author | masputra |
Sat, 22 Oct 2005 22:50:14 -0700 | |
changeset 741 | 40027a3621ac |
parent 740 | 70e4862c9a1a |
child 898 | 64b2a371a6bd |
permissions | -rw-r--r-- |
0 | 1 |
/* |
2 |
* CDDL HEADER START |
|
3 |
* |
|
4 |
* The contents of this file are subject to the terms of the |
|
5 |
* Common Development and Distribution License, Version 1.0 only |
|
6 |
* (the "License"). You may not use this file except in compliance |
|
7 |
* with the License. |
|
8 |
* |
|
9 |
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE |
|
10 |
* or http://www.opensolaris.org/os/licensing. |
|
11 |
* See the License for the specific language governing permissions |
|
12 |
* and limitations under the License. |
|
13 |
* |
|
14 |
* When distributing Covered Code, include this CDDL HEADER in each |
|
15 |
* file and include the License file at usr/src/OPENSOLARIS.LICENSE. |
|
16 |
* If applicable, add the following below this CDDL HEADER, with the |
|
17 |
* fields enclosed by brackets "[]" replaced with your own identifying |
|
18 |
* information: Portions Copyright [yyyy] [name of copyright owner] |
|
19 |
* |
|
20 |
* CDDL HEADER END |
|
21 |
*/ |
|
22 |
/* |
|
23 |
* Copyright 2005 Sun Microsystems, Inc. All rights reserved. |
|
24 |
* Use is subject to license terms. |
|
25 |
*/ |
|
26 |
/* Copyright (c) 1990 Mentat Inc. */ |
|
27 |
||
28 |
#ifndef _INET_TCP_H |
|
29 |
#define _INET_TCP_H |
|
30 |
||
31 |
#pragma ident "%Z%%M% %I% %E% SMI" |
|
32 |
||
33 |
#ifdef __cplusplus |
|
34 |
extern "C" { |
|
35 |
#endif |
|
36 |
||
37 |
#include <sys/inttypes.h> |
|
38 |
#include <netinet/ip6.h> |
|
39 |
#include <netinet/tcp.h> |
|
40 |
#include <inet/tcp_sack.h> |
|
41 |
#include <sys/socket.h> |
|
42 |
#include <sys/multidata.h> |
|
43 |
||
44 |
/* |
|
45 |
* Private (and possibly temporary) ioctl used by configuration code |
|
46 |
* to lock in the "default" stream for detached closes. |
|
47 |
*/ |
|
48 |
#define TCP_IOC_DEFAULT_Q (('T' << 8) + 51) |
|
49 |
||
50 |
/* TCP states */ |
|
51 |
#define TCPS_CLOSED -6 |
|
52 |
#define TCPS_IDLE -5 /* idle (opened, but not bound) */ |
|
53 |
#define TCPS_BOUND -4 /* bound, ready to connect or accept */ |
|
54 |
#define TCPS_LISTEN -3 /* listening for connection */ |
|
55 |
#define TCPS_SYN_SENT -2 /* active, have sent syn */ |
|
56 |
#define TCPS_SYN_RCVD -1 /* have received syn (and sent ours) */ |
|
57 |
/* states < TCPS_ESTABLISHED are those where connections not established */ |
|
58 |
#define TCPS_ESTABLISHED 0 /* established */ |
|
59 |
#define TCPS_CLOSE_WAIT 1 /* rcvd fin, waiting for close */ |
|
60 |
/* states > TCPS_CLOSE_WAIT are those where user has closed */ |
|
61 |
#define TCPS_FIN_WAIT_1 2 /* have closed and sent fin */ |
|
62 |
#define TCPS_CLOSING 3 /* closed, xchd FIN, await FIN ACK */ |
|
63 |
#define TCPS_LAST_ACK 4 /* had fin and close; await FIN ACK */ |
|
64 |
/* states > TCPS_CLOSE_WAIT && < TCPS_FIN_WAIT_2 await ACK of FIN */ |
|
65 |
#define TCPS_FIN_WAIT_2 5 /* have closed, fin is acked */ |
|
66 |
#define TCPS_TIME_WAIT 6 /* in 2*msl quiet wait after close */ |
|
67 |
||
68 |
/* |
|
69 |
* Internal flags used in conjunction with the packet header flags. |
|
70 |
* Used in tcp_rput_data to keep track of what needs to be done. |
|
71 |
*/ |
|
72 |
#define TH_LIMIT_XMIT 0x0400 /* Limited xmit is needed */ |
|
73 |
#define TH_XMIT_NEEDED 0x0800 /* Window opened - send queued data */ |
|
74 |
#define TH_REXMIT_NEEDED 0x1000 /* Time expired for unacked data */ |
|
75 |
#define TH_ACK_NEEDED 0x2000 /* Send an ack now. */ |
|
76 |
#define TH_NEED_SACK_REXMIT 0x4000 /* Use SACK info to retransmission */ |
|
77 |
#define TH_ACK_TIMER_NEEDED 0x8000 /* Start the delayed ACK timer */ |
|
78 |
#define TH_ORDREL_NEEDED 0x10000 /* Generate an ordrel indication */ |
|
79 |
#define TH_MARKNEXT_NEEDED 0x20000 /* Data should have MSGMARKNEXT */ |
|
80 |
#define TH_SEND_URP_MARK 0x40000 /* Send up tcp_urp_mark_mp */ |
|
81 |
||
82 |
/* |
|
83 |
* TCP sequence numbers are 32 bit integers operated |
|
84 |
* on with modular arithmetic. These macros can be |
|
85 |
* used to compare such integers. |
|
86 |
*/ |
|
87 |
#define SEQ_LT(a, b) ((int32_t)((a)-(b)) < 0) |
|
88 |
#define SEQ_LEQ(a, b) ((int32_t)((a)-(b)) <= 0) |
|
89 |
#define SEQ_GT(a, b) ((int32_t)((a)-(b)) > 0) |
|
90 |
#define SEQ_GEQ(a, b) ((int32_t)((a)-(b)) >= 0) |
|
91 |
||
92 |
/* TCP Protocol header */ |
|
93 |
typedef struct tcphdr_s { |
|
94 |
uint8_t th_lport[2]; /* Source port */ |
|
95 |
uint8_t th_fport[2]; /* Destination port */ |
|
96 |
uint8_t th_seq[4]; /* Sequence number */ |
|
97 |
uint8_t th_ack[4]; /* Acknowledgement number */ |
|
98 |
uint8_t th_offset_and_rsrvd[1]; /* Offset to the packet data */ |
|
99 |
uint8_t th_flags[1]; |
|
100 |
uint8_t th_win[2]; /* Allocation number */ |
|
101 |
uint8_t th_sum[2]; /* TCP checksum */ |
|
102 |
uint8_t th_urp[2]; /* Urgent pointer */ |
|
103 |
} tcph_t; |
|
104 |
||
105 |
#define TCP_HDR_LENGTH(tcph) (((tcph)->th_offset_and_rsrvd[0] >>2) &(0xF << 2)) |
|
106 |
#define TCP_MAX_COMBINED_HEADER_LENGTH (60 + 60) /* Maxed out ip + tcp */ |
|
107 |
#define TCP_MAX_IP_OPTIONS_LENGTH (60 - IP_SIMPLE_HDR_LENGTH) |
|
108 |
#define TCP_MAX_HDR_LENGTH 60 |
|
109 |
#define TCP_MAX_TCP_OPTIONS_LENGTH (60 - sizeof (tcph_t)) |
|
110 |
#define TCP_MIN_HEADER_LENGTH 20 |
|
111 |
#define TCP_MAXWIN 65535 |
|
112 |
#define TCP_PORT_LEN sizeof (in_port_t) |
|
113 |
#define TCP_MAX_WINSHIFT 14 |
|
114 |
#define TCP_MAX_LARGEWIN (TCP_MAXWIN << TCP_MAX_WINSHIFT) |
|
115 |
||
116 |
#define TCPIP_HDR_LENGTH(mp, n) \ |
|
117 |
(n) = IPH_HDR_LENGTH((mp)->b_rptr), \ |
|
118 |
(n) += TCP_HDR_LENGTH((tcph_t *)&(mp)->b_rptr[(n)]) |
|
119 |
||
120 |
/* TCP Protocol header (used if the header is known to be 32-bit aligned) */ |
|
121 |
typedef struct tcphdra_s { |
|
122 |
in_port_t tha_lport; /* Source port */ |
|
123 |
in_port_t tha_fport; /* Destination port */ |
|
124 |
uint32_t tha_seq; /* Sequence number */ |
|
125 |
uint32_t tha_ack; /* Acknowledgement number */ |
|
126 |
uint8_t tha_offset_and_reserved; /* Offset to the packet data */ |
|
127 |
uint8_t tha_flags; |
|
128 |
uint16_t tha_win; /* Allocation number */ |
|
129 |
uint16_t tha_sum; /* TCP checksum */ |
|
130 |
uint16_t tha_urp; /* Urgent pointer */ |
|
131 |
} tcpha_t; |
|
132 |
||
133 |
struct conn_s; |
|
134 |
||
135 |
/* |
|
136 |
* Control structure for each open TCP stream, |
|
137 |
* defined only within the kernel or for a kmem user. |
|
138 |
* NOTE: tcp_reinit_values MUST have a line for each field in this structure! |
|
139 |
*/ |
|
140 |
#if (defined(_KERNEL) || defined(_KMEMUSER)) |
|
141 |
||
142 |
typedef struct tcp_s { |
|
143 |
/* Pointer to previous bind hash next. */ |
|
144 |
struct tcp_s *tcp_time_wait_next; |
|
145 |
/* Pointer to next T/W block */ |
|
146 |
struct tcp_s *tcp_time_wait_prev; |
|
147 |
/* Pointer to previous T/W next */ |
|
148 |
clock_t tcp_time_wait_expire; |
|
149 |
struct conn_s *tcp_connp; |
|
150 |
||
151 |
int32_t tcp_state; |
|
152 |
int32_t tcp_rcv_ws; /* My window scale power */ |
|
153 |
int32_t tcp_snd_ws; /* Sender's window scale power */ |
|
154 |
uint32_t tcp_ts_recent; /* Timestamp of earliest unacked */ |
|
155 |
/* data segment */ |
|
156 |
clock_t tcp_rto; /* Round trip timeout */ |
|
157 |
clock_t tcp_last_rcv_lbolt; |
|
158 |
/* lbolt on last packet, used for PAWS */ |
|
159 |
||
160 |
uint32_t tcp_snxt; /* Senders next seq num */ |
|
161 |
uint32_t tcp_swnd; /* Senders window (relative to suna) */ |
|
162 |
uint32_t tcp_mss; /* Max segment size */ |
|
163 |
uint32_t tcp_iss; /* Initial send seq num */ |
|
164 |
uint32_t tcp_rnxt; /* Seq we expect to recv next */ |
|
165 |
uint32_t tcp_rwnd; |
|
166 |
||
167 |
queue_t *tcp_rq; /* Our upstream neighbor (client) */ |
|
168 |
queue_t *tcp_wq; /* Our downstream neighbor */ |
|
169 |
||
170 |
/* Fields arranged in approximate access order along main paths */ |
|
171 |
mblk_t *tcp_xmit_head; /* Head of rexmit list */ |
|
172 |
mblk_t *tcp_xmit_last; /* last valid data seen by tcp_wput */ |
|
173 |
mblk_t *tcp_xmit_tail; /* Last rexmit data sent */ |
|
174 |
uint32_t tcp_unsent; /* # of bytes in hand that are unsent */ |
|
175 |
uint32_t tcp_xmit_tail_unsent; /* # of unsent bytes in xmit_tail */ |
|
176 |
||
177 |
uint32_t tcp_suna; /* Sender unacknowledged */ |
|
178 |
uint32_t tcp_rexmit_nxt; /* Next rexmit seq num */ |
|
179 |
uint32_t tcp_rexmit_max; /* Max retran seq num */ |
|
180 |
int32_t tcp_snd_burst; /* Send burst factor */ |
|
181 |
uint32_t tcp_cwnd; /* Congestion window */ |
|
182 |
int32_t tcp_cwnd_cnt; /* cwnd cnt in congestion avoidance */ |
|
183 |
||
184 |
uint32_t tcp_ibsegs; /* Inbound segments on this stream */ |
|
185 |
uint32_t tcp_obsegs; /* Outbound segments on this stream */ |
|
186 |
||
187 |
uint32_t tcp_naglim; /* Tunable nagle limit */ |
|
188 |
uint32_t tcp_valid_bits; |
|
189 |
#define TCP_ISS_VALID 0x1 /* Is the tcp_iss seq num active? */ |
|
190 |
#define TCP_FSS_VALID 0x2 /* Is the tcp_fss seq num active? */ |
|
191 |
#define TCP_URG_VALID 0x4 /* Is the tcp_urg seq num active? */ |
|
192 |
#define TCP_OFO_FIN_VALID 0x8 /* Has TCP received an out of order FIN? */ |
|
193 |
||
194 |
||
195 |
int32_t tcp_xmit_hiwater; /* Send buffer high water mark. */ |
|
196 |
||
197 |
timeout_id_t tcp_timer_tid; /* Control block for timer service */ |
|
198 |
uchar_t tcp_timer_backoff; /* Backoff shift count. */ |
|
199 |
int64_t tcp_last_recv_time; /* Last time we receive a segment. */ |
|
200 |
uint32_t tcp_init_cwnd; /* Initial cwnd (start/restart) */ |
|
201 |
||
202 |
/* |
|
203 |
* Following socket options are set by sockfs outside the squeue |
|
204 |
* and we want to separate these bit fields from the other bit fields |
|
205 |
* set by TCP to avoid grabbing locks. sockfs ensures that only one |
|
206 |
* thread in sockfs can set a socket option at a time on a conn_t. |
|
207 |
* However TCP may read these options concurrently. The linger option |
|
208 |
* needs atomicity since tcp_lingertime also needs to be in sync. |
|
209 |
* However TCP uses it only during close, and by then no socket option |
|
210 |
* can come down. So we don't need any locks, instead just separating |
|
211 |
* the sockfs settable bit fields from the other bit fields is |
|
212 |
* sufficient. |
|
213 |
*/ |
|
214 |
uint32_t |
|
215 |
tcp_debug : 1, /* SO_DEBUG "socket" option. */ |
|
216 |
tcp_dontroute : 1, /* SO_DONTROUTE "socket" option. */ |
|
217 |
tcp_broadcast : 1, /* SO_BROADCAST "socket" option. */ |
|
218 |
tcp_useloopback : 1, /* SO_USELOOPBACK "socket" option. */ |
|
219 |
||
220 |
tcp_oobinline : 1, /* SO_OOBINLINE "socket" option. */ |
|
221 |
tcp_dgram_errind : 1, /* SO_DGRAM_ERRIND option */ |
|
222 |
tcp_linger : 1, /* SO_LINGER turned on */ |
|
223 |
tcp_reuseaddr : 1, /* SO_REUSEADDR "socket" option. */ |
|
224 |
||
225 |
tcp_junk_to_bit_31 : 24; |
|
226 |
||
227 |
/* Following manipulated by TCP under squeue protection */ |
|
228 |
uint32_t |
|
229 |
tcp_urp_last_valid : 1, /* Is tcp_urp_last valid? */ |
|
230 |
tcp_hard_binding : 1, /* If we've started a full bind */ |
|
231 |
tcp_hard_bound : 1, /* If we've done a full bind with IP */ |
|
232 |
tcp_fin_acked : 1, /* Has our FIN been acked? */ |
|
233 |
||
234 |
tcp_fin_rcvd : 1, /* Have we seen a FIN? */ |
|
235 |
tcp_fin_sent : 1, /* Have we sent our FIN yet? */ |
|
236 |
tcp_ordrel_done : 1, /* Have we sent the ord_rel upstream? */ |
|
237 |
tcp_flow_stopped : 1, /* Have we flow controlled xmitter? */ |
|
238 |
||
239 |
tcp_detached : 1, /* If we're detached from a stream */ |
|
240 |
tcp_bind_pending : 1, /* Client is waiting for bind ack */ |
|
241 |
tcp_unbind_pending : 1, /* Client sent T_UNBIND_REQ */ |
|
242 |
tcp_deferred_clean_death : 1, |
|
243 |
/* defer tcp endpoint cleanup etc. */ |
|
244 |
||
245 |
tcp_conn_def_q0: 1, /* move from q0 to q deferred */ |
|
246 |
tcp_ka_enabled: 1, /* Connection KeepAlive Timer needed */ |
|
247 |
tcp_zero_win_probe: 1, /* Zero win probing is in progress */ |
|
248 |
tcp_loopback: 1, /* src and dst are the same machine */ |
|
249 |
||
250 |
tcp_localnet: 1, /* src and dst are on the same subnet */ |
|
251 |
tcp_syn_defense: 1, /* For defense against SYN attack */ |
|
252 |
#define tcp_dontdrop tcp_syn_defense |
|
253 |
tcp_set_timer : 1, |
|
254 |
tcp_active_open: 1, /* This is a active open */ |
|
255 |
||
256 |
tcp_timeout : 1, /* qbufcall failed, qtimeout pending */ |
|
257 |
tcp_rexmit : 1, /* TCP is retransmitting */ |
|
258 |
tcp_snd_sack_ok : 1, /* Can use SACK for this connection */ |
|
259 |
tcp_empty_flag : 1, /* Empty flag for future use */ |
|
260 |
||
261 |
tcp_recvdstaddr : 1, /* return T_EXTCONN_IND with dst addr */ |
|
262 |
tcp_hwcksum : 1, /* The NIC is capable of hwcksum */ |
|
263 |
tcp_ip_forward_progress : 1, |
|
264 |
tcp_anon_priv_bind : 1, |
|
265 |
||
266 |
tcp_ecn_ok : 1, /* Can use ECN for this connection */ |
|
267 |
tcp_ecn_echo_on : 1, /* Need to do ECN echo */ |
|
268 |
tcp_ecn_cwr_sent : 1, /* ECN_CWR has been sent */ |
|
269 |
tcp_cwr : 1; /* Cwnd has reduced recently */ |
|
270 |
||
271 |
/* Following manipulated by TCP under squeue protection */ |
|
272 |
uint32_t |
|
273 |
tcp_mdt : 1, /* Lower layer is capable of MDT */ |
|
274 |
tcp_snd_ts_ok : 1, |
|
275 |
tcp_snd_ws_ok : 1, |
|
276 |
tcp_exclbind : 1, /* ``exclusive'' binding */ |
|
277 |
||
278 |
tcp_reserved_port : 1, |
|
279 |
tcp_hdr_grown : 1, |
|
280 |
tcp_in_free_list : 1, |
|
281 |
tcp_snd_zcopy_on : 1, /* xmit zero-copy enabled */ |
|
282 |
||
283 |
tcp_snd_zcopy_aware : 1, /* client is zero-copy aware */ |
|
284 |
tcp_xmit_zc_clean : 1, /* the xmit list is free of zc-mblk */ |
|
285 |
tcp_wait_for_eagers : 1, /* Wait for eagers to disappear */ |
|
286 |
tcp_accept_error : 1, /* Error during TLI accept */ |
|
287 |
||
288 |
tcp_send_discon_ind : 1, /* TLI accept err, send discon ind */ |
|
289 |
tcp_cork : 1, /* tcp_cork option */ |
|
741
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
290 |
tcp_pad_to_bit_31 : 18; |
0 | 291 |
|
292 |
uint32_t tcp_if_mtu; /* Outgoing interface MTU. */ |
|
293 |
||
294 |
mblk_t *tcp_reass_head; /* Out of order reassembly list head */ |
|
295 |
mblk_t *tcp_reass_tail; /* Out of order reassembly list tail */ |
|
296 |
||
297 |
tcp_sack_info_t *tcp_sack_info; |
|
298 |
||
299 |
#define tcp_pipe tcp_sack_info->tcp_pipe |
|
300 |
#define tcp_fack tcp_sack_info->tcp_fack |
|
301 |
#define tcp_sack_snxt tcp_sack_info->tcp_sack_snxt |
|
302 |
#define tcp_max_sack_blk tcp_sack_info->tcp_max_sack_blk |
|
303 |
#define tcp_num_sack_blk tcp_sack_info->tcp_num_sack_blk |
|
304 |
#define tcp_sack_list tcp_sack_info->tcp_sack_list |
|
305 |
#define tcp_num_notsack_blk tcp_sack_info->tcp_num_notsack_blk |
|
306 |
#define tcp_cnt_notsack_list tcp_sack_info->tcp_cnt_notsack_list |
|
307 |
#define tcp_notsack_list tcp_sack_info->tcp_notsack_list |
|
308 |
||
309 |
mblk_t *tcp_rcv_list; /* Queued until push, urgent data, */ |
|
310 |
mblk_t *tcp_rcv_last_head; /* optdata, or the count exceeds */ |
|
311 |
mblk_t *tcp_rcv_last_tail; /* tcp_rcv_push_wait. */ |
|
312 |
uint32_t tcp_rcv_cnt; /* tcp_rcv_list is b_next chain. */ |
|
313 |
||
314 |
uint32_t tcp_cwnd_ssthresh; /* Congestion window */ |
|
315 |
uint32_t tcp_cwnd_max; |
|
316 |
uint32_t tcp_csuna; /* Clear (no rexmits in window) suna */ |
|
317 |
||
318 |
clock_t tcp_rtt_sa; /* Round trip smoothed average */ |
|
319 |
clock_t tcp_rtt_sd; /* Round trip smoothed deviation */ |
|
320 |
clock_t tcp_rtt_update; /* Round trip update(s) */ |
|
321 |
clock_t tcp_ms_we_have_waited; /* Total retrans time */ |
|
322 |
||
323 |
uint32_t tcp_swl1; /* These help us avoid using stale */ |
|
324 |
uint32_t tcp_swl2; /* packets to update state */ |
|
325 |
||
326 |
uint32_t tcp_rack; /* Seq # we have acked */ |
|
327 |
uint32_t tcp_rack_cnt; /* # of segs we have deferred ack */ |
|
328 |
uint32_t tcp_rack_cur_max; /* # of segs we may defer ack for now */ |
|
329 |
uint32_t tcp_rack_abs_max; /* # of segs we may defer ack ever */ |
|
330 |
timeout_id_t tcp_ack_tid; /* Delayed ACK timer ID */ |
|
331 |
timeout_id_t tcp_push_tid; /* Push timer ID */ |
|
332 |
||
333 |
uint32_t tcp_max_swnd; /* Maximum swnd we have seen */ |
|
334 |
||
335 |
struct tcp_s *tcp_listener; /* Our listener */ |
|
336 |
||
337 |
int32_t tcp_xmit_lowater; /* Send buffer low water mark. */ |
|
338 |
||
339 |
uint32_t tcp_irs; /* Initial recv seq num */ |
|
340 |
uint32_t tcp_fss; /* Final/fin send seq num */ |
|
341 |
uint32_t tcp_urg; /* Urgent data seq num */ |
|
342 |
||
343 |
clock_t tcp_first_timer_threshold; /* When to prod IP */ |
|
344 |
clock_t tcp_second_timer_threshold; /* When to give up completely */ |
|
345 |
clock_t tcp_first_ctimer_threshold; /* 1st threshold while connecting */ |
|
346 |
clock_t tcp_second_ctimer_threshold; /* 2nd ... while connecting */ |
|
347 |
||
348 |
int tcp_lingertime; /* Close linger time (in seconds) */ |
|
349 |
||
350 |
uint32_t tcp_urp_last; /* Last urp for which signal sent */ |
|
351 |
mblk_t *tcp_urp_mp; /* T_EXDATA_IND for urgent byte */ |
|
352 |
mblk_t *tcp_urp_mark_mp; /* zero-length marked/unmarked msg */ |
|
353 |
||
354 |
int tcp_conn_req_cnt_q0; /* # of conn reqs in SYN_RCVD */ |
|
355 |
int tcp_conn_req_cnt_q; /* # of conn reqs in ESTABLISHED */ |
|
356 |
int tcp_conn_req_max; /* # of ESTABLISHED conn reqs allowed */ |
|
357 |
t_scalar_t tcp_conn_req_seqnum; /* Incrementing pending conn req ID */ |
|
358 |
#define tcp_ip_addr_cache tcp_reass_tail |
|
359 |
/* Cache ip addresses that */ |
|
360 |
/* complete the 3-way handshake */ |
|
361 |
kmutex_t tcp_eager_lock; |
|
362 |
struct tcp_s *tcp_eager_next_q; /* next eager in ESTABLISHED state */ |
|
363 |
struct tcp_s *tcp_eager_last_q; /* last eager in ESTABLISHED state */ |
|
364 |
struct tcp_s *tcp_eager_next_q0; /* next eager in SYN_RCVD state */ |
|
365 |
struct tcp_s *tcp_eager_prev_q0; /* prev eager in SYN_RCVD state */ |
|
366 |
/* all eagers form a circular list */ |
|
367 |
union { |
|
368 |
mblk_t *tcp_eager_conn_ind; /* T_CONN_IND waiting for 3rd ack. */ |
|
369 |
mblk_t *tcp_opts_conn_req; /* T_CONN_REQ w/ options processed */ |
|
370 |
} tcp_conn; |
|
371 |
uint32_t tcp_syn_rcvd_timeout; /* How many SYN_RCVD timeout in q0 */ |
|
372 |
||
373 |
/* TCP Keepalive Timer members */ |
|
374 |
int32_t tcp_ka_last_intrvl; /* Last probe interval */ |
|
375 |
timeout_id_t tcp_ka_tid; /* Keepalive timer ID */ |
|
376 |
uint32_t tcp_ka_interval; /* Keepalive interval */ |
|
377 |
uint32_t tcp_ka_abort_thres; /* Keepalive abort threshold */ |
|
378 |
||
379 |
int32_t tcp_client_errno; /* How the client screwed up */ |
|
380 |
||
381 |
char *tcp_iphc; /* Buffer holding tcp/ip hdr template */ |
|
382 |
int tcp_iphc_len; /* actual allocated buffer size */ |
|
383 |
int32_t tcp_hdr_len; /* Byte len of combined TCP/IP hdr */ |
|
384 |
ipha_t *tcp_ipha; /* IPv4 header in the buffer */ |
|
385 |
ip6_t *tcp_ip6h; /* IPv6 header in the buffer */ |
|
386 |
int tcp_ip_hdr_len; /* Byte len of our current IPvx hdr */ |
|
387 |
tcph_t *tcp_tcph; /* tcp header within combined hdr */ |
|
388 |
int32_t tcp_tcp_hdr_len; /* tcp header len within combined */ |
|
389 |
||
390 |
uint32_t tcp_sum; /* checksum to compensate for source */ |
|
391 |
/* routed packets. Host byte order */ |
|
392 |
uint16_t tcp_last_sent_len; /* Record length for nagle */ |
|
393 |
uint16_t tcp_dupack_cnt; /* # of consequtive duplicate acks */ |
|
394 |
||
395 |
kmutex_t *tcp_acceptor_lockp; /* Ptr to tf_lock */ |
|
396 |
||
397 |
timeout_id_t tcp_ordrelid; /* qbufcall/qtimeout id */ |
|
398 |
t_uscalar_t tcp_acceptor_id; /* ACCEPTOR_id */ |
|
399 |
||
400 |
int tcp_ipsec_overhead; |
|
401 |
/* |
|
402 |
* Address family that app wishes returned addrsses to be in. |
|
403 |
* Currently taken from address family used in T_BIND_REQ, but |
|
404 |
* should really come from family used in original socket() call. |
|
405 |
* Value can be AF_INET or AF_INET6. |
|
406 |
*/ |
|
407 |
uint_t tcp_family; |
|
408 |
/* |
|
409 |
* used for a quick test to determine if any ancillary bits are |
|
410 |
* set |
|
411 |
*/ |
|
412 |
uint_t tcp_ipv6_recvancillary; /* Flags */ |
|
413 |
#define TCP_IPV6_RECVPKTINFO 0x01 /* IPV6_RECVPKTINFO option */ |
|
414 |
#define TCP_IPV6_RECVHOPLIMIT 0x02 /* IPV6_RECVHOPLIMIT option */ |
|
415 |
#define TCP_IPV6_RECVHOPOPTS 0x04 /* IPV6_RECVHOPOPTS option */ |
|
416 |
#define TCP_IPV6_RECVDSTOPTS 0x08 /* IPV6_RECVDSTOPTS option */ |
|
417 |
#define TCP_IPV6_RECVRTHDR 0x10 /* IPV6_RECVRTHDR option */ |
|
418 |
#define TCP_IPV6_RECVRTDSTOPTS 0x20 /* IPV6_RECVRTHDRDSTOPTS option */ |
|
419 |
#define TCP_IPV6_RECVTCLASS 0x40 /* IPV6_RECVTCLASS option */ |
|
420 |
#define TCP_OLD_IPV6_RECVDSTOPTS 0x80 /* old IPV6_RECVDSTOPTS option */ |
|
421 |
||
422 |
uint_t tcp_recvifindex; /* Last received IPV6_RCVPKTINFO */ |
|
423 |
uint_t tcp_recvhops; /* Last received IPV6_RECVHOPLIMIT */ |
|
424 |
uint_t tcp_recvtclass; /* Last received IPV6_RECVTCLASS */ |
|
425 |
ip6_hbh_t *tcp_hopopts; /* Last received IPV6_RECVHOPOPTS */ |
|
426 |
ip6_dest_t *tcp_dstopts; /* Last received IPV6_RECVDSTOPTS */ |
|
427 |
ip6_dest_t *tcp_rtdstopts; /* Last recvd IPV6_RECVRTHDRDSTOPTS */ |
|
428 |
ip6_rthdr_t *tcp_rthdr; /* Last received IPV6_RECVRTHDR */ |
|
429 |
uint_t tcp_hopoptslen; |
|
430 |
uint_t tcp_dstoptslen; |
|
431 |
uint_t tcp_rtdstoptslen; |
|
432 |
uint_t tcp_rthdrlen; |
|
433 |
||
434 |
mblk_t *tcp_timercache; |
|
435 |
cred_t *tcp_cred; /* Credentials when this was opened */ |
|
436 |
pid_t tcp_cpid; /* Process id when this was opened */ |
|
437 |
||
438 |
||
439 |
union { |
|
440 |
struct { |
|
441 |
uchar_t v4_ttl; |
|
442 |
/* Dup of tcp_ipha.iph_type_of_service */ |
|
443 |
uchar_t v4_tos; /* Dup of tcp_ipha.iph_ttl */ |
|
444 |
} v4_hdr_info; |
|
445 |
struct { |
|
446 |
uint_t v6_vcf; /* Dup of tcp_ip6h.ip6h_vcf */ |
|
447 |
uchar_t v6_hops; /* Dup of tcp_ip6h.ip6h_hops */ |
|
448 |
} v6_hdr_info; |
|
449 |
} tcp_hdr_info; |
|
450 |
#define tcp_ttl tcp_hdr_info.v4_hdr_info.v4_ttl |
|
451 |
#define tcp_tos tcp_hdr_info.v4_hdr_info.v4_tos |
|
452 |
#define tcp_ip6_vcf tcp_hdr_info.v6_hdr_info.v6_vcf |
|
453 |
#define tcp_ip6_hops tcp_hdr_info.v6_hdr_info.v6_hops |
|
454 |
||
455 |
ushort_t tcp_ipversion; |
|
456 |
uint_t tcp_bound_if; /* IPV6_BOUND_IF */ |
|
457 |
||
458 |
#define tcp_lport tcp_connp->conn_lport |
|
459 |
#define tcp_fport tcp_connp->conn_fport |
|
460 |
#define tcp_ports tcp_connp->conn_ports |
|
461 |
||
462 |
#define tcp_remote tcp_connp->conn_rem |
|
463 |
#define tcp_ip_src tcp_connp->conn_src |
|
464 |
||
465 |
#define tcp_remote_v6 tcp_connp->conn_remv6 |
|
466 |
#define tcp_ip_src_v6 tcp_connp->conn_srcv6 |
|
467 |
#define tcp_bound_source_v6 tcp_connp->conn_bound_source_v6 |
|
468 |
#define tcp_bound_source tcp_connp->conn_bound_source |
|
469 |
||
470 |
kmutex_t tcp_closelock; |
|
471 |
kcondvar_t tcp_closecv; |
|
472 |
uint8_t tcp_closed; |
|
473 |
uint8_t tcp_closeflags; |
|
474 |
uint8_t tcp_cleandeathtag; |
|
475 |
mblk_t tcp_closemp; |
|
476 |
timeout_id_t tcp_linger_tid; /* Linger timer ID */ |
|
477 |
void *tcp_tracebuf; |
|
478 |
||
479 |
struct tcp_s *tcp_acceptor_hash; /* Acceptor hash chain */ |
|
480 |
struct tcp_s **tcp_ptpahn; /* Pointer to previous accept hash next. */ |
|
481 |
struct tcp_s *tcp_bind_hash; /* Bind hash chain */ |
|
482 |
struct tcp_s **tcp_ptpbhn; |
|
483 |
||
484 |
boolean_t tcp_ire_ill_check_done; |
|
485 |
uint_t tcp_maxpsz; |
|
486 |
||
487 |
/* |
|
488 |
* used for Multidata Transmit |
|
489 |
*/ |
|
490 |
uint_t tcp_mdt_hdr_head; /* leading header fragment extra space */ |
|
491 |
uint_t tcp_mdt_hdr_tail; /* trailing header fragment extra space */ |
|
492 |
int tcp_mdt_max_pld; /* maximum payload buffers per Multidata */ |
|
493 |
||
494 |
uint32_t tcp_ofo_fin_seq; /* Recv out of order FIN seq num */ |
|
495 |
uint32_t tcp_cwr_snd_max; |
|
496 |
uint_t tcp_drop_opt_ack_cnt; /* # tcp generated optmgmt */ |
|
497 |
ip6_pkt_t tcp_sticky_ipp; /* Sticky options */ |
|
498 |
#define tcp_ipp_fields tcp_sticky_ipp.ipp_fields /* valid fields */ |
|
499 |
#define tcp_ipp_ifindex tcp_sticky_ipp.ipp_ifindex /* pktinfo ifindex */ |
|
500 |
#define tcp_ipp_addr tcp_sticky_ipp.ipp_addr /* pktinfo src/dst addr */ |
|
501 |
#define tcp_ipp_hoplimit tcp_sticky_ipp.ipp_hoplimit |
|
502 |
#define tcp_ipp_hopoptslen tcp_sticky_ipp.ipp_hopoptslen |
|
503 |
#define tcp_ipp_rtdstoptslen tcp_sticky_ipp.ipp_rtdstoptslen |
|
504 |
#define tcp_ipp_rthdrlen tcp_sticky_ipp.ipp_rthdrlen |
|
505 |
#define tcp_ipp_dstoptslen tcp_sticky_ipp.ipp_dstoptslen |
|
506 |
#define tcp_ipp_hopopts tcp_sticky_ipp.ipp_hopopts |
|
507 |
#define tcp_ipp_rtdstopts tcp_sticky_ipp.ipp_rtdstopts |
|
508 |
#define tcp_ipp_rthdr tcp_sticky_ipp.ipp_rthdr |
|
509 |
#define tcp_ipp_dstopts tcp_sticky_ipp.ipp_dstopts |
|
510 |
#define tcp_ipp_nexthop tcp_sticky_ipp.ipp_nexthop |
|
511 |
#define tcp_ipp_use_min_mtu tcp_sticky_ipp.ipp_use_min_mtu |
|
512 |
struct tcp_s *tcp_saved_listener; /* saved value of listener */ |
|
513 |
||
741
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
514 |
uint32_t tcp_in_ack_unsent; /* ACK for unsent data cnt. */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
515 |
|
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
516 |
/* |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
517 |
* The following fusion-related fields are protected by squeue. |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
518 |
*/ |
0 | 519 |
struct tcp_s *tcp_loopback_peer; /* peer tcp for loopback */ |
520 |
mblk_t *tcp_fused_sigurg_mp; /* M_PCSIG mblk for SIGURG */ |
|
741
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
521 |
size_t tcp_fuse_rcv_hiwater; /* fusion receive queue size */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
522 |
uint_t tcp_fuse_rcv_unread_hiwater; /* max # of outstanding pkts */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
523 |
/* |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
524 |
* The following fusion-related fields and bit fields are to be |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
525 |
* manipulated with squeue protection or with tcp_fuse_lock held. |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
526 |
*/ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
527 |
kmutex_t tcp_fuse_lock; |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
528 |
uint_t tcp_fuse_rcv_unread_cnt; /* # of outstanding pkts */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
529 |
uint32_t |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
530 |
tcp_fused : 1, /* loopback tcp in fusion mode */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
531 |
tcp_unfusable : 1, /* fusion not allowed on endpoint */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
532 |
tcp_fused_sigurg : 1, /* send SIGURG upon draining */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
533 |
tcp_direct_sockfs : 1, /* direct calls to sockfs */ |
0 | 534 |
|
741
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
535 |
tcp_fuse_syncstr_stopped : 1, /* synchronous streams stopped */ |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
536 |
tcp_fuse_to_bit_31 : 27; |
0 | 537 |
|
538 |
/* |
|
539 |
* This variable is accessed without any lock protection |
|
540 |
* and therefore must not be declared as a bit field along |
|
541 |
* with the rest which require such condition. |
|
542 |
*/ |
|
543 |
boolean_t tcp_issocket; /* this is a socket tcp */ |
|
741
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
544 |
|
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
545 |
uint32_t tcp_squeue_bytes; |
0 | 546 |
} tcp_t; |
547 |
||
548 |
extern void tcp_free(tcp_t *tcp); |
|
549 |
extern void tcp_ddi_init(void); |
|
550 |
extern void tcp_ddi_destroy(void); |
|
551 |
extern void tcp_xmit_listeners_reset(mblk_t *mp, uint_t ip_hdr_len); |
|
552 |
extern void tcp_conn_request(void *arg, mblk_t *mp, void *arg2); |
|
553 |
extern void tcp_conn_request_unbound(void *arg, mblk_t *mp, void *arg2); |
|
554 |
extern void tcp_input(void *arg, mblk_t *mp, void *arg2); |
|
555 |
extern void tcp_rput_data(void *arg, mblk_t *mp, void *arg2); |
|
556 |
extern void *tcp_get_conn(void *arg); |
|
557 |
extern void tcp_time_wait_collector(void *arg); |
|
741
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
558 |
extern int tcp_snmp_get(queue_t *, mblk_t *); |
40027a3621ac
PSARC 2005/082 Yosemite: UDP Performance Enhancement
masputra
parents:
740
diff
changeset
|
559 |
extern int tcp_snmp_set(queue_t *, int, int, uchar_t *, int len); |
0 | 560 |
/* |
561 |
* The TCP Fanout structure. |
|
562 |
* The hash tables and their linkage (tcp_*_hash_next, tcp_ptp*hn) are |
|
563 |
* protected by the per-bucket tf_lock. Each tcp_t |
|
564 |
* inserted in the list points back at this lock using tcp_*_lockp. |
|
565 |
* |
|
566 |
* The listener and acceptor hash queues are lists of tcp_t. |
|
567 |
*/ |
|
568 |
||
569 |
/* listener hash and acceptor hash queue head */ |
|
570 |
typedef struct tf_s { |
|
571 |
tcp_t *tf_tcp; |
|
572 |
kmutex_t tf_lock; |
|
573 |
} tf_t; |
|
574 |
||
575 |
extern mib2_tcp_t tcp_mib; |
|
576 |
||
577 |
#endif /* (defined(_KERNEL) || defined(_KMEMUSER)) */ |
|
578 |
||
579 |
/* Contract private interface between TCP and Clustering. */ |
|
580 |
||
581 |
#define CL_TCPI_V1 1 /* cl_tcpi_version number */ |
|
582 |
||
583 |
typedef struct cl_tcp_info_s { |
|
584 |
ushort_t cl_tcpi_version; /* cl_tcp_info_t's version no */ |
|
585 |
ushort_t cl_tcpi_ipversion; /* IP version */ |
|
586 |
int32_t cl_tcpi_state; /* TCP state */ |
|
587 |
in_port_t cl_tcpi_lport; /* Local port */ |
|
588 |
in_port_t cl_tcpi_fport; /* Remote port */ |
|
589 |
in6_addr_t cl_tcpi_laddr_v6; /* Local IP address */ |
|
590 |
in6_addr_t cl_tcpi_faddr_v6; /* Remote IP address */ |
|
591 |
#ifdef _KERNEL |
|
592 |
/* Note: V4_PART_OF_V6 is meant to be used only for _KERNEL defined stuff */ |
|
593 |
#define cl_tcpi_laddr V4_PART_OF_V6(cl_tcpi_laddr_v6) |
|
594 |
#define cl_tcpi_faddr V4_PART_OF_V6(cl_tcpi_faddr_v6) |
|
595 |
||
596 |
#endif /* _KERNEL */ |
|
597 |
} cl_tcp_info_t; |
|
598 |
||
599 |
/* |
|
600 |
* Hook functions to enable cluster networking |
|
601 |
* On non-clustered systems these vectors must always be NULL. |
|
602 |
*/ |
|
603 |
||
604 |
extern void (*cl_inet_listen)(uint8_t, sa_family_t, uint8_t *, in_port_t); |
|
605 |
extern void (*cl_inet_unlisten)(uint8_t, sa_family_t, uint8_t *, |
|
606 |
in_port_t); |
|
607 |
||
608 |
/* |
|
609 |
* Contracted Consolidation Private ioctl for aborting TCP connections. |
|
610 |
* In order to keep the offsets and size of the structure the same between |
|
611 |
* a 32-bit application and a 64-bit amd64 kernel, we use a #pragma |
|
612 |
* pack(4). |
|
613 |
*/ |
|
614 |
#define TCP_IOC_ABORT_CONN (('T' << 8) + 91) |
|
615 |
||
616 |
#if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4 |
|
617 |
#pragma pack(4) |
|
618 |
#endif |
|
619 |
||
620 |
typedef struct tcp_ioc_abort_conn_s { |
|
621 |
struct sockaddr_storage ac_local; /* local addr and port */ |
|
622 |
struct sockaddr_storage ac_remote; /* remote addr and port */ |
|
623 |
int32_t ac_start; /* start state */ |
|
624 |
int32_t ac_end; /* end state */ |
|
625 |
int32_t ac_zoneid; /* zoneid */ |
|
626 |
} tcp_ioc_abort_conn_t; |
|
627 |
||
628 |
#if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4 |
|
629 |
#pragma pack() |
|
630 |
#endif |
|
631 |
||
632 |
#if (defined(_KERNEL) || defined(_KMEMUSER)) |
|
633 |
extern void tcp_rput_other(tcp_t *tcp, mblk_t *mp); |
|
634 |
#endif |
|
635 |
||
636 |
#ifdef __cplusplus |
|
637 |
} |
|
638 |
#endif |
|
639 |
||
640 |
#endif /* _INET_TCP_H */ |