xref: /illumos-gate/usr/src/uts/common/inet/tcp.h (revision 68ac2337c38c8af06edcf32a72e42de36ec72a9d)
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 (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 /* Copyright (c) 1990 Mentat Inc. */
26 
27 #ifndef	_INET_TCP_H
28 #define	_INET_TCP_H
29 
30 #pragma ident	"%Z%%M%	%I%	%E% SMI"
31 
32 #ifdef	__cplusplus
33 extern "C" {
34 #endif
35 
36 #include <sys/inttypes.h>
37 #include <netinet/ip6.h>
38 #include <netinet/tcp.h>
39 #include <inet/tcp_sack.h>
40 #include <sys/socket.h>
41 #include <sys/multidata.h>
42 #include <inet/kssl/ksslapi.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 #define	TCP_MAX_LSO_LENGTH	(IP_MAXPACKET - TCP_MAX_COMBINED_HEADER_LENGTH)
116 
117 #define	TCPIP_HDR_LENGTH(mp, n)					\
118 	(n) = IPH_HDR_LENGTH((mp)->b_rptr),			\
119 	(n) += TCP_HDR_LENGTH((tcph_t *)&(mp)->b_rptr[(n)])
120 
121 /* TCP Protocol header (used if the header is known to be 32-bit aligned) */
122 typedef	struct tcphdra_s {
123 	in_port_t	tha_lport;	/* Source port */
124 	in_port_t	tha_fport;	/* Destination port */
125 	uint32_t	tha_seq;	/* Sequence number */
126 	uint32_t	tha_ack;	/* Acknowledgement number */
127 	uint8_t tha_offset_and_reserved; /* Offset to the packet data */
128 	uint8_t		tha_flags;
129 	uint16_t	tha_win;	/* Allocation number */
130 	uint16_t	tha_sum;	/* TCP checksum */
131 	uint16_t	tha_urp;	/* Urgent pointer */
132 } tcpha_t;
133 
134 struct conn_s;
135 
136 /*
137  * Control structure for each open TCP stream,
138  * defined only within the kernel or for a kmem user.
139  * NOTE: tcp_reinit_values MUST have a line for each field in this structure!
140  */
141 #if (defined(_KERNEL) || defined(_KMEMUSER))
142 
143 typedef struct tcp_s {
144 				/* Pointer to previous bind hash next. */
145 	struct tcp_s *tcp_time_wait_next;
146 				/* Pointer to next T/W block */
147 	struct tcp_s *tcp_time_wait_prev;
148 				/* Pointer to previous T/W next */
149 	clock_t	tcp_time_wait_expire;
150 	struct conn_s *tcp_connp;
151 
152 	int32_t	tcp_state;
153 	int32_t	tcp_rcv_ws;		/* My window scale power */
154 	int32_t	tcp_snd_ws;		/* Sender's window scale power */
155 	uint32_t tcp_ts_recent;		/* Timestamp of earliest unacked */
156 					/*  data segment */
157 	clock_t	tcp_rto;		/* Round trip timeout */
158 	clock_t	tcp_last_rcv_lbolt;
159 				/* lbolt on last packet, used for PAWS */
160 
161 	uint32_t tcp_snxt;		/* Senders next seq num */
162 	uint32_t tcp_swnd;		/* Senders window (relative to suna) */
163 	uint32_t tcp_mss;		/* Max segment size */
164 	uint32_t tcp_iss;		/* Initial send seq num */
165 	uint32_t tcp_rnxt;		/* Seq we expect to recv next */
166 	uint32_t tcp_rwnd;
167 
168 	queue_t	*tcp_rq;		/* Our upstream neighbor (client) */
169 	queue_t	*tcp_wq;		/* Our downstream neighbor */
170 
171 	/* Fields arranged in approximate access order along main paths */
172 	mblk_t	*tcp_xmit_head;		/* Head of rexmit list */
173 	mblk_t	*tcp_xmit_last;		/* last valid data seen by tcp_wput */
174 	mblk_t	*tcp_xmit_tail;		/* Last rexmit data sent */
175 	uint32_t tcp_unsent;		/* # of bytes in hand that are unsent */
176 	uint32_t tcp_xmit_tail_unsent;	/* # of unsent bytes in xmit_tail */
177 
178 	uint32_t tcp_suna;		/* Sender unacknowledged */
179 	uint32_t tcp_rexmit_nxt;	/* Next rexmit seq num */
180 	uint32_t tcp_rexmit_max;	/* Max retran seq num */
181 	int32_t	tcp_snd_burst;		/* Send burst factor */
182 	uint32_t tcp_cwnd;		/* Congestion window */
183 	int32_t tcp_cwnd_cnt;		/* cwnd cnt in congestion avoidance */
184 
185 	uint32_t tcp_ibsegs;		/* Inbound segments on this stream */
186 	uint32_t tcp_obsegs;		/* Outbound segments on this stream */
187 
188 	uint32_t tcp_naglim;		/* Tunable nagle limit */
189 	uint32_t	tcp_valid_bits;
190 #define	TCP_ISS_VALID	0x1	/* Is the tcp_iss seq num active? */
191 #define	TCP_FSS_VALID	0x2	/* Is the tcp_fss seq num active? */
192 #define	TCP_URG_VALID	0x4	/* Is the tcp_urg seq num active? */
193 #define	TCP_OFO_FIN_VALID 0x8	/* Has TCP received an out of order FIN? */
194 
195 
196 	int32_t	tcp_xmit_hiwater;	/* Send buffer high water mark. */
197 
198 	timeout_id_t	tcp_timer_tid;	/* Control block for timer service */
199 	uchar_t	tcp_timer_backoff;	/* Backoff shift count. */
200 	int64_t tcp_last_recv_time;	/* Last time we receive a segment. */
201 	uint32_t tcp_init_cwnd;		/* Initial cwnd (start/restart) */
202 
203 	/*
204 	 * Following socket options are set by sockfs outside the squeue
205 	 * and we want to separate these bit fields from the other bit fields
206 	 * set by TCP to avoid grabbing locks. sockfs ensures that only one
207 	 * thread in sockfs can set a socket option at a time on a conn_t.
208 	 * However TCP may read these options concurrently. The linger option
209 	 * needs atomicity since tcp_lingertime also needs to be in sync.
210 	 * However TCP uses it only during close, and by then no socket option
211 	 * can come down. So we don't need any locks, instead just separating
212 	 * the sockfs settable bit fields from the other bit fields is
213 	 * sufficient.
214 	 */
215 	uint32_t
216 		tcp_debug : 1,		/* SO_DEBUG "socket" option. */
217 		tcp_dontroute : 1,	/* SO_DONTROUTE "socket" option. */
218 		tcp_broadcast : 1,	/* SO_BROADCAST "socket" option. */
219 		tcp_useloopback : 1,	/* SO_USELOOPBACK "socket" option. */
220 
221 		tcp_oobinline : 1,	/* SO_OOBINLINE "socket" option. */
222 		tcp_dgram_errind : 1,	/* SO_DGRAM_ERRIND option */
223 		tcp_linger : 1,		/* SO_LINGER turned on */
224 		tcp_reuseaddr	: 1,	/* SO_REUSEADDR "socket" option. */
225 
226 		tcp_junk_to_bit_31 : 24;
227 
228 	/* Following manipulated by TCP under squeue protection */
229 	uint32_t
230 		tcp_urp_last_valid : 1,	/* Is tcp_urp_last valid? */
231 		tcp_hard_binding : 1,	/* If we've started a full bind */
232 		tcp_hard_bound : 1,	/* If we've done a full bind with IP */
233 		tcp_fin_acked : 1,	/* Has our FIN been acked? */
234 
235 		tcp_fin_rcvd : 1,	/* Have we seen a FIN? */
236 		tcp_fin_sent : 1,	/* Have we sent our FIN yet? */
237 		tcp_ordrel_done : 1,	/* Have we sent the ord_rel upstream? */
238 		tcp_flow_stopped : 1,	/* Have we flow controlled xmitter? */
239 
240 		tcp_detached : 1,	/* If we're detached from a stream */
241 		tcp_bind_pending : 1,	/* Client is waiting for bind ack */
242 		tcp_unbind_pending : 1, /* Client sent T_UNBIND_REQ */
243 		tcp_deferred_clean_death : 1,
244 					/* defer tcp endpoint cleanup etc. */
245 
246 		tcp_conn_def_q0: 1,	/* move from q0 to q deferred */
247 		tcp_ka_enabled: 1,	/* Connection KeepAlive Timer needed */
248 		tcp_zero_win_probe: 1,	/* Zero win probing is in progress */
249 		tcp_loopback: 1,	/* src and dst are the same machine */
250 
251 		tcp_localnet: 1,	/* src and dst are on the same subnet */
252 		tcp_syn_defense: 1,	/* For defense against SYN attack */
253 #define	tcp_dontdrop	tcp_syn_defense
254 		tcp_set_timer : 1,
255 		tcp_active_open: 1,	/* This is a active open */
256 
257 		tcp_timeout : 1,	/* qbufcall failed, qtimeout pending */
258 		tcp_rexmit : 1,		/* TCP is retransmitting */
259 		tcp_snd_sack_ok : 1,	/* Can use SACK for this connection */
260 		tcp_empty_flag : 1,	/* Empty flag for future use */
261 
262 		tcp_recvdstaddr : 1,	/* return T_EXTCONN_IND with dst addr */
263 		tcp_hwcksum : 1,	/* The NIC is capable of hwcksum */
264 		tcp_ip_forward_progress : 1,
265 		tcp_anon_priv_bind : 1,
266 
267 		tcp_ecn_ok : 1,		/* Can use ECN for this connection */
268 		tcp_ecn_echo_on : 1,	/* Need to do ECN echo */
269 		tcp_ecn_cwr_sent : 1,	/* ECN_CWR has been sent */
270 		tcp_cwr : 1;		/* Cwnd has reduced recently */
271 
272 	/* Following manipulated by TCP under squeue protection */
273 	uint32_t
274 		tcp_mdt : 1,		/* Lower layer is capable of MDT */
275 		tcp_snd_ts_ok  : 1,
276 		tcp_snd_ws_ok  : 1,
277 		tcp_exclbind	: 1,	/* ``exclusive'' binding */
278 
279 		tcp_reserved_port : 1,
280 		tcp_hdr_grown	: 1,
281 		tcp_in_free_list : 1,
282 		tcp_snd_zcopy_on : 1,	/* xmit zero-copy enabled */
283 
284 		tcp_snd_zcopy_aware : 1, /* client is zero-copy aware */
285 		tcp_xmit_zc_clean : 1,	/* the xmit list is free of zc-mblk */
286 		tcp_wait_for_eagers : 1, /* Wait for eagers to disappear */
287 		tcp_accept_error : 1,	/* Error during TLI accept */
288 
289 		tcp_send_discon_ind : 1, /* TLI accept err, send discon ind */
290 		tcp_cork : 1,		/* tcp_cork option */
291 		tcp_tconnind_started : 1, /* conn_ind message is being sent */
292 		tcp_lso :1,		/* Lower layer is capable of LSO */
293 		tcp_pad_to_bit_31 : 16;
294 
295 	uint32_t	tcp_if_mtu;	/* Outgoing interface MTU. */
296 
297 	mblk_t	*tcp_reass_head;	/* Out of order reassembly list head */
298 	mblk_t	*tcp_reass_tail;	/* Out of order reassembly list tail */
299 
300 	tcp_sack_info_t	*tcp_sack_info;
301 
302 #define	tcp_pipe	tcp_sack_info->tcp_pipe
303 #define	tcp_fack	tcp_sack_info->tcp_fack
304 #define	tcp_sack_snxt	tcp_sack_info->tcp_sack_snxt
305 #define	tcp_max_sack_blk	tcp_sack_info->tcp_max_sack_blk
306 #define	tcp_num_sack_blk	tcp_sack_info->tcp_num_sack_blk
307 #define	tcp_sack_list		tcp_sack_info->tcp_sack_list
308 #define	tcp_num_notsack_blk	tcp_sack_info->tcp_num_notsack_blk
309 #define	tcp_cnt_notsack_list	tcp_sack_info->tcp_cnt_notsack_list
310 #define	tcp_notsack_list		tcp_sack_info->tcp_notsack_list
311 
312 	mblk_t	*tcp_rcv_list;		/* Queued until push, urgent data, */
313 	mblk_t	*tcp_rcv_last_head;	/* optdata, or the count exceeds */
314 	mblk_t	*tcp_rcv_last_tail;	/* tcp_rcv_push_wait. */
315 	uint32_t tcp_rcv_cnt;		/* tcp_rcv_list is b_next chain. */
316 
317 	uint32_t tcp_cwnd_ssthresh;	/* Congestion window */
318 	uint32_t tcp_cwnd_max;
319 	uint32_t tcp_csuna;		/* Clear (no rexmits in window) suna */
320 
321 	clock_t	tcp_rtt_sa;		/* Round trip smoothed average */
322 	clock_t	tcp_rtt_sd;		/* Round trip smoothed deviation */
323 	clock_t	tcp_rtt_update;		/* Round trip update(s) */
324 	clock_t tcp_ms_we_have_waited;	/* Total retrans time */
325 
326 	uint32_t tcp_swl1;		/* These help us avoid using stale */
327 	uint32_t tcp_swl2;		/*  packets to update state */
328 
329 	uint32_t tcp_rack;		/* Seq # we have acked */
330 	uint32_t tcp_rack_cnt;		/* # of segs we have deferred ack */
331 	uint32_t tcp_rack_cur_max;	/* # of segs we may defer ack for now */
332 	uint32_t tcp_rack_abs_max;	/* # of segs we may defer ack ever */
333 	timeout_id_t	tcp_ack_tid;	/* Delayed ACK timer ID */
334 	timeout_id_t	tcp_push_tid;	/* Push timer ID */
335 
336 	uint32_t tcp_max_swnd;		/* Maximum swnd we have seen */
337 
338 	struct tcp_s *tcp_listener;	/* Our listener */
339 
340 	int32_t	tcp_xmit_lowater;	/* Send buffer low water mark. */
341 
342 	uint32_t tcp_irs;		/* Initial recv seq num */
343 	uint32_t tcp_fss;		/* Final/fin send seq num */
344 	uint32_t tcp_urg;		/* Urgent data seq num */
345 
346 	clock_t	tcp_first_timer_threshold;  /* When to prod IP */
347 	clock_t	tcp_second_timer_threshold; /* When to give up completely */
348 	clock_t	tcp_first_ctimer_threshold; /* 1st threshold while connecting */
349 	clock_t tcp_second_ctimer_threshold; /* 2nd ... while connecting */
350 
351 	int	tcp_lingertime;		/* Close linger time (in seconds) */
352 
353 	uint32_t tcp_urp_last;		/* Last urp for which signal sent */
354 	mblk_t	*tcp_urp_mp;		/* T_EXDATA_IND for urgent byte */
355 	mblk_t	*tcp_urp_mark_mp;	/* zero-length marked/unmarked msg */
356 
357 	int tcp_conn_req_cnt_q0;	/* # of conn reqs in SYN_RCVD */
358 	int tcp_conn_req_cnt_q;	/* # of conn reqs in ESTABLISHED */
359 	int tcp_conn_req_max;	/* # of ESTABLISHED conn reqs allowed */
360 	t_scalar_t tcp_conn_req_seqnum;	/* Incrementing pending conn req ID */
361 #define	tcp_ip_addr_cache	tcp_reass_tail
362 					/* Cache ip addresses that */
363 					/* complete the 3-way handshake */
364 	kmutex_t  tcp_eager_lock;
365 	struct tcp_s *tcp_eager_next_q; /* next eager in ESTABLISHED state */
366 	struct tcp_s *tcp_eager_last_q;	/* last eager in ESTABLISHED state */
367 	struct tcp_s *tcp_eager_next_q0; /* next eager in SYN_RCVD state */
368 	struct tcp_s *tcp_eager_prev_q0; /* prev eager in SYN_RCVD state */
369 					/* all eagers form a circular list */
370 	union {
371 	    mblk_t *tcp_eager_conn_ind; /* T_CONN_IND waiting for 3rd ack. */
372 	    mblk_t *tcp_opts_conn_req; /* T_CONN_REQ w/ options processed */
373 	} tcp_conn;
374 	uint32_t tcp_syn_rcvd_timeout;	/* How many SYN_RCVD timeout in q0 */
375 
376 	/* TCP Keepalive Timer members */
377 	int32_t	tcp_ka_last_intrvl;	/* Last probe interval */
378 	timeout_id_t tcp_ka_tid;	/* Keepalive timer ID */
379 	uint32_t tcp_ka_interval;	/* Keepalive interval */
380 	uint32_t tcp_ka_abort_thres;	/* Keepalive abort threshold */
381 
382 	int32_t	tcp_client_errno;	/* How the client screwed up */
383 
384 	char	*tcp_iphc;		/* Buffer holding tcp/ip hdr template */
385 	int	tcp_iphc_len;		/* actual allocated buffer size */
386 	int32_t	tcp_hdr_len;		/* Byte len of combined TCP/IP hdr */
387 	ipha_t	*tcp_ipha;		/* IPv4 header in the buffer */
388 	ip6_t	*tcp_ip6h;		/* IPv6 header in the buffer */
389 	int	tcp_ip_hdr_len;		/* Byte len of our current IPvx hdr */
390 	tcph_t	*tcp_tcph;		/* tcp header within combined hdr */
391 	int32_t	tcp_tcp_hdr_len;	/* tcp header len within combined */
392 
393 	uint32_t tcp_sum;		/* checksum to compensate for source */
394 					/* routed packets. Host byte order */
395 	uint16_t tcp_last_sent_len;	/* Record length for nagle */
396 	uint16_t tcp_dupack_cnt;	/* # of consequtive duplicate acks */
397 
398 	kmutex_t	*tcp_acceptor_lockp;	/* Ptr to tf_lock */
399 
400 	timeout_id_t	tcp_ordrelid;		/* qbufcall/qtimeout id */
401 	t_uscalar_t	tcp_acceptor_id;	/* ACCEPTOR_id */
402 
403 	int		tcp_ipsec_overhead;
404 	/*
405 	 * Address family that app wishes returned addrsses to be in.
406 	 * Currently taken from address family used in T_BIND_REQ, but
407 	 * should really come from family used in original socket() call.
408 	 * Value can be AF_INET or AF_INET6.
409 	 */
410 	uint_t	tcp_family;
411 	/*
412 	 * used for a quick test to determine if any ancillary bits are
413 	 * set
414 	 */
415 	uint_t		tcp_ipv6_recvancillary;		/* Flags */
416 #define	TCP_IPV6_RECVPKTINFO	0x01	/* IPV6_RECVPKTINFO option  */
417 #define	TCP_IPV6_RECVHOPLIMIT	0x02	/* IPV6_RECVHOPLIMIT option */
418 #define	TCP_IPV6_RECVHOPOPTS	0x04	/* IPV6_RECVHOPOPTS option */
419 #define	TCP_IPV6_RECVDSTOPTS	0x08	/* IPV6_RECVDSTOPTS option */
420 #define	TCP_IPV6_RECVRTHDR	0x10	/* IPV6_RECVRTHDR option */
421 #define	TCP_IPV6_RECVRTDSTOPTS	0x20	/* IPV6_RECVRTHDRDSTOPTS option */
422 #define	TCP_IPV6_RECVTCLASS	0x40	/* IPV6_RECVTCLASS option */
423 #define	TCP_OLD_IPV6_RECVDSTOPTS 0x80	/* old IPV6_RECVDSTOPTS option */
424 
425 	uint_t		tcp_recvifindex; /* Last received IPV6_RCVPKTINFO */
426 	uint_t		tcp_recvhops;	/* Last received IPV6_RECVHOPLIMIT */
427 	uint_t		tcp_recvtclass;	/* Last received IPV6_RECVTCLASS */
428 	ip6_hbh_t	*tcp_hopopts;	/* Last received IPV6_RECVHOPOPTS */
429 	ip6_dest_t	*tcp_dstopts;	/* Last received IPV6_RECVDSTOPTS */
430 	ip6_dest_t	*tcp_rtdstopts;	/* Last recvd IPV6_RECVRTHDRDSTOPTS */
431 	ip6_rthdr_t	*tcp_rthdr;	/* Last received IPV6_RECVRTHDR */
432 	uint_t		tcp_hopoptslen;
433 	uint_t		tcp_dstoptslen;
434 	uint_t		tcp_rtdstoptslen;
435 	uint_t		tcp_rthdrlen;
436 
437 	mblk_t		*tcp_timercache;
438 	cred_t		*tcp_cred;	/* Credentials when this was opened */
439 	pid_t		tcp_cpid;	/* Process id when this was opened */
440 	uint64_t	tcp_open_time;	/* time when this was opened */
441 
442 
443 	union {
444 		struct {
445 			uchar_t	v4_ttl;
446 				/* Dup of tcp_ipha.iph_type_of_service */
447 			uchar_t	v4_tos; /* Dup of tcp_ipha.iph_ttl */
448 		} v4_hdr_info;
449 		struct {
450 			uint_t	v6_vcf;		/* Dup of tcp_ip6h.ip6h_vcf */
451 			uchar_t	v6_hops;	/* Dup of tcp_ip6h.ip6h_hops */
452 		} v6_hdr_info;
453 	} tcp_hdr_info;
454 #define	tcp_ttl	tcp_hdr_info.v4_hdr_info.v4_ttl
455 #define	tcp_tos	tcp_hdr_info.v4_hdr_info.v4_tos
456 #define	tcp_ip6_vcf	tcp_hdr_info.v6_hdr_info.v6_vcf
457 #define	tcp_ip6_hops	tcp_hdr_info.v6_hdr_info.v6_hops
458 
459 	ushort_t	tcp_ipversion;
460 	uint_t		tcp_bound_if;	/* IPV6_BOUND_IF */
461 
462 #define	tcp_lport	tcp_connp->conn_lport
463 #define	tcp_fport	tcp_connp->conn_fport
464 #define	tcp_ports	tcp_connp->conn_ports
465 
466 #define	tcp_remote	tcp_connp->conn_rem
467 #define	tcp_ip_src	tcp_connp->conn_src
468 
469 #define	tcp_remote_v6	tcp_connp->conn_remv6
470 #define	tcp_ip_src_v6	tcp_connp->conn_srcv6
471 #define	tcp_bound_source_v6	tcp_connp->conn_bound_source_v6
472 #define	tcp_bound_source	tcp_connp->conn_bound_source
473 
474 	kmutex_t	tcp_closelock;
475 	kcondvar_t	tcp_closecv;
476 	uint8_t		tcp_closed;
477 	uint8_t		tcp_closeflags;
478 	uint8_t		tcp_cleandeathtag;
479 	mblk_t		tcp_closemp;
480 	timeout_id_t	tcp_linger_tid;	/* Linger timer ID */
481 	void		*tcp_tracebuf;
482 
483 	struct tcp_s *tcp_acceptor_hash; /* Acceptor hash chain */
484 	struct tcp_s **tcp_ptpahn; /* Pointer to previous accept hash next. */
485 	struct tcp_s *tcp_bind_hash; /* Bind hash chain */
486 	struct tcp_s **tcp_ptpbhn;
487 
488 	boolean_t	tcp_ire_ill_check_done;
489 	uint_t		tcp_maxpsz;
490 
491 	/*
492 	 * used for Multidata Transmit
493 	 */
494 	uint_t	tcp_mdt_hdr_head; /* leading header fragment extra space */
495 	uint_t	tcp_mdt_hdr_tail; /* trailing header fragment extra space */
496 	int	tcp_mdt_max_pld;  /* maximum payload buffers per Multidata */
497 
498 	uint32_t	tcp_lso_max; /* maximum LSO payload */
499 
500 	uint32_t	tcp_ofo_fin_seq; /* Recv out of order FIN seq num */
501 	uint32_t	tcp_cwr_snd_max;
502 	uint_t		tcp_drop_opt_ack_cnt; /* # tcp generated optmgmt */
503 	ip6_pkt_t	tcp_sticky_ipp;			/* Sticky options */
504 #define	tcp_ipp_fields	tcp_sticky_ipp.ipp_fields	/* valid fields */
505 #define	tcp_ipp_ifindex	tcp_sticky_ipp.ipp_ifindex	/* pktinfo ifindex */
506 #define	tcp_ipp_addr	tcp_sticky_ipp.ipp_addr	/* pktinfo src/dst addr */
507 #define	tcp_ipp_hoplimit	tcp_sticky_ipp.ipp_hoplimit
508 #define	tcp_ipp_hopoptslen	tcp_sticky_ipp.ipp_hopoptslen
509 #define	tcp_ipp_rtdstoptslen	tcp_sticky_ipp.ipp_rtdstoptslen
510 #define	tcp_ipp_rthdrlen	tcp_sticky_ipp.ipp_rthdrlen
511 #define	tcp_ipp_dstoptslen	tcp_sticky_ipp.ipp_dstoptslen
512 #define	tcp_ipp_hopopts		tcp_sticky_ipp.ipp_hopopts
513 #define	tcp_ipp_rtdstopts	tcp_sticky_ipp.ipp_rtdstopts
514 #define	tcp_ipp_rthdr		tcp_sticky_ipp.ipp_rthdr
515 #define	tcp_ipp_dstopts		tcp_sticky_ipp.ipp_dstopts
516 #define	tcp_ipp_nexthop		tcp_sticky_ipp.ipp_nexthop
517 #define	tcp_ipp_use_min_mtu	tcp_sticky_ipp.ipp_use_min_mtu
518 	struct tcp_s *tcp_saved_listener;	/* saved value of listener */
519 
520 	uint32_t	tcp_in_ack_unsent;	/* ACK for unsent data cnt. */
521 
522 	/*
523 	 * The following fusion-related fields are protected by squeue.
524 	 */
525 	struct tcp_s *tcp_loopback_peer;	/* peer tcp for loopback */
526 	mblk_t	*tcp_fused_sigurg_mp;		/* M_PCSIG mblk for SIGURG */
527 	size_t	tcp_fuse_rcv_hiwater;		/* fusion receive queue size */
528 	uint_t	tcp_fuse_rcv_unread_hiwater;	/* max # of outstanding pkts */
529 	/*
530 	 * The following fusion-related fields and bit fields are to be
531 	 * manipulated with squeue protection or with tcp_fuse_lock held.
532 	 */
533 	kmutex_t tcp_fuse_lock;
534 	kcondvar_t tcp_fuse_plugcv;
535 	uint_t tcp_fuse_rcv_unread_cnt;	/* # of outstanding pkts */
536 	uint32_t
537 		tcp_fused : 1,		/* loopback tcp in fusion mode */
538 		tcp_unfusable : 1,	/* fusion not allowed on endpoint */
539 		tcp_fused_sigurg : 1,	/* send SIGURG upon draining */
540 		tcp_direct_sockfs : 1,	/* direct calls to sockfs */
541 
542 		tcp_fuse_syncstr_stopped : 1, /* synchronous streams stopped */
543 		tcp_fuse_syncstr_plugged : 1, /* synchronous streams plugged */
544 		tcp_fuse_to_bit_31 : 26;
545 
546 	/*
547 	 * This variable is accessed without any lock protection
548 	 * and therefore must not be declared as a bit field along
549 	 * with the rest which require such condition.
550 	 */
551 	boolean_t	tcp_issocket;	/* this is a socket tcp */
552 
553 	uint32_t	tcp_squeue_bytes;
554 	/*
555 	 * Kernel SSL session information
556 	 */
557 	boolean_t		tcp_kssl_pending; /* waiting for 1st SSL rec. */
558 	boolean_t		tcp_kssl_inhandshake; /* during SSL handshake */
559 	kssl_ent_t		tcp_kssl_ent;	/* SSL table entry */
560 	kssl_ctx_t		tcp_kssl_ctx;	/* SSL session */
561 	uint_t	tcp_label_len;	/* length of cached label */
562 
563 	/*
564 	 * tcp_closemp_used is protected by listener's tcp_eager_lock
565 	 * when used for eagers. When used for a tcp in TIME_WAIT state
566 	 * or in tcp_close(), it is not protected by any lock as we
567 	 * do not expect any other thread to use it concurrently.
568 	 * Since we do allow re-use of tcp_closemp at certain places,
569 	 * tcp_closemp_used is declared as uint32_t instead of boolean_t
570 	 * to record any attempt to re-use tcp_closemp while it is still
571 	 * in use. This would facilitate debugging in non-debug kernels.
572 	 */
573 	uint32_t tcp_closemp_used;
574 
575 	/*
576 	 * previous and next eagers in the list of droppable eagers. See
577 	 * the comments before MAKE_DROPPABLE(). These pointers are
578 	 * protected by listener's tcp_eager_lock.
579 	 */
580 	struct tcp_s	*tcp_eager_prev_drop_q0;
581 	struct tcp_s	*tcp_eager_next_drop_q0;
582 #ifdef DEBUG
583 	pc_t			tcmp_stk[15];
584 #endif
585 } tcp_t;
586 
587 #ifdef DEBUG
588 #define	TCP_DEBUG_GETPCSTACK(buffer, depth)	((void) getpcstack(buffer, \
589 						    depth))
590 #else
591 #define	TCP_DEBUG_GETPCSTACK(buffer, depth)
592 #endif
593 
594 extern void 	tcp_free(tcp_t *tcp);
595 extern void	tcp_ddi_init(void);
596 extern void	tcp_ddi_destroy(void);
597 extern void	tcp_xmit_listeners_reset(mblk_t *mp, uint_t ip_hdr_len,
598 		    zoneid_t zoneid);
599 extern void	tcp_conn_request(void *arg, mblk_t *mp, void *arg2);
600 extern void	tcp_conn_request_unbound(void *arg, mblk_t *mp, void *arg2);
601 extern void 	tcp_input(void *arg, mblk_t *mp, void *arg2);
602 extern void	tcp_rput_data(void *arg, mblk_t *mp, void *arg2);
603 extern void 	*tcp_get_conn(void *arg);
604 extern void	tcp_time_wait_collector(void *arg);
605 extern int	tcp_snmp_get(queue_t *, mblk_t *);
606 extern int	tcp_snmp_set(queue_t *, int, int, uchar_t *, int len);
607 extern mblk_t	*tcp_xmit_mp(tcp_t *tcp, mblk_t *mp, int32_t max_to_send,
608 		    int32_t *offset, mblk_t **end_mp, uint32_t seq,
609 		    boolean_t sendall, uint32_t *seg_len, boolean_t rexmit);
610 /*
611  * The TCP Fanout structure.
612  * The hash tables and their linkage (tcp_*_hash_next, tcp_ptp*hn) are
613  * protected by the per-bucket tf_lock. Each tcp_t
614  * inserted in the list points back at this lock using tcp_*_lockp.
615  *
616  * The listener and acceptor hash queues are lists of tcp_t.
617  */
618 
619 /* listener hash and acceptor hash queue head */
620 typedef struct tf_s {
621 	tcp_t		*tf_tcp;
622 	kmutex_t	tf_lock;
623 } tf_t;
624 
625 extern mib2_tcp_t tcp_mib;
626 
627 #endif	/* (defined(_KERNEL) || defined(_KMEMUSER)) */
628 
629 /* Contract private interface between TCP and Clustering. */
630 
631 #define	CL_TCPI_V1	1	/* cl_tcpi_version number */
632 
633 typedef struct cl_tcp_info_s {
634 	ushort_t	cl_tcpi_version;	/* cl_tcp_info_t's version no */
635 	ushort_t	cl_tcpi_ipversion;	/* IP version */
636 	int32_t		cl_tcpi_state;		/* TCP state */
637 	in_port_t	cl_tcpi_lport;		/* Local port */
638 	in_port_t	cl_tcpi_fport;		/* Remote port */
639 	in6_addr_t	cl_tcpi_laddr_v6;	/* Local IP address */
640 	in6_addr_t	cl_tcpi_faddr_v6;	/* Remote IP address */
641 #ifdef _KERNEL
642 /* Note: V4_PART_OF_V6 is meant to be used only for _KERNEL defined stuff */
643 #define	cl_tcpi_laddr	V4_PART_OF_V6(cl_tcpi_laddr_v6)
644 #define	cl_tcpi_faddr	V4_PART_OF_V6(cl_tcpi_faddr_v6)
645 
646 #endif	/* _KERNEL */
647 } cl_tcp_info_t;
648 
649 /*
650  * Hook functions to enable cluster networking
651  * On non-clustered systems these vectors must always be NULL.
652  */
653 
654 extern void	(*cl_inet_listen)(uint8_t, sa_family_t, uint8_t *, in_port_t);
655 extern void	(*cl_inet_unlisten)(uint8_t, sa_family_t, uint8_t *,
656 		    in_port_t);
657 
658 /*
659  * Contracted Consolidation Private ioctl for aborting TCP connections.
660  * In order to keep the offsets and size of the structure the same between
661  * a 32-bit application and a 64-bit amd64 kernel, we use a #pragma
662  * pack(4).
663  */
664 #define	TCP_IOC_ABORT_CONN	(('T' << 8) + 91)
665 
666 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
667 #pragma pack(4)
668 #endif
669 
670 typedef struct tcp_ioc_abort_conn_s {
671 	struct sockaddr_storage ac_local;	/* local addr and port */
672 	struct sockaddr_storage ac_remote;	/* remote addr and port */
673 	int32_t ac_start;			/* start state */
674 	int32_t ac_end;				/* end state  */
675 	int32_t ac_zoneid;			/* zoneid */
676 } tcp_ioc_abort_conn_t;
677 
678 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
679 #pragma pack()
680 #endif
681 
682 #if (defined(_KERNEL) || defined(_KMEMUSER))
683 extern void tcp_rput_other(tcp_t *tcp, mblk_t *mp);
684 #endif
685 
686 #ifdef	__cplusplus
687 }
688 #endif
689 
690 #endif	/* _INET_TCP_H */
691