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 2008 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 <sys/socket.h>
40 #include <sys/sodirect.h>
41 #include <sys/multidata.h>
42 #include <sys/md5.h>
43 #include <inet/common.h>
44 #include <inet/ip.h>
45 #include <inet/ip6.h>
46 #include <inet/mi.h>
47 #include <inet/mib2.h>
48 #include <inet/tcp_stack.h>
49 #include <inet/tcp_sack.h>
50 #include <inet/kssl/ksslapi.h>
51
52 /*
53 * Private (and possibly temporary) ioctl used by configuration code
54 * to lock in the "default" stream for detached closes.
55 */
56 #define TCP_IOC_DEFAULT_Q (('T' << 8) + 51)
57
58 /* TCP states */
59 #define TCPS_CLOSED -6
60 #define TCPS_IDLE -5 /* idle (opened, but not bound) */
61 #define TCPS_BOUND -4 /* bound, ready to connect or accept */
62 #define TCPS_LISTEN -3 /* listening for connection */
63 #define TCPS_SYN_SENT -2 /* active, have sent syn */
64 #define TCPS_SYN_RCVD -1 /* have received syn (and sent ours) */
65 /* states < TCPS_ESTABLISHED are those where connections not established */
66 #define TCPS_ESTABLISHED 0 /* established */
67 #define TCPS_CLOSE_WAIT 1 /* rcvd fin, waiting for close */
68 /* states > TCPS_CLOSE_WAIT are those where user has closed */
69 #define TCPS_FIN_WAIT_1 2 /* have closed and sent fin */
70 #define TCPS_CLOSING 3 /* closed, xchd FIN, await FIN ACK */
71 #define TCPS_LAST_ACK 4 /* had fin and close; await FIN ACK */
72 /* states > TCPS_CLOSE_WAIT && < TCPS_FIN_WAIT_2 await ACK of FIN */
73 #define TCPS_FIN_WAIT_2 5 /* have closed, fin is acked */
74 #define TCPS_TIME_WAIT 6 /* in 2*msl quiet wait after close */
75
76 /*
77 * Internal flags used in conjunction with the packet header flags.
78 * Used in tcp_rput_data to keep track of what needs to be done.
79 */
80 #define TH_LIMIT_XMIT 0x0400 /* Limited xmit is needed */
81 #define TH_XMIT_NEEDED 0x0800 /* Window opened - send queued data */
82 #define TH_REXMIT_NEEDED 0x1000 /* Time expired for unacked data */
83 #define TH_ACK_NEEDED 0x2000 /* Send an ack now. */
84 #define TH_NEED_SACK_REXMIT 0x4000 /* Use SACK info to retransmission */
85 #define TH_ACK_TIMER_NEEDED 0x8000 /* Start the delayed ACK timer */
86 #define TH_ORDREL_NEEDED 0x10000 /* Generate an ordrel indication */
87 #define TH_MARKNEXT_NEEDED 0x20000 /* Data should have MSGMARKNEXT */
88 #define TH_SEND_URP_MARK 0x40000 /* Send up tcp_urp_mark_mp */
89
90 /*
91 * TCP sequence numbers are 32 bit integers operated
92 * on with modular arithmetic. These macros can be
93 * used to compare such integers.
94 */
95 #define SEQ_LT(a, b) ((int32_t)((a)-(b)) < 0)
96 #define SEQ_LEQ(a, b) ((int32_t)((a)-(b)) <= 0)
97 #define SEQ_GT(a, b) ((int32_t)((a)-(b)) > 0)
98 #define SEQ_GEQ(a, b) ((int32_t)((a)-(b)) >= 0)
99
100 /* TCP Protocol header */
101 typedef struct tcphdr_s {
102 uint8_t th_lport[2]; /* Source port */
103 uint8_t th_fport[2]; /* Destination port */
104 uint8_t th_seq[4]; /* Sequence number */
105 uint8_t th_ack[4]; /* Acknowledgement number */
106 uint8_t th_offset_and_rsrvd[1]; /* Offset to the packet data */
107 uint8_t th_flags[1];
108 uint8_t th_win[2]; /* Allocation number */
109 uint8_t th_sum[2]; /* TCP checksum */
110 uint8_t th_urp[2]; /* Urgent pointer */
111 } tcph_t;
112
113 #define TCP_HDR_LENGTH(tcph) (((tcph)->th_offset_and_rsrvd[0] >>2) &(0xF << 2))
114 #define TCP_MAX_COMBINED_HEADER_LENGTH (60 + 60) /* Maxed out ip + tcp */
115 #define TCP_MAX_IP_OPTIONS_LENGTH (60 - IP_SIMPLE_HDR_LENGTH)
116 #define TCP_MAX_HDR_LENGTH 60
117 #define TCP_MAX_TCP_OPTIONS_LENGTH (60 - sizeof (tcph_t))
118 #define TCP_MIN_HEADER_LENGTH 20
119 #define TCP_MAXWIN 65535
120 #define TCP_PORT_LEN sizeof (in_port_t)
121 #define TCP_MAX_WINSHIFT 14
122 #define TCP_MAX_LARGEWIN (TCP_MAXWIN << TCP_MAX_WINSHIFT)
123 #define TCP_MAX_LSO_LENGTH (IP_MAXPACKET - TCP_MAX_COMBINED_HEADER_LENGTH)
124
125 #define TCPIP_HDR_LENGTH(mp, n) \
126 (n) = IPH_HDR_LENGTH((mp)->b_rptr), \
127 (n) += TCP_HDR_LENGTH((tcph_t *)&(mp)->b_rptr[(n)])
128
129 /* TCP Protocol header (used if the header is known to be 32-bit aligned) */
130 typedef struct tcphdra_s {
131 in_port_t tha_lport; /* Source port */
132 in_port_t tha_fport; /* Destination port */
133 uint32_t tha_seq; /* Sequence number */
134 uint32_t tha_ack; /* Acknowledgement number */
135 uint8_t tha_offset_and_reserved; /* Offset to the packet data */
136 uint8_t tha_flags;
137 uint16_t tha_win; /* Allocation number */
138 uint16_t tha_sum; /* TCP checksum */
139 uint16_t tha_urp; /* Urgent pointer */
140 } tcpha_t;
141
142 struct conn_s;
143
144 /*
145 * Control structure for each open TCP stream,
146 * defined only within the kernel or for a kmem user.
147 * NOTE: tcp_reinit_values MUST have a line for each field in this structure!
148 */
149 #if (defined(_KERNEL) || defined(_KMEMUSER))
150
151 typedef struct tcp_s {
152 /* Pointer to previous bind hash next. */
153 struct tcp_s *tcp_time_wait_next;
154 /* Pointer to next T/W block */
155 struct tcp_s *tcp_time_wait_prev;
156 /* Pointer to previous T/W next */
157 clock_t tcp_time_wait_expire;
158
159 struct conn_s *tcp_connp;
160 tcp_stack_t *tcp_tcps; /* Shortcut via conn_netstack */
161
162 int32_t tcp_state;
163 int32_t tcp_rcv_ws; /* My window scale power */
164 int32_t tcp_snd_ws; /* Sender's window scale power */
165 uint32_t tcp_ts_recent; /* Timestamp of earliest unacked */
166 /* data segment */
167 clock_t tcp_rto; /* Round trip timeout */
168 clock_t tcp_last_rcv_lbolt;
169 /* lbolt on last packet, used for PAWS */
170
171 uint32_t tcp_snxt; /* Senders next seq num */
172 uint32_t tcp_swnd; /* Senders window (relative to suna) */
173 uint32_t tcp_mss; /* Max segment size */
174 uint32_t tcp_iss; /* Initial send seq num */
175 uint32_t tcp_rnxt; /* Seq we expect to recv next */
176 uint32_t tcp_rwnd;
177
178 queue_t *tcp_rq; /* Our upstream neighbor (client) */
179 queue_t *tcp_wq; /* Our downstream neighbor */
180
181 /* Fields arranged in approximate access order along main paths */
182 mblk_t *tcp_xmit_head; /* Head of rexmit list */
183 mblk_t *tcp_xmit_last; /* last valid data seen by tcp_wput */
184 mblk_t *tcp_xmit_tail; /* Last rexmit data sent */
185 uint32_t tcp_unsent; /* # of bytes in hand that are unsent */
186 uint32_t tcp_xmit_tail_unsent; /* # of unsent bytes in xmit_tail */
187
188 uint32_t tcp_suna; /* Sender unacknowledged */
189 uint32_t tcp_rexmit_nxt; /* Next rexmit seq num */
190 uint32_t tcp_rexmit_max; /* Max retran seq num */
191 int32_t tcp_snd_burst; /* Send burst factor */
192 uint32_t tcp_cwnd; /* Congestion window */
193 int32_t tcp_cwnd_cnt; /* cwnd cnt in congestion avoidance */
194
195 uint32_t tcp_ibsegs; /* Inbound segments on this stream */
196 uint32_t tcp_obsegs; /* Outbound segments on this stream */
197
198 uint32_t tcp_naglim; /* Tunable nagle limit */
199 uint32_t tcp_valid_bits;
200 #define TCP_ISS_VALID 0x1 /* Is the tcp_iss seq num active? */
201 #define TCP_FSS_VALID 0x2 /* Is the tcp_fss seq num active? */
202 #define TCP_URG_VALID 0x4 /* Is the tcp_urg seq num active? */
203 #define TCP_OFO_FIN_VALID 0x8 /* Has TCP received an out of order FIN? */
204
205
206 int32_t tcp_xmit_hiwater; /* Send buffer high water mark. */
207
208 timeout_id_t tcp_timer_tid; /* Control block for timer service */
209 uchar_t tcp_timer_backoff; /* Backoff shift count. */
210 int64_t tcp_last_recv_time; /* Last time we receive a segment. */
211 uint32_t tcp_init_cwnd; /* Initial cwnd (start/restart) */
212
213 /*
214 * Following socket options are set by sockfs outside the squeue
215 * and we want to separate these bit fields from the other bit fields
216 * set by TCP to avoid grabbing locks. sockfs ensures that only one
217 * thread in sockfs can set a socket option at a time on a conn_t.
218 * However TCP may read these options concurrently. The linger option
219 * needs atomicity since tcp_lingertime also needs to be in sync.
220 * However TCP uses it only during close, and by then no socket option
221 * can come down. So we don't need any locks, instead just separating
222 * the sockfs settable bit fields from the other bit fields is
223 * sufficient.
224 */
225 uint32_t
226 tcp_debug : 1, /* SO_DEBUG "socket" option. */
227 tcp_dontroute : 1, /* SO_DONTROUTE "socket" option. */
228 tcp_broadcast : 1, /* SO_BROADCAST "socket" option. */
229 tcp_useloopback : 1, /* SO_USELOOPBACK "socket" option. */
230
231 tcp_oobinline : 1, /* SO_OOBINLINE "socket" option. */
232 tcp_dgram_errind : 1, /* SO_DGRAM_ERRIND option */
233 tcp_linger : 1, /* SO_LINGER turned on */
234 tcp_reuseaddr : 1, /* SO_REUSEADDR "socket" option. */
235
236 tcp_junk_to_bit_31 : 24;
237
238 /* Following manipulated by TCP under squeue protection */
239 uint32_t
240 tcp_urp_last_valid : 1, /* Is tcp_urp_last valid? */
241 tcp_hard_binding : 1, /* If we've started a full bind */
242 tcp_hard_bound : 1, /* If we've done a full bind with IP */
243 tcp_fin_acked : 1, /* Has our FIN been acked? */
244
245 tcp_fin_rcvd : 1, /* Have we seen a FIN? */
246 tcp_fin_sent : 1, /* Have we sent our FIN yet? */
247 tcp_ordrel_done : 1, /* Have we sent the ord_rel upstream? */
248 tcp_detached : 1, /* If we're detached from a stream */
249
250 tcp_bind_pending : 1, /* Client is waiting for bind ack */
251 tcp_unbind_pending : 1, /* Client sent T_UNBIND_REQ */
252 tcp_deferred_clean_death : 1,
253 /* defer tcp endpoint cleanup etc. */
254 tcp_conn_def_q0: 1, /* move from q0 to q deferred */
255
256 tcp_ka_enabled: 1, /* Connection KeepAlive Timer needed */
257 tcp_zero_win_probe: 1, /* Zero win probing is in progress */
258 tcp_loopback: 1, /* src and dst are the same machine */
259 tcp_localnet: 1, /* src and dst are on the same subnet */
260
261 tcp_syn_defense: 1, /* For defense against SYN attack */
262 #define tcp_dontdrop tcp_syn_defense
263 tcp_set_timer : 1,
264 tcp_active_open: 1, /* This is a active open */
265 tcp_timeout : 1, /* qbufcall failed, qtimeout pending */
266
267 tcp_rexmit : 1, /* TCP is retransmitting */
268 tcp_snd_sack_ok : 1, /* Can use SACK for this connection */
269 tcp_empty_flag : 1, /* Empty flag for future use */
270 tcp_recvdstaddr : 1, /* return T_EXTCONN_IND with dst addr */
271
272 tcp_hwcksum : 1, /* The NIC is capable of hwcksum */
273 tcp_ip_forward_progress : 1,
274 tcp_anon_priv_bind : 1,
275 tcp_ecn_ok : 1, /* Can use ECN for this connection */
276
277 tcp_ecn_echo_on : 1, /* Need to do ECN echo */
278 tcp_ecn_cwr_sent : 1, /* ECN_CWR has been sent */
279 tcp_cwr : 1, /* Cwnd has reduced recently */
280 tcp_pad_to_bit31 : 1;
281 /* Following manipulated by TCP under squeue protection */
282 uint32_t
283 tcp_mdt : 1, /* Lower layer is capable of MDT */
284 tcp_snd_ts_ok : 1,
285 tcp_snd_ws_ok : 1,
286 tcp_exclbind : 1, /* ``exclusive'' binding */
287
288 tcp_reserved_port : 1,
289 tcp_hdr_grown : 1,
290 tcp_in_free_list : 1,
291 tcp_snd_zcopy_on : 1, /* xmit zero-copy enabled */
292
293 tcp_snd_zcopy_aware : 1, /* client is zero-copy aware */
294 tcp_xmit_zc_clean : 1, /* the xmit list is free of zc-mblk */
295 tcp_wait_for_eagers : 1, /* Wait for eagers to disappear */
296 tcp_accept_error : 1, /* Error during TLI accept */
297
298 tcp_send_discon_ind : 1, /* TLI accept err, send discon ind */
299 tcp_cork : 1, /* tcp_cork option */
300 tcp_tconnind_started : 1, /* conn_ind message is being sent */
301 tcp_lso :1, /* Lower layer is capable of LSO */
302 tcp_pad_to_bit_31 : 16;
303
304 uint32_t tcp_if_mtu; /* Outgoing interface MTU. */
305
306 mblk_t *tcp_reass_head; /* Out of order reassembly list head */
307 mblk_t *tcp_reass_tail; /* Out of order reassembly list tail */
308
309 tcp_sack_info_t *tcp_sack_info;
310
311 #define tcp_pipe tcp_sack_info->tcp_pipe
312 #define tcp_fack tcp_sack_info->tcp_fack
313 #define tcp_sack_snxt tcp_sack_info->tcp_sack_snxt
314 #define tcp_max_sack_blk tcp_sack_info->tcp_max_sack_blk
315 #define tcp_num_sack_blk tcp_sack_info->tcp_num_sack_blk
316 #define tcp_sack_list tcp_sack_info->tcp_sack_list
317 #define tcp_num_notsack_blk tcp_sack_info->tcp_num_notsack_blk
318 #define tcp_cnt_notsack_list tcp_sack_info->tcp_cnt_notsack_list
319 #define tcp_notsack_list tcp_sack_info->tcp_notsack_list
320
321 mblk_t *tcp_rcv_list; /* Queued until push, urgent data, */
322 mblk_t *tcp_rcv_last_head; /* optdata, or the count exceeds */
323 mblk_t *tcp_rcv_last_tail; /* tcp_rcv_push_wait. */
324 uint32_t tcp_rcv_cnt; /* tcp_rcv_list is b_next chain. */
325
326 uint32_t tcp_cwnd_ssthresh; /* Congestion window */
327 uint32_t tcp_cwnd_max;
328 uint32_t tcp_csuna; /* Clear (no rexmits in window) suna */
329
330 clock_t tcp_rtt_sa; /* Round trip smoothed average */
331 clock_t tcp_rtt_sd; /* Round trip smoothed deviation */
332 clock_t tcp_rtt_update; /* Round trip update(s) */
333 clock_t tcp_ms_we_have_waited; /* Total retrans time */
334
335 uint32_t tcp_swl1; /* These help us avoid using stale */
336 uint32_t tcp_swl2; /* packets to update state */
337
338 uint32_t tcp_rack; /* Seq # we have acked */
339 uint32_t tcp_rack_cnt; /* # of segs we have deferred ack */
340 uint32_t tcp_rack_cur_max; /* # of segs we may defer ack for now */
341 uint32_t tcp_rack_abs_max; /* # of segs we may defer ack ever */
342 timeout_id_t tcp_ack_tid; /* Delayed ACK timer ID */
343 timeout_id_t tcp_push_tid; /* Push timer ID */
344
345 uint32_t tcp_max_swnd; /* Maximum swnd we have seen */
346
347 struct tcp_s *tcp_listener; /* Our listener */
348
349 int32_t tcp_xmit_lowater; /* Send buffer low water mark. */
350
351 uint32_t tcp_irs; /* Initial recv seq num */
352 uint32_t tcp_fss; /* Final/fin send seq num */
353 uint32_t tcp_urg; /* Urgent data seq num */
354
355 clock_t tcp_first_timer_threshold; /* When to prod IP */
356 clock_t tcp_second_timer_threshold; /* When to give up completely */
357 clock_t tcp_first_ctimer_threshold; /* 1st threshold while connecting */
358 clock_t tcp_second_ctimer_threshold; /* 2nd ... while connecting */
359
360 int tcp_lingertime; /* Close linger time (in seconds) */
361
362 uint32_t tcp_urp_last; /* Last urp for which signal sent */
363 mblk_t *tcp_urp_mp; /* T_EXDATA_IND for urgent byte */
364 mblk_t *tcp_urp_mark_mp; /* zero-length marked/unmarked msg */
365
366 int tcp_conn_req_cnt_q0; /* # of conn reqs in SYN_RCVD */
367 int tcp_conn_req_cnt_q; /* # of conn reqs in ESTABLISHED */
368 int tcp_conn_req_max; /* # of ESTABLISHED conn reqs allowed */
369 t_scalar_t tcp_conn_req_seqnum; /* Incrementing pending conn req ID */
370 #define tcp_ip_addr_cache tcp_reass_tail
371 /* Cache ip addresses that */
372 /* complete the 3-way handshake */
373 kmutex_t tcp_eager_lock;
374 struct tcp_s *tcp_eager_next_q; /* next eager in ESTABLISHED state */
375 struct tcp_s *tcp_eager_last_q; /* last eager in ESTABLISHED state */
376 struct tcp_s *tcp_eager_next_q0; /* next eager in SYN_RCVD state */
377 struct tcp_s *tcp_eager_prev_q0; /* prev eager in SYN_RCVD state */
378 /* all eagers form a circular list */
379 union {
380 mblk_t *tcp_eager_conn_ind; /* T_CONN_IND waiting for 3rd ack. */
381 mblk_t *tcp_opts_conn_req; /* T_CONN_REQ w/ options processed */
382 } tcp_conn;
383 uint32_t tcp_syn_rcvd_timeout; /* How many SYN_RCVD timeout in q0 */
384
385 /* TCP Keepalive Timer members */
386 int32_t tcp_ka_last_intrvl; /* Last probe interval */
387 timeout_id_t tcp_ka_tid; /* Keepalive timer ID */
388 uint32_t tcp_ka_interval; /* Keepalive interval */
389 uint32_t tcp_ka_abort_thres; /* Keepalive abort threshold */
390
391 int32_t tcp_client_errno; /* How the client screwed up */
392
393 char *tcp_iphc; /* Buffer holding tcp/ip hdr template */
394 int tcp_iphc_len; /* actual allocated buffer size */
395 int32_t tcp_hdr_len; /* Byte len of combined TCP/IP hdr */
396 ipha_t *tcp_ipha; /* IPv4 header in the buffer */
397 ip6_t *tcp_ip6h; /* IPv6 header in the buffer */
398 int tcp_ip_hdr_len; /* Byte len of our current IPvx hdr */
399 tcph_t *tcp_tcph; /* tcp header within combined hdr */
400 int32_t tcp_tcp_hdr_len; /* tcp header len within combined */
401
402 uint32_t tcp_sum; /* checksum to compensate for source */
403 /* routed packets. Host byte order */
404 uint16_t tcp_last_sent_len; /* Record length for nagle */
405 uint16_t tcp_dupack_cnt; /* # of consequtive duplicate acks */
406
407 kmutex_t *tcp_acceptor_lockp; /* Ptr to tf_lock */
408
409 timeout_id_t tcp_ordrelid; /* qbufcall/qtimeout id */
410 t_uscalar_t tcp_acceptor_id; /* ACCEPTOR_id */
411
412 int tcp_ipsec_overhead;
413 /*
414 * Address family that app wishes returned addrsses to be in.
415 * Currently taken from address family used in T_BIND_REQ, but
416 * should really come from family used in original socket() call.
417 * Value can be AF_INET or AF_INET6.
418 */
419 uint_t tcp_family;
420 /*
421 * used for a quick test to determine if any ancillary bits are
422 * set
423 */
424 uint_t tcp_ipv6_recvancillary; /* Flags */
425 #define TCP_IPV6_RECVPKTINFO 0x01 /* IPV6_RECVPKTINFO option */
426 #define TCP_IPV6_RECVHOPLIMIT 0x02 /* IPV6_RECVHOPLIMIT option */
427 #define TCP_IPV6_RECVHOPOPTS 0x04 /* IPV6_RECVHOPOPTS option */
428 #define TCP_IPV6_RECVDSTOPTS 0x08 /* IPV6_RECVDSTOPTS option */
429 #define TCP_IPV6_RECVRTHDR 0x10 /* IPV6_RECVRTHDR option */
430 #define TCP_IPV6_RECVRTDSTOPTS 0x20 /* IPV6_RECVRTHDRDSTOPTS option */
431 #define TCP_IPV6_RECVTCLASS 0x40 /* IPV6_RECVTCLASS option */
432 #define TCP_OLD_IPV6_RECVDSTOPTS 0x80 /* old IPV6_RECVDSTOPTS option */
433
434 uint_t tcp_recvifindex; /* Last received IPV6_RCVPKTINFO */
435 uint_t tcp_recvhops; /* Last received IPV6_RECVHOPLIMIT */
436 uint_t tcp_recvtclass; /* Last received IPV6_RECVTCLASS */
437 ip6_hbh_t *tcp_hopopts; /* Last received IPV6_RECVHOPOPTS */
438 ip6_dest_t *tcp_dstopts; /* Last received IPV6_RECVDSTOPTS */
439 ip6_dest_t *tcp_rtdstopts; /* Last recvd IPV6_RECVRTHDRDSTOPTS */
440 ip6_rthdr_t *tcp_rthdr; /* Last received IPV6_RECVRTHDR */
441 uint_t tcp_hopoptslen;
442 uint_t tcp_dstoptslen;
443 uint_t tcp_rtdstoptslen;
444 uint_t tcp_rthdrlen;
445
446 mblk_t *tcp_timercache;
447 cred_t *tcp_cred; /* Credentials when this was opened */
448 pid_t tcp_cpid; /* Process id when this was opened */
449 uint64_t tcp_open_time; /* time when this was opened */
450
451
452 union {
453 struct {
454 uchar_t v4_ttl;
455 /* Dup of tcp_ipha.iph_type_of_service */
456 uchar_t v4_tos; /* Dup of tcp_ipha.iph_ttl */
457 } v4_hdr_info;
458 struct {
459 uint_t v6_vcf; /* Dup of tcp_ip6h.ip6h_vcf */
460 uchar_t v6_hops; /* Dup of tcp_ip6h.ip6h_hops */
461 } v6_hdr_info;
462 } tcp_hdr_info;
463 #define tcp_ttl tcp_hdr_info.v4_hdr_info.v4_ttl
464 #define tcp_tos tcp_hdr_info.v4_hdr_info.v4_tos
465 #define tcp_ip6_vcf tcp_hdr_info.v6_hdr_info.v6_vcf
466 #define tcp_ip6_hops tcp_hdr_info.v6_hdr_info.v6_hops
467
468 ushort_t tcp_ipversion;
469 uint_t tcp_bound_if; /* IPV6_BOUND_IF */
470
471 #define tcp_lport tcp_connp->conn_lport
472 #define tcp_fport tcp_connp->conn_fport
473 #define tcp_ports tcp_connp->conn_ports
474
475 #define tcp_remote tcp_connp->conn_rem
476 #define tcp_ip_src tcp_connp->conn_src
477
478 #define tcp_remote_v6 tcp_connp->conn_remv6
479 #define tcp_ip_src_v6 tcp_connp->conn_srcv6
480 #define tcp_bound_source_v6 tcp_connp->conn_bound_source_v6
481 #define tcp_bound_source tcp_connp->conn_bound_source
482
483 kmutex_t tcp_closelock;
484 kcondvar_t tcp_closecv;
485 uint8_t tcp_closed;
486 uint8_t tcp_closeflags;
487 uint8_t tcp_cleandeathtag;
488 mblk_t tcp_closemp;
489 timeout_id_t tcp_linger_tid; /* Linger timer ID */
490 void *tcp_tracebuf;
491
492 struct tcp_s *tcp_acceptor_hash; /* Acceptor hash chain */
493 struct tcp_s **tcp_ptpahn; /* Pointer to previous accept hash next. */
494 struct tcp_s *tcp_bind_hash; /* Bind hash chain */
495 struct tcp_s **tcp_ptpbhn;
496
497 boolean_t tcp_ire_ill_check_done;
498 uint_t tcp_maxpsz;
499
500 /*
501 * used for Multidata Transmit
502 */
503 uint_t tcp_mdt_hdr_head; /* leading header fragment extra space */
504 uint_t tcp_mdt_hdr_tail; /* trailing header fragment extra space */
505 int tcp_mdt_max_pld; /* maximum payload buffers per Multidata */
506
507 uint32_t tcp_lso_max; /* maximum LSO payload */
508
509 uint32_t tcp_ofo_fin_seq; /* Recv out of order FIN seq num */
510 uint32_t tcp_cwr_snd_max;
511 uint_t tcp_drop_opt_ack_cnt; /* # tcp generated optmgmt */
512 ip6_pkt_t tcp_sticky_ipp; /* Sticky options */
513 #define tcp_ipp_fields tcp_sticky_ipp.ipp_fields /* valid fields */
514 #define tcp_ipp_ifindex tcp_sticky_ipp.ipp_ifindex /* pktinfo ifindex */
515 #define tcp_ipp_addr tcp_sticky_ipp.ipp_addr /* pktinfo src/dst addr */
516 #define tcp_ipp_hoplimit tcp_sticky_ipp.ipp_hoplimit
517 #define tcp_ipp_hopoptslen tcp_sticky_ipp.ipp_hopoptslen
518 #define tcp_ipp_rtdstoptslen tcp_sticky_ipp.ipp_rtdstoptslen
519 #define tcp_ipp_rthdrlen tcp_sticky_ipp.ipp_rthdrlen
520 #define tcp_ipp_dstoptslen tcp_sticky_ipp.ipp_dstoptslen
521 #define tcp_ipp_hopopts tcp_sticky_ipp.ipp_hopopts
522 #define tcp_ipp_rtdstopts tcp_sticky_ipp.ipp_rtdstopts
523 #define tcp_ipp_rthdr tcp_sticky_ipp.ipp_rthdr
524 #define tcp_ipp_dstopts tcp_sticky_ipp.ipp_dstopts
525 #define tcp_ipp_nexthop tcp_sticky_ipp.ipp_nexthop
526 #define tcp_ipp_use_min_mtu tcp_sticky_ipp.ipp_use_min_mtu
527 struct tcp_s *tcp_saved_listener; /* saved value of listener */
528
529 uint32_t tcp_in_ack_unsent; /* ACK for unsent data cnt. */
530
531 /*
532 * The following fusion-related fields are protected by squeue.
533 */
534 struct tcp_s *tcp_loopback_peer; /* peer tcp for loopback */
535 mblk_t *tcp_fused_sigurg_mp; /* M_PCSIG mblk for SIGURG */
536 size_t tcp_fuse_rcv_hiwater; /* fusion receive queue size */
537 uint_t tcp_fuse_rcv_unread_hiwater; /* max # of outstanding pkts */
538 /*
539 * The following fusion-related fields and bit fields are to be
540 * manipulated with squeue protection or with tcp_non_sq_lock held.
541 * tcp_non_sq_lock is used to protect fields that may be modified
542 * accessed outside the squeue.
543 */
544 kmutex_t tcp_non_sq_lock;
545 kcondvar_t tcp_fuse_plugcv;
546 uint_t tcp_fuse_rcv_unread_cnt; /* # of outstanding pkts */
547 uint32_t
548 tcp_fused : 1, /* loopback tcp in fusion mode */
549 tcp_unfusable : 1, /* fusion not allowed on endpoint */
550 tcp_fused_sigurg : 1, /* send SIGURG upon draining */
551 tcp_direct_sockfs : 1, /* direct calls to sockfs */
552
553 tcp_fuse_syncstr_stopped : 1, /* synchronous streams stopped */
554 tcp_fuse_syncstr_plugged : 1, /* synchronous streams plugged */
555 tcp_fuse_to_bit_31 : 26;
556
557 /*
558 * This variable is accessed without any lock protection
559 * and therefore must not be declared as a bit field along
560 * with the rest which require such condition.
561 */
562 boolean_t tcp_issocket; /* this is a socket tcp */
563
564 /* protected by the tcp_non_sq_lock lock */
565 uint32_t tcp_squeue_bytes;
566 /*
567 * Kernel SSL session information
568 */
569 boolean_t tcp_kssl_pending; /* waiting for 1st SSL rec. */
570 boolean_t tcp_kssl_inhandshake; /* during SSL handshake */
571 kssl_ent_t tcp_kssl_ent; /* SSL table entry */
572 kssl_ctx_t tcp_kssl_ctx; /* SSL session */
573 uint_t tcp_label_len; /* length of cached label */
574
575 /*
576 * tcp_closemp_used is protected by listener's tcp_eager_lock
577 * when used for eagers. When used for a tcp in TIME_WAIT state
578 * or in tcp_close(), it is not protected by any lock as we
579 * do not expect any other thread to use it concurrently.
580 * We do allow re-use of tcp_closemp in tcp_time_wait_collector()
581 * and tcp_close() but not concurrently.
582 */
583 boolean_t tcp_closemp_used;
584
585 /*
586 * previous and next eagers in the list of droppable eagers. See
587 * the comments before MAKE_DROPPABLE(). These pointers are
588 * protected by listener's tcp_eager_lock.
589 */
590 struct tcp_s *tcp_eager_prev_drop_q0;
591 struct tcp_s *tcp_eager_next_drop_q0;
592
593 /*
594 * Have we flow controlled xmitter?
595 * This variable can be modified outside the squeue and hence must
596 * not be declared as a bit field along with the rest that are
597 * modified only within the squeue.
598 * protected by the tcp_non_sq_lock lock.
599 */
600 boolean_t tcp_flow_stopped;
601
602 /*
603 * tcp_sodirect is used by tcp on the receive side to push mblk_t(s)
604 * directly to sockfs. Also, to schedule asynchronous copyout directly
605 * to a pending user-land uio buffer.
606 */
607 sodirect_t *tcp_sodirect;
608
609 #ifdef DEBUG
610 pc_t tcmp_stk[15];
611 #endif
612 } tcp_t;
613
614 #ifdef DEBUG
615 #define TCP_DEBUG_GETPCSTACK(buffer, depth) ((void) getpcstack(buffer, \
616 depth))
617 #else
618 #define TCP_DEBUG_GETPCSTACK(buffer, depth)
619 #endif
620
621 /*
622 * Track a reference count on the tcps in order to know when
623 * the tcps_g_q can be removed. As long as there is any
624 * tcp_t, other that the tcps_g_q itself, in the tcp_stack_t we
625 * need to keep tcps_g_q around so that a closing connection can
626 * switch to using tcps_g_q as part of it closing.
627 */
628 #define TCPS_REFHOLD(tcps) { \
629 atomic_add_32(&(tcps)->tcps_refcnt, 1); \
630 ASSERT((tcps)->tcps_refcnt != 0); \
631 DTRACE_PROBE1(tcps__refhold, tcp_stack_t, tcps); \
632 }
633
634 /*
635 * Decrement the reference count on the tcp_stack_t.
636 * In architectures e.g sun4u, where atomic_add_32_nv is just
637 * a cas, we need to maintain the right memory barrier semantics
638 * as that of mutex_exit i.e all the loads and stores should complete
639 * before the cas is executed. membar_exit() does that here.
640 */
641 #define TCPS_REFRELE(tcps) { \
642 ASSERT((tcps)->tcps_refcnt != 0); \
643 membar_exit(); \
644 DTRACE_PROBE1(tcps__refrele, tcp_stack_t, tcps); \
645 if (atomic_add_32_nv(&(tcps)->tcps_refcnt, -1) == 0 && \
646 (tcps)->tcps_g_q != NULL) { \
647 /* Only tcps_g_q left */ \
648 tcp_g_q_inactive(tcps); \
649 } \
650 }
651
652 extern void tcp_free(tcp_t *tcp);
653 extern void tcp_ddi_g_init(void);
654 extern void tcp_ddi_g_destroy(void);
655 extern void tcp_g_q_inactive(tcp_stack_t *);
656 extern void tcp_xmit_listeners_reset(mblk_t *mp, uint_t ip_hdr_len,
657 zoneid_t zoneid, tcp_stack_t *, conn_t *connp);
658 extern void tcp_conn_request(void *arg, mblk_t *mp, void *arg2);
659 extern void tcp_conn_request_unbound(void *arg, mblk_t *mp, void *arg2);
660 extern void tcp_input(void *arg, mblk_t *mp, void *arg2);
661 extern void tcp_rput_data(void *arg, mblk_t *mp, void *arg2);
662 extern void *tcp_get_conn(void *arg, tcp_stack_t *);
663 extern void tcp_time_wait_collector(void *arg);
664 extern mblk_t *tcp_snmp_get(queue_t *, mblk_t *);
665 extern int tcp_snmp_set(queue_t *, int, int, uchar_t *, int len);
666 extern mblk_t *tcp_xmit_mp(tcp_t *tcp, mblk_t *mp, int32_t max_to_send,
667 int32_t *offset, mblk_t **end_mp, uint32_t seq,
668 boolean_t sendall, uint32_t *seg_len, boolean_t rexmit);
669 /*
670 * The TCP Fanout structure.
671 * The hash tables and their linkage (tcp_*_hash_next, tcp_ptp*hn) are
672 * protected by the per-bucket tf_lock. Each tcp_t
673 * inserted in the list points back at this lock using tcp_*_lockp.
674 *
675 * The listener and acceptor hash queues are lists of tcp_t.
676 */
677 /* listener hash and acceptor hash queue head */
678 typedef struct tf_s {
679 tcp_t *tf_tcp;
680 kmutex_t tf_lock;
681 } tf_t;
682 #endif /* (defined(_KERNEL) || defined(_KMEMUSER)) */
683
684 /* Contract private interface between TCP and Clustering. */
685
686 #define CL_TCPI_V1 1 /* cl_tcpi_version number */
687
688 typedef struct cl_tcp_info_s {
689 ushort_t cl_tcpi_version; /* cl_tcp_info_t's version no */
690 ushort_t cl_tcpi_ipversion; /* IP version */
691 int32_t cl_tcpi_state; /* TCP state */
692 in_port_t cl_tcpi_lport; /* Local port */
693 in_port_t cl_tcpi_fport; /* Remote port */
694 in6_addr_t cl_tcpi_laddr_v6; /* Local IP address */
695 in6_addr_t cl_tcpi_faddr_v6; /* Remote IP address */
696 #ifdef _KERNEL
697 /* Note: V4_PART_OF_V6 is meant to be used only for _KERNEL defined stuff */
698 #define cl_tcpi_laddr V4_PART_OF_V6(cl_tcpi_laddr_v6)
699 #define cl_tcpi_faddr V4_PART_OF_V6(cl_tcpi_faddr_v6)
700
701 #endif /* _KERNEL */
702 } cl_tcp_info_t;
703
704 /*
705 * Hook functions to enable cluster networking
706 * On non-clustered systems these vectors must always be NULL.
707 */
708
709 extern void (*cl_inet_listen)(uint8_t, sa_family_t, uint8_t *, in_port_t);
710 extern void (*cl_inet_unlisten)(uint8_t, sa_family_t, uint8_t *,
711 in_port_t);
712
713 /*
714 * Contracted Consolidation Private ioctl for aborting TCP connections.
715 * In order to keep the offsets and size of the structure the same between
716 * a 32-bit application and a 64-bit amd64 kernel, we use a #pragma
717 * pack(4).
718 */
719 #define TCP_IOC_ABORT_CONN (('T' << 8) + 91)
720
721 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
722 #pragma pack(4)
723 #endif
724
725 typedef struct tcp_ioc_abort_conn_s {
726 struct sockaddr_storage ac_local; /* local addr and port */
727 struct sockaddr_storage ac_remote; /* remote addr and port */
728 int32_t ac_start; /* start state */
729 int32_t ac_end; /* end state */
730 int32_t ac_zoneid; /* zoneid */
731 } tcp_ioc_abort_conn_t;
732
733 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4
734 #pragma pack()
735 #endif
736
737 #if (defined(_KERNEL) || defined(_KMEMUSER))
738 extern void tcp_rput_other(tcp_t *tcp, mblk_t *mp);
739 #endif
740
741 #ifdef __cplusplus
742 }
743 #endif
744
745 #endif /* _INET_TCP_H */