]> bbs.cooldavid.org Git - net-next-2.6.git/blame - include/net/tcp.h
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[NET]: Do not allocate unneeded memory for dev->priv alignment.
[net-next-2.6.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>
c6aefafb 32#include <linux/cryptohash.h>
3f421baa
ACM		 33
34#include <net/inet_connection_sock.h>
295ff7ed 35#include <net/inet_timewait_sock.h>
77d8bf9c 36#include <net/inet_hashtables.h>
1da177e4 37#include <net/checksum.h>
2e6599cb 38#include <net/request_sock.h>
1da177e4
LT		 39#include <net/sock.h>
40#include <net/snmp.h>
41#include <net/ip.h>
c752f073 42#include <net/tcp_states.h>
bdf1ee5d 43#include <net/inet_ecn.h>
c752f073 44
1da177e4
LT		 45#include <linux/seq_file.h>
46
6e04e021 47extern struct inet_hashinfo tcp_hashinfo;
1da177e4 48
1da177e4 49extern atomic_t tcp_orphan_count;
1da177e4 50extern void tcp_time_wait(struct sock *sk, int state, int timeo);
1da177e4 51
1da177e4
LT		 52#define MAX_TCP_HEADER (128 + MAX_HEADER)
53
54/*
55 * Never offer a window over 32767 without using window scaling. Some
56 * poor stacks do signed 16bit maths!
57 */
58#define MAX_TCP_WINDOW 32767U
59
60/* Minimal accepted MSS. It is (60+60+8) - (20+20). */
61#define TCP_MIN_MSS 88U
62
63/* Minimal RCV_MSS. */
64#define TCP_MIN_RCVMSS 536U
65
5d424d5a
JH		 66/* The least MTU to use for probing */
67#define TCP_BASE_MSS 512
68
1da177e4
LT		 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
caa20d9a 98 * connection: ~180sec is RFC minimum */
1da177e4
LT		 99
100#define TCP_SYNACK_RETRIES 5 /* number of times to retry passive opening a
caa20d9a 101 * connection: ~180sec is RFC minimum */
1da177e4
LT		 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
ec3c0982 142#define MAX_TCP_ACCEPT_DEFERRED 65535
1da177e4
LT		 143
144#define TCP_SYNQ_INTERVAL (HZ/5) /* Period of SYNACK timer */
1da177e4
LT		 145
146#define TCP_PAWS_24DAYS (60 * 60 * 24 * 24)
147#define TCP_PAWS_MSL 60 /* Per-host timestamps are invalidated
148 * after this time. It should be equal
149 * (or greater than) TCP_TIMEWAIT_LEN
150 * to provide reliability equal to one
151 * provided by timewait state.
152 */
153#define TCP_PAWS_WINDOW 1 /* Replay window for per-host
154 * timestamps. It must be less than
155 * minimal timewait lifetime.
156 */
1da177e4
LT		 157/*
158 * TCP option
159 */
160
161#define TCPOPT_NOP 1 /* Padding */
162#define TCPOPT_EOL 0 /* End of options */
163#define TCPOPT_MSS 2 /* Segment size negotiating */
164#define TCPOPT_WINDOW 3 /* Window scaling */
165#define TCPOPT_SACK_PERM 4 /* SACK Permitted */
166#define TCPOPT_SACK 5 /* SACK Block */
167#define TCPOPT_TIMESTAMP 8 /* Better RTT estimations/PAWS */
cfb6eeb4 168#define TCPOPT_MD5SIG 19 /* MD5 Signature (RFC2385) */
1da177e4
LT		 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
cfb6eeb4 178#define TCPOLEN_MD5SIG 18
1da177e4
LT		 179
180/* But this is what stacks really send out. */
181#define TCPOLEN_TSTAMP_ALIGNED 12
182#define TCPOLEN_WSCALE_ALIGNED 4
183#define TCPOLEN_SACKPERM_ALIGNED 4
184#define TCPOLEN_SACK_BASE 2
185#define TCPOLEN_SACK_BASE_ALIGNED 4
186#define TCPOLEN_SACK_PERBLOCK 8
cfb6eeb4 187#define TCPOLEN_MD5SIG_ALIGNED 20
1da177e4 188
1da177e4
LT		 189/* Flags in tp->nonagle */
190#define TCP_NAGLE_OFF 1 /* Nagle's algo is disabled */
191#define TCP_NAGLE_CORK 2 /* Socket is corked */
caa20d9a 192#define TCP_NAGLE_PUSH 4 /* Cork is overridden for already queued data */
1da177e4 193
295ff7ed
ACM		 194extern struct inet_timewait_death_row tcp_death_row;
195
1da177e4 196/* sysctl variables for tcp */
1da177e4
LT		 197extern int sysctl_tcp_timestamps;
198extern int sysctl_tcp_window_scaling;
199extern int sysctl_tcp_sack;
200extern int sysctl_tcp_fin_timeout;
1da177e4
LT		 201extern int sysctl_tcp_keepalive_time;
202extern int sysctl_tcp_keepalive_probes;
203extern int sysctl_tcp_keepalive_intvl;
204extern int sysctl_tcp_syn_retries;
205extern int sysctl_tcp_synack_retries;
206extern int sysctl_tcp_retries1;
207extern int sysctl_tcp_retries2;
208extern int sysctl_tcp_orphan_retries;
209extern int sysctl_tcp_syncookies;
210extern int sysctl_tcp_retrans_collapse;
211extern int sysctl_tcp_stdurg;
212extern int sysctl_tcp_rfc1337;
213extern int sysctl_tcp_abort_on_overflow;
214extern int sysctl_tcp_max_orphans;
1da177e4
LT		 215extern int sysctl_tcp_fack;
216extern int sysctl_tcp_reordering;
217extern int sysctl_tcp_ecn;
218extern int sysctl_tcp_dsack;
219extern int sysctl_tcp_mem[3];
220extern int sysctl_tcp_wmem[3];
221extern int sysctl_tcp_rmem[3];
222extern int sysctl_tcp_app_win;
223extern int sysctl_tcp_adv_win_scale;
224extern int sysctl_tcp_tw_reuse;
225extern int sysctl_tcp_frto;
3cfe3baa 226extern int sysctl_tcp_frto_response;
1da177e4 227extern int sysctl_tcp_low_latency;
95937825 228extern int sysctl_tcp_dma_copybreak;
1da177e4 229extern int sysctl_tcp_nometrics_save;
1da177e4
LT		 230extern int sysctl_tcp_moderate_rcvbuf;
231extern int sysctl_tcp_tso_win_divisor;
9772efb9 232extern int sysctl_tcp_abc;
5d424d5a
JH		 233extern int sysctl_tcp_mtu_probing;
234extern int sysctl_tcp_base_mss;
15d99e02 235extern int sysctl_tcp_workaround_signed_windows;
35089bb2 236extern int sysctl_tcp_slow_start_after_idle;
886236c1 237extern int sysctl_tcp_max_ssthresh;
1da177e4
LT		 238
239extern atomic_t tcp_memory_allocated;
240extern atomic_t tcp_sockets_allocated;
241extern int tcp_memory_pressure;
242
1da177e4
LT		 243/*
244 * The next routines deal with comparing 32 bit unsigned ints
245 * and worry about wraparound (automatic with unsigned arithmetic).
246 */
247
248static inline int before(__u32 seq1, __u32 seq2)
249{
0d630cc0 250 return (__s32)(seq1-seq2) < 0;
1da177e4 251}
9a036b9c 252#define after(seq2, seq1) before(seq1, seq2)
1da177e4
LT		 253
254/* is s2<=s1<=s3 ? */
255static inline int between(__u32 seq1, __u32 seq2, __u32 seq3)
256{
257 return seq3 - seq2 >= seq1 - seq2;
258}
259
e4fd5da3
PE		 260static inline int tcp_too_many_orphans(struct sock *sk, int num)
261{
262 return (num > sysctl_tcp_max_orphans) ||
263 (sk->sk_wmem_queued > SOCK_MIN_SNDBUF &&
264 atomic_read(&tcp_memory_allocated) > sysctl_tcp_mem[2]);
265}
1da177e4
LT		 266
267extern struct proto tcp_prot;
268
269DECLARE_SNMP_STAT(struct tcp_mib, tcp_statistics);
270#define TCP_INC_STATS(field) SNMP_INC_STATS(tcp_statistics, field)
271#define TCP_INC_STATS_BH(field) SNMP_INC_STATS_BH(tcp_statistics, field)
272#define TCP_INC_STATS_USER(field) SNMP_INC_STATS_USER(tcp_statistics, field)
273#define TCP_DEC_STATS(field) SNMP_DEC_STATS(tcp_statistics, field)
274#define TCP_ADD_STATS_BH(field, val) SNMP_ADD_STATS_BH(tcp_statistics, field, val)
275#define TCP_ADD_STATS_USER(field, val) SNMP_ADD_STATS_USER(tcp_statistics, field, val)
276
1da177e4
LT		 277extern void tcp_v4_err(struct sk_buff *skb, u32);
278
279extern void tcp_shutdown (struct sock *sk, int how);
280
281extern int tcp_v4_rcv(struct sk_buff *skb);
282
283extern int tcp_v4_remember_stamp(struct sock *sk);
284
8feaf0c0 285extern int tcp_v4_tw_remember_stamp(struct inet_timewait_sock *tw);
1da177e4 286
3516ffb0 287extern int tcp_sendmsg(struct kiocb *iocb, struct socket *sock,
1da177e4
LT		 288 struct msghdr *msg, size_t size);
289extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
290
291extern int tcp_ioctl(struct sock *sk,
292 int cmd,
293 unsigned long arg);
294
295extern int tcp_rcv_state_process(struct sock *sk,
296 struct sk_buff *skb,
297 struct tcphdr *th,
298 unsigned len);
299
300extern int tcp_rcv_established(struct sock *sk,
301 struct sk_buff *skb,
302 struct tcphdr *th,
303 unsigned len);
304
305extern void tcp_rcv_space_adjust(struct sock *sk);
306
0e4b4992
CL		 307extern void tcp_cleanup_rbuf(struct sock *sk, int copied);
308
6d6ee43e
ACM		 309extern int tcp_twsk_unique(struct sock *sk,
310 struct sock *sktw, void *twp);
311
cfb6eeb4
YH		 312extern void tcp_twsk_destructor(struct sock *sk);
313
9c55e01c
JA		 314extern ssize_t tcp_splice_read(struct socket *sk, loff_t *ppos,
315 struct pipe_inode_info *pipe, size_t len, unsigned int flags);
316
463c84b9
ACM		 317static inline void tcp_dec_quickack_mode(struct sock *sk,
318 const unsigned int pkts)
1da177e4 319{
463c84b9 320 struct inet_connection_sock *icsk = inet_csk(sk);
fc6415bc 321
463c84b9
ACM		 322 if (icsk->icsk_ack.quick) {
323 if (pkts >= icsk->icsk_ack.quick) {
324 icsk->icsk_ack.quick = 0;
fc6415bc 325 /* Leaving quickack mode we deflate ATO. */
463c84b9 326 icsk->icsk_ack.ato = TCP_ATO_MIN;
fc6415bc 327 } else
463c84b9 328 icsk->icsk_ack.quick -= pkts;
1da177e4
LT		 329 }
330}
331
463c84b9 332extern void tcp_enter_quickack_mode(struct sock *sk);
1da177e4 333
1da177e4
LT		 334static inline void tcp_clear_options(struct tcp_options_received *rx_opt)
335{
336 rx_opt->tstamp_ok = rx_opt->sack_ok = rx_opt->wscale_ok = rx_opt->snd_wscale = 0;
337}
338
bdf1ee5d
IJ		 339#define TCP_ECN_OK 1
340#define TCP_ECN_QUEUE_CWR 2
341#define TCP_ECN_DEMAND_CWR 4
342
343static __inline__ void
344TCP_ECN_create_request(struct request_sock *req, struct tcphdr *th)
345{
346 if (sysctl_tcp_ecn && th->ece && th->cwr)
347 inet_rsk(req)->ecn_ok = 1;
348}
349
1da177e4
LT		 350enum tcp_tw_status
351{
352 TCP_TW_SUCCESS = 0,
353 TCP_TW_RST = 1,
354 TCP_TW_ACK = 2,
355 TCP_TW_SYN = 3
356};
357
358
8feaf0c0 359extern enum tcp_tw_status tcp_timewait_state_process(struct inet_timewait_sock *tw,
1da177e4 360 struct sk_buff *skb,
8feaf0c0 361 const struct tcphdr *th);
1da177e4
LT		 362
363extern struct sock * tcp_check_req(struct sock *sk,struct sk_buff *skb,
60236fdd
ACM		 364 struct request_sock *req,
365 struct request_sock **prev);
1da177e4
LT		 366extern int tcp_child_process(struct sock *parent,
367 struct sock *child,
368 struct sk_buff *skb);
46d0de4e 369extern int tcp_use_frto(struct sock *sk);
1da177e4
LT		 370extern void tcp_enter_frto(struct sock *sk);
371extern void tcp_enter_loss(struct sock *sk, int how);
372extern void tcp_clear_retrans(struct tcp_sock *tp);
373extern void tcp_update_metrics(struct sock *sk);
374
375extern void tcp_close(struct sock *sk,
376 long timeout);
1da177e4
LT		 377extern unsigned int tcp_poll(struct file * file, struct socket *sock, struct poll_table_struct *wait);
378
379extern int tcp_getsockopt(struct sock *sk, int level,
380 int optname,
381 char __user *optval,
382 int __user *optlen);
383extern int tcp_setsockopt(struct sock *sk, int level,
384 int optname, char __user *optval,
385 int optlen);
3fdadf7d
DM		 386extern int compat_tcp_getsockopt(struct sock *sk,
387 int level, int optname,
388 char __user *optval, int __user *optlen);
389extern int compat_tcp_setsockopt(struct sock *sk,
390 int level, int optname,
391 char __user *optval, int optlen);
1da177e4
LT		 392extern void tcp_set_keepalive(struct sock *sk, int val);
393extern int tcp_recvmsg(struct kiocb *iocb, struct sock *sk,
394 struct msghdr *msg,
395 size_t len, int nonblock,
396 int flags, int *addr_len);
397
1da177e4
LT		 398extern void tcp_parse_options(struct sk_buff *skb,
399 struct tcp_options_received *opt_rx,
400 int estab);
401
402/*
403 * TCP v4 functions exported for the inet6 API
404 */
405
8292a17a 406extern void tcp_v4_send_check(struct sock *sk, int len,
1da177e4
LT		 407 struct sk_buff *skb);
408
409extern int tcp_v4_conn_request(struct sock *sk,
410 struct sk_buff *skb);
411
412extern struct sock * tcp_create_openreq_child(struct sock *sk,
60236fdd 413 struct request_sock *req,
1da177e4
LT		 414 struct sk_buff *skb);
415
416extern struct sock * tcp_v4_syn_recv_sock(struct sock *sk,
417 struct sk_buff *skb,
60236fdd 418 struct request_sock *req,
1da177e4
LT		 419 struct dst_entry *dst);
420
421extern int tcp_v4_do_rcv(struct sock *sk,
422 struct sk_buff *skb);
423
424extern int tcp_v4_connect(struct sock *sk,
425 struct sockaddr *uaddr,
426 int addr_len);
427
428extern int tcp_connect(struct sock *sk);
429
430extern struct sk_buff * tcp_make_synack(struct sock *sk,
431 struct dst_entry *dst,
60236fdd 432 struct request_sock *req);
1da177e4
LT		 433
434extern int tcp_disconnect(struct sock *sk, int flags);
435
436extern void tcp_unhash(struct sock *sk);
437
1da177e4 438/* From syncookies.c */
2051f11f 439extern __u32 syncookie_secret[2][16-4+SHA_DIGEST_WORDS];
1da177e4
LT		 440extern struct sock *cookie_v4_check(struct sock *sk, struct sk_buff *skb,
441 struct ip_options *opt);
442extern __u32 cookie_v4_init_sequence(struct sock *sk, struct sk_buff *skb,
443 __u16 *mss);
444
4dfc2817
FW		 445extern __u32 cookie_init_timestamp(struct request_sock *req);
446extern void cookie_check_timestamp(struct tcp_options_received *tcp_opt);
447
c6aefafb
GG		 448/* From net/ipv6/syncookies.c */
449extern struct sock *cookie_v6_check(struct sock *sk, struct sk_buff *skb);
450extern __u32 cookie_v6_init_sequence(struct sock *sk, struct sk_buff *skb,
451 __u16 *mss);
452
1da177e4
LT		 453/* tcp_output.c */
454
9e412ba7
IJ		 455extern void __tcp_push_pending_frames(struct sock *sk, unsigned int cur_mss,
456 int nonagle);
457extern int tcp_may_send_now(struct sock *sk);
1da177e4
LT		 458extern int tcp_retransmit_skb(struct sock *, struct sk_buff *);
459extern void tcp_xmit_retransmit_queue(struct sock *);
460extern void tcp_simple_retransmit(struct sock *);
461extern int tcp_trim_head(struct sock *, struct sk_buff *, u32);
6475be16 462extern int tcp_fragment(struct sock *, struct sk_buff *, u32, unsigned int);
1da177e4
LT		 463
464extern void tcp_send_probe0(struct sock *);
465extern void tcp_send_partial(struct sock *);
466extern int tcp_write_wakeup(struct sock *);
467extern void tcp_send_fin(struct sock *sk);
dd0fc66f 468extern void tcp_send_active_reset(struct sock *sk, gfp_t priority);
1da177e4 469extern int tcp_send_synack(struct sock *);
c1b4a7e6 470extern void tcp_push_one(struct sock *, unsigned int mss_now);
1da177e4
LT		 471extern void tcp_send_ack(struct sock *sk);
472extern void tcp_send_delayed_ack(struct sock *sk);
473
a762a980
DM		 474/* tcp_input.c */
475extern void tcp_cwnd_application_limited(struct sock *sk);
476
1da177e4
LT		 477/* tcp_timer.c */
478extern void tcp_init_xmit_timers(struct sock *);
463c84b9
ACM		 479static inline void tcp_clear_xmit_timers(struct sock *sk)
480{
481 inet_csk_clear_xmit_timers(sk);
482}
1da177e4 483
1da177e4
LT		 484extern unsigned int tcp_sync_mss(struct sock *sk, u32 pmtu);
485extern unsigned int tcp_current_mss(struct sock *sk, int large);
486
17b085ea 487/* tcp.c */
1da177e4
LT		 488extern void tcp_get_info(struct sock *, struct tcp_info *);
489
490/* Read 'sendfile()'-style from a TCP socket */
491typedef int (*sk_read_actor_t)(read_descriptor_t *, struct sk_buff *,
492 unsigned int, size_t);
493extern int tcp_read_sock(struct sock *sk, read_descriptor_t *desc,
494 sk_read_actor_t recv_actor);
495
40efc6fa 496extern void tcp_initialize_rcv_mss(struct sock *sk);
1da177e4 497
5d424d5a
JH		 498extern int tcp_mtu_to_mss(struct sock *sk, int pmtu);
499extern int tcp_mss_to_mtu(struct sock *sk, int mss);
500extern void tcp_mtup_init(struct sock *sk);
501
40efc6fa 502static inline void __tcp_fast_path_on(struct tcp_sock *tp, u32 snd_wnd)
1da177e4
LT		 503{
504 tp->pred_flags = htonl((tp->tcp_header_len << 26) |
505 ntohl(TCP_FLAG_ACK) |
506 snd_wnd);
507}
508
40efc6fa 509static inline void tcp_fast_path_on(struct tcp_sock *tp)
1da177e4
LT		 510{
511 __tcp_fast_path_on(tp, tp->snd_wnd >> tp->rx_opt.snd_wscale);
512}
513
9e412ba7 514static inline void tcp_fast_path_check(struct sock *sk)
1da177e4 515{
9e412ba7
IJ		 516 struct tcp_sock *tp = tcp_sk(sk);
517
b03efcfb 518 if (skb_queue_empty(&tp->out_of_order_queue) &&
1da177e4
LT		 519 tp->rcv_wnd &&
520 atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf &&
521 !tp->urg_data)
522 tcp_fast_path_on(tp);
523}
524
525/* Compute the actual receive window we are currently advertising.
526 * Rcv_nxt can be after the window if our peer push more data
527 * than the offered window.
528 */
40efc6fa 529static inline u32 tcp_receive_window(const struct tcp_sock *tp)
1da177e4
LT		 530{
531 s32 win = tp->rcv_wup + tp->rcv_wnd - tp->rcv_nxt;
532
533 if (win < 0)
534 win = 0;
535 return (u32) win;
536}
537
538/* Choose a new window, without checks for shrinking, and without
539 * scaling applied to the result. The caller does these things
540 * if necessary. This is a "raw" window selection.
541 */
542extern u32 __tcp_select_window(struct sock *sk);
543
544/* TCP timestamps are only 32-bits, this causes a slight
545 * complication on 64-bit systems since we store a snapshot
31f34269
SH		 546 * of jiffies in the buffer control blocks below. We decided
547 * to use only the low 32-bits of jiffies and hide the ugly
1da177e4
LT		 548 * casts with the following macro.
549 */
550#define tcp_time_stamp ((__u32)(jiffies))
551
caa20d9a 552/* This is what the send packet queuing engine uses to pass
1da177e4
LT		 553 * TCP per-packet control information to the transmission
554 * code. We also store the host-order sequence numbers in
555 * here too. This is 36 bytes on 32-bit architectures,
556 * 40 bytes on 64-bit machines, if this grows please adjust
557 * skbuff.h:skbuff->cb[xxx] size appropriately.
558 */
559struct tcp_skb_cb {
560 union {
561 struct inet_skb_parm h4;
562#if defined(CONFIG_IPV6) || defined (CONFIG_IPV6_MODULE)
563 struct inet6_skb_parm h6;
564#endif
565 } header; /* For incoming frames */
566 __u32 seq; /* Starting sequence number */
567 __u32 end_seq; /* SEQ + FIN + SYN + datalen */
568 __u32 when; /* used to compute rtt's */
569 __u8 flags; /* TCP header flags. */
570
571 /* NOTE: These must match up to the flags byte in a
572 * real TCP header.
573 */
574#define TCPCB_FLAG_FIN 0x01
575#define TCPCB_FLAG_SYN 0x02
576#define TCPCB_FLAG_RST 0x04
577#define TCPCB_FLAG_PSH 0x08
578#define TCPCB_FLAG_ACK 0x10
579#define TCPCB_FLAG_URG 0x20
580#define TCPCB_FLAG_ECE 0x40
581#define TCPCB_FLAG_CWR 0x80
582
583 __u8 sacked; /* State flags for SACK/FACK. */
584#define TCPCB_SACKED_ACKED 0x01 /* SKB ACK'd by a SACK block */
585#define TCPCB_SACKED_RETRANS 0x02 /* SKB retransmitted */
586#define TCPCB_LOST 0x04 /* SKB is lost */
587#define TCPCB_TAGBITS 0x07 /* All tag bits */
588
589#define TCPCB_EVER_RETRANS 0x80 /* Ever retransmitted frame */
590#define TCPCB_RETRANS (TCPCB_SACKED_RETRANS|TCPCB_EVER_RETRANS)
591
1da177e4
LT		 592 __u16 urg_ptr; /* Valid w/URG flags is set. */
593 __u32 ack_seq; /* Sequence number ACK'd */
594};
595
596#define TCP_SKB_CB(__skb) ((struct tcp_skb_cb *)&((__skb)->cb[0]))
597
1da177e4
LT		 598/* Due to TSO, an SKB can be composed of multiple actual
599 * packets. To keep these tracked properly, we use this.
600 */
601static inline int tcp_skb_pcount(const struct sk_buff *skb)
602{
7967168c 603 return skb_shinfo(skb)->gso_segs;
1da177e4
LT		 604}
605
606/* This is valid iff tcp_skb_pcount() > 1. */
607static inline int tcp_skb_mss(const struct sk_buff *skb)
608{
7967168c 609 return skb_shinfo(skb)->gso_size;
1da177e4
LT		 610}
611
af610b4c 612static inline void tcp_dec_pcount_approx_int(__u32 *count, const int decr)
1da177e4
LT		 613{
614 if (*count) {
af610b4c 615 *count -= decr;
1da177e4
LT		 616 if ((int)*count < 0)
617 *count = 0;
618 }
619}
620
af610b4c
IJ		 621static inline void tcp_dec_pcount_approx(__u32 *count,
622 const struct sk_buff *skb)
623{
624 tcp_dec_pcount_approx_int(count, tcp_skb_pcount(skb));
625}
626
317a76f9
SH		 627/* Events passed to congestion control interface */
628enum tcp_ca_event {
629 CA_EVENT_TX_START, /* first transmit when no packets in flight */
630 CA_EVENT_CWND_RESTART, /* congestion window restart */
631 CA_EVENT_COMPLETE_CWR, /* end of congestion recovery */
632 CA_EVENT_FRTO, /* fast recovery timeout */
633 CA_EVENT_LOSS, /* loss timeout */
634 CA_EVENT_FAST_ACK, /* in sequence ack */
635 CA_EVENT_SLOW_ACK, /* other ack */
636};
637
638/*
639 * Interface for adding new TCP congestion control handlers
640 */
641#define TCP_CA_NAME_MAX 16
3ff825b2
SH		 642#define TCP_CA_MAX 128
643#define TCP_CA_BUF_MAX (TCP_CA_NAME_MAX*TCP_CA_MAX)
644
164891aa
SH		 645#define TCP_CONG_NON_RESTRICTED 0x1
646#define TCP_CONG_RTT_STAMP 0x2
647
317a76f9
SH		 648struct tcp_congestion_ops {
649 struct list_head list;
164891aa 650 unsigned long flags;
317a76f9
SH		 651
652 /* initialize private data (optional) */
6687e988 653 void (*init)(struct sock *sk);
317a76f9 654 /* cleanup private data (optional) */
6687e988 655 void (*release)(struct sock *sk);
317a76f9
SH		 656
657 /* return slow start threshold (required) */
6687e988 658 u32 (*ssthresh)(struct sock *sk);
317a76f9 659 /* lower bound for congestion window (optional) */
72dc5b92 660 u32 (*min_cwnd)(const struct sock *sk);
317a76f9 661 /* do new cwnd calculation (required) */
c3a05c60 662 void (*cong_avoid)(struct sock *sk, u32 ack, u32 in_flight);
317a76f9 663 /* call before changing ca_state (optional) */
6687e988 664 void (*set_state)(struct sock *sk, u8 new_state);
317a76f9 665 /* call when cwnd event occurs (optional) */
6687e988 666 void (*cwnd_event)(struct sock *sk, enum tcp_ca_event ev);
317a76f9 667 /* new value of cwnd after loss (optional) */
6687e988 668 u32 (*undo_cwnd)(struct sock *sk);
317a76f9 669 /* hook for packet ack accounting (optional) */
30cfd0ba 670 void (*pkts_acked)(struct sock *sk, u32 num_acked, s32 rtt_us);
73c1f4a0 671 /* get info for inet_diag (optional) */
6687e988 672 void (*get_info)(struct sock *sk, u32 ext, struct sk_buff *skb);
317a76f9
SH		 673
674 char name[TCP_CA_NAME_MAX];
675 struct module *owner;
676};
677
678extern int tcp_register_congestion_control(struct tcp_congestion_ops *type);
679extern void tcp_unregister_congestion_control(struct tcp_congestion_ops *type);
680
6687e988
ACM		 681extern void tcp_init_congestion_control(struct sock *sk);
682extern void tcp_cleanup_congestion_control(struct sock *sk);
317a76f9
SH		 683extern int tcp_set_default_congestion_control(const char *name);
684extern void tcp_get_default_congestion_control(char *name);
3ff825b2 685extern void tcp_get_available_congestion_control(char *buf, size_t len);
ce7bc3bf
SH		 686extern void tcp_get_allowed_congestion_control(char *buf, size_t len);
687extern int tcp_set_allowed_congestion_control(char *allowed);
6687e988 688extern int tcp_set_congestion_control(struct sock *sk, const char *name);
40efc6fa 689extern void tcp_slow_start(struct tcp_sock *tp);
317a76f9 690
5f8ef48d 691extern struct tcp_congestion_ops tcp_init_congestion_ops;
6687e988 692extern u32 tcp_reno_ssthresh(struct sock *sk);
c3a05c60 693extern void tcp_reno_cong_avoid(struct sock *sk, u32 ack, u32 in_flight);
72dc5b92 694extern u32 tcp_reno_min_cwnd(const struct sock *sk);
a8acfbac 695extern struct tcp_congestion_ops tcp_reno;
317a76f9 696
6687e988 697static inline void tcp_set_ca_state(struct sock *sk, const u8 ca_state)
317a76f9 698{
6687e988
ACM		 699 struct inet_connection_sock *icsk = inet_csk(sk);
700
701 if (icsk->icsk_ca_ops->set_state)
702 icsk->icsk_ca_ops->set_state(sk, ca_state);
703 icsk->icsk_ca_state = ca_state;
317a76f9
SH		 704}
705
6687e988 706static inline void tcp_ca_event(struct sock *sk, const enum tcp_ca_event event)
317a76f9 707{
6687e988
ACM		 708 const struct inet_connection_sock *icsk = inet_csk(sk);
709
710 if (icsk->icsk_ca_ops->cwnd_event)
711 icsk->icsk_ca_ops->cwnd_event(sk, event);
317a76f9
SH		 712}
713
e60402d0
IJ		 714/* These functions determine how the current flow behaves in respect of SACK
715 * handling. SACK is negotiated with the peer, and therefore it can vary
716 * between different flows.
717 *
718 * tcp_is_sack - SACK enabled
719 * tcp_is_reno - No SACK
720 * tcp_is_fack - FACK enabled, implies SACK enabled
721 */
722static inline int tcp_is_sack(const struct tcp_sock *tp)
723{
724 return tp->rx_opt.sack_ok;
725}
726
727static inline int tcp_is_reno(const struct tcp_sock *tp)
728{
729 return !tcp_is_sack(tp);
730}
731
732static inline int tcp_is_fack(const struct tcp_sock *tp)
733{
734 return tp->rx_opt.sack_ok & 2;
735}
736
737static inline void tcp_enable_fack(struct tcp_sock *tp)
738{
739 tp->rx_opt.sack_ok |= 2;
740}
741
83ae4088
IJ		 742static inline unsigned int tcp_left_out(const struct tcp_sock *tp)
743{
744 return tp->sacked_out + tp->lost_out;
745}
746
1da177e4
LT		 747/* This determines how many packets are "in the network" to the best
748 * of our knowledge. In many cases it is conservative, but where
749 * detailed information is available from the receiver (via SACK
750 * blocks etc.) we can make more aggressive calculations.
751 *
752 * Use this for decisions involving congestion control, use just
753 * tp->packets_out to determine if the send queue is empty or not.
754 *
755 * Read this equation as:
756 *
757 * "Packets sent once on transmission queue" MINUS
758 * "Packets left network, but not honestly ACKed yet" PLUS
759 * "Packets fast retransmitted"
760 */
40efc6fa 761static inline unsigned int tcp_packets_in_flight(const struct tcp_sock *tp)
1da177e4 762{
83ae4088 763 return tp->packets_out - tcp_left_out(tp) + tp->retrans_out;
1da177e4
LT		 764}
765
882bebaa
IJ		 766extern int tcp_limit_reno_sacked(struct tcp_sock *tp);
767
1da177e4
LT		 768/* If cwnd > ssthresh, we may raise ssthresh to be half-way to cwnd.
769 * The exception is rate halving phase, when cwnd is decreasing towards
770 * ssthresh.
771 */
6687e988 772static inline __u32 tcp_current_ssthresh(const struct sock *sk)
1da177e4 773{
6687e988
ACM		 774 const struct tcp_sock *tp = tcp_sk(sk);
775 if ((1 << inet_csk(sk)->icsk_ca_state) & (TCPF_CA_CWR | TCPF_CA_Recovery))
1da177e4
LT		 776 return tp->snd_ssthresh;
777 else
778 return max(tp->snd_ssthresh,
779 ((tp->snd_cwnd >> 1) +
780 (tp->snd_cwnd >> 2)));
781}
782
b9c4595b
IJ		 783/* Use define here intentionally to get WARN_ON location shown at the caller */
784#define tcp_verify_left_out(tp) WARN_ON(tcp_left_out(tp) > tp->packets_out)
1da177e4 785
3cfe3baa 786extern void tcp_enter_cwr(struct sock *sk, const int set_ssthresh);
1da177e4
LT		 787extern __u32 tcp_init_cwnd(struct tcp_sock *tp, struct dst_entry *dst);
788
789/* Slow start with delack produces 3 packets of burst, so that
790 * it is safe "de facto".
791 */
792static __inline__ __u32 tcp_max_burst(const struct tcp_sock *tp)
793{
794 return 3;
795}
796
90840def
IJ		 797/* Returns end sequence number of the receiver's advertised window */
798static inline u32 tcp_wnd_end(const struct tcp_sock *tp)
799{
800 return tp->snd_una + tp->snd_wnd;
801}
cea14e0e 802extern int tcp_is_cwnd_limited(const struct sock *sk, u32 in_flight);
f4805ede 803
c1bd24b7 804static inline void tcp_minshall_update(struct tcp_sock *tp, unsigned int mss,
40efc6fa 805 const struct sk_buff *skb)
1da177e4
LT		 806{
807 if (skb->len < mss)
808 tp->snd_sml = TCP_SKB_CB(skb)->end_seq;
809}
810
9e412ba7 811static inline void tcp_check_probe_timer(struct sock *sk)
1da177e4 812{
9e412ba7 813 struct tcp_sock *tp = tcp_sk(sk);
463c84b9 814 const struct inet_connection_sock *icsk = inet_csk(sk);
9e412ba7 815
463c84b9 816 if (!tp->packets_out && !icsk->icsk_pending)
3f421baa
ACM		 817 inet_csk_reset_xmit_timer(sk, ICSK_TIME_PROBE0,
818 icsk->icsk_rto, TCP_RTO_MAX);
1da177e4
LT		 819}
820
9e412ba7 821static inline void tcp_push_pending_frames(struct sock *sk)
1da177e4 822{
9e412ba7
IJ		 823 struct tcp_sock *tp = tcp_sk(sk);
824
825 __tcp_push_pending_frames(sk, tcp_current_mss(sk, 1), tp->nonagle);
1da177e4
LT		 826}
827
40efc6fa 828static inline void tcp_init_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 829{
830 tp->snd_wl1 = seq;
831}
832
40efc6fa 833static inline void tcp_update_wl(struct tcp_sock *tp, u32 ack, u32 seq)
1da177e4
LT		 834{
835 tp->snd_wl1 = seq;
836}
837
1da177e4
LT		 838/*
839 * Calculate(/check) TCP checksum
840 */
ba7808ea
FD		 841static inline __sum16 tcp_v4_check(int len, __be32 saddr,
842 __be32 daddr, __wsum base)
1da177e4
LT		 843{
844 return csum_tcpudp_magic(saddr,daddr,len,IPPROTO_TCP,base);
845}
846
b51655b9 847static inline __sum16 __tcp_checksum_complete(struct sk_buff *skb)
1da177e4 848{
fb286bb2 849 return __skb_checksum_complete(skb);
1da177e4
LT		 850}
851
40efc6fa 852static inline int tcp_checksum_complete(struct sk_buff *skb)
1da177e4 853{
60476372 854 return !skb_csum_unnecessary(skb) &&
1da177e4
LT		 855 __tcp_checksum_complete(skb);
856}
857
858/* Prequeue for VJ style copy to user, combined with checksumming. */
859
40efc6fa 860static inline void tcp_prequeue_init(struct tcp_sock *tp)
1da177e4
LT		 861{
862 tp->ucopy.task = NULL;
863 tp->ucopy.len = 0;
864 tp->ucopy.memory = 0;
865 skb_queue_head_init(&tp->ucopy.prequeue);
97fc2f08
CL		 866#ifdef CONFIG_NET_DMA
867 tp->ucopy.dma_chan = NULL;
868 tp->ucopy.wakeup = 0;
869 tp->ucopy.pinned_list = NULL;
870 tp->ucopy.dma_cookie = 0;
871#endif
1da177e4
LT		 872}
873
874/* Packet is added to VJ-style prequeue for processing in process
875 * context, if a reader task is waiting. Apparently, this exciting
876 * idea (VJ's mail "Re: query about TCP header on tcp-ip" of 07 Sep 93)
877 * failed somewhere. Latency? Burstiness? Well, at least now we will
878 * see, why it failed. 8)8) --ANK
879 *
880 * NOTE: is this not too big to inline?
881 */
40efc6fa 882static inline int tcp_prequeue(struct sock *sk, struct sk_buff *skb)
1da177e4
LT		 883{
884 struct tcp_sock *tp = tcp_sk(sk);
885
886 if (!sysctl_tcp_low_latency && tp->ucopy.task) {
887 __skb_queue_tail(&tp->ucopy.prequeue, skb);
888 tp->ucopy.memory += skb->truesize;
889 if (tp->ucopy.memory > sk->sk_rcvbuf) {
890 struct sk_buff *skb1;
891
892 BUG_ON(sock_owned_by_user(sk));
893
894 while ((skb1 = __skb_dequeue(&tp->ucopy.prequeue)) != NULL) {
895 sk->sk_backlog_rcv(sk, skb1);
896 NET_INC_STATS_BH(LINUX_MIB_TCPPREQUEUEDROPPED);
897 }
898
899 tp->ucopy.memory = 0;
900 } else if (skb_queue_len(&tp->ucopy.prequeue) == 1) {
901 wake_up_interruptible(sk->sk_sleep);
463c84b9
ACM		 902 if (!inet_csk_ack_scheduled(sk))
903 inet_csk_reset_xmit_timer(sk, ICSK_TIME_DACK,
3f421baa
ACM		 904 (3 * TCP_RTO_MIN) / 4,
905 TCP_RTO_MAX);
1da177e4
LT		 906 }
907 return 1;
908 }
909 return 0;
910}
911
912
913#undef STATE_TRACE
914
915#ifdef STATE_TRACE
916static const char *statename[]={
917 "Unused","Established","Syn Sent","Syn Recv",
918 "Fin Wait 1","Fin Wait 2","Time Wait", "Close",
919 "Close Wait","Last ACK","Listen","Closing"
920};
921#endif
490d5046 922extern void tcp_set_state(struct sock *sk, int state);
1da177e4 923
4ac02bab 924extern void tcp_done(struct sock *sk);
1da177e4 925
40efc6fa 926static inline void tcp_sack_reset(struct tcp_options_received *rx_opt)
1da177e4
LT		 927{
928 rx_opt->dsack = 0;
929 rx_opt->eff_sacks = 0;
930 rx_opt->num_sacks = 0;
931}
932
1da177e4
LT		 933/* Determine a window scaling and initial window to offer. */
934extern void tcp_select_initial_window(int __space, __u32 mss,
935 __u32 *rcv_wnd, __u32 *window_clamp,
936 int wscale_ok, __u8 *rcv_wscale);
937
938static inline int tcp_win_from_space(int space)
939{
940 return sysctl_tcp_adv_win_scale<=0 ?
941 (space>>(-sysctl_tcp_adv_win_scale)) :
942 space - (space>>sysctl_tcp_adv_win_scale);
943}
944
945/* Note: caller must be prepared to deal with negative returns */
946static inline int tcp_space(const struct sock *sk)
947{
948 return tcp_win_from_space(sk->sk_rcvbuf -
949 atomic_read(&sk->sk_rmem_alloc));
950}
951
952static inline int tcp_full_space(const struct sock *sk)
953{
954 return tcp_win_from_space(sk->sk_rcvbuf);
955}
956
40efc6fa
SH		 957static inline void tcp_openreq_init(struct request_sock *req,
958 struct tcp_options_received *rx_opt,
959 struct sk_buff *skb)
1da177e4 960{
2e6599cb
ACM		 961 struct inet_request_sock *ireq = inet_rsk(req);
962
1da177e4 963 req->rcv_wnd = 0; /* So that tcp_send_synack() knows! */
4dfc2817 964 req->cookie_ts = 0;
2e6599cb 965 tcp_rsk(req)->rcv_isn = TCP_SKB_CB(skb)->seq;
1da177e4
LT		 966 req->mss = rx_opt->mss_clamp;
967 req->ts_recent = rx_opt->saw_tstamp ? rx_opt->rcv_tsval : 0;
2e6599cb
ACM		 968 ireq->tstamp_ok = rx_opt->tstamp_ok;
969 ireq->sack_ok = rx_opt->sack_ok;
970 ireq->snd_wscale = rx_opt->snd_wscale;
971 ireq->wscale_ok = rx_opt->wscale_ok;
972 ireq->acked = 0;
973 ireq->ecn_ok = 0;
aa8223c7 974 ireq->rmt_port = tcp_hdr(skb)->source;
1da177e4
LT		 975}
976
977extern void tcp_enter_memory_pressure(void);
978
1da177e4
LT		 979static inline int keepalive_intvl_when(const struct tcp_sock *tp)
980{
981 return tp->keepalive_intvl ? : sysctl_tcp_keepalive_intvl;
982}
983
984static inline int keepalive_time_when(const struct tcp_sock *tp)
985{
986 return tp->keepalive_time ? : sysctl_tcp_keepalive_time;
987}
988
463c84b9 989static inline int tcp_fin_time(const struct sock *sk)
1da177e4 990{
463c84b9
ACM		 991 int fin_timeout = tcp_sk(sk)->linger2 ? : sysctl_tcp_fin_timeout;
992 const int rto = inet_csk(sk)->icsk_rto;
1da177e4 993
463c84b9
ACM		 994 if (fin_timeout < (rto << 2) - (rto >> 1))
995 fin_timeout = (rto << 2) - (rto >> 1);
1da177e4
LT		 996
997 return fin_timeout;
998}
999
1000static inline int tcp_paws_check(const struct tcp_options_received *rx_opt, int rst)
1001{
1002 if ((s32)(rx_opt->rcv_tsval - rx_opt->ts_recent) >= 0)
1003 return 0;
9d729f72 1004 if (get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_24DAYS)
1da177e4
LT		 1005 return 0;
1006
1007 /* RST segments are not recommended to carry timestamp,
1008 and, if they do, it is recommended to ignore PAWS because
1009 "their cleanup function should take precedence over timestamps."
1010 Certainly, it is mistake. It is necessary to understand the reasons
1011 of this constraint to relax it: if peer reboots, clock may go
1012 out-of-sync and half-open connections will not be reset.
1013 Actually, the problem would be not existing if all
1014 the implementations followed draft about maintaining clock
1015 via reboots. Linux-2.2 DOES NOT!
1016
1017 However, we can relax time bounds for RST segments to MSL.
1018 */
9d729f72 1019 if (rst && get_seconds() >= rx_opt->ts_recent_stamp + TCP_PAWS_MSL)
1da177e4
LT		 1020 return 0;
1021 return 1;
1022}
1023
1da177e4
LT		 1024#define TCP_CHECK_TIMER(sk) do { } while (0)
1025
1da177e4
LT		 1026static inline void tcp_mib_init(void)
1027{
1028 /* See RFC 2012 */
1029 TCP_ADD_STATS_USER(TCP_MIB_RTOALGORITHM, 1);
1030 TCP_ADD_STATS_USER(TCP_MIB_RTOMIN, TCP_RTO_MIN*1000/HZ);
1031 TCP_ADD_STATS_USER(TCP_MIB_RTOMAX, TCP_RTO_MAX*1000/HZ);
1032 TCP_ADD_STATS_USER(TCP_MIB_MAXCONN, -1);
1033}
1034
5af4ec23 1035/* from STCP */
0800f170
DM		 1036static inline void tcp_clear_retrans_hints_partial(struct tcp_sock *tp)
1037{
6a438bbe
SH		 1038 tp->lost_skb_hint = NULL;
1039 tp->scoreboard_skb_hint = NULL;
1040 tp->retransmit_skb_hint = NULL;
1041 tp->forward_skb_hint = NULL;
b7689205
IJ		 1042}
1043
0800f170
DM		 1044static inline void tcp_clear_all_retrans_hints(struct tcp_sock *tp)
1045{
b7689205 1046 tcp_clear_retrans_hints_partial(tp);
6a438bbe
SH		 1047}
1048
cfb6eeb4
YH		 1049/* MD5 Signature */
1050struct crypto_hash;
1051
1052/* - key database */
1053struct tcp_md5sig_key {
1054 u8 *key;
1055 u8 keylen;
1056};
1057
1058struct tcp4_md5sig_key {
f8ab18d2 1059 struct tcp_md5sig_key base;
cfb6eeb4
YH		 1060 __be32 addr;
1061};
1062
1063struct tcp6_md5sig_key {
f8ab18d2 1064 struct tcp_md5sig_key base;
cfb6eeb4
YH		 1065#if 0
1066 u32 scope_id; /* XXX */
1067#endif
1068 struct in6_addr addr;
1069};
1070
1071/* - sock block */
1072struct tcp_md5sig_info {
1073 struct tcp4_md5sig_key *keys4;
1074#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1075 struct tcp6_md5sig_key *keys6;
1076 u32 entries6;
1077 u32 alloced6;
1078#endif
1079 u32 entries4;
1080 u32 alloced4;
1081};
1082
1083/* - pseudo header */
1084struct tcp4_pseudohdr {
1085 __be32 saddr;
1086 __be32 daddr;
1087 __u8 pad;
1088 __u8 protocol;
1089 __be16 len;
1090};
1091
1092struct tcp6_pseudohdr {
1093 struct in6_addr saddr;
1094 struct in6_addr daddr;
1095 __be32 len;
1096 __be32 protocol; /* including padding */
1097};
1098
1099union tcp_md5sum_block {
1100 struct tcp4_pseudohdr ip4;
1101#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
1102 struct tcp6_pseudohdr ip6;
1103#endif
1104};
1105
1106/* - pool: digest algorithm, hash description and scratch buffer */
1107struct tcp_md5sig_pool {
1108 struct hash_desc md5_desc;
1109 union tcp_md5sum_block md5_blk;
1110};
1111
1112#define TCP_MD5SIG_MAXKEYS (~(u32)0) /* really?! */
1113
1114/* - functions */
1115extern int tcp_v4_calc_md5_hash(char *md5_hash,
1116 struct tcp_md5sig_key *key,
1117 struct sock *sk,
1118 struct dst_entry *dst,
1119 struct request_sock *req,
1120 struct tcphdr *th,
9cb5734e
YH		 1121 int protocol,
1122 unsigned int tcplen);
cfb6eeb4
YH		 1123extern struct tcp_md5sig_key *tcp_v4_md5_lookup(struct sock *sk,
1124 struct sock *addr_sk);
1125
1126extern int tcp_v4_md5_do_add(struct sock *sk,
1127 __be32 addr,
1128 u8 *newkey,
1129 u8 newkeylen);
1130
1131extern int tcp_v4_md5_do_del(struct sock *sk,
8e5200f5 1132 __be32 addr);
cfb6eeb4
YH		 1133
1134extern struct tcp_md5sig_pool **tcp_alloc_md5sig_pool(void);
1135extern void tcp_free_md5sig_pool(void);
1136
1137extern struct tcp_md5sig_pool *__tcp_get_md5sig_pool(int cpu);
1138extern void __tcp_put_md5sig_pool(void);
1139
1140static inline
1141struct tcp_md5sig_pool *tcp_get_md5sig_pool(void)
1142{
1143 int cpu = get_cpu();
1144 struct tcp_md5sig_pool *ret = __tcp_get_md5sig_pool(cpu);
1145 if (!ret)
1146 put_cpu();
1147 return ret;
1148}
1149
1150static inline void tcp_put_md5sig_pool(void)
1151{
1152 __tcp_put_md5sig_pool();
1153 put_cpu();
1154}
1155
fe067e8a
DM		 1156/* write queue abstraction */
1157static inline void tcp_write_queue_purge(struct sock *sk)
1158{
1159 struct sk_buff *skb;
1160
1161 while ((skb = __skb_dequeue(&sk->sk_write_queue)) != NULL)
3ab224be
HA		 1162 sk_wmem_free_skb(sk, skb);
1163 sk_mem_reclaim(sk);
fe067e8a
DM		 1164}
1165
1166static inline struct sk_buff *tcp_write_queue_head(struct sock *sk)
1167{
1168 struct sk_buff *skb = sk->sk_write_queue.next;
1169 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1170 return NULL;
1171 return skb;
1172}
1173
1174static inline struct sk_buff *tcp_write_queue_tail(struct sock *sk)
1175{
1176 struct sk_buff *skb = sk->sk_write_queue.prev;
1177 if (skb == (struct sk_buff *) &sk->sk_write_queue)
1178 return NULL;
1179 return skb;
1180}
1181
1182static inline struct sk_buff *tcp_write_queue_next(struct sock *sk, struct sk_buff *skb)
1183{
1184 return skb->next;
1185}
1186
1187#define tcp_for_write_queue(skb, sk) \
1188 for (skb = (sk)->sk_write_queue.next; \
1189 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1190 skb = skb->next)
1191
1192#define tcp_for_write_queue_from(skb, sk) \
1193 for (; (skb != (struct sk_buff *)&(sk)->sk_write_queue);\
1194 skb = skb->next)
1195
234b6860
IJ		 1196#define tcp_for_write_queue_from_safe(skb, tmp, sk) \
1197 for (tmp = skb->next; \
1198 (skb != (struct sk_buff *)&(sk)->sk_write_queue); \
1199 skb = tmp, tmp = skb->next)
1200
fe067e8a
DM		 1201static inline struct sk_buff *tcp_send_head(struct sock *sk)
1202{
1203 return sk->sk_send_head;
1204}
1205
1206static inline void tcp_advance_send_head(struct sock *sk, struct sk_buff *skb)
1207{
1208 sk->sk_send_head = skb->next;
1209 if (sk->sk_send_head == (struct sk_buff *)&sk->sk_write_queue)
1210 sk->sk_send_head = NULL;
1211}
1212
1213static inline void tcp_check_send_head(struct sock *sk, struct sk_buff *skb_unlinked)
1214{
1215 if (sk->sk_send_head == skb_unlinked)
1216 sk->sk_send_head = NULL;
1217}
1218
1219static inline void tcp_init_send_head(struct sock *sk)
1220{
1221 sk->sk_send_head = NULL;
1222}
1223
1224static inline void __tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1225{
1226 __skb_queue_tail(&sk->sk_write_queue, skb);
1227}
1228
1229static inline void tcp_add_write_queue_tail(struct sock *sk, struct sk_buff *skb)
1230{
1231 __tcp_add_write_queue_tail(sk, skb);
1232
1233 /* Queue it, remembering where we must start sending. */
6859d494 1234 if (sk->sk_send_head == NULL) {
fe067e8a 1235 sk->sk_send_head = skb;
6859d494
IJ		 1236
1237 if (tcp_sk(sk)->highest_sack == NULL)
1238 tcp_sk(sk)->highest_sack = skb;
1239 }
fe067e8a
DM		 1240}
1241
1242static inline void __tcp_add_write_queue_head(struct sock *sk, struct sk_buff *skb)
1243{
1244 __skb_queue_head(&sk->sk_write_queue, skb);
1245}
1246
1247/* Insert buff after skb on the write queue of sk. */
1248static inline void tcp_insert_write_queue_after(struct sk_buff *skb,
1249 struct sk_buff *buff,
1250 struct sock *sk)
1251{
7de6c033 1252 __skb_queue_after(&sk->sk_write_queue, skb, buff);
fe067e8a
DM		 1253}
1254
1255/* Insert skb between prev and next on the write queue of sk. */
1256static inline void tcp_insert_write_queue_before(struct sk_buff *new,
1257 struct sk_buff *skb,
1258 struct sock *sk)
1259{
1260 __skb_insert(new, skb->prev, skb, &sk->sk_write_queue);