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 2007 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #ifndef _INET_TCP_IMPL_H 27 #define _INET_TCP_IMPL_H 28 29 #pragma ident "%Z%%M% %I% %E% SMI" 30 31 /* 32 * TCP implementation private declarations. These interfaces are 33 * used to build the IP module and are not meant to be accessed 34 * by any modules except IP itself. They are undocumented and are 35 * subject to change without notice. 36 */ 37 38 #ifdef __cplusplus 39 extern "C" { 40 #endif 41 42 #ifdef _KERNEL 43 44 #include <inet/tcp.h> 45 46 #define TCP_MOD_ID 5105 47 48 /* 49 * Was this tcp created via socket() interface? 50 */ 51 #define TCP_IS_SOCKET(tcp) ((tcp)->tcp_issocket) 52 53 /* 54 * Is this tcp not attached to any upper client? 55 */ 56 #define TCP_IS_DETACHED(tcp) ((tcp)->tcp_detached) 57 58 #define TCP_TIMER(tcp, f, tim) \ 59 tcp_timeout(tcp->tcp_connp, f, tim) 60 #define TCP_TIMER_CANCEL(tcp, id) \ 61 tcp_timeout_cancel(tcp->tcp_connp, id) 62 63 /* 64 * To restart the TCP retransmission timer. 65 */ 66 #define TCP_TIMER_RESTART(tcp, intvl) { \ 67 if ((tcp)->tcp_timer_tid != 0) \ 68 (void) TCP_TIMER_CANCEL((tcp), (tcp)->tcp_timer_tid); \ 69 (tcp)->tcp_timer_tid = TCP_TIMER((tcp), tcp_timer, \ 70 MSEC_TO_TICK(intvl)); \ 71 } 72 73 /* 74 * This stops synchronous streams for a fused tcp endpoint 75 * and prevents tcp_fuse_rrw() from pulling data from it. 76 */ 77 #define TCP_FUSE_SYNCSTR_STOP(tcp) { \ 78 if ((tcp)->tcp_direct_sockfs) { \ 79 mutex_enter(&(tcp)->tcp_non_sq_lock); \ 80 (tcp)->tcp_fuse_syncstr_stopped = B_TRUE; \ 81 mutex_exit(&(tcp)->tcp_non_sq_lock); \ 82 } \ 83 } 84 85 /* 86 * This causes all calls to tcp_fuse_rrw() to block until 87 * TCP_FUSE_SYNCSTR_UNPLUG_DRAIN() is called. 88 */ 89 #define TCP_FUSE_SYNCSTR_PLUG_DRAIN(tcp) { \ 90 if ((tcp)->tcp_direct_sockfs) { \ 91 mutex_enter(&(tcp)->tcp_non_sq_lock); \ 92 ASSERT(!(tcp)->tcp_fuse_syncstr_plugged); \ 93 (tcp)->tcp_fuse_syncstr_plugged = B_TRUE; \ 94 mutex_exit(&(tcp)->tcp_non_sq_lock); \ 95 } \ 96 } 97 98 /* 99 * This unplugs the draining of data through tcp_fuse_rrw(); see 100 * the comments in tcp_fuse_rrw() for how we preserve ordering. 101 */ 102 #define TCP_FUSE_SYNCSTR_UNPLUG_DRAIN(tcp) { \ 103 if ((tcp)->tcp_direct_sockfs) { \ 104 mutex_enter(&(tcp)->tcp_non_sq_lock); \ 105 (tcp)->tcp_fuse_syncstr_plugged = B_FALSE; \ 106 (void) cv_broadcast(&(tcp)->tcp_fuse_plugcv); \ 107 mutex_exit(&(tcp)->tcp_non_sq_lock); \ 108 } \ 109 } 110 111 /* 112 * Before caching the conn IRE, we need to make sure certain TCP 113 * states are in sync with the ire. The mismatch could occur if the 114 * TCP state has been set in tcp_adapt_ire() using a different IRE, 115 * e.g if an address was not present during an initial connect(), 116 * tcp_adapt_ire() will set the state using the interface route. 117 * Subsequently, if the address is added to the local machine, the 118 * retransmitted SYN will get the correct (loopback) IRE, but the TCP 119 * state (tcp_loopback and tcp_localnet) will remain out of sync. 120 * This is especially an issue with TCP fusion which relies on the 121 * TCP state to be accurate. 122 * 123 * This check/change should be made only if the TCP is not yet in 124 * the established state, else it would lead to inconsistencies. 125 */ 126 #define TCP_CHECK_IREINFO(tcp, ire) { \ 127 if ((tcp)->tcp_state < TCPS_ESTABLISHED) { \ 128 if (((ire)->ire_type & (IRE_LOOPBACK | \ 129 IRE_LOCAL)) && !(tcp)->tcp_loopback) { \ 130 (tcp)->tcp_loopback = B_TRUE; \ 131 } else if ((tcp)->tcp_loopback && \ 132 !((ire)->ire_type & (IRE_LOOPBACK | IRE_LOCAL))) { \ 133 (tcp)->tcp_loopback = B_FALSE; \ 134 } \ 135 if ((tcp)->tcp_ipversion == IPV4_VERSION) { \ 136 (tcp)->tcp_localnet = \ 137 ((ire)->ire_gateway_addr == 0); \ 138 } else { \ 139 (tcp)->tcp_localnet = \ 140 IN6_IS_ADDR_UNSPECIFIED( \ 141 &(ire)->ire_gateway_addr_v6); \ 142 } \ 143 } \ 144 } 145 146 /* 147 * Write-side flow-control is implemented via the per instance STREAMS 148 * write-side Q by explicitly setting QFULL to stop the flow of mblk_t(s) 149 * and clearing QFULL and calling qbackenable() to restart the flow based 150 * on the number of TCP unsent bytes (i.e. those not on the wire waiting 151 * for a remote ACK). 152 * 153 * This is different than a standard STREAMS kmod which when using the 154 * STREAMS Q the framework would automatictly flow-control based on the 155 * defined hiwat/lowat values as mblk_t's are enqueued/dequeued. 156 * 157 * As of FireEngine TCP write-side flow-control needs to take into account 158 * both the unsent tcp_xmit list bytes but also any squeue_t enqueued bytes 159 * (i.e. from tcp_wput() -> tcp_output()). 160 * 161 * This is accomplished by adding a new tcp_t fields, tcp_squeue_bytes, to 162 * count the number of bytes enqueued by tcp_wput() and the number of bytes 163 * dequeued and processed by tcp_output(). 164 * 165 * So, the total number of bytes unsent is (squeue_bytes + unsent) with all 166 * flow-control uses of unsent replaced with the macro TCP_UNSENT_BYTES. 167 */ 168 extern void tcp_clrqfull(tcp_t *); 169 extern void tcp_setqfull(tcp_t *); 170 171 #define TCP_UNSENT_BYTES(tcp) \ 172 ((tcp)->tcp_squeue_bytes + (tcp)->tcp_unsent) 173 174 /* Named Dispatch Parameter Management Structure */ 175 typedef struct tcpparam_s { 176 uint32_t tcp_param_min; 177 uint32_t tcp_param_max; 178 uint32_t tcp_param_val; 179 char *tcp_param_name; 180 } tcpparam_t; 181 182 183 #define tcps_time_wait_interval tcps_params[0].tcp_param_val 184 #define tcps_conn_req_max_q tcps_params[1].tcp_param_val 185 #define tcps_conn_req_max_q0 tcps_params[2].tcp_param_val 186 #define tcps_conn_req_min tcps_params[3].tcp_param_val 187 #define tcps_conn_grace_period tcps_params[4].tcp_param_val 188 #define tcps_cwnd_max_ tcps_params[5].tcp_param_val 189 #define tcps_dbg tcps_params[6].tcp_param_val 190 #define tcps_smallest_nonpriv_port tcps_params[7].tcp_param_val 191 #define tcps_ip_abort_cinterval tcps_params[8].tcp_param_val 192 #define tcps_ip_abort_linterval tcps_params[9].tcp_param_val 193 #define tcps_ip_abort_interval tcps_params[10].tcp_param_val 194 #define tcps_ip_notify_cinterval tcps_params[11].tcp_param_val 195 #define tcps_ip_notify_interval tcps_params[12].tcp_param_val 196 #define tcps_ipv4_ttl tcps_params[13].tcp_param_val 197 #define tcps_keepalive_interval_high tcps_params[14].tcp_param_max 198 #define tcps_keepalive_interval tcps_params[14].tcp_param_val 199 #define tcps_keepalive_interval_low tcps_params[14].tcp_param_min 200 #define tcps_maxpsz_multiplier tcps_params[15].tcp_param_val 201 #define tcps_mss_def_ipv4 tcps_params[16].tcp_param_val 202 #define tcps_mss_max_ipv4 tcps_params[17].tcp_param_val 203 #define tcps_mss_min tcps_params[18].tcp_param_val 204 #define tcps_naglim_def tcps_params[19].tcp_param_val 205 #define tcps_rexmit_interval_initial tcps_params[20].tcp_param_val 206 #define tcps_rexmit_interval_max tcps_params[21].tcp_param_val 207 #define tcps_rexmit_interval_min tcps_params[22].tcp_param_val 208 #define tcps_deferred_ack_interval tcps_params[23].tcp_param_val 209 #define tcps_snd_lowat_fraction tcps_params[24].tcp_param_val 210 #define tcps_sth_rcv_hiwat tcps_params[25].tcp_param_val 211 #define __tcps_not_used1 tcps_params[26].tcp_param_val 212 #define tcps_dupack_fast_retransmit tcps_params[27].tcp_param_val 213 #define tcps_ignore_path_mtu tcps_params[28].tcp_param_val 214 #define tcps_smallest_anon_port tcps_params[29].tcp_param_val 215 #define tcps_largest_anon_port tcps_params[30].tcp_param_val 216 #define tcps_xmit_hiwat tcps_params[31].tcp_param_val 217 #define tcps_xmit_lowat tcps_params[32].tcp_param_val 218 #define tcps_recv_hiwat tcps_params[33].tcp_param_val 219 #define tcps_recv_hiwat_minmss tcps_params[34].tcp_param_val 220 #define tcps_fin_wait_2_flush_interval tcps_params[35].tcp_param_val 221 #define __tcps_not_used2 tcps_params[36].tcp_param_val 222 #define tcps_max_buf tcps_params[37].tcp_param_val 223 #define tcps_strong_iss tcps_params[38].tcp_param_val 224 #define tcps_rtt_updates tcps_params[39].tcp_param_val 225 #define tcps_wscale_always tcps_params[40].tcp_param_val 226 #define tcps_tstamp_always tcps_params[41].tcp_param_val 227 #define tcps_tstamp_if_wscale tcps_params[42].tcp_param_val 228 #define tcps_rexmit_interval_extra tcps_params[43].tcp_param_val 229 #define tcps_deferred_acks_max tcps_params[44].tcp_param_val 230 #define tcps_slow_start_after_idle tcps_params[45].tcp_param_val 231 #define tcps_slow_start_initial tcps_params[46].tcp_param_val 232 #define tcps_co_timer_interval tcps_params[47].tcp_param_val 233 #define tcps_sack_permitted tcps_params[48].tcp_param_val 234 #define tcps_trace tcps_params[49].tcp_param_val 235 #define __tcps_not_used4 tcps_params[50].tcp_param_val 236 #define tcps_ipv6_hoplimit tcps_params[51].tcp_param_val 237 #define tcps_mss_def_ipv6 tcps_params[52].tcp_param_val 238 #define tcps_mss_max_ipv6 tcps_params[53].tcp_param_val 239 #define tcps_rev_src_routes tcps_params[54].tcp_param_val 240 #define tcps_local_dack_interval tcps_params[55].tcp_param_val 241 #define tcps_ndd_get_info_interval tcps_params[56].tcp_param_val 242 #define tcps_local_dacks_max tcps_params[57].tcp_param_val 243 #define tcps_ecn_permitted tcps_params[58].tcp_param_val 244 #define tcps_rst_sent_rate_enabled tcps_params[59].tcp_param_val 245 #define tcps_rst_sent_rate tcps_params[60].tcp_param_val 246 #define tcps_push_timer_interval tcps_params[61].tcp_param_val 247 #define tcps_use_smss_as_mss_opt tcps_params[62].tcp_param_val 248 #define tcps_keepalive_abort_interval_high tcps_params[63].tcp_param_max 249 #define tcps_keepalive_abort_interval tcps_params[63].tcp_param_val 250 #define tcps_keepalive_abort_interval_low tcps_params[63].tcp_param_min 251 252 extern struct qinit tcp_loopback_rinit, tcp_rinit; 253 extern boolean_t do_tcp_fusion; 254 255 extern int tcp_maxpsz_set(tcp_t *, boolean_t); 256 extern void tcp_timers_stop(tcp_t *); 257 extern void tcp_rcv_enqueue(tcp_t *, mblk_t *, uint_t); 258 extern void tcp_push_timer(void *); 259 extern timeout_id_t tcp_timeout(conn_t *, void (*)(void *), clock_t); 260 extern clock_t tcp_timeout_cancel(conn_t *, timeout_id_t); 261 262 extern void tcp_fuse(tcp_t *, uchar_t *, tcph_t *); 263 extern void tcp_unfuse(tcp_t *); 264 extern boolean_t tcp_fuse_output(tcp_t *, mblk_t *, uint32_t); 265 extern void tcp_fuse_output_urg(tcp_t *, mblk_t *); 266 extern boolean_t tcp_fuse_rcv_drain(queue_t *, tcp_t *, mblk_t **); 267 extern void tcp_fuse_syncstr_enable_pair(tcp_t *); 268 extern void tcp_fuse_disable_pair(tcp_t *, boolean_t); 269 extern int tcp_fuse_rrw(queue_t *, struiod_t *); 270 extern int tcp_fuse_rinfop(queue_t *, infod_t *); 271 extern size_t tcp_fuse_set_rcv_hiwat(tcp_t *, size_t); 272 extern int tcp_fuse_maxpsz_set(tcp_t *); 273 274 #endif /* _KERNEL */ 275 276 #ifdef __cplusplus 277 } 278 #endif 279 280 #endif /* _INET_TCP_IMPL_H */ 281