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, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 /* Copyright (c) 1990 Mentat Inc. */ 27 28 29 #pragma ident "%Z%%M% %I% %E% SMI" 30 31 #include <sys/types.h> 32 #include <sys/stream.h> 33 #include <sys/stropts.h> 34 #include <sys/strlog.h> 35 #include <sys/strsun.h> 36 #define _SUN_TPI_VERSION 2 37 #include <sys/tihdr.h> 38 #include <sys/timod.h> 39 #include <sys/ddi.h> 40 #include <sys/sunddi.h> 41 #include <sys/cmn_err.h> 42 #include <sys/debug.h> 43 #include <sys/kmem.h> 44 #include <sys/policy.h> 45 #include <sys/zone.h> 46 47 #include <sys/socket.h> 48 #include <sys/isa_defs.h> 49 #include <sys/suntpi.h> 50 #include <sys/xti_inet.h> 51 52 #include <net/route.h> 53 #include <net/if.h> 54 55 #include <netinet/in.h> 56 #include <netinet/ip6.h> 57 #include <netinet/icmp6.h> 58 #include <inet/common.h> 59 #include <inet/ip.h> 60 #include <inet/ip6.h> 61 #include <inet/ip_ire.h> 62 #include <inet/mi.h> 63 #include <inet/nd.h> 64 #include <inet/optcom.h> 65 #include <inet/snmpcom.h> 66 #include <inet/kstatcom.h> 67 #include <inet/rawip_impl.h> 68 69 #include <netinet/ip_mroute.h> 70 #include <inet/tcp.h> 71 #include <net/pfkeyv2.h> 72 #include <inet/ipsec_info.h> 73 #include <inet/ipclassifier.h> 74 75 #define ICMP6 "icmp6" 76 major_t ICMP6_MAJ; 77 78 /* 79 * Object to represent database of options to search passed to 80 * {sock,tpi}optcom_req() interface routine to take care of option 81 * management and associated methods. 82 * XXX These and other extern's should really move to a icmp header. 83 */ 84 extern optdb_obj_t icmp_opt_obj; 85 extern uint_t icmp_max_optsize; 86 87 /* 88 * Synchronization notes: 89 * 90 * At all points in this code where exclusive access is required, we 91 * pass a message to a subroutine by invoking qwriter(..., PERIM_OUTER) 92 * which will arrange to call the routine only after all threads have 93 * exited the shared resource. 94 */ 95 96 /* Named Dispatch Parameter Management Structure */ 97 typedef struct icmpparam_s { 98 uint_t icmp_param_min; 99 uint_t icmp_param_max; 100 uint_t icmp_param_value; 101 char *icmp_param_name; 102 } icmpparam_t; 103 104 static void icmp_addr_req(queue_t *q, mblk_t *mp); 105 static void icmp_bind(queue_t *q, mblk_t *mp); 106 static void icmp_bind_proto(queue_t *q); 107 static int icmp_build_hdrs(queue_t *q, icmp_t *icmp); 108 static void icmp_capability_req(queue_t *q, mblk_t *mp); 109 static int icmp_close(queue_t *q); 110 static void icmp_connect(queue_t *q, mblk_t *mp); 111 static void icmp_disconnect(queue_t *q, mblk_t *mp); 112 static void icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, 113 int sys_error); 114 static void icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive, 115 t_scalar_t t_error, int sys_error); 116 static void icmp_icmp_error(queue_t *q, mblk_t *mp); 117 static void icmp_icmp_error_ipv6(queue_t *q, mblk_t *mp); 118 static void icmp_info_req(queue_t *q, mblk_t *mp); 119 static mblk_t *icmp_ip_bind_mp(icmp_t *icmp, t_scalar_t bind_prim, 120 t_scalar_t addr_length, in_port_t); 121 static int icmp_open(queue_t *q, dev_t *devp, int flag, 122 int sflag, cred_t *credp); 123 static int icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, 124 int *errorp, void *thisdg_attrs); 125 static boolean_t icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name); 126 int icmp_opt_set(queue_t *q, uint_t optset_context, 127 int level, int name, uint_t inlen, 128 uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, 129 void *thisdg_attrs, cred_t *cr, mblk_t *mblk); 130 int icmp_opt_get(queue_t *q, int level, int name, 131 uchar_t *ptr); 132 static int icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr); 133 static boolean_t icmp_param_register(icmpparam_t *icmppa, int cnt); 134 static int icmp_param_set(queue_t *q, mblk_t *mp, char *value, 135 caddr_t cp, cred_t *cr); 136 static int icmp_pkt_set(uchar_t *invalp, uint_t inlen, boolean_t sticky, 137 uchar_t **optbufp, uint_t *optlenp); 138 static void icmp_rput(queue_t *q, mblk_t *mp); 139 static void icmp_rput_bind_ack(queue_t *q, mblk_t *mp); 140 static int icmp_snmp_get(queue_t *q, mblk_t *mpctl); 141 static int icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name, 142 uchar_t *ptr, int len); 143 static int icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, 144 cred_t *cr); 145 static void icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err); 146 static void icmp_unbind(queue_t *q, mblk_t *mp); 147 static void icmp_wput(queue_t *q, mblk_t *mp); 148 static void icmp_wput_ipv6(queue_t *q, mblk_t *mp, sin6_t *sin6, 149 t_scalar_t tudr_optlen); 150 static void icmp_wput_other(queue_t *q, mblk_t *mp); 151 static void icmp_wput_iocdata(queue_t *q, mblk_t *mp); 152 static void icmp_wput_restricted(queue_t *q, mblk_t *mp); 153 154 static void rawip_kstat_init(void); 155 static void rawip_kstat_fini(void); 156 static int rawip_kstat_update(kstat_t *kp, int rw); 157 158 159 static struct module_info info = { 160 5707, "icmp", 1, INFPSZ, 512, 128 161 }; 162 163 static struct qinit rinit = { 164 (pfi_t)icmp_rput, NULL, icmp_open, icmp_close, NULL, &info 165 }; 166 167 static struct qinit winit = { 168 (pfi_t)icmp_wput, NULL, NULL, NULL, NULL, &info 169 }; 170 171 struct streamtab icmpinfo = { 172 &rinit, &winit 173 }; 174 175 static sin_t sin_null; /* Zero address for quick clears */ 176 static sin6_t sin6_null; /* Zero address for quick clears */ 177 static void *icmp_g_head; /* Head for list of open icmp streams. */ 178 static IDP icmp_g_nd; /* Points to table of ICMP ND variables. */ 179 180 /* MIB-2 stuff for SNMP */ 181 static mib2_rawip_t rawip_mib; /* SNMP fixed size info */ 182 static kstat_t *rawip_mibkp; /* kstat exporting rawip_mib data */ 183 184 /* Default structure copied into T_INFO_ACK messages */ 185 static struct T_info_ack icmp_g_t_info_ack = { 186 T_INFO_ACK, 187 IP_MAXPACKET, /* TSDU_size. icmp allows maximum size messages. */ 188 T_INVALID, /* ETSDU_size. icmp does not support expedited data. */ 189 T_INVALID, /* CDATA_size. icmp does not support connect data. */ 190 T_INVALID, /* DDATA_size. icmp does not support disconnect data. */ 191 0, /* ADDR_size - filled in later. */ 192 0, /* OPT_size - not initialized here */ 193 IP_MAXPACKET, /* TIDU_size. icmp allows maximum size messages. */ 194 T_CLTS, /* SERV_type. icmp supports connection-less. */ 195 TS_UNBND, /* CURRENT_state. This is set from icmp_state. */ 196 (XPG4_1|SENDZERO) /* PROVIDER_flag */ 197 }; 198 199 /* 200 * Table of ND variables supported by icmp. These are loaded into icmp_g_nd 201 * in icmp_open. 202 * All of these are alterable, within the min/max values given, at run time. 203 */ 204 static icmpparam_t icmp_param_arr[] = { 205 /* min max value name */ 206 { 0, 128, 32, "icmp_wroff_extra" }, 207 { 1, 255, 255, "icmp_ipv4_ttl" }, 208 { 0, IPV6_MAX_HOPS, IPV6_DEFAULT_HOPS, "icmp_ipv6_hoplimit"}, 209 { 0, 1, 1, "icmp_bsd_compat" }, 210 { 4096, 65536, 8192, "icmp_xmit_hiwat"}, 211 { 0, 65536, 1024, "icmp_xmit_lowat"}, 212 { 4096, 65536, 8192, "icmp_recv_hiwat"}, 213 { 65536, 1024*1024*1024, 256*1024, "icmp_max_buf"}, 214 }; 215 #define icmp_wroff_extra icmp_param_arr[0].icmp_param_value 216 #define icmp_ipv4_ttl icmp_param_arr[1].icmp_param_value 217 #define icmp_ipv6_hoplimit icmp_param_arr[2].icmp_param_value 218 #define icmp_bsd_compat icmp_param_arr[3].icmp_param_value 219 #define icmp_xmit_hiwat icmp_param_arr[4].icmp_param_value 220 #define icmp_xmit_lowat icmp_param_arr[5].icmp_param_value 221 #define icmp_recv_hiwat icmp_param_arr[6].icmp_param_value 222 #define icmp_max_buf icmp_param_arr[7].icmp_param_value 223 224 /* 225 * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message 226 * passed to icmp_wput. 227 * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the ICMP 228 * protocol type placed in the message following the address. A T_BIND_ACK 229 * message is passed upstream when ip acknowledges the request. 230 * (Called as writer.) 231 */ 232 static void 233 icmp_bind(queue_t *q, mblk_t *mp) 234 { 235 sin_t *sin; 236 sin6_t *sin6; 237 mblk_t *mp1; 238 struct T_bind_req *tbr; 239 icmp_t *icmp; 240 241 icmp = (icmp_t *)q->q_ptr; 242 if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) { 243 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 244 "icmp_bind: bad req, len %u", 245 (uint_t)(mp->b_wptr - mp->b_rptr)); 246 icmp_err_ack(q, mp, TPROTO, 0); 247 return; 248 } 249 if (icmp->icmp_state != TS_UNBND) { 250 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 251 "icmp_bind: bad state, %d", icmp->icmp_state); 252 icmp_err_ack(q, mp, TOUTSTATE, 0); 253 return; 254 } 255 /* 256 * Reallocate the message to make sure we have enough room for an 257 * address and the protocol type. 258 */ 259 mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1); 260 if (!mp1) { 261 icmp_err_ack(q, mp, TSYSERR, ENOMEM); 262 return; 263 } 264 mp = mp1; 265 tbr = (struct T_bind_req *)mp->b_rptr; 266 switch (tbr->ADDR_length) { 267 case 0: /* Generic request */ 268 tbr->ADDR_offset = sizeof (struct T_bind_req); 269 if (icmp->icmp_family == AF_INET) { 270 tbr->ADDR_length = sizeof (sin_t); 271 sin = (sin_t *)&tbr[1]; 272 *sin = sin_null; 273 sin->sin_family = AF_INET; 274 mp->b_wptr = (uchar_t *)&sin[1]; 275 } else { 276 ASSERT(icmp->icmp_family == AF_INET6); 277 tbr->ADDR_length = sizeof (sin6_t); 278 sin6 = (sin6_t *)&tbr[1]; 279 *sin6 = sin6_null; 280 sin6->sin6_family = AF_INET6; 281 mp->b_wptr = (uchar_t *)&sin6[1]; 282 } 283 break; 284 case sizeof (sin_t): /* Complete IP address */ 285 sin = (sin_t *)mi_offset_param(mp, tbr->ADDR_offset, 286 sizeof (sin_t)); 287 if (sin == NULL || !OK_32PTR((char *)sin)) { 288 icmp_err_ack(q, mp, TSYSERR, EINVAL); 289 return; 290 } 291 if (icmp->icmp_family != AF_INET || 292 sin->sin_family != AF_INET) { 293 icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 294 return; 295 } 296 break; 297 case sizeof (sin6_t): /* Complete IP address */ 298 sin6 = (sin6_t *)mi_offset_param(mp, tbr->ADDR_offset, 299 sizeof (sin6_t)); 300 if (sin6 == NULL || !OK_32PTR((char *)sin6)) { 301 icmp_err_ack(q, mp, TSYSERR, EINVAL); 302 return; 303 } 304 if (icmp->icmp_family != AF_INET6 || 305 sin6->sin6_family != AF_INET6) { 306 icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 307 return; 308 } 309 /* No support for mapped addresses on raw sockets */ 310 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 311 icmp_err_ack(q, mp, TSYSERR, EADDRNOTAVAIL); 312 return; 313 } 314 break; 315 default: 316 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 317 "icmp_bind: bad ADDR_length %d", tbr->ADDR_length); 318 icmp_err_ack(q, mp, TBADADDR, 0); 319 return; 320 } 321 /* 322 * Copy the source address into our icmp structure. This address 323 * may still be zero; if so, ip will fill in the correct address 324 * each time an outbound packet is passed to it. 325 * If we are binding to a broadcast or multicast address icmp_rput 326 * will clear the source address when it receives the T_BIND_ACK. 327 */ 328 icmp->icmp_state = TS_IDLE; 329 330 if (icmp->icmp_family == AF_INET) { 331 ASSERT(sin != NULL); 332 ASSERT(icmp->icmp_ipversion == IPV4_VERSION); 333 IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr, 334 &icmp->icmp_v6src); 335 icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + 336 icmp->icmp_ip_snd_options_len; 337 icmp->icmp_bound_v6src = icmp->icmp_v6src; 338 } else { 339 int error; 340 341 ASSERT(sin6 != NULL); 342 ASSERT(icmp->icmp_ipversion == IPV6_VERSION); 343 icmp->icmp_v6src = sin6->sin6_addr; 344 icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len; 345 icmp->icmp_bound_v6src = icmp->icmp_v6src; 346 347 /* Rebuild the header template */ 348 error = icmp_build_hdrs(q, icmp); 349 if (error != 0) { 350 icmp_err_ack(q, mp, TSYSERR, error); 351 return; 352 } 353 } 354 /* 355 * Place protocol type in the O_T_BIND_REQ/T_BIND_REQ following 356 * the address. 357 */ 358 *mp->b_wptr++ = icmp->icmp_proto; 359 if (!(V6_OR_V4_INADDR_ANY(icmp->icmp_v6src))) { 360 /* 361 * Append a request for an IRE if src not 0 (INADDR_ANY) 362 */ 363 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 364 if (!mp->b_cont) { 365 icmp_err_ack(q, mp, TSYSERR, ENOMEM); 366 return; 367 } 368 mp->b_cont->b_wptr += sizeof (ire_t); 369 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 370 } 371 372 /* Pass the O_T_BIND_REQ/T_BIND_REQ to ip. */ 373 putnext(q, mp); 374 } 375 376 /* 377 * Send message to IP to just bind to the protocol. 378 */ 379 static void 380 icmp_bind_proto(queue_t *q) 381 { 382 mblk_t *mp; 383 struct T_bind_req *tbr; 384 icmp_t *icmp; 385 386 icmp = (icmp_t *)q->q_ptr; 387 mp = allocb(sizeof (struct T_bind_req) + sizeof (sin6_t) + 1, 388 BPRI_MED); 389 if (!mp) { 390 return; 391 } 392 mp->b_datap->db_type = M_PROTO; 393 tbr = (struct T_bind_req *)mp->b_rptr; 394 tbr->PRIM_type = O_T_BIND_REQ; /* change to T_BIND_REQ ? */ 395 tbr->ADDR_offset = sizeof (struct T_bind_req); 396 if (icmp->icmp_ipversion == IPV4_VERSION) { 397 sin_t *sin; 398 399 tbr->ADDR_length = sizeof (sin_t); 400 sin = (sin_t *)&tbr[1]; 401 *sin = sin_null; 402 sin->sin_family = AF_INET; 403 mp->b_wptr = (uchar_t *)&sin[1]; 404 } else { 405 sin6_t *sin6; 406 407 ASSERT(icmp->icmp_ipversion == IPV6_VERSION); 408 tbr->ADDR_length = sizeof (sin6_t); 409 sin6 = (sin6_t *)&tbr[1]; 410 *sin6 = sin6_null; 411 sin6->sin6_family = AF_INET6; 412 mp->b_wptr = (uchar_t *)&sin6[1]; 413 } 414 415 /* Place protocol type in the O_T_BIND_REQ following the address. */ 416 *mp->b_wptr++ = icmp->icmp_proto; 417 418 /* Pass the O_T_BIND_REQ to ip. */ 419 putnext(q, mp); 420 } 421 422 /* 423 * This routine handles each T_CONN_REQ message passed to icmp. It 424 * associates a default destination address with the stream. 425 * 426 * This routine sends down a T_BIND_REQ to IP with the following mblks: 427 * T_BIND_REQ - specifying local and remote address. 428 * IRE_DB_REQ_TYPE - to get an IRE back containing ire_type and src 429 * T_OK_ACK - for the T_CONN_REQ 430 * T_CONN_CON - to keep the TPI user happy 431 * 432 * The connect completes in icmp_rput. 433 * When a T_BIND_ACK is received information is extracted from the IRE 434 * and the two appended messages are sent to the TPI user. 435 * Should icmp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert 436 * it to an error ack for the appropriate primitive. 437 */ 438 static void 439 icmp_connect(queue_t *q, mblk_t *mp) 440 { 441 sin_t *sin; 442 sin6_t *sin6; 443 mblk_t *mp1, *mp2; 444 struct T_conn_req *tcr; 445 icmp_t *icmp; 446 ipaddr_t v4dst; 447 in6_addr_t v6dst; 448 uint32_t flowinfo; 449 450 icmp = (icmp_t *)q->q_ptr; 451 tcr = (struct T_conn_req *)mp->b_rptr; 452 /* Sanity checks */ 453 if ((mp->b_wptr - mp->b_rptr < sizeof (struct T_conn_req))) { 454 icmp_err_ack(q, mp, TPROTO, 0); 455 return; 456 } 457 458 if (icmp->icmp_state == TS_DATA_XFER) { 459 /* Already connected - clear out state */ 460 icmp->icmp_v6src = icmp->icmp_bound_v6src; 461 icmp->icmp_state = TS_IDLE; 462 } 463 464 465 if (tcr->OPT_length != 0) { 466 icmp_err_ack(q, mp, TBADOPT, 0); 467 return; 468 } 469 switch (tcr->DEST_length) { 470 default: 471 icmp_err_ack(q, mp, TBADADDR, 0); 472 return; 473 474 case sizeof (sin_t): 475 sin = (sin_t *)mi_offset_param(mp, tcr->DEST_offset, 476 sizeof (sin_t)); 477 if (sin == NULL || !OK_32PTR((char *)sin)) { 478 icmp_err_ack(q, mp, TSYSERR, EINVAL); 479 return; 480 } 481 if (icmp->icmp_family != AF_INET || 482 sin->sin_family != AF_INET) { 483 icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 484 return; 485 } 486 v4dst = sin->sin_addr.s_addr; 487 IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst); 488 ASSERT(icmp->icmp_ipversion == IPV4_VERSION); 489 icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + 490 icmp->icmp_ip_snd_options_len; 491 break; 492 493 case sizeof (sin6_t): 494 sin6 = (sin6_t *)mi_offset_param(mp, tcr->DEST_offset, 495 sizeof (sin6_t)); 496 if (sin6 == NULL || !OK_32PTR((char *)sin6)) { 497 icmp_err_ack(q, mp, TSYSERR, EINVAL); 498 return; 499 } 500 if (icmp->icmp_family != AF_INET6 || 501 sin6->sin6_family != AF_INET6) { 502 icmp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 503 return; 504 } 505 /* No support for mapped addresses on raw sockets */ 506 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 507 icmp_err_ack(q, mp, TSYSERR, EADDRNOTAVAIL); 508 return; 509 } 510 v6dst = sin6->sin6_addr; 511 ASSERT(icmp->icmp_ipversion == IPV6_VERSION); 512 icmp->icmp_max_hdr_len = icmp->icmp_sticky_hdrs_len; 513 flowinfo = sin6->sin6_flowinfo; 514 break; 515 } 516 if (icmp->icmp_ipversion == IPV4_VERSION) { 517 /* 518 * Interpret a zero destination to mean loopback. 519 * Update the T_CONN_REQ (sin/sin6) since it is used to 520 * generate the T_CONN_CON. 521 */ 522 if (v4dst == INADDR_ANY) { 523 v4dst = htonl(INADDR_LOOPBACK); 524 IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst); 525 if (icmp->icmp_family == AF_INET) { 526 sin->sin_addr.s_addr = v4dst; 527 } else { 528 sin6->sin6_addr = v6dst; 529 } 530 } 531 icmp->icmp_v6dst = v6dst; 532 icmp->icmp_flowinfo = 0; 533 534 /* 535 * If the destination address is multicast and 536 * an outgoing multicast interface has been set, 537 * use the address of that interface as our 538 * source address if no source address has been set. 539 */ 540 if (V4_PART_OF_V6(icmp->icmp_v6src) == INADDR_ANY && 541 CLASSD(v4dst) && 542 icmp->icmp_multicast_if_addr != INADDR_ANY) { 543 IN6_IPADDR_TO_V4MAPPED(icmp->icmp_multicast_if_addr, 544 &icmp->icmp_v6src); 545 } 546 } else { 547 ASSERT(icmp->icmp_ipversion == IPV6_VERSION); 548 /* 549 * Interpret a zero destination to mean loopback. 550 * Update the T_CONN_REQ (sin/sin6) since it is used to 551 * generate the T_CONN_CON. 552 */ 553 if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) { 554 v6dst = ipv6_loopback; 555 sin6->sin6_addr = v6dst; 556 } 557 icmp->icmp_v6dst = v6dst; 558 icmp->icmp_flowinfo = flowinfo; 559 /* 560 * If the destination address is multicast and 561 * an outgoing multicast interface has been set, 562 * then the ip bind logic will pick the correct source 563 * address (i.e. matching the outgoing multicast interface). 564 */ 565 } 566 567 /* 568 * Send down bind to IP to verify that there is a route 569 * and to determine the source address. 570 * This will come back as T_BIND_ACK with an IRE_DB_TYPE in rput. 571 */ 572 if (icmp->icmp_family == AF_INET) { 573 mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (ipa_conn_t), 574 sin->sin_port); 575 } else { 576 ASSERT(icmp->icmp_family == AF_INET6); 577 mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (ipa6_conn_t), 578 sin6->sin6_port); 579 } 580 if (mp1 == NULL) { 581 icmp_err_ack(q, mp, TSYSERR, ENOMEM); 582 return; 583 } 584 585 /* 586 * We also have to send a connection confirmation to 587 * keep TLI happy. Prepare it for icmp_rput. 588 */ 589 if (icmp->icmp_family == AF_INET) { 590 mp2 = mi_tpi_conn_con(NULL, (char *)sin, sizeof (*sin), NULL, 591 0); 592 } else { 593 ASSERT(icmp->icmp_family == AF_INET6); 594 mp2 = mi_tpi_conn_con(NULL, (char *)sin6, sizeof (*sin6), NULL, 595 0); 596 } 597 if (mp2 == NULL) { 598 freemsg(mp1); 599 icmp_err_ack(q, mp, TSYSERR, ENOMEM); 600 return; 601 } 602 603 mp = mi_tpi_ok_ack_alloc(mp); 604 if (mp == NULL) { 605 /* Unable to reuse the T_CONN_REQ for the ack. */ 606 freemsg(mp2); 607 icmp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM); 608 return; 609 } 610 611 icmp->icmp_state = TS_DATA_XFER; 612 613 /* Hang onto the T_OK_ACK and T_CONN_CON for later. */ 614 linkb(mp1, mp); 615 linkb(mp1, mp2); 616 617 putnext(q, mp1); 618 } 619 620 static int 621 icmp_close(queue_t *q) 622 { 623 icmp_t *icmp = (icmp_t *)q->q_ptr; 624 int i1; 625 626 qprocsoff(q); 627 628 /* If there are any options associated with the stream, free them. */ 629 if (icmp->icmp_ip_snd_options) 630 mi_free((char *)icmp->icmp_ip_snd_options); 631 632 if (icmp->icmp_filter != NULL) 633 kmem_free(icmp->icmp_filter, sizeof (icmp6_filter_t)); 634 635 /* Free memory associated with sticky options */ 636 if (icmp->icmp_sticky_hdrs_len != 0) { 637 kmem_free(icmp->icmp_sticky_hdrs, 638 icmp->icmp_sticky_hdrs_len); 639 icmp->icmp_sticky_hdrs = NULL; 640 icmp->icmp_sticky_hdrs_len = 0; 641 } 642 if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) { 643 kmem_free(icmp->icmp_sticky_ipp.ipp_hopopts, 644 icmp->icmp_sticky_ipp.ipp_hopoptslen); 645 } 646 if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) { 647 kmem_free(icmp->icmp_sticky_ipp.ipp_rtdstopts, 648 icmp->icmp_sticky_ipp.ipp_rtdstoptslen); 649 } 650 if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) { 651 kmem_free(icmp->icmp_sticky_ipp.ipp_rthdr, 652 icmp->icmp_sticky_ipp.ipp_rthdrlen); 653 } 654 if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) { 655 kmem_free(icmp->icmp_sticky_ipp.ipp_dstopts, 656 icmp->icmp_sticky_ipp.ipp_dstoptslen); 657 } 658 if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_PATHMTU) { 659 kmem_free(icmp->icmp_sticky_ipp.ipp_pathmtu, 660 icmp->icmp_sticky_ipp.ipp_pathmtulen); 661 } 662 icmp->icmp_sticky_ipp.ipp_fields &= 663 ~(IPPF_HOPOPTS|IPPF_RTDSTOPTS|IPPF_RTHDR|IPPF_DSTOPTS); 664 665 crfree(icmp->icmp_credp); 666 667 /* Free the icmp structure and release the minor device number. */ 668 i1 = mi_close_comm(&icmp_g_head, q); 669 670 return (i1); 671 } 672 673 /* 674 * This routine handles each T_DISCON_REQ message passed to icmp 675 * as an indicating that ICMP is no longer connected. This results 676 * in sending a T_BIND_REQ to IP to restore the binding to just 677 * the local address. 678 * 679 * This routine sends down a T_BIND_REQ to IP with the following mblks: 680 * T_BIND_REQ - specifying just the local address. 681 * T_OK_ACK - for the T_DISCON_REQ 682 * 683 * The disconnect completes in icmp_rput. 684 * When a T_BIND_ACK is received the appended T_OK_ACK is sent to the TPI user. 685 * Should icmp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert 686 * it to an error ack for the appropriate primitive. 687 */ 688 static void 689 icmp_disconnect(queue_t *q, mblk_t *mp) 690 { 691 icmp_t *icmp; 692 mblk_t *mp1; 693 694 icmp = (icmp_t *)q->q_ptr; 695 696 if (icmp->icmp_state != TS_DATA_XFER) { 697 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 698 "icmp_disconnect: bad state, %d", icmp->icmp_state); 699 icmp_err_ack(q, mp, TOUTSTATE, 0); 700 return; 701 } 702 icmp->icmp_v6src = icmp->icmp_bound_v6src; 703 icmp->icmp_state = TS_IDLE; 704 705 /* 706 * Send down bind to IP to remove the full binding and revert 707 * to the local address binding. 708 */ 709 if (icmp->icmp_family == AF_INET) { 710 mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (sin_t), 0); 711 } else { 712 ASSERT(icmp->icmp_family == AF_INET6); 713 mp1 = icmp_ip_bind_mp(icmp, O_T_BIND_REQ, sizeof (sin6_t), 0); 714 } 715 if (mp1 == NULL) { 716 icmp_err_ack(q, mp, TSYSERR, ENOMEM); 717 return; 718 } 719 mp = mi_tpi_ok_ack_alloc(mp); 720 if (mp == NULL) { 721 /* Unable to reuse the T_DISCON_REQ for the ack. */ 722 icmp_err_ack_prim(q, mp1, T_DISCON_REQ, TSYSERR, ENOMEM); 723 return; 724 } 725 726 if (icmp->icmp_family == AF_INET6) { 727 int error; 728 729 /* Rebuild the header template */ 730 error = icmp_build_hdrs(q, icmp); 731 if (error != 0) { 732 icmp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, error); 733 freemsg(mp1); 734 return; 735 } 736 } 737 icmp->icmp_discon_pending = 1; 738 739 /* Append the T_OK_ACK to the T_BIND_REQ for icmp_rput */ 740 linkb(mp1, mp); 741 putnext(q, mp1); 742 } 743 744 /* This routine creates a T_ERROR_ACK message and passes it upstream. */ 745 static void 746 icmp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error) 747 { 748 if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL) 749 qreply(q, mp); 750 } 751 752 /* Shorthand to generate and send TPI error acks to our client */ 753 static void 754 icmp_err_ack_prim(queue_t *q, mblk_t *mp, t_scalar_t primitive, 755 t_scalar_t t_error, int sys_error) 756 { 757 struct T_error_ack *teackp; 758 759 if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack), 760 M_PCPROTO, T_ERROR_ACK)) != NULL) { 761 teackp = (struct T_error_ack *)mp->b_rptr; 762 teackp->ERROR_prim = primitive; 763 teackp->TLI_error = t_error; 764 teackp->UNIX_error = sys_error; 765 qreply(q, mp); 766 } 767 } 768 769 /* 770 * icmp_icmp_error is called by icmp_rput to process ICMP 771 * messages passed up by IP. 772 * Generates the appropriate T_UDERROR_IND for permanent 773 * (non-transient) errors. 774 * Assumes that IP has pulled up everything up to and including 775 * the ICMP header. 776 */ 777 static void 778 icmp_icmp_error(queue_t *q, mblk_t *mp) 779 { 780 icmph_t *icmph; 781 ipha_t *ipha; 782 int iph_hdr_length; 783 sin_t sin; 784 sin6_t sin6; 785 mblk_t *mp1; 786 int error = 0; 787 icmp_t *icmp = (icmp_t *)q->q_ptr; 788 789 /* 790 * Deliver T_UDERROR_IND when the application has asked for it. 791 * The socket layer enables this automatically when connected. 792 */ 793 if (!icmp->icmp_dgram_errind) { 794 freemsg(mp); 795 return; 796 } 797 798 ipha = (ipha_t *)mp->b_rptr; 799 800 if (IPH_HDR_VERSION(ipha) != IPV4_VERSION) { 801 ASSERT(IPH_HDR_VERSION(ipha) == IPV6_VERSION); 802 icmp_icmp_error_ipv6(q, mp); 803 return; 804 } 805 ASSERT(IPH_HDR_VERSION(ipha) == IPV4_VERSION); 806 807 iph_hdr_length = IPH_HDR_LENGTH(ipha); 808 icmph = (icmph_t *)(&mp->b_rptr[iph_hdr_length]); 809 ipha = (ipha_t *)&icmph[1]; 810 iph_hdr_length = IPH_HDR_LENGTH(ipha); 811 812 switch (icmph->icmph_type) { 813 case ICMP_DEST_UNREACHABLE: 814 switch (icmph->icmph_code) { 815 case ICMP_FRAGMENTATION_NEEDED: 816 /* 817 * IP has already adjusted the path MTU. 818 * XXX Somehow pass MTU indication to application? 819 */ 820 break; 821 case ICMP_PORT_UNREACHABLE: 822 case ICMP_PROTOCOL_UNREACHABLE: 823 error = ECONNREFUSED; 824 break; 825 default: 826 /* Transient errors */ 827 break; 828 } 829 break; 830 default: 831 /* Transient errors */ 832 break; 833 } 834 if (error == 0) { 835 freemsg(mp); 836 return; 837 } 838 839 switch (icmp->icmp_family) { 840 case AF_INET: 841 sin = sin_null; 842 sin.sin_family = AF_INET; 843 sin.sin_addr.s_addr = ipha->ipha_dst; 844 mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL, 0, 845 error); 846 break; 847 case AF_INET6: 848 sin6 = sin6_null; 849 sin6.sin6_family = AF_INET6; 850 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr); 851 852 mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), 853 NULL, 0, error); 854 break; 855 } 856 if (mp1) 857 putnext(q, mp1); 858 freemsg(mp); 859 } 860 861 /* 862 * icmp_icmp_error_ipv6 is called by icmp_icmp_error to process ICMPv6 863 * for IPv6 packets. 864 * Send permanent (non-transient) errors upstream. 865 * Assumes that IP has pulled up all the extension headers as well 866 * as the ICMPv6 header. 867 */ 868 static void 869 icmp_icmp_error_ipv6(queue_t *q, mblk_t *mp) 870 { 871 icmp6_t *icmp6; 872 ip6_t *ip6h, *outer_ip6h; 873 uint16_t iph_hdr_length; 874 uint8_t *nexthdrp; 875 sin6_t sin6; 876 mblk_t *mp1; 877 int error = 0; 878 icmp_t *icmp = (icmp_t *)q->q_ptr; 879 880 outer_ip6h = (ip6_t *)mp->b_rptr; 881 if (outer_ip6h->ip6_nxt != IPPROTO_ICMPV6) 882 iph_hdr_length = ip_hdr_length_v6(mp, outer_ip6h); 883 else 884 iph_hdr_length = IPV6_HDR_LEN; 885 886 icmp6 = (icmp6_t *)&mp->b_rptr[iph_hdr_length]; 887 ip6h = (ip6_t *)&icmp6[1]; 888 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &iph_hdr_length, &nexthdrp)) { 889 freemsg(mp); 890 return; 891 } 892 if (*nexthdrp != icmp->icmp_proto) { 893 /* 894 * Could have switched icmp_proto after while ip did fanout of 895 * this message 896 */ 897 freemsg(mp); 898 return; 899 } 900 switch (icmp6->icmp6_type) { 901 case ICMP6_DST_UNREACH: 902 switch (icmp6->icmp6_code) { 903 case ICMP6_DST_UNREACH_NOPORT: 904 error = ECONNREFUSED; 905 break; 906 case ICMP6_DST_UNREACH_ADMIN: 907 case ICMP6_DST_UNREACH_NOROUTE: 908 case ICMP6_DST_UNREACH_BEYONDSCOPE: 909 case ICMP6_DST_UNREACH_ADDR: 910 /* Transient errors */ 911 break; 912 default: 913 break; 914 } 915 break; 916 case ICMP6_PACKET_TOO_BIG: { 917 struct T_unitdata_ind *tudi; 918 struct T_opthdr *toh; 919 size_t udi_size; 920 mblk_t *newmp; 921 t_scalar_t opt_length = sizeof (struct T_opthdr) + 922 sizeof (struct ip6_mtuinfo); 923 sin6_t *sin6; 924 struct ip6_mtuinfo *mtuinfo; 925 926 /* 927 * If the application has requested to receive path mtu 928 * information, send up an empty message containing an 929 * IPV6_PATHMTU ancillary data item. 930 */ 931 if (!icmp->icmp_ipv6_recvpathmtu) 932 break; 933 934 udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) + 935 opt_length; 936 if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) { 937 BUMP_MIB(&rawip_mib, rawipInErrors); 938 break; 939 } 940 941 /* 942 * newmp->b_cont is left to NULL on purpose. This is an 943 * empty message containing only ancillary data. 944 */ 945 newmp->b_datap->db_type = M_PROTO; 946 tudi = (struct T_unitdata_ind *)newmp->b_rptr; 947 newmp->b_wptr = (uchar_t *)tudi + udi_size; 948 tudi->PRIM_type = T_UNITDATA_IND; 949 tudi->SRC_length = sizeof (sin6_t); 950 tudi->SRC_offset = sizeof (struct T_unitdata_ind); 951 tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t); 952 tudi->OPT_length = opt_length; 953 954 sin6 = (sin6_t *)&tudi[1]; 955 bzero(sin6, sizeof (sin6_t)); 956 sin6->sin6_family = AF_INET6; 957 sin6->sin6_addr = icmp->icmp_v6dst; 958 959 toh = (struct T_opthdr *)&sin6[1]; 960 toh->level = IPPROTO_IPV6; 961 toh->name = IPV6_PATHMTU; 962 toh->len = opt_length; 963 toh->status = 0; 964 965 mtuinfo = (struct ip6_mtuinfo *)&toh[1]; 966 bzero(mtuinfo, sizeof (struct ip6_mtuinfo)); 967 mtuinfo->ip6m_addr.sin6_family = AF_INET6; 968 mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst; 969 mtuinfo->ip6m_mtu = icmp6->icmp6_mtu; 970 /* 971 * We've consumed everything we need from the original 972 * message. Free it, then send our empty message. 973 */ 974 freemsg(mp); 975 putnext(q, newmp); 976 return; 977 } 978 case ICMP6_TIME_EXCEEDED: 979 /* Transient errors */ 980 break; 981 case ICMP6_PARAM_PROB: 982 /* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */ 983 if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER && 984 (uchar_t *)ip6h + icmp6->icmp6_pptr == 985 (uchar_t *)nexthdrp) { 986 error = ECONNREFUSED; 987 break; 988 } 989 break; 990 } 991 if (error == 0) { 992 freemsg(mp); 993 return; 994 } 995 996 sin6 = sin6_null; 997 sin6.sin6_family = AF_INET6; 998 sin6.sin6_addr = ip6h->ip6_dst; 999 sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK; 1000 1001 mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), NULL, 0, 1002 error); 1003 if (mp1) 1004 putnext(q, mp1); 1005 freemsg(mp); 1006 } 1007 1008 /* 1009 * This routine responds to T_ADDR_REQ messages. It is called by icmp_wput. 1010 * The local address is filled in if endpoint is bound. The remote address 1011 * is filled in if remote address has been precified ("connected endpoint") 1012 * (The concept of connected CLTS sockets is alien to published TPI 1013 * but we support it anyway). 1014 */ 1015 static void 1016 icmp_addr_req(queue_t *q, mblk_t *mp) 1017 { 1018 icmp_t *icmp = (icmp_t *)q->q_ptr; 1019 mblk_t *ackmp; 1020 struct T_addr_ack *taa; 1021 1022 /* Make it large enough for worst case */ 1023 ackmp = reallocb(mp, sizeof (struct T_addr_ack) + 1024 2 * sizeof (sin6_t), 1); 1025 if (ackmp == NULL) { 1026 icmp_err_ack(q, mp, TSYSERR, ENOMEM); 1027 return; 1028 } 1029 taa = (struct T_addr_ack *)ackmp->b_rptr; 1030 1031 bzero(taa, sizeof (struct T_addr_ack)); 1032 ackmp->b_wptr = (uchar_t *)&taa[1]; 1033 1034 taa->PRIM_type = T_ADDR_ACK; 1035 ackmp->b_datap->db_type = M_PCPROTO; 1036 1037 /* 1038 * Note: Following code assumes 32 bit alignment of basic 1039 * data structures like sin_t and struct T_addr_ack. 1040 */ 1041 if (icmp->icmp_state != TS_UNBND) { 1042 /* 1043 * Fill in local address 1044 */ 1045 taa->LOCADDR_offset = sizeof (*taa); 1046 if (icmp->icmp_family == AF_INET) { 1047 sin_t *sin; 1048 1049 taa->LOCADDR_length = sizeof (sin_t); 1050 sin = (sin_t *)&taa[1]; 1051 /* Fill zeroes and then intialize non-zero fields */ 1052 *sin = sin_null; 1053 sin->sin_family = AF_INET; 1054 if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) && 1055 !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) { 1056 IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, 1057 sin->sin_addr.s_addr); 1058 } else { 1059 /* 1060 * INADDR_ANY 1061 * icmp_v6src is not set, we might be bound to 1062 * broadcast/multicast. Use icmp_bound_v6src as 1063 * local address instead (that could 1064 * also still be INADDR_ANY) 1065 */ 1066 IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_bound_v6src, 1067 sin->sin_addr.s_addr); 1068 } 1069 ackmp->b_wptr = (uchar_t *)&sin[1]; 1070 } else { 1071 sin6_t *sin6; 1072 1073 ASSERT(icmp->icmp_family == AF_INET6); 1074 taa->LOCADDR_length = sizeof (sin6_t); 1075 sin6 = (sin6_t *)&taa[1]; 1076 /* Fill zeroes and then intialize non-zero fields */ 1077 *sin6 = sin6_null; 1078 sin6->sin6_family = AF_INET6; 1079 if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) { 1080 sin6->sin6_addr = icmp->icmp_v6src; 1081 } else { 1082 /* 1083 * UNSPECIFIED 1084 * icmp_v6src is not set, we might be bound to 1085 * broadcast/multicast. Use icmp_bound_v6src as 1086 * local address instead (that could 1087 * also still be UNSPECIFIED) 1088 */ 1089 sin6->sin6_addr = icmp->icmp_bound_v6src; 1090 } 1091 ackmp->b_wptr = (uchar_t *)&sin6[1]; 1092 } 1093 } 1094 ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim); 1095 qreply(q, ackmp); 1096 } 1097 1098 static void 1099 icmp_copy_info(struct T_info_ack *tap, icmp_t *icmp) 1100 { 1101 *tap = icmp_g_t_info_ack; 1102 1103 if (icmp->icmp_family == AF_INET6) 1104 tap->ADDR_size = sizeof (sin6_t); 1105 else 1106 tap->ADDR_size = sizeof (sin_t); 1107 tap->CURRENT_state = icmp->icmp_state; 1108 tap->OPT_size = icmp_max_optsize; 1109 } 1110 1111 /* 1112 * This routine responds to T_CAPABILITY_REQ messages. It is called by 1113 * icmp_wput. Much of the T_CAPABILITY_ACK information is copied from 1114 * icmp_g_t_info_ack. The current state of the stream is copied from 1115 * icmp_state. 1116 */ 1117 static void 1118 icmp_capability_req(queue_t *q, mblk_t *mp) 1119 { 1120 icmp_t *icmp = (icmp_t *)q->q_ptr; 1121 t_uscalar_t cap_bits1; 1122 struct T_capability_ack *tcap; 1123 1124 cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1; 1125 1126 mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack), 1127 mp->b_datap->db_type, T_CAPABILITY_ACK); 1128 if (!mp) 1129 return; 1130 1131 tcap = (struct T_capability_ack *)mp->b_rptr; 1132 tcap->CAP_bits1 = 0; 1133 1134 if (cap_bits1 & TC1_INFO) { 1135 icmp_copy_info(&tcap->INFO_ack, icmp); 1136 tcap->CAP_bits1 |= TC1_INFO; 1137 } 1138 1139 qreply(q, mp); 1140 } 1141 1142 /* 1143 * This routine responds to T_INFO_REQ messages. It is called by icmp_wput. 1144 * Most of the T_INFO_ACK information is copied from icmp_g_t_info_ack. 1145 * The current state of the stream is copied from icmp_state. 1146 */ 1147 static void 1148 icmp_info_req(queue_t *q, mblk_t *mp) 1149 { 1150 icmp_t *icmp = (icmp_t *)q->q_ptr; 1151 1152 mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO, 1153 T_INFO_ACK); 1154 if (!mp) 1155 return; 1156 icmp_copy_info((struct T_info_ack *)mp->b_rptr, icmp); 1157 qreply(q, mp); 1158 } 1159 1160 /* 1161 * IP recognizes seven kinds of bind requests: 1162 * 1163 * - A zero-length address binds only to the protocol number. 1164 * 1165 * - A 4-byte address is treated as a request to 1166 * validate that the address is a valid local IPv4 1167 * address, appropriate for an application to bind to. 1168 * IP does the verification, but does not make any note 1169 * of the address at this time. 1170 * 1171 * - A 16-byte address contains is treated as a request 1172 * to validate a local IPv6 address, as the 4-byte 1173 * address case above. 1174 * 1175 * - A 16-byte sockaddr_in to validate the local IPv4 address and also 1176 * use it for the inbound fanout of packets. 1177 * 1178 * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also 1179 * use it for the inbound fanout of packets. 1180 * 1181 * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout 1182 * information consisting of local and remote addresses 1183 * and ports (unused for raw sockets). In this case, the addresses are both 1184 * validated as appropriate for this operation, and, if 1185 * so, the information is retained for use in the 1186 * inbound fanout. 1187 * 1188 * - A 36-byte address address (ipa6_conn_t) containing complete IPv6 1189 * fanout information, like the 12-byte case above. 1190 * 1191 * IP will also fill in the IRE request mblk with information 1192 * regarding our peer. In all cases, we notify IP of our protocol 1193 * type by appending a single protocol byte to the bind request. 1194 */ 1195 static mblk_t * 1196 icmp_ip_bind_mp(icmp_t *icmp, t_scalar_t bind_prim, t_scalar_t addr_length, 1197 in_port_t fport) 1198 { 1199 char *cp; 1200 mblk_t *mp; 1201 struct T_bind_req *tbr; 1202 ipa_conn_t *ac; 1203 ipa6_conn_t *ac6; 1204 sin_t *sin; 1205 sin6_t *sin6; 1206 1207 ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ); 1208 1209 mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI); 1210 if (mp == NULL) 1211 return (NULL); 1212 mp->b_datap->db_type = M_PROTO; 1213 tbr = (struct T_bind_req *)mp->b_rptr; 1214 tbr->PRIM_type = bind_prim; 1215 tbr->ADDR_offset = sizeof (*tbr); 1216 tbr->CONIND_number = 0; 1217 tbr->ADDR_length = addr_length; 1218 cp = (char *)&tbr[1]; 1219 switch (addr_length) { 1220 case sizeof (ipa_conn_t): 1221 ASSERT(icmp->icmp_family == AF_INET); 1222 /* Append a request for an IRE */ 1223 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 1224 if (mp->b_cont == NULL) { 1225 freemsg(mp); 1226 return (NULL); 1227 } 1228 mp->b_cont->b_wptr += sizeof (ire_t); 1229 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 1230 1231 /* cp known to be 32 bit aligned */ 1232 ac = (ipa_conn_t *)cp; 1233 ac->ac_laddr = V4_PART_OF_V6(icmp->icmp_v6src); 1234 ac->ac_faddr = V4_PART_OF_V6(icmp->icmp_v6dst); 1235 ac->ac_fport = fport; 1236 ac->ac_lport = 0; 1237 break; 1238 1239 case sizeof (ipa6_conn_t): 1240 ASSERT(icmp->icmp_family == AF_INET6); 1241 /* Append a request for an IRE */ 1242 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 1243 if (mp->b_cont == NULL) { 1244 freemsg(mp); 1245 return (NULL); 1246 } 1247 mp->b_cont->b_wptr += sizeof (ire_t); 1248 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 1249 1250 /* cp known to be 32 bit aligned */ 1251 ac6 = (ipa6_conn_t *)cp; 1252 ac6->ac6_laddr = icmp->icmp_v6src; 1253 ac6->ac6_faddr = icmp->icmp_v6dst; 1254 ac6->ac6_fport = fport; 1255 ac6->ac6_lport = 0; 1256 break; 1257 1258 case sizeof (sin_t): 1259 ASSERT(icmp->icmp_family == AF_INET); 1260 /* Append a request for an IRE */ 1261 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 1262 if (!mp->b_cont) { 1263 freemsg(mp); 1264 return (NULL); 1265 } 1266 mp->b_cont->b_wptr += sizeof (ire_t); 1267 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 1268 1269 sin = (sin_t *)cp; 1270 *sin = sin_null; 1271 sin->sin_family = AF_INET; 1272 sin->sin_addr.s_addr = V4_PART_OF_V6(icmp->icmp_bound_v6src); 1273 break; 1274 1275 case sizeof (sin6_t): 1276 ASSERT(icmp->icmp_family == AF_INET6); 1277 /* Append a request for an IRE */ 1278 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 1279 if (!mp->b_cont) { 1280 freemsg(mp); 1281 return (NULL); 1282 } 1283 mp->b_cont->b_wptr += sizeof (ire_t); 1284 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 1285 1286 sin6 = (sin6_t *)cp; 1287 *sin6 = sin6_null; 1288 sin6->sin6_family = AF_INET6; 1289 sin6->sin6_addr = icmp->icmp_bound_v6src; 1290 break; 1291 } 1292 /* Add protocol number to end */ 1293 cp[addr_length] = icmp->icmp_proto; 1294 mp->b_wptr = (uchar_t *)&cp[addr_length + 1]; 1295 return (mp); 1296 } 1297 1298 /* 1299 * This is the open routine for icmp. It allocates a icmp_t structure for 1300 * the stream and, on the first open of the module, creates an ND table. 1301 */ 1302 static int 1303 icmp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp) 1304 { 1305 int err; 1306 icmp_t *icmp; 1307 1308 /* If the stream is already open, return immediately. */ 1309 if (q->q_ptr != NULL) 1310 return (0); 1311 1312 /* If this is not a push of icmp as a module, fail. */ 1313 if (sflag != MODOPEN) 1314 return (EINVAL); 1315 1316 /* 1317 * Defer the qprocson until everything is initialized since 1318 * we are D_MTPERQ and after qprocson the rput routine can 1319 * run. (Could do qprocson earlier since icmp currently 1320 * has an outer perimeter.) 1321 */ 1322 1323 /* 1324 * Create a icmp_t structure for this stream and link into the 1325 * list of open streams. 1326 */ 1327 err = mi_open_comm(&icmp_g_head, sizeof (icmp_t), q, devp, 1328 flag, sflag, credp); 1329 if (err) 1330 return (err); 1331 1332 /* 1333 * The receive hiwat is only looked at on the stream head queue. 1334 * Store in q_hiwat in order to return on SO_RCVBUF getsockopts. 1335 */ 1336 q->q_hiwat = icmp_recv_hiwat; 1337 1338 /* Set the initial state of the stream and the privilege status. */ 1339 icmp = (icmp_t *)q->q_ptr; 1340 icmp->icmp_state = TS_UNBND; 1341 icmp->icmp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 1342 icmp->icmp_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 1343 icmp->icmp_filter = NULL; 1344 1345 icmp->icmp_credp = credp; 1346 crhold(credp); 1347 1348 icmp->icmp_zoneid = getzoneid(); 1349 1350 if (getmajor(*devp) == (major_t)ICMP6_MAJ) { 1351 icmp->icmp_ipversion = IPV6_VERSION; 1352 icmp->icmp_family = AF_INET6; 1353 /* May be changed by a SO_PROTOTYPE socket option. */ 1354 icmp->icmp_proto = IPPROTO_ICMPV6; 1355 icmp->icmp_checksum_off = 2; /* Offset for icmp6_cksum */ 1356 icmp->icmp_max_hdr_len = IPV6_HDR_LEN; 1357 icmp->icmp_ttl = (uint8_t)icmp_ipv6_hoplimit; 1358 } else { 1359 icmp->icmp_ipversion = IPV4_VERSION; 1360 icmp->icmp_family = AF_INET; 1361 /* May be changed by a SO_PROTOTYPE socket option. */ 1362 icmp->icmp_proto = IPPROTO_ICMP; 1363 icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH; 1364 icmp->icmp_ttl = (uint8_t)icmp_ipv4_ttl; 1365 } 1366 qprocson(q); 1367 1368 /* 1369 * Check if icmp is being I_PUSHed by a non-privileged user. 1370 * If so, we set icmp_restricted to indicate that only MIB 1371 * traffic may pass. 1372 */ 1373 if (secpolicy_net_icmpaccess(credp) != 0) { 1374 icmp->icmp_restricted = 1; 1375 } 1376 1377 /* 1378 * The transmit hiwat is only looked at on IP's queue. 1379 * Store in q_hiwat in order to return on SO_SNDBUF 1380 * getsockopts. 1381 */ 1382 WR(q)->q_hiwat = icmp_xmit_hiwat; 1383 WR(q)->q_next->q_hiwat = WR(q)->q_hiwat; 1384 WR(q)->q_lowat = icmp_xmit_lowat; 1385 WR(q)->q_next->q_lowat = WR(q)->q_lowat; 1386 1387 if (icmp->icmp_family == AF_INET6) { 1388 /* Build initial header template for transmit */ 1389 int error; 1390 1391 error = icmp_build_hdrs(q, icmp); 1392 if (error != 0) { 1393 (void) icmp_close(q); 1394 return (error); 1395 } 1396 } 1397 /* Set the Stream head write offset. */ 1398 (void) mi_set_sth_wroff(q, icmp->icmp_max_hdr_len + icmp_wroff_extra); 1399 (void) mi_set_sth_hiwat(q, q->q_hiwat); 1400 1401 return (0); 1402 } 1403 1404 /* 1405 * Which ICMP options OK to set through T_UNITDATA_REQ... 1406 */ 1407 /* ARGSUSED */ 1408 static boolean_t 1409 icmp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name) 1410 { 1411 return (B_TRUE); 1412 } 1413 1414 /* 1415 * This routine gets default values of certain options whose default 1416 * values are maintained by protcol specific code 1417 */ 1418 /* ARGSUSED */ 1419 int 1420 icmp_opt_default(queue_t *q, int level, int name, uchar_t *ptr) 1421 { 1422 int *i1 = (int *)ptr; 1423 1424 switch (level) { 1425 case IPPROTO_IP: 1426 switch (name) { 1427 case IP_MULTICAST_TTL: 1428 *ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL; 1429 return (sizeof (uchar_t)); 1430 case IP_MULTICAST_LOOP: 1431 *ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP; 1432 return (sizeof (uchar_t)); 1433 } 1434 break; 1435 case IPPROTO_IPV6: 1436 switch (name) { 1437 case IPV6_MULTICAST_HOPS: 1438 *i1 = IP_DEFAULT_MULTICAST_TTL; 1439 return (sizeof (int)); 1440 case IPV6_MULTICAST_LOOP: 1441 *i1 = IP_DEFAULT_MULTICAST_LOOP; 1442 return (sizeof (int)); 1443 case IPV6_UNICAST_HOPS: 1444 *i1 = icmp_ipv6_hoplimit; 1445 return (sizeof (int)); 1446 } 1447 break; 1448 case IPPROTO_ICMPV6: 1449 switch (name) { 1450 case ICMP6_FILTER: 1451 /* Make it look like "pass all" */ 1452 ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr); 1453 return (sizeof (icmp6_filter_t)); 1454 } 1455 break; 1456 } 1457 return (-1); 1458 } 1459 1460 /* 1461 * This routine retrieves the current status of socket options. 1462 * It returns the size of the option retrieved. 1463 */ 1464 int 1465 icmp_opt_get(queue_t *q, int level, int name, uchar_t *ptr) 1466 { 1467 icmp_t *icmp = (icmp_t *)q->q_ptr; 1468 int *i1 = (int *)ptr; 1469 ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp; 1470 1471 switch (level) { 1472 case SOL_SOCKET: 1473 switch (name) { 1474 case SO_DEBUG: 1475 *i1 = icmp->icmp_debug; 1476 break; 1477 case SO_TYPE: 1478 *i1 = SOCK_RAW; 1479 break; 1480 case SO_PROTOTYPE: 1481 *i1 = icmp->icmp_proto; 1482 break; 1483 case SO_REUSEADDR: 1484 *i1 = icmp->icmp_reuseaddr; 1485 break; 1486 1487 /* 1488 * The following three items are available here, 1489 * but are only meaningful to IP. 1490 */ 1491 case SO_DONTROUTE: 1492 *i1 = icmp->icmp_dontroute; 1493 break; 1494 case SO_USELOOPBACK: 1495 *i1 = icmp->icmp_useloopback; 1496 break; 1497 case SO_BROADCAST: 1498 *i1 = icmp->icmp_broadcast; 1499 break; 1500 1501 case SO_SNDBUF: 1502 ASSERT(q->q_hiwat <= INT_MAX); 1503 *i1 = (int)q->q_hiwat; 1504 break; 1505 case SO_RCVBUF: 1506 ASSERT(RD(q)->q_hiwat <= INT_MAX); 1507 *i1 = (int)RD(q)->q_hiwat; 1508 break; 1509 case SO_DGRAM_ERRIND: 1510 *i1 = icmp->icmp_dgram_errind; 1511 break; 1512 /* 1513 * Following three not meaningful for icmp 1514 * Action is same as "default" to which we fallthrough 1515 * so we keep them in comments. 1516 * case SO_LINGER: 1517 * case SO_KEEPALIVE: 1518 * case SO_OOBINLINE: 1519 */ 1520 default: 1521 return (-1); 1522 } 1523 break; 1524 case IPPROTO_IP: 1525 /* 1526 * Only allow IPv4 option processing on IPv4 sockets. 1527 */ 1528 if (icmp->icmp_family != AF_INET) 1529 return (-1); 1530 1531 switch (name) { 1532 case IP_OPTIONS: 1533 case T_IP_OPTIONS: 1534 /* Options are passed up with each packet */ 1535 return (0); 1536 case IP_HDRINCL: 1537 *i1 = (int)icmp->icmp_hdrincl; 1538 break; 1539 case IP_TOS: 1540 case T_IP_TOS: 1541 *i1 = (int)icmp->icmp_type_of_service; 1542 break; 1543 case IP_TTL: 1544 *i1 = (int)icmp->icmp_ttl; 1545 break; 1546 case IP_MULTICAST_IF: 1547 /* 0 address if not set */ 1548 *(ipaddr_t *)ptr = icmp->icmp_multicast_if_addr; 1549 return (sizeof (ipaddr_t)); 1550 case IP_MULTICAST_TTL: 1551 *(uchar_t *)ptr = icmp->icmp_multicast_ttl; 1552 return (sizeof (uchar_t)); 1553 case IP_MULTICAST_LOOP: 1554 *ptr = icmp->icmp_multicast_loop; 1555 return (sizeof (uint8_t)); 1556 case IP_BOUND_IF: 1557 /* Zero if not set */ 1558 *i1 = icmp->icmp_bound_if; 1559 break; /* goto sizeof (int) option return */ 1560 case IP_UNSPEC_SRC: 1561 *ptr = icmp->icmp_unspec_source; 1562 break; /* goto sizeof (int) option return */ 1563 case IP_XMIT_IF: 1564 *i1 = icmp->icmp_xmit_if; 1565 break; /* goto sizeof (int) option return */ 1566 case IP_RECVIF: 1567 *ptr = icmp->icmp_recvif; 1568 break; /* goto sizeof (int) option return */ 1569 /* 1570 * Cannot "get" the value of following options 1571 * at this level. Action is same as "default" to 1572 * which we fallthrough so we keep them in comments. 1573 * 1574 * case IP_ADD_MEMBERSHIP: 1575 * case IP_DROP_MEMBERSHIP: 1576 * case IP_BLOCK_SOURCE: 1577 * case IP_UNBLOCK_SOURCE: 1578 * case IP_ADD_SOURCE_MEMBERSHIP: 1579 * case IP_DROP_SOURCE_MEMBERSHIP: 1580 * case MCAST_JOIN_GROUP: 1581 * case MCAST_LEAVE_GROUP: 1582 * case MCAST_BLOCK_SOURCE: 1583 * case MCAST_UNBLOCK_SOURCE: 1584 * case MCAST_JOIN_SOURCE_GROUP: 1585 * case MCAST_LEAVE_SOURCE_GROUP: 1586 * case MRT_INIT: 1587 * case MRT_DONE: 1588 * case MRT_ADD_VIF: 1589 * case MRT_DEL_VIF: 1590 * case MRT_ADD_MFC: 1591 * case MRT_DEL_MFC: 1592 * case MRT_VERSION: 1593 * case MRT_ASSERT: 1594 * case IP_SEC_OPT: 1595 * case IP_DONTFAILOVER_IF: 1596 */ 1597 default: 1598 return (-1); 1599 } 1600 break; 1601 case IPPROTO_IPV6: 1602 /* 1603 * Only allow IPv6 option processing on native IPv6 sockets. 1604 */ 1605 if (icmp->icmp_family != AF_INET6) 1606 return (-1); 1607 switch (name) { 1608 case IPV6_UNICAST_HOPS: 1609 *i1 = (unsigned int)icmp->icmp_ttl; 1610 break; 1611 case IPV6_MULTICAST_IF: 1612 /* 0 index if not set */ 1613 *i1 = icmp->icmp_multicast_if_index; 1614 break; 1615 case IPV6_MULTICAST_HOPS: 1616 *i1 = icmp->icmp_multicast_ttl; 1617 break; 1618 case IPV6_MULTICAST_LOOP: 1619 *i1 = icmp->icmp_multicast_loop; 1620 break; 1621 case IPV6_BOUND_IF: 1622 /* Zero if not set */ 1623 *i1 = icmp->icmp_bound_if; 1624 break; 1625 case IPV6_UNSPEC_SRC: 1626 *i1 = icmp->icmp_unspec_source; 1627 break; 1628 case IPV6_CHECKSUM: 1629 /* 1630 * Return offset or -1 if no checksum offset. 1631 * Does not apply to IPPROTO_ICMPV6 1632 */ 1633 if (icmp->icmp_proto == IPPROTO_ICMPV6) 1634 return (-1); 1635 1636 if (icmp->icmp_raw_checksum) { 1637 *i1 = icmp->icmp_checksum_off; 1638 } else { 1639 *i1 = -1; 1640 } 1641 break; 1642 case IPV6_JOIN_GROUP: 1643 case IPV6_LEAVE_GROUP: 1644 case MCAST_JOIN_GROUP: 1645 case MCAST_LEAVE_GROUP: 1646 case MCAST_BLOCK_SOURCE: 1647 case MCAST_UNBLOCK_SOURCE: 1648 case MCAST_JOIN_SOURCE_GROUP: 1649 case MCAST_LEAVE_SOURCE_GROUP: 1650 /* cannot "get" the value for these */ 1651 return (-1); 1652 case IPV6_RECVPKTINFO: 1653 *i1 = icmp->icmp_ipv6_recvpktinfo; 1654 break; 1655 case IPV6_RECVTCLASS: 1656 *i1 = icmp->icmp_ipv6_recvtclass; 1657 break; 1658 case IPV6_RECVPATHMTU: 1659 *i1 = icmp->icmp_ipv6_recvpathmtu; 1660 break; 1661 case IPV6_V6ONLY: 1662 *i1 = 1; 1663 break; 1664 case IPV6_RECVHOPLIMIT: 1665 *i1 = icmp->icmp_ipv6_recvhoplimit; 1666 break; 1667 case IPV6_RECVHOPOPTS: 1668 *i1 = icmp->icmp_ipv6_recvhopopts; 1669 break; 1670 case IPV6_RECVDSTOPTS: 1671 *i1 = icmp->icmp_ipv6_recvdstopts; 1672 break; 1673 case _OLD_IPV6_RECVDSTOPTS: 1674 *i1 = icmp->icmp_old_ipv6_recvdstopts; 1675 break; 1676 case IPV6_RECVRTHDRDSTOPTS: 1677 *i1 = icmp->icmp_ipv6_recvrtdstopts; 1678 break; 1679 case IPV6_RECVRTHDR: 1680 *i1 = icmp->icmp_ipv6_recvrthdr; 1681 break; 1682 case IPV6_PKTINFO: { 1683 /* XXX assumes that caller has room for max size! */ 1684 struct in6_pktinfo *pkti; 1685 1686 pkti = (struct in6_pktinfo *)ptr; 1687 if (ipp->ipp_fields & IPPF_IFINDEX) 1688 pkti->ipi6_ifindex = ipp->ipp_ifindex; 1689 else 1690 pkti->ipi6_ifindex = 0; 1691 if (ipp->ipp_fields & IPPF_ADDR) 1692 pkti->ipi6_addr = ipp->ipp_addr; 1693 else 1694 pkti->ipi6_addr = ipv6_all_zeros; 1695 return (sizeof (struct in6_pktinfo)); 1696 } 1697 case IPV6_NEXTHOP: { 1698 sin6_t *sin6 = (sin6_t *)ptr; 1699 1700 if (!(ipp->ipp_fields & IPPF_NEXTHOP)) 1701 return (0); 1702 *sin6 = sin6_null; 1703 sin6->sin6_family = AF_INET6; 1704 sin6->sin6_addr = ipp->ipp_nexthop; 1705 return (sizeof (sin6_t)); 1706 } 1707 case IPV6_HOPOPTS: 1708 if (!(ipp->ipp_fields & IPPF_HOPOPTS)) 1709 return (0); 1710 bcopy(ipp->ipp_hopopts, ptr, ipp->ipp_hopoptslen); 1711 return (ipp->ipp_hopoptslen); 1712 case IPV6_RTHDRDSTOPTS: 1713 if (!(ipp->ipp_fields & IPPF_RTDSTOPTS)) 1714 return (0); 1715 bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen); 1716 return (ipp->ipp_rtdstoptslen); 1717 case IPV6_RTHDR: 1718 if (!(ipp->ipp_fields & IPPF_RTHDR)) 1719 return (0); 1720 bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen); 1721 return (ipp->ipp_rthdrlen); 1722 case IPV6_DSTOPTS: 1723 if (!(ipp->ipp_fields & IPPF_DSTOPTS)) 1724 return (0); 1725 bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen); 1726 return (ipp->ipp_dstoptslen); 1727 case IPV6_PATHMTU: 1728 if (!(ipp->ipp_fields & IPPF_PATHMTU)) 1729 return (0); 1730 1731 return (ip_fill_mtuinfo(&icmp->icmp_v6dst, 0, 1732 (struct ip6_mtuinfo *)ptr)); 1733 case IPV6_TCLASS: 1734 if (ipp->ipp_fields & IPPF_TCLASS) 1735 *i1 = ipp->ipp_tclass; 1736 else 1737 *i1 = IPV6_FLOW_TCLASS( 1738 IPV6_DEFAULT_VERS_AND_FLOW); 1739 break; 1740 default: 1741 return (-1); 1742 } 1743 break; 1744 case IPPROTO_ICMPV6: 1745 /* 1746 * Only allow IPv6 option processing on native IPv6 sockets. 1747 */ 1748 if (icmp->icmp_family != AF_INET6) 1749 return (-1); 1750 1751 if (icmp->icmp_proto != IPPROTO_ICMPV6) 1752 return (-1); 1753 1754 switch (name) { 1755 case ICMP6_FILTER: 1756 if (icmp->icmp_filter == NULL) { 1757 /* Make it look like "pass all" */ 1758 ICMP6_FILTER_SETPASSALL((icmp6_filter_t *)ptr); 1759 } else { 1760 (void) bcopy(icmp->icmp_filter, ptr, 1761 sizeof (icmp6_filter_t)); 1762 } 1763 return (sizeof (icmp6_filter_t)); 1764 default: 1765 return (-1); 1766 } 1767 default: 1768 return (-1); 1769 } 1770 return (sizeof (int)); 1771 } 1772 1773 /* This routine sets socket options. */ 1774 /* ARGSUSED */ 1775 int 1776 icmp_opt_set(queue_t *q, uint_t optset_context, int level, int name, 1777 uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp, 1778 void *thisdg_attrs, cred_t *cr, mblk_t *mblk) 1779 { 1780 icmp_t *icmp = (icmp_t *)q->q_ptr; 1781 int *i1 = (int *)invalp; 1782 boolean_t onoff = (*i1 == 0) ? 0 : 1; 1783 boolean_t checkonly; 1784 int error; 1785 1786 switch (optset_context) { 1787 case SETFN_OPTCOM_CHECKONLY: 1788 checkonly = B_TRUE; 1789 /* 1790 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ 1791 * inlen != 0 implies value supplied and 1792 * we have to "pretend" to set it. 1793 * inlen == 0 implies that there is no 1794 * value part in T_CHECK request and just validation 1795 * done elsewhere should be enough, we just return here. 1796 */ 1797 if (inlen == 0) { 1798 *outlenp = 0; 1799 return (0); 1800 } 1801 break; 1802 case SETFN_OPTCOM_NEGOTIATE: 1803 checkonly = B_FALSE; 1804 break; 1805 case SETFN_UD_NEGOTIATE: 1806 case SETFN_CONN_NEGOTIATE: 1807 checkonly = B_FALSE; 1808 /* 1809 * Negotiating local and "association-related" options 1810 * through T_UNITDATA_REQ. 1811 * 1812 * Following routine can filter out ones we do not 1813 * want to be "set" this way. 1814 */ 1815 if (!icmp_opt_allow_udr_set(level, name)) { 1816 *outlenp = 0; 1817 return (EINVAL); 1818 } 1819 break; 1820 default: 1821 /* 1822 * We should never get here 1823 */ 1824 *outlenp = 0; 1825 return (EINVAL); 1826 } 1827 1828 ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) || 1829 (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0)); 1830 1831 /* 1832 * For fixed length options, no sanity check 1833 * of passed in length is done. It is assumed *_optcom_req() 1834 * routines do the right thing. 1835 */ 1836 1837 switch (level) { 1838 case SOL_SOCKET: 1839 switch (name) { 1840 case SO_DEBUG: 1841 if (!checkonly) 1842 icmp->icmp_debug = onoff; 1843 break; 1844 case SO_PROTOTYPE: 1845 if ((*i1 & 0xFF) != IPPROTO_ICMP && 1846 (*i1 & 0xFF) != IPPROTO_ICMPV6 && 1847 secpolicy_net_rawaccess(cr) != 0) { 1848 *outlenp = 0; 1849 return (EACCES); 1850 } 1851 /* Can't use IPPROTO_RAW with IPv6 */ 1852 if ((*i1 & 0xFF) == IPPROTO_RAW && 1853 icmp->icmp_family == AF_INET6) { 1854 *outlenp = 0; 1855 return (EPROTONOSUPPORT); 1856 } 1857 if (checkonly) { 1858 /* T_CHECK case */ 1859 *(int *)outvalp = (*i1 & 0xFF); 1860 break; 1861 } 1862 icmp->icmp_proto = *i1 & 0xFF; 1863 if ((icmp->icmp_proto == IPPROTO_RAW || 1864 icmp->icmp_proto == IPPROTO_IGMP) && 1865 icmp->icmp_family == AF_INET) 1866 icmp->icmp_hdrincl = 1; 1867 else 1868 icmp->icmp_hdrincl = 0; 1869 1870 if (icmp->icmp_family == AF_INET6 && 1871 icmp->icmp_proto == IPPROTO_ICMPV6) { 1872 /* Set offset for icmp6_cksum */ 1873 icmp->icmp_raw_checksum = 0; 1874 icmp->icmp_checksum_off = 2; 1875 } 1876 if (icmp->icmp_proto == IPPROTO_UDP || 1877 icmp->icmp_proto == IPPROTO_TCP || 1878 icmp->icmp_proto == IPPROTO_SCTP) { 1879 icmp->icmp_no_tp_cksum = 1; 1880 icmp->icmp_sticky_ipp.ipp_fields |= 1881 IPPF_NO_CKSUM; 1882 } else { 1883 icmp->icmp_no_tp_cksum = 0; 1884 icmp->icmp_sticky_ipp.ipp_fields &= 1885 ~IPPF_NO_CKSUM; 1886 } 1887 1888 if (icmp->icmp_filter != NULL && 1889 icmp->icmp_proto != IPPROTO_ICMPV6) { 1890 kmem_free(icmp->icmp_filter, 1891 sizeof (icmp6_filter_t)); 1892 icmp->icmp_filter = NULL; 1893 } 1894 1895 /* Rebuild the header template */ 1896 error = icmp_build_hdrs(q, icmp); 1897 if (error != 0) { 1898 *outlenp = 0; 1899 return (error); 1900 } 1901 1902 /* 1903 * For SCTP, we don't use icmp_bind_proto() for 1904 * raw socket binding. Note that we do not need 1905 * to set *outlenp. 1906 */ 1907 if (icmp->icmp_proto == IPPROTO_SCTP) 1908 return (0); 1909 1910 icmp_bind_proto(q); 1911 *outlenp = sizeof (int); 1912 *(int *)outvalp = *i1 & 0xFF; 1913 return (0); 1914 case SO_REUSEADDR: 1915 if (!checkonly) 1916 icmp->icmp_reuseaddr = onoff; 1917 break; 1918 1919 /* 1920 * The following three items are available here, 1921 * but are only meaningful to IP. 1922 */ 1923 case SO_DONTROUTE: 1924 if (!checkonly) 1925 icmp->icmp_dontroute = onoff; 1926 break; 1927 case SO_USELOOPBACK: 1928 if (!checkonly) 1929 icmp->icmp_useloopback = onoff; 1930 break; 1931 case SO_BROADCAST: 1932 if (!checkonly) 1933 icmp->icmp_broadcast = onoff; 1934 break; 1935 1936 case SO_SNDBUF: 1937 if (*i1 > icmp_max_buf) { 1938 *outlenp = 0; 1939 return (ENOBUFS); 1940 } 1941 if (!checkonly) { 1942 q->q_hiwat = *i1; 1943 q->q_next->q_hiwat = *i1; 1944 } 1945 break; 1946 case SO_RCVBUF: 1947 if (*i1 > icmp_max_buf) { 1948 *outlenp = 0; 1949 return (ENOBUFS); 1950 } 1951 if (!checkonly) { 1952 RD(q)->q_hiwat = *i1; 1953 (void) mi_set_sth_hiwat(RD(q), *i1); 1954 } 1955 break; 1956 case SO_DGRAM_ERRIND: 1957 if (!checkonly) 1958 icmp->icmp_dgram_errind = onoff; 1959 break; 1960 /* 1961 * Following three not meaningful for icmp 1962 * Action is same as "default" so we keep them 1963 * in comments. 1964 * case SO_LINGER: 1965 * case SO_KEEPALIVE: 1966 * case SO_OOBINLINE: 1967 */ 1968 default: 1969 *outlenp = 0; 1970 return (EINVAL); 1971 } 1972 break; 1973 case IPPROTO_IP: 1974 /* 1975 * Only allow IPv4 option processing on IPv4 sockets. 1976 */ 1977 if (icmp->icmp_family != AF_INET) { 1978 *outlenp = 0; 1979 return (ENOPROTOOPT); 1980 } 1981 switch (name) { 1982 case IP_OPTIONS: 1983 case T_IP_OPTIONS: 1984 /* Save options for use by IP. */ 1985 if (inlen & 0x3) { 1986 *outlenp = 0; 1987 return (EINVAL); 1988 } 1989 if (checkonly) 1990 break; 1991 1992 if (icmp->icmp_ip_snd_options) { 1993 mi_free((char *)icmp->icmp_ip_snd_options); 1994 icmp->icmp_ip_snd_options_len = 0; 1995 icmp->icmp_ip_snd_options = NULL; 1996 } 1997 if (inlen) { 1998 icmp->icmp_ip_snd_options = 1999 (uchar_t *)mi_alloc(inlen, BPRI_HI); 2000 if (icmp->icmp_ip_snd_options) { 2001 bcopy(invalp, 2002 icmp->icmp_ip_snd_options, inlen); 2003 icmp->icmp_ip_snd_options_len = inlen; 2004 } 2005 } 2006 icmp->icmp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + 2007 icmp->icmp_ip_snd_options_len; 2008 (void) mi_set_sth_wroff(RD(q), icmp->icmp_max_hdr_len + 2009 icmp_wroff_extra); 2010 break; 2011 case IP_HDRINCL: 2012 if (!checkonly) 2013 icmp->icmp_hdrincl = onoff; 2014 break; 2015 case IP_TOS: 2016 case T_IP_TOS: 2017 if (!checkonly) { 2018 icmp->icmp_type_of_service = (uint8_t)*i1; 2019 } 2020 break; 2021 case IP_TTL: 2022 if (!checkonly) { 2023 icmp->icmp_ttl = (uint8_t)*i1; 2024 } 2025 break; 2026 case IP_MULTICAST_IF: 2027 /* 2028 * TODO should check OPTMGMT reply and undo this if 2029 * there is an error. 2030 */ 2031 if (!checkonly) 2032 icmp->icmp_multicast_if_addr = *i1; 2033 break; 2034 case IP_MULTICAST_TTL: 2035 if (!checkonly) 2036 icmp->icmp_multicast_ttl = *invalp; 2037 break; 2038 case IP_MULTICAST_LOOP: 2039 if (!checkonly) { 2040 icmp->icmp_multicast_loop = 2041 (*invalp == 0) ? 0 : 1; 2042 } 2043 break; 2044 case IP_BOUND_IF: 2045 if (!checkonly) 2046 icmp->icmp_bound_if = *i1; 2047 break; 2048 case IP_UNSPEC_SRC: 2049 if (!checkonly) 2050 icmp->icmp_unspec_source = onoff; 2051 break; 2052 case IP_XMIT_IF: 2053 if (!checkonly) 2054 icmp->icmp_xmit_if = *i1; 2055 break; 2056 case IP_RECVIF: 2057 if (!checkonly) 2058 icmp->icmp_recvif = onoff; 2059 break; 2060 case IP_ADD_MEMBERSHIP: 2061 case IP_DROP_MEMBERSHIP: 2062 case IP_BLOCK_SOURCE: 2063 case IP_UNBLOCK_SOURCE: 2064 case IP_ADD_SOURCE_MEMBERSHIP: 2065 case IP_DROP_SOURCE_MEMBERSHIP: 2066 case MCAST_JOIN_GROUP: 2067 case MCAST_LEAVE_GROUP: 2068 case MCAST_BLOCK_SOURCE: 2069 case MCAST_UNBLOCK_SOURCE: 2070 case MCAST_JOIN_SOURCE_GROUP: 2071 case MCAST_LEAVE_SOURCE_GROUP: 2072 case MRT_INIT: 2073 case MRT_DONE: 2074 case MRT_ADD_VIF: 2075 case MRT_DEL_VIF: 2076 case MRT_ADD_MFC: 2077 case MRT_DEL_MFC: 2078 case MRT_VERSION: 2079 case MRT_ASSERT: 2080 case IP_SEC_OPT: 2081 case IP_DONTFAILOVER_IF: 2082 /* 2083 * "soft" error (negative) 2084 * option not handled at this level 2085 * Note: Do not modify *outlenp 2086 */ 2087 return (-EINVAL); 2088 default: 2089 *outlenp = 0; 2090 return (EINVAL); 2091 } 2092 break; 2093 case IPPROTO_IPV6: { 2094 ip6_pkt_t *ipp; 2095 boolean_t sticky; 2096 2097 if (icmp->icmp_family != AF_INET6) { 2098 *outlenp = 0; 2099 return (ENOPROTOOPT); 2100 } 2101 /* 2102 * Deal with both sticky options and ancillary data 2103 */ 2104 if (thisdg_attrs == NULL) { 2105 /* sticky options, or none */ 2106 ipp = &icmp->icmp_sticky_ipp; 2107 sticky = B_TRUE; 2108 } else { 2109 /* ancillary data */ 2110 ipp = (ip6_pkt_t *)thisdg_attrs; 2111 sticky = B_FALSE; 2112 } 2113 2114 switch (name) { 2115 case IPV6_MULTICAST_IF: 2116 if (!checkonly) 2117 icmp->icmp_multicast_if_index = *i1; 2118 break; 2119 case IPV6_UNICAST_HOPS: 2120 /* -1 means use default */ 2121 if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) { 2122 *outlenp = 0; 2123 return (EINVAL); 2124 } 2125 if (!checkonly) { 2126 if (*i1 == -1) { 2127 icmp->icmp_ttl = ipp->ipp_unicast_hops = 2128 icmp_ipv6_hoplimit; 2129 ipp->ipp_fields &= ~IPPF_UNICAST_HOPS; 2130 /* Pass modified value to IP. */ 2131 *i1 = ipp->ipp_hoplimit; 2132 } else { 2133 icmp->icmp_ttl = ipp->ipp_unicast_hops = 2134 (uint8_t)*i1; 2135 ipp->ipp_fields |= IPPF_UNICAST_HOPS; 2136 } 2137 /* Rebuild the header template */ 2138 error = icmp_build_hdrs(q, icmp); 2139 if (error != 0) { 2140 *outlenp = 0; 2141 return (error); 2142 } 2143 } 2144 break; 2145 case IPV6_MULTICAST_HOPS: 2146 /* -1 means use default */ 2147 if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) { 2148 *outlenp = 0; 2149 return (EINVAL); 2150 } 2151 if (!checkonly) { 2152 if (*i1 == -1) { 2153 icmp->icmp_multicast_ttl = 2154 ipp->ipp_multicast_hops = 2155 IP_DEFAULT_MULTICAST_TTL; 2156 ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS; 2157 /* Pass modified value to IP. */ 2158 *i1 = icmp->icmp_multicast_ttl; 2159 } else { 2160 icmp->icmp_multicast_ttl = 2161 ipp->ipp_multicast_hops = 2162 (uint8_t)*i1; 2163 ipp->ipp_fields |= IPPF_MULTICAST_HOPS; 2164 } 2165 } 2166 break; 2167 case IPV6_MULTICAST_LOOP: 2168 if (*i1 != 0 && *i1 != 1) { 2169 *outlenp = 0; 2170 return (EINVAL); 2171 } 2172 if (!checkonly) 2173 icmp->icmp_multicast_loop = *i1; 2174 break; 2175 case IPV6_CHECKSUM: 2176 /* 2177 * Integer offset into the user data of where the 2178 * checksum is located. 2179 * Offset of -1 disables option. 2180 * Does not apply to IPPROTO_ICMPV6. 2181 */ 2182 if (icmp->icmp_proto == IPPROTO_ICMPV6 || !sticky) { 2183 *outlenp = 0; 2184 return (EINVAL); 2185 } 2186 if ((*i1 != -1) && ((*i1 < 0) || (*i1 & 0x1) != 0)) { 2187 /* Negative or not 16 bit aligned offset */ 2188 *outlenp = 0; 2189 return (EINVAL); 2190 } 2191 if (checkonly) 2192 break; 2193 2194 if (*i1 == -1) { 2195 icmp->icmp_raw_checksum = 0; 2196 ipp->ipp_fields &= ~IPPF_RAW_CKSUM; 2197 } else { 2198 icmp->icmp_raw_checksum = 1; 2199 icmp->icmp_checksum_off = *i1; 2200 ipp->ipp_fields |= IPPF_RAW_CKSUM; 2201 } 2202 /* Rebuild the header template */ 2203 error = icmp_build_hdrs(q, icmp); 2204 if (error != 0) { 2205 *outlenp = 0; 2206 return (error); 2207 } 2208 break; 2209 case IPV6_JOIN_GROUP: 2210 case IPV6_LEAVE_GROUP: 2211 case MCAST_JOIN_GROUP: 2212 case MCAST_LEAVE_GROUP: 2213 case MCAST_BLOCK_SOURCE: 2214 case MCAST_UNBLOCK_SOURCE: 2215 case MCAST_JOIN_SOURCE_GROUP: 2216 case MCAST_LEAVE_SOURCE_GROUP: 2217 /* 2218 * "soft" error (negative) 2219 * option not handled at this level 2220 * Note: Do not modify *outlenp 2221 */ 2222 return (-EINVAL); 2223 case IPV6_BOUND_IF: 2224 if (!checkonly) 2225 icmp->icmp_bound_if = *i1; 2226 break; 2227 case IPV6_UNSPEC_SRC: 2228 if (!checkonly) 2229 icmp->icmp_unspec_source = onoff; 2230 break; 2231 case IPV6_RECVTCLASS: 2232 if (!checkonly) 2233 icmp->icmp_ipv6_recvtclass = onoff; 2234 break; 2235 /* 2236 * Set boolean switches for ancillary data delivery 2237 */ 2238 case IPV6_RECVPKTINFO: 2239 if (!checkonly) 2240 icmp->icmp_ipv6_recvpktinfo = onoff; 2241 break; 2242 case IPV6_RECVPATHMTU: 2243 if (!checkonly) 2244 icmp->icmp_ipv6_recvpathmtu = onoff; 2245 break; 2246 case IPV6_RECVHOPLIMIT: 2247 if (!checkonly) 2248 icmp->icmp_ipv6_recvhoplimit = onoff; 2249 break; 2250 case IPV6_RECVHOPOPTS: 2251 if (!checkonly) 2252 icmp->icmp_ipv6_recvhopopts = onoff; 2253 break; 2254 case IPV6_RECVDSTOPTS: 2255 if (!checkonly) 2256 icmp->icmp_ipv6_recvdstopts = onoff; 2257 break; 2258 case _OLD_IPV6_RECVDSTOPTS: 2259 if (!checkonly) 2260 icmp->icmp_old_ipv6_recvdstopts = onoff; 2261 break; 2262 case IPV6_RECVRTHDRDSTOPTS: 2263 if (!checkonly) 2264 icmp->icmp_ipv6_recvrtdstopts = onoff; 2265 break; 2266 case IPV6_RECVRTHDR: 2267 if (!checkonly) 2268 icmp->icmp_ipv6_recvrthdr = onoff; 2269 break; 2270 /* 2271 * Set sticky options or ancillary data. 2272 * If sticky options, (re)build any extension headers 2273 * that might be needed as a result. 2274 */ 2275 case IPV6_PKTINFO: 2276 /* 2277 * The source address and ifindex are verified 2278 * in ip_opt_set(). For ancillary data the 2279 * source address is checked in ip_wput_v6. 2280 */ 2281 if (inlen != 0 && inlen != sizeof (struct in6_pktinfo)) 2282 return (EINVAL); 2283 if (checkonly) 2284 break; 2285 2286 if (inlen == 0) { 2287 ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR); 2288 ipp->ipp_sticky_ignored |= 2289 (IPPF_IFINDEX|IPPF_ADDR); 2290 } else { 2291 struct in6_pktinfo *pkti; 2292 2293 pkti = (struct in6_pktinfo *)invalp; 2294 ipp->ipp_ifindex = pkti->ipi6_ifindex; 2295 ipp->ipp_addr = pkti->ipi6_addr; 2296 if (ipp->ipp_ifindex != 0) 2297 ipp->ipp_fields |= IPPF_IFINDEX; 2298 else 2299 ipp->ipp_fields &= ~IPPF_IFINDEX; 2300 if (!IN6_IS_ADDR_UNSPECIFIED( 2301 &ipp->ipp_addr)) 2302 ipp->ipp_fields |= IPPF_ADDR; 2303 else 2304 ipp->ipp_fields &= ~IPPF_ADDR; 2305 } 2306 if (sticky) { 2307 error = icmp_build_hdrs(q, icmp); 2308 if (error != 0) 2309 return (error); 2310 } 2311 break; 2312 case IPV6_HOPLIMIT: 2313 /* This option can only be used as ancillary data. */ 2314 if (sticky) 2315 return (EINVAL); 2316 if (inlen != 0 && inlen != sizeof (int)) 2317 return (EINVAL); 2318 if (checkonly) 2319 break; 2320 2321 if (inlen == 0) { 2322 ipp->ipp_fields &= ~IPPF_HOPLIMIT; 2323 ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT; 2324 } else { 2325 if (*i1 > 255 || *i1 < -1) 2326 return (EINVAL); 2327 if (*i1 == -1) 2328 ipp->ipp_hoplimit = icmp_ipv6_hoplimit; 2329 else 2330 ipp->ipp_hoplimit = *i1; 2331 ipp->ipp_fields |= IPPF_HOPLIMIT; 2332 } 2333 break; 2334 case IPV6_TCLASS: 2335 /* 2336 * IPV6_RECVTCLASS accepts -1 as use kernel default 2337 * and [0, 255] as the actualy traffic class. 2338 */ 2339 if (inlen != 0 && inlen != sizeof (int)) 2340 return (EINVAL); 2341 if (checkonly) 2342 break; 2343 2344 if (inlen == 0) { 2345 ipp->ipp_fields &= ~IPPF_TCLASS; 2346 ipp->ipp_sticky_ignored |= IPPF_TCLASS; 2347 } else { 2348 if (*i1 >= 256 || *i1 < -1) 2349 return (EINVAL); 2350 if (*i1 == -1) { 2351 ipp->ipp_tclass = 2352 IPV6_FLOW_TCLASS( 2353 IPV6_DEFAULT_VERS_AND_FLOW); 2354 } else { 2355 ipp->ipp_tclass = *i1; 2356 } 2357 ipp->ipp_fields |= IPPF_TCLASS; 2358 } 2359 if (sticky) { 2360 error = icmp_build_hdrs(q, icmp); 2361 if (error != 0) 2362 return (error); 2363 } 2364 break; 2365 case IPV6_NEXTHOP: 2366 /* 2367 * IP will verify that the nexthop is reachable 2368 * and fail for sticky options. 2369 */ 2370 if (inlen != 0 && inlen != sizeof (sin6_t)) 2371 return (EINVAL); 2372 if (checkonly) 2373 break; 2374 2375 if (inlen == 0) { 2376 ipp->ipp_fields &= ~IPPF_NEXTHOP; 2377 ipp->ipp_sticky_ignored |= IPPF_NEXTHOP; 2378 } else { 2379 sin6_t *sin6 = (sin6_t *)invalp; 2380 2381 if (sin6->sin6_family != AF_INET6) 2382 return (EAFNOSUPPORT); 2383 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) 2384 return (EADDRNOTAVAIL); 2385 ipp->ipp_nexthop = sin6->sin6_addr; 2386 if (!IN6_IS_ADDR_UNSPECIFIED( 2387 &ipp->ipp_nexthop)) 2388 ipp->ipp_fields |= IPPF_NEXTHOP; 2389 else 2390 ipp->ipp_fields &= ~IPPF_NEXTHOP; 2391 } 2392 if (sticky) { 2393 error = icmp_build_hdrs(q, icmp); 2394 if (error != 0) 2395 return (error); 2396 } 2397 break; 2398 case IPV6_HOPOPTS: { 2399 ip6_hbh_t *hopts = (ip6_hbh_t *)invalp; 2400 /* 2401 * Sanity checks - minimum size, size a multiple of 2402 * eight bytes, and matching size passed in. 2403 */ 2404 if (inlen != 0 && 2405 inlen != (8 * (hopts->ip6h_len + 1))) 2406 return (EINVAL); 2407 2408 if (checkonly) 2409 break; 2410 if (inlen == 0) { 2411 if (sticky && 2412 (ipp->ipp_fields & IPPF_HOPOPTS) != 0) { 2413 kmem_free(ipp->ipp_hopopts, 2414 ipp->ipp_hopoptslen); 2415 ipp->ipp_hopopts = NULL; 2416 ipp->ipp_hopoptslen = 0; 2417 } 2418 ipp->ipp_fields &= ~IPPF_HOPOPTS; 2419 ipp->ipp_sticky_ignored |= IPPF_HOPOPTS; 2420 } else { 2421 error = icmp_pkt_set(invalp, inlen, sticky, 2422 (uchar_t **)&ipp->ipp_hopopts, 2423 &ipp->ipp_hopoptslen); 2424 if (error != 0) 2425 return (error); 2426 ipp->ipp_fields |= IPPF_HOPOPTS; 2427 } 2428 if (sticky) { 2429 error = icmp_build_hdrs(q, icmp); 2430 if (error != 0) 2431 return (error); 2432 } 2433 break; 2434 } 2435 case IPV6_RTHDRDSTOPTS: { 2436 ip6_dest_t *dopts = (ip6_dest_t *)invalp; 2437 2438 /* 2439 * Sanity checks - minimum size, size a multiple of 2440 * eight bytes, and matching size passed in. 2441 */ 2442 if (inlen != 0 && 2443 inlen != (8 * (dopts->ip6d_len + 1))) 2444 return (EINVAL); 2445 2446 if (checkonly) 2447 break; 2448 2449 if (inlen == 0) { 2450 if (sticky && 2451 (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) { 2452 kmem_free(ipp->ipp_rtdstopts, 2453 ipp->ipp_rtdstoptslen); 2454 ipp->ipp_rtdstopts = NULL; 2455 ipp->ipp_rtdstoptslen = 0; 2456 } 2457 ipp->ipp_fields &= ~IPPF_RTDSTOPTS; 2458 ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS; 2459 } else { 2460 error = icmp_pkt_set(invalp, inlen, sticky, 2461 (uchar_t **)&ipp->ipp_rtdstopts, 2462 &ipp->ipp_rtdstoptslen); 2463 if (error != 0) 2464 return (error); 2465 ipp->ipp_fields |= IPPF_RTDSTOPTS; 2466 } 2467 if (sticky) { 2468 error = icmp_build_hdrs(q, icmp); 2469 if (error != 0) 2470 return (error); 2471 } 2472 break; 2473 } 2474 case IPV6_DSTOPTS: { 2475 ip6_dest_t *dopts = (ip6_dest_t *)invalp; 2476 2477 /* 2478 * Sanity checks - minimum size, size a multiple of 2479 * eight bytes, and matching size passed in. 2480 */ 2481 if (inlen != 0 && 2482 inlen != (8 * (dopts->ip6d_len + 1))) 2483 return (EINVAL); 2484 2485 if (checkonly) 2486 break; 2487 2488 if (inlen == 0) { 2489 if (sticky && 2490 (ipp->ipp_fields & IPPF_DSTOPTS) != 0) { 2491 kmem_free(ipp->ipp_dstopts, 2492 ipp->ipp_dstoptslen); 2493 ipp->ipp_dstopts = NULL; 2494 ipp->ipp_dstoptslen = 0; 2495 } 2496 ipp->ipp_fields &= ~IPPF_DSTOPTS; 2497 ipp->ipp_sticky_ignored |= IPPF_DSTOPTS; 2498 } else { 2499 error = icmp_pkt_set(invalp, inlen, sticky, 2500 (uchar_t **)&ipp->ipp_dstopts, 2501 &ipp->ipp_dstoptslen); 2502 if (error != 0) 2503 return (error); 2504 ipp->ipp_fields |= IPPF_DSTOPTS; 2505 } 2506 if (sticky) { 2507 error = icmp_build_hdrs(q, icmp); 2508 if (error != 0) 2509 return (error); 2510 } 2511 break; 2512 } 2513 case IPV6_RTHDR: { 2514 ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp; 2515 2516 /* 2517 * Sanity checks - minimum size, size a multiple of 2518 * eight bytes, and matching size passed in. 2519 */ 2520 if (inlen != 0 && 2521 inlen != (8 * (rt->ip6r_len + 1))) 2522 return (EINVAL); 2523 2524 if (checkonly) 2525 break; 2526 2527 if (inlen == 0) { 2528 if (sticky && 2529 (ipp->ipp_fields & IPPF_RTHDR) != 0) { 2530 kmem_free(ipp->ipp_rthdr, 2531 ipp->ipp_rthdrlen); 2532 ipp->ipp_rthdr = NULL; 2533 ipp->ipp_rthdrlen = 0; 2534 } 2535 ipp->ipp_fields &= ~IPPF_RTHDR; 2536 ipp->ipp_sticky_ignored |= IPPF_RTHDR; 2537 } else { 2538 error = icmp_pkt_set(invalp, inlen, sticky, 2539 (uchar_t **)&ipp->ipp_rthdr, 2540 &ipp->ipp_rthdrlen); 2541 if (error != 0) 2542 return (error); 2543 ipp->ipp_fields |= IPPF_RTHDR; 2544 } 2545 if (sticky) { 2546 error = icmp_build_hdrs(q, icmp); 2547 if (error != 0) 2548 return (error); 2549 } 2550 break; 2551 } 2552 2553 case IPV6_DONTFRAG: 2554 if (checkonly) 2555 break; 2556 2557 if (onoff) { 2558 ipp->ipp_fields |= IPPF_DONTFRAG; 2559 } else { 2560 ipp->ipp_fields &= ~IPPF_DONTFRAG; 2561 } 2562 break; 2563 2564 case IPV6_USE_MIN_MTU: 2565 if (inlen != sizeof (int)) 2566 return (EINVAL); 2567 2568 if (*i1 < -1 || *i1 > 1) 2569 return (EINVAL); 2570 2571 if (checkonly) 2572 break; 2573 2574 ipp->ipp_fields |= IPPF_USE_MIN_MTU; 2575 ipp->ipp_use_min_mtu = *i1; 2576 break; 2577 2578 /* 2579 * This option can't be set. Its only returned via 2580 * getsockopt() or ancillary data. 2581 */ 2582 case IPV6_PATHMTU: 2583 return (EINVAL); 2584 2585 case IPV6_BOUND_PIF: 2586 case IPV6_SEC_OPT: 2587 case IPV6_DONTFAILOVER_IF: 2588 case IPV6_SRC_PREFERENCES: 2589 case IPV6_V6ONLY: 2590 /* Handled at IP level */ 2591 return (-EINVAL); 2592 default: 2593 *outlenp = 0; 2594 return (EINVAL); 2595 } 2596 break; 2597 } /* end IPPROTO_IPV6 */ 2598 2599 case IPPROTO_ICMPV6: 2600 /* 2601 * Only allow IPv6 option processing on IPv6 sockets. 2602 */ 2603 if (icmp->icmp_family != AF_INET6) { 2604 *outlenp = 0; 2605 return (ENOPROTOOPT); 2606 } 2607 if (icmp->icmp_proto != IPPROTO_ICMPV6) { 2608 *outlenp = 0; 2609 return (ENOPROTOOPT); 2610 } 2611 switch (name) { 2612 case ICMP6_FILTER: 2613 if (!checkonly) { 2614 if ((inlen != 0) && 2615 (inlen != sizeof (icmp6_filter_t))) 2616 return (EINVAL); 2617 2618 if (inlen == 0) { 2619 if (icmp->icmp_filter != NULL) { 2620 kmem_free(icmp->icmp_filter, 2621 sizeof (icmp6_filter_t)); 2622 icmp->icmp_filter = NULL; 2623 } 2624 } else { 2625 if (icmp->icmp_filter == NULL) { 2626 icmp->icmp_filter = kmem_alloc( 2627 sizeof (icmp6_filter_t), 2628 KM_NOSLEEP); 2629 if (icmp->icmp_filter == NULL) { 2630 *outlenp = 0; 2631 return (ENOBUFS); 2632 } 2633 } 2634 (void) bcopy(invalp, icmp->icmp_filter, 2635 inlen); 2636 } 2637 } 2638 break; 2639 2640 default: 2641 *outlenp = 0; 2642 return (EINVAL); 2643 } 2644 break; 2645 default: 2646 *outlenp = 0; 2647 return (EINVAL); 2648 } 2649 /* 2650 * Common case of OK return with outval same as inval. 2651 */ 2652 if (invalp != outvalp) { 2653 /* don't trust bcopy for identical src/dst */ 2654 (void) bcopy(invalp, outvalp, inlen); 2655 } 2656 *outlenp = inlen; 2657 return (0); 2658 } 2659 2660 /* 2661 * Update icmp_sticky_hdrs based on icmp_sticky_ipp, icmp_v6src, icmp_ttl, 2662 * icmp_proto, icmp_raw_checksum and icmp_no_tp_cksum. 2663 * The headers include ip6i_t (if needed), ip6_t, and any sticky extension 2664 * headers. 2665 * Returns failure if can't allocate memory. 2666 */ 2667 static int 2668 icmp_build_hdrs(queue_t *q, icmp_t *icmp) 2669 { 2670 uchar_t *hdrs; 2671 uint_t hdrs_len; 2672 ip6_t *ip6h; 2673 ip6i_t *ip6i; 2674 ip6_pkt_t *ipp = &icmp->icmp_sticky_ipp; 2675 2676 hdrs_len = ip_total_hdrs_len_v6(ipp); 2677 ASSERT(hdrs_len != 0); 2678 if (hdrs_len != icmp->icmp_sticky_hdrs_len) { 2679 /* Need to reallocate */ 2680 if (hdrs_len != 0) { 2681 hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP); 2682 if (hdrs == NULL) 2683 return (ENOMEM); 2684 } else { 2685 hdrs = NULL; 2686 } 2687 if (icmp->icmp_sticky_hdrs_len != 0) { 2688 kmem_free(icmp->icmp_sticky_hdrs, 2689 icmp->icmp_sticky_hdrs_len); 2690 } 2691 icmp->icmp_sticky_hdrs = hdrs; 2692 icmp->icmp_sticky_hdrs_len = hdrs_len; 2693 } 2694 ip_build_hdrs_v6(icmp->icmp_sticky_hdrs, 2695 icmp->icmp_sticky_hdrs_len, ipp, icmp->icmp_proto); 2696 2697 /* Set header fields not in ipp */ 2698 if (ipp->ipp_fields & IPPF_HAS_IP6I) { 2699 ip6i = (ip6i_t *)icmp->icmp_sticky_hdrs; 2700 ip6h = (ip6_t *)&ip6i[1]; 2701 2702 if (ipp->ipp_fields & IPPF_RAW_CKSUM) { 2703 ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM; 2704 ip6i->ip6i_checksum_off = icmp->icmp_checksum_off; 2705 } 2706 if (ipp->ipp_fields & IPPF_NO_CKSUM) { 2707 ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM; 2708 } 2709 } else { 2710 ip6h = (ip6_t *)icmp->icmp_sticky_hdrs; 2711 } 2712 2713 if (!(ipp->ipp_fields & IPPF_ADDR)) 2714 ip6h->ip6_src = icmp->icmp_v6src; 2715 2716 /* Try to get everything in a single mblk */ 2717 if (hdrs_len > icmp->icmp_max_hdr_len) { 2718 icmp->icmp_max_hdr_len = hdrs_len; 2719 (void) mi_set_sth_wroff(RD(q), icmp->icmp_max_hdr_len + 2720 icmp_wroff_extra); 2721 } 2722 return (0); 2723 } 2724 2725 /* 2726 * Set optbuf and optlen for the option. 2727 * If sticky is set allocate memory (if not already present). 2728 * Otherwise just point optbuf and optlen at invalp and inlen. 2729 * Returns failure if memory can not be allocated. 2730 */ 2731 static int 2732 icmp_pkt_set(uchar_t *invalp, uint_t inlen, boolean_t sticky, 2733 uchar_t **optbufp, uint_t *optlenp) 2734 { 2735 uchar_t *optbuf; 2736 2737 if (!sticky) { 2738 *optbufp = invalp; 2739 *optlenp = inlen; 2740 return (0); 2741 } 2742 if (inlen == *optlenp) { 2743 /* Unchanged length - no need to realocate */ 2744 bcopy(invalp, *optbufp, inlen); 2745 return (0); 2746 } 2747 if (inlen != 0) { 2748 /* Allocate new buffer before free */ 2749 optbuf = kmem_alloc(inlen, KM_NOSLEEP); 2750 if (optbuf == NULL) 2751 return (ENOMEM); 2752 } else { 2753 optbuf = NULL; 2754 } 2755 /* Free old buffer */ 2756 if (*optlenp != 0) 2757 kmem_free(*optbufp, *optlenp); 2758 2759 bcopy(invalp, optbuf, inlen); 2760 *optbufp = optbuf; 2761 *optlenp = inlen; 2762 return (0); 2763 } 2764 2765 /* 2766 * This routine retrieves the value of an ND variable in a icmpparam_t 2767 * structure. It is called through nd_getset when a user reads the 2768 * variable. 2769 */ 2770 /* ARGSUSED */ 2771 static int 2772 icmp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr) 2773 { 2774 icmpparam_t *icmppa = (icmpparam_t *)cp; 2775 2776 (void) mi_mpprintf(mp, "%d", icmppa->icmp_param_value); 2777 return (0); 2778 } 2779 2780 /* 2781 * Walk through the param array specified registering each element with the 2782 * named dispatch (ND) handler. 2783 */ 2784 static boolean_t 2785 icmp_param_register(icmpparam_t *icmppa, int cnt) 2786 { 2787 for (; cnt-- > 0; icmppa++) { 2788 if (icmppa->icmp_param_name && icmppa->icmp_param_name[0]) { 2789 if (!nd_load(&icmp_g_nd, icmppa->icmp_param_name, 2790 icmp_param_get, icmp_param_set, 2791 (caddr_t)icmppa)) { 2792 nd_free(&icmp_g_nd); 2793 return (B_FALSE); 2794 } 2795 } 2796 } 2797 if (!nd_load(&icmp_g_nd, "icmp_status", icmp_status_report, NULL, 2798 NULL)) { 2799 nd_free(&icmp_g_nd); 2800 return (B_FALSE); 2801 } 2802 return (B_TRUE); 2803 } 2804 2805 /* This routine sets an ND variable in a icmpparam_t structure. */ 2806 /* ARGSUSED */ 2807 static int 2808 icmp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr) 2809 { 2810 long new_value; 2811 icmpparam_t *icmppa = (icmpparam_t *)cp; 2812 2813 /* 2814 * Fail the request if the new value does not lie within the 2815 * required bounds. 2816 */ 2817 if (ddi_strtol(value, NULL, 10, &new_value) != 0 || 2818 new_value < icmppa->icmp_param_min || 2819 new_value > icmppa->icmp_param_max) { 2820 return (EINVAL); 2821 } 2822 /* Set the new value */ 2823 icmppa->icmp_param_value = new_value; 2824 return (0); 2825 } 2826 2827 static void 2828 icmp_rput(queue_t *q, mblk_t *mp) 2829 { 2830 struct T_unitdata_ind *tudi; 2831 uchar_t *rptr; 2832 struct T_error_ack *tea; 2833 icmp_t *icmp; 2834 sin_t *sin; 2835 sin6_t *sin6; 2836 ip6_t *ip6h; 2837 ip6i_t *ip6i; 2838 mblk_t *mp1; 2839 int hdr_len; 2840 ipha_t *ipha; 2841 int udi_size; /* Size of T_unitdata_ind */ 2842 uint_t ipvers; 2843 ip6_pkt_t ipp; 2844 uint8_t nexthdr; 2845 boolean_t recvif = B_FALSE; 2846 in_pktinfo_t *pinfo; 2847 mblk_t *options_mp = NULL; 2848 uint_t icmp_opt = 0; 2849 boolean_t icmp_ipv6_recvhoplimit = B_FALSE; 2850 2851 icmp = (icmp_t *)q->q_ptr; 2852 if (icmp->icmp_restricted) { 2853 putnext(q, mp); 2854 return; 2855 } 2856 2857 if (mp->b_datap->db_type == M_CTL) { 2858 /* 2859 * IP sends up the IPSEC_IN message for handling IPSEC 2860 * policy at the TCP level. We don't need it here. 2861 */ 2862 if (*(uint32_t *)(mp->b_rptr) == IPSEC_IN) { 2863 mp1 = mp->b_cont; 2864 freeb(mp); 2865 mp = mp1; 2866 } else { 2867 pinfo = (in_pktinfo_t *)mp->b_rptr; 2868 if ((icmp->icmp_recvif != 0) && 2869 (pinfo->in_pkt_ulp_type == IN_PKTINFO)) { 2870 /* 2871 * IP has passed the options in mp and the 2872 * actual data is in b_cont. 2873 */ 2874 recvif = B_TRUE; 2875 /* 2876 * We are here bcos IP_RECVIF is set so we need 2877 * to extract the options mblk and adjust the 2878 * rptr 2879 */ 2880 options_mp = mp; 2881 mp = mp->b_cont; 2882 } 2883 } 2884 } 2885 2886 rptr = mp->b_rptr; 2887 switch (mp->b_datap->db_type) { 2888 case M_DATA: 2889 /* 2890 * M_DATA messages contain IP packets. They are handled 2891 * following the switch. 2892 */ 2893 break; 2894 case M_PROTO: 2895 case M_PCPROTO: 2896 /* M_PROTO messages contain some type of TPI message. */ 2897 if ((mp->b_wptr - rptr) < sizeof (t_scalar_t)) { 2898 freemsg(mp); 2899 return; 2900 } 2901 tea = (struct T_error_ack *)rptr; 2902 switch (tea->PRIM_type) { 2903 case T_ERROR_ACK: 2904 switch (tea->ERROR_prim) { 2905 case O_T_BIND_REQ: 2906 case T_BIND_REQ: 2907 /* 2908 * If our O_T_BIND_REQ/T_BIND_REQ fails, 2909 * clear out the source address before 2910 * passing the message upstream. 2911 * If this was caused by a T_CONN_REQ 2912 * revert back to bound state. 2913 */ 2914 if (icmp->icmp_state == TS_UNBND) { 2915 /* 2916 * TPI has not yet bound - bind sent by 2917 * icmp_bind_proto. 2918 */ 2919 freemsg(mp); 2920 return; 2921 } 2922 if (icmp->icmp_state == TS_DATA_XFER) { 2923 /* Connect failed */ 2924 tea->ERROR_prim = T_CONN_REQ; 2925 icmp->icmp_v6src = 2926 icmp->icmp_bound_v6src; 2927 icmp->icmp_state = TS_IDLE; 2928 if (icmp->icmp_family == AF_INET6) 2929 (void) icmp_build_hdrs(q, icmp); 2930 break; 2931 } 2932 2933 if (icmp->icmp_discon_pending) { 2934 tea->ERROR_prim = T_DISCON_REQ; 2935 icmp->icmp_discon_pending = 0; 2936 } 2937 V6_SET_ZERO(icmp->icmp_v6src); 2938 V6_SET_ZERO(icmp->icmp_bound_v6src); 2939 icmp->icmp_state = TS_UNBND; 2940 if (icmp->icmp_family == AF_INET6) 2941 (void) icmp_build_hdrs(q, icmp); 2942 break; 2943 default: 2944 break; 2945 } 2946 break; 2947 case T_BIND_ACK: 2948 icmp_rput_bind_ack(q, mp); 2949 return; 2950 2951 case T_OPTMGMT_ACK: 2952 case T_OK_ACK: 2953 if (tea->PRIM_type == T_OK_ACK) { 2954 struct T_ok_ack *toa; 2955 toa = (struct T_ok_ack *)rptr; 2956 if (toa->CORRECT_prim == T_UNBIND_REQ) { 2957 /* 2958 * If somebody sets IPSEC options, IP 2959 * sends some IPSEC info which is used 2960 * by the TCP for detached connections. 2961 * We don't need it here. 2962 */ 2963 if ((mp1 = mp->b_cont) != NULL) { 2964 freemsg(mp1); 2965 mp->b_cont = NULL; 2966 } 2967 } 2968 } 2969 break; 2970 default: 2971 freemsg(mp); 2972 return; 2973 } 2974 putnext(q, mp); 2975 return; 2976 case M_CTL: 2977 if (recvif) { 2978 /* 2979 * IP has passed the options in mp and the actual data 2980 * is in b_cont. Jump to normal data processing. 2981 */ 2982 break; 2983 } 2984 2985 /* Contains ICMP packet from IP */ 2986 icmp_icmp_error(q, mp); 2987 return; 2988 default: 2989 putnext(q, mp); 2990 return; 2991 } 2992 2993 /* 2994 * Discard message if it is misaligned or smaller than the IP header. 2995 */ 2996 if (!OK_32PTR(rptr) || (mp->b_wptr - rptr) < sizeof (ipha_t)) { 2997 freemsg(mp); 2998 if (options_mp != NULL) 2999 freeb(options_mp); 3000 BUMP_MIB(&rawip_mib, rawipInErrors); 3001 return; 3002 } 3003 ipvers = IPH_HDR_VERSION((ipha_t *)rptr); 3004 3005 /* Handle M_DATA messages containing IP packets messages */ 3006 if (ipvers == IPV4_VERSION) { 3007 /* 3008 * Special case where IP attaches 3009 * the IRE needs to be handled so that we don't send up 3010 * IRE to the user land. 3011 */ 3012 ipha = (ipha_t *)rptr; 3013 hdr_len = IPH_HDR_LENGTH(ipha); 3014 3015 if (ipha->ipha_protocol == IPPROTO_TCP) { 3016 tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len]; 3017 3018 if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == 3019 TH_SYN) && mp->b_cont != NULL) { 3020 mp1 = mp->b_cont; 3021 if (mp1->b_datap->db_type == IRE_DB_TYPE) { 3022 freeb(mp1); 3023 mp->b_cont = NULL; 3024 } 3025 } 3026 } 3027 if (icmp_bsd_compat) { 3028 ushort_t len; 3029 len = ntohs(ipha->ipha_length); 3030 3031 if (mp->b_datap->db_ref > 1) { 3032 /* 3033 * Allocate a new IP header so that we can 3034 * modify ipha_length. 3035 */ 3036 mblk_t *mp1; 3037 3038 mp1 = allocb(hdr_len, BPRI_MED); 3039 if (!mp1) { 3040 freemsg(mp); 3041 if (options_mp != NULL) 3042 freeb(options_mp); 3043 BUMP_MIB(&rawip_mib, rawipInErrors); 3044 return; 3045 } 3046 bcopy(rptr, mp1->b_rptr, hdr_len); 3047 mp->b_rptr = rptr + hdr_len; 3048 rptr = mp1->b_rptr; 3049 ipha = (ipha_t *)rptr; 3050 mp1->b_cont = mp; 3051 mp1->b_wptr = rptr + hdr_len; 3052 mp = mp1; 3053 } 3054 len -= hdr_len; 3055 ipha->ipha_length = htons(len); 3056 } 3057 } 3058 3059 /* 3060 * This is the inbound data path. Packets are passed upstream as 3061 * T_UNITDATA_IND messages with full IP headers still attached. 3062 */ 3063 if (icmp->icmp_family == AF_INET) { 3064 ASSERT(ipvers == IPV4_VERSION); 3065 udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t); 3066 if (recvif) { 3067 udi_size += sizeof (struct T_opthdr) + 3068 sizeof (uint_t); 3069 } 3070 mp1 = allocb(udi_size, BPRI_MED); 3071 if (mp1 == NULL) { 3072 freemsg(mp); 3073 if (options_mp != NULL) 3074 freeb(options_mp); 3075 BUMP_MIB(&rawip_mib, rawipInErrors); 3076 return; 3077 } 3078 mp1->b_cont = mp; 3079 mp = mp1; 3080 tudi = (struct T_unitdata_ind *)mp->b_rptr; 3081 mp->b_datap->db_type = M_PROTO; 3082 mp->b_wptr = (uchar_t *)tudi + udi_size; 3083 tudi->PRIM_type = T_UNITDATA_IND; 3084 tudi->SRC_length = sizeof (sin_t); 3085 tudi->SRC_offset = sizeof (struct T_unitdata_ind); 3086 sin = (sin_t *)&tudi[1]; 3087 *sin = sin_null; 3088 sin->sin_family = AF_INET; 3089 sin->sin_addr.s_addr = ipha->ipha_src; 3090 tudi->OPT_offset = sizeof (struct T_unitdata_ind) + 3091 sizeof (sin_t); 3092 udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t)); 3093 tudi->OPT_length = udi_size; 3094 3095 /* 3096 * Add options if IP_RECVIF is set 3097 */ 3098 if (udi_size != 0) { 3099 char *dstopt; 3100 3101 dstopt = (char *)&sin[1]; 3102 if (recvif) { 3103 3104 struct T_opthdr *toh; 3105 uint_t *dstptr; 3106 3107 toh = (struct T_opthdr *)dstopt; 3108 toh->level = IPPROTO_IP; 3109 toh->name = IP_RECVIF; 3110 toh->len = sizeof (struct T_opthdr) + 3111 sizeof (uint_t); 3112 toh->status = 0; 3113 dstopt += sizeof (struct T_opthdr); 3114 dstptr = (uint_t *)dstopt; 3115 *dstptr = pinfo->in_pkt_ifindex; 3116 dstopt += sizeof (uint_t); 3117 freeb(options_mp); 3118 udi_size -= toh->len; 3119 } 3120 3121 /* Consumed all of allocated space */ 3122 ASSERT(udi_size == 0); 3123 } 3124 3125 BUMP_MIB(&rawip_mib, rawipInDatagrams); 3126 putnext(q, mp); 3127 return; 3128 } 3129 3130 /* 3131 * We don't need options_mp in the IPv6 path. 3132 */ 3133 if (options_mp != NULL) { 3134 freeb(options_mp); 3135 options_mp = NULL; 3136 } 3137 3138 /* 3139 * Discard message if it is smaller than the IPv6 header 3140 * or if the header is malformed. 3141 */ 3142 if ((mp->b_wptr - rptr) < sizeof (ip6_t) || 3143 IPH_HDR_VERSION((ipha_t *)rptr) != IPV6_VERSION || 3144 icmp->icmp_family != AF_INET6) { 3145 freemsg(mp); 3146 BUMP_MIB(&rawip_mib, rawipInErrors); 3147 return; 3148 } 3149 3150 /* Initialize */ 3151 ipp.ipp_fields = 0; 3152 3153 ip6h = (ip6_t *)rptr; 3154 /* 3155 * Call on ip_find_hdr_v6 which gets the total hdr len 3156 * as well as individual lenghts of ext hdrs (and ptrs to 3157 * them). 3158 */ 3159 if (ip6h->ip6_nxt != icmp->icmp_proto) { 3160 /* Look for ifindex information */ 3161 if (ip6h->ip6_nxt == IPPROTO_RAW) { 3162 ip6i = (ip6i_t *)ip6h; 3163 if (ip6i->ip6i_flags & IP6I_IFINDEX) { 3164 ASSERT(ip6i->ip6i_ifindex != 0); 3165 ipp.ipp_fields |= IPPF_IFINDEX; 3166 ipp.ipp_ifindex = ip6i->ip6i_ifindex; 3167 } 3168 rptr = (uchar_t *)&ip6i[1]; 3169 mp->b_rptr = rptr; 3170 if (rptr == mp->b_wptr) { 3171 mp1 = mp->b_cont; 3172 freeb(mp); 3173 mp = mp1; 3174 rptr = mp->b_rptr; 3175 } 3176 ASSERT(mp->b_wptr - rptr >= IPV6_HDR_LEN); 3177 ip6h = (ip6_t *)rptr; 3178 } 3179 hdr_len = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdr); 3180 } else { 3181 hdr_len = IPV6_HDR_LEN; 3182 ip6i = NULL; 3183 nexthdr = ip6h->ip6_nxt; 3184 } 3185 /* 3186 * One special case where IP attaches the IRE needs to 3187 * be handled so that we don't send up IRE to the user land. 3188 */ 3189 if (nexthdr == IPPROTO_TCP) { 3190 tcph_t *tcph = (tcph_t *)&mp->b_rptr[hdr_len]; 3191 3192 if (((tcph->th_flags[0] & (TH_SYN|TH_ACK)) == TH_SYN) && 3193 mp->b_cont != NULL) { 3194 mp1 = mp->b_cont; 3195 if (mp1->b_datap->db_type == IRE_DB_TYPE) { 3196 freeb(mp1); 3197 mp->b_cont = NULL; 3198 } 3199 } 3200 } 3201 /* 3202 * Check a filter for ICMPv6 types if needed. 3203 * Verify raw checksums if needed. 3204 */ 3205 if (icmp->icmp_filter != NULL || icmp->icmp_raw_checksum) { 3206 if (icmp->icmp_filter != NULL) { 3207 int type; 3208 3209 /* Assumes that IP has done the pullupmsg */ 3210 type = mp->b_rptr[hdr_len]; 3211 3212 ASSERT(mp->b_rptr + hdr_len <= mp->b_wptr); 3213 if (ICMP6_FILTER_WILLBLOCK(type, icmp->icmp_filter)) { 3214 freemsg(mp); 3215 return; 3216 } 3217 } else { 3218 /* Checksum */ 3219 uint16_t *up; 3220 uint32_t sum; 3221 int remlen; 3222 3223 up = (uint16_t *)&ip6h->ip6_src; 3224 3225 remlen = msgdsize(mp) - hdr_len; 3226 sum = htons(icmp->icmp_proto + remlen) 3227 + up[0] + up[1] + up[2] + up[3] 3228 + up[4] + up[5] + up[6] + up[7] 3229 + up[8] + up[9] + up[10] + up[11] 3230 + up[12] + up[13] + up[14] + up[15]; 3231 sum = (sum & 0xffff) + (sum >> 16); 3232 sum = IP_CSUM(mp, hdr_len, sum); 3233 if (sum != 0) { 3234 /* IPv6 RAW checksum failed */ 3235 ip0dbg(("icmp_rput: RAW checksum " 3236 "failed %x\n", sum)); 3237 freemsg(mp); 3238 BUMP_MIB(&rawip_mib, rawipInCksumErrs); 3239 return; 3240 } 3241 } 3242 } 3243 /* Skip all the IPv6 headers per API */ 3244 mp->b_rptr += hdr_len; 3245 3246 udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t); 3247 3248 /* 3249 * We use local variables icmp_opt and icmp_ipv6_recvhoplimit to 3250 * maintain state information, instead of relying on icmp_t 3251 * structure, since there arent any locks protecting these members 3252 * and there is a window where there might be a race between a 3253 * thread setting options on the write side and a thread reading 3254 * these options on the read size. 3255 */ 3256 if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS| 3257 IPPF_RTHDR|IPPF_IFINDEX)) { 3258 if (icmp->icmp_ipv6_recvhopopts && 3259 (ipp.ipp_fields & IPPF_HOPOPTS)) { 3260 udi_size += sizeof (struct T_opthdr) + 3261 ipp.ipp_hopoptslen; 3262 icmp_opt |= IPPF_HOPOPTS; 3263 } 3264 if ((icmp->icmp_ipv6_recvdstopts || 3265 icmp->icmp_old_ipv6_recvdstopts) && 3266 (ipp.ipp_fields & IPPF_DSTOPTS)) { 3267 udi_size += sizeof (struct T_opthdr) + 3268 ipp.ipp_dstoptslen; 3269 icmp_opt |= IPPF_DSTOPTS; 3270 } 3271 if (((icmp->icmp_ipv6_recvdstopts && 3272 icmp->icmp_ipv6_recvrthdr && 3273 (ipp.ipp_fields & IPPF_RTHDR)) || 3274 icmp->icmp_ipv6_recvrtdstopts) && 3275 (ipp.ipp_fields & IPPF_RTDSTOPTS)) { 3276 udi_size += sizeof (struct T_opthdr) + 3277 ipp.ipp_rtdstoptslen; 3278 icmp_opt |= IPPF_RTDSTOPTS; 3279 } 3280 if (icmp->icmp_ipv6_recvrthdr && 3281 (ipp.ipp_fields & IPPF_RTHDR)) { 3282 udi_size += sizeof (struct T_opthdr) + 3283 ipp.ipp_rthdrlen; 3284 icmp_opt |= IPPF_RTHDR; 3285 } 3286 if (icmp->icmp_ipv6_recvpktinfo && 3287 (ipp.ipp_fields & IPPF_IFINDEX)) { 3288 udi_size += sizeof (struct T_opthdr) + 3289 sizeof (struct in6_pktinfo); 3290 icmp_opt |= IPPF_IFINDEX; 3291 } 3292 } 3293 if (icmp->icmp_ipv6_recvhoplimit) { 3294 udi_size += sizeof (struct T_opthdr) + sizeof (int); 3295 icmp_ipv6_recvhoplimit = B_TRUE; 3296 } 3297 3298 if (icmp->icmp_ipv6_recvtclass) 3299 udi_size += sizeof (struct T_opthdr) + sizeof (int); 3300 3301 mp1 = allocb(udi_size, BPRI_MED); 3302 if (mp1 == NULL) { 3303 freemsg(mp); 3304 BUMP_MIB(&rawip_mib, rawipInErrors); 3305 return; 3306 } 3307 mp1->b_cont = mp; 3308 mp = mp1; 3309 mp->b_datap->db_type = M_PROTO; 3310 tudi = (struct T_unitdata_ind *)mp->b_rptr; 3311 mp->b_wptr = (uchar_t *)tudi + udi_size; 3312 tudi->PRIM_type = T_UNITDATA_IND; 3313 tudi->SRC_length = sizeof (sin6_t); 3314 tudi->SRC_offset = sizeof (struct T_unitdata_ind); 3315 tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin6_t); 3316 udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t)); 3317 tudi->OPT_length = udi_size; 3318 sin6 = (sin6_t *)&tudi[1]; 3319 sin6->sin6_port = 0; 3320 sin6->sin6_family = AF_INET6; 3321 3322 sin6->sin6_addr = ip6h->ip6_src; 3323 /* No sin6_flowinfo per API */ 3324 sin6->sin6_flowinfo = 0; 3325 /* For link-scope source pass up scope id */ 3326 if ((ipp.ipp_fields & IPPF_IFINDEX) && 3327 IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src)) 3328 sin6->sin6_scope_id = ipp.ipp_ifindex; 3329 else 3330 sin6->sin6_scope_id = 0; 3331 3332 sin6->__sin6_src_id = ip_srcid_find_addr(&ip6h->ip6_dst, 3333 icmp->icmp_zoneid); 3334 3335 if (udi_size != 0) { 3336 uchar_t *dstopt; 3337 3338 dstopt = (uchar_t *)&sin6[1]; 3339 if (icmp_opt & IPPF_IFINDEX) { 3340 struct T_opthdr *toh; 3341 struct in6_pktinfo *pkti; 3342 3343 toh = (struct T_opthdr *)dstopt; 3344 toh->level = IPPROTO_IPV6; 3345 toh->name = IPV6_PKTINFO; 3346 toh->len = sizeof (struct T_opthdr) + 3347 sizeof (*pkti); 3348 toh->status = 0; 3349 dstopt += sizeof (struct T_opthdr); 3350 pkti = (struct in6_pktinfo *)dstopt; 3351 pkti->ipi6_addr = ip6h->ip6_dst; 3352 pkti->ipi6_ifindex = ipp.ipp_ifindex; 3353 dstopt += sizeof (*pkti); 3354 udi_size -= toh->len; 3355 } 3356 if (icmp_ipv6_recvhoplimit) { 3357 struct T_opthdr *toh; 3358 3359 toh = (struct T_opthdr *)dstopt; 3360 toh->level = IPPROTO_IPV6; 3361 toh->name = IPV6_HOPLIMIT; 3362 toh->len = sizeof (struct T_opthdr) + 3363 sizeof (uint_t); 3364 toh->status = 0; 3365 dstopt += sizeof (struct T_opthdr); 3366 *(uint_t *)dstopt = ip6h->ip6_hops; 3367 dstopt += sizeof (uint_t); 3368 udi_size -= toh->len; 3369 } 3370 if (icmp->icmp_ipv6_recvtclass) { 3371 struct T_opthdr *toh; 3372 3373 toh = (struct T_opthdr *)dstopt; 3374 toh->level = IPPROTO_IPV6; 3375 toh->name = IPV6_TCLASS; 3376 toh->len = sizeof (struct T_opthdr) + 3377 sizeof (uint_t); 3378 toh->status = 0; 3379 dstopt += sizeof (struct T_opthdr); 3380 *(uint_t *)dstopt = IPV6_FLOW_TCLASS(ip6h->ip6_flow); 3381 dstopt += sizeof (uint_t); 3382 udi_size -= toh->len; 3383 } 3384 if (icmp_opt & IPPF_HOPOPTS) { 3385 struct T_opthdr *toh; 3386 3387 toh = (struct T_opthdr *)dstopt; 3388 toh->level = IPPROTO_IPV6; 3389 toh->name = IPV6_HOPOPTS; 3390 toh->len = sizeof (struct T_opthdr) + 3391 ipp.ipp_hopoptslen; 3392 toh->status = 0; 3393 dstopt += sizeof (struct T_opthdr); 3394 bcopy(ipp.ipp_hopopts, dstopt, 3395 ipp.ipp_hopoptslen); 3396 dstopt += ipp.ipp_hopoptslen; 3397 udi_size -= toh->len; 3398 } 3399 if (icmp_opt & IPPF_RTDSTOPTS) { 3400 struct T_opthdr *toh; 3401 3402 toh = (struct T_opthdr *)dstopt; 3403 toh->level = IPPROTO_IPV6; 3404 toh->name = IPV6_DSTOPTS; 3405 toh->len = sizeof (struct T_opthdr) + 3406 ipp.ipp_rtdstoptslen; 3407 toh->status = 0; 3408 dstopt += sizeof (struct T_opthdr); 3409 bcopy(ipp.ipp_rtdstopts, dstopt, 3410 ipp.ipp_rtdstoptslen); 3411 dstopt += ipp.ipp_rtdstoptslen; 3412 udi_size -= toh->len; 3413 } 3414 if (icmp_opt & IPPF_RTHDR) { 3415 struct T_opthdr *toh; 3416 3417 toh = (struct T_opthdr *)dstopt; 3418 toh->level = IPPROTO_IPV6; 3419 toh->name = IPV6_RTHDR; 3420 toh->len = sizeof (struct T_opthdr) + 3421 ipp.ipp_rthdrlen; 3422 toh->status = 0; 3423 dstopt += sizeof (struct T_opthdr); 3424 bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen); 3425 dstopt += ipp.ipp_rthdrlen; 3426 udi_size -= toh->len; 3427 } 3428 if (icmp_opt & IPPF_DSTOPTS) { 3429 struct T_opthdr *toh; 3430 3431 toh = (struct T_opthdr *)dstopt; 3432 toh->level = IPPROTO_IPV6; 3433 toh->name = IPV6_DSTOPTS; 3434 toh->len = sizeof (struct T_opthdr) + 3435 ipp.ipp_dstoptslen; 3436 toh->status = 0; 3437 dstopt += sizeof (struct T_opthdr); 3438 bcopy(ipp.ipp_dstopts, dstopt, 3439 ipp.ipp_dstoptslen); 3440 dstopt += ipp.ipp_dstoptslen; 3441 udi_size -= toh->len; 3442 } 3443 /* Consumed all of allocated space */ 3444 ASSERT(udi_size == 0); 3445 } 3446 BUMP_MIB(&rawip_mib, rawipInDatagrams); 3447 putnext(q, mp); 3448 } 3449 3450 /* 3451 * Process a T_BIND_ACK 3452 */ 3453 static void 3454 icmp_rput_bind_ack(queue_t *q, mblk_t *mp) 3455 { 3456 icmp_t *icmp = (icmp_t *)q->q_ptr; 3457 mblk_t *mp1; 3458 ire_t *ire; 3459 struct T_bind_ack *tba; 3460 uchar_t *addrp; 3461 ipa_conn_t *ac; 3462 ipa6_conn_t *ac6; 3463 3464 /* 3465 * We know if headers are included or not so we can 3466 * safely do this. 3467 */ 3468 if (icmp->icmp_state == TS_UNBND) { 3469 /* 3470 * TPI has not yet bound - bind sent by 3471 * icmp_bind_proto. 3472 */ 3473 freemsg(mp); 3474 return; 3475 } 3476 if (icmp->icmp_discon_pending) 3477 icmp->icmp_discon_pending = 0; 3478 3479 /* 3480 * If a broadcast/multicast address was bound set 3481 * the source address to 0. 3482 * This ensures no datagrams with broadcast address 3483 * as source address are emitted (which would violate 3484 * RFC1122 - Hosts requirements) 3485 * 3486 * Note that when connecting the returned IRE is 3487 * for the destination address and we only perform 3488 * the broadcast check for the source address (it 3489 * is OK to connect to a broadcast/multicast address.) 3490 */ 3491 mp1 = mp->b_cont; 3492 if (mp1 != NULL && mp1->b_datap->db_type == IRE_DB_TYPE) { 3493 ire = (ire_t *)mp1->b_rptr; 3494 3495 /* 3496 * Note: we get IRE_BROADCAST for IPv6 to "mark" a multicast 3497 * local address. 3498 */ 3499 if (ire->ire_type == IRE_BROADCAST && 3500 icmp->icmp_state != TS_DATA_XFER) { 3501 /* This was just a local bind to a MC/broadcast addr */ 3502 V6_SET_ZERO(icmp->icmp_v6src); 3503 if (icmp->icmp_family == AF_INET6) 3504 (void) icmp_build_hdrs(q, icmp); 3505 } else if (V6_OR_V4_INADDR_ANY(icmp->icmp_v6src)) { 3506 /* 3507 * Local address not yet set - pick it from the 3508 * T_bind_ack 3509 */ 3510 tba = (struct T_bind_ack *)mp->b_rptr; 3511 addrp = &mp->b_rptr[tba->ADDR_offset]; 3512 switch (icmp->icmp_family) { 3513 case AF_INET: 3514 if (tba->ADDR_length == sizeof (ipa_conn_t)) { 3515 ac = (ipa_conn_t *)addrp; 3516 } else { 3517 ASSERT(tba->ADDR_length == 3518 sizeof (ipa_conn_x_t)); 3519 ac = &((ipa_conn_x_t *)addrp)->acx_conn; 3520 } 3521 IN6_IPADDR_TO_V4MAPPED(ac->ac_laddr, 3522 &icmp->icmp_v6src); 3523 break; 3524 case AF_INET6: 3525 if (tba->ADDR_length == sizeof (ipa6_conn_t)) { 3526 ac6 = (ipa6_conn_t *)addrp; 3527 } else { 3528 ASSERT(tba->ADDR_length == 3529 sizeof (ipa6_conn_x_t)); 3530 ac6 = &((ipa6_conn_x_t *) 3531 addrp)->ac6x_conn; 3532 } 3533 icmp->icmp_v6src = ac6->ac6_laddr; 3534 (void) icmp_build_hdrs(q, icmp); 3535 } 3536 } 3537 mp1 = mp1->b_cont; 3538 } 3539 /* 3540 * Look for one or more appended ACK message added by 3541 * icmp_connect or icmp_disconnect. 3542 * If none found just send up the T_BIND_ACK. 3543 * icmp_connect has appended a T_OK_ACK and a 3544 * T_CONN_CON. 3545 * icmp_disconnect has appended a T_OK_ACK. 3546 */ 3547 if (mp1 != NULL) { 3548 if (mp->b_cont == mp1) 3549 mp->b_cont = NULL; 3550 else { 3551 ASSERT(mp->b_cont->b_cont == mp1); 3552 mp->b_cont->b_cont = NULL; 3553 } 3554 freemsg(mp); 3555 mp = mp1; 3556 while (mp != NULL) { 3557 mp1 = mp->b_cont; 3558 mp->b_cont = NULL; 3559 putnext(q, mp); 3560 mp = mp1; 3561 } 3562 return; 3563 } 3564 freemsg(mp->b_cont); 3565 mp->b_cont = NULL; 3566 putnext(q, mp); 3567 } 3568 3569 /* 3570 * return SNMP stuff in buffer in mpdata 3571 */ 3572 static int 3573 icmp_snmp_get(queue_t *q, mblk_t *mpctl) 3574 { 3575 mblk_t *mpdata; 3576 struct opthdr *optp; 3577 3578 if (mpctl == NULL || 3579 (mpdata = mpctl->b_cont) == NULL) { 3580 return (0); 3581 } 3582 3583 /* fixed length structure for IPv4 and IPv6 counters */ 3584 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)]; 3585 optp->level = EXPER_RAWIP; 3586 optp->name = 0; 3587 (void) snmp_append_data(mpdata, (char *)&rawip_mib, sizeof (rawip_mib)); 3588 optp->len = msgdsize(mpdata); 3589 qreply(q, mpctl); 3590 3591 return (1); 3592 } 3593 3594 /* 3595 * Return 0 if invalid set request, 1 otherwise, including non-rawip requests. 3596 * TODO: If this ever actually tries to set anything, it needs to be 3597 * to do the appropriate locking. 3598 */ 3599 /* ARGSUSED */ 3600 static int 3601 icmp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name, 3602 uchar_t *ptr, int len) 3603 { 3604 switch (level) { 3605 case EXPER_RAWIP: 3606 return (0); 3607 default: 3608 return (1); 3609 } 3610 } 3611 3612 /* Report for ndd "icmp_status" */ 3613 /* ARGSUSED */ 3614 static int 3615 icmp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr) 3616 { 3617 IDP idp; 3618 icmp_t *icmp; 3619 char *state; 3620 char laddrbuf[INET6_ADDRSTRLEN]; 3621 char faddrbuf[INET6_ADDRSTRLEN]; 3622 3623 (void) mi_mpprintf(mp, 3624 "RAWIP " MI_COL_HDRPAD_STR 3625 /* 01234567[89ABCDEF] */ 3626 " src addr dest addr state"); 3627 /* xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx UNBOUND */ 3628 3629 3630 for (idp = mi_first_ptr(&icmp_g_head); 3631 (icmp = (icmp_t *)idp) != NULL; 3632 idp = mi_next_ptr(&icmp_g_head, idp)) { 3633 if (icmp->icmp_state == TS_UNBND) 3634 state = "UNBOUND"; 3635 else if (icmp->icmp_state == TS_IDLE) 3636 state = "IDLE"; 3637 else if (icmp->icmp_state == TS_DATA_XFER) 3638 state = "CONNECTED"; 3639 else 3640 state = "UnkState"; 3641 3642 (void) mi_mpprintf(mp, 3643 MI_COL_PTRFMT_STR "%s %s %s", 3644 (void *)icmp, 3645 inet_ntop(AF_INET6, &icmp->icmp_v6dst, faddrbuf, 3646 sizeof (faddrbuf)), 3647 inet_ntop(AF_INET6, &icmp->icmp_v6src, laddrbuf, 3648 sizeof (laddrbuf)), 3649 state); 3650 } 3651 return (0); 3652 } 3653 3654 /* 3655 * This routine creates a T_UDERROR_IND message and passes it upstream. 3656 * The address and options are copied from the T_UNITDATA_REQ message 3657 * passed in mp. This message is freed. 3658 */ 3659 static void 3660 icmp_ud_err(queue_t *q, mblk_t *mp, t_scalar_t err) 3661 { 3662 mblk_t *mp1; 3663 uchar_t *rptr = mp->b_rptr; 3664 struct T_unitdata_req *tudr = (struct T_unitdata_req *)rptr; 3665 3666 mp1 = mi_tpi_uderror_ind((char *)&rptr[tudr->DEST_offset], 3667 tudr->DEST_length, (char *)&rptr[tudr->OPT_offset], 3668 tudr->OPT_length, err); 3669 if (mp1) 3670 qreply(q, mp1); 3671 freemsg(mp); 3672 } 3673 3674 /* 3675 * This routine is called by icmp_wput to handle T_UNBIND_REQ messages. 3676 * After some error checking, the message is passed downstream to ip. 3677 */ 3678 static void 3679 icmp_unbind(queue_t *q, mblk_t *mp) 3680 { 3681 icmp_t *icmp = (icmp_t *)q->q_ptr; 3682 3683 /* If a bind has not been done, we can't unbind. */ 3684 if (icmp->icmp_state == TS_UNBND) { 3685 icmp_err_ack(q, mp, TOUTSTATE, 0); 3686 return; 3687 } 3688 V6_SET_ZERO(icmp->icmp_v6src); 3689 V6_SET_ZERO(icmp->icmp_bound_v6src); 3690 icmp->icmp_state = TS_UNBND; 3691 3692 if (icmp->icmp_family == AF_INET6) { 3693 int error; 3694 3695 /* Rebuild the header template */ 3696 error = icmp_build_hdrs(q, icmp); 3697 if (error != 0) { 3698 icmp_err_ack(q, mp, TSYSERR, error); 3699 return; 3700 } 3701 } 3702 /* Pass the unbind to IP. */ 3703 putnext(q, mp); 3704 } 3705 3706 /* 3707 * Process IPv4 packets that already include an IP header. 3708 * Used when IP_HDRINCL has been set (implicit for IPPROTO_RAW and 3709 * IPPROTO_IGMP). 3710 */ 3711 static void 3712 icmp_wput_hdrincl(queue_t *q, mblk_t *mp, icmp_t *icmp) 3713 { 3714 ipha_t *ipha; 3715 int ip_hdr_length; 3716 int tp_hdr_len; 3717 mblk_t *mp1; 3718 uint_t pkt_len; 3719 3720 ipha = (ipha_t *)mp->b_rptr; 3721 ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len; 3722 if ((mp->b_wptr - mp->b_rptr) < IP_SIMPLE_HDR_LENGTH) { 3723 if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH)) { 3724 BUMP_MIB(&rawip_mib, rawipOutErrors); 3725 freemsg(mp); 3726 return; 3727 } 3728 ipha = (ipha_t *)mp->b_rptr; 3729 } 3730 ipha->ipha_version_and_hdr_length = 3731 (IP_VERSION<<4) | (ip_hdr_length>>2); 3732 3733 /* 3734 * For the socket of SOCK_RAW type, the checksum is provided in the 3735 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to 3736 * tell IP that the application has sent a complete IP header and not 3737 * to compute the transport checksum nor change the DF flag. 3738 */ 3739 ipha->ipha_ident = IP_HDR_INCLUDED; 3740 ipha->ipha_hdr_checksum = 0; 3741 ipha->ipha_fragment_offset_and_flags &= htons(IPH_DF); 3742 /* Insert options if any */ 3743 if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) { 3744 /* 3745 * Put the IP header plus any transport header that is 3746 * checksumed by ip_wput into the first mblk. (ip_wput assumes 3747 * that at least the checksum field is in the first mblk.) 3748 */ 3749 switch (ipha->ipha_protocol) { 3750 case IPPROTO_UDP: 3751 tp_hdr_len = 8; 3752 break; 3753 case IPPROTO_TCP: 3754 tp_hdr_len = 20; 3755 break; 3756 default: 3757 tp_hdr_len = 0; 3758 break; 3759 } 3760 /* 3761 * The code below assumes that IP_SIMPLE_HDR_LENGTH plus 3762 * tp_hdr_len bytes will be in a single mblk. 3763 */ 3764 if ((mp->b_wptr - mp->b_rptr) < (IP_SIMPLE_HDR_LENGTH + 3765 tp_hdr_len)) { 3766 if (!pullupmsg(mp, IP_SIMPLE_HDR_LENGTH + 3767 tp_hdr_len)) { 3768 BUMP_MIB(&rawip_mib, rawipOutErrors); 3769 freemsg(mp); 3770 return; 3771 } 3772 ipha = (ipha_t *)mp->b_rptr; 3773 } 3774 3775 /* 3776 * if the length is larger then the max allowed IP packet, 3777 * then send an error and abort the processing. 3778 */ 3779 pkt_len = ntohs(ipha->ipha_length) 3780 + icmp->icmp_ip_snd_options_len; 3781 if (pkt_len > IP_MAXPACKET) { 3782 icmp_ud_err(q, mp, EMSGSIZE); 3783 return; 3784 } 3785 if (!(mp1 = allocb(ip_hdr_length + icmp_wroff_extra + 3786 tp_hdr_len, BPRI_LO))) { 3787 icmp_ud_err(q, mp, ENOMEM); 3788 return; 3789 } 3790 mp1->b_rptr += icmp_wroff_extra; 3791 mp1->b_wptr = mp1->b_rptr + ip_hdr_length; 3792 3793 ipha->ipha_length = htons((uint16_t)pkt_len); 3794 bcopy(ipha, mp1->b_rptr, IP_SIMPLE_HDR_LENGTH); 3795 3796 /* Copy transport header if any */ 3797 bcopy(&ipha[1], mp1->b_wptr, tp_hdr_len); 3798 mp1->b_wptr += tp_hdr_len; 3799 3800 /* Add options */ 3801 ipha = (ipha_t *)mp1->b_rptr; 3802 bcopy(icmp->icmp_ip_snd_options, &ipha[1], 3803 icmp->icmp_ip_snd_options_len); 3804 3805 /* Drop IP header and transport header from original */ 3806 (void) adjmsg(mp, IP_SIMPLE_HDR_LENGTH + tp_hdr_len); 3807 3808 mp1->b_cont = mp; 3809 mp = mp1; 3810 /* 3811 * Massage source route putting first source 3812 * route in ipha_dst. 3813 */ 3814 (void) ip_massage_options(ipha); 3815 } 3816 putnext(q, mp); 3817 } 3818 3819 /* 3820 * This routine handles all messages passed downstream. It either 3821 * consumes the message or passes it downstream; it never queues a 3822 * a message. 3823 */ 3824 static void 3825 icmp_wput(queue_t *q, mblk_t *mp) 3826 { 3827 uchar_t *rptr = mp->b_rptr; 3828 ipha_t *ipha; 3829 mblk_t *mp1; 3830 int ip_hdr_length; 3831 #define tudr ((struct T_unitdata_req *)rptr) 3832 size_t ip_len; 3833 icmp_t *icmp; 3834 sin6_t *sin6; 3835 sin_t *sin; 3836 ipaddr_t v4dst; 3837 3838 icmp = (icmp_t *)q->q_ptr; 3839 if (icmp->icmp_restricted) { 3840 icmp_wput_restricted(q, mp); 3841 return; 3842 } 3843 3844 switch (mp->b_datap->db_type) { 3845 case M_DATA: 3846 if (icmp->icmp_hdrincl) { 3847 ASSERT(icmp->icmp_ipversion == IPV4_VERSION); 3848 icmp_wput_hdrincl(q, mp, icmp); 3849 return; 3850 } 3851 freemsg(mp); 3852 return; 3853 case M_PROTO: 3854 case M_PCPROTO: 3855 ip_len = mp->b_wptr - rptr; 3856 if (ip_len >= sizeof (struct T_unitdata_req)) { 3857 /* Expedite valid T_UNITDATA_REQ to below the switch */ 3858 if (((union T_primitives *)rptr)->type 3859 == T_UNITDATA_REQ) 3860 break; 3861 } 3862 /* FALLTHRU */ 3863 default: 3864 icmp_wput_other(q, mp); 3865 return; 3866 } 3867 3868 /* Handle T_UNITDATA_REQ messages here. */ 3869 3870 if (icmp->icmp_state == TS_UNBND) { 3871 /* If a port has not been bound to the stream, fail. */ 3872 BUMP_MIB(&rawip_mib, rawipOutErrors); 3873 icmp_ud_err(q, mp, EPROTO); 3874 return; 3875 } 3876 mp1 = mp->b_cont; 3877 if (mp1 == NULL) { 3878 BUMP_MIB(&rawip_mib, rawipOutErrors); 3879 icmp_ud_err(q, mp, EPROTO); 3880 return; 3881 } 3882 3883 if ((rptr + tudr->DEST_offset + tudr->DEST_length) > mp->b_wptr) { 3884 BUMP_MIB(&rawip_mib, rawipOutErrors); 3885 icmp_ud_err(q, mp, EADDRNOTAVAIL); 3886 return; 3887 } 3888 3889 switch (icmp->icmp_family) { 3890 case AF_INET6: 3891 sin6 = (sin6_t *)&rptr[tudr->DEST_offset]; 3892 if (!OK_32PTR((char *)sin6) || 3893 tudr->DEST_length != sizeof (sin6_t) || 3894 sin6->sin6_family != AF_INET6) { 3895 BUMP_MIB(&rawip_mib, rawipOutErrors); 3896 icmp_ud_err(q, mp, EADDRNOTAVAIL); 3897 return; 3898 } 3899 3900 /* No support for mapped addresses on raw sockets */ 3901 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 3902 BUMP_MIB(&rawip_mib, rawipOutErrors); 3903 icmp_ud_err(q, mp, EADDRNOTAVAIL); 3904 return; 3905 } 3906 3907 /* 3908 * Destination is a native IPv6 address. 3909 * Send out an IPv6 format packet. 3910 */ 3911 icmp_wput_ipv6(q, mp, sin6, tudr->OPT_length); 3912 return; 3913 3914 case AF_INET: 3915 sin = (sin_t *)&rptr[tudr->DEST_offset]; 3916 if (!OK_32PTR((char *)sin) || 3917 tudr->DEST_length != sizeof (sin_t) || 3918 sin->sin_family != AF_INET) { 3919 BUMP_MIB(&rawip_mib, rawipOutErrors); 3920 icmp_ud_err(q, mp, EADDRNOTAVAIL); 3921 return; 3922 } 3923 /* Extract and ipaddr */ 3924 v4dst = sin->sin_addr.s_addr; 3925 break; 3926 } 3927 3928 /* 3929 * If options passed in, feed it for verification and handling 3930 */ 3931 if (tudr->OPT_length != 0) { 3932 int error; 3933 3934 if (icmp_unitdata_opt_process(q, mp, &error, 3935 (uchar_t *)0) < 0) { 3936 /* failure */ 3937 BUMP_MIB(&rawip_mib, rawipOutErrors); 3938 icmp_ud_err(q, mp, error); 3939 return; 3940 } 3941 /* 3942 * Note: Success in processing options. 3943 * mp option buffer represented by 3944 * OPT_length/offset now potentially modified 3945 * and contain option setting results 3946 */ 3947 } 3948 3949 /* Protocol 255 contains full IP headers */ 3950 if (icmp->icmp_hdrincl) { 3951 freeb(mp); 3952 icmp_wput_hdrincl(q, mp1, icmp); 3953 return; 3954 } 3955 /* Add an IP header */ 3956 ip_hdr_length = IP_SIMPLE_HDR_LENGTH + icmp->icmp_ip_snd_options_len; 3957 ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length]; 3958 if ((uchar_t *)ipha < mp1->b_datap->db_base || 3959 mp1->b_datap->db_ref != 1 || 3960 !OK_32PTR(ipha)) { 3961 if (!(mp1 = allocb(ip_hdr_length + icmp_wroff_extra, 3962 BPRI_LO))) { 3963 BUMP_MIB(&rawip_mib, rawipOutErrors); 3964 icmp_ud_err(q, mp1, ENOMEM); 3965 return; 3966 } 3967 mp1->b_cont = mp->b_cont; 3968 ipha = (ipha_t *)mp1->b_datap->db_lim; 3969 mp1->b_wptr = (uchar_t *)ipha; 3970 ipha = (ipha_t *)((uchar_t *)ipha - ip_hdr_length); 3971 } 3972 #ifdef _BIG_ENDIAN 3973 /* Set version, header length, and tos */ 3974 *(uint16_t *)&ipha->ipha_version_and_hdr_length = 3975 ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) | 3976 icmp->icmp_type_of_service); 3977 /* Set ttl and protocol */ 3978 *(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_ttl << 8) | icmp->icmp_proto; 3979 #else 3980 /* Set version, header length, and tos */ 3981 *(uint16_t *)&ipha->ipha_version_and_hdr_length = 3982 ((icmp->icmp_type_of_service << 8) | 3983 ((IP_VERSION << 4) | (ip_hdr_length>>2))); 3984 /* Set ttl and protocol */ 3985 *(uint16_t *)&ipha->ipha_ttl = (icmp->icmp_proto << 8) | icmp->icmp_ttl; 3986 #endif 3987 /* 3988 * Copy our address into the packet. If this is zero, 3989 * ip will fill in the real source address. 3990 */ 3991 IN6_V4MAPPED_TO_IPADDR(&icmp->icmp_v6src, ipha->ipha_src); 3992 ipha->ipha_fragment_offset_and_flags = 0; 3993 3994 /* 3995 * For the socket of SOCK_RAW type, the checksum is provided in the 3996 * pre-built packet. We set the ipha_ident field to IP_HDR_INCLUDED to 3997 * tell IP that the application has sent a complete IP header and not 3998 * to compute the transport checksum nor change the DF flag. 3999 */ 4000 ipha->ipha_ident = IP_HDR_INCLUDED; 4001 4002 /* Finish common formatting of the packet. */ 4003 mp1->b_rptr = (uchar_t *)ipha; 4004 4005 ip_len = mp1->b_wptr - (uchar_t *)ipha; 4006 if (mp1->b_cont != NULL) 4007 ip_len += msgdsize(mp1->b_cont); 4008 4009 /* 4010 * Set the length into the IP header. 4011 * If the length is greater than the maximum allowed by IP, 4012 * then free the message and return. Do not try and send it 4013 * as this can cause problems in layers below. 4014 */ 4015 if (ip_len > IP_MAXPACKET) { 4016 BUMP_MIB(&rawip_mib, rawipOutErrors); 4017 icmp_ud_err(q, mp, EMSGSIZE); 4018 return; 4019 } 4020 ipha->ipha_length = htons((uint16_t)ip_len); 4021 /* 4022 * Copy in the destination address from the T_UNITDATA 4023 * request 4024 */ 4025 if (v4dst == INADDR_ANY) 4026 ipha->ipha_dst = htonl(INADDR_LOOPBACK); 4027 else 4028 ipha->ipha_dst = v4dst; 4029 4030 /* 4031 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic. 4032 */ 4033 if (CLASSD(v4dst)) 4034 ipha->ipha_ttl = icmp->icmp_multicast_ttl; 4035 4036 /* Copy in options if any */ 4037 if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) { 4038 bcopy(icmp->icmp_ip_snd_options, 4039 &ipha[1], icmp->icmp_ip_snd_options_len); 4040 /* 4041 * Massage source route putting first source route in ipha_dst. 4042 * Ignore the destination in the T_unitdata_req. 4043 */ 4044 (void) ip_massage_options(ipha); 4045 } 4046 freeb(mp); 4047 BUMP_MIB(&rawip_mib, rawipOutDatagrams); 4048 putnext(q, mp1); 4049 #undef ipha 4050 #undef tudr 4051 } 4052 4053 /* 4054 * icmp_wput_ipv6(): 4055 * Assumes that icmp_wput did some sanity checking on the destination 4056 * address. 4057 */ 4058 void 4059 icmp_wput_ipv6(queue_t *q, mblk_t *mp, sin6_t *sin6, t_scalar_t tudr_optlen) 4060 { 4061 ip6_t *ip6h; 4062 ip6i_t *ip6i; /* mp1->b_rptr even if no ip6i_t */ 4063 mblk_t *mp1; 4064 int ip_hdr_len = IPV6_HDR_LEN; 4065 size_t ip_len; 4066 icmp_t *icmp; 4067 ip6_pkt_t ipp_s; /* For ancillary data options */ 4068 ip6_pkt_t *ipp = &ipp_s; 4069 ip6_pkt_t *tipp; 4070 uint32_t csum = 0; 4071 uint_t ignore = 0; 4072 uint_t option_exists = 0, is_sticky = 0; 4073 uint8_t *cp; 4074 uint8_t *nxthdr_ptr; 4075 4076 icmp = (icmp_t *)q->q_ptr; 4077 4078 /* 4079 * If the local address is a mapped address return 4080 * an error. 4081 * It would be possible to send an IPv6 packet but the 4082 * response would never make it back to the application 4083 * since it is bound to a mapped address. 4084 */ 4085 if (IN6_IS_ADDR_V4MAPPED(&icmp->icmp_v6src)) { 4086 BUMP_MIB(&rawip_mib, rawipOutErrors); 4087 icmp_ud_err(q, mp, EADDRNOTAVAIL); 4088 return; 4089 } 4090 4091 ipp->ipp_fields = 0; 4092 ipp->ipp_sticky_ignored = 0; 4093 4094 /* 4095 * If TPI options passed in, feed it for verification and handling 4096 */ 4097 if (tudr_optlen != 0) { 4098 int error; 4099 4100 if (icmp_unitdata_opt_process(q, mp, &error, 4101 (void *)ipp) < 0) { 4102 /* failure */ 4103 BUMP_MIB(&rawip_mib, rawipOutErrors); 4104 icmp_ud_err(q, mp, error); 4105 return; 4106 } 4107 ignore = ipp->ipp_sticky_ignored; 4108 ASSERT(error == 0); 4109 } 4110 4111 if (sin6->sin6_scope_id != 0 && 4112 IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) { 4113 /* 4114 * IPPF_SCOPE_ID is special. It's neither a sticky 4115 * option nor ancillary data. It needs to be 4116 * explicitly set in options_exists. 4117 */ 4118 option_exists |= IPPF_SCOPE_ID; 4119 } 4120 4121 if ((icmp->icmp_sticky_ipp.ipp_fields == 0) && 4122 (ipp->ipp_fields == 0)) { 4123 /* No sticky options nor ancillary data. */ 4124 goto no_options; 4125 } 4126 4127 /* 4128 * Go through the options figuring out where each is going to 4129 * come from and build two masks. The first mask indicates if 4130 * the option exists at all. The second mask indicates if the 4131 * option is sticky or ancillary. 4132 */ 4133 if (!(ignore & IPPF_HOPOPTS)) { 4134 if (ipp->ipp_fields & IPPF_HOPOPTS) { 4135 option_exists |= IPPF_HOPOPTS; 4136 ip_hdr_len += ipp->ipp_hopoptslen; 4137 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) { 4138 option_exists |= IPPF_HOPOPTS; 4139 is_sticky |= IPPF_HOPOPTS; 4140 ip_hdr_len += icmp->icmp_sticky_ipp.ipp_hopoptslen; 4141 } 4142 } 4143 4144 if (!(ignore & IPPF_RTHDR)) { 4145 if (ipp->ipp_fields & IPPF_RTHDR) { 4146 option_exists |= IPPF_RTHDR; 4147 ip_hdr_len += ipp->ipp_rthdrlen; 4148 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTHDR) { 4149 option_exists |= IPPF_RTHDR; 4150 is_sticky |= IPPF_RTHDR; 4151 ip_hdr_len += icmp->icmp_sticky_ipp.ipp_rthdrlen; 4152 } 4153 } 4154 4155 if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) { 4156 /* 4157 * Need to have a router header to use these. 4158 */ 4159 if (ipp->ipp_fields & IPPF_RTDSTOPTS) { 4160 option_exists |= IPPF_RTDSTOPTS; 4161 ip_hdr_len += ipp->ipp_rtdstoptslen; 4162 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) { 4163 option_exists |= IPPF_RTDSTOPTS; 4164 is_sticky |= IPPF_RTDSTOPTS; 4165 ip_hdr_len += 4166 icmp->icmp_sticky_ipp.ipp_rtdstoptslen; 4167 } 4168 } 4169 4170 if (!(ignore & IPPF_DSTOPTS)) { 4171 if (ipp->ipp_fields & IPPF_DSTOPTS) { 4172 option_exists |= IPPF_DSTOPTS; 4173 ip_hdr_len += ipp->ipp_dstoptslen; 4174 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) { 4175 option_exists |= IPPF_DSTOPTS; 4176 is_sticky |= IPPF_DSTOPTS; 4177 ip_hdr_len += icmp->icmp_sticky_ipp.ipp_dstoptslen; 4178 } 4179 } 4180 4181 if (!(ignore & IPPF_IFINDEX)) { 4182 if (ipp->ipp_fields & IPPF_IFINDEX) { 4183 option_exists |= IPPF_IFINDEX; 4184 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_IFINDEX) { 4185 option_exists |= IPPF_IFINDEX; 4186 is_sticky |= IPPF_IFINDEX; 4187 } 4188 } 4189 4190 if (!(ignore & IPPF_ADDR)) { 4191 if (ipp->ipp_fields & IPPF_ADDR) { 4192 option_exists |= IPPF_ADDR; 4193 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_ADDR) { 4194 option_exists |= IPPF_ADDR; 4195 is_sticky |= IPPF_ADDR; 4196 } 4197 } 4198 4199 if (!(ignore & IPPF_DONTFRAG)) { 4200 if (ipp->ipp_fields & IPPF_DONTFRAG) { 4201 option_exists |= IPPF_DONTFRAG; 4202 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) { 4203 option_exists |= IPPF_DONTFRAG; 4204 is_sticky |= IPPF_DONTFRAG; 4205 } 4206 } 4207 4208 if (!(ignore & IPPF_USE_MIN_MTU)) { 4209 if (ipp->ipp_fields & IPPF_USE_MIN_MTU) { 4210 option_exists |= IPPF_USE_MIN_MTU; 4211 } else if (icmp->icmp_sticky_ipp.ipp_fields & 4212 IPPF_USE_MIN_MTU) { 4213 option_exists |= IPPF_USE_MIN_MTU; 4214 is_sticky |= IPPF_USE_MIN_MTU; 4215 } 4216 } 4217 4218 if (!(ignore & IPPF_NEXTHOP)) { 4219 if (ipp->ipp_fields & IPPF_NEXTHOP) { 4220 option_exists |= IPPF_NEXTHOP; 4221 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NEXTHOP) { 4222 option_exists |= IPPF_NEXTHOP; 4223 is_sticky |= IPPF_NEXTHOP; 4224 } 4225 } 4226 4227 if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT)) 4228 option_exists |= IPPF_HOPLIMIT; 4229 /* IPV6_HOPLIMIT can never be sticky */ 4230 ASSERT(!(icmp->icmp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT)); 4231 4232 if (!(ignore & IPPF_UNICAST_HOPS) && 4233 (icmp->icmp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) { 4234 option_exists |= IPPF_UNICAST_HOPS; 4235 is_sticky |= IPPF_UNICAST_HOPS; 4236 } 4237 4238 if (!(ignore & IPPF_MULTICAST_HOPS) && 4239 (icmp->icmp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) { 4240 option_exists |= IPPF_MULTICAST_HOPS; 4241 is_sticky |= IPPF_MULTICAST_HOPS; 4242 } 4243 4244 if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_NO_CKSUM) { 4245 /* This is a sticky socket option only */ 4246 option_exists |= IPPF_NO_CKSUM; 4247 is_sticky |= IPPF_NO_CKSUM; 4248 } 4249 4250 if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_RAW_CKSUM) { 4251 /* This is a sticky socket option only */ 4252 option_exists |= IPPF_RAW_CKSUM; 4253 is_sticky |= IPPF_RAW_CKSUM; 4254 } 4255 4256 if (!(ignore & IPPF_TCLASS)) { 4257 if (ipp->ipp_fields & IPPF_TCLASS) { 4258 option_exists |= IPPF_TCLASS; 4259 } else if (icmp->icmp_sticky_ipp.ipp_fields & IPPF_TCLASS) { 4260 option_exists |= IPPF_TCLASS; 4261 is_sticky |= IPPF_TCLASS; 4262 } 4263 } 4264 4265 no_options: 4266 4267 /* 4268 * If any options carried in the ip6i_t were specified, we 4269 * need to account for the ip6i_t in the data we'll be sending 4270 * down. 4271 */ 4272 if (option_exists & IPPF_HAS_IP6I) 4273 ip_hdr_len += sizeof (ip6i_t); 4274 4275 /* check/fix buffer config, setup pointers into it */ 4276 mp1 = mp->b_cont; 4277 ip6h = (ip6_t *)&mp1->b_rptr[-ip_hdr_len]; 4278 if ((mp1->b_datap->db_ref != 1) || 4279 ((unsigned char *)ip6h < mp1->b_datap->db_base) || 4280 !OK_32PTR(ip6h)) { 4281 /* Try to get everything in a single mblk next time */ 4282 if (ip_hdr_len > icmp->icmp_max_hdr_len) { 4283 icmp->icmp_max_hdr_len = ip_hdr_len; 4284 (void) mi_set_sth_wroff(RD(q), 4285 icmp->icmp_max_hdr_len + icmp_wroff_extra); 4286 } 4287 mp1 = allocb(ip_hdr_len + icmp_wroff_extra, BPRI_LO); 4288 if (!mp1) { 4289 BUMP_MIB(&rawip_mib, rawipOutErrors); 4290 icmp_ud_err(q, mp, ENOMEM); 4291 return; 4292 } 4293 mp1->b_cont = mp->b_cont; 4294 mp1->b_wptr = mp1->b_datap->db_lim; 4295 ip6h = (ip6_t *)(mp1->b_wptr - ip_hdr_len); 4296 } 4297 mp1->b_rptr = (unsigned char *)ip6h; 4298 ip6i = (ip6i_t *)ip6h; 4299 4300 #define ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &icmp->icmp_sticky_ipp : ipp) 4301 if (option_exists & IPPF_HAS_IP6I) { 4302 ip6h = (ip6_t *)&ip6i[1]; 4303 ip6i->ip6i_flags = 0; 4304 ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW; 4305 4306 /* sin6_scope_id takes precendence over IPPF_IFINDEX */ 4307 if (option_exists & IPPF_SCOPE_ID) { 4308 ip6i->ip6i_flags |= IP6I_IFINDEX; 4309 ip6i->ip6i_ifindex = sin6->sin6_scope_id; 4310 } else if (option_exists & IPPF_IFINDEX) { 4311 tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX); 4312 ASSERT(tipp->ipp_ifindex != 0); 4313 ip6i->ip6i_flags |= IP6I_IFINDEX; 4314 ip6i->ip6i_ifindex = tipp->ipp_ifindex; 4315 } 4316 4317 if (option_exists & IPPF_RAW_CKSUM) { 4318 ip6i->ip6i_flags |= IP6I_RAW_CHECKSUM; 4319 ip6i->ip6i_checksum_off = icmp->icmp_checksum_off; 4320 } 4321 4322 if (option_exists & IPPF_NO_CKSUM) { 4323 ip6i->ip6i_flags |= IP6I_NO_ULP_CKSUM; 4324 } 4325 4326 if (option_exists & IPPF_ADDR) { 4327 /* 4328 * Enable per-packet source address verification if 4329 * IPV6_PKTINFO specified the source address. 4330 * ip6_src is set in the transport's _wput function. 4331 */ 4332 ip6i->ip6i_flags |= IP6I_VERIFY_SRC; 4333 } 4334 4335 if (option_exists & IPPF_DONTFRAG) { 4336 ip6i->ip6i_flags |= IP6I_DONTFRAG; 4337 } 4338 4339 if (option_exists & IPPF_USE_MIN_MTU) { 4340 ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU( 4341 ip6i->ip6i_flags, ipp->ipp_use_min_mtu); 4342 } 4343 4344 if (option_exists & IPPF_NEXTHOP) { 4345 tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP); 4346 ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop)); 4347 ip6i->ip6i_flags |= IP6I_NEXTHOP; 4348 ip6i->ip6i_nexthop = tipp->ipp_nexthop; 4349 } 4350 4351 /* 4352 * tell IP this is an ip6i_t private header 4353 */ 4354 ip6i->ip6i_nxt = IPPROTO_RAW; 4355 } 4356 4357 /* Initialize IPv6 header */ 4358 ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW; 4359 bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src)); 4360 4361 /* Set the hoplimit of the outgoing packet. */ 4362 if (option_exists & IPPF_HOPLIMIT) { 4363 /* IPV6_HOPLIMIT ancillary data overrides all other settings. */ 4364 ip6h->ip6_hops = ipp->ipp_hoplimit; 4365 ip6i->ip6i_flags |= IP6I_HOPLIMIT; 4366 } else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 4367 ip6h->ip6_hops = icmp->icmp_multicast_ttl; 4368 if (option_exists & IPPF_MULTICAST_HOPS) 4369 ip6i->ip6i_flags |= IP6I_HOPLIMIT; 4370 } else { 4371 ip6h->ip6_hops = icmp->icmp_ttl; 4372 if (option_exists & IPPF_UNICAST_HOPS) 4373 ip6i->ip6i_flags |= IP6I_HOPLIMIT; 4374 } 4375 4376 if (option_exists & IPPF_ADDR) { 4377 tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR); 4378 ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr)); 4379 ip6h->ip6_src = tipp->ipp_addr; 4380 } else { 4381 /* 4382 * The source address was not set using IPV6_PKTINFO. 4383 * First look at the bound source. 4384 * If unspecified fallback to __sin6_src_id. 4385 */ 4386 ip6h->ip6_src = icmp->icmp_v6src; 4387 if (sin6->__sin6_src_id != 0 && 4388 IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) { 4389 ip_srcid_find_id(sin6->__sin6_src_id, 4390 &ip6h->ip6_src, icmp->icmp_zoneid); 4391 } 4392 } 4393 4394 nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt; 4395 cp = (uint8_t *)&ip6h[1]; 4396 4397 /* 4398 * Here's where we have to start stringing together 4399 * any extension headers in the right order: 4400 * Hop-by-hop, destination, routing, and final destination opts. 4401 */ 4402 if (option_exists & IPPF_HOPOPTS) { 4403 /* Hop-by-hop options */ 4404 ip6_hbh_t *hbh = (ip6_hbh_t *)cp; 4405 tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS); 4406 4407 *nxthdr_ptr = IPPROTO_HOPOPTS; 4408 nxthdr_ptr = &hbh->ip6h_nxt; 4409 4410 bcopy(tipp->ipp_hopopts, cp, tipp->ipp_hopoptslen); 4411 cp += tipp->ipp_hopoptslen; 4412 } 4413 /* 4414 * En-route destination options 4415 * Only do them if there's a routing header as well 4416 */ 4417 if (option_exists & IPPF_RTDSTOPTS) { 4418 ip6_dest_t *dst = (ip6_dest_t *)cp; 4419 tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS); 4420 4421 *nxthdr_ptr = IPPROTO_DSTOPTS; 4422 nxthdr_ptr = &dst->ip6d_nxt; 4423 4424 bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen); 4425 cp += tipp->ipp_rtdstoptslen; 4426 } 4427 /* 4428 * Routing header next 4429 */ 4430 if (option_exists & IPPF_RTHDR) { 4431 ip6_rthdr_t *rt = (ip6_rthdr_t *)cp; 4432 tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR); 4433 4434 *nxthdr_ptr = IPPROTO_ROUTING; 4435 nxthdr_ptr = &rt->ip6r_nxt; 4436 4437 bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen); 4438 cp += tipp->ipp_rthdrlen; 4439 } 4440 /* 4441 * Do ultimate destination options 4442 */ 4443 if (option_exists & IPPF_DSTOPTS) { 4444 ip6_dest_t *dest = (ip6_dest_t *)cp; 4445 tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS); 4446 4447 *nxthdr_ptr = IPPROTO_DSTOPTS; 4448 nxthdr_ptr = &dest->ip6d_nxt; 4449 4450 bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen); 4451 cp += tipp->ipp_dstoptslen; 4452 } 4453 4454 /* 4455 * Now set the last header pointer to the proto passed in 4456 */ 4457 ASSERT((int)(cp - (uint8_t *)ip6i) == ip_hdr_len); 4458 *nxthdr_ptr = icmp->icmp_proto; 4459 4460 /* 4461 * Copy in the destination address 4462 */ 4463 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 4464 ip6h->ip6_dst = ipv6_loopback; 4465 else 4466 ip6h->ip6_dst = sin6->sin6_addr; 4467 4468 ip6h->ip6_vcf = 4469 (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) | 4470 (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK); 4471 4472 if (option_exists & IPPF_TCLASS) { 4473 tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS); 4474 ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf, 4475 tipp->ipp_tclass); 4476 } 4477 if (option_exists & IPPF_RTHDR) { 4478 ip6_rthdr_t *rth; 4479 4480 /* 4481 * Perform any processing needed for source routing. 4482 * We know that all extension headers will be in the same mblk 4483 * as the IPv6 header. 4484 */ 4485 rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr); 4486 if (rth != NULL && rth->ip6r_segleft != 0) { 4487 if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) { 4488 /* 4489 * Drop packet - only support Type 0 routing. 4490 * Notify the application as well. 4491 */ 4492 icmp_ud_err(q, mp, EPROTO); 4493 BUMP_MIB(&rawip_mib, rawipOutErrors); 4494 return; 4495 } 4496 /* 4497 * rth->ip6r_len is twice the number of 4498 * addresses in the header 4499 */ 4500 if (rth->ip6r_len & 0x1) { 4501 icmp_ud_err(q, mp, EPROTO); 4502 BUMP_MIB(&rawip_mib, rawipOutErrors); 4503 return; 4504 } 4505 /* 4506 * Shuffle the routing header and ip6_dst 4507 * addresses, and get the checksum difference 4508 * between the first hop (in ip6_dst) and 4509 * the destination (in the last routing hdr entry). 4510 */ 4511 csum = ip_massage_options_v6(ip6h, rth); 4512 /* 4513 * Verify that the first hop isn't a mapped address. 4514 * Routers along the path need to do this verification 4515 * for subsequent hops. 4516 */ 4517 if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) { 4518 icmp_ud_err(q, mp, EADDRNOTAVAIL); 4519 BUMP_MIB(&rawip_mib, rawipOutErrors); 4520 return; 4521 } 4522 } 4523 } 4524 4525 ip_len = mp1->b_wptr - (uchar_t *)ip6h - IPV6_HDR_LEN; 4526 if (mp1->b_cont != NULL) 4527 ip_len += msgdsize(mp1->b_cont); 4528 4529 /* 4530 * Set the length into the IP header. 4531 * If the length is greater than the maximum allowed by IP, 4532 * then free the message and return. Do not try and send it 4533 * as this can cause problems in layers below. 4534 */ 4535 if (ip_len > IP_MAXPACKET) { 4536 BUMP_MIB(&rawip_mib, rawipOutErrors); 4537 icmp_ud_err(q, mp1, EMSGSIZE); 4538 return; 4539 } 4540 if (icmp->icmp_proto == IPPROTO_ICMPV6 || icmp->icmp_raw_checksum) { 4541 uint_t cksum_off; /* From ip6i == mp1->b_rptr */ 4542 uint16_t *cksum_ptr; 4543 uint_t ext_hdrs_len; 4544 4545 /* ICMPv6 must have an offset matching icmp6_cksum offset */ 4546 ASSERT(icmp->icmp_proto != IPPROTO_ICMPV6 || 4547 icmp->icmp_checksum_off == 2); 4548 4549 /* 4550 * We make it easy for IP to include our pseudo header 4551 * by putting our length in uh_checksum, modified (if 4552 * we have a routing header) by the checksum difference 4553 * between the ultimate destination and first hop addresses. 4554 * Note: ICMPv6 must always checksum the packet. 4555 */ 4556 cksum_off = ip_hdr_len + icmp->icmp_checksum_off; 4557 if (cksum_off + sizeof (uint16_t) > mp1->b_wptr - mp1->b_rptr) { 4558 if (!pullupmsg(mp1, cksum_off + sizeof (uint16_t))) { 4559 BUMP_MIB(&rawip_mib, rawipOutErrors); 4560 freemsg(mp); 4561 return; 4562 } 4563 ip6i = (ip6i_t *)mp1->b_rptr; 4564 if (ip6i->ip6i_nxt == IPPROTO_RAW) 4565 ip6h = (ip6_t *)&ip6i[1]; 4566 else 4567 ip6h = (ip6_t *)ip6i; 4568 } 4569 /* Add payload length to checksum */ 4570 ext_hdrs_len = ip_hdr_len - IPV6_HDR_LEN - 4571 (int)((uchar_t *)ip6h - (uchar_t *)ip6i); 4572 csum += htons(ip_len - ext_hdrs_len); 4573 4574 cksum_ptr = (uint16_t *)((uchar_t *)ip6i + cksum_off); 4575 csum = (csum & 0xFFFF) + (csum >> 16); 4576 *cksum_ptr = (uint16_t)csum; 4577 } 4578 4579 #ifdef _LITTLE_ENDIAN 4580 ip_len = htons(ip_len); 4581 #endif 4582 ip6h->ip6_plen = (uint16_t)ip_len; 4583 4584 freeb(mp); 4585 4586 /* We're done. Pass the packet to IP */ 4587 BUMP_MIB(&rawip_mib, rawipOutDatagrams); 4588 putnext(q, mp1); 4589 } 4590 4591 static void 4592 icmp_wput_other(queue_t *q, mblk_t *mp) 4593 { 4594 uchar_t *rptr = mp->b_rptr; 4595 struct iocblk *iocp; 4596 #define tudr ((struct T_unitdata_req *)rptr) 4597 icmp_t *icmp; 4598 cred_t *cr; 4599 4600 icmp = (icmp_t *)q->q_ptr; 4601 4602 cr = DB_CREDDEF(mp, icmp->icmp_credp); 4603 4604 switch (mp->b_datap->db_type) { 4605 case M_PROTO: 4606 case M_PCPROTO: 4607 if (mp->b_wptr - rptr < sizeof (t_scalar_t)) { 4608 /* 4609 * If the message does not contain a PRIM_type, 4610 * throw it away. 4611 */ 4612 freemsg(mp); 4613 return; 4614 } 4615 switch (((union T_primitives *)rptr)->type) { 4616 case T_ADDR_REQ: 4617 icmp_addr_req(q, mp); 4618 return; 4619 case O_T_BIND_REQ: 4620 case T_BIND_REQ: 4621 qwriter(q, mp, icmp_bind, PERIM_OUTER); 4622 return; 4623 case T_CONN_REQ: 4624 icmp_connect(q, mp); 4625 return; 4626 case T_CAPABILITY_REQ: 4627 icmp_capability_req(q, mp); 4628 return; 4629 case T_INFO_REQ: 4630 icmp_info_req(q, mp); 4631 return; 4632 case T_UNITDATA_REQ: 4633 /* 4634 * If a T_UNITDATA_REQ gets here, the address must 4635 * be bad. Valid T_UNITDATA_REQs are found above 4636 * and break to below this switch. 4637 */ 4638 icmp_ud_err(q, mp, EADDRNOTAVAIL); 4639 return; 4640 case T_UNBIND_REQ: 4641 icmp_unbind(q, mp); 4642 return; 4643 4644 case T_SVR4_OPTMGMT_REQ: 4645 if (!snmpcom_req(q, mp, icmp_snmp_set, icmp_snmp_get, 4646 cr)) 4647 /* Only IP can return anything meaningful */ 4648 (void) svr4_optcom_req(q, mp, cr, 4649 &icmp_opt_obj); 4650 return; 4651 4652 case T_OPTMGMT_REQ: 4653 /* Only IP can return anything meaningful */ 4654 (void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj); 4655 return; 4656 4657 case T_DISCON_REQ: 4658 icmp_disconnect(q, mp); 4659 return; 4660 4661 /* The following TPI message is not supported by icmp. */ 4662 case O_T_CONN_RES: 4663 case T_CONN_RES: 4664 icmp_err_ack(q, mp, TNOTSUPPORT, 0); 4665 return; 4666 4667 /* The following 3 TPI requests are illegal for icmp. */ 4668 case T_DATA_REQ: 4669 case T_EXDATA_REQ: 4670 case T_ORDREL_REQ: 4671 freemsg(mp); 4672 (void) putctl1(RD(q), M_ERROR, EPROTO); 4673 return; 4674 default: 4675 break; 4676 } 4677 break; 4678 case M_IOCTL: 4679 iocp = (struct iocblk *)mp->b_rptr; 4680 switch (iocp->ioc_cmd) { 4681 case TI_GETPEERNAME: 4682 if (icmp->icmp_state != TS_DATA_XFER) { 4683 /* 4684 * If a default destination address has not 4685 * been associated with the stream, then we 4686 * don't know the peer's name. 4687 */ 4688 iocp->ioc_error = ENOTCONN; 4689 err_ret:; 4690 iocp->ioc_count = 0; 4691 mp->b_datap->db_type = M_IOCACK; 4692 qreply(q, mp); 4693 return; 4694 } 4695 /* FALLTHRU */ 4696 case TI_GETMYNAME: 4697 /* 4698 * For TI_GETPEERNAME and TI_GETMYNAME, we first 4699 * need to copyin the user's strbuf structure. 4700 * Processing will continue in the M_IOCDATA case 4701 * below. 4702 */ 4703 mi_copyin(q, mp, NULL, 4704 SIZEOF_STRUCT(strbuf, iocp->ioc_flag)); 4705 return; 4706 case ND_SET: 4707 /* nd_getset performs the necessary error checking */ 4708 case ND_GET: 4709 if (nd_getset(q, icmp_g_nd, mp)) { 4710 qreply(q, mp); 4711 return; 4712 } 4713 break; 4714 default: 4715 break; 4716 } 4717 break; 4718 case M_IOCDATA: 4719 icmp_wput_iocdata(q, mp); 4720 return; 4721 default: 4722 break; 4723 } 4724 putnext(q, mp); 4725 } 4726 4727 /* 4728 * icmp_wput_iocdata is called by icmp_wput_slow to handle all M_IOCDATA 4729 * messages. 4730 */ 4731 static void 4732 icmp_wput_iocdata(queue_t *q, mblk_t *mp) 4733 { 4734 mblk_t *mp1; 4735 STRUCT_HANDLE(strbuf, sb); 4736 icmp_t *icmp; 4737 in6_addr_t v6addr; 4738 ipaddr_t v4addr; 4739 uint32_t flowinfo = 0; 4740 int addrlen; 4741 4742 /* Make sure it is one of ours. */ 4743 switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) { 4744 case TI_GETMYNAME: 4745 case TI_GETPEERNAME: 4746 break; 4747 default: 4748 putnext(q, mp); 4749 return; 4750 } 4751 switch (mi_copy_state(q, mp, &mp1)) { 4752 case -1: 4753 return; 4754 case MI_COPY_CASE(MI_COPY_IN, 1): 4755 break; 4756 case MI_COPY_CASE(MI_COPY_OUT, 1): 4757 /* 4758 * The address has been copied out, so now 4759 * copyout the strbuf. 4760 */ 4761 mi_copyout(q, mp); 4762 return; 4763 case MI_COPY_CASE(MI_COPY_OUT, 2): 4764 /* 4765 * The address and strbuf have been copied out. 4766 * We're done, so just acknowledge the original 4767 * M_IOCTL. 4768 */ 4769 mi_copy_done(q, mp, 0); 4770 return; 4771 default: 4772 /* 4773 * Something strange has happened, so acknowledge 4774 * the original M_IOCTL with an EPROTO error. 4775 */ 4776 mi_copy_done(q, mp, EPROTO); 4777 return; 4778 } 4779 /* 4780 * Now we have the strbuf structure for TI_GETMYNAME 4781 * and TI_GETPEERNAME. Next we copyout the requested 4782 * address and then we'll copyout the strbuf. 4783 */ 4784 STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag, 4785 (void *)mp1->b_rptr); 4786 icmp = (icmp_t *)q->q_ptr; 4787 if (icmp->icmp_family == AF_INET) 4788 addrlen = sizeof (sin_t); 4789 else 4790 addrlen = sizeof (sin6_t); 4791 4792 if (STRUCT_FGET(sb, maxlen) < addrlen) { 4793 mi_copy_done(q, mp, EINVAL); 4794 return; 4795 } 4796 switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) { 4797 case TI_GETMYNAME: 4798 if (icmp->icmp_family == AF_INET) { 4799 ASSERT(icmp->icmp_ipversion == IPV4_VERSION); 4800 if (!IN6_IS_ADDR_V4MAPPED_ANY(&icmp->icmp_v6src) && 4801 !IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) { 4802 v4addr = V4_PART_OF_V6(icmp->icmp_v6src); 4803 } else { 4804 /* 4805 * INADDR_ANY 4806 * icmp_v6src is not set, we might be bound to 4807 * broadcast/multicast. Use icmp_bound_v6src as 4808 * local address instead (that could 4809 * also still be INADDR_ANY) 4810 */ 4811 v4addr = V4_PART_OF_V6(icmp->icmp_bound_v6src); 4812 } 4813 } else { 4814 /* icmp->icmp_family == AF_INET6 */ 4815 if (!IN6_IS_ADDR_UNSPECIFIED(&icmp->icmp_v6src)) { 4816 v6addr = icmp->icmp_v6src; 4817 } else { 4818 /* 4819 * UNSPECIFIED 4820 * icmp_v6src is not set, we might be bound to 4821 * broadcast/multicast. Use icmp_bound_v6src as 4822 * local address instead (that could 4823 * also still be UNSPECIFIED) 4824 */ 4825 v6addr = icmp->icmp_bound_v6src; 4826 } 4827 } 4828 break; 4829 case TI_GETPEERNAME: 4830 if (icmp->icmp_family == AF_INET) { 4831 ASSERT(icmp->icmp_ipversion == IPV4_VERSION); 4832 v4addr = V4_PART_OF_V6(icmp->icmp_v6dst); 4833 } else { 4834 /* icmp->icmp_family == AF_INET6) */ 4835 v6addr = icmp->icmp_v6dst; 4836 flowinfo = icmp->icmp_flowinfo; 4837 } 4838 break; 4839 default: 4840 mi_copy_done(q, mp, EPROTO); 4841 return; 4842 } 4843 mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE); 4844 if (!mp1) 4845 return; 4846 4847 if (icmp->icmp_family == AF_INET) { 4848 sin_t *sin; 4849 4850 STRUCT_FSET(sb, len, (int)sizeof (sin_t)); 4851 sin = (sin_t *)mp1->b_rptr; 4852 mp1->b_wptr = (uchar_t *)&sin[1]; 4853 *sin = sin_null; 4854 sin->sin_family = AF_INET; 4855 sin->sin_addr.s_addr = v4addr; 4856 } else { 4857 /* icmp->icmp_family == AF_INET6 */ 4858 sin6_t *sin6; 4859 4860 ASSERT(icmp->icmp_family == AF_INET6); 4861 STRUCT_FSET(sb, len, (int)sizeof (sin6_t)); 4862 sin6 = (sin6_t *)mp1->b_rptr; 4863 mp1->b_wptr = (uchar_t *)&sin6[1]; 4864 *sin6 = sin6_null; 4865 sin6->sin6_family = AF_INET6; 4866 sin6->sin6_flowinfo = flowinfo; 4867 sin6->sin6_addr = v6addr; 4868 } 4869 /* Copy out the address */ 4870 mi_copyout(q, mp); 4871 } 4872 4873 /* 4874 * Only allow MIB requests and M_FLUSHes to pass. 4875 * All other messages are nacked or dropped. 4876 */ 4877 static void 4878 icmp_wput_restricted(queue_t *q, mblk_t *mp) 4879 { 4880 cred_t *cr; 4881 icmp_t *icmp; 4882 4883 switch (DB_TYPE(mp)) { 4884 case M_PROTO: 4885 case M_PCPROTO: 4886 if (MBLKL(mp) < sizeof (t_scalar_t)) { 4887 freemsg(mp); 4888 return; 4889 } 4890 icmp = (icmp_t *)q->q_ptr; 4891 cr = DB_CREDDEF(mp, icmp->icmp_credp); 4892 4893 switch (((union T_primitives *)mp->b_rptr)->type) { 4894 case T_SVR4_OPTMGMT_REQ: 4895 if (!snmpcom_req(q, mp, 4896 icmp_snmp_set, icmp_snmp_get, cr)) 4897 (void) svr4_optcom_req(q, mp, cr, 4898 &icmp_opt_obj); 4899 return; 4900 case T_OPTMGMT_REQ: 4901 (void) tpi_optcom_req(q, mp, cr, &icmp_opt_obj); 4902 return; 4903 default: 4904 icmp_err_ack(q, mp, TSYSERR, ENOTSUP); 4905 return; 4906 } 4907 /* NOTREACHED */ 4908 case M_IOCTL: 4909 miocnak(q, mp, 0, ENOTSUP); 4910 break; 4911 case M_FLUSH: 4912 putnext(q, mp); 4913 break; 4914 default: 4915 freemsg(mp); 4916 break; 4917 } 4918 } 4919 4920 static int 4921 icmp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp, 4922 void *thisdg_attrs) 4923 { 4924 icmp_t *icmp; 4925 struct T_unitdata_req *udreqp; 4926 int is_absreq_failure; 4927 cred_t *cr; 4928 4929 icmp = (icmp_t *)q->q_ptr; 4930 4931 udreqp = (struct T_unitdata_req *)mp->b_rptr; 4932 *errorp = 0; 4933 4934 cr = DB_CREDDEF(mp, icmp->icmp_credp); 4935 4936 *errorp = tpi_optcom_buf(q, mp, &udreqp->OPT_length, 4937 udreqp->OPT_offset, cr, &icmp_opt_obj, 4938 thisdg_attrs, &is_absreq_failure); 4939 4940 if (*errorp != 0) { 4941 /* 4942 * Note: No special action needed in this 4943 * module for "is_absreq_failure" 4944 */ 4945 return (-1); /* failure */ 4946 } 4947 ASSERT(is_absreq_failure == 0); 4948 return (0); /* success */ 4949 } 4950 4951 void 4952 icmp_ddi_init(void) 4953 { 4954 ICMP6_MAJ = ddi_name_to_major(ICMP6); 4955 icmp_max_optsize = 4956 optcom_max_optsize(icmp_opt_obj.odb_opt_des_arr, 4957 icmp_opt_obj.odb_opt_arr_cnt); 4958 4959 (void) icmp_param_register(icmp_param_arr, A_CNT(icmp_param_arr)); 4960 4961 rawip_kstat_init(); 4962 } 4963 4964 void 4965 icmp_ddi_destroy(void) 4966 { 4967 nd_free(&icmp_g_nd); 4968 4969 rawip_kstat_fini(); 4970 } 4971 4972 static void 4973 rawip_kstat_init(void) { 4974 4975 rawip_named_kstat_t template = { 4976 { "inDatagrams", KSTAT_DATA_UINT32, 0 }, 4977 { "inCksumErrs", KSTAT_DATA_UINT32, 0 }, 4978 { "inErrors", KSTAT_DATA_UINT32, 0 }, 4979 { "outDatagrams", KSTAT_DATA_UINT32, 0 }, 4980 { "outErrors", KSTAT_DATA_UINT32, 0 }, 4981 }; 4982 4983 rawip_mibkp = kstat_create("icmp", 0, "rawip", "mib2", 4984 KSTAT_TYPE_NAMED, 4985 NUM_OF_FIELDS(rawip_named_kstat_t), 4986 0); 4987 if (rawip_mibkp == NULL) 4988 return; 4989 4990 bcopy(&template, rawip_mibkp->ks_data, sizeof (template)); 4991 4992 rawip_mibkp->ks_update = rawip_kstat_update; 4993 4994 kstat_install(rawip_mibkp); 4995 } 4996 4997 static void 4998 rawip_kstat_fini(void) { 4999 if (rawip_mibkp) { 5000 kstat_delete(rawip_mibkp); 5001 rawip_mibkp = NULL; 5002 } 5003 } 5004 5005 static int 5006 rawip_kstat_update(kstat_t *kp, int rw) { 5007 rawip_named_kstat_t *rawipkp; 5008 5009 if ((kp == NULL) || (kp->ks_data == NULL)) 5010 return (EIO); 5011 5012 if (rw == KSTAT_WRITE) 5013 return (EACCES); 5014 5015 rawipkp = (rawip_named_kstat_t *)kp->ks_data; 5016 5017 rawipkp->inDatagrams.value.ui32 = rawip_mib.rawipInDatagrams; 5018 rawipkp->inCksumErrs.value.ui32 = rawip_mib.rawipInCksumErrs; 5019 rawipkp->inErrors.value.ui32 = rawip_mib.rawipInErrors; 5020 rawipkp->outDatagrams.value.ui32 = rawip_mib.rawipOutDatagrams; 5021 rawipkp->outErrors.value.ui32 = rawip_mib.rawipOutErrors; 5022 5023 return (0); 5024 } 5025