struct sock *sk, int flags);
451
452
453 /* From syncookies.c */
454 extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
455 extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
456 struct ip_options *opt);
457 extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
458 __u16 *mss);
459
460 extern __u32 cookie_init_timestamp(struct request_sock *req);
461 extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt);
462
463 /* From net/ipv6/syncookies.c */
464 extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
465 extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
466 __u16 *mss);
467
468 /* tcp_output.c */
469
470 extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
471 int nonagle);
472 extern int tcp_may_send_now(struct sock *sk);
473 extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
474 extern void tcp_retransmit_timer(struct sock *sk);
475 extern void tcp_xmit_retransmit_queue(struct sock *);
476 extern void tcp_simple_retransmit(struct sock *);
477 extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
478 extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
479
480 extern void tcp_send_probe0(struct sock *);
481 extern void tcp_send_partial(struct sock *);
482 extern int tcp_write_wakeup(struct sock *);
483 extern void tcp_send_fin(struct sock *sk);
484 extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
485 extern int tcp_send_synack(struct sock *);
486 extern void tcp_push_one(struct sock *, unsigned int mss_now);
487 extern void tcp_send_ack(struct sock *sk);
488 extern void tcp_send_delayed_ack(struct sock *sk);
489
490 /* tcp_input.c */
491 extern void tcp_cwnd_application_limited(struct sock *sk);
492
493 /* tcp_timer.c */
494 extern void tcp_init_xmit_timers(struct sock *);
495 static inline void tcp_clear_xmit_timers(struct sock *sk)
496 {
497 inet_csk_clear_xmit_timers(sk);
498 }
499
500 extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
501 extern unsigned int tcp_current_mss(struct sock *sk);
502
503 /* Bound MSS / TSO packet size with the half of the window */
504 static inline int tcp_bound_to_half_wnd(struct tcp_sock *tp, int pktsize)
505 {
506 if (tp->max_window && pktsize > (tp->max_window >> 1))
507 return max(tp->max_window >> 1, 68U - tp->tcp_header_len);
508 else
509 return pktsize;
510 }
511
512 /* tcp.c */
513 extern void tcp_get_info(struct sock *, struct tcp_info *);
514
515 /* Read 'sendfile()'-style from a TCP socket */
516 typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
517 unsigned int, size_t);
518 extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
519 sk_read_actor_t recv_actor);
520
521 extern void tcp_initialize_rcv_mss(struct sock *sk);
522
523 extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
524 extern int tcp_mss_to_mtu(struct sock *sk, int mss);
525 extern void tcp_mtup_init(struct sock *sk);
526
527 static inline void tcp_bound_rto(const struct sock *sk)
528 {
529 if (inet_csk(sk)->icsk_rto > TCP_RTO_MAX)
530 inet_csk(sk)->icsk_rto = TCP_RTO_MAX;
531 }
532
533 static inline u32 __tcp_set_rto(const struct tcp_sock *tp)
534 {
535 return (tp->srtt >> 3) + tp->rttvar;
536 }
537
538 static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
539 {
540 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
541 ntohl(TCP_FLAG_ACK) |
542 snd_wnd);
543 }
544
545 static inline void tcp_fast_path_on(struct tcp_sock *tp)
546 {
547 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
548 }
549
550 static inline void tcp_fast_path_check(struct sock *sk)
551 {
552 struct tcp_sock *tp = tcp_sk(sk);
553
554 if (skb_queue_empty(&tp->out_of_order_queue) &&
555 tp->rcv_wnd &&
556 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
557 !tp->urg_data)
558 tcp_fast_path_on(tp);
559 }
560
561 /* Compute the actual rto_min value */
562 static inline u32 tcp_rto_min(struct sock *sk)
563 {
564 struct dst_entry *dst = __sk_dst_get(sk);
565 u32 rto_min = TCP_RTO_MIN;
566
567 if (dst && dst_metric_locked(dst, RTAX_RTO_MIN))
568 rto_min = dst_metric_rtt(dst, RTAX_RTO_MIN);
569 return rto_min;
570 }
571
572 /* Compute the actual receive window we are currently advertising.
573 * Rcv_nxt can be after the window if our peer push more data
574 * than the offered window.
575 */
576 static inline u32 tcp_receive_window(const struct tcp_sock *tp)
577 {
578 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
579
580 if (win < 0)
581 win = 0;
582 return (u32) win;
583 }
584
585 /* Choose a new window, without checks for shrinking, and without
586 * scaling applied to the result. The caller does these things
587 * if necessary. This is a "raw" window selection.
588 */
589 extern u32 __tcp_select_window(struct sock *sk);
590
591 /* TCP timestamps are only 32-bits, this causes a slight
592 * complication on 64-bit systems since we store a snapshot
593 * of jiffies in the buffer control blocks below. We decided
594 * to use only the low 32-bits of jiffies and hide the ugly
595 * casts with the following macro.
596 */
597 #define tcp_time_stamp ((__u32)(jiffies))
598
599 /* This is what the send packet queuing engine uses to pass
600 * TCP per-packet control information to the transmission
601 * code. We also store the host-order sequence numbers in
602 * here too. This is 36 bytes on 32-bit architectures,
603 * 40 bytes on 64-bit machines, if this grows please adjust
604 * skbuff.h:skbuff->cb[xxx] size appropriately.
605 */
606 struct tcp_skb_cb {
607 union {
608 struct inet_skb_parm h4;
609 #if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
610 struct inet6_skb_parm h6;
611 #endif
612 } header; /* For incoming frames */
613 __u32 seq; /* Starting sequence number */
614 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
615 __u32 when; /* used to compute rtt's */
616 __u8 flags; /* TCP header flags. */
617
618 /* NOTE: These must match up to the flags byte in a
619 * real TCP header.
620 */
621 #define TCPCB_FLAG_FIN 0x01
622 #define TCPCB_FLAG_SYN 0x02
623 #define TCPCB_FLAG_RST 0x04
624 #define TCPCB_FLAG_PSH 0x08
625 #define TCPCB_FLAG_ACK 0x10
626 #define TCPCB_FLAG_URG 0x20
627 #define TCPCB_FLAG_ECE 0x40
628 #define TCPCB_FLAG_CWR 0x80
629
630 __u8 sacked; /* State flags for SACK/FACK. */
631 #define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
632 #define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
633 #define TCPCB_LOST 0x04 /* SKB is lost */
634 #define TCPCB_TAGBITS 0x07 /* All tag bits */
635
636 #define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
637 #define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
638
639 __u32 ack_seq; /* Sequence number ACK'd */
640 };
641
642 #define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
643
644 /* Due to TSO, an SKB can be composed of multiple actual
645 * packets. To keep these tracked properly, we use this.
646 */
647 static inline int tcp_skb_pcount(const struct sk_buff *skb)
648 {
649 return skb_shinfo(skb)->gso_segs;
650 }
651
652 /* This is valid iff tcp_skb_pcount() > 1. */
653 static inline int tcp_skb_mss(const struct sk_buff *skb)
654 {
655 return skb_shinfo(skb)->gso_size;
656 }
657
658 /* Events passed to congestion control interface */
659 enum tcp_ca_event {
660 CA_EVENT_TX_START, /* first transmit when no packets in flight */
661 CA_EVENT_CWND_RESTART, /* congestion window restart */
662 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
663 CA_EVENT_FRTO, /* fast recovery timeout */
664 CA_EVENT_LOSS, /* loss timeout */
665 CA_EVENT_FAST_ACK, /* in sequence ack */
666 CA_EVENT_SLOW_ACK, /* other ack */
667 };
668
669 /*
670 * Interface for adding new TCP congestion control handlers
671 */
672 #define TCP_CA_NAME_MAX 16
673 #define TCP_CA_MAX 128
674 #define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
675
676 #define TCP_CONG_NON_RESTRICTED 0x1
677 #define TCP_CONG_RTT_STAMP 0x2
678
679 struct tcp_congestion_ops {
680 struct list_head list;
681 unsigned long flags;
682
683 /* initialize private data (optional) */
684 void (*init)(struct sock *sk);
685 /* cleanup private data (optional) */
686 void (*release)(struct sock *sk);
687
688 /* return slow start threshold (required) */
689 u32 (*ssthresh)(struct sock *sk);
690 /* lower bound for congestion window (optional) */
691 u32 (*min_cwnd)(const struct sock *sk);
692 /* do new cwnd calculation (required) */
693 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
694 /* call before changing ca_state (optional) */
695 void (*set_state)(struct sock *sk, u8 new_state);
696 /* call when cwnd event occurs (optional) */
697 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
698 /* new value of cwnd after loss (optional) */
699 u32 (*undo_cwnd)(struct sock *sk);
700 /* hook for packet ack accounting (optional) */
701 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
702 /* get info for inet_diag (optional) */
703 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
704
705 char name[TCP_CA_NAME_MAX];
706 struct module *owner;
707 };
708
709 extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
710 extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
711
712 extern void tcp_init_congestion_control(struct sock *sk);
713 extern void tcp_cleanup_congestion_control(struct sock *sk);
714 extern int tcp_set_default_congestion_control(const char *name);
715 extern void tcp_get_default_congestion_control(char *name);
716 extern void tcp_get_available_congestion_control(char *buf, size_t len);
717 extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
718 extern int tcp_set_allowed_congestion_control(char *allowed);
719 extern int tcp_set_congestion_control(struct sock *sk, const char *name);
720 extern void tcp_slow_start(struct tcp_sock *tp);
721 extern void tcp_cong_avoid_ai(struct tcp_sock *tp, u32 w);
722
723 extern struct tcp_congestion_ops tcp_init_congestion_ops;
724 extern u32 tcp_reno_ssthresh(struct sock *sk);
725 extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
726 extern u32 tcp_reno_min_cwnd(const struct sock *sk);
727 extern struct tcp_congestion_ops tcp_reno;
728
729 static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
730 {
731 struct inet_connection_sock *icsk = inet_csk(sk);
732
733 if (icsk->icsk_ca_ops->set_state)
734 icsk->icsk_ca_ops->set_state(sk, ca_state);
735 icsk->icsk_ca_state = ca_state;
736 }
737
738 static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
739 {
740 const struct inet_connection_sock *icsk = inet_csk(sk);
741
742 if (icsk->icsk_ca_ops->cwnd_event)
743 icsk->icsk_ca_ops->cwnd_event(sk, event);
744 }
745
746 /* These functions determine how the current flow behaves in respect of SACK
747 * handling. SACK is negotiated with the peer, and therefore it can vary
748 * between different flows.
749 *
750 * tcp_is_sack - SACK enabled
751 * tcp_is_reno - No SACK
752 * tcp_is_fack - FACK enabled, implies SACK enabled
753 */
754 static inline int tcp_is_sack(const struct tcp_sock *tp)
755 {
756 return tp->rx_opt.sack_ok;
757 }
758
759 static inline int tcp_is_reno(const struct tcp_sock *tp)
760 {
761 return !tcp_is_sack(tp);
762 }
763
764 static inline int tcp_is_fack(const struct tcp_sock *tp)
765 {
766 return tp->rx_opt.sack_ok & 2;
767 }
768
769 static inline void tcp_enable_fack(struct tcp_sock *tp)
770 {
771 tp->rx_opt.sack_ok |= 2;
772 }
773
774 static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
775 {
776 return tp->sacked_out + tp->lost_out;
777 }
778
779 /* This determines how many packets are "in the network" to the best
780 * of our knowledge. In many cases it is conservative, but where
781 * detailed information is available from the receiver (via SACK
782 * blocks etc.) we can make more aggressive calculations.
783 *
784 * Use this for decisions involving congestion control, use just
785 * tp->packets_out to determine if the send queue is empty or not.
786 *
787 * Read this equation as:
788 *
789 * "Packets sent once on transmission queue" MINUS
790 * "Packets left network, but not honestly ACKed yet" PLUS
791 * "Packets fast retransmitted"
792 */
793 static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
794 {
795 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
796 }
797
798 #define TCP_INFINITE_SSTHRESH 0x7fffffff
799
800 static inline bool tcp_in_initial_slowstart(const struct tcp_sock *tp)
801 {
802 return tp->snd_ssthresh >= TCP_INFINITE_SSTHRESH;
803 }
804
805 /* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
806 * The exception is rate halving phase, when cwnd is decreasing towards
807 * ssthresh.
808 */
809 static inline __u32 tcp_current_ssthresh(const struct sock *sk)
810 {
811 const struct tcp_sock *tp = tcp_sk(sk);
812 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
813 return tp->snd_ssthresh;
814 else
815 return max(tp->snd_ssthresh,
816 ((tp->snd_cwnd >> 1) +
817 (tp->snd_cwnd >> 2)));
818 }
819
820 /* Use define here intentionally to get WARN_ON location shown at the caller */
821 #define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
822
823 extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
824 extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
825
826 /* Slow start with delack produces 3 packets of burst, so that
827 * it is safe "de facto". This will be the default - same as
828 * the default reordering threshold - but if reordering increases,
829 * we must be able to allow cwnd to burst at least this much in order
830 * to not pull it back when holes are filled.
831 */
832 static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
833 {
834 return tp->reordering;
835 }
836
837 /* Returns end sequence number of the receiver's advertised window */
838 static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
839 {
840 return tp->snd_una + tp->snd_wnd;
841 }
842 extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
843
844 static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
845 const struct sk_buff *skb)
846 {
847 if (skb->len < mss)
848 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
849 }
850
851 static inline void tcp_check_probe_timer(struct sock *sk)
852 {
853 struct tcp_sock *tp = tcp_sk(sk);
854 const struct inet_connection_sock *icsk = inet_csk(sk);
855
856 if (!tp->packets_out && !icsk->icsk_pending)
857 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
858 icsk->icsk_rto, TCP_RTO_MAX);
859 }
860
861 static inline void tcp_init_wl(struct tcp_sock *tp, u32 seq)
862 {
863 tp->snd_wl1 = seq;
864 }
865
866 static inline void tcp_update_wl(struct tcp_sock *tp, u32 seq)
867 {
868 tp->snd_wl1 = seq;
869 }
870
871 /*
872 * Calculate(/check) TCP checksum
873 */
874 static inline __sum16 tcp_v4_check(int len, __be32 saddr,
875 __be32 daddr, __wsum base)
876 {
877 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
878 }
879
880 static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
881 {
882 return __skb_checksum_complete(skb);
883 }
884
885 static inline int tcp_checksum_complete(struct sk_buff *skb)
886 {
887 return !skb_csum_unnecessary(skb) &&
888 __tcp_checksum_complete(skb);
889 }
890
891 /* Prequeue for VJ style copy to user, combined with checksumming. */
892
893 static inline void tcp_prequeue_init(struct tcp_sock *tp)
894 {
895 tp->ucopy.task = NULL;
896 tp->ucopy.len = 0;
897 tp->ucopy.memory = 0;
898 skb_queue_head_init(&tp->ucopy.prequeue);
899 #ifdef CONFIG_NET_DMA
900 tp->ucopy.dma_chan = NULL;
901 tp->ucopy.wakeup = 0;
902 tp->ucopy.pinned_list = NULL;
903 tp->ucopy.dma_cookie = 0;
904 #endif
905 }
906
907 /* Packet is added to VJ-style prequeue for processing in process
908 * context, if a reader task is waiting. Apparently, this exciting
909 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
910 * failed somewhere. Latency? Burstiness? Well, at least now we will
911 * see, why it failed. 8)8) --ANK
912 *
913 * NOTE: is this not too big to inline?
914 */
915 static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
916 {
917 struct tcp_sock *tp = tcp_sk(sk);
918
919 if (sysctl_tcp_low_latency || !tp->ucopy.task)
920 return 0;
921
922 __skb_queue_tail(&tp->ucopy.prequeue, skb);
923 tp->ucopy.memory += skb->truesize;
924 if (tp->ucopy.memory > sk->sk_rcvbuf) {
925 struct sk_buff *skb1;
926
927 BUG_ON(sock_owned_by_user(sk));
928
929 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
930 sk_backlog_rcv(sk, skb1);
931 NET_INC_STATS_BH(sock_net(sk),
932 LINUX_MIB_TCPPREQUEUEDROPPED);
933 }
934
935 tp->ucopy.memory = 0;
936 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
937 wake_up_interruptible_poll(sk->sk_sleep,
938 POLLIN | POLLRDNORM | POLLRDBAND);
939 if (!inet_csk_ack_scheduled(sk))
940 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
941 (3 * tcp_rto_min(sk)) / 4,
942 TCP_RTO_MAX);
943 }
944 return 1;
945 }
946
947
948 #undef STATE_TRACE
949
950 #ifdef STATE_TRACE
951 static const char *statename[]={
952 "Unused","Established","Syn Sent","Syn Recv",
953 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
954 "Close Wait","Last ACK","Listen","Closing"
955 };
956 #endif
957 extern void tcp_set_state(struct sock *sk, int state);
958
959 extern void tcp_done(struct sock *sk);
960
961 static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
962 {
963 rx_opt->dsack = 0;
964 rx_opt->num_sacks = 0;
965 }
966
967 /* Determine a window scaling and initial window to offer. */
968 extern void tcp_select_initial_window(int __space, __u32 mss,
969 __u32 *rcv_wnd, __u32 *window_clamp,
970 int wscale_ok, __u8 *rcv_wscale,
971 __u32 init_rcv_wnd);
972
973 static inline int tcp_win_from_space(int space)
974 {
975 return sysctl_tcp_adv_win_scale<=0 ?
976 (space>>(-sysctl_tcp_adv_win_scale)) :
977 space - (space>>sysctl_tcp_adv_win_scale);
978 }
979
980 /* Note: caller must be prepared to deal with negative returns */
981 static inline int tcp_space(const struct sock *sk)
982 {
983 return tcp_win_from_space(sk->sk_rcvbuf -
984 atomic_read(&sk->sk_rmem_alloc));
985 }
986
987 static inline int tcp_full_space(const struct sock *sk)
988 {
989 return tcp_win_from_space(sk->sk_rcvbuf);
990 }
991
992 static inline void tcp_openreq_init(struct request_sock *req,
993 struct tcp_options_received *rx_opt,
994 struct sk_buff *skb)
995 {
996 struct inet_request_sock *ireq = inet_rsk(req);
997
998 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
999 req->cookie_ts = 0;
1000 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1001 req->mss = rx_opt->mss_clamp;
1002 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
1003 ireq->tstamp_ok = rx_opt->tstamp_ok;
1004 ireq->sack_ok = rx_opt->sack_ok;
1005 ireq->snd_wscale = rx_opt->snd_wscale;
1006 ireq->wscale_ok = rx_opt->wscale_ok;
1007 ireq->acked = 0;
1008 ireq->ecn_ok = 0;
1009 ireq->rmt_port = tcp_hdr(skb)->source;
1010 ireq->loc_port = tcp_hdr(skb)->dest;
1011 }
1012
1013 extern void tcp_enter_memory_pressure(struct sock *sk);
1014
1015 static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1016 {
1017 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1018 }
1019
1020 static inline int keepalive_time_when(const struct tcp_sock *tp)
1021 {
1022 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1023 }
1024
1025 static inline int keepalive_probes(const struct tcp_sock *tp)
1026 {
1027 return tp->keepalive_probes ? : sysctl_tcp_keepalive_probes;
1028 }
1029
1030 static inline int tcp_fin_time(const struct sock *sk)
1031 {
1032 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1033 const int rto = inet_csk(sk)->icsk_rto;
1034
1035 if (fin_timeout < (rto << 2) - (rto >> 1))
1036 fin_timeout = (rto << 2) - (rto >> 1);
1037
1038 return fin_timeout;
1039 }
1040
1041 static inline int tcp_paws_check(const struct tcp_options_received *rx_opt,
1042 int paws_win)
1043 {
1044 if ((s32)(rx_opt->ts_recent - rx_opt->rcv_tsval) <= paws_win)
1045 return 1;
1046 if (unlikely(get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS))
1047 return 1;
1048
1049 return 0;
1050 }
1051
1052 static inline int tcp_paws_reject(const struct tcp_options_received *rx_opt,
1053 int rst)
1054 {
1055 if (tcp_paws_check(rx_opt, 0))
1056 return 0;
1057
1058 /* RST segments are not recommended to carry timestamp,
1059 and, if they do, it is recommended to ignore PAWS because
1060 "their cleanup function should take precedence over timestamps."
1061 Certainly, it is mistake. It is necessary to understand the reasons
1062 of this constraint to relax it: if peer reboots, clock may go
1063 out-of-sync and half-open connections will not be reset.
1064 Actually, the problem would be not existing if all
1065 the implementations followed draft about maintaining clock
1066 via reboots. Linux-2.2 DOES NOT!
1067
1068 However, we can relax time bounds for RST segments to MSL.
1069 */
1070 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1071 return 0;
1072 return 1;
1073 }
1074
1075 #define TCP_CHECK_TIMER(sk) do { } while (0)
1076
1077 static inline void tcp_mib_init(struct net *net)
1078 {
1079 /* See RFC 2012 */
1080 TCP_ADD_STATS_USER(net, TCP_MIB_RTOALGORITHM, 1);
1081 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1082 TCP_ADD_STATS_USER(net, TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1083 TCP_ADD_STATS_USER(net, TCP_MIB_MAXCONN, -1);
1084 }
1085
1086 /* from STCP */
1087 static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1088 {
1089 tp->lost_skb_hint = NULL;
1090 tp->scoreboard_skb_hint = NULL;
1091 }
1092
1093 static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1094 {
1095 tcp_clear_retrans_hints_partial(tp);
1096 tp->retransmit_skb_hint = NULL;
1097 }
1098
1099 /* MD5 Signature */
1100 struct crypto_hash;
1101
1102 /* - key database */
1103 struct tcp_md5sig_key {
1104 u8 *key;
1105 u8 keylen;
1106 };
1107
1108 struct tcp4_md5sig_key {
1109 struct tcp_md5sig_key base;
1110 __be32 addr;
1111 };
1112
1113 struct tcp6_md5sig_key {
1114 struct tcp_md5sig_key base;
1115 #if 0
1116 u32 scope_id; /* XXX */
1117 #endif
1118 struct in6_addr addr;
1119 };
1120
1121 /* - sock block */
1122 struct tcp_md5sig_info {
1123 struct tcp4_md5sig_key *keys4;
1124 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1125 struct tcp6_md5sig_key *keys6;
1126 u32 entries6;
1127 u32 alloced6;
1128 #endif
1129 u32 entries4;
1130 u32 alloced4;
1131 };
1132
1133 /* - pseudo header */
1134 struct tcp4_pseudohdr {
1135 __be32 saddr;
1136 __be32 daddr;
1137 __u8 pad;
1138 __u8 protocol;
1139 __be16 len;
1140 };
1141
1142 struct tcp6_pseudohdr {
1143 struct in6_addr saddr;
1144 struct in6_addr daddr;
1145 __be32 len;
1146 __be32 protocol; /* including padding */
1147 };
1148
1149 union tcp_md5sum_block {
1150 struct tcp4_pseudohdr ip4;
1151 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1152 struct tcp6_pseudohdr ip6;
1153 #endif
1154 };
1155
1156 /* - pool: digest algorithm, hash description and scratch buffer */
1157 struct tcp_md5sig_pool {
1158 struct hash_desc md5_desc;
1159 union tcp_md5sum_block md5_blk;
1160 };
1161
1162 #define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1163
1164 /* - functions */
1165 extern int tcp_v4_md5_hash_skb(char *md5_hash,
1166 struct tcp_md5sig_key *key,
1167 struct sock *sk,
1168 struct request_sock *req,
1169 struct sk_buff *skb);
1170
1171 extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1172 struct sock *addr_sk);
1173
1174 extern int tcp_v4_md5_do_add(struct sock *sk,
1175 __be32 addr,
1176 u8 *newkey,
1177 u8 newkeylen);
1178
1179 extern int tcp_v4_md5_do_del(struct sock *sk,
1180 __be32 addr);
1181
1182 #ifdef CONFIG_TCP_MD5SIG
1183 #define tcp_twsk_md5_key(twsk) ((twsk)->tw_md5_keylen ? \
1184 &(struct tcp_md5sig_key) { \
1185 .key = (twsk)->tw_md5_key, \
1186 .keylen = (twsk)->tw_md5_keylen, \
1187 } : NULL)
1188 #else
1189 #define tcp_twsk_md5_key(twsk) NULL
1190 #endif
1191
1192 extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(struct sock *);
1193 extern void tcp_free_md5sig_pool(void);
1194
1195 extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1196 extern void __tcp_put_md5sig_pool(void);
1197 extern int tcp_md5_hash_header(struct tcp_md5sig_pool *, struct tcphdr *);
1198 extern int tcp_md5_hash_skb_data(struct tcp_md5sig_pool *, struct sk_buff *,
1199 unsigned header_len);
1200 extern int tcp_md5_hash_key(struct tcp_md5sig_pool *hp,
1201 struct tcp_md5sig_key *key);
1202
1203 static inline
1204 struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1205 {
1206 int cpu = get_cpu();
1207 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1208 if (!ret)
1209 put_cpu();
1210 return ret;
1211 }
1212
1213 static inline void tcp_put_md5sig_pool(void)
1214 {
1215 __tcp_put_md5sig_pool();
1216 put_cpu();
1217 }
1218
1219 /* write queue abstraction */
1220 static inline void tcp_write_queue_purge(struct sock *sk)
1221 {
1222 struct sk_buff *skb;
1223
1224 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1225 sk_wmem_free_skb(sk, skb);
1226 sk_mem_reclaim(sk);
1227 tcp_clear_all_retrans_hints(tcp_sk(sk));
1228 }
1229
1230 static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1231 {
1232 return skb_peek(&sk->sk_write_queue);
1233 }
1234
1235 static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1236 {
1237 return skb_peek_tail(&sk->sk_write_queue);
1238 }
1239
1240 static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1241 {
1242 return skb_queue_next(&sk->sk_write_queue, skb);
1243 }
1244
1245 static inline struct sk_buff *tcp_write_queue_prev(struct sock *sk, struct sk_buff *skb)
1246 {
1247 return skb_queue_prev(&sk->sk_write_queue, skb);
1248 }
1249
1250 #define tcp_for_write_queue(skb, sk) \
1251 skb_queue_walk(&(sk)->sk_write_queue, skb)
1252
1253 #define tcp_for_write_queue_from(skb, sk) \
1254 skb_queue_walk_from(&(sk)->sk_write_queue, skb)
1255
1256 #define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1257 skb_queue_walk_from_safe(&(sk)->sk_write_queue, skb, tmp)
1258
1259 static inline struct sk_buff *tcp_send_head(struct sock *sk)
1260 {
1261 return sk->sk_send_head;
1262 }
1263
1264 static inline bool tcp_skb_is_last(const struct sock *sk,
1265 const struct sk_buff *skb)
1266 {
1267 return skb_queue_is_last(&sk->sk_write_queue, skb);
1268 }
1269
1270 static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1271 {
1272 if (tcp_skb_is_last(sk, skb))
1273 sk->sk_send_head = NULL;
1274 else
1275 sk->sk_send_head = tcp_write_queue_next(sk, skb);
1276 }
1277
1278 static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1279 {
1280 if (sk->sk_send_head == skb_unlinked)
1281 sk->sk_send_head = NULL;
1282 }
1283
1284 static inline void tcp_init_send_head(struct sock *sk)
1285 {
1286 sk->sk_send_head = NULL;
1287 }
1288
1289 static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1290 {
1291 __skb_queue_tail(&sk->sk_write_queue, skb);
1292 }
1293
1294 static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1295 {
1296 __tcp_add_write_queue_tail(sk, skb);
1297
1298 /* Queue it, remembering where we must start sending. */
1299 if (sk->sk_send_head == NULL) {
1300 sk->sk_send_head = skb;
1301
1302 if (tcp_sk(sk)->highest_sack == NULL)
1303 tcp_sk(sk)->highest_sack = skb;
1304 }
1305 }
1306
1307 static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1308 {
1309 __skb_queue_head(&sk->sk_write_queue, skb);
1310 }
1311
1312 /* Insert buff after skb on the write queue of sk. */
1313 static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1314 struct sk_buff *buff,
1315 struct sock *sk)
1316 {
1317 __skb_queue_after(&sk->sk_write_queue, skb, buff);
1318 }
1319
1320 /* Insert new before skb on the write queue of sk. */
1321 static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1322 struct sk_buff *skb,
1323 struct sock *sk)
1324 {
1325 __skb_queue_before(&sk->sk_write_queue, skb, new);
1326
1327 if (sk->sk_send_head == skb)
1328 sk->sk_send_head = new;
1329 }
1330
1331 static inline void tcp_unlink_write_queue(struct sk_buff *skb, struct sock *sk)
1332 {
1333 __skb_unlink(skb, &sk->sk_write_queue);
1334 }
1335
1336 static inline int tcp_write_queue_empty(struct sock *sk)
1337 {
1338 return skb_queue_empty(&sk->sk_write_queue);
1339 }
1340
1341 static inline void tcp_push_pending_frames(struct sock *sk)
1342 {
1343 if (tcp_send_head(sk)) {
1344 struct tcp_sock *tp = tcp_sk(sk);
1345
1346 __tcp_push_pending_frames(sk, tcp_current_mss(sk), tp->nonagle);
1347 }
1348 }
1349
1350 /* Start sequence of the highest skb with SACKed bit, valid only if
1351 * sacked > 0 or when the caller has ensured validity by itself.
1352 */
1353 static inline u32 tcp_highest_sack_seq(struct tcp_sock *tp)
1354 {
1355 if (!tp->sacked_out)
1356 return tp->snd_una;
1357
1358 if (tp->highest_sack == NULL)
1359 return tp->snd_nxt;
1360
1361 return TCP_SKB_CB(tp->highest_sack)->seq;
1362 }
1363
1364 static inline void tcp_advance_highest_sack(struct sock *sk, struct sk_buff *skb)
1365 {
1366 tcp_sk(sk)->highest_sack = tcp_skb_is_last(sk, skb) ? NULL :
1367 tcp_write_queue_next(sk, skb);
1368 }
1369
1370 static inline struct sk_buff *tcp_highest_sack(struct sock *sk)
1371 {
1372 return tcp_sk(sk)->highest_sack;
1373 }
1374
1375 static inline void tcp_highest_sack_reset(struct sock *sk)
1376 {
1377 tcp_sk(sk)->highest_sack = tcp_write_queue_head(sk);
1378 }
1379
1380 /* Called when old skb is about to be deleted (to be combined with new skb) */
1381 static inline void tcp_highest_sack_combine(struct sock *sk,
1382 struct sk_buff *old,
1383 struct sk_buff *new)
1384 {
1385 if (tcp_sk(sk)->sacked_out && (old == tcp_sk(sk)->highest_sack))
1386 tcp_sk(sk)->highest_sack = new;
1387 }
1388
1389 /* /proc */
1390 enum tcp_seq_states {
1391 TCP_SEQ_STATE_LISTENING,
1392 TCP_SEQ_STATE_OPENREQ,
1393 TCP_SEQ_STATE_ESTABLISHED,
1394 TCP_SEQ_STATE_TIME_WAIT,
1395 };
1396
1397 struct tcp_seq_afinfo {
1398 char *name;
1399 sa_family_t family;
1400 struct file_operations seq_fops;
1401 struct seq_operations seq_ops;
1402 };
1403
1404 struct tcp_iter_state {
1405 struct seq_net_private p;
1406 sa_family_t family;
1407 enum tcp_seq_states state;
1408 struct sock *syn_wait_sk;
1409 int bucket, sbucket, num, uid;
1410 };
1411
1412 extern int tcp_proc_register(struct net *net, struct tcp_seq_afinfo *afinfo);
1413 extern void tcp_proc_unregister(struct net *net, struct tcp_seq_afinfo *afinfo);
1414
1415 extern struct request_sock_ops tcp_request_sock_ops;
1416 extern struct request_sock_ops tcp6_request_sock_ops;
1417
1418 extern void tcp_v4_destroy_sock(struct sock *sk);
1419
1420 extern int tcp_v4_gso_send_check(struct sk_buff *skb);
1421 extern struct sk_buff *tcp_tso_segment(struct sk_buff *skb, int features);
1422 extern struct sk_buff **tcp_gro_receive(struct sk_buff **head,
1423 struct sk_buff *skb);
1424 extern struct sk_buff **tcp4_gro_receive(struct sk_buff **head,
1425 struct sk_buff *skb);
1426 extern int tcp_gro_complete(struct sk_buff *skb);
1427 extern int tcp4_gro_complete(struct sk_buff *skb);
1428
1429 #ifdef CONFIG_PROC_FS
1430 extern int tcp4_proc_init(void);
1431 extern void tcp4_proc_exit(void);
1432 #endif
1433
1434 /* TCP af-specific functions */
1435 struct tcp_sock_af_ops {
1436 #ifdef CONFIG_TCP_MD5SIG
1437 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1438 struct sock *addr_sk);
1439 int (*calc_md5_hash) (char *location,
1440 struct tcp_md5sig_key *md5,
1441 struct sock *sk,
1442 struct request_sock *req,
1443 struct sk_buff *skb);
1444 int (*md5_add) (struct sock *sk,
1445 struct sock *addr_sk,
1446 u8 *newkey,
1447 u8 len);
1448 int (*md5_parse) (struct sock *sk,
1449 char __user *optval,
1450 int optlen);
1451 #endif
1452 };
1453
1454 struct tcp_request_sock_ops {
1455 #ifdef CONFIG_TCP_MD5SIG
1456 struct tcp_md5sig_key *(*md5_lookup) (struct sock *sk,
1457 struct request_sock *req);
1458 int (*calc_md5_hash) (char *location,
1459 struct tcp_md5sig_key *md5,
1460 struct sock *sk,
1461 struct request_sock *req,
1462 struct sk_buff *skb);
1463 #endif
1464 };
1465
1466 /* Using SHA1 for now, define some constants.
1467 */
1468 #define COOKIE_DIGEST_WORDS (SHA_DIGEST_WORDS)
1469 #define COOKIE_MESSAGE_WORDS (SHA_MESSAGE_BYTES / 4)
1470 #define COOKIE_WORKSPACE_WORDS (COOKIE_DIGEST_WORDS + COOKIE_MESSAGE_WORDS)
1471
1472 extern int tcp_cookie_generator(u32 *bakery);
1473
1474 /**
1475 * struct tcp_cookie_values - each socket needs extra space for the
1476 * cookies, together with (optional) space for any SYN data.
1477 *
1478 * A tcp_sock contains a pointer to the current value, and this is
1479 * cloned to the tcp_timewait_sock.
1480 *
1481 * @cookie_pair: variable data from the option exchange.
1482 *
1483 * @cookie_desired: user specified tcpct_cookie_desired. Zero
1484 * indicates default (sysctl_tcp_cookie_size).
1485 * After cookie sent, remembers size of cookie.
1486 * Range 0, TCP_COOKIE_MIN to TCP_COOKIE_MAX.
1487 *
1488 * @s_data_desired: user specified tcpct_s_data_desired. When the
1489 * constant payload is specified (@s_data_constant),
1490 * holds its length instead.
1491 * Range 0 to TCP_MSS_DESIRED.
1492 *
1493 * @s_data_payload: constant data that is to be included in the
1494 * payload of SYN or SYNACK segments when the
1495 * cookie option is present.
1496 */
1497 struct tcp_cookie_values {
1498 struct kref kref;
1499 u8 cookie_pair[TCP_COOKIE_PAIR_SIZE];
1500 u8 cookie_pair_size;
1501 u8 cookie_desired;
1502 u16 s_data_desired:11,
1503 s_data_constant:1,
1504 s_data_in:1,
1505 s_data_out:1,
1506 s_data_unused:2;
1507 u8 s_data_payload[0];
1508 };
1509
1510 static inline void tcp_cookie_values_release(struct kref *kref)
1511 {
1512 kfree(container_of(kref, struct tcp_cookie_values, kref));
1513 }
1514
1515 /* The length of constant payload data. Note that s_data_desired is
1516 * overloaded, depending on s_data_constant: either the length of constant
1517 * data (returned here) or the limit on variable data.
1518 */
1519 static inline int tcp_s_data_size(const struct tcp_sock *tp)
1520 {
1521 return (tp->cookie_values != NULL && tp->cookie_values->s_data_constant)
1522 ? tp->cookie_values->s_data_desired
1523 : 0;
1524 }
1525
1526 /**
1527 * struct tcp_extend_values - tcp_ipv?.c to tcp_output.c workspace.
1528 *
1529 * As tcp_request_sock has already been extended in other places, the
1530 * only remaining method is to pass stack values along as function
1531 * parameters. These parameters are not needed after sending SYNACK.
1532 *
1533 * @cookie_bakery: cryptographic secret and message workspace.
1534 *
1535 * @cookie_plus: bytes in authenticator/cookie option, copied from
1536 * struct tcp_options_received (above).
1537 */
1538 struct tcp_extend_values {
1539 struct request_values rv;
1540 u32 cookie_bakery[COOKIE_WORKSPACE_WORDS];
1541 u8 cookie_plus:6,
1542 cookie_out_never:1,
1543 cookie_in_always:1;
1544 };
1545
1546 static inline struct tcp_extend_values *tcp_xv(struct request_values *rvp)
1547 {
1548 return (struct tcp_extend_values *)rvp;
1549 }
1550
1551 extern void tcp_v4_init(void);
1552 extern void tcp_init(void);
1553
1554 #endif /* _TCP_H */
Linux kernel - Deliverables
This page took 0.083312 seconds and 5 git commands to generate.