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 2009 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 #ifdef __cplusplus 31 extern "C" { 32 #endif 33 34 #include <sys/inttypes.h> 35 #include <netinet/ip6.h> 36 #include <netinet/tcp.h> 37 #include <sys/socket.h> 38 #include <sys/socket_proto.h> 39 #include <sys/sodirect.h> 40 #include <sys/multidata.h> 41 #include <sys/md5.h> 42 #include <inet/common.h> 43 #include <inet/ip.h> 44 #include <inet/ip6.h> 45 #include <inet/mi.h> 46 #include <inet/mib2.h> 47 #include <inet/tcp_stack.h> 48 #include <inet/tcp_sack.h> 49 #include <inet/kssl/ksslapi.h> 50 51 /* 52 * Private (and possibly temporary) ioctl used by configuration code 53 * to lock in the "default" stream for detached closes. 54 */ 55 #define TCP_IOC_DEFAULT_Q (('T' << 8) + 51) 56 57 /* TCP states */ 58 #define TCPS_CLOSED -6 59 #define TCPS_IDLE -5 /* idle (opened, but not bound) */ 60 #define TCPS_BOUND -4 /* bound, ready to connect or accept */ 61 #define TCPS_LISTEN -3 /* listening for connection */ 62 #define TCPS_SYN_SENT -2 /* active, have sent syn */ 63 #define TCPS_SYN_RCVD -1 /* have received syn (and sent ours) */ 64 /* states < TCPS_ESTABLISHED are those where connections not established */ 65 #define TCPS_ESTABLISHED 0 /* established */ 66 #define TCPS_CLOSE_WAIT 1 /* rcvd fin, waiting for close */ 67 /* states > TCPS_CLOSE_WAIT are those where user has closed */ 68 #define TCPS_FIN_WAIT_1 2 /* have closed and sent fin */ 69 #define TCPS_CLOSING 3 /* closed, xchd FIN, await FIN ACK */ 70 #define TCPS_LAST_ACK 4 /* had fin and close; await FIN ACK */ 71 /* states > TCPS_CLOSE_WAIT && < TCPS_FIN_WAIT_2 await ACK of FIN */ 72 #define TCPS_FIN_WAIT_2 5 /* have closed, fin is acked */ 73 #define TCPS_TIME_WAIT 6 /* in 2*msl quiet wait after close */ 74 75 /* 76 * Internal flags used in conjunction with the packet header flags. 77 * Used in tcp_rput_data to keep track of what needs to be done. 78 */ 79 #define TH_LIMIT_XMIT 0x0400 /* Limited xmit is needed */ 80 #define TH_XMIT_NEEDED 0x0800 /* Window opened - send queued data */ 81 #define TH_REXMIT_NEEDED 0x1000 /* Time expired for unacked data */ 82 #define TH_ACK_NEEDED 0x2000 /* Send an ack now. */ 83 #define TH_NEED_SACK_REXMIT 0x4000 /* Use SACK info to retransmission */ 84 #define TH_ACK_TIMER_NEEDED 0x8000 /* Start the delayed ACK timer */ 85 #define TH_ORDREL_NEEDED 0x10000 /* Generate an ordrel indication */ 86 #define TH_MARKNEXT_NEEDED 0x20000 /* Data should have MSGMARKNEXT */ 87 #define TH_SEND_URP_MARK 0x40000 /* Send up tcp_urp_mark_mp */ 88 89 /* 90 * TCP sequence numbers are 32 bit integers operated 91 * on with modular arithmetic. These macros can be 92 * used to compare such integers. 93 */ 94 #define SEQ_LT(a, b) ((int32_t)((a)-(b)) < 0) 95 #define SEQ_LEQ(a, b) ((int32_t)((a)-(b)) <= 0) 96 #define SEQ_GT(a, b) ((int32_t)((a)-(b)) > 0) 97 #define SEQ_GEQ(a, b) ((int32_t)((a)-(b)) >= 0) 98 99 /* TCP Protocol header */ 100 typedef struct tcphdr_s { 101 uint8_t th_lport[2]; /* Source port */ 102 uint8_t th_fport[2]; /* Destination port */ 103 uint8_t th_seq[4]; /* Sequence number */ 104 uint8_t th_ack[4]; /* Acknowledgement number */ 105 uint8_t th_offset_and_rsrvd[1]; /* Offset to the packet data */ 106 uint8_t th_flags[1]; 107 uint8_t th_win[2]; /* Allocation number */ 108 uint8_t th_sum[2]; /* TCP checksum */ 109 uint8_t th_urp[2]; /* Urgent pointer */ 110 } tcph_t; 111 112 #define TCP_HDR_LENGTH(tcph) (((tcph)->th_offset_and_rsrvd[0] >>2) &(0xF << 2)) 113 #define TCP_MAX_COMBINED_HEADER_LENGTH (60 + 60) /* Maxed out ip + tcp */ 114 #define TCP_MAX_IP_OPTIONS_LENGTH (60 - IP_SIMPLE_HDR_LENGTH) 115 #define TCP_MAX_HDR_LENGTH 60 116 #define TCP_MAX_TCP_OPTIONS_LENGTH (60 - sizeof (tcph_t)) 117 #define TCP_MIN_HEADER_LENGTH 20 118 #define TCP_MAXWIN 65535 119 #define TCP_PORT_LEN sizeof (in_port_t) 120 #define TCP_MAX_WINSHIFT 14 121 #define TCP_MAX_LARGEWIN (TCP_MAXWIN << TCP_MAX_WINSHIFT) 122 #define TCP_MAX_LSO_LENGTH (IP_MAXPACKET - TCP_MAX_COMBINED_HEADER_LENGTH) 123 124 #define TCPIP_HDR_LENGTH(mp, n) \ 125 (n) = IPH_HDR_LENGTH((mp)->b_rptr), \ 126 (n) += TCP_HDR_LENGTH((tcph_t *)&(mp)->b_rptr[(n)]) 127 128 /* TCP Protocol header (used if the header is known to be 32-bit aligned) */ 129 typedef struct tcphdra_s { 130 in_port_t tha_lport; /* Source port */ 131 in_port_t tha_fport; /* Destination port */ 132 uint32_t tha_seq; /* Sequence number */ 133 uint32_t tha_ack; /* Acknowledgement number */ 134 uint8_t tha_offset_and_reserved; /* Offset to the packet data */ 135 uint8_t tha_flags; 136 uint16_t tha_win; /* Allocation number */ 137 uint16_t tha_sum; /* TCP checksum */ 138 uint16_t tha_urp; /* Urgent pointer */ 139 } tcpha_t; 140 141 struct conn_s; 142 143 /* 144 * Control structure for each open TCP stream, 145 * defined only within the kernel or for a kmem user. 146 * NOTE: tcp_reinit_values MUST have a line for each field in this structure! 147 */ 148 #if (defined(_KERNEL) || defined(_KMEMUSER)) 149 150 typedef struct tcp_s { 151 /* Pointer to previous bind hash next. */ 152 struct tcp_s *tcp_time_wait_next; 153 /* Pointer to next T/W block */ 154 struct tcp_s *tcp_time_wait_prev; 155 /* Pointer to previous T/W next */ 156 clock_t tcp_time_wait_expire; 157 158 struct conn_s *tcp_connp; 159 tcp_stack_t *tcp_tcps; /* Shortcut via conn_netstack */ 160 161 int32_t tcp_state; 162 int32_t tcp_rcv_ws; /* My window scale power */ 163 int32_t tcp_snd_ws; /* Sender's window scale power */ 164 uint32_t tcp_ts_recent; /* Timestamp of earliest unacked */ 165 /* data segment */ 166 clock_t tcp_rto; /* Round trip timeout */ 167 clock_t tcp_last_rcv_lbolt; 168 /* lbolt on last packet, used for PAWS */ 169 170 uint32_t tcp_snxt; /* Senders next seq num */ 171 uint32_t tcp_swnd; /* Senders window (relative to suna) */ 172 uint32_t tcp_mss; /* Max segment size */ 173 uint32_t tcp_iss; /* Initial send seq num */ 174 uint32_t tcp_rnxt; /* Seq we expect to recv next */ 175 uint32_t tcp_rwnd; 176 177 queue_t *tcp_rq; /* Our upstream neighbor (client) */ 178 queue_t *tcp_wq; /* Our downstream neighbor */ 179 180 /* Fields arranged in approximate access order along main paths */ 181 mblk_t *tcp_xmit_head; /* Head of rexmit list */ 182 mblk_t *tcp_xmit_last; /* last valid data seen by tcp_wput */ 183 mblk_t *tcp_xmit_tail; /* Last rexmit data sent */ 184 uint32_t tcp_unsent; /* # of bytes in hand that are unsent */ 185 uint32_t tcp_xmit_tail_unsent; /* # of unsent bytes in xmit_tail */ 186 187 uint32_t tcp_suna; /* Sender unacknowledged */ 188 uint32_t tcp_rexmit_nxt; /* Next rexmit seq num */ 189 uint32_t tcp_rexmit_max; /* Max retran seq num */ 190 int32_t tcp_snd_burst; /* Send burst factor */ 191 uint32_t tcp_cwnd; /* Congestion window */ 192 int32_t tcp_cwnd_cnt; /* cwnd cnt in congestion avoidance */ 193 194 uint32_t tcp_ibsegs; /* Inbound segments on this stream */ 195 uint32_t tcp_obsegs; /* Outbound segments on this stream */ 196 197 uint32_t tcp_naglim; /* Tunable nagle limit */ 198 uint32_t tcp_valid_bits; 199 #define TCP_ISS_VALID 0x1 /* Is the tcp_iss seq num active? */ 200 #define TCP_FSS_VALID 0x2 /* Is the tcp_fss seq num active? */ 201 #define TCP_URG_VALID 0x4 /* Is the tcp_urg seq num active? */ 202 #define TCP_OFO_FIN_VALID 0x8 /* Has TCP received an out of order FIN? */ 203 204 205 206 timeout_id_t tcp_timer_tid; /* Control block for timer service */ 207 uchar_t tcp_timer_backoff; /* Backoff shift count. */ 208 int64_t tcp_last_recv_time; /* Last time we receive a segment. */ 209 uint32_t tcp_init_cwnd; /* Initial cwnd (start/restart) */ 210 211 /* 212 * Following socket options are set by sockfs outside the squeue 213 * and we want to separate these bit fields from the other bit fields 214 * set by TCP to avoid grabbing locks. sockfs ensures that only one 215 * thread in sockfs can set a socket option at a time on a conn_t. 216 * However TCP may read these options concurrently. The linger option 217 * needs atomicity since tcp_lingertime also needs to be in sync. 218 * However TCP uses it only during close, and by then no socket option 219 * can come down. So we don't need any locks, instead just separating 220 * the sockfs settable bit fields from the other bit fields is 221 * sufficient. 222 */ 223 uint32_t 224 tcp_debug : 1, /* SO_DEBUG "socket" option. */ 225 tcp_dontroute : 1, /* SO_DONTROUTE "socket" option. */ 226 tcp_broadcast : 1, /* SO_BROADCAST "socket" option. */ 227 tcp_useloopback : 1, /* SO_USELOOPBACK "socket" option. */ 228 229 tcp_oobinline : 1, /* SO_OOBINLINE "socket" option. */ 230 tcp_dgram_errind : 1, /* SO_DGRAM_ERRIND option */ 231 tcp_linger : 1, /* SO_LINGER turned on */ 232 tcp_reuseaddr : 1, /* SO_REUSEADDR "socket" option. */ 233 234 tcp_junk_to_bit_31 : 24; 235 236 /* Following manipulated by TCP under squeue protection */ 237 uint32_t 238 tcp_urp_last_valid : 1, /* Is tcp_urp_last valid? */ 239 tcp_hard_binding : 1, /* If we've started a full bind */ 240 tcp_hard_bound : 1, /* If we've done a full bind with IP */ 241 tcp_fin_acked : 1, /* Has our FIN been acked? */ 242 243 tcp_fin_rcvd : 1, /* Have we seen a FIN? */ 244 tcp_fin_sent : 1, /* Have we sent our FIN yet? */ 245 tcp_ordrel_done : 1, /* Have we sent the ord_rel upstream? */ 246 tcp_detached : 1, /* If we're detached from a stream */ 247 248 tcp_bind_pending : 1, /* Client is waiting for bind ack */ 249 tcp_unbind_pending : 1, /* Client sent T_UNBIND_REQ */ 250 tcp_ka_enabled: 1, /* Connection KeepAlive Timer needed */ 251 tcp_zero_win_probe: 1, /* Zero win probing is in progress */ 252 253 tcp_loopback: 1, /* src and dst are the same machine */ 254 tcp_localnet: 1, /* src and dst are on the same subnet */ 255 tcp_syn_defense: 1, /* For defense against SYN attack */ 256 #define tcp_dontdrop tcp_syn_defense 257 tcp_set_timer : 1, 258 259 tcp_active_open: 1, /* This is a active open */ 260 tcp_rexmit : 1, /* TCP is retransmitting */ 261 tcp_snd_sack_ok : 1, /* Can use SACK for this connection */ 262 tcp_empty_flag : 1, /* Empty flag for future use */ 263 264 tcp_recvdstaddr : 1, /* return T_EXTCONN_IND with dst addr */ 265 tcp_hwcksum : 1, /* The NIC is capable of hwcksum */ 266 tcp_ip_forward_progress : 1, 267 tcp_anon_priv_bind : 1, 268 269 tcp_ecn_ok : 1, /* Can use ECN for this connection */ 270 tcp_ecn_echo_on : 1, /* Need to do ECN echo */ 271 tcp_ecn_cwr_sent : 1, /* ECN_CWR has been sent */ 272 tcp_cwr : 1, /* Cwnd has reduced recently */ 273 274 tcp_pad_to_bit31 : 4; 275 /* Following manipulated by TCP under squeue protection */ 276 uint32_t 277 tcp_mdt : 1, /* Lower layer is capable of MDT */ 278 tcp_snd_ts_ok : 1, 279 tcp_snd_ws_ok : 1, 280 tcp_exclbind : 1, /* ``exclusive'' binding */ 281 282 tcp_hdr_grown : 1, 283 tcp_in_free_list : 1, 284 tcp_snd_zcopy_on : 1, /* xmit zero-copy enabled */ 285 286 tcp_snd_zcopy_aware : 1, /* client is zero-copy aware */ 287 tcp_xmit_zc_clean : 1, /* the xmit list is free of zc-mblk */ 288 tcp_wait_for_eagers : 1, /* Wait for eagers to disappear */ 289 tcp_accept_error : 1, /* Error during TLI accept */ 290 291 tcp_send_discon_ind : 1, /* TLI accept err, send discon ind */ 292 tcp_cork : 1, /* tcp_cork option */ 293 tcp_tconnind_started : 1, /* conn_ind message is being sent */ 294 tcp_lso :1, /* Lower layer is capable of LSO */ 295 tcp_refuse :1, /* Connection needs refusing */ 296 tcp_pad_to_bit_31 : 16; 297 298 uint32_t tcp_if_mtu; /* Outgoing interface MTU. */ 299 300 mblk_t *tcp_reass_head; /* Out of order reassembly list head */ 301 mblk_t *tcp_reass_tail; /* Out of order reassembly list tail */ 302 303 tcp_sack_info_t *tcp_sack_info; 304 305 #define tcp_pipe tcp_sack_info->tcp_pipe 306 #define tcp_fack tcp_sack_info->tcp_fack 307 #define tcp_sack_snxt tcp_sack_info->tcp_sack_snxt 308 #define tcp_max_sack_blk tcp_sack_info->tcp_max_sack_blk 309 #define tcp_num_sack_blk tcp_sack_info->tcp_num_sack_blk 310 #define tcp_sack_list tcp_sack_info->tcp_sack_list 311 #define tcp_num_notsack_blk tcp_sack_info->tcp_num_notsack_blk 312 #define tcp_cnt_notsack_list tcp_sack_info->tcp_cnt_notsack_list 313 #define tcp_notsack_list tcp_sack_info->tcp_notsack_list 314 315 mblk_t *tcp_rcv_list; /* Queued until push, urgent data, */ 316 mblk_t *tcp_rcv_last_head; /* optdata, or the count exceeds */ 317 mblk_t *tcp_rcv_last_tail; /* tcp_rcv_push_wait. */ 318 uint32_t tcp_rcv_cnt; /* tcp_rcv_list is b_next chain. */ 319 320 uint32_t tcp_cwnd_ssthresh; /* Congestion window */ 321 uint32_t tcp_cwnd_max; 322 uint32_t tcp_csuna; /* Clear (no rexmits in window) suna */ 323 324 clock_t tcp_rtt_sa; /* Round trip smoothed average */ 325 clock_t tcp_rtt_sd; /* Round trip smoothed deviation */ 326 clock_t tcp_rtt_update; /* Round trip update(s) */ 327 clock_t tcp_ms_we_have_waited; /* Total retrans time */ 328 329 uint32_t tcp_swl1; /* These help us avoid using stale */ 330 uint32_t tcp_swl2; /* packets to update state */ 331 332 uint32_t tcp_rack; /* Seq # we have acked */ 333 uint32_t tcp_rack_cnt; /* # of segs we have deferred ack */ 334 uint32_t tcp_rack_cur_max; /* # of segs we may defer ack for now */ 335 uint32_t tcp_rack_abs_max; /* # of segs we may defer ack ever */ 336 timeout_id_t tcp_ack_tid; /* Delayed ACK timer ID */ 337 timeout_id_t tcp_push_tid; /* Push timer ID */ 338 339 uint32_t tcp_max_swnd; /* Maximum swnd we have seen */ 340 341 struct tcp_s *tcp_listener; /* Our listener */ 342 343 size_t tcp_xmit_hiwater; /* Send buffer high water mark. */ 344 size_t tcp_xmit_lowater; /* Send buffer low water mark. */ 345 size_t tcp_recv_hiwater; /* Recv high water mark */ 346 size_t tcp_recv_lowater; /* Recv low water mark */ 347 348 uint32_t tcp_irs; /* Initial recv seq num */ 349 uint32_t tcp_fss; /* Final/fin send seq num */ 350 uint32_t tcp_urg; /* Urgent data seq num */ 351 352 clock_t tcp_first_timer_threshold; /* When to prod IP */ 353 clock_t tcp_second_timer_threshold; /* When to give up completely */ 354 clock_t tcp_first_ctimer_threshold; /* 1st threshold while connecting */ 355 clock_t tcp_second_ctimer_threshold; /* 2nd ... while connecting */ 356 357 int tcp_lingertime; /* Close linger time (in seconds) */ 358 359 uint32_t tcp_urp_last; /* Last urp for which signal sent */ 360 mblk_t *tcp_urp_mp; /* T_EXDATA_IND for urgent byte */ 361 mblk_t *tcp_urp_mark_mp; /* zero-length marked/unmarked msg */ 362 363 int tcp_conn_req_cnt_q0; /* # of conn reqs in SYN_RCVD */ 364 int tcp_conn_req_cnt_q; /* # of conn reqs in ESTABLISHED */ 365 int tcp_conn_req_max; /* # of ESTABLISHED conn reqs allowed */ 366 t_scalar_t tcp_conn_req_seqnum; /* Incrementing pending conn req ID */ 367 #define tcp_ip_addr_cache tcp_reass_tail 368 /* Cache ip addresses that */ 369 /* complete the 3-way handshake */ 370 kmutex_t tcp_eager_lock; 371 struct tcp_s *tcp_eager_next_q; /* next eager in ESTABLISHED state */ 372 struct tcp_s *tcp_eager_last_q; /* last eager in ESTABLISHED state */ 373 struct tcp_s *tcp_eager_next_q0; /* next eager in SYN_RCVD state */ 374 struct tcp_s *tcp_eager_prev_q0; /* prev eager in SYN_RCVD state */ 375 /* all eagers form a circular list */ 376 boolean_t tcp_conn_def_q0; /* move from q0 to q deferred */ 377 378 union { 379 mblk_t *tcp_eager_conn_ind; /* T_CONN_IND waiting for 3rd ack. */ 380 mblk_t *tcp_opts_conn_req; /* T_CONN_REQ w/ options processed */ 381 } tcp_conn; 382 uint32_t tcp_syn_rcvd_timeout; /* How many SYN_RCVD timeout in q0 */ 383 384 /* TCP Keepalive Timer members */ 385 int32_t tcp_ka_last_intrvl; /* Last probe interval */ 386 timeout_id_t tcp_ka_tid; /* Keepalive timer ID */ 387 uint32_t tcp_ka_interval; /* Keepalive interval */ 388 uint32_t tcp_ka_abort_thres; /* Keepalive abort threshold */ 389 390 int32_t tcp_client_errno; /* How the client screwed up */ 391 392 char *tcp_iphc; /* Buffer holding tcp/ip hdr template */ 393 int tcp_iphc_len; /* actual allocated buffer size */ 394 int32_t tcp_hdr_len; /* Byte len of combined TCP/IP hdr */ 395 ipha_t *tcp_ipha; /* IPv4 header in the buffer */ 396 ip6_t *tcp_ip6h; /* IPv6 header in the buffer */ 397 int tcp_ip_hdr_len; /* Byte len of our current IPvx hdr */ 398 tcph_t *tcp_tcph; /* tcp header within combined hdr */ 399 int32_t tcp_tcp_hdr_len; /* tcp header len within combined */ 400 /* Saved peer headers in the case of re-fusion */ 401 ipha_t tcp_saved_ipha; 402 ip6_t tcp_saved_ip6h; 403 tcph_t tcp_saved_tcph; 404 405 uint32_t tcp_sum; /* checksum to compensate for source */ 406 /* routed packets. Host byte order */ 407 uint16_t tcp_last_sent_len; /* Record length for nagle */ 408 uint16_t tcp_dupack_cnt; /* # of consequtive duplicate acks */ 409 410 kmutex_t *tcp_acceptor_lockp; /* Ptr to tf_lock */ 411 412 mblk_t *tcp_ordrel_mp; /* T_ordrel_ind mblk */ 413 t_uscalar_t tcp_acceptor_id; /* ACCEPTOR_id */ 414 415 int tcp_ipsec_overhead; 416 /* 417 * Address family that app wishes returned addrsses to be in. 418 * Currently taken from address family used in T_BIND_REQ, but 419 * should really come from family used in original socket() call. 420 * Value can be AF_INET or AF_INET6. 421 */ 422 uint_t tcp_family; 423 /* 424 * used for a quick test to determine if any ancillary bits are 425 * set 426 */ 427 uint_t tcp_ipv6_recvancillary; /* Flags */ 428 #define TCP_IPV6_RECVPKTINFO 0x01 /* IPV6_RECVPKTINFO option */ 429 #define TCP_IPV6_RECVHOPLIMIT 0x02 /* IPV6_RECVHOPLIMIT option */ 430 #define TCP_IPV6_RECVHOPOPTS 0x04 /* IPV6_RECVHOPOPTS option */ 431 #define TCP_IPV6_RECVDSTOPTS 0x08 /* IPV6_RECVDSTOPTS option */ 432 #define TCP_IPV6_RECVRTHDR 0x10 /* IPV6_RECVRTHDR option */ 433 #define TCP_IPV6_RECVRTDSTOPTS 0x20 /* IPV6_RECVRTHDRDSTOPTS option */ 434 #define TCP_IPV6_RECVTCLASS 0x40 /* IPV6_RECVTCLASS option */ 435 #define TCP_OLD_IPV6_RECVDSTOPTS 0x80 /* old IPV6_RECVDSTOPTS option */ 436 437 uint_t tcp_recvifindex; /* Last received IPV6_RCVPKTINFO */ 438 uint_t tcp_recvhops; /* Last received IPV6_RECVHOPLIMIT */ 439 uint_t tcp_recvtclass; /* Last received IPV6_RECVTCLASS */ 440 ip6_hbh_t *tcp_hopopts; /* Last received IPV6_RECVHOPOPTS */ 441 ip6_dest_t *tcp_dstopts; /* Last received IPV6_RECVDSTOPTS */ 442 ip6_dest_t *tcp_rtdstopts; /* Last recvd IPV6_RECVRTHDRDSTOPTS */ 443 ip6_rthdr_t *tcp_rthdr; /* Last received IPV6_RECVRTHDR */ 444 uint_t tcp_hopoptslen; 445 uint_t tcp_dstoptslen; 446 uint_t tcp_rtdstoptslen; 447 uint_t tcp_rthdrlen; 448 449 mblk_t *tcp_timercache; 450 cred_t *tcp_cred; /* Credentials when this was opened */ 451 pid_t tcp_cpid; /* Process id when this was opened */ 452 uint64_t tcp_open_time; /* time when this was opened */ 453 454 455 union { 456 struct { 457 uchar_t v4_ttl; 458 /* Dup of tcp_ipha.iph_type_of_service */ 459 uchar_t v4_tos; /* Dup of tcp_ipha.iph_ttl */ 460 } v4_hdr_info; 461 struct { 462 uint_t v6_vcf; /* Dup of tcp_ip6h.ip6h_vcf */ 463 uchar_t v6_hops; /* Dup of tcp_ip6h.ip6h_hops */ 464 } v6_hdr_info; 465 } tcp_hdr_info; 466 #define tcp_ttl tcp_hdr_info.v4_hdr_info.v4_ttl 467 #define tcp_tos tcp_hdr_info.v4_hdr_info.v4_tos 468 #define tcp_ip6_vcf tcp_hdr_info.v6_hdr_info.v6_vcf 469 #define tcp_ip6_hops tcp_hdr_info.v6_hdr_info.v6_hops 470 471 ushort_t tcp_ipversion; 472 uint_t tcp_bound_if; /* IPV6_BOUND_IF */ 473 474 #define tcp_lport tcp_connp->conn_lport 475 #define tcp_fport tcp_connp->conn_fport 476 #define tcp_ports tcp_connp->conn_ports 477 478 #define tcp_remote tcp_connp->conn_rem 479 #define tcp_ip_src tcp_connp->conn_src 480 481 #define tcp_remote_v6 tcp_connp->conn_remv6 482 #define tcp_ip_src_v6 tcp_connp->conn_srcv6 483 #define tcp_bound_source_v6 tcp_connp->conn_bound_source_v6 484 #define tcp_bound_source tcp_connp->conn_bound_source 485 486 kmutex_t tcp_closelock; 487 kcondvar_t tcp_closecv; 488 uint8_t tcp_closed; 489 uint8_t tcp_closeflags; 490 uint8_t tcp_cleandeathtag; 491 mblk_t tcp_closemp; 492 timeout_id_t tcp_linger_tid; /* Linger timer ID */ 493 494 struct tcp_s *tcp_acceptor_hash; /* Acceptor hash chain */ 495 struct tcp_s **tcp_ptpahn; /* Pointer to previous accept hash next. */ 496 struct tcp_s *tcp_bind_hash; /* Bind hash chain */ 497 struct tcp_s *tcp_bind_hash_port; /* tcp_t's bound to the same lport */ 498 struct tcp_s **tcp_ptpbhn; 499 500 boolean_t tcp_ire_ill_check_done; 501 uint_t tcp_maxpsz; 502 503 /* 504 * used for Multidata Transmit 505 */ 506 uint_t tcp_mdt_hdr_head; /* leading header fragment extra space */ 507 uint_t tcp_mdt_hdr_tail; /* trailing header fragment extra space */ 508 int tcp_mdt_max_pld; /* maximum payload buffers per Multidata */ 509 510 uint32_t tcp_lso_max; /* maximum LSO payload */ 511 512 uint32_t tcp_ofo_fin_seq; /* Recv out of order FIN seq num */ 513 uint32_t tcp_cwr_snd_max; 514 uint_t tcp_drop_opt_ack_cnt; /* # tcp generated optmgmt */ 515 ip6_pkt_t tcp_sticky_ipp; /* Sticky options */ 516 #define tcp_ipp_fields tcp_sticky_ipp.ipp_fields /* valid fields */ 517 #define tcp_ipp_ifindex tcp_sticky_ipp.ipp_ifindex /* pktinfo ifindex */ 518 #define tcp_ipp_addr tcp_sticky_ipp.ipp_addr /* pktinfo src/dst addr */ 519 #define tcp_ipp_hoplimit tcp_sticky_ipp.ipp_hoplimit 520 #define tcp_ipp_hopoptslen tcp_sticky_ipp.ipp_hopoptslen 521 #define tcp_ipp_rtdstoptslen tcp_sticky_ipp.ipp_rtdstoptslen 522 #define tcp_ipp_rthdrlen tcp_sticky_ipp.ipp_rthdrlen 523 #define tcp_ipp_dstoptslen tcp_sticky_ipp.ipp_dstoptslen 524 #define tcp_ipp_hopopts tcp_sticky_ipp.ipp_hopopts 525 #define tcp_ipp_rtdstopts tcp_sticky_ipp.ipp_rtdstopts 526 #define tcp_ipp_rthdr tcp_sticky_ipp.ipp_rthdr 527 #define tcp_ipp_dstopts tcp_sticky_ipp.ipp_dstopts 528 #define tcp_ipp_nexthop tcp_sticky_ipp.ipp_nexthop 529 #define tcp_ipp_use_min_mtu tcp_sticky_ipp.ipp_use_min_mtu 530 struct tcp_s *tcp_saved_listener; /* saved value of listener */ 531 532 uint32_t tcp_in_ack_unsent; /* ACK for unsent data cnt. */ 533 534 /* 535 * The following fusion-related fields are protected by squeue. 536 */ 537 struct tcp_s *tcp_loopback_peer; /* peer tcp for loopback */ 538 mblk_t *tcp_fused_sigurg_mp; /* M_PCSIG mblk for SIGURG */ 539 size_t tcp_fuse_rcv_hiwater; /* fusion receive queue size */ 540 uint_t tcp_fuse_rcv_unread_hiwater; /* max # of outstanding pkts */ 541 /* 542 * The following fusion-related fields and bit fields are to be 543 * manipulated with squeue protection or with tcp_non_sq_lock held. 544 * tcp_non_sq_lock is used to protect fields that may be modified 545 * accessed outside the squeue. 546 */ 547 kmutex_t tcp_non_sq_lock; 548 kcondvar_t tcp_fuse_plugcv; 549 uint_t tcp_fuse_rcv_unread_cnt; /* # of outstanding pkts */ 550 uint32_t 551 tcp_fused : 1, /* loopback tcp in fusion mode */ 552 tcp_unfusable : 1, /* fusion not allowed on endpoint */ 553 tcp_fused_sigurg : 1, /* send SIGURG upon draining */ 554 tcp_direct_sockfs : 1, /* direct calls to sockfs */ 555 556 tcp_fuse_syncstr_stopped : 1, /* synchronous streams stopped */ 557 tcp_fuse_syncstr_plugged : 1, /* synchronous streams plugged */ 558 tcp_fuse_to_bit_31 : 26; 559 560 /* 561 * This variable is accessed without any lock protection 562 * and therefore must not be declared as a bit field along 563 * with the rest which require such condition. 564 */ 565 boolean_t tcp_issocket; /* this is a socket tcp */ 566 567 /* protected by the tcp_non_sq_lock lock */ 568 uint32_t tcp_squeue_bytes; 569 /* 570 * Kernel SSL session information 571 */ 572 boolean_t tcp_kssl_pending; /* waiting for 1st SSL rec. */ 573 boolean_t tcp_kssl_inhandshake; /* during SSL handshake */ 574 kssl_ent_t tcp_kssl_ent; /* SSL table entry */ 575 kssl_ctx_t tcp_kssl_ctx; /* SSL session */ 576 uint_t tcp_label_len; /* length of cached label */ 577 578 /* 579 * tcp_closemp_used is protected by listener's tcp_eager_lock 580 * when used for eagers. When used for a tcp in TIME_WAIT state 581 * or in tcp_close(), it is not protected by any lock as we 582 * do not expect any other thread to use it concurrently. 583 * We do allow re-use of tcp_closemp in tcp_time_wait_collector() 584 * and tcp_close() but not concurrently. 585 */ 586 boolean_t tcp_closemp_used; 587 588 /* 589 * previous and next eagers in the list of droppable eagers. See 590 * the comments before MAKE_DROPPABLE(). These pointers are 591 * protected by listener's tcp_eager_lock. 592 */ 593 struct tcp_s *tcp_eager_prev_drop_q0; 594 struct tcp_s *tcp_eager_next_drop_q0; 595 596 /* 597 * Have we flow controlled xmitter? 598 * This variable can be modified outside the squeue and hence must 599 * not be declared as a bit field along with the rest that are 600 * modified only within the squeue. 601 * protected by the tcp_non_sq_lock lock. 602 */ 603 boolean_t tcp_flow_stopped; 604 605 /* 606 * The socket generation number is bumped when an outgoing connection 607 * attempts is made, and it sent up to the socket when the 608 * connection was successfully established, or an error occured. The 609 * generation is used to ensure that the socket does not miss the 610 * asynchronous notification. 611 */ 612 sock_connid_t tcp_connid; 613 614 /* 615 * tcp_sodirect is used by tcp on the receive side to push mblk_t(s) 616 * directly to sockfs. Also, to schedule asynchronous copyout directly 617 * to a pending user-land uio buffer. 618 */ 619 sodirect_t *tcp_sodirect; 620 621 /* mblk_t used to enter TCP's squeue from the service routine. */ 622 mblk_t *tcp_rsrv_mp; 623 /* Mutex for accessing tcp_rsrv_mp */ 624 kmutex_t tcp_rsrv_mp_lock; 625 626 #ifdef DEBUG 627 pc_t tcmp_stk[15]; 628 #endif 629 } tcp_t; 630 631 #ifdef DEBUG 632 #define TCP_DEBUG_GETPCSTACK(buffer, depth) ((void) getpcstack(buffer, \ 633 depth)) 634 #else 635 #define TCP_DEBUG_GETPCSTACK(buffer, depth) 636 #endif 637 638 /* 639 * Track a reference count on the tcps in order to know when 640 * the tcps_g_q can be removed. As long as there is any 641 * tcp_t, other that the tcps_g_q itself, in the tcp_stack_t we 642 * need to keep tcps_g_q around so that a closing connection can 643 * switch to using tcps_g_q as part of it closing. 644 */ 645 #define TCPS_REFHOLD(tcps) { \ 646 atomic_add_32(&(tcps)->tcps_refcnt, 1); \ 647 ASSERT((tcps)->tcps_refcnt != 0); \ 648 DTRACE_PROBE1(tcps__refhold, tcp_stack_t, tcps); \ 649 } 650 651 /* 652 * Decrement the reference count on the tcp_stack_t. 653 * In architectures e.g sun4u, where atomic_add_32_nv is just 654 * a cas, we need to maintain the right memory barrier semantics 655 * as that of mutex_exit i.e all the loads and stores should complete 656 * before the cas is executed. membar_exit() does that here. 657 */ 658 #define TCPS_REFRELE(tcps) { \ 659 ASSERT((tcps)->tcps_refcnt != 0); \ 660 membar_exit(); \ 661 DTRACE_PROBE1(tcps__refrele, tcp_stack_t, tcps); \ 662 if (atomic_add_32_nv(&(tcps)->tcps_refcnt, -1) == 0 && \ 663 (tcps)->tcps_g_q != NULL) { \ 664 /* Only tcps_g_q left */ \ 665 tcp_g_q_inactive(tcps); \ 666 } \ 667 } 668 669 extern void tcp_free(tcp_t *tcp); 670 extern void tcp_ddi_g_init(void); 671 extern void tcp_ddi_g_destroy(void); 672 extern void tcp_g_q_inactive(tcp_stack_t *); 673 extern void tcp_xmit_listeners_reset(mblk_t *mp, uint_t ip_hdr_len, 674 zoneid_t zoneid, tcp_stack_t *, conn_t *connp); 675 extern void tcp_conn_request(void *arg, mblk_t *mp, void *arg2); 676 extern void tcp_conn_request_unbound(void *arg, mblk_t *mp, void *arg2); 677 extern void tcp_input(void *arg, mblk_t *mp, void *arg2); 678 extern void tcp_rput_data(void *arg, mblk_t *mp, void *arg2); 679 extern void *tcp_get_conn(void *arg, tcp_stack_t *); 680 extern void tcp_time_wait_collector(void *arg); 681 extern mblk_t *tcp_snmp_get(queue_t *, mblk_t *); 682 extern int tcp_snmp_set(queue_t *, int, int, uchar_t *, int len); 683 extern mblk_t *tcp_xmit_mp(tcp_t *tcp, mblk_t *mp, int32_t max_to_send, 684 int32_t *offset, mblk_t **end_mp, uint32_t seq, 685 boolean_t sendall, uint32_t *seg_len, boolean_t rexmit); 686 extern void tcp_xmit_reset(void *arg, mblk_t *mp, void *arg2); 687 688 /* 689 * The TCP Fanout structure. 690 * The hash tables and their linkage (tcp_*_hash_next, tcp_ptp*hn) are 691 * protected by the per-bucket tf_lock. Each tcp_t 692 * inserted in the list points back at this lock using tcp_*_lockp. 693 * 694 * The listener and acceptor hash queues are lists of tcp_t. 695 */ 696 /* listener hash and acceptor hash queue head */ 697 typedef struct tf_s { 698 tcp_t *tf_tcp; 699 kmutex_t tf_lock; 700 } tf_t; 701 #endif /* (defined(_KERNEL) || defined(_KMEMUSER)) */ 702 703 /* Contract private interface between TCP and Clustering. */ 704 705 #define CL_TCPI_V1 1 /* cl_tcpi_version number */ 706 707 typedef struct cl_tcp_info_s { 708 ushort_t cl_tcpi_version; /* cl_tcp_info_t's version no */ 709 ushort_t cl_tcpi_ipversion; /* IP version */ 710 int32_t cl_tcpi_state; /* TCP state */ 711 in_port_t cl_tcpi_lport; /* Local port */ 712 in_port_t cl_tcpi_fport; /* Remote port */ 713 in6_addr_t cl_tcpi_laddr_v6; /* Local IP address */ 714 in6_addr_t cl_tcpi_faddr_v6; /* Remote IP address */ 715 #ifdef _KERNEL 716 /* Note: V4_PART_OF_V6 is meant to be used only for _KERNEL defined stuff */ 717 #define cl_tcpi_laddr V4_PART_OF_V6(cl_tcpi_laddr_v6) 718 #define cl_tcpi_faddr V4_PART_OF_V6(cl_tcpi_faddr_v6) 719 720 #endif /* _KERNEL */ 721 } cl_tcp_info_t; 722 723 /* 724 * Contracted Consolidation Private ioctl for aborting TCP connections. 725 * In order to keep the offsets and size of the structure the same between 726 * a 32-bit application and a 64-bit amd64 kernel, we use a #pragma 727 * pack(4). 728 */ 729 #define TCP_IOC_ABORT_CONN (('T' << 8) + 91) 730 731 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4 732 #pragma pack(4) 733 #endif 734 735 typedef struct tcp_ioc_abort_conn_s { 736 struct sockaddr_storage ac_local; /* local addr and port */ 737 struct sockaddr_storage ac_remote; /* remote addr and port */ 738 int32_t ac_start; /* start state */ 739 int32_t ac_end; /* end state */ 740 int32_t ac_zoneid; /* zoneid */ 741 } tcp_ioc_abort_conn_t; 742 743 #if _LONG_LONG_ALIGNMENT == 8 && _LONG_LONG_ALIGNMENT_32 == 4 744 #pragma pack() 745 #endif 746 747 #if (defined(_KERNEL) || defined(_KMEMUSER)) 748 extern void tcp_rput_other(tcp_t *tcp, mblk_t *mp); 749 #endif 750 751 #if (defined(_KERNEL)) 752 #define TCP_XRE_EVENT_IP_FANOUT_TCP 1 753 754 /* 755 * This is a private structure used to pass data to an squeue function during 756 * tcp's listener reset sending path. 757 */ 758 typedef struct tcp_xmit_reset_event { 759 int tcp_xre_event; 760 int tcp_xre_iphdrlen; 761 zoneid_t tcp_xre_zoneid; 762 tcp_stack_t *tcp_xre_tcps; 763 } tcp_xmit_reset_event_t; 764 #endif 765 766 #ifdef __cplusplus 767 } 768 #endif 769 770 #endif /* _INET_TCP_H */ 771