projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[TCP]: Cleanup local variables of clean_rtx_queue
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / include / net / tcp.h
CommitLineData
1da177e4
LT		 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 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 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
1da177e4
LT		 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>
fb286bb2 29#include <linux/skbuff.h>
97fc2f08 30#include <linux/dmaengine.h>
cfb6eeb4 31#include <linux/crypto.h>
3f421baa
ACM		 32
33#include <net/inet_connection_sock.h>
295ff7ed 34#include <net/inet_timewait_sock.h>
77d8bf9c 35#include <net/inet_hashtables.h>
1da177e4 36#include <net/checksum.h>
2e6599cb 37#include <net/request_sock.h>
1da177e4
LT		 38#include <net/sock.h>
39#include <net/snmp.h>
40#include <net/ip.h>
c752f073 41#include <net/tcp_states.h>
bdf1ee5d 42#include <net/inet_ecn.h>
c752f073 43
1da177e4
LT		 44#include <linux/seq_file.h>
45
6e04e021 46extern struct inet_hashinfo tcp_hashinfo;
1da177e4 47
1da177e4 48extern atomic_t tcp_orphan_count;
1da177e4 49extern void tcp_time_wait(struct sock *sk, int state, int timeo);
1da177e4 50
1da177e4
LT		 51#define MAX_TCP_HEADER (128 + MAX_HEADER)
52
53/*
54 * Never offer a window over 32767 without using window scaling. Some
55 * poor stacks do signed 16bit maths!
56 */
57#define MAX_TCP_WINDOW 32767U
58
59/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
60#define TCP_MIN_MSS 88U
61
62/* Minimal RCV_MSS. */
63#define TCP_MIN_RCVMSS 536U
64
5d424d5a
JH		 65/* The least MTU to use for probing */
66#define TCP_BASE_MSS 512
67
1da177e4
LT		 68/* After receiving this amount of duplicate ACKs fast retransmit starts. */
69#define TCP_FASTRETRANS_THRESH 3
70
71/* Maximal reordering. */
72#define TCP_MAX_REORDERING 127
73
74/* Maximal number of ACKs sent quickly to accelerate slow-start. */
75#define TCP_MAX_QUICKACKS 16U
76
77/* urg_data states */
78#define TCP_URG_VALID 0x0100
79#define TCP_URG_NOTYET 0x0200
80#define TCP_URG_READ 0x0400
81
82#define TCP_RETR1 3 /*
83 * This is how many retries it does before it
84 * tries to figure out if the gateway is
85 * down. Minimal RFC value is 3; it corresponds
86 * to ~3sec-8min depending on RTO.
87 */
88
89#define TCP_RETR2 15 /*
90 * This should take at least
91 * 90 minutes to time out.
92 * RFC1122 says that the limit is 100 sec.
93 * 15 is ~13-30min depending on RTO.
94 */
95
96#define TCP_SYN_RETRIES 5 /* number of times to retry active opening a
caa20d9a 97 * connection: ~180sec is RFC minimum */
1da177e4
LT		 98
99#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
caa20d9a 100 * connection: ~180sec is RFC minimum */
1da177e4
LT		 101
102
103#define TCP_ORPHAN_RETRIES 7 /* number of times to retry on an orphaned
104 * socket. 7 is ~50sec-16min.
105 */
106
107
108#define TCP_TIMEWAIT_LEN (60*HZ) /* how long to wait to destroy TIME-WAIT
109 * state, about 60 seconds */
110#define TCP_FIN_TIMEOUT TCP_TIMEWAIT_LEN
111 /* BSD style FIN_WAIT2 deadlock breaker.
112 * It used to be 3min, new value is 60sec,
113 * to combine FIN-WAIT-2 timeout with
114 * TIME-WAIT timer.
115 */
116
117#define TCP_DELACK_MAX ((unsigned)(HZ/5)) /* maximal time to delay before sending an ACK */
118#if HZ >= 100
119#define TCP_DELACK_MIN ((unsigned)(HZ/25)) /* minimal time to delay before sending an ACK */
120#define TCP_ATO_MIN ((unsigned)(HZ/25))
121#else
122#define TCP_DELACK_MIN 4U
123#define TCP_ATO_MIN 4U
124#endif
125#define TCP_RTO_MAX ((unsigned)(120*HZ))
126#define TCP_RTO_MIN ((unsigned)(HZ/5))
127#define TCP_TIMEOUT_INIT ((unsigned)(3*HZ)) /* RFC 1122 initial RTO value */
128
129#define TCP_RESOURCE_PROBE_INTERVAL ((unsigned)(HZ/2U)) /* Maximal interval between probes
130 * for local resources.
131 */
132
133#define TCP_KEEPALIVE_TIME (120*60*HZ) /* two hours */
134#define TCP_KEEPALIVE_PROBES 9 /* Max of 9 keepalive probes */
135#define TCP_KEEPALIVE_INTVL (75*HZ)
136
137#define MAX_TCP_KEEPIDLE 32767
138#define MAX_TCP_KEEPINTVL 32767
139#define MAX_TCP_KEEPCNT 127
140#define MAX_TCP_SYNCNT 127
141
142#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
1da177e4
LT		 143
144#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
145#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
146 * after this time. It should be equal
147 * (or greater than) TCP_TIMEWAIT_LEN
148 * to provide reliability equal to one
149 * provided by timewait state.
150 */
151#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
152 * timestamps. It must be less than
153 * minimal timewait lifetime.
154 */
1da177e4
LT		 155/*
156 * TCP option
157 */
158
159#define TCPOPT_NOP 1 /* Padding */
160#define TCPOPT_EOL 0 /* End of options */
161#define TCPOPT_MSS 2 /* Segment size negotiating */
162#define TCPOPT_WINDOW 3 /* Window scaling */
163#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
164#define TCPOPT_SACK 5 /* SACK Block */
165#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
cfb6eeb4 166#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
1da177e4
LT		 167
168/*
169 * TCP option lengths
170 */
171
172#define TCPOLEN_MSS 4
173#define TCPOLEN_WINDOW 3
174#define TCPOLEN_SACK_PERM 2
175#define TCPOLEN_TIMESTAMP 10
cfb6eeb4 176#define TCPOLEN_MD5SIG 18
1da177e4
LT		 177
178/* But this is what stacks really send out. */
179#define TCPOLEN_TSTAMP_ALIGNED 12
180#define TCPOLEN_WSCALE_ALIGNED 4
181#define TCPOLEN_SACKPERM_ALIGNED 4
182#define TCPOLEN_SACK_BASE 2
183#define TCPOLEN_SACK_BASE_ALIGNED 4
184#define TCPOLEN_SACK_PERBLOCK 8
cfb6eeb4 185#define TCPOLEN_MD5SIG_ALIGNED 20
1da177e4 186
1da177e4
LT		 187/* Flags in tp->nonagle */
188#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
189#define TCP_NAGLE_CORK 2 /* Socket is corked */
caa20d9a 190#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
1da177e4 191
295ff7ed
ACM		 192extern struct inet_timewait_death_row tcp_death_row;
193
1da177e4 194/* sysctl variables for tcp */
1da177e4
LT		 195extern int sysctl_tcp_timestamps;
196extern int sysctl_tcp_window_scaling;
197extern int sysctl_tcp_sack;
198extern int sysctl_tcp_fin_timeout;
1da177e4
LT		 199extern int sysctl_tcp_keepalive_time;
200extern int sysctl_tcp_keepalive_probes;
201extern int sysctl_tcp_keepalive_intvl;
202extern int sysctl_tcp_syn_retries;
203extern int sysctl_tcp_synack_retries;
204extern int sysctl_tcp_retries1;
205extern int sysctl_tcp_retries2;
206extern int sysctl_tcp_orphan_retries;
207extern int sysctl_tcp_syncookies;
208extern int sysctl_tcp_retrans_collapse;
209extern int sysctl_tcp_stdurg;
210extern int sysctl_tcp_rfc1337;
211extern int sysctl_tcp_abort_on_overflow;
212extern int sysctl_tcp_max_orphans;
1da177e4
LT		 213extern int sysctl_tcp_fack;
214extern int sysctl_tcp_reordering;
215extern int sysctl_tcp_ecn;
216extern int sysctl_tcp_dsack;
217extern int sysctl_tcp_mem[3];
218extern int sysctl_tcp_wmem[3];
219extern int sysctl_tcp_rmem[3];
220extern int sysctl_tcp_app_win;
221extern int sysctl_tcp_adv_win_scale;
222extern int sysctl_tcp_tw_reuse;
223extern int sysctl_tcp_frto;
3cfe3baa 224extern int sysctl_tcp_frto_response;
1da177e4 225extern int sysctl_tcp_low_latency;
95937825 226extern int sysctl_tcp_dma_copybreak;
1da177e4 227extern int sysctl_tcp_nometrics_save;
1da177e4
LT		 228extern int sysctl_tcp_moderate_rcvbuf;
229extern int sysctl_tcp_tso_win_divisor;
9772efb9 230extern int sysctl_tcp_abc;
5d424d5a
JH		 231extern int sysctl_tcp_mtu_probing;
232extern int sysctl_tcp_base_mss;
15d99e02 233extern int sysctl_tcp_workaround_signed_windows;
35089bb2 234extern int sysctl_tcp_slow_start_after_idle;
886236c1 235extern int sysctl_tcp_max_ssthresh;
1da177e4
LT		 236
237extern atomic_t tcp_memory_allocated;
238extern atomic_t tcp_sockets_allocated;
239extern int tcp_memory_pressure;
240
1da177e4
LT		 241/*
242 * The next routines deal with comparing 32 bit unsigned ints
243 * and worry about wraparound (automatic with unsigned arithmetic).
244 */
245
246static inline int before(__u32 seq1, __u32 seq2)
247{
0d630cc0 248 return (__s32)(seq1-seq2) < 0;
1da177e4 249}
9a036b9c 250#define after(seq2, seq1) before(seq1, seq2)
1da177e4
LT		 251
252/* is s2<=s1<=s3 ? */
253static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
254{
255 return seq3 - seq2 >= seq1 - seq2;
256}
257
e4fd5da3
PE		 258static inline int tcp_too_many_orphans(struct sock *sk, int num)
259{
260 return (num > sysctl_tcp_max_orphans) ||
261 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
262 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
263}
1da177e4
LT		 264
265extern struct proto tcp_prot;
266
267DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
268#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
269#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
270#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
271#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
272#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
273#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
274
1da177e4
LT		 275extern void tcp_v4_err(struct sk_buff *skb, u32);
276
277extern void tcp_shutdown (struct sock *sk, int how);
278
279extern int tcp_v4_rcv(struct sk_buff *skb);
280
281extern int tcp_v4_remember_stamp(struct sock *sk);
282
8feaf0c0 283extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
1da177e4 284
3516ffb0 285extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
1da177e4
LT		 286 struct msghdr *msg, size_t size);
287extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
288
289extern int tcp_ioctl(struct sock *sk,
290 int cmd,
291 unsigned long arg);
292
293extern int tcp_rcv_state_process(struct sock *sk,
294 struct sk_buff *skb,
295 struct tcphdr *th,
296 unsigned len);
297
298extern int tcp_rcv_established(struct sock *sk,
299 struct sk_buff *skb,
300 struct tcphdr *th,
301 unsigned len);
302
303extern void tcp_rcv_space_adjust(struct sock *sk);
304
0e4b4992
CL		 305extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
306
6d6ee43e
ACM		 307extern int tcp_twsk_unique(struct sock *sk,
308 struct sock *sktw, void *twp);
309
cfb6eeb4
YH		 310extern void tcp_twsk_destructor(struct sock *sk);
311
9c55e01c
JA		 312extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
313 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
314
463c84b9
ACM		 315static inline void tcp_dec_quickack_mode(struct sock *sk,
316 const unsigned int pkts)
1da177e4 317{
463c84b9 318 struct inet_connection_sock *icsk = inet_csk(sk);
fc6415bc 319
463c84b9
ACM		 320 if (icsk->icsk_ack.quick) {
321 if (pkts >= icsk->icsk_ack.quick) {
322 icsk->icsk_ack.quick = 0;
fc6415bc 323 /* Leaving quickack mode we deflate ATO. */
463c84b9 324 icsk->icsk_ack.ato = TCP_ATO_MIN;
fc6415bc 325 } else
463c84b9 326 icsk->icsk_ack.quick -= pkts;
1da177e4
LT		 327 }
328}
329
463c84b9 330extern void tcp_enter_quickack_mode(struct sock *sk);
1da177e4 331
1da177e4
LT		 332static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
333{
334 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
335}
336
bdf1ee5d
IJ		 337#define TCP_ECN_OK 1
338#define TCP_ECN_QUEUE_CWR 2
339#define TCP_ECN_DEMAND_CWR 4
340
341static __inline__ void
342TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
343{
344 if (sysctl_tcp_ecn && th->ece && th->cwr)
345 inet_rsk(req)->ecn_ok = 1;
346}
347
1da177e4
LT		 348enum tcp_tw_status
349{
350 TCP_TW_SUCCESS = 0,
351 TCP_TW_RST = 1,
352 TCP_TW_ACK = 2,
353 TCP_TW_SYN = 3
354};
355
356
8feaf0c0 357extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
1da177e4 358 struct sk_buff *skb,
8feaf0c0 359 const struct tcphdr *th);
1da177e4
LT		 360
361extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
60236fdd
ACM		 362 struct request_sock *req,
363 struct request_sock **prev);
1da177e4
LT		 364extern int tcp_child_process(struct sock *parent,
365 struct sock *child,
366 struct sk_buff *skb);
46d0de4e 367extern int tcp_use_frto(struct sock *sk);
1da177e4
LT		 368extern void tcp_enter_frto(struct sock *sk);
369extern void tcp_enter_loss(struct sock *sk, int how);
370extern void tcp_clear_retrans(struct tcp_sock *tp);
371extern void tcp_update_metrics(struct sock *sk);
372
373extern void tcp_close(struct sock *sk,
374 long timeout);
1da177e4
LT		 375extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
376
377extern int tcp_getsockopt(struct sock *sk, int level,
378 int optname,
379 char __user *optval,
380 int __user *optlen);
381extern int tcp_setsockopt(struct sock *sk, int level,
382 int optname, char __user *optval,
383 int optlen);
3fdadf7d
DM		 384extern int compat_tcp_getsockopt(struct sock *sk,
385 int level, int optname,
386 char __user *optval, int __user *optlen);
387extern int compat_tcp_setsockopt(struct sock *sk,
388 int level, int optname,
389 char __user *optval, int optlen);
1da177e4
LT		 390extern void tcp_set_keepalive(struct sock *sk, int val);
391extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
392 struct msghdr *msg,
393 size_t len, int nonblock,
394 int flags, int *addr_len);
395
1da177e4
LT		 396extern void tcp_parse_options(struct sk_buff *skb,
397 struct tcp_options_received *opt_rx,
398 int estab);
399
400/*
401 * TCP v4 functions exported for the inet6 API
402 */
403
8292a17a 404extern void tcp_v4_send_check(struct sock *sk, int len,
1da177e4
LT		 405 struct sk_buff *skb);
406
407extern int tcp_v4_conn_request(struct sock *sk,
408 struct sk_buff *skb);
409
410extern struct sock * tcp_create_openreq_child(struct sock *sk,
60236fdd 411 struct request_sock *req,
1da177e4
LT		 412 struct sk_buff *skb);
413
414extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
415 struct sk_buff *skb,
60236fdd 416 struct request_sock *req,
1da177e4
LT		 417 struct dst_entry *dst);
418
419extern int tcp_v4_do_rcv(struct sock *sk,
420 struct sk_buff *skb);
421
422extern int tcp_v4_connect(struct sock *sk,
423 struct sockaddr *uaddr,
424 int addr_len);
425
426extern int tcp_connect(struct sock *sk);
427
428extern struct sk_buff * tcp_make_synack(struct sock *sk,
429 struct dst_entry *dst,
60236fdd 430 struct request_sock *req);
1da177e4
LT		 431
432extern int tcp_disconnect(struct sock *sk, int flags);
433
434extern void tcp_unhash(struct sock *sk);
435
1da177e4
LT		 436/* From syncookies.c */
437extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
438 struct ip_options *opt);
439extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
440 __u16 *mss);
441
442/* tcp_output.c */
443
9e412ba7
IJ		 444extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
445 int nonagle);
446extern int tcp_may_send_now(struct sock *sk);
1da177e4
LT		 447extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
448extern void tcp_xmit_retransmit_queue(struct sock *);
449extern void tcp_simple_retransmit(struct sock *);
450extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
6475be16 451extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
1da177e4
LT		 452
453extern void tcp_send_probe0(struct sock *);
454extern void tcp_send_partial(struct sock *);
455extern int tcp_write_wakeup(struct sock *);
456extern void tcp_send_fin(struct sock *sk);
dd0fc66f 457extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
1da177e4 458extern int tcp_send_synack(struct sock *);
c1b4a7e6 459extern void tcp_push_one(struct sock *, unsigned int mss_now);
1da177e4
LT		 460extern void tcp_send_ack(struct sock *sk);
461extern void tcp_send_delayed_ack(struct sock *sk);
462
a762a980
DM		 463/* tcp_input.c */
464extern void tcp_cwnd_application_limited(struct sock *sk);
465
1da177e4
LT		 466/* tcp_timer.c */
467extern void tcp_init_xmit_timers(struct sock *);
463c84b9
ACM		 468static inline void tcp_clear_xmit_timers(struct sock *sk)
469{
470 inet_csk_clear_xmit_timers(sk);
471}
1da177e4 472
1da177e4
LT		 473extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
474extern unsigned int tcp_current_mss(struct sock *sk, int large);
475
17b085ea 476/* tcp.c */
1da177e4
LT		 477extern void tcp_get_info(struct sock *, struct tcp_info *);
478
479/* Read 'sendfile()'-style from a TCP socket */
480typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
481 unsigned int, size_t);
482extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
483 sk_read_actor_t recv_actor);
484
40efc6fa 485extern void tcp_initialize_rcv_mss(struct sock *sk);
1da177e4 486
5d424d5a
JH		 487extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
488extern int tcp_mss_to_mtu(struct sock *sk, int mss);
489extern void tcp_mtup_init(struct sock *sk);
490
40efc6fa 491static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
1da177e4
LT		 492{
493 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
494 ntohl(TCP_FLAG_ACK) |
495 snd_wnd);
496}
497
40efc6fa 498static inline void tcp_fast_path_on(struct tcp_sock *tp)
1da177e4
LT		 499{
500 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
501}
502
9e412ba7 503static inline void tcp_fast_path_check(struct sock *sk)
1da177e4 504{
9e412ba7
IJ		 505 struct tcp_sock *tp = tcp_sk(sk);
506
b03efcfb 507 if (skb_queue_empty(&tp->out_of_order_queue) &&
1da177e4
LT		 508 tp->rcv_wnd &&
509 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
510 !tp->urg_data)
511 tcp_fast_path_on(tp);
512}
513
514/* Compute the actual receive window we are currently advertising.
515 * Rcv_nxt can be after the window if our peer push more data
516 * than the offered window.
517 */
40efc6fa 518static inline u32 tcp_receive_window(const struct tcp_sock *tp)
1da177e4
LT		 519{
520 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
521
522 if (win < 0)
523 win = 0;
524 return (u32) win;
525}
526
527/* Choose a new window, without checks for shrinking, and without
528 * scaling applied to the result. The caller does these things
529 * if necessary. This is a "raw" window selection.
530 */
531extern u32 __tcp_select_window(struct sock *sk);
532
533/* TCP timestamps are only 32-bits, this causes a slight
534 * complication on 64-bit systems since we store a snapshot
31f34269
SH		 535 * of jiffies in the buffer control blocks below. We decided
536 * to use only the low 32-bits of jiffies and hide the ugly
1da177e4
LT		 537 * casts with the following macro.
538 */
539#define tcp_time_stamp ((__u32)(jiffies))
540
caa20d9a 541/* This is what the send packet queuing engine uses to pass
1da177e4
LT		 542 * TCP per-packet control information to the transmission
543 * code. We also store the host-order sequence numbers in
544 * here too. This is 36 bytes on 32-bit architectures,
545 * 40 bytes on 64-bit machines, if this grows please adjust
546 * skbuff.h:skbuff->cb[xxx] size appropriately.
547 */
548struct tcp_skb_cb {
549 union {
550 struct inet_skb_parm h4;
551#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
552 struct inet6_skb_parm h6;
553#endif
554 } header; /* For incoming frames */
555 __u32 seq; /* Starting sequence number */
556 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
557 __u32 when; /* used to compute rtt's */
558 __u8 flags; /* TCP header flags. */
559
560 /* NOTE: These must match up to the flags byte in a
561 * real TCP header.
562 */
563#define TCPCB_FLAG_FIN 0x01
564#define TCPCB_FLAG_SYN 0x02
565#define TCPCB_FLAG_RST 0x04
566#define TCPCB_FLAG_PSH 0x08
567#define TCPCB_FLAG_ACK 0x10
568#define TCPCB_FLAG_URG 0x20
569#define TCPCB_FLAG_ECE 0x40
570#define TCPCB_FLAG_CWR 0x80
571
572 __u8 sacked; /* State flags for SACK/FACK. */
573#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
574#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
575#define TCPCB_LOST 0x04 /* SKB is lost */
576#define TCPCB_TAGBITS 0x07 /* All tag bits */
577
578#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
579#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
580
caa20d9a 581#define TCPCB_URG 0x20 /* Urgent pointer advanced here */
1da177e4
LT		 582
583#define TCPCB_AT_TAIL (TCPCB_URG)
584
585 __u16 urg_ptr; /* Valid w/URG flags is set. */
586 __u32 ack_seq; /* Sequence number ACK'd */
587};
588
589#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
590
1da177e4
LT		 591/* Due to TSO, an SKB can be composed of multiple actual
592 * packets. To keep these tracked properly, we use this.
593 */
594static inline int tcp_skb_pcount(const struct sk_buff *skb)
595{
7967168c 596 return skb_shinfo(skb)->gso_segs;
1da177e4
LT		 597}
598
599/* This is valid iff tcp_skb_pcount() > 1. */
600static inline int tcp_skb_mss(const struct sk_buff *skb)
601{
7967168c 602 return skb_shinfo(skb)->gso_size;
1da177e4
LT		 603}
604
af610b4c 605static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr)
1da177e4
LT		 606{
607 if (*count) {
af610b4c 608 *count -= decr;
1da177e4
LT		 609 if ((int)*count < 0)
610 *count = 0;
611 }
612}
613
af610b4c
IJ		 614static inline void tcp_dec_pcount_approx(__u32 *count,
615 const struct sk_buff *skb)
616{
617 tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb));
618}
619
317a76f9
SH		 620/* Events passed to congestion control interface */
621enum tcp_ca_event {
622 CA_EVENT_TX_START, /* first transmit when no packets in flight */
623 CA_EVENT_CWND_RESTART, /* congestion window restart */
624 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
625 CA_EVENT_FRTO, /* fast recovery timeout */
626 CA_EVENT_LOSS, /* loss timeout */
627 CA_EVENT_FAST_ACK, /* in sequence ack */
628 CA_EVENT_SLOW_ACK, /* other ack */
629};
630
631/*
632 * Interface for adding new TCP congestion control handlers
633 */
634#define TCP_CA_NAME_MAX 16
3ff825b2
SH		 635#define TCP_CA_MAX 128
636#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
637
164891aa
SH		 638#define TCP_CONG_NON_RESTRICTED 0x1
639#define TCP_CONG_RTT_STAMP 0x2
640
317a76f9
SH		 641struct tcp_congestion_ops {
642 struct list_head list;
164891aa 643 unsigned long flags;
317a76f9
SH		 644
645 /* initialize private data (optional) */
6687e988 646 void (*init)(struct sock *sk);
317a76f9 647 /* cleanup private data (optional) */
6687e988 648 void (*release)(struct sock *sk);
317a76f9
SH		 649
650 /* return slow start threshold (required) */
6687e988 651 u32 (*ssthresh)(struct sock *sk);
317a76f9 652 /* lower bound for congestion window (optional) */
72dc5b92 653 u32 (*min_cwnd)(const struct sock *sk);
317a76f9 654 /* do new cwnd calculation (required) */
c3a05c60 655 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
317a76f9 656 /* call before changing ca_state (optional) */
6687e988 657 void (*set_state)(struct sock *sk, u8 new_state);
317a76f9 658 /* call when cwnd event occurs (optional) */
6687e988 659 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
317a76f9 660 /* new value of cwnd after loss (optional) */
6687e988 661 u32 (*undo_cwnd)(struct sock *sk);
317a76f9 662 /* hook for packet ack accounting (optional) */
30cfd0ba 663 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
73c1f4a0 664 /* get info for inet_diag (optional) */
6687e988 665 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
317a76f9
SH		 666
667 char name[TCP_CA_NAME_MAX];
668 struct module *owner;
669};
670
671extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
672extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
673
6687e988
ACM		 674extern void tcp_init_congestion_control(struct sock *sk);
675extern void tcp_cleanup_congestion_control(struct sock *sk);
317a76f9
SH		 676extern int tcp_set_default_congestion_control(const char *name);
677extern void tcp_get_default_congestion_control(char *name);
3ff825b2 678extern void tcp_get_available_congestion_control(char *buf, size_t len);
ce7bc3bf
SH		 679extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
680extern int tcp_set_allowed_congestion_control(char *allowed);
6687e988 681extern int tcp_set_congestion_control(struct sock *sk, const char *name);
40efc6fa 682extern void tcp_slow_start(struct tcp_sock *tp);
317a76f9 683
5f8ef48d 684extern struct tcp_congestion_ops tcp_init_congestion_ops;
6687e988 685extern u32 tcp_reno_ssthresh(struct sock *sk);
c3a05c60 686extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
72dc5b92 687extern u32 tcp_reno_min_cwnd(const struct sock *sk);
a8acfbac 688extern struct tcp_congestion_ops tcp_reno;
317a76f9 689
6687e988 690static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
317a76f9 691{
6687e988
ACM		 692 struct inet_connection_sock *icsk = inet_csk(sk);
693
694 if (icsk->icsk_ca_ops->set_state)
695 icsk->icsk_ca_ops->set_state(sk, ca_state);
696 icsk->icsk_ca_state = ca_state;
317a76f9
SH		 697}
698
6687e988 699static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
317a76f9 700{
6687e988
ACM		 701 const struct inet_connection_sock *icsk = inet_csk(sk);
702
703 if (icsk->icsk_ca_ops->cwnd_event)
704 icsk->icsk_ca_ops->cwnd_event(sk, event);
317a76f9
SH		 705}
706
e60402d0
IJ		 707/* These functions determine how the current flow behaves in respect of SACK
708 * handling. SACK is negotiated with the peer, and therefore it can vary
709 * between different flows.
710 *
711 * tcp_is_sack - SACK enabled
712 * tcp_is_reno - No SACK
713 * tcp_is_fack - FACK enabled, implies SACK enabled
714 */
715static inline int tcp_is_sack(const struct tcp_sock *tp)
716{
717 return tp->rx_opt.sack_ok;
718}
719
720static inline int tcp_is_reno(const struct tcp_sock *tp)
721{
722 return !tcp_is_sack(tp);
723}
724
725static inline int tcp_is_fack(const struct tcp_sock *tp)
726{
727 return tp->rx_opt.sack_ok & 2;
728}
729
730static inline void tcp_enable_fack(struct tcp_sock *tp)
731{
732 tp->rx_opt.sack_ok |= 2;
733}
734
83ae4088
IJ		 735static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
736{
737 return tp->sacked_out + tp->lost_out;
738}
739
1da177e4
LT		 740/* This determines how many packets are "in the network" to the best
741 * of our knowledge. In many cases it is conservative, but where
742 * detailed information is available from the receiver (via SACK
743 * blocks etc.) we can make more aggressive calculations.
744 *
745 * Use this for decisions involving congestion control, use just
746 * tp->packets_out to determine if the send queue is empty or not.
747 *
748 * Read this equation as:
749 *
750 * "Packets sent once on transmission queue" MINUS
751 * "Packets left network, but not honestly ACKed yet" PLUS
752 * "Packets fast retransmitted"
753 */
40efc6fa 754static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
1da177e4 755{
83ae4088 756 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
1da177e4
LT		 757}
758
1da177e4
LT		 759/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
760 * The exception is rate halving phase, when cwnd is decreasing towards
761 * ssthresh.
762 */
6687e988 763static inline __u32 tcp_current_ssthresh(const struct sock *sk)
1da177e4 764{
6687e988
ACM		 765 const struct tcp_sock *tp = tcp_sk(sk);
766 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
1da177e4
LT		 767 return tp->snd_ssthresh;
768 else
769 return max(tp->snd_ssthresh,
770 ((tp->snd_cwnd >> 1) +
771 (tp->snd_cwnd >> 2)));
772}
773
b9c4595b
IJ		 774/* Use define here intentionally to get WARN_ON location shown at the caller */
775#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
1da177e4 776
3cfe3baa 777extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
1da177e4
LT		 778extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
779
780/* Slow start with delack produces 3 packets of burst, so that
781 * it is safe "de facto".
782 */
783static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
784{
785 return 3;
786}
787
f4805ede
SH		 788/* RFC2861 Check whether we are limited by application or congestion window
789 * This is the inverse of cwnd check in tcp_tso_should_defer
790 */
791static inline int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight)
792{
793 const struct tcp_sock *tp = tcp_sk(sk);
794 u32 left;
795
796 if (in_flight >= tp->snd_cwnd)
797 return 1;
798
bcd76111 799 if (!sk_can_gso(sk))
f4805ede
SH		 800 return 0;
801
802 left = tp->snd_cwnd - in_flight;
803 if (sysctl_tcp_tso_win_divisor)
804 return left * sysctl_tcp_tso_win_divisor < tp->snd_cwnd;
805 else
806 return left <= tcp_max_burst(tp);
807}
808
c1bd24b7 809static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
40efc6fa 810 const struct sk_buff *skb)
1da177e4
LT		 811{
812 if (skb->len < mss)
813 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
814}
815
9e412ba7 816static inline void tcp_check_probe_timer(struct sock *sk)
1da177e4 817{
9e412ba7 818 struct tcp_sock *tp = tcp_sk(sk);
463c84b9 819 const struct inet_connection_sock *icsk = inet_csk(sk);
9e412ba7 820
463c84b9 821 if (!tp->packets_out && !icsk->icsk_pending)
3f421baa
ACM		 822 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
823 icsk->icsk_rto, TCP_RTO_MAX);
1da177e4
LT		 824}
825
9e412ba7 826static inline void tcp_push_pending_frames(struct sock *sk)
1da177e4 827{
9e412ba7
IJ		 828 struct tcp_sock *tp = tcp_sk(sk);
829
830 __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle);
1da177e4
LT		 831}
832
40efc6fa 833static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 834{
835 tp->snd_wl1 = seq;
836}
837
40efc6fa 838static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 839{
840 tp->snd_wl1 = seq;
841}
842
1da177e4
LT		 843/*
844 * Calculate(/check) TCP checksum
845 */
ba7808ea
FD		 846static inline __sum16 tcp_v4_check(int len, __be32 saddr,
847 __be32 daddr, __wsum base)
1da177e4
LT		 848{
849 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
850}
851
b51655b9 852static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1da177e4 853{
fb286bb2 854 return __skb_checksum_complete(skb);
1da177e4
LT		 855}
856
40efc6fa 857static inline int tcp_checksum_complete(struct sk_buff *skb)
1da177e4 858{
60476372 859 return !skb_csum_unnecessary(skb) &&
1da177e4
LT		 860 __tcp_checksum_complete(skb);
861}
862
863/* Prequeue for VJ style copy to user, combined with checksumming. */
864
40efc6fa 865static inline void tcp_prequeue_init(struct tcp_sock *tp)
1da177e4
LT		 866{
867 tp->ucopy.task = NULL;
868 tp->ucopy.len = 0;
869 tp->ucopy.memory = 0;
870 skb_queue_head_init(&tp->ucopy.prequeue);
97fc2f08
CL		 871#ifdef CONFIG_NET_DMA
872 tp->ucopy.dma_chan = NULL;
873 tp->ucopy.wakeup = 0;
874 tp->ucopy.pinned_list = NULL;
875 tp->ucopy.dma_cookie = 0;
876#endif
1da177e4
LT		 877}
878
879/* Packet is added to VJ-style prequeue for processing in process
880 * context, if a reader task is waiting. Apparently, this exciting
881 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
882 * failed somewhere. Latency? Burstiness? Well, at least now we will
883 * see, why it failed. 8)8) --ANK
884 *
885 * NOTE: is this not too big to inline?
886 */
40efc6fa 887static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
1da177e4
LT		 888{
889 struct tcp_sock *tp = tcp_sk(sk);
890
891 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
892 __skb_queue_tail(&tp->ucopy.prequeue, skb);
893 tp->ucopy.memory += skb->truesize;
894 if (tp->ucopy.memory > sk->sk_rcvbuf) {
895 struct sk_buff *skb1;
896
897 BUG_ON(sock_owned_by_user(sk));
898
899 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
900 sk->sk_backlog_rcv(sk, skb1);
901 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
902 }
903
904 tp->ucopy.memory = 0;
905 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
906 wake_up_interruptible(sk->sk_sleep);
463c84b9
ACM		 907 if (!inet_csk_ack_scheduled(sk))
908 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
3f421baa
ACM		 909 (3 * TCP_RTO_MIN) / 4,
910 TCP_RTO_MAX);
1da177e4
LT		 911 }
912 return 1;
913 }
914 return 0;
915}
916
917
918#undef STATE_TRACE
919
920#ifdef STATE_TRACE
921static const char *statename[]={
922 "Unused","Established","Syn Sent","Syn Recv",
923 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
924 "Close Wait","Last ACK","Listen","Closing"
925};
926#endif
927
40efc6fa 928static inline void tcp_set_state(struct sock *sk, int state)
1da177e4
LT		 929{
930 int oldstate = sk->sk_state;
931
932 switch (state) {
933 case TCP_ESTABLISHED:
934 if (oldstate != TCP_ESTABLISHED)
935 TCP_INC_STATS(TCP_MIB_CURRESTAB);
936 break;
937
938 case TCP_CLOSE:
939 if (oldstate == TCP_CLOSE_WAIT || oldstate == TCP_ESTABLISHED)
940 TCP_INC_STATS(TCP_MIB_ESTABRESETS);
941
942 sk->sk_prot->unhash(sk);
463c84b9 943 if (inet_csk(sk)->icsk_bind_hash &&
1da177e4 944 !(sk->sk_userlocks & SOCK_BINDPORT_LOCK))
2d8c4ce5 945 inet_put_port(&tcp_hashinfo, sk);
1da177e4
LT		 946 /* fall through */
947 default:
948 if (oldstate==TCP_ESTABLISHED)
949 TCP_DEC_STATS(TCP_MIB_CURRESTAB);
950 }
951
952 /* Change state AFTER socket is unhashed to avoid closed
953 * socket sitting in hash tables.
954 */
955 sk->sk_state = state;
956
957#ifdef STATE_TRACE
958 SOCK_DEBUG(sk, "TCP sk=%p, State %s -> %s\n",sk, statename[oldstate],statename[state]);
959#endif
960}
961
4ac02bab 962extern void tcp_done(struct sock *sk);
1da177e4 963
40efc6fa 964static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1da177e4
LT		 965{
966 rx_opt->dsack = 0;
967 rx_opt->eff_sacks = 0;
968 rx_opt->num_sacks = 0;
969}
970
1da177e4
LT		 971/* Determine a window scaling and initial window to offer. */
972extern void tcp_select_initial_window(int __space, __u32 mss,
973 __u32 *rcv_wnd, __u32 *window_clamp,
974 int wscale_ok, __u8 *rcv_wscale);
975
976static inline int tcp_win_from_space(int space)
977{
978 return sysctl_tcp_adv_win_scale<=0 ?
979 (space>>(-sysctl_tcp_adv_win_scale)) :
980 space - (space>>sysctl_tcp_adv_win_scale);
981}
982
983/* Note: caller must be prepared to deal with negative returns */
984static inline int tcp_space(const struct sock *sk)
985{
986 return tcp_win_from_space(sk->sk_rcvbuf -
987 atomic_read(&sk->sk_rmem_alloc));
988}
989
990static inline int tcp_full_space(const struct sock *sk)
991{
992 return tcp_win_from_space(sk->sk_rcvbuf);
993}
994
40efc6fa
SH		 995static inline void tcp_openreq_init(struct request_sock *req,
996 struct tcp_options_received *rx_opt,
997 struct sk_buff *skb)
1da177e4 998{
2e6599cb
ACM		 999 struct inet_request_sock *ireq = inet_rsk(req);
1000
1da177e4 1001 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
2e6599cb 1002 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1da177e4
LT		 1003 req->mss = rx_opt->mss_clamp;
1004 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
2e6599cb
ACM		 1005 ireq->tstamp_ok = rx_opt->tstamp_ok;
1006 ireq->sack_ok = rx_opt->sack_ok;
1007 ireq->snd_wscale = rx_opt->snd_wscale;
1008 ireq->wscale_ok = rx_opt->wscale_ok;
1009 ireq->acked = 0;
1010 ireq->ecn_ok = 0;
aa8223c7 1011 ireq->rmt_port = tcp_hdr(skb)->source;
1da177e4
LT		 1012}
1013
1014extern void tcp_enter_memory_pressure(void);
1015
1da177e4
LT		 1016static inline int keepalive_intvl_when(const struct tcp_sock *tp)
1017{
1018 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
1019}
1020
1021static inline int keepalive_time_when(const struct tcp_sock *tp)
1022{
1023 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
1024}
1025
463c84b9 1026static inline int tcp_fin_time(const struct sock *sk)
1da177e4 1027{
463c84b9
ACM		 1028 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
1029 const int rto = inet_csk(sk)->icsk_rto;
1da177e4 1030
463c84b9
ACM		 1031 if (fin_timeout < (rto << 2) - (rto >> 1))
1032 fin_timeout = (rto << 2) - (rto >> 1);
1da177e4
LT		 1033
1034 return fin_timeout;
1035}
1036
1037static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1038{
1039 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1040 return 0;
9d729f72 1041 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
1da177e4
LT		 1042 return 0;
1043
1044 /* RST segments are not recommended to carry timestamp,
1045 and, if they do, it is recommended to ignore PAWS because
1046 "their cleanup function should take precedence over timestamps."
1047 Certainly, it is mistake. It is necessary to understand the reasons
1048 of this constraint to relax it: if peer reboots, clock may go
1049 out-of-sync and half-open connections will not be reset.
1050 Actually, the problem would be not existing if all
1051 the implementations followed draft about maintaining clock
1052 via reboots. Linux-2.2 DOES NOT!
1053
1054 However, we can relax time bounds for RST segments to MSL.
1055 */
9d729f72 1056 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1da177e4
LT		 1057 return 0;
1058 return 1;
1059}
1060
1da177e4
LT		 1061#define TCP_CHECK_TIMER(sk) do { } while (0)
1062
1da177e4
LT		 1063static inline void tcp_mib_init(void)
1064{
1065 /* See RFC 2012 */
1066 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1067 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1068 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1069 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1070}
1071
5af4ec23 1072/* from STCP */
0800f170
DM		 1073static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1074{
6a438bbe
SH		 1075 tp->lost_skb_hint = NULL;
1076 tp->scoreboard_skb_hint = NULL;
1077 tp->retransmit_skb_hint = NULL;
1078 tp->forward_skb_hint = NULL;
b7689205
IJ		 1079}
1080
0800f170
DM		 1081static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1082{
b7689205 1083 tcp_clear_retrans_hints_partial(tp);
6a438bbe
SH		 1084}
1085
cfb6eeb4
YH		 1086/* MD5 Signature */
1087struct crypto_hash;
1088
1089/* - key database */
1090struct tcp_md5sig_key {
1091 u8 *key;
1092 u8 keylen;
1093};
1094
1095struct tcp4_md5sig_key {
f8ab18d2 1096 struct tcp_md5sig_key base;
cfb6eeb4
YH		 1097 __be32 addr;
1098};
1099
1100struct tcp6_md5sig_key {
f8ab18d2 1101 struct tcp_md5sig_key base;
cfb6eeb4
YH		 1102#if 0
1103 u32 scope_id; /* XXX */
1104#endif
1105 struct in6_addr addr;
1106};
1107
1108/* - sock block */
1109struct tcp_md5sig_info {
1110 struct tcp4_md5sig_key *keys4;
1111#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1112 struct tcp6_md5sig_key *keys6;
1113 u32 entries6;
1114 u32 alloced6;
1115#endif
1116 u32 entries4;
1117 u32 alloced4;
1118};
1119
1120/* - pseudo header */
1121struct tcp4_pseudohdr {
1122 __be32 saddr;
1123 __be32 daddr;
1124 __u8 pad;
1125 __u8 protocol;
1126 __be16 len;
1127};
1128
1129struct tcp6_pseudohdr {
1130 struct in6_addr saddr;
1131 struct in6_addr daddr;
1132 __be32 len;
1133 __be32 protocol; /* including padding */
1134};
1135
1136union tcp_md5sum_block {
1137 struct tcp4_pseudohdr ip4;
1138#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1139 struct tcp6_pseudohdr ip6;
1140#endif
1141};
1142
1143/* - pool: digest algorithm, hash description and scratch buffer */
1144struct tcp_md5sig_pool {
1145 struct hash_desc md5_desc;
1146 union tcp_md5sum_block md5_blk;
1147};
1148
1149#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1150
1151/* - functions */
1152extern int tcp_v4_calc_md5_hash(char *md5_hash,
1153 struct tcp_md5sig_key *key,
1154 struct sock *sk,
1155 struct dst_entry *dst,
1156 struct request_sock *req,
1157 struct tcphdr *th,
1158 int protocol, int tcplen);
1159extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1160 struct sock *addr_sk);
1161
1162extern int tcp_v4_md5_do_add(struct sock *sk,
1163 __be32 addr,
1164 u8 *newkey,
1165 u8 newkeylen);
1166
1167extern int tcp_v4_md5_do_del(struct sock *sk,
8e5200f5 1168 __be32 addr);
cfb6eeb4
YH		 1169
1170extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1171extern void tcp_free_md5sig_pool(void);
1172
1173extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1174extern void __tcp_put_md5sig_pool(void);
1175
1176static inline
1177struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1178{
1179 int cpu = get_cpu();
1180 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1181 if (!ret)
1182 put_cpu();
1183 return ret;
1184}
1185
1186static inline void tcp_put_md5sig_pool(void)
1187{
1188 __tcp_put_md5sig_pool();
1189 put_cpu();
1190}
1191
fe067e8a
DM		 1192/* write queue abstraction */
1193static inline void tcp_write_queue_purge(struct sock *sk)
1194{
1195 struct sk_buff *skb;
1196
1197 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
1198 sk_stream_free_skb(sk, skb);
1199 sk_stream_mem_reclaim(sk);
1200}
1201
1202static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1203{
1204 struct sk_buff *skb = sk->sk_write_queue.next;
1205 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1206 return NULL;
1207 return skb;
1208}
1209
1210static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1211{
1212 struct sk_buff *skb = sk->sk_write_queue.prev;
1213 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1214 return NULL;
1215 return skb;
1216}
1217
1218static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1219{
1220 return skb->next;
1221}
1222
1223#define tcp_for_write_queue(skb, sk) \
1224 for (skb = (sk)->sk_write_queue.next; \
1225 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1226 skb = skb->next)
1227
1228#define tcp_for_write_queue_from(skb, sk) \
1229 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1230 skb = skb->next)
1231
234b6860
IJ		 1232#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1233 for (tmp = skb->next; \
1234 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1235 skb = tmp, tmp = skb->next)
1236
fe067e8a
DM		 1237static inline struct sk_buff *tcp_send_head(struct sock *sk)
1238{
1239 return sk->sk_send_head;
1240}
1241
1242static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1243{
1244 sk->sk_send_head = skb->next;
1245 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1246 sk->sk_send_head = NULL;
1247}
1248
1249static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1250{
1251 if (sk->sk_send_head == skb_unlinked)
1252 sk->sk_send_head = NULL;
1253}
1254
1255static inline void tcp_init_send_head(struct sock *sk)
1256{
1257 sk->sk_send_head = NULL;
1258}
1259
1260static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1261{
1262 __skb_queue_tail(&sk->sk_write_queue, skb);
1263}
1264
1265static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1266{
1267 __tcp_add_write_queue_tail(sk, skb);
1268
1269 /* Queue it, remembering where we must start sending. */
1270 if (sk->sk_send_head == NULL)
1271 sk->sk_send_head = skb;
1272}
1273
1274static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1275{
1276 __skb_queue_head(&sk->sk_write_queue, skb);
1277}
1278
1279/* Insert buff after skb on the write queue of sk. */
1280static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1281 struct sk_buff *buff,
1282 struct sock *sk)
1283{
1284 __skb_append(skb, buff, &sk->sk_write_queue);
1285}
1286
1287/* Insert skb between prev and next on the write queue of sk. */
1288static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1289 struct sk_buff *skb,
1290 struct sock *sk)
1291{
1292 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);