1 0 stevel /* 2 0 stevel * CDDL HEADER START 3 0 stevel * 4 0 stevel * The contents of this file are subject to the terms of the 5 3388 kcpoon * Common Development and Distribution License (the "License"). 6 3388 kcpoon * You may not use this file except in compliance with the License. 7 0 stevel * 8 0 stevel * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 0 stevel * or http://www.opensolaris.org/os/licensing. 10 0 stevel * See the License for the specific language governing permissions 11 0 stevel * and limitations under the License. 12 0 stevel * 13 0 stevel * When distributing Covered Code, include this CDDL HEADER in each 14 0 stevel * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 0 stevel * If applicable, add the following below this CDDL HEADER, with the 16 0 stevel * fields enclosed by brackets "[]" replaced with your own identifying 17 0 stevel * information: Portions Copyright [yyyy] [name of copyright owner] 18 0 stevel * 19 0 stevel * CDDL HEADER END 20 0 stevel */ 21 0 stevel /* 22 8477 Rao * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 23 0 stevel * Use is subject to license terms. 24 0 stevel */ 25 0 stevel 26 0 stevel #include <sys/types.h> 27 0 stevel #include <sys/stream.h> 28 0 stevel #include <sys/strsubr.h> 29 0 stevel #include <sys/stropts.h> 30 0 stevel #include <sys/strsun.h> 31 0 stevel #include <sys/strlog.h> 32 0 stevel #define _SUN_TPI_VERSION 2 33 0 stevel #include <sys/tihdr.h> 34 0 stevel #include <sys/timod.h> 35 0 stevel #include <sys/ddi.h> 36 0 stevel #include <sys/sunddi.h> 37 0 stevel #include <sys/cmn_err.h> 38 0 stevel #include <sys/proc.h> 39 0 stevel #include <sys/suntpi.h> 40 0 stevel #include <sys/policy.h> 41 3448 dh155122 #include <sys/zone.h> 42 8348 Eric #include <sys/disp.h> 43 0 stevel 44 0 stevel #include <sys/socket.h> 45 8348 Eric #include <sys/socketvar.h> 46 0 stevel #include <netinet/in.h> 47 0 stevel 48 0 stevel #include <inet/common.h> 49 0 stevel #include <netinet/ip6.h> 50 0 stevel #include <inet/ip.h> 51 5240 nordmark #include <inet/ipclassifier.h> 52 8348 Eric #include <inet/proto_set.h> 53 0 stevel #include <inet/nd.h> 54 0 stevel #include <inet/optcom.h> 55 0 stevel #include <netinet/ip_mroute.h> 56 0 stevel #include <sys/isa_defs.h> 57 0 stevel #include <net/route.h> 58 5240 nordmark 59 5240 nordmark #include <inet/rts_impl.h> 60 5240 nordmark #include <inet/ip_rts.h> 61 0 stevel 62 0 stevel /* 63 0 stevel * This is a transport provider for routing sockets. Downstream messages are 64 0 stevel * wrapped with a IP_IOCTL header, and ip_wput_ioctl calls the appropriate entry 65 0 stevel * in the ip_ioctl_ftbl callout table to pass the routing socket data into IP. 66 0 stevel * Upstream messages are generated for listeners of the routing socket as well 67 0 stevel * as the message sender (unless they have turned off their end using 68 0 stevel * SO_USELOOPBACK or shutdown(3n)). Upstream messages may also be generated 69 0 stevel * asynchronously when: 70 0 stevel * 71 0 stevel * Interfaces are brought up or down. 72 0 stevel * Addresses are assigned to interfaces. 73 4365 nordmark * ICMP redirects are processed and a IRE_HOST/RTF_DYNAMIC is installed. 74 0 stevel * No route is found while sending a packet. 75 0 stevel * 76 0 stevel * Since all we do is reformat the messages between routing socket and 77 0 stevel * ioctl forms, no synchronization is necessary in this module; all 78 0 stevel * the dirty work is done down in ip. 79 0 stevel */ 80 0 stevel 81 0 stevel /* Default structure copied into T_INFO_ACK messages */ 82 0 stevel static struct T_info_ack rts_g_t_info_ack = { 83 0 stevel T_INFO_ACK, 84 0 stevel T_INFINITE, /* TSDU_size. Maximum size messages. */ 85 0 stevel T_INVALID, /* ETSDU_size. No expedited data. */ 86 0 stevel T_INVALID, /* CDATA_size. No connect data. */ 87 0 stevel T_INVALID, /* DDATA_size. No disconnect data. */ 88 0 stevel 0, /* ADDR_size. */ 89 0 stevel 0, /* OPT_size - not initialized here */ 90 0 stevel 64 * 1024, /* TIDU_size. rts allows maximum size messages. */ 91 0 stevel T_COTS, /* SERV_type. rts supports connection oriented. */ 92 0 stevel TS_UNBND, /* CURRENT_state. This is set from rts_state. */ 93 0 stevel (XPG4_1) /* PROVIDER_flag */ 94 0 stevel }; 95 0 stevel 96 0 stevel /* 97 0 stevel * Table of ND variables supported by rts. These are loaded into rts_g_nd 98 0 stevel * in rts_open. 99 0 stevel * All of these are alterable, within the min/max values given, at run time. 100 0 stevel */ 101 3448 dh155122 static rtsparam_t lcl_param_arr[] = { 102 0 stevel /* min max value name */ 103 0 stevel { 4096, 65536, 8192, "rts_xmit_hiwat"}, 104 0 stevel { 0, 65536, 1024, "rts_xmit_lowat"}, 105 0 stevel { 4096, 65536, 8192, "rts_recv_hiwat"}, 106 0 stevel { 65536, 1024*1024*1024, 256*1024, "rts_max_buf"}, 107 0 stevel }; 108 3448 dh155122 #define rtss_xmit_hiwat rtss_params[0].rts_param_value 109 3448 dh155122 #define rtss_xmit_lowat rtss_params[1].rts_param_value 110 3448 dh155122 #define rtss_recv_hiwat rtss_params[2].rts_param_value 111 5240 nordmark #define rtss_max_buf rtss_params[3].rts_param_value 112 0 stevel 113 0 stevel static void rts_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, 114 0 stevel int sys_error); 115 11042 Erik static void rts_input(void *, mblk_t *, void *, ip_recv_attr_t *); 116 11042 Erik static void rts_icmp_input(void *, mblk_t *, void *, ip_recv_attr_t *); 117 8778 Erik static mblk_t *rts_ioctl_alloc(mblk_t *data); 118 0 stevel static int rts_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr); 119 3448 dh155122 static boolean_t rts_param_register(IDP *ndp, rtsparam_t *rtspa, int cnt); 120 0 stevel static int rts_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, 121 0 stevel cred_t *cr); 122 5240 nordmark static void rts_rsrv(queue_t *q); 123 3448 dh155122 static void *rts_stack_init(netstackid_t stackid, netstack_t *ns); 124 3448 dh155122 static void rts_stack_fini(netstackid_t stackid, void *arg); 125 0 stevel static void rts_wput(queue_t *q, mblk_t *mp); 126 0 stevel static void rts_wput_iocdata(queue_t *q, mblk_t *mp); 127 0 stevel static void rts_wput_other(queue_t *q, mblk_t *mp); 128 0 stevel static int rts_wrw(queue_t *q, struiod_t *dp); 129 0 stevel 130 8348 Eric static int rts_stream_open(queue_t *q, dev_t *devp, int flag, int sflag, 131 8348 Eric cred_t *credp); 132 8348 Eric static conn_t *rts_open(int flag, cred_t *credp); 133 8348 Eric 134 8348 Eric static int rts_stream_close(queue_t *q); 135 8348 Eric static int rts_close(sock_lower_handle_t proto_handle, int flags, 136 8348 Eric cred_t *cr); 137 8348 Eric 138 5240 nordmark static struct module_info rts_mod_info = { 139 0 stevel 129, "rts", 1, INFPSZ, 512, 128 140 0 stevel }; 141 0 stevel 142 5240 nordmark static struct qinit rtsrinit = { 143 8348 Eric NULL, (pfi_t)rts_rsrv, rts_stream_open, rts_stream_close, NULL, 144 8348 Eric &rts_mod_info 145 0 stevel }; 146 0 stevel 147 5240 nordmark static struct qinit rtswinit = { 148 5240 nordmark (pfi_t)rts_wput, NULL, NULL, NULL, NULL, &rts_mod_info, 149 0 stevel NULL, (pfi_t)rts_wrw, NULL, STRUIOT_STANDARD 150 0 stevel }; 151 0 stevel 152 0 stevel struct streamtab rtsinfo = { 153 5240 nordmark &rtsrinit, &rtswinit 154 0 stevel }; 155 0 stevel 156 0 stevel /* 157 0 stevel * This routine allocates the necessary 158 0 stevel * message blocks for IOCTL wrapping the 159 0 stevel * user data. 160 0 stevel */ 161 0 stevel static mblk_t * 162 8778 Erik rts_ioctl_alloc(mblk_t *data) 163 0 stevel { 164 0 stevel mblk_t *mp = NULL; 165 0 stevel mblk_t *mp1 = NULL; 166 0 stevel ipllc_t *ipllc; 167 0 stevel struct iocblk *ioc; 168 0 stevel 169 8778 Erik mp = allocb_tmpl(sizeof (ipllc_t), data); 170 0 stevel if (mp == NULL) 171 0 stevel return (NULL); 172 8778 Erik mp1 = allocb_tmpl(sizeof (struct iocblk), data); 173 0 stevel if (mp1 == NULL) { 174 0 stevel freeb(mp); 175 0 stevel return (NULL); 176 0 stevel } 177 0 stevel 178 0 stevel ipllc = (ipllc_t *)mp->b_rptr; 179 0 stevel ipllc->ipllc_cmd = IP_IOC_RTS_REQUEST; 180 0 stevel ipllc->ipllc_name_offset = 0; 181 0 stevel ipllc->ipllc_name_length = 0; 182 0 stevel mp->b_wptr += sizeof (ipllc_t); 183 0 stevel mp->b_cont = data; 184 0 stevel 185 0 stevel ioc = (struct iocblk *)mp1->b_rptr; 186 0 stevel ioc->ioc_cmd = IP_IOCTL; 187 0 stevel ioc->ioc_error = 0; 188 0 stevel ioc->ioc_cr = NULL; 189 0 stevel ioc->ioc_count = msgdsize(mp); 190 0 stevel mp1->b_wptr += sizeof (struct iocblk); 191 0 stevel mp1->b_datap->db_type = M_IOCTL; 192 0 stevel mp1->b_cont = mp; 193 0 stevel 194 0 stevel return (mp1); 195 0 stevel } 196 0 stevel 197 0 stevel /* 198 0 stevel * This routine closes rts stream, by disabling 199 0 stevel * put/srv routines and freeing the this module 200 0 stevel * internal datastructure. 201 0 stevel */ 202 0 stevel static int 203 8348 Eric rts_common_close(queue_t *q, conn_t *connp) 204 0 stevel { 205 5240 nordmark 206 5240 nordmark ASSERT(connp != NULL && IPCL_IS_RTS(connp)); 207 5240 nordmark 208 5240 nordmark ip_rts_unregister(connp); 209 5240 nordmark 210 5240 nordmark ip_quiesce_conn(connp); 211 3448 dh155122 212 8348 Eric if (!IPCL_IS_NONSTR(connp)) { 213 8348 Eric qprocsoff(q); 214 11042 Erik } 215 0 stevel 216 11042 Erik /* 217 11042 Erik * Now we are truly single threaded on this stream, and can 218 11042 Erik * delete the things hanging off the connp, and finally the connp. 219 11042 Erik * We removed this connp from the fanout list, it cannot be 220 11042 Erik * accessed thru the fanouts, and we already waited for the 221 11042 Erik * conn_ref to drop to 0. We are already in close, so 222 11042 Erik * there cannot be any other thread from the top. qprocsoff 223 11042 Erik * has completed, and service has completed or won't run in 224 11042 Erik * future. 225 11042 Erik */ 226 11042 Erik ASSERT(connp->conn_ref == 1); 227 11042 Erik 228 11042 Erik if (!IPCL_IS_NONSTR(connp)) { 229 8348 Eric inet_minor_free(connp->conn_minor_arena, connp->conn_dev); 230 8348 Eric } else { 231 8477 Rao ip_free_helper_stream(connp); 232 8348 Eric } 233 5240 nordmark 234 5240 nordmark connp->conn_ref--; 235 5240 nordmark ipcl_conn_destroy(connp); 236 8348 Eric return (0); 237 8348 Eric } 238 8348 Eric 239 8348 Eric static int 240 8348 Eric rts_stream_close(queue_t *q) 241 8348 Eric { 242 8348 Eric conn_t *connp = Q_TO_CONN(q); 243 8348 Eric 244 8348 Eric (void) rts_common_close(q, connp); 245 5240 nordmark q->q_ptr = WR(q)->q_ptr = NULL; 246 0 stevel return (0); 247 0 stevel } 248 0 stevel 249 0 stevel /* 250 0 stevel * This is the open routine for routing socket. It allocates 251 5240 nordmark * rts_t structure for the stream and tells IP that it is a routing socket. 252 0 stevel */ 253 0 stevel /* ARGSUSED */ 254 0 stevel static int 255 8348 Eric rts_stream_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp) 256 0 stevel { 257 5240 nordmark conn_t *connp; 258 5240 nordmark dev_t conn_dev; 259 8348 Eric rts_t *rts; 260 0 stevel 261 0 stevel /* If the stream is already open, return immediately. */ 262 0 stevel if (q->q_ptr != NULL) 263 0 stevel return (0); 264 0 stevel 265 5240 nordmark if (sflag == MODOPEN) 266 0 stevel return (EINVAL); 267 8348 Eric 268 8348 Eric /* 269 8348 Eric * Since RTS is not used so heavily, allocating from the small 270 8348 Eric * arena should be sufficient. 271 8348 Eric */ 272 8348 Eric if ((conn_dev = inet_minor_alloc(ip_minor_arena_sa)) == 0) { 273 8348 Eric return (EBUSY); 274 8348 Eric } 275 8348 Eric 276 8348 Eric connp = rts_open(flag, credp); 277 8348 Eric ASSERT(connp != NULL); 278 8348 Eric 279 8348 Eric *devp = makedevice(getemajor(*devp), (minor_t)conn_dev); 280 8348 Eric 281 8348 Eric rts = connp->conn_rts; 282 8348 Eric rw_enter(&rts->rts_rwlock, RW_WRITER); 283 8348 Eric connp->conn_dev = conn_dev; 284 8348 Eric connp->conn_minor_arena = ip_minor_arena_sa; 285 8348 Eric 286 8348 Eric q->q_ptr = connp; 287 8348 Eric WR(q)->q_ptr = connp; 288 8348 Eric connp->conn_rq = q; 289 8348 Eric connp->conn_wq = WR(q); 290 8348 Eric 291 11042 Erik WR(q)->q_hiwat = connp->conn_sndbuf; 292 11042 Erik WR(q)->q_lowat = connp->conn_sndlowat; 293 8348 Eric 294 8348 Eric mutex_enter(&connp->conn_lock); 295 8348 Eric connp->conn_state_flags &= ~CONN_INCIPIENT; 296 8348 Eric mutex_exit(&connp->conn_lock); 297 11042 Erik rw_exit(&rts->rts_rwlock); 298 11042 Erik 299 11042 Erik /* Indicate to IP that this is a routing socket client */ 300 11042 Erik ip_rts_register(connp); 301 8348 Eric 302 8348 Eric qprocson(q); 303 8348 Eric 304 8348 Eric return (0); 305 8348 Eric } 306 8348 Eric 307 8348 Eric /* ARGSUSED */ 308 8348 Eric static conn_t * 309 8348 Eric rts_open(int flag, cred_t *credp) 310 8348 Eric { 311 8348 Eric netstack_t *ns; 312 8348 Eric rts_stack_t *rtss; 313 8348 Eric rts_t *rts; 314 8348 Eric conn_t *connp; 315 8348 Eric zoneid_t zoneid; 316 0 stevel 317 3448 dh155122 ns = netstack_find_by_cred(credp); 318 3448 dh155122 ASSERT(ns != NULL); 319 3448 dh155122 rtss = ns->netstack_rts; 320 3448 dh155122 ASSERT(rtss != NULL); 321 3448 dh155122 322 5240 nordmark /* 323 5240 nordmark * For exclusive stacks we set the zoneid to zero 324 5240 nordmark * to make RTS operate as if in the global zone. 325 5240 nordmark */ 326 5240 nordmark if (ns->netstack_stackid != GLOBAL_NETSTACKID) 327 5240 nordmark zoneid = GLOBAL_ZONEID; 328 5240 nordmark else 329 5240 nordmark zoneid = crgetzoneid(credp); 330 5240 nordmark 331 5240 nordmark connp = ipcl_conn_create(IPCL_RTSCONN, KM_SLEEP, ns); 332 5240 nordmark rts = connp->conn_rts; 333 5240 nordmark 334 5240 nordmark /* 335 5240 nordmark * ipcl_conn_create did a netstack_hold. Undo the hold that was 336 5240 nordmark * done by netstack_find_by_cred() 337 5240 nordmark */ 338 5240 nordmark netstack_rele(ns); 339 5240 nordmark 340 5240 nordmark rw_enter(&rts->rts_rwlock, RW_WRITER); 341 5240 nordmark ASSERT(connp->conn_rts == rts); 342 5240 nordmark ASSERT(rts->rts_connp == connp); 343 5240 nordmark 344 11042 Erik connp->conn_ixa->ixa_flags |= IXAF_MULTICAST_LOOP | IXAF_SET_ULP_CKSUM; 345 11042 Erik /* conn_allzones can not be set this early, hence no IPCL_ZONEID */ 346 11042 Erik connp->conn_ixa->ixa_zoneid = zoneid; 347 5240 nordmark connp->conn_zoneid = zoneid; 348 8348 Eric connp->conn_flow_cntrld = B_FALSE; 349 5240 nordmark 350 11042 Erik rts->rts_rtss = rtss; 351 3448 dh155122 352 11042 Erik connp->conn_rcvbuf = rtss->rtss_recv_hiwat; 353 11042 Erik connp->conn_sndbuf = rtss->rtss_xmit_hiwat; 354 11042 Erik connp->conn_sndlowat = rtss->rtss_xmit_lowat; 355 11042 Erik connp->conn_rcvlowat = rts_mod_info.mi_lowat; 356 11042 Erik 357 11042 Erik connp->conn_family = PF_ROUTE; 358 11042 Erik connp->conn_so_type = SOCK_RAW; 359 11042 Erik /* SO_PROTOTYPE is always sent down by sockfs setting conn_proto */ 360 5240 nordmark 361 5240 nordmark connp->conn_recv = rts_input; 362 11042 Erik connp->conn_recvicmp = rts_icmp_input; 363 11042 Erik 364 5240 nordmark crhold(credp); 365 5240 nordmark connp->conn_cred = credp; 366 11042 Erik connp->conn_cpid = curproc->p_pid; 367 11042 Erik /* Cache things in ixa without an extra refhold */ 368 11042 Erik connp->conn_ixa->ixa_cred = connp->conn_cred; 369 11042 Erik connp->conn_ixa->ixa_cpid = connp->conn_cpid; 370 11042 Erik if (is_system_labeled()) 371 11042 Erik connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred); 372 5240 nordmark 373 8348 Eric /* 374 8348 Eric * rts sockets start out as bound and connected 375 8348 Eric * For streams based sockets, socket state is set to 376 8348 Eric * SS_ISBOUND | SS_ISCONNECTED in so_strinit. 377 8348 Eric */ 378 8348 Eric rts->rts_state = TS_DATA_XFER; 379 5240 nordmark rw_exit(&rts->rts_rwlock); 380 5240 nordmark 381 8348 Eric return (connp); 382 0 stevel } 383 0 stevel 384 0 stevel /* 385 0 stevel * This routine creates a T_ERROR_ACK message and passes it upstream. 386 0 stevel */ 387 0 stevel static void 388 0 stevel rts_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error) 389 0 stevel { 390 0 stevel if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL) 391 0 stevel qreply(q, mp); 392 0 stevel } 393 0 stevel 394 0 stevel /* 395 0 stevel * This routine creates a T_OK_ACK message and passes it upstream. 396 0 stevel */ 397 0 stevel static void 398 0 stevel rts_ok_ack(queue_t *q, mblk_t *mp) 399 0 stevel { 400 0 stevel if ((mp = mi_tpi_ok_ack_alloc(mp)) != NULL) 401 0 stevel qreply(q, mp); 402 0 stevel } 403 0 stevel 404 0 stevel /* 405 0 stevel * This routine is called by rts_wput to handle T_UNBIND_REQ messages. 406 0 stevel */ 407 0 stevel static void 408 8348 Eric rts_tpi_unbind(queue_t *q, mblk_t *mp) 409 0 stevel { 410 5240 nordmark conn_t *connp = Q_TO_CONN(q); 411 5240 nordmark rts_t *rts = connp->conn_rts; 412 0 stevel 413 0 stevel /* If a bind has not been done, we can't unbind. */ 414 0 stevel if (rts->rts_state != TS_IDLE) { 415 0 stevel rts_err_ack(q, mp, TOUTSTATE, 0); 416 0 stevel return; 417 0 stevel } 418 0 stevel rts->rts_state = TS_UNBND; 419 0 stevel rts_ok_ack(q, mp); 420 0 stevel } 421 0 stevel 422 0 stevel /* 423 0 stevel * This routine is called to handle each 424 0 stevel * O_T_BIND_REQ/T_BIND_REQ message passed to 425 0 stevel * rts_wput. Note: This routine works with both 426 0 stevel * O_T_BIND_REQ and T_BIND_REQ semantics. 427 0 stevel */ 428 0 stevel static void 429 8348 Eric rts_tpi_bind(queue_t *q, mblk_t *mp) 430 0 stevel { 431 5240 nordmark conn_t *connp = Q_TO_CONN(q); 432 5240 nordmark rts_t *rts = connp->conn_rts; 433 0 stevel struct T_bind_req *tbr; 434 0 stevel 435 0 stevel if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) { 436 0 stevel (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 437 8348 Eric "rts_tpi_bind: bad data, %d", rts->rts_state); 438 0 stevel rts_err_ack(q, mp, TBADADDR, 0); 439 0 stevel return; 440 0 stevel } 441 0 stevel if (rts->rts_state != TS_UNBND) { 442 0 stevel (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 443 8348 Eric "rts_tpi_bind: bad state, %d", rts->rts_state); 444 0 stevel rts_err_ack(q, mp, TOUTSTATE, 0); 445 0 stevel return; 446 0 stevel } 447 0 stevel tbr = (struct T_bind_req *)mp->b_rptr; 448 0 stevel if (tbr->ADDR_length != 0) { 449 0 stevel (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 450 8348 Eric "rts_tpi_bind: bad ADDR_length %d", tbr->ADDR_length); 451 0 stevel rts_err_ack(q, mp, TBADADDR, 0); 452 0 stevel return; 453 0 stevel } 454 0 stevel /* Generic request */ 455 0 stevel tbr->ADDR_offset = (t_scalar_t)sizeof (struct T_bind_req); 456 0 stevel tbr->ADDR_length = 0; 457 0 stevel tbr->PRIM_type = T_BIND_ACK; 458 11042 Erik mp->b_datap->db_type = M_PCPROTO; 459 0 stevel rts->rts_state = TS_IDLE; 460 0 stevel qreply(q, mp); 461 0 stevel } 462 0 stevel 463 0 stevel static void 464 0 stevel rts_copy_info(struct T_info_ack *tap, rts_t *rts) 465 0 stevel { 466 0 stevel *tap = rts_g_t_info_ack; 467 0 stevel tap->CURRENT_state = rts->rts_state; 468 0 stevel tap->OPT_size = rts_max_optsize; 469 0 stevel } 470 0 stevel 471 0 stevel /* 472 0 stevel * This routine responds to T_CAPABILITY_REQ messages. It is called by 473 0 stevel * rts_wput. Much of the T_CAPABILITY_ACK information is copied from 474 0 stevel * rts_g_t_info_ack. The current state of the stream is copied from 475 0 stevel * rts_state. 476 0 stevel */ 477 0 stevel static void 478 0 stevel rts_capability_req(queue_t *q, mblk_t *mp) 479 0 stevel { 480 5240 nordmark conn_t *connp = Q_TO_CONN(q); 481 5240 nordmark rts_t *rts = connp->conn_rts; 482 0 stevel t_uscalar_t cap_bits1; 483 0 stevel struct T_capability_ack *tcap; 484 0 stevel 485 0 stevel cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1; 486 0 stevel 487 0 stevel mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack), 488 5240 nordmark mp->b_datap->db_type, T_CAPABILITY_ACK); 489 0 stevel if (mp == NULL) 490 0 stevel return; 491 0 stevel 492 0 stevel tcap = (struct T_capability_ack *)mp->b_rptr; 493 0 stevel tcap->CAP_bits1 = 0; 494 0 stevel 495 0 stevel if (cap_bits1 & TC1_INFO) { 496 0 stevel rts_copy_info(&tcap->INFO_ack, rts); 497 0 stevel tcap->CAP_bits1 |= TC1_INFO; 498 0 stevel } 499 0 stevel 500 0 stevel qreply(q, mp); 501 0 stevel } 502 0 stevel 503 0 stevel /* 504 0 stevel * This routine responds to T_INFO_REQ messages. It is called by rts_wput. 505 0 stevel * Most of the T_INFO_ACK information is copied from rts_g_t_info_ack. 506 0 stevel * The current state of the stream is copied from rts_state. 507 0 stevel */ 508 0 stevel static void 509 0 stevel rts_info_req(queue_t *q, mblk_t *mp) 510 0 stevel { 511 5240 nordmark conn_t *connp = Q_TO_CONN(q); 512 5240 nordmark rts_t *rts = connp->conn_rts; 513 0 stevel 514 0 stevel mp = tpi_ack_alloc(mp, sizeof (rts_g_t_info_ack), M_PCPROTO, 515 0 stevel T_INFO_ACK); 516 0 stevel if (mp == NULL) 517 0 stevel return; 518 0 stevel rts_copy_info((struct T_info_ack *)mp->b_rptr, rts); 519 0 stevel qreply(q, mp); 520 0 stevel } 521 0 stevel 522 0 stevel /* 523 0 stevel * This routine gets default values of certain options whose default 524 0 stevel * values are maintained by protcol specific code 525 0 stevel */ 526 0 stevel /* ARGSUSED */ 527 0 stevel int 528 0 stevel rts_opt_default(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr) 529 0 stevel { 530 0 stevel /* no default value processed by protocol specific code currently */ 531 0 stevel return (-1); 532 0 stevel } 533 0 stevel 534 8348 Eric 535 8348 Eric static int 536 8348 Eric rts_opt_get(conn_t *connp, int level, int name, uchar_t *ptr) 537 0 stevel { 538 8348 Eric rts_t *rts = connp->conn_rts; 539 11042 Erik conn_opt_arg_t coas; 540 11042 Erik int retval; 541 8348 Eric 542 8348 Eric ASSERT(RW_READ_HELD(&rts->rts_rwlock)); 543 0 stevel 544 0 stevel switch (level) { 545 11042 Erik /* do this in conn_opt_get? */ 546 8485 Peter case SOL_ROUTE: 547 8485 Peter switch (name) { 548 8485 Peter case RT_AWARE: 549 8485 Peter mutex_enter(&connp->conn_lock); 550 11042 Erik *(int *)ptr = connp->conn_rtaware; 551 8485 Peter mutex_exit(&connp->conn_lock); 552 11042 Erik return (0); 553 0 stevel } 554 0 stevel break; 555 0 stevel } 556 11042 Erik coas.coa_connp = connp; 557 11042 Erik coas.coa_ixa = connp->conn_ixa; 558 11042 Erik coas.coa_ipp = &connp->conn_xmit_ipp; 559 11042 Erik mutex_enter(&connp->conn_lock); 560 11042 Erik retval = conn_opt_get(&coas, level, name, ptr); 561 11042 Erik mutex_exit(&connp->conn_lock); 562 11042 Erik return (retval); 563 0 stevel } 564 0 stevel 565 8348 Eric /* ARGSUSED */ 566 8348 Eric static int 567 8348 Eric rts_do_opt_set(conn_t *connp, int level, int name, uint_t inlen, 568 8348 Eric uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, cred_t *cr, 569 8348 Eric void *thisdg_attrs, boolean_t checkonly) 570 0 stevel { 571 0 stevel int *i1 = (int *)invalp; 572 5240 nordmark rts_t *rts = connp->conn_rts; 573 3448 dh155122 rts_stack_t *rtss = rts->rts_rtss; 574 11042 Erik int error; 575 11042 Erik conn_opt_arg_t coas; 576 11042 Erik 577 11042 Erik coas.coa_connp = connp; 578 11042 Erik coas.coa_ixa = connp->conn_ixa; 579 11042 Erik coas.coa_ipp = &connp->conn_xmit_ipp; 580 0 stevel 581 8348 Eric ASSERT(RW_WRITE_HELD(&rts->rts_rwlock)); 582 0 stevel 583 0 stevel /* 584 0 stevel * For rts, we should have no ancillary data sent down 585 0 stevel * (rts_wput doesn't handle options). 586 0 stevel */ 587 0 stevel ASSERT(thisdg_attrs == NULL); 588 0 stevel 589 0 stevel /* 590 0 stevel * For fixed length options, no sanity check 591 0 stevel * of passed in length is done. It is assumed *_optcom_req() 592 0 stevel * routines do the right thing. 593 0 stevel */ 594 0 stevel 595 0 stevel switch (level) { 596 0 stevel case SOL_SOCKET: 597 0 stevel switch (name) { 598 0 stevel case SO_PROTOTYPE: 599 0 stevel /* 600 0 stevel * Routing socket applications that call socket() with 601 0 stevel * a third argument can filter which messages will be 602 0 stevel * sent upstream thanks to sockfs. so_socket() sends 603 0 stevel * down the SO_PROTOTYPE and rts_queue_input() 604 0 stevel * implements the filtering. 605 0 stevel */ 606 11042 Erik if (*i1 != AF_INET && *i1 != AF_INET6) { 607 11042 Erik *outlenp = 0; 608 0 stevel return (EPROTONOSUPPORT); 609 11042 Erik } 610 11042 Erik if (!checkonly) 611 8552 Rao connp->conn_proto = *i1; 612 11042 Erik *outlenp = inlen; 613 11042 Erik return (0); 614 11042 Erik 615 0 stevel /* 616 0 stevel * The following two items can be manipulated, 617 0 stevel * but changing them should do nothing. 618 0 stevel */ 619 0 stevel case SO_SNDBUF: 620 3448 dh155122 if (*i1 > rtss->rtss_max_buf) { 621 0 stevel *outlenp = 0; 622 0 stevel return (ENOBUFS); 623 0 stevel } 624 0 stevel break; /* goto sizeof (int) option return */ 625 0 stevel case SO_RCVBUF: 626 3448 dh155122 if (*i1 > rtss->rtss_max_buf) { 627 0 stevel *outlenp = 0; 628 0 stevel return (ENOBUFS); 629 0 stevel } 630 8485 Peter break; /* goto sizeof (int) option return */ 631 8485 Peter } 632 8485 Peter break; 633 8485 Peter case SOL_ROUTE: 634 8485 Peter switch (name) { 635 8485 Peter case RT_AWARE: 636 8485 Peter if (!checkonly) { 637 8485 Peter mutex_enter(&connp->conn_lock); 638 8485 Peter connp->conn_rtaware = *i1; 639 8485 Peter mutex_exit(&connp->conn_lock); 640 8485 Peter } 641 11042 Erik *outlenp = inlen; 642 11042 Erik return (0); 643 0 stevel } 644 0 stevel break; 645 11042 Erik } 646 11042 Erik /* Serialized setsockopt since we are D_MTQPAIR */ 647 11042 Erik error = conn_opt_set(&coas, level, name, inlen, invalp, 648 11042 Erik checkonly, cr); 649 11042 Erik if (error != 0) { 650 0 stevel *outlenp = 0; 651 11042 Erik return (error); 652 0 stevel } 653 0 stevel /* 654 0 stevel * Common case of return from an option that is sizeof (int) 655 0 stevel */ 656 8348 Eric if (invalp != outvalp) { 657 8348 Eric /* don't trust bcopy for identical src/dst */ 658 8348 Eric (void) bcopy(invalp, outvalp, inlen); 659 8348 Eric } 660 0 stevel *outlenp = (t_uscalar_t)sizeof (int); 661 0 stevel return (0); 662 8348 Eric } 663 8348 Eric 664 8348 Eric static int 665 8348 Eric rts_opt_set(conn_t *connp, uint_t optset_context, int level, int name, 666 8348 Eric uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, 667 8348 Eric void *thisdg_attrs, cred_t *cr) 668 8348 Eric { 669 8348 Eric boolean_t checkonly = B_FALSE; 670 8348 Eric 671 8348 Eric if (optset_context) { 672 8348 Eric switch (optset_context) { 673 8348 Eric case SETFN_OPTCOM_CHECKONLY: 674 8348 Eric checkonly = B_TRUE; 675 8348 Eric /* 676 8348 Eric * Note: Implies T_CHECK semantics for T_OPTCOM_REQ 677 8348 Eric * inlen != 0 implies value supplied and 678 8348 Eric * we have to "pretend" to set it. 679 8348 Eric * inlen == 0 implies that there is no value part 680 8348 Eric * in T_CHECK request and just validation 681 8348 Eric * done elsewhere should be enough, we just return here. 682 8348 Eric */ 683 8348 Eric if (inlen == 0) { 684 8348 Eric *outlenp = 0; 685 8348 Eric return (0); 686 8348 Eric } 687 8348 Eric break; 688 8348 Eric case SETFN_OPTCOM_NEGOTIATE: 689 8348 Eric checkonly = B_FALSE; 690 8348 Eric break; 691 8348 Eric case SETFN_UD_NEGOTIATE: 692 8348 Eric case SETFN_CONN_NEGOTIATE: 693 8348 Eric checkonly = B_FALSE; 694 8348 Eric /* 695 8348 Eric * Negotiating local and "association-related" options 696 8348 Eric * through T_UNITDATA_REQ or T_CONN_{REQ,CON} 697 8348 Eric * Not allowed in this module. 698 8348 Eric */ 699 8348 Eric return (EINVAL); 700 8348 Eric default: 701 8348 Eric /* 702 8348 Eric * We should never get here 703 8348 Eric */ 704 8348 Eric *outlenp = 0; 705 8348 Eric return (EINVAL); 706 8348 Eric } 707 8348 Eric 708 8348 Eric ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) || 709 8348 Eric (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0)); 710 8348 Eric 711 8348 Eric } 712 8348 Eric return (rts_do_opt_set(connp, level, name, inlen, invalp, outlenp, 713 8348 Eric outvalp, cr, thisdg_attrs, checkonly)); 714 8348 Eric 715 8348 Eric } 716 8348 Eric 717 8348 Eric /* 718 8348 Eric * This routine retrieves the current status of socket options. 719 8348 Eric * It returns the size of the option retrieved. 720 8348 Eric */ 721 8348 Eric int 722 8348 Eric rts_tpi_opt_get(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr) 723 8348 Eric { 724 8348 Eric rts_t *rts; 725 8348 Eric int err; 726 8348 Eric 727 8348 Eric rts = Q_TO_RTS(q); 728 8348 Eric rw_enter(&rts->rts_rwlock, RW_READER); 729 8348 Eric err = rts_opt_get(Q_TO_CONN(q), level, name, ptr); 730 8348 Eric rw_exit(&rts->rts_rwlock); 731 8348 Eric return (err); 732 8348 Eric } 733 8348 Eric 734 8348 Eric /* 735 8348 Eric * This routine sets socket options. 736 8348 Eric */ 737 8348 Eric /*ARGSUSED*/ 738 8348 Eric int 739 8348 Eric rts_tpi_opt_set(queue_t *q, uint_t optset_context, int level, 740 8348 Eric int name, uint_t inlen, uchar_t *invalp, uint_t *outlenp, 741 11042 Erik uchar_t *outvalp, void *thisdg_attrs, cred_t *cr) 742 8348 Eric { 743 8348 Eric conn_t *connp = Q_TO_CONN(q); 744 8348 Eric int error; 745 8348 Eric rts_t *rts = connp->conn_rts; 746 8348 Eric 747 8348 Eric 748 8348 Eric rw_enter(&rts->rts_rwlock, RW_WRITER); 749 8348 Eric error = rts_opt_set(connp, optset_context, level, name, inlen, invalp, 750 8348 Eric outlenp, outvalp, thisdg_attrs, cr); 751 8348 Eric rw_exit(&rts->rts_rwlock); 752 8348 Eric return (error); 753 0 stevel } 754 0 stevel 755 0 stevel /* 756 0 stevel * This routine retrieves the value of an ND variable in a rtsparam_t 757 0 stevel * structure. It is called through nd_getset when a user reads the 758 0 stevel * variable. 759 0 stevel */ 760 0 stevel /* ARGSUSED */ 761 0 stevel static int 762 0 stevel rts_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr) 763 0 stevel { 764 0 stevel rtsparam_t *rtspa = (rtsparam_t *)cp; 765 0 stevel 766 0 stevel (void) mi_mpprintf(mp, "%u", rtspa->rts_param_value); 767 0 stevel return (0); 768 0 stevel } 769 0 stevel 770 0 stevel /* 771 0 stevel * Walk through the param array specified registering each element with the 772 0 stevel * named dispatch (ND) handler. 773 0 stevel */ 774 0 stevel static boolean_t 775 3448 dh155122 rts_param_register(IDP *ndp, rtsparam_t *rtspa, int cnt) 776 0 stevel { 777 0 stevel for (; cnt-- > 0; rtspa++) { 778 0 stevel if (rtspa->rts_param_name != NULL && rtspa->rts_param_name[0]) { 779 3448 dh155122 if (!nd_load(ndp, rtspa->rts_param_name, 780 0 stevel rts_param_get, rts_param_set, (caddr_t)rtspa)) { 781 3448 dh155122 nd_free(ndp); 782 0 stevel return (B_FALSE); 783 0 stevel } 784 0 stevel } 785 0 stevel } 786 0 stevel return (B_TRUE); 787 0 stevel } 788 0 stevel 789 0 stevel /* This routine sets an ND variable in a rtsparam_t structure. */ 790 0 stevel /* ARGSUSED */ 791 0 stevel static int 792 0 stevel rts_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr) 793 0 stevel { 794 0 stevel ulong_t new_value; 795 0 stevel rtsparam_t *rtspa = (rtsparam_t *)cp; 796 0 stevel 797 0 stevel /* 798 0 stevel * Fail the request if the new value does not lie within the 799 0 stevel * required bounds. 800 0 stevel */ 801 0 stevel if (ddi_strtoul(value, NULL, 10, &new_value) != 0 || 802 0 stevel new_value < rtspa->rts_param_min || 803 0 stevel new_value > rtspa->rts_param_max) { 804 0 stevel return (EINVAL); 805 0 stevel } 806 0 stevel 807 0 stevel /* Set the new value */ 808 0 stevel rtspa->rts_param_value = new_value; 809 0 stevel return (0); 810 0 stevel } 811 0 stevel 812 0 stevel /* 813 5240 nordmark * Empty rsrv routine which is used by rts_input to cause a wakeup 814 5240 nordmark * of a thread in qwait. 815 5240 nordmark */ 816 5240 nordmark /*ARGSUSED*/ 817 5240 nordmark static void 818 5240 nordmark rts_rsrv(queue_t *q) 819 5240 nordmark { 820 5240 nordmark } 821 5240 nordmark 822 5240 nordmark /* 823 0 stevel * This routine handles synchronous messages passed downstream. It either 824 0 stevel * consumes the message or passes it downstream; it never queues a 825 0 stevel * a message. The data messages that go down are wrapped in an IOCTL 826 0 stevel * message. 827 0 stevel * 828 0 stevel * Since it is synchronous, it waits for the M_IOCACK/M_IOCNAK so that 829 0 stevel * it can return an immediate error (such as ENETUNREACH when adding a route). 830 0 stevel * It uses the RTS_WRW_PENDING to ensure that each rts instance has only 831 0 stevel * one M_IOCTL outstanding at any given time. 832 0 stevel */ 833 0 stevel static int 834 0 stevel rts_wrw(queue_t *q, struiod_t *dp) 835 0 stevel { 836 0 stevel mblk_t *mp = dp->d_mp; 837 0 stevel mblk_t *mp1; 838 0 stevel int error; 839 0 stevel rt_msghdr_t *rtm; 840 5240 nordmark conn_t *connp = Q_TO_CONN(q); 841 5240 nordmark rts_t *rts = connp->conn_rts; 842 0 stevel 843 0 stevel while (rts->rts_flag & RTS_WRW_PENDING) { 844 0 stevel if (qwait_rw(q)) { 845 0 stevel rts->rts_error = EINTR; 846 0 stevel goto err_ret; 847 0 stevel } 848 8348 Eric } 849 0 stevel rts->rts_flag |= RTS_WRW_PENDING; 850 0 stevel 851 0 stevel if (isuioq(q) && (error = struioget(q, mp, dp, 0))) { 852 0 stevel /* 853 0 stevel * Uio error of some sort, so just return the error. 854 0 stevel */ 855 0 stevel rts->rts_error = error; 856 0 stevel goto err_ret; 857 0 stevel } 858 0 stevel /* 859 0 stevel * Pass the mblk (chain) onto wput(). 860 0 stevel */ 861 0 stevel dp->d_mp = 0; 862 0 stevel 863 0 stevel switch (mp->b_datap->db_type) { 864 0 stevel case M_PROTO: 865 0 stevel case M_PCPROTO: 866 0 stevel /* Expedite other than T_DATA_REQ to below the switch */ 867 0 stevel if (((mp->b_wptr - mp->b_rptr) != 868 0 stevel sizeof (struct T_data_req)) || 869 0 stevel (((union T_primitives *)mp->b_rptr)->type != T_DATA_REQ)) 870 0 stevel break; 871 0 stevel if ((mp1 = mp->b_cont) == NULL) { 872 0 stevel rts->rts_error = EINVAL; 873 8752 Peter freemsg(mp); 874 0 stevel goto err_ret; 875 0 stevel } 876 0 stevel freeb(mp); 877 0 stevel mp = mp1; 878 0 stevel /* FALLTHRU */ 879 0 stevel case M_DATA: 880 0 stevel /* 881 0 stevel * The semantics of the routing socket is such that the rtm_pid 882 0 stevel * field is automatically filled in during requests with the 883 0 stevel * current process' pid. We do this here (where we still have 884 0 stevel * user context) after checking we have at least a message the 885 0 stevel * size of a routing message header. 886 0 stevel */ 887 0 stevel if ((mp->b_wptr - mp->b_rptr) < sizeof (rt_msghdr_t)) { 888 0 stevel if (!pullupmsg(mp, sizeof (rt_msghdr_t))) { 889 0 stevel rts->rts_error = EINVAL; 890 8752 Peter freemsg(mp); 891 0 stevel goto err_ret; 892 0 stevel } 893 0 stevel } 894 0 stevel rtm = (rt_msghdr_t *)mp->b_rptr; 895 0 stevel rtm->rtm_pid = curproc->p_pid; 896 0 stevel break; 897 0 stevel default: 898 0 stevel break; 899 0 stevel } 900 0 stevel rts->rts_flag |= RTS_WPUT_PENDING; 901 0 stevel rts_wput(q, mp); 902 0 stevel while (rts->rts_flag & RTS_WPUT_PENDING) 903 0 stevel if (qwait_rw(q)) { 904 0 stevel /* RTS_WPUT_PENDING will be cleared below */ 905 0 stevel rts->rts_error = EINTR; 906 0 stevel break; 907 0 stevel } 908 0 stevel err_ret: 909 0 stevel rts->rts_flag &= ~(RTS_WPUT_PENDING | RTS_WRW_PENDING); 910 0 stevel return (rts->rts_error); 911 0 stevel } 912 0 stevel 913 0 stevel /* 914 0 stevel * This routine handles all messages passed downstream. It either 915 0 stevel * consumes the message or passes it downstream; it never queues a 916 0 stevel * a message. The data messages that go down are wrapped in an IOCTL 917 0 stevel * message. 918 0 stevel */ 919 0 stevel static void 920 0 stevel rts_wput(queue_t *q, mblk_t *mp) 921 0 stevel { 922 0 stevel uchar_t *rptr = mp->b_rptr; 923 0 stevel mblk_t *mp1; 924 5240 nordmark conn_t *connp = Q_TO_CONN(q); 925 5240 nordmark rts_t *rts = connp->conn_rts; 926 0 stevel 927 0 stevel switch (mp->b_datap->db_type) { 928 0 stevel case M_DATA: 929 0 stevel break; 930 0 stevel case M_PROTO: 931 0 stevel case M_PCPROTO: 932 0 stevel if ((mp->b_wptr - rptr) == sizeof (struct T_data_req)) { 933 0 stevel /* Expedite valid T_DATA_REQ to below the switch */ 934 0 stevel if (((union T_primitives *)rptr)->type == T_DATA_REQ) { 935 0 stevel mp1 = mp->b_cont; 936 0 stevel freeb(mp); 937 0 stevel if (mp1 == NULL) 938 0 stevel return; 939 0 stevel mp = mp1; 940 0 stevel break; 941 0 stevel } 942 0 stevel } 943 0 stevel /* FALLTHRU */ 944 0 stevel default: 945 0 stevel rts_wput_other(q, mp); 946 0 stevel return; 947 0 stevel } 948 0 stevel 949 0 stevel 950 8778 Erik ASSERT(msg_getcred(mp, NULL) != NULL); 951 8778 Erik 952 8778 Erik mp1 = rts_ioctl_alloc(mp); 953 0 stevel if (mp1 == NULL) { 954 0 stevel ASSERT(rts != NULL); 955 0 stevel freemsg(mp); 956 0 stevel if (rts->rts_flag & RTS_WPUT_PENDING) { 957 0 stevel rts->rts_error = ENOMEM; 958 0 stevel rts->rts_flag &= ~RTS_WPUT_PENDING; 959 0 stevel } 960 0 stevel return; 961 0 stevel } 962 11042 Erik ip_wput_nondata(q, mp1); 963 0 stevel } 964 0 stevel 965 0 stevel 966 0 stevel /* 967 0 stevel * Handles all the control message, if it 968 0 stevel * can not understand it, it will 969 0 stevel * pass down stream. 970 0 stevel */ 971 0 stevel static void 972 0 stevel rts_wput_other(queue_t *q, mblk_t *mp) 973 0 stevel { 974 5240 nordmark conn_t *connp = Q_TO_CONN(q); 975 5240 nordmark rts_t *rts = connp->conn_rts; 976 0 stevel uchar_t *rptr = mp->b_rptr; 977 0 stevel struct iocblk *iocp; 978 0 stevel cred_t *cr; 979 3448 dh155122 rts_stack_t *rtss; 980 0 stevel 981 3448 dh155122 rtss = rts->rts_rtss; 982 0 stevel 983 0 stevel switch (mp->b_datap->db_type) { 984 0 stevel case M_PROTO: 985 0 stevel case M_PCPROTO: 986 0 stevel if ((mp->b_wptr - rptr) < sizeof (t_scalar_t)) { 987 0 stevel /* 988 0 stevel * If the message does not contain a PRIM_type, 989 0 stevel * throw it away. 990 0 stevel */ 991 0 stevel freemsg(mp); 992 0 stevel return; 993 0 stevel } 994 0 stevel switch (((union T_primitives *)rptr)->type) { 995 0 stevel case T_BIND_REQ: 996 0 stevel case O_T_BIND_REQ: 997 8348 Eric rts_tpi_bind(q, mp); 998 0 stevel return; 999 0 stevel case T_UNBIND_REQ: 1000 8348 Eric rts_tpi_unbind(q, mp); 1001 0 stevel return; 1002 0 stevel case T_CAPABILITY_REQ: 1003 0 stevel rts_capability_req(q, mp); 1004 0 stevel return; 1005 0 stevel case T_INFO_REQ: 1006 0 stevel rts_info_req(q, mp); 1007 0 stevel return; 1008 0 stevel case T_SVR4_OPTMGMT_REQ: 1009 0 stevel case T_OPTMGMT_REQ: 1010 8778 Erik /* 1011 8778 Erik * All Solaris components should pass a db_credp 1012 8778 Erik * for this TPI message, hence we ASSERT. 1013 8778 Erik * But in case there is some other M_PROTO that looks 1014 8778 Erik * like a TPI message sent by some other kernel 1015 8778 Erik * component, we check and return an error. 1016 8778 Erik */ 1017 8778 Erik cr = msg_getcred(mp, NULL); 1018 8778 Erik ASSERT(cr != NULL); 1019 8778 Erik if (cr == NULL) { 1020 8778 Erik rts_err_ack(q, mp, TSYSERR, EINVAL); 1021 8778 Erik return; 1022 8778 Erik } 1023 8778 Erik if (((union T_primitives *)rptr)->type == 1024 8778 Erik T_SVR4_OPTMGMT_REQ) { 1025 11042 Erik svr4_optcom_req(q, mp, cr, &rts_opt_obj); 1026 8778 Erik } else { 1027 11042 Erik tpi_optcom_req(q, mp, cr, &rts_opt_obj); 1028 8778 Erik } 1029 0 stevel return; 1030 0 stevel case O_T_CONN_RES: 1031 0 stevel case T_CONN_RES: 1032 0 stevel case T_DISCON_REQ: 1033 0 stevel /* Not supported by rts. */ 1034 0 stevel rts_err_ack(q, mp, TNOTSUPPORT, 0); 1035 0 stevel return; 1036 0 stevel case T_DATA_REQ: 1037 0 stevel case T_EXDATA_REQ: 1038 0 stevel case T_ORDREL_REQ: 1039 0 stevel /* Illegal for rts. */ 1040 0 stevel freemsg(mp); 1041 0 stevel (void) putnextctl1(RD(q), M_ERROR, EPROTO); 1042 0 stevel return; 1043 8348 Eric 1044 0 stevel default: 1045 0 stevel break; 1046 0 stevel } 1047 0 stevel break; 1048 0 stevel case M_IOCTL: 1049 0 stevel iocp = (struct iocblk *)mp->b_rptr; 1050 0 stevel switch (iocp->ioc_cmd) { 1051 0 stevel case ND_SET: 1052 0 stevel case ND_GET: 1053 3448 dh155122 if (nd_getset(q, rtss->rtss_g_nd, mp)) { 1054 0 stevel qreply(q, mp); 1055 0 stevel return; 1056 0 stevel } 1057 0 stevel break; 1058 0 stevel case TI_GETPEERNAME: 1059 0 stevel mi_copyin(q, mp, NULL, 1060 0 stevel SIZEOF_STRUCT(strbuf, iocp->ioc_flag)); 1061 0 stevel return; 1062 0 stevel default: 1063 0 stevel break; 1064 0 stevel } 1065 0 stevel case M_IOCDATA: 1066 0 stevel rts_wput_iocdata(q, mp); 1067 0 stevel return; 1068 0 stevel default: 1069 0 stevel break; 1070 0 stevel } 1071 11042 Erik ip_wput_nondata(q, mp); 1072 0 stevel } 1073 0 stevel 1074 0 stevel /* 1075 0 stevel * Called by rts_wput_other to handle all M_IOCDATA messages. 1076 0 stevel */ 1077 0 stevel static void 1078 0 stevel rts_wput_iocdata(queue_t *q, mblk_t *mp) 1079 0 stevel { 1080 0 stevel struct sockaddr *rtsaddr; 1081 0 stevel mblk_t *mp1; 1082 0 stevel STRUCT_HANDLE(strbuf, sb); 1083 0 stevel struct iocblk *iocp = (struct iocblk *)mp->b_rptr; 1084 0 stevel 1085 0 stevel /* Make sure it is one of ours. */ 1086 0 stevel switch (iocp->ioc_cmd) { 1087 0 stevel case TI_GETPEERNAME: 1088 0 stevel break; 1089 0 stevel default: 1090 11042 Erik ip_wput_nondata(q, mp); 1091 0 stevel return; 1092 0 stevel } 1093 0 stevel switch (mi_copy_state(q, mp, &mp1)) { 1094 0 stevel case -1: 1095 0 stevel return; 1096 0 stevel case MI_COPY_CASE(MI_COPY_IN, 1): 1097 0 stevel break; 1098 0 stevel case MI_COPY_CASE(MI_COPY_OUT, 1): 1099 0 stevel /* Copy out the strbuf. */ 1100 0 stevel mi_copyout(q, mp); 1101 0 stevel return; 1102 0 stevel case MI_COPY_CASE(MI_COPY_OUT, 2): 1103 0 stevel /* All done. */ 1104 0 stevel mi_copy_done(q, mp, 0); 1105 0 stevel return; 1106 0 stevel default: 1107 0 stevel mi_copy_done(q, mp, EPROTO); 1108 0 stevel return; 1109 0 stevel } 1110 0 stevel STRUCT_SET_HANDLE(sb, iocp->ioc_flag, (void *)mp1->b_rptr); 1111 0 stevel if (STRUCT_FGET(sb, maxlen) < (int)sizeof (sin_t)) { 1112 0 stevel mi_copy_done(q, mp, EINVAL); 1113 0 stevel return; 1114 0 stevel } 1115 0 stevel switch (iocp->ioc_cmd) { 1116 0 stevel case TI_GETPEERNAME: 1117 0 stevel break; 1118 0 stevel default: 1119 0 stevel mi_copy_done(q, mp, EPROTO); 1120 0 stevel return; 1121 0 stevel } 1122 0 stevel mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), sizeof (sin_t), 1123 0 stevel B_TRUE); 1124 0 stevel if (mp1 == NULL) 1125 0 stevel return; 1126 0 stevel STRUCT_FSET(sb, len, (int)sizeof (sin_t)); 1127 0 stevel rtsaddr = (struct sockaddr *)mp1->b_rptr; 1128 0 stevel mp1->b_wptr = (uchar_t *)&rtsaddr[1]; 1129 0 stevel bzero(rtsaddr, sizeof (struct sockaddr)); 1130 0 stevel rtsaddr->sa_family = AF_ROUTE; 1131 0 stevel /* Copy out the address */ 1132 0 stevel mi_copyout(q, mp); 1133 0 stevel } 1134 0 stevel 1135 11042 Erik /* 1136 11042 Erik * IP passes up a NULL ira. 1137 11042 Erik */ 1138 5240 nordmark /*ARGSUSED2*/ 1139 0 stevel static void 1140 11042 Erik rts_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira) 1141 0 stevel { 1142 5240 nordmark conn_t *connp = (conn_t *)arg1; 1143 5240 nordmark rts_t *rts = connp->conn_rts; 1144 0 stevel struct iocblk *iocp; 1145 0 stevel mblk_t *mp1; 1146 0 stevel struct T_data_ind *tdi; 1147 8348 Eric int error; 1148 0 stevel 1149 0 stevel switch (mp->b_datap->db_type) { 1150 0 stevel case M_IOCACK: 1151 0 stevel case M_IOCNAK: 1152 0 stevel iocp = (struct iocblk *)mp->b_rptr; 1153 11042 Erik ASSERT(!IPCL_IS_NONSTR(connp)); 1154 11042 Erik if (rts->rts_flag & (RTS_WPUT_PENDING)) { 1155 11042 Erik rts->rts_flag &= ~RTS_WPUT_PENDING; 1156 0 stevel rts->rts_error = iocp->ioc_error; 1157 11042 Erik /* 1158 11042 Erik * Tell rts_wvw/qwait that we are done. 1159 11042 Erik * Note: there is no qwait_wakeup() we can use. 1160 11042 Erik */ 1161 11042 Erik qenable(connp->conn_rq); 1162 0 stevel freemsg(mp); 1163 0 stevel return; 1164 0 stevel } 1165 0 stevel break; 1166 0 stevel case M_DATA: 1167 0 stevel /* 1168 0 stevel * Prepend T_DATA_IND to prevent the stream head from 1169 0 stevel * consolidating multiple messages together. 1170 0 stevel * If the allocation fails just send up the M_DATA. 1171 0 stevel */ 1172 0 stevel mp1 = allocb(sizeof (*tdi), BPRI_MED); 1173 0 stevel if (mp1 != NULL) { 1174 0 stevel mp1->b_cont = mp; 1175 0 stevel mp = mp1; 1176 0 stevel 1177 0 stevel mp->b_datap->db_type = M_PROTO; 1178 0 stevel mp->b_wptr += sizeof (*tdi); 1179 0 stevel tdi = (struct T_data_ind *)mp->b_rptr; 1180 0 stevel tdi->PRIM_type = T_DATA_IND; 1181 0 stevel tdi->MORE_flag = 0; 1182 0 stevel } 1183 0 stevel break; 1184 0 stevel default: 1185 0 stevel break; 1186 0 stevel } 1187 8348 Eric 1188 8348 Eric if (IPCL_IS_NONSTR(connp)) { 1189 8348 Eric if ((*connp->conn_upcalls->su_recv) 1190 8348 Eric (connp->conn_upper_handle, mp, msgdsize(mp), 0, 1191 8348 Eric &error, NULL) < 0) { 1192 8348 Eric ASSERT(error == ENOSPC); 1193 8348 Eric /* 1194 8348 Eric * Let's confirm hoding the lock that 1195 8348 Eric * we are out of recv space. 1196 8348 Eric */ 1197 8348 Eric mutex_enter(&rts->rts_recv_mutex); 1198 8348 Eric if ((*connp->conn_upcalls->su_recv) 1199 8348 Eric (connp->conn_upper_handle, NULL, 0, 0, 1200 8348 Eric &error, NULL) < 0) { 1201 8348 Eric ASSERT(error == ENOSPC); 1202 8348 Eric connp->conn_flow_cntrld = B_TRUE; 1203 8348 Eric } 1204 8348 Eric mutex_exit(&rts->rts_recv_mutex); 1205 8348 Eric } 1206 8348 Eric } else { 1207 8348 Eric putnext(connp->conn_rq, mp); 1208 8348 Eric } 1209 0 stevel } 1210 0 stevel 1211 11042 Erik /*ARGSUSED*/ 1212 11042 Erik static void 1213 11042 Erik rts_icmp_input(void *arg1, mblk_t *mp, void *arg2, ip_recv_attr_t *ira) 1214 11042 Erik { 1215 11042 Erik freemsg(mp); 1216 11042 Erik } 1217 0 stevel 1218 0 stevel void 1219 8348 Eric rts_ddi_g_init(void) 1220 0 stevel { 1221 0 stevel rts_max_optsize = optcom_max_optsize(rts_opt_obj.odb_opt_des_arr, 1222 0 stevel rts_opt_obj.odb_opt_arr_cnt); 1223 3448 dh155122 1224 3448 dh155122 /* 1225 3448 dh155122 * We want to be informed each time a stack is created or 1226 3448 dh155122 * destroyed in the kernel, so we can maintain the 1227 3448 dh155122 * set of rts_stack_t's. 1228 3448 dh155122 */ 1229 3448 dh155122 netstack_register(NS_RTS, rts_stack_init, NULL, rts_stack_fini); 1230 0 stevel } 1231 3448 dh155122 1232 3448 dh155122 void 1233 8348 Eric rts_ddi_g_destroy(void) 1234 3448 dh155122 { 1235 3448 dh155122 netstack_unregister(NS_RTS); 1236 3448 dh155122 } 1237 8348 Eric 1238 8348 Eric #define INET_NAME "ip" 1239 3448 dh155122 1240 3448 dh155122 /* 1241 3448 dh155122 * Initialize the RTS stack instance. 1242 3448 dh155122 */ 1243 3448 dh155122 /* ARGSUSED */ 1244 3448 dh155122 static void * 1245 3448 dh155122 rts_stack_init(netstackid_t stackid, netstack_t *ns) 1246 3448 dh155122 { 1247 3448 dh155122 rts_stack_t *rtss; 1248 3448 dh155122 rtsparam_t *pa; 1249 8348 Eric int error = 0; 1250 8348 Eric major_t major; 1251 3448 dh155122 1252 3448 dh155122 rtss = (rts_stack_t *)kmem_zalloc(sizeof (*rtss), KM_SLEEP); 1253 3448 dh155122 rtss->rtss_netstack = ns; 1254 3448 dh155122 1255 3448 dh155122 pa = (rtsparam_t *)kmem_alloc(sizeof (lcl_param_arr), KM_SLEEP); 1256 3448 dh155122 rtss->rtss_params = pa; 1257 3448 dh155122 bcopy(lcl_param_arr, rtss->rtss_params, sizeof (lcl_param_arr)); 1258 3448 dh155122 1259 3448 dh155122 (void) rts_param_register(&rtss->rtss_g_nd, 1260 3448 dh155122 rtss->rtss_params, A_CNT(lcl_param_arr)); 1261 8348 Eric 1262 8348 Eric major = mod_name_to_major(INET_NAME); 1263 8348 Eric error = ldi_ident_from_major(major, &rtss->rtss_ldi_ident); 1264 8348 Eric ASSERT(error == 0); 1265 3448 dh155122 return (rtss); 1266 3448 dh155122 } 1267 3448 dh155122 1268 3448 dh155122 /* 1269 3448 dh155122 * Free the RTS stack instance. 1270 3448 dh155122 */ 1271 3448 dh155122 /* ARGSUSED */ 1272 3448 dh155122 static void 1273 3448 dh155122 rts_stack_fini(netstackid_t stackid, void *arg) 1274 3448 dh155122 { 1275 3448 dh155122 rts_stack_t *rtss = (rts_stack_t *)arg; 1276 3448 dh155122 1277 5240 nordmark nd_free(&rtss->rtss_g_nd); 1278 3448 dh155122 kmem_free(rtss->rtss_params, sizeof (lcl_param_arr)); 1279 3448 dh155122 rtss->rtss_params = NULL; 1280 8348 Eric ldi_ident_release(rtss->rtss_ldi_ident); 1281 3448 dh155122 kmem_free(rtss, sizeof (*rtss)); 1282 3448 dh155122 } 1283 8348 Eric 1284 8348 Eric /* ARGSUSED */ 1285 8348 Eric int 1286 8348 Eric rts_accept(sock_lower_handle_t lproto_handle, 1287 8348 Eric sock_lower_handle_t eproto_handle, sock_upper_handle_t sock_handle, 1288 8348 Eric cred_t *cr) 1289 8348 Eric { 1290 8348 Eric return (EINVAL); 1291 8348 Eric } 1292 8348 Eric 1293 8348 Eric /* ARGSUSED */ 1294 8348 Eric static int 1295 8348 Eric rts_bind(sock_lower_handle_t proto_handle, struct sockaddr *sa, 1296 8348 Eric socklen_t len, cred_t *cr) 1297 8348 Eric { 1298 8348 Eric /* 1299 8348 Eric * rebind not allowed 1300 8348 Eric */ 1301 8348 Eric return (EINVAL); 1302 8348 Eric } 1303 8348 Eric 1304 8348 Eric /* ARGSUSED */ 1305 8348 Eric int 1306 8348 Eric rts_listen(sock_lower_handle_t proto_handle, int backlog, cred_t *cr) 1307 8348 Eric { 1308 8348 Eric return (EINVAL); 1309 8348 Eric } 1310 8348 Eric 1311 8348 Eric /* ARGSUSED */ 1312 8348 Eric int 1313 8348 Eric rts_connect(sock_lower_handle_t proto_handle, const struct sockaddr *sa, 1314 8348 Eric socklen_t len, sock_connid_t *id, cred_t *cr) 1315 8348 Eric { 1316 8348 Eric /* 1317 8348 Eric * rts sockets start out as bound and connected 1318 8348 Eric */ 1319 8348 Eric *id = 0; 1320 8348 Eric return (EISCONN); 1321 8348 Eric } 1322 8348 Eric 1323 8348 Eric /* ARGSUSED */ 1324 8348 Eric int 1325 8348 Eric rts_getpeername(sock_lower_handle_t proto_handle, struct sockaddr *addr, 1326 8348 Eric socklen_t *addrlen, cred_t *cr) 1327 8348 Eric { 1328 8348 Eric bzero(addr, sizeof (struct sockaddr)); 1329 8348 Eric addr->sa_family = AF_ROUTE; 1330 8348 Eric *addrlen = sizeof (struct sockaddr); 1331 8348 Eric 1332 8348 Eric return (0); 1333 8348 Eric } 1334 8348 Eric 1335 8348 Eric /* ARGSUSED */ 1336 8348 Eric int 1337 8348 Eric rts_getsockname(sock_lower_handle_t proto_handle, struct sockaddr *addr, 1338 8348 Eric socklen_t *addrlen, cred_t *cr) 1339 8348 Eric { 1340 11042 Erik bzero(addr, sizeof (struct sockaddr)); 1341 11042 Erik addr->sa_family = AF_ROUTE; 1342 11042 Erik *addrlen = sizeof (struct sockaddr); 1343 11042 Erik 1344 11042 Erik return (0); 1345 8348 Eric } 1346 8348 Eric 1347 8348 Eric static int 1348 8348 Eric rts_getsockopt(sock_lower_handle_t proto_handle, int level, int option_name, 1349 8348 Eric void *optvalp, socklen_t *optlen, cred_t *cr) 1350 8348 Eric { 1351 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1352 8348 Eric rts_t *rts = connp->conn_rts; 1353 8348 Eric int error; 1354 8348 Eric t_uscalar_t max_optbuf_len; 1355 8348 Eric void *optvalp_buf; 1356 8348 Eric int len; 1357 8348 Eric 1358 8348 Eric error = proto_opt_check(level, option_name, *optlen, &max_optbuf_len, 1359 8348 Eric rts_opt_obj.odb_opt_des_arr, 1360 8348 Eric rts_opt_obj.odb_opt_arr_cnt, 1361 8348 Eric B_FALSE, B_TRUE, cr); 1362 8348 Eric if (error != 0) { 1363 8348 Eric if (error < 0) 1364 8348 Eric error = proto_tlitosyserr(-error); 1365 8348 Eric return (error); 1366 8348 Eric } 1367 8348 Eric 1368 8348 Eric optvalp_buf = kmem_alloc(max_optbuf_len, KM_SLEEP); 1369 8348 Eric rw_enter(&rts->rts_rwlock, RW_READER); 1370 8348 Eric len = rts_opt_get(connp, level, option_name, optvalp_buf); 1371 8348 Eric rw_exit(&rts->rts_rwlock); 1372 11042 Erik if (len == -1) { 1373 11042 Erik kmem_free(optvalp_buf, max_optbuf_len); 1374 11042 Erik return (EINVAL); 1375 8348 Eric } 1376 8348 Eric 1377 11042 Erik /* 1378 11042 Erik * update optlen and copy option value 1379 11042 Erik */ 1380 11042 Erik t_uscalar_t size = MIN(len, *optlen); 1381 11042 Erik 1382 11042 Erik bcopy(optvalp_buf, optvalp, size); 1383 11042 Erik bcopy(&size, optlen, sizeof (size)); 1384 8348 Eric kmem_free(optvalp_buf, max_optbuf_len); 1385 11042 Erik return (0); 1386 8348 Eric } 1387 8348 Eric 1388 8348 Eric static int 1389 8348 Eric rts_setsockopt(sock_lower_handle_t proto_handle, int level, int option_name, 1390 8348 Eric const void *optvalp, socklen_t optlen, cred_t *cr) 1391 8348 Eric { 1392 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1393 8348 Eric rts_t *rts = connp->conn_rts; 1394 8348 Eric int error; 1395 8348 Eric 1396 8348 Eric error = proto_opt_check(level, option_name, optlen, NULL, 1397 8348 Eric rts_opt_obj.odb_opt_des_arr, 1398 8348 Eric rts_opt_obj.odb_opt_arr_cnt, 1399 8348 Eric B_TRUE, B_FALSE, cr); 1400 8348 Eric 1401 8348 Eric if (error != 0) { 1402 8348 Eric if (error < 0) 1403 8348 Eric error = proto_tlitosyserr(-error); 1404 8348 Eric return (error); 1405 8348 Eric } 1406 8348 Eric 1407 8348 Eric rw_enter(&rts->rts_rwlock, RW_WRITER); 1408 8348 Eric error = rts_opt_set(connp, SETFN_OPTCOM_NEGOTIATE, level, option_name, 1409 8348 Eric optlen, (uchar_t *)optvalp, (uint_t *)&optlen, (uchar_t *)optvalp, 1410 8348 Eric NULL, cr); 1411 8348 Eric rw_exit(&rts->rts_rwlock); 1412 8348 Eric 1413 8348 Eric ASSERT(error >= 0); 1414 8348 Eric 1415 8348 Eric return (error); 1416 8348 Eric } 1417 8348 Eric 1418 8348 Eric /* ARGSUSED */ 1419 8348 Eric static int 1420 8348 Eric rts_send(sock_lower_handle_t proto_handle, mblk_t *mp, 1421 8348 Eric struct nmsghdr *msg, cred_t *cr) 1422 8348 Eric { 1423 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1424 8348 Eric rt_msghdr_t *rtm; 1425 8348 Eric int error; 1426 8348 Eric 1427 8348 Eric ASSERT(DB_TYPE(mp) == M_DATA); 1428 8348 Eric /* 1429 8348 Eric * The semantics of the routing socket is such that the rtm_pid 1430 8348 Eric * field is automatically filled in during requests with the 1431 8348 Eric * current process' pid. We do this here (where we still have 1432 8348 Eric * user context) after checking we have at least a message the 1433 8348 Eric * size of a routing message header. 1434 8348 Eric */ 1435 8348 Eric if ((mp->b_wptr - mp->b_rptr) < sizeof (rt_msghdr_t)) { 1436 8348 Eric if (!pullupmsg(mp, sizeof (rt_msghdr_t))) { 1437 8348 Eric freemsg(mp); 1438 11042 Erik return (EINVAL); 1439 8348 Eric } 1440 8348 Eric } 1441 8348 Eric rtm = (rt_msghdr_t *)mp->b_rptr; 1442 8348 Eric rtm->rtm_pid = curproc->p_pid; 1443 8348 Eric 1444 8348 Eric /* 1445 11042 Erik * We are not constrained by the ioctl interface and 1446 11042 Erik * ip_rts_request_common processing requests synchronously hence 1447 11042 Erik * we can send them down concurrently. 1448 8348 Eric */ 1449 11042 Erik error = ip_rts_request_common(mp, connp, cr); 1450 8348 Eric return (error); 1451 8348 Eric } 1452 8348 Eric 1453 8348 Eric /* ARGSUSED */ 1454 8348 Eric sock_lower_handle_t 1455 8348 Eric rts_create(int family, int type, int proto, sock_downcalls_t **sock_downcalls, 1456 8348 Eric uint_t *smodep, int *errorp, int flags, cred_t *credp) 1457 8348 Eric { 1458 8348 Eric conn_t *connp; 1459 8348 Eric 1460 8348 Eric if (family != AF_ROUTE || type != SOCK_RAW || 1461 8348 Eric (proto != 0 && proto != AF_INET && proto != AF_INET6)) { 1462 8348 Eric *errorp = EPROTONOSUPPORT; 1463 8348 Eric return (NULL); 1464 8348 Eric } 1465 8348 Eric 1466 8348 Eric connp = rts_open(flags, credp); 1467 8348 Eric ASSERT(connp != NULL); 1468 8348 Eric connp->conn_flags |= IPCL_NONSTR; 1469 8348 Eric 1470 11042 Erik connp->conn_proto = proto; 1471 8348 Eric 1472 8348 Eric mutex_enter(&connp->conn_lock); 1473 8348 Eric connp->conn_state_flags &= ~CONN_INCIPIENT; 1474 8348 Eric mutex_exit(&connp->conn_lock); 1475 8348 Eric 1476 8348 Eric *errorp = 0; 1477 8348 Eric *smodep = SM_ATOMIC; 1478 8348 Eric *sock_downcalls = &sock_rts_downcalls; 1479 8348 Eric return ((sock_lower_handle_t)connp); 1480 8348 Eric } 1481 8348 Eric 1482 8348 Eric /* ARGSUSED */ 1483 8348 Eric void 1484 8348 Eric rts_activate(sock_lower_handle_t proto_handle, sock_upper_handle_t sock_handle, 1485 8348 Eric sock_upcalls_t *sock_upcalls, int flags, cred_t *cr) 1486 8348 Eric { 1487 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1488 8348 Eric struct sock_proto_props sopp; 1489 8348 Eric 1490 8348 Eric connp->conn_upcalls = sock_upcalls; 1491 8348 Eric connp->conn_upper_handle = sock_handle; 1492 8348 Eric 1493 8348 Eric sopp.sopp_flags = SOCKOPT_WROFF | SOCKOPT_RCVHIWAT | SOCKOPT_RCVLOWAT | 1494 8348 Eric SOCKOPT_MAXBLK | SOCKOPT_MAXPSZ | SOCKOPT_MINPSZ; 1495 8348 Eric sopp.sopp_wroff = 0; 1496 11042 Erik sopp.sopp_rxhiwat = connp->conn_rcvbuf; 1497 11042 Erik sopp.sopp_rxlowat = connp->conn_rcvlowat; 1498 8348 Eric sopp.sopp_maxblk = INFPSZ; 1499 8348 Eric sopp.sopp_maxpsz = rts_mod_info.mi_maxpsz; 1500 8348 Eric sopp.sopp_minpsz = (rts_mod_info.mi_minpsz == 1) ? 0 : 1501 8348 Eric rts_mod_info.mi_minpsz; 1502 8348 Eric 1503 8348 Eric (*connp->conn_upcalls->su_set_proto_props) 1504 8348 Eric (connp->conn_upper_handle, &sopp); 1505 8348 Eric 1506 8348 Eric /* 1507 8348 Eric * We treat it as already connected for routing socket. 1508 8348 Eric */ 1509 8348 Eric (*connp->conn_upcalls->su_connected) 1510 8348 Eric (connp->conn_upper_handle, 0, NULL, -1); 1511 8348 Eric 1512 11042 Erik /* Indicate to IP that this is a routing socket client */ 1513 8348 Eric ip_rts_register(connp); 1514 8348 Eric } 1515 8348 Eric 1516 8348 Eric /* ARGSUSED */ 1517 8348 Eric int 1518 8348 Eric rts_close(sock_lower_handle_t proto_handle, int flags, cred_t *cr) 1519 8348 Eric { 1520 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1521 8348 Eric 1522 8348 Eric ASSERT(connp != NULL && IPCL_IS_RTS(connp)); 1523 8348 Eric return (rts_common_close(NULL, connp)); 1524 8348 Eric } 1525 8348 Eric 1526 8348 Eric /* ARGSUSED */ 1527 8348 Eric int 1528 8348 Eric rts_shutdown(sock_lower_handle_t proto_handle, int how, cred_t *cr) 1529 8348 Eric { 1530 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1531 8348 Eric 1532 8348 Eric /* shut down the send side */ 1533 8348 Eric if (how != SHUT_RD) 1534 8348 Eric (*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle, 1535 8348 Eric SOCK_OPCTL_SHUT_SEND, 0); 1536 8348 Eric /* shut down the recv side */ 1537 8348 Eric if (how != SHUT_WR) 1538 8348 Eric (*connp->conn_upcalls->su_opctl)(connp->conn_upper_handle, 1539 8348 Eric SOCK_OPCTL_SHUT_RECV, 0); 1540 8348 Eric return (0); 1541 8348 Eric } 1542 8348 Eric 1543 8348 Eric void 1544 8348 Eric rts_clr_flowctrl(sock_lower_handle_t proto_handle) 1545 8348 Eric { 1546 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1547 8348 Eric rts_t *rts = connp->conn_rts; 1548 8348 Eric 1549 8348 Eric mutex_enter(&rts->rts_recv_mutex); 1550 8348 Eric connp->conn_flow_cntrld = B_FALSE; 1551 8348 Eric mutex_exit(&rts->rts_recv_mutex); 1552 8348 Eric } 1553 8348 Eric 1554 8348 Eric int 1555 8348 Eric rts_ioctl(sock_lower_handle_t proto_handle, int cmd, intptr_t arg, 1556 8348 Eric int mode, int32_t *rvalp, cred_t *cr) 1557 8348 Eric { 1558 8348 Eric conn_t *connp = (conn_t *)proto_handle; 1559 8348 Eric int error; 1560 8348 Eric 1561 11042 Erik /* 1562 11042 Erik * If we don't have a helper stream then create one. 1563 11042 Erik * ip_create_helper_stream takes care of locking the conn_t, 1564 11042 Erik * so this check for NULL is just a performance optimization. 1565 11042 Erik */ 1566 11042 Erik if (connp->conn_helper_info == NULL) { 1567 11042 Erik rts_stack_t *rtss = connp->conn_rts->rts_rtss; 1568 11042 Erik 1569 11042 Erik ASSERT(rtss->rtss_ldi_ident != NULL); 1570 11042 Erik 1571 11042 Erik /* 1572 11042 Erik * Create a helper stream for non-STREAMS socket. 1573 11042 Erik */ 1574 11042 Erik error = ip_create_helper_stream(connp, rtss->rtss_ldi_ident); 1575 11042 Erik if (error != 0) { 1576 11042 Erik ip0dbg(("rts_ioctl: create of IP helper stream " 1577 11042 Erik "failed %d\n", error)); 1578 11042 Erik return (error); 1579 11042 Erik } 1580 11042 Erik } 1581 11042 Erik 1582 8348 Eric switch (cmd) { 1583 8348 Eric case ND_SET: 1584 8348 Eric case ND_GET: 1585 8348 Eric case TI_GETPEERNAME: 1586 8348 Eric case TI_GETMYNAME: 1587 8348 Eric #ifdef DEUG 1588 8348 Eric cmn_err(CE_CONT, "rts_ioctl cmd 0x%x on non sreams" 1589 8348 Eric " socket", cmd); 1590 8348 Eric #endif 1591 8348 Eric error = EINVAL; 1592 8348 Eric break; 1593 8348 Eric default: 1594 8348 Eric /* 1595 8348 Eric * Pass on to IP using helper stream 1596 8348 Eric */ 1597 8444 Rao error = ldi_ioctl(connp->conn_helper_info->iphs_handle, 1598 8348 Eric cmd, arg, mode, cr, rvalp); 1599 8348 Eric break; 1600 8348 Eric } 1601 8348 Eric 1602 8348 Eric return (error); 1603 8348 Eric } 1604 8348 Eric 1605 8348 Eric sock_downcalls_t sock_rts_downcalls = { 1606 8348 Eric rts_activate, 1607 8348 Eric rts_accept, 1608 8348 Eric rts_bind, 1609 8348 Eric rts_listen, 1610 8348 Eric rts_connect, 1611 8348 Eric rts_getpeername, 1612 8348 Eric rts_getsockname, 1613 8348 Eric rts_getsockopt, 1614 8348 Eric rts_setsockopt, 1615 8348 Eric rts_send, 1616 8348 Eric NULL, 1617 8348 Eric NULL, 1618 8348 Eric NULL, 1619 8348 Eric rts_shutdown, 1620 8348 Eric rts_clr_flowctrl, 1621 8348 Eric rts_ioctl, 1622 8348 Eric rts_close 1623 8348 Eric }; 1624