2 * Copyright 2004 Sun Microsystems, Inc. All rights reserved.
3 * Use is subject to license terms.
7 * Copyright (c) 1982, 1986 Regents of the University of California.
8 * All rights reserved. The Berkeley software License Agreement
9 * specifies the terms and conditions for redistribution.
15 #pragma ident "%Z%%M% %I% %E% SMI"
24 #define TCPS_ALL_ACKED -7 /* Internal state for retransmissions */
25 #define TCPS_CLOSED -6
26 #define TCPS_IDLE -5 /* idle (opened, but not bound) */
27 #define TCPS_BOUND -4 /* bound, ready to connect or accept */
28 #define TCPS_LISTEN -3 /* listening for connection */
29 #define TCPS_SYN_SENT -2 /* active, have sent syn */
30 #define TCPS_SYN_RCVD -1 /* have received syn (and sent ours) */
31 /* states < TCPS_ESTABLISHED are those where connections not established */
32 #define TCPS_ESTABLISHED 0 /* established */
33 #define TCPS_CLOSE_WAIT 1 /* rcvd fin, waiting for close */
34 /* states > TCPS_CLOSE_WAIT are those where user has closed */
35 #define TCPS_FIN_WAIT_1 2 /* have closed and sent fin */
36 #define TCPS_CLOSING 3 /* closed, xchd FIN, await FIN ACK */
37 #define TCPS_LAST_ACK 4 /* had fin and close; await FIN ACK */
38 /* states > TCPS_CLOSE_WAIT && < TCPS_FIN_WAIT_2 await ACK of FIN */
39 #define TCPS_FIN_WAIT_2 5 /* have closed, fin is acked */
40 #define TCPS_TIME_WAIT 6 /* in 2*msl quiet wait after close */
43 * Internal flags used in conjunction with the packet header flags.
44 * Used in tcp_rput_data to keep track of what needs to be done.
46 #define TH_LIMIT_XMIT 0x0400 /* Limited xmit is needed */
47 #define TH_XMIT_NEEDED 0x0800 /* Window opened - send queued data */
48 #define TH_REXMIT_NEEDED 0x1000 /* Time expired for unacked data */
49 #define TH_ACK_NEEDED 0x2000 /* Send an ack now. */
50 #define TH_NEED_SACK_REXMIT 0x4000 /* Use SACK info to retransmission */
51 #define TH_TIMER_NEEDED 0x8000 /* Start the delayed ack/push bit timer */
54 * Special Magic number used by TCP to issue a callback to recvfrom() in the
55 * form of a dummy inetgram.
57 #define TCP_CALLB_MAGIC_ID 0xFFFF
60 * TCP sequence numbers are 32 bit integers operated
61 * on with modular arithmetic. These macros can be
62 * used to compare such integers.
64 #define SEQ_LT(a, b) ((int32_t)((a)-(b)) < 0)
65 #define SEQ_LEQ(a, b) ((int32_t)((a)-(b)) <= 0)
66 #define SEQ_GT(a, b) ((int32_t)((a)-(b)) > 0)
67 #define SEQ_GEQ(a, b) ((int32_t)((a)-(b)) >= 0)
69 /* TCP Protocol header */
70 typedef struct tcphdr_s
{
71 uint8_t th_lport
[2]; /* Source port */
72 uint8_t th_fport
[2]; /* Destination port */
73 uint8_t th_seq
[4]; /* Sequence number */
74 uint8_t th_ack
[4]; /* Acknowledgement number */
75 uint8_t th_offset_and_rsrvd
[1]; /* Offset to the packet data */
77 uint8_t th_win
[2]; /* Allocation number */
78 uint8_t th_sum
[2]; /* TCP checksum */
79 uint8_t th_urp
[2]; /* Urgent pointer */
82 #define TCP_HDR_LENGTH(tcph) (((tcph)->th_offset_and_rsrvd[0] >>2) &(0xF << 2))
83 #define TCP_MAX_COMBINED_HEADER_LENGTH (60 + 60) /* Maxed out ip + tcp */
84 #define TCP_MAX_IP_OPTIONS_LENGTH (60 - IP_SIMPLE_HDR_LENGTH)
85 #define TCP_MAX_HDR_LENGTH 60
86 #define TCP_MAX_TCP_OPTIONS_LENGTH (60 - sizeof (tcph_t))
87 #define TCP_MIN_HEADER_LENGTH 20
88 #define TCP_MAXWIN 65535
89 #define TCP_PORT_LEN sizeof (in_port_t)
90 #define TCP_MAX_WINSHIFT 14
91 #define TCP_MAX_LARGEWIN (TCP_MAXWIN << TCP_MAX_WINSHIFT)
93 #define TCPIP_HDR_LENGTH(mp, n) \
94 (n) = IPH_HDR_LENGTH((mp)->b_rptr), \
95 (n) += TCP_HDR_LENGTH((tcph_t *)&(mp)->b_rptr[(n)])
97 /* TCP Protocol header (used if the header is known to be 32-bit aligned) */
98 typedef struct tcphdra_s
{
99 in_port_t tha_lport
; /* Source port */
100 in_port_t tha_fport
; /* Destination port */
101 uint32_t tha_seq
; /* Sequence number */
102 uint32_t tha_ack
; /* Acknowledgement number */
103 uint8_t tha_offset_and_reserved
; /* Offset to the packet data */
105 uint16_t tha_win
; /* Allocation number */
106 uint16_t tha_sum
; /* TCP checksum */
107 uint16_t tha_urp
; /* Urgent pointer */
110 typedef struct tcp_s
{
111 struct tcp_s
*tcp_time_wait_next
; /* Next TCP in TIME_WAIT list */
112 struct tcp_s
*tcp_time_wait_prev
; /* Prev TCP in TIME_WAIT list */
113 uint32_t tcp_max_swnd
; /* Maximum swnd we have seen */
114 uint32_t tcp_suna
; /* Sender unacknowledged */
115 uint32_t tcp_csuna
; /* Clear (no rexmits in window) suna */
116 uint32_t tcp_snxt
; /* Senders next seq num */
117 uint32_t tcp_swnd
; /* Senders window (relative to suna) */
118 uint32_t tcp_mss
; /* Max segment size */
119 uint32_t tcp_iss
; /* Initial send seq num */
120 uint32_t tcp_rnxt
; /* Seq we expect to recv next */
121 uint32_t tcp_rwnd
; /* Current receive window */
122 uint32_t tcp_rwnd_max
; /* Max receive window */
124 uint32_t tcp_irs
; /* Initial recv seq num */
125 uint32_t tcp_fss
; /* Final/fin send seq num */
127 uint32_t tcp_swl1
; /* These help us avoid using stale */
128 uint32_t tcp_swl2
; /* packets to update state */
130 uint32_t tcp_cwnd
; /* Congestion window */
131 int32_t tcp_cwnd_cnt
; /* cwnd cnt in congestion avoidance */
132 uint32_t tcp_cwnd_ssthresh
; /* Congestion window */
133 uint32_t tcp_cwnd_max
;
135 int32_t tcp_snd_burst
; /* Send burst factor */
138 int32_t tcp_rcv_ws
; /* My window scale power */
139 int32_t tcp_snd_ws
; /* Sender's window scale power */
140 uint32_t tcp_ts_recent
; /* Timestamp of earliest unacked */
143 uint32_t tcp_if_mtu
; /* Outgoing interface MTU. */
145 uint32_t tcp_rtt_sa
; /* Round trip smoothed average */
146 uint32_t tcp_rtt_sd
; /* Round trip smoothed deviation */
147 uint32_t tcp_rtt_update
; /* Round trip update(s) */
148 uint32_t tcp_ms_we_have_waited
; /* Total retrans time */
149 uint32_t tcp_rto
; /* Round trip timeout */
150 uint32_t tcp_rto_timeout
; /* RTT timeout time */
151 uint32_t tcp_time_wait_expire
;
152 /* time in hz when t/w expires */
153 uint32_t tcp_last_rcv_lbolt
;
154 /* lbolt on last packet, used for PAWS */
155 int tcp_lingertime
; /* Close linger time (in seconds) */
157 uint32_t tcp_first_timer_threshold
; /* When to prod IP */
158 uint32_t tcp_second_timer_threshold
; /* When to give up completely */
159 uint32_t tcp_first_ctimer_threshold
; /* 1st threshold when connecting */
160 uint32_t tcp_second_ctimer_threshold
; /* 2nd ... while connecting */
162 uint32_t tcp_rexmit_nxt
; /* Next rexmit seq num */
163 uint32_t tcp_rexmit_max
; /* Max retran seq num */
165 uint32_t tcp_naglim
; /* Tunable nagle limit */
166 uint32_t tcp_valid_bits
;
167 #define TCP_ISS_VALID 0x1 /* Is the tcp_iss seq num active? */
168 #define TCP_FSS_VALID 0x2 /* Is the tcp_fss seq num active? */
169 #define TCP_URG_VALID 0x4 /* Is the tcp_urg seq num active? */
170 #define TCP_OFO_FIN_VALID 0x8 /* Has TCP received an out of order FIN? */
172 int32_t tcp_xmit_hiwater
; /* Send buffer high water mark. */
173 int32_t tcp_xmit_lowater
; /* Send buffer low water mark. */
175 uchar_t tcp_timer_backoff
; /* Backoff shift count. */
176 uint32_t tcp_last_recv_time
; /* Last time we receive a segment. */
178 /* Fields arranged in approximate access order along main paths */
179 mblk_t
*tcp_xmit_head
; /* Head of rexmit list */
180 mblk_t
*tcp_xmit_last
; /* last valid data seen by tcp_wput */
181 uint32_t tcp_unsent
; /* # of bytes in hand that are unsent */
182 mblk_t
*tcp_xmit_tail
; /* Last rexmit data sent */
183 uint32_t tcp_xmit_tail_unsent
; /* # of unsent bytes in xmit_tail */
184 mblk_t
*tcp_rcv_list
; /* Queued until push or exceed */
185 mblk_t
*tcp_rcv_last_tail
; /* tcp_rcv_push_wait. */
186 uint32_t tcp_rcv_cnt
; /* tcp_rcv_list is b_next chain. */
188 uint32_t tcp_rack
; /* Seq # we have acked */
189 uint32_t tcp_rack_cnt
; /* # of bytes we have deferred ack */
191 mblk_t
*tcp_reass_head
; /* Out of order reassembly list head */
192 mblk_t
*tcp_reass_tail
; /* Out of order reassembly list tail */
194 tcp_sack_info_t
*tcp_sack_info
;
196 #define tcp_pipe tcp_sack_info->tcp_pipe
197 #define tcp_fack tcp_sack_info->tcp_fack
198 #define tcp_sack_snxt tcp_sack_info->tcp_sack_snxt
199 #define tcp_max_sack_blk tcp_sack_info->tcp_max_sack_blk
200 #define tcp_num_sack_blk tcp_sack_info->tcp_num_sack_blk
201 #define tcp_sack_list tcp_sack_info->tcp_sack_list
202 #define tcp_num_notsack_blk tcp_sack_info->tcp_num_notsack_blk
203 #define tcp_cnt_notsack_list tcp_sack_info->tcp_cnt_notsack_list
204 #define tcp_notsack_list tcp_sack_info->tcp_notsack_list
206 int tcp_conn_req_cnt_q0
; /* # of conn reqs in SYN_RCVD */
207 int tcp_conn_req_cnt_q
; /* # of conn reqs in ESTABLISHED */
208 int tcp_conn_req_max
; /* # of ESTABLISHED conn reqs allowed */
210 int tcp_iphc_len
; /* actual allocated buffer size */
211 int32_t tcp_hdr_len
; /* Byte len of combined TCP/IP hdr */
212 struct ip
*tcp_ipha
; /* IPv4 header in the buffer */
213 int tcp_ip_hdr_len
; /* Byte len of our current IPvx hdr */
214 tcph_t
*tcp_tcph
; /* tcp header within combined hdr */
215 int32_t tcp_tcp_hdr_len
; /* tcp header len within combined */
217 uint16_t tcp_last_sent_len
; /* Record length for nagle */
218 uint16_t tcp_dupack_cnt
; /* # of consequtive duplicate acks */
221 * Address family that app wishes returned addrsses to be in.
222 * Currently taken from address family used in T_BIND_REQ, but
223 * should really come from family used in original socket() call.
224 * Value can be AF_INET or AF_INET6.
228 uint32_t tcp_ofo_fin_seq
; /* Recv out of order FIN seq num */
229 uint32_t tcp_cwr_snd_max
;
232 tcp_fin_acked
: 1, /* Has our FIN been acked? */
233 tcp_fin_rcvd
: 1, /* Have we seen a FIN? */
234 tcp_fin_sent
: 1, /* Have we sent our FIN yet? */
235 tcp_useloopback
: 1, /* SO_USELOOPBACK "socket" option. */
237 tcp_conn_def_q0
: 1, /* move from q0 to q deferred */
238 tcp_zero_win_probe
: 1, /* Zero win probing is in progress */
240 tcp_active_open
: 1, /* This is a active open */
242 tcp_timer_running
: 1, /* Retransmission timer running */
243 tcp_rexmit
: 1, /* TCP is retransmitting */
244 tcp_snd_sack_ok
: 1, /* Can use SACK for this connection */
245 tcp_ecn_ok
: 1, /* Can use ECN for this connection */
247 tcp_ecn_echo_on
: 1, /* Need to do ECN echo */
248 tcp_ecn_cwr_sent
: 1, /* ECN_CWR has been sent */
249 tcp_cwr
: 1, /* Cwnd has reduced recently */
255 tcp_junk_fill_thru_bit_31
: 13;
257 char *tcp_iphc
; /* Buffer holding tcp/ip hdr template */
259 in_addr_t tcp_remote
; /* true remote address - needed for */
260 /* source routing. */
261 in_addr_t tcp_bound_source
; /* IP address in bind_req */
263 * These fields contain the same information as tcp_tcph->th_*port.
264 * However, the lookup functions can not use the header fields
265 * since during IP option manipulation the tcp_tcph pointer
270 in_port_t tcpu_fport
; /* Remote port */
271 in_port_t tcpu_lport
; /* Local port */
273 uint32_t tcpu_ports2
; /* Rem port, */
275 /* Used for TCP_MATCH performance */
277 #define tcp_fport tcp_tcpu.tcpu_ports1.tcpu_fport
278 #define tcp_lport tcp_tcpu.tcpu_ports1.tcpu_lport
279 #define tcp_ports tcp_tcpu.tcpu_ports2
281 * IP format that packets transmitted from this struct should use.
282 * Value can be IPV4_VERSION or IPV6_VERSION. Determines whether
283 * IP+TCP header template above stores an IPv4 or IPv6 header.
285 ushort_t tcp_ipversion
;
286 int tcp_client_errno
;
287 struct tcp_s
*tcp_eager_next_q
; /* next eager in ESTABLISHED state */
288 struct tcp_s
*tcp_eager_last_q
; /* last eager in ESTABLISHED state */
289 struct tcp_s
*tcp_eager_next_q0
; /* next eager in SYN_RCVD state */
290 struct tcp_s
*tcp_eager_prev_q0
; /* prev eager in SYN_RCVD state */
291 struct tcp_s
*tcp_listener
; /* Our listener */
294 /* External TCP functions. */
295 extern void tcp_socket_init(struct inetboot_socket
*);
296 extern int tcp_connect(int);
297 extern int tcp_listen(int, int);
298 extern int tcp_bind(int);
299 extern int tcp_send(int, tcp_t
*, const void *, int);
300 extern int tcp_opt_set(tcp_t
*, int, int, const void *, socklen_t
);
301 extern int tcp_accept(int, struct sockaddr
*, socklen_t
*);
302 extern int tcp_shutdown(int);
304 /* Exported for recvfrom */
305 extern void tcp_rcv_drain_sock(int);
307 /* Exported for stand/lib/sock/sock_test.c */
308 extern void tcp_time_wait_report(void);
314 #endif /* _TCP_INET_H */