1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 /* Copyright (c) 1990 Mentat Inc. */ 26 27 #pragma ident "%Z%%M% %I% %E% SMI" 28 29 const char udp_version[] = "%Z%%M% %I% %E% SMI"; 30 31 #include <sys/types.h> 32 #include <sys/stream.h> 33 #include <sys/dlpi.h> 34 #include <sys/pattr.h> 35 #include <sys/stropts.h> 36 #include <sys/strlog.h> 37 #include <sys/strsun.h> 38 #include <sys/time.h> 39 #define _SUN_TPI_VERSION 2 40 #include <sys/tihdr.h> 41 #include <sys/timod.h> 42 #include <sys/ddi.h> 43 #include <sys/sunddi.h> 44 #include <sys/strsubr.h> 45 #include <sys/suntpi.h> 46 #include <sys/xti_inet.h> 47 #include <sys/cmn_err.h> 48 #include <sys/kmem.h> 49 #include <sys/policy.h> 50 #include <sys/ucred.h> 51 #include <sys/zone.h> 52 53 #include <sys/socket.h> 54 #include <sys/sockio.h> 55 #include <sys/vtrace.h> 56 #include <sys/debug.h> 57 #include <sys/isa_defs.h> 58 #include <sys/random.h> 59 #include <netinet/in.h> 60 #include <netinet/ip6.h> 61 #include <netinet/icmp6.h> 62 #include <netinet/udp.h> 63 #include <net/if.h> 64 #include <net/route.h> 65 66 #include <inet/common.h> 67 #include <inet/ip.h> 68 #include <inet/ip_impl.h> 69 #include <inet/ip6.h> 70 #include <inet/ip_ire.h> 71 #include <inet/ip_if.h> 72 #include <inet/ip_multi.h> 73 #include <inet/ip_ndp.h> 74 #include <inet/mi.h> 75 #include <inet/mib2.h> 76 #include <inet/nd.h> 77 #include <inet/optcom.h> 78 #include <inet/snmpcom.h> 79 #include <inet/kstatcom.h> 80 #include <inet/udp_impl.h> 81 #include <inet/ipclassifier.h> 82 #include <inet/ipsec_impl.h> 83 #include <inet/ipp_common.h> 84 85 /* 86 * The ipsec_info.h header file is here since it has the definition for the 87 * M_CTL message types used by IP to convey information to the ULP. The 88 * ipsec_info.h needs the pfkeyv2.h, hence the latter's presence. 89 */ 90 #include <net/pfkeyv2.h> 91 #include <inet/ipsec_info.h> 92 93 #include <sys/tsol/label.h> 94 #include <sys/tsol/tnet.h> 95 #include <rpc/pmap_prot.h> 96 97 /* 98 * Synchronization notes: 99 * 100 * UDP uses a combination of its internal perimeter, a global lock and 101 * a set of bind hash locks to protect its data structures. Please see 102 * the note above udp_mode_assertions for details about the internal 103 * perimeter. 104 * 105 * When a UDP endpoint is bound to a local port, it is inserted into 106 * a bind hash list. The list consists of an array of udp_fanout_t buckets. 107 * The size of the array is controlled by the udp_bind_fanout_size variable. 108 * This variable can be changed in /etc/system if the default value is 109 * not large enough. Each bind hash bucket is protected by a per bucket 110 * lock. It protects the udp_bind_hash and udp_ptpbhn fields in the udp_t 111 * structure. An UDP endpoint is removed from the bind hash list only 112 * when it is being unbound or being closed. The per bucket lock also 113 * protects a UDP endpoint's state changes. 114 * 115 * Plumbing notes: 116 * 117 * Both udp and ip are merged, but the streams plumbing is kept unchanged 118 * in that udp is always pushed atop /dev/ip. This is done to preserve 119 * backwards compatibility for certain applications which rely on such 120 * plumbing geometry to do things such as issuing I_POP on the stream 121 * in order to obtain direct access to /dev/ip, etc. 122 * 123 * All UDP processings happen in the /dev/ip instance; the udp module 124 * instance does not possess any state about the endpoint, and merely 125 * acts as a dummy module whose presence is to keep the streams plumbing 126 * appearance unchanged. At open time /dev/ip allocates a conn_t that 127 * happens to embed a udp_t. This stays dormant until the time udp is 128 * pushed, which indicates to /dev/ip that it must convert itself from 129 * an IP to a UDP endpoint. 130 * 131 * We only allow for the following plumbing cases: 132 * 133 * Normal: 134 * /dev/ip is first opened and later udp is pushed directly on top. 135 * This is the default action that happens when a udp socket or 136 * /dev/udp is opened. The conn_t created by /dev/ip instance is 137 * now shared and is marked with IPCL_UDP. 138 * 139 * SNMP-only: 140 * udp is pushed on top of a module other than /dev/ip. When this 141 * happens it will support only SNMP semantics. A new conn_t is 142 * allocated and marked with IPCL_UDPMOD. 143 * 144 * The above cases imply that we don't support any intermediate module to 145 * reside in between /dev/ip and udp -- in fact, we never supported such 146 * scenario in the past as the inter-layer communication semantics have 147 * always been private. Also note that the normal case allows for SNMP 148 * requests to be processed in addition to the rest of UDP operations. 149 * 150 * The normal case plumbing is depicted by the following diagram: 151 * 152 * +---------------+---------------+ 153 * | | | udp 154 * | udp_wq | udp_rq | 155 * | | UDP_RD | 156 * | | | 157 * +---------------+---------------+ 158 * | ^ 159 * v | 160 * +---------------+---------------+ 161 * | | | /dev/ip 162 * | ip_wq | ip_rq | conn_t 163 * | UDP_WR | | 164 * | | | 165 * +---------------+---------------+ 166 * 167 * Messages arriving at udp_wq from above will end up in ip_wq before 168 * it gets processed, i.e. udp write entry points will advance udp_wq 169 * and use its q_next value as ip_wq in order to use the conn_t that 170 * is stored in its q_ptr. Likewise, messages generated by ip to the 171 * module above udp will appear as if they are originated from udp_rq, 172 * i.e. putnext() calls to the module above udp is done using the 173 * udp_rq instead of ip_rq in order to avoid udp_rput() which does 174 * nothing more than calling putnext(). 175 * 176 * The above implies the following rule of thumb: 177 * 178 * 1. udp_t is obtained from conn_t, which is created by the /dev/ip 179 * instance and is stored in q_ptr of both ip_wq and ip_rq. There 180 * is no direct reference to conn_t from either udp_wq or udp_rq. 181 * 182 * 2. Write-side entry points of udp can obtain the conn_t via the 183 * Q_TO_CONN() macro, using the queue value obtain from UDP_WR(). 184 * 185 * 3. While in /dev/ip context, putnext() to the module above udp can 186 * be done by supplying the queue value obtained from UDP_RD(). 187 * 188 */ 189 190 static queue_t *UDP_WR(queue_t *); 191 static queue_t *UDP_RD(queue_t *); 192 193 udp_stat_t udp_statistics = { 194 { "udp_ip_send", KSTAT_DATA_UINT64 }, 195 { "udp_ip_ire_send", KSTAT_DATA_UINT64 }, 196 { "udp_ire_null", KSTAT_DATA_UINT64 }, 197 { "udp_drain", KSTAT_DATA_UINT64 }, 198 { "udp_sock_fallback", KSTAT_DATA_UINT64 }, 199 { "udp_rrw_busy", KSTAT_DATA_UINT64 }, 200 { "udp_rrw_msgcnt", KSTAT_DATA_UINT64 }, 201 { "udp_out_sw_cksum", KSTAT_DATA_UINT64 }, 202 { "udp_out_sw_cksum_bytes", KSTAT_DATA_UINT64 }, 203 { "udp_out_opt", KSTAT_DATA_UINT64 }, 204 { "udp_out_err_notconn", KSTAT_DATA_UINT64 }, 205 { "udp_out_err_output", KSTAT_DATA_UINT64 }, 206 { "udp_out_err_tudr", KSTAT_DATA_UINT64 }, 207 { "udp_in_pktinfo", KSTAT_DATA_UINT64 }, 208 { "udp_in_recvdstaddr", KSTAT_DATA_UINT64 }, 209 { "udp_in_recvopts", KSTAT_DATA_UINT64 }, 210 { "udp_in_recvif", KSTAT_DATA_UINT64 }, 211 { "udp_in_recvslla", KSTAT_DATA_UINT64 }, 212 { "udp_in_recvucred", KSTAT_DATA_UINT64 }, 213 { "udp_in_recvttl", KSTAT_DATA_UINT64 }, 214 { "udp_in_recvhopopts", KSTAT_DATA_UINT64 }, 215 { "udp_in_recvhoplimit", KSTAT_DATA_UINT64 }, 216 { "udp_in_recvdstopts", KSTAT_DATA_UINT64 }, 217 { "udp_in_recvrtdstopts", KSTAT_DATA_UINT64 }, 218 { "udp_in_recvrthdr", KSTAT_DATA_UINT64 }, 219 { "udp_in_recvpktinfo", KSTAT_DATA_UINT64 }, 220 { "udp_in_recvtclass", KSTAT_DATA_UINT64 }, 221 { "udp_in_timestamp", KSTAT_DATA_UINT64 }, 222 #ifdef DEBUG 223 { "udp_data_conn", KSTAT_DATA_UINT64 }, 224 { "udp_data_notconn", KSTAT_DATA_UINT64 }, 225 #endif 226 }; 227 228 static kstat_t *udp_ksp; 229 struct kmem_cache *udp_cache; 230 231 /* 232 * Bind hash list size and hash function. It has to be a power of 2 for 233 * hashing. 234 */ 235 #define UDP_BIND_FANOUT_SIZE 512 236 #define UDP_BIND_HASH(lport) \ 237 ((ntohs((uint16_t)lport)) & (udp_bind_fanout_size - 1)) 238 239 /* UDP bind fanout hash structure. */ 240 typedef struct udp_fanout_s { 241 udp_t *uf_udp; 242 kmutex_t uf_lock; 243 #if defined(_LP64) || defined(_I32LPx) 244 char uf_pad[48]; 245 #else 246 char uf_pad[56]; 247 #endif 248 } udp_fanout_t; 249 250 uint_t udp_bind_fanout_size = UDP_BIND_FANOUT_SIZE; 251 /* udp_fanout_t *udp_bind_fanout. */ 252 static udp_fanout_t *udp_bind_fanout; 253 254 /* 255 * This controls the rate some ndd info report functions can be used 256 * by non-privileged users. It stores the last time such info is 257 * requested. When those report functions are called again, this 258 * is checked with the current time and compare with the ndd param 259 * udp_ndd_get_info_interval. 260 */ 261 static clock_t udp_last_ndd_get_info_time; 262 #define NDD_TOO_QUICK_MSG \ 263 "ndd get info rate too high for non-privileged users, try again " \ 264 "later.\n" 265 #define NDD_OUT_OF_BUF_MSG "<< Out of buffer >>\n" 266 267 /* Option processing attrs */ 268 typedef struct udpattrs_s { 269 ip6_pkt_t *udpattr_ipp; 270 mblk_t *udpattr_mb; 271 boolean_t udpattr_credset; 272 } udpattrs_t; 273 274 static void udp_addr_req(queue_t *q, mblk_t *mp); 275 static void udp_bind(queue_t *q, mblk_t *mp); 276 static void udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp); 277 static void udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock); 278 static int udp_build_hdrs(queue_t *q, udp_t *udp); 279 static void udp_capability_req(queue_t *q, mblk_t *mp); 280 static int udp_close(queue_t *q); 281 static void udp_connect(queue_t *q, mblk_t *mp); 282 static void udp_disconnect(queue_t *q, mblk_t *mp); 283 static void udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, 284 int sys_error); 285 static void udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive, 286 t_scalar_t tlierr, int unixerr); 287 static int udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp, 288 cred_t *cr); 289 static int udp_extra_priv_ports_add(queue_t *q, mblk_t *mp, 290 char *value, caddr_t cp, cred_t *cr); 291 static int udp_extra_priv_ports_del(queue_t *q, mblk_t *mp, 292 char *value, caddr_t cp, cred_t *cr); 293 static void udp_icmp_error(queue_t *q, mblk_t *mp); 294 static void udp_icmp_error_ipv6(queue_t *q, mblk_t *mp); 295 static void udp_info_req(queue_t *q, mblk_t *mp); 296 static mblk_t *udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim, 297 t_scalar_t addr_length); 298 static int udp_open(queue_t *q, dev_t *devp, int flag, int sflag, 299 cred_t *credp); 300 static int udp_unitdata_opt_process(queue_t *q, mblk_t *mp, 301 int *errorp, udpattrs_t *udpattrs); 302 static boolean_t udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name); 303 static int udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr); 304 static boolean_t udp_param_register(udpparam_t *udppa, int cnt); 305 static int udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, 306 cred_t *cr); 307 static void udp_report_item(mblk_t *mp, udp_t *udp); 308 static void udp_rput(queue_t *q, mblk_t *mp); 309 static void udp_rput_other(queue_t *, mblk_t *); 310 static int udp_rinfop(queue_t *q, infod_t *dp); 311 static int udp_rrw(queue_t *q, struiod_t *dp); 312 static void udp_rput_bind_ack(queue_t *q, mblk_t *mp); 313 static int udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, 314 cred_t *cr); 315 static void udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp, ipha_t *ipha); 316 static void udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr, 317 t_scalar_t destlen, t_scalar_t err); 318 static void udp_unbind(queue_t *q, mblk_t *mp); 319 static in_port_t udp_update_next_port(udp_t *udp, in_port_t port, 320 boolean_t random); 321 static void udp_wput(queue_t *q, mblk_t *mp); 322 static mblk_t *udp_output_v4(conn_t *, mblk_t *mp, ipaddr_t v4dst, 323 uint16_t port, uint_t srcid, int *error); 324 static mblk_t *udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6, 325 int *error); 326 static void udp_wput_other(queue_t *q, mblk_t *mp); 327 static void udp_wput_iocdata(queue_t *q, mblk_t *mp); 328 static void udp_output(conn_t *connp, mblk_t *mp, struct sockaddr *addr, 329 socklen_t addrlen); 330 static size_t udp_set_rcv_hiwat(udp_t *udp, size_t size); 331 332 static void udp_kstat_init(void); 333 static void udp_kstat_fini(void); 334 static int udp_kstat_update(kstat_t *kp, int rw); 335 static void udp_input_wrapper(void *arg, mblk_t *mp, void *arg2); 336 static void udp_rput_other_wrapper(void *arg, mblk_t *mp, void *arg2); 337 static void udp_wput_other_wrapper(void *arg, mblk_t *mp, void *arg2); 338 static void udp_resume_bind_cb(void *arg, mblk_t *mp, void *arg2); 339 340 static void udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp, 341 uint_t pkt_len); 342 static void udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing); 343 static void udp_enter(conn_t *, mblk_t *, sqproc_t, uint8_t); 344 static void udp_exit(conn_t *); 345 static void udp_become_writer(conn_t *, mblk_t *, sqproc_t, uint8_t); 346 #ifdef DEBUG 347 static void udp_mode_assertions(udp_t *, int); 348 #endif /* DEBUG */ 349 350 major_t UDP6_MAJ; 351 #define UDP6 "udp6" 352 353 #define UDP_RECV_HIWATER (56 * 1024) 354 #define UDP_RECV_LOWATER 128 355 #define UDP_XMIT_HIWATER (56 * 1024) 356 #define UDP_XMIT_LOWATER 1024 357 358 static struct module_info udp_info = { 359 UDP_MOD_ID, UDP_MOD_NAME, 1, INFPSZ, UDP_RECV_HIWATER, UDP_RECV_LOWATER 360 }; 361 362 static struct qinit udp_rinit = { 363 (pfi_t)udp_rput, NULL, udp_open, udp_close, NULL, 364 &udp_info, NULL, udp_rrw, udp_rinfop, STRUIOT_STANDARD 365 }; 366 367 static struct qinit udp_winit = { 368 (pfi_t)udp_wput, NULL, NULL, NULL, NULL, 369 &udp_info, NULL, NULL, NULL, STRUIOT_NONE 370 }; 371 372 static struct qinit winit = { 373 (pfi_t)putnext, NULL, NULL, NULL, NULL, 374 &udp_info, NULL, NULL, NULL, STRUIOT_NONE 375 }; 376 377 /* Support for just SNMP if UDP is not pushed directly over device IP */ 378 struct qinit udp_snmp_rinit = { 379 (pfi_t)putnext, NULL, udp_open, ip_snmpmod_close, NULL, 380 &udp_info, NULL, NULL, NULL, STRUIOT_NONE 381 }; 382 383 struct qinit udp_snmp_winit = { 384 (pfi_t)ip_snmpmod_wput, NULL, udp_open, ip_snmpmod_close, NULL, 385 &udp_info, NULL, NULL, NULL, STRUIOT_NONE 386 }; 387 388 struct streamtab udpinfo = { 389 &udp_rinit, &winit 390 }; 391 392 static sin_t sin_null; /* Zero address for quick clears */ 393 static sin6_t sin6_null; /* Zero address for quick clears */ 394 395 /* Hint not protected by any lock */ 396 static in_port_t udp_g_next_port_to_try; 397 398 /* 399 * Extra privileged ports. In host byte order. 400 */ 401 #define UDP_NUM_EPRIV_PORTS 64 402 static int udp_g_num_epriv_ports = UDP_NUM_EPRIV_PORTS; 403 static in_port_t udp_g_epriv_ports[UDP_NUM_EPRIV_PORTS] = { 2049, 4045 }; 404 405 /* Only modified during _init and _fini thus no locking is needed. */ 406 static IDP udp_g_nd; /* Points to table of UDP ND variables. */ 407 408 /* MIB-2 stuff for SNMP */ 409 static mib2_udp_t udp_mib; /* SNMP fixed size info */ 410 static kstat_t *udp_mibkp; /* kstat exporting udp_mib data */ 411 412 #define UDP_MAXPACKET_IPV4 (IP_MAXPACKET - UDPH_SIZE - IP_SIMPLE_HDR_LENGTH) 413 414 /* Default structure copied into T_INFO_ACK messages */ 415 static struct T_info_ack udp_g_t_info_ack_ipv4 = { 416 T_INFO_ACK, 417 UDP_MAXPACKET_IPV4, /* TSDU_size. Excl. headers */ 418 T_INVALID, /* ETSU_size. udp does not support expedited data. */ 419 T_INVALID, /* CDATA_size. udp does not support connect data. */ 420 T_INVALID, /* DDATA_size. udp does not support disconnect data. */ 421 sizeof (sin_t), /* ADDR_size. */ 422 0, /* OPT_size - not initialized here */ 423 UDP_MAXPACKET_IPV4, /* TIDU_size. Excl. headers */ 424 T_CLTS, /* SERV_type. udp supports connection-less. */ 425 TS_UNBND, /* CURRENT_state. This is set from udp_state. */ 426 (XPG4_1|SENDZERO) /* PROVIDER_flag */ 427 }; 428 429 #define UDP_MAXPACKET_IPV6 (IP_MAXPACKET - UDPH_SIZE - IPV6_HDR_LEN) 430 431 static struct T_info_ack udp_g_t_info_ack_ipv6 = { 432 T_INFO_ACK, 433 UDP_MAXPACKET_IPV6, /* TSDU_size. Excl. headers */ 434 T_INVALID, /* ETSU_size. udp does not support expedited data. */ 435 T_INVALID, /* CDATA_size. udp does not support connect data. */ 436 T_INVALID, /* DDATA_size. udp does not support disconnect data. */ 437 sizeof (sin6_t), /* ADDR_size. */ 438 0, /* OPT_size - not initialized here */ 439 UDP_MAXPACKET_IPV6, /* TIDU_size. Excl. headers */ 440 T_CLTS, /* SERV_type. udp supports connection-less. */ 441 TS_UNBND, /* CURRENT_state. This is set from udp_state. */ 442 (XPG4_1|SENDZERO) /* PROVIDER_flag */ 443 }; 444 445 /* largest UDP port number */ 446 #define UDP_MAX_PORT 65535 447 448 /* 449 * Table of ND variables supported by udp. These are loaded into udp_g_nd 450 * in udp_open. 451 * All of these are alterable, within the min/max values given, at run time. 452 */ 453 /* BEGIN CSTYLED */ 454 udpparam_t udp_param_arr[] = { 455 /*min max value name */ 456 { 0L, 256, 32, "udp_wroff_extra" }, 457 { 1L, 255, 255, "udp_ipv4_ttl" }, 458 { 0, IPV6_MAX_HOPS, IPV6_DEFAULT_HOPS, "udp_ipv6_hoplimit"}, 459 { 1024, (32 * 1024), 1024, "udp_smallest_nonpriv_port" }, 460 { 0, 1, 1, "udp_do_checksum" }, 461 { 1024, UDP_MAX_PORT, (32 * 1024), "udp_smallest_anon_port" }, 462 { 1024, UDP_MAX_PORT, UDP_MAX_PORT, "udp_largest_anon_port" }, 463 { UDP_XMIT_LOWATER, (1<<30), UDP_XMIT_HIWATER, "udp_xmit_hiwat"}, 464 { 0, (1<<30), UDP_XMIT_LOWATER, "udp_xmit_lowat"}, 465 { UDP_RECV_LOWATER, (1<<30), UDP_RECV_HIWATER, "udp_recv_hiwat"}, 466 { 65536, (1<<30), 2*1024*1024, "udp_max_buf"}, 467 { 100, 60000, 1000, "udp_ndd_get_info_interval"}, 468 }; 469 /* END CSTYLED */ 470 471 /* 472 * The smallest anonymous port in the privileged port range which UDP 473 * looks for free port. Use in the option UDP_ANONPRIVBIND. 474 */ 475 static in_port_t udp_min_anonpriv_port = 512; 476 477 /* If set to 0, pick ephemeral port sequentially; otherwise randomly. */ 478 uint32_t udp_random_anon_port = 1; 479 480 /* 481 * Hook functions to enable cluster networking. 482 * On non-clustered systems these vectors must always be NULL 483 */ 484 485 void (*cl_inet_bind)(uchar_t protocol, sa_family_t addr_family, 486 uint8_t *laddrp, in_port_t lport) = NULL; 487 void (*cl_inet_unbind)(uint8_t protocol, sa_family_t addr_family, 488 uint8_t *laddrp, in_port_t lport) = NULL; 489 490 typedef union T_primitives *t_primp_t; 491 492 #define UDP_ENQUEUE_MP(udp, mp, proc, tag) { \ 493 ASSERT((mp)->b_prev == NULL && (mp)->b_queue == NULL); \ 494 ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock)); \ 495 (mp)->b_queue = (queue_t *)((uintptr_t)tag); \ 496 (mp)->b_prev = (mblk_t *)proc; \ 497 if ((udp)->udp_mphead == NULL) \ 498 (udp)->udp_mphead = (mp); \ 499 else \ 500 (udp)->udp_mptail->b_next = (mp); \ 501 (udp)->udp_mptail = (mp); \ 502 (udp)->udp_mpcount++; \ 503 } 504 505 #define UDP_READERS_INCREF(udp) { \ 506 ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock)); \ 507 (udp)->udp_reader_count++; \ 508 } 509 510 #define UDP_READERS_DECREF(udp) { \ 511 ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock)); \ 512 (udp)->udp_reader_count--; \ 513 if ((udp)->udp_reader_count == 0) \ 514 cv_broadcast(&(udp)->udp_connp->conn_cv); \ 515 } 516 517 #define UDP_SQUEUE_DECREF(udp) { \ 518 ASSERT(MUTEX_HELD(&(udp)->udp_connp->conn_lock)); \ 519 (udp)->udp_squeue_count--; \ 520 if ((udp)->udp_squeue_count == 0) \ 521 cv_broadcast(&(udp)->udp_connp->conn_cv); \ 522 } 523 524 /* 525 * Notes on UDP endpoint synchronization: 526 * 527 * UDP needs exclusive operation on a per endpoint basis, when executing 528 * functions that modify the endpoint state. udp_rput_other() deals with 529 * packets with IP options, and processing these packets end up having 530 * to update the endpoint's option related state. udp_wput_other() deals 531 * with control operations from the top, e.g. connect() that needs to 532 * update the endpoint state. These could be synchronized using locks, 533 * but the current version uses squeues for this purpose. squeues may 534 * give performance improvement for certain cases such as connected UDP 535 * sockets; thus the framework allows for using squeues. 536 * 537 * The perimeter routines are described as follows: 538 * 539 * udp_enter(): 540 * Enter the UDP endpoint perimeter. 541 * 542 * udp_become_writer(): 543 * Become exclusive on the UDP endpoint. Specifies a function 544 * that will be called exclusively either immediately or later 545 * when the perimeter is available exclusively. 546 * 547 * udp_exit(): 548 * Exit the UDP perimeter. 549 * 550 * Entering UDP from the top or from the bottom must be done using 551 * udp_enter(). No lock must be held while attempting to enter the UDP 552 * perimeter. When finished, udp_exit() must be called to get out of 553 * the perimeter. 554 * 555 * UDP operates in either MT_HOT mode or in SQUEUE mode. In MT_HOT mode, 556 * multiple threads may enter a UDP endpoint concurrently. This is used 557 * for sending and/or receiving normal data. Control operations and other 558 * special cases call udp_become_writer() to become exclusive on a per 559 * endpoint basis and this results in transitioning to SQUEUE mode. squeue 560 * by definition serializes access to the conn_t. When there are no more 561 * pending messages on the squeue for the UDP connection, the endpoint 562 * reverts to MT_HOT mode. During the interregnum when not all MT threads 563 * of an endpoint have finished, messages are queued in the UDP endpoint 564 * and the UDP is in UDP_MT_QUEUED mode or UDP_QUEUED_SQUEUE mode. 565 * 566 * These modes have the following analogs: 567 * 568 * UDP_MT_HOT/udp_reader_count==0 none 569 * UDP_MT_HOT/udp_reader_count>0 RW_READ_LOCK 570 * UDP_MT_QUEUED RW_WRITE_WANTED 571 * UDP_SQUEUE or UDP_QUEUED_SQUEUE RW_WRITE_LOCKED 572 * 573 * Stable modes: UDP_MT_HOT, UDP_SQUEUE 574 * Transient modes: UDP_MT_QUEUED, UDP_QUEUED_SQUEUE 575 * 576 * While in stable modes, UDP keeps track of the number of threads 577 * operating on the endpoint. The udp_reader_count variable represents 578 * the number of threads entering the endpoint as readers while it is 579 * in UDP_MT_HOT mode. Transitioning to UDP_SQUEUE happens when there 580 * is only a single reader, i.e. when this counter drops to 1. Likewise, 581 * udp_squeue_count represents the number of threads operating on the 582 * endpoint's squeue while it is in UDP_SQUEUE mode. The mode transition 583 * to UDP_MT_HOT happens after the last thread exits the endpoint, i.e. 584 * when this counter drops to 0. 585 * 586 * The default mode is set to UDP_MT_HOT and UDP alternates between 587 * UDP_MT_HOT and UDP_SQUEUE as shown in the state transition below. 588 * 589 * Mode transition: 590 * ---------------------------------------------------------------- 591 * old mode Event New mode 592 * ---------------------------------------------------------------- 593 * UDP_MT_HOT Call to udp_become_writer() UDP_SQUEUE 594 * and udp_reader_count == 1 595 * 596 * UDP_MT_HOT Call to udp_become_writer() UDP_MT_QUEUED 597 * and udp_reader_count > 1 598 * 599 * UDP_MT_QUEUED udp_reader_count drops to zero UDP_QUEUED_SQUEUE 600 * 601 * UDP_QUEUED_SQUEUE All messages enqueued on the UDP_SQUEUE 602 * internal UDP queue successfully 603 * moved to squeue AND udp_squeue_count != 0 604 * 605 * UDP_QUEUED_SQUEUE All messages enqueued on the UDP_MT_HOT 606 * internal UDP queue successfully 607 * moved to squeue AND udp_squeue_count 608 * drops to zero 609 * 610 * UDP_SQUEUE udp_squeue_count drops to zero UDP_MT_HOT 611 * ---------------------------------------------------------------- 612 */ 613 614 static queue_t * 615 UDP_WR(queue_t *q) 616 { 617 ASSERT(q->q_ptr == NULL && _OTHERQ(q)->q_ptr == NULL); 618 ASSERT(WR(q)->q_next != NULL && WR(q)->q_next->q_ptr != NULL); 619 ASSERT(IPCL_IS_UDP(Q_TO_CONN(WR(q)->q_next))); 620 621 return (_WR(q)->q_next); 622 } 623 624 static queue_t * 625 UDP_RD(queue_t *q) 626 { 627 ASSERT(q->q_ptr != NULL && _OTHERQ(q)->q_ptr != NULL); 628 ASSERT(IPCL_IS_UDP(Q_TO_CONN(q))); 629 ASSERT(RD(q)->q_next != NULL && RD(q)->q_next->q_ptr == NULL); 630 631 return (_RD(q)->q_next); 632 } 633 634 #ifdef DEBUG 635 #define UDP_MODE_ASSERTIONS(udp, caller) udp_mode_assertions(udp, caller) 636 #else 637 #define UDP_MODE_ASSERTIONS(udp, caller) 638 #endif 639 640 /* Invariants */ 641 #ifdef DEBUG 642 643 uint32_t udp_count[4]; 644 645 /* Context of udp_mode_assertions */ 646 #define UDP_ENTER 1 647 #define UDP_BECOME_WRITER 2 648 #define UDP_EXIT 3 649 650 static void 651 udp_mode_assertions(udp_t *udp, int caller) 652 { 653 ASSERT(MUTEX_HELD(&udp->udp_connp->conn_lock)); 654 655 switch (udp->udp_mode) { 656 case UDP_MT_HOT: 657 /* 658 * Messages have not yet been enqueued on the internal queue, 659 * otherwise we would have switched to UDP_MT_QUEUED. Likewise 660 * by definition, there can't be any messages enqueued on the 661 * squeue. The UDP could be quiescent, so udp_reader_count 662 * could be zero at entry. 663 */ 664 ASSERT(udp->udp_mphead == NULL && udp->udp_mpcount == 0 && 665 udp->udp_squeue_count == 0); 666 ASSERT(caller == UDP_ENTER || udp->udp_reader_count != 0); 667 udp_count[0]++; 668 break; 669 670 case UDP_MT_QUEUED: 671 /* 672 * The last MT thread to exit the udp perimeter empties the 673 * internal queue and then switches the UDP to 674 * UDP_QUEUED_SQUEUE mode. Since we are still in UDP_MT_QUEUED 675 * mode, it means there must be at least 1 MT thread still in 676 * the perimeter and at least 1 message on the internal queue. 677 */ 678 ASSERT(udp->udp_reader_count >= 1 && udp->udp_mphead != NULL && 679 udp->udp_mpcount != 0 && udp->udp_squeue_count == 0); 680 udp_count[1]++; 681 break; 682 683 case UDP_QUEUED_SQUEUE: 684 /* 685 * The switch has happened from MT to SQUEUE. So there can't 686 * any MT threads. Messages could still pile up on the internal 687 * queue until the transition is complete and we move to 688 * UDP_SQUEUE mode. We can't assert on nonzero udp_squeue_count 689 * since the squeue could drain any time. 690 */ 691 ASSERT(udp->udp_reader_count == 0); 692 udp_count[2]++; 693 break; 694 695 case UDP_SQUEUE: 696 /* 697 * The transition is complete. Thre can't be any messages on 698 * the internal queue. The udp could be quiescent or the squeue 699 * could drain any time, so we can't assert on nonzero 700 * udp_squeue_count during entry. Nor can we assert that 701 * udp_reader_count is zero, since, a reader thread could have 702 * directly become writer in line by calling udp_become_writer 703 * without going through the queued states. 704 */ 705 ASSERT(udp->udp_mphead == NULL && udp->udp_mpcount == 0); 706 ASSERT(caller == UDP_ENTER || udp->udp_squeue_count != 0); 707 udp_count[3]++; 708 break; 709 } 710 } 711 #endif 712 713 #define _UDP_ENTER(connp, mp, proc, tag) { \ 714 udp_t *_udp = (connp)->conn_udp; \ 715 \ 716 mutex_enter(&(connp)->conn_lock); \ 717 if ((connp)->conn_state_flags & CONN_CLOSING) { \ 718 mutex_exit(&(connp)->conn_lock); \ 719 freemsg(mp); \ 720 } else { \ 721 UDP_MODE_ASSERTIONS(_udp, UDP_ENTER); \ 722 \ 723 switch (_udp->udp_mode) { \ 724 case UDP_MT_HOT: \ 725 /* We can execute as reader right away. */ \ 726 UDP_READERS_INCREF(_udp); \ 727 mutex_exit(&(connp)->conn_lock); \ 728 (*(proc))(connp, mp, (connp)->conn_sqp); \ 729 break; \ 730 \ 731 case UDP_SQUEUE: \ 732 /* \ 733 * We are in squeue mode, send the \ 734 * packet to the squeue \ 735 */ \ 736 _udp->udp_squeue_count++; \ 737 CONN_INC_REF_LOCKED(connp); \ 738 mutex_exit(&(connp)->conn_lock); \ 739 squeue_enter((connp)->conn_sqp, mp, proc, \ 740 connp, tag); \ 741 break; \ 742 \ 743 case UDP_MT_QUEUED: \ 744 case UDP_QUEUED_SQUEUE: \ 745 /* \ 746 * Some messages may have been enqueued \ 747 * ahead of us. Enqueue the new message \ 748 * at the tail of the internal queue to \ 749 * preserve message ordering. \ 750 */ \ 751 UDP_ENQUEUE_MP(_udp, mp, proc, tag); \ 752 mutex_exit(&(connp)->conn_lock); \ 753 break; \ 754 } \ 755 } \ 756 } 757 758 static void 759 udp_enter(conn_t *connp, mblk_t *mp, sqproc_t proc, uint8_t tag) 760 { 761 _UDP_ENTER(connp, mp, proc, tag); 762 } 763 764 static void 765 udp_become_writer(conn_t *connp, mblk_t *mp, sqproc_t proc, uint8_t tag) 766 { 767 udp_t *udp; 768 769 udp = connp->conn_udp; 770 771 mutex_enter(&connp->conn_lock); 772 773 UDP_MODE_ASSERTIONS(udp, UDP_BECOME_WRITER); 774 775 switch (udp->udp_mode) { 776 case UDP_MT_HOT: 777 if (udp->udp_reader_count == 1) { 778 /* 779 * We are the only MT thread. Switch to squeue mode 780 * immediately. 781 */ 782 udp->udp_mode = UDP_SQUEUE; 783 udp->udp_squeue_count = 1; 784 CONN_INC_REF_LOCKED(connp); 785 mutex_exit(&connp->conn_lock); 786 squeue_enter(connp->conn_sqp, mp, proc, connp, tag); 787 return; 788 } 789 /* FALLTHRU */ 790 791 case UDP_MT_QUEUED: 792 /* Enqueue the packet internally in UDP */ 793 udp->udp_mode = UDP_MT_QUEUED; 794 UDP_ENQUEUE_MP(udp, mp, proc, tag); 795 mutex_exit(&connp->conn_lock); 796 return; 797 798 case UDP_SQUEUE: 799 case UDP_QUEUED_SQUEUE: 800 /* 801 * We are already exclusive. i.e. we are already 802 * writer. Simply call the desired function. 803 */ 804 udp->udp_squeue_count++; 805 mutex_exit(&connp->conn_lock); 806 (*proc)(connp, mp, connp->conn_sqp); 807 return; 808 } 809 } 810 811 /* 812 * Transition from MT mode to SQUEUE mode, when the last MT thread 813 * is exiting the UDP perimeter. Move all messages from the internal 814 * udp queue to the squeue. A better way would be to move all the 815 * messages in one shot, this needs more support from the squeue framework 816 */ 817 static void 818 udp_switch_to_squeue(udp_t *udp) 819 { 820 mblk_t *mp; 821 mblk_t *mp_next; 822 sqproc_t proc; 823 uint8_t tag; 824 conn_t *connp = udp->udp_connp; 825 826 ASSERT(MUTEX_HELD(&connp->conn_lock)); 827 ASSERT(udp->udp_mode == UDP_MT_QUEUED); 828 while (udp->udp_mphead != NULL) { 829 mp = udp->udp_mphead; 830 udp->udp_mphead = NULL; 831 udp->udp_mptail = NULL; 832 udp->udp_mpcount = 0; 833 udp->udp_mode = UDP_QUEUED_SQUEUE; 834 mutex_exit(&connp->conn_lock); 835 /* 836 * It is best not to hold any locks across the calls 837 * to squeue functions. Since we drop the lock we 838 * need to go back and check the udp_mphead once again 839 * after the squeue_fill and hence the while loop at 840 * the top of this function 841 */ 842 for (; mp != NULL; mp = mp_next) { 843 mp_next = mp->b_next; 844 proc = (sqproc_t)mp->b_prev; 845 tag = (uint8_t)((uintptr_t)mp->b_queue); 846 mp->b_next = NULL; 847 mp->b_prev = NULL; 848 mp->b_queue = NULL; 849 CONN_INC_REF(connp); 850 udp->udp_squeue_count++; 851 squeue_fill(connp->conn_sqp, mp, proc, connp, 852 tag); 853 } 854 mutex_enter(&connp->conn_lock); 855 } 856 /* 857 * udp_squeue_count of zero implies that the squeue has drained 858 * even before we arrived here (i.e. after the squeue_fill above) 859 */ 860 udp->udp_mode = (udp->udp_squeue_count != 0) ? 861 UDP_SQUEUE : UDP_MT_HOT; 862 } 863 864 #define _UDP_EXIT(connp) { \ 865 udp_t *_udp = (connp)->conn_udp; \ 866 \ 867 mutex_enter(&(connp)->conn_lock); \ 868 UDP_MODE_ASSERTIONS(_udp, UDP_EXIT); \ 869 \ 870 switch (_udp->udp_mode) { \ 871 case UDP_MT_HOT: \ 872 UDP_READERS_DECREF(_udp); \ 873 mutex_exit(&(connp)->conn_lock); \ 874 break; \ 875 \ 876 case UDP_SQUEUE: \ 877 UDP_SQUEUE_DECREF(_udp); \ 878 if (_udp->udp_squeue_count == 0) \ 879 _udp->udp_mode = UDP_MT_HOT; \ 880 mutex_exit(&(connp)->conn_lock); \ 881 break; \ 882 \ 883 case UDP_MT_QUEUED: \ 884 /* \ 885 * If this is the last MT thread, we need to \ 886 * switch to squeue mode \ 887 */ \ 888 UDP_READERS_DECREF(_udp); \ 889 if (_udp->udp_reader_count == 0) \ 890 udp_switch_to_squeue(_udp); \ 891 mutex_exit(&(connp)->conn_lock); \ 892 break; \ 893 \ 894 case UDP_QUEUED_SQUEUE: \ 895 UDP_SQUEUE_DECREF(_udp); \ 896 /* \ 897 * Even if the udp_squeue_count drops to zero, we \ 898 * don't want to change udp_mode to UDP_MT_HOT here. \ 899 * The thread in udp_switch_to_squeue will take care \ 900 * of the transition to UDP_MT_HOT, after emptying \ 901 * any more new messages that have been enqueued in \ 902 * udp_mphead. \ 903 */ \ 904 mutex_exit(&(connp)->conn_lock); \ 905 break; \ 906 } \ 907 } 908 909 static void 910 udp_exit(conn_t *connp) 911 { 912 _UDP_EXIT(connp); 913 } 914 915 /* 916 * Return the next anonymous port in the privileged port range for 917 * bind checking. 918 * 919 * Trusted Extension (TX) notes: TX allows administrator to mark or 920 * reserve ports as Multilevel ports (MLP). MLP has special function 921 * on TX systems. Once a port is made MLP, it's not available as 922 * ordinary port. This creates "holes" in the port name space. It 923 * may be necessary to skip the "holes" find a suitable anon port. 924 */ 925 static in_port_t 926 udp_get_next_priv_port(udp_t *udp) 927 { 928 static in_port_t next_priv_port = IPPORT_RESERVED - 1; 929 in_port_t nextport; 930 boolean_t restart = B_FALSE; 931 932 retry: 933 if (next_priv_port < udp_min_anonpriv_port || 934 next_priv_port >= IPPORT_RESERVED) { 935 next_priv_port = IPPORT_RESERVED - 1; 936 if (restart) 937 return (0); 938 restart = B_TRUE; 939 } 940 941 if (is_system_labeled() && 942 (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred), 943 next_priv_port, IPPROTO_UDP, B_FALSE)) != 0) { 944 next_priv_port = nextport; 945 goto retry; 946 } 947 948 return (next_priv_port--); 949 } 950 951 /* UDP bind hash report triggered via the Named Dispatch mechanism. */ 952 /* ARGSUSED */ 953 static int 954 udp_bind_hash_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr) 955 { 956 udp_fanout_t *udpf; 957 int i; 958 zoneid_t zoneid; 959 conn_t *connp; 960 udp_t *udp; 961 962 connp = Q_TO_CONN(q); 963 udp = connp->conn_udp; 964 965 /* Refer to comments in udp_status_report(). */ 966 if (cr == NULL || secpolicy_net_config(cr, B_TRUE) != 0) { 967 if (ddi_get_lbolt() - udp_last_ndd_get_info_time < 968 drv_usectohz(udp_ndd_get_info_interval * 1000)) { 969 (void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG); 970 return (0); 971 } 972 } 973 if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) { 974 /* The following may work even if we cannot get a large buf. */ 975 (void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG); 976 return (0); 977 } 978 979 (void) mi_mpprintf(mp, 980 "UDP " MI_COL_HDRPAD_STR 981 /* 12345678[89ABCDEF] */ 982 " zone lport src addr dest addr port state"); 983 /* 1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */ 984 985 zoneid = connp->conn_zoneid; 986 987 for (i = 0; i < udp_bind_fanout_size; i++) { 988 udpf = &udp_bind_fanout[i]; 989 mutex_enter(&udpf->uf_lock); 990 991 /* Print the hash index. */ 992 udp = udpf->uf_udp; 993 if (zoneid != GLOBAL_ZONEID) { 994 /* skip to first entry in this zone; might be none */ 995 while (udp != NULL && 996 udp->udp_connp->conn_zoneid != zoneid) 997 udp = udp->udp_bind_hash; 998 } 999 if (udp != NULL) { 1000 uint_t print_len, buf_len; 1001 1002 buf_len = mp->b_cont->b_datap->db_lim - 1003 mp->b_cont->b_wptr; 1004 print_len = snprintf((char *)mp->b_cont->b_wptr, 1005 buf_len, "%d\n", i); 1006 if (print_len < buf_len) { 1007 mp->b_cont->b_wptr += print_len; 1008 } else { 1009 mp->b_cont->b_wptr += buf_len; 1010 } 1011 for (; udp != NULL; udp = udp->udp_bind_hash) { 1012 if (zoneid == GLOBAL_ZONEID || 1013 zoneid == udp->udp_connp->conn_zoneid) 1014 udp_report_item(mp->b_cont, udp); 1015 } 1016 } 1017 mutex_exit(&udpf->uf_lock); 1018 } 1019 udp_last_ndd_get_info_time = ddi_get_lbolt(); 1020 return (0); 1021 } 1022 1023 /* 1024 * Hash list removal routine for udp_t structures. 1025 */ 1026 static void 1027 udp_bind_hash_remove(udp_t *udp, boolean_t caller_holds_lock) 1028 { 1029 udp_t *udpnext; 1030 kmutex_t *lockp; 1031 1032 if (udp->udp_ptpbhn == NULL) 1033 return; 1034 1035 /* 1036 * Extract the lock pointer in case there are concurrent 1037 * hash_remove's for this instance. 1038 */ 1039 ASSERT(udp->udp_port != 0); 1040 if (!caller_holds_lock) { 1041 lockp = &udp_bind_fanout[UDP_BIND_HASH(udp->udp_port)].uf_lock; 1042 ASSERT(lockp != NULL); 1043 mutex_enter(lockp); 1044 } 1045 if (udp->udp_ptpbhn != NULL) { 1046 udpnext = udp->udp_bind_hash; 1047 if (udpnext != NULL) { 1048 udpnext->udp_ptpbhn = udp->udp_ptpbhn; 1049 udp->udp_bind_hash = NULL; 1050 } 1051 *udp->udp_ptpbhn = udpnext; 1052 udp->udp_ptpbhn = NULL; 1053 } 1054 if (!caller_holds_lock) { 1055 mutex_exit(lockp); 1056 } 1057 } 1058 1059 static void 1060 udp_bind_hash_insert(udp_fanout_t *uf, udp_t *udp) 1061 { 1062 udp_t **udpp; 1063 udp_t *udpnext; 1064 1065 ASSERT(MUTEX_HELD(&uf->uf_lock)); 1066 if (udp->udp_ptpbhn != NULL) { 1067 udp_bind_hash_remove(udp, B_TRUE); 1068 } 1069 udpp = &uf->uf_udp; 1070 udpnext = udpp[0]; 1071 if (udpnext != NULL) { 1072 /* 1073 * If the new udp bound to the INADDR_ANY address 1074 * and the first one in the list is not bound to 1075 * INADDR_ANY we skip all entries until we find the 1076 * first one bound to INADDR_ANY. 1077 * This makes sure that applications binding to a 1078 * specific address get preference over those binding to 1079 * INADDR_ANY. 1080 */ 1081 if (V6_OR_V4_INADDR_ANY(udp->udp_bound_v6src) && 1082 !V6_OR_V4_INADDR_ANY(udpnext->udp_bound_v6src)) { 1083 while ((udpnext = udpp[0]) != NULL && 1084 !V6_OR_V4_INADDR_ANY( 1085 udpnext->udp_bound_v6src)) { 1086 udpp = &(udpnext->udp_bind_hash); 1087 } 1088 if (udpnext != NULL) 1089 udpnext->udp_ptpbhn = &udp->udp_bind_hash; 1090 } else { 1091 udpnext->udp_ptpbhn = &udp->udp_bind_hash; 1092 } 1093 } 1094 udp->udp_bind_hash = udpnext; 1095 udp->udp_ptpbhn = udpp; 1096 udpp[0] = udp; 1097 } 1098 1099 /* 1100 * This routine is called to handle each O_T_BIND_REQ/T_BIND_REQ message 1101 * passed to udp_wput. 1102 * It associates a port number and local address with the stream. 1103 * The O_T_BIND_REQ/T_BIND_REQ is passed downstream to ip with the UDP 1104 * protocol type (IPPROTO_UDP) placed in the message following the address. 1105 * A T_BIND_ACK message is passed upstream when ip acknowledges the request. 1106 * (Called as writer.) 1107 * 1108 * Note that UDP over IPv4 and IPv6 sockets can use the same port number 1109 * without setting SO_REUSEADDR. This is needed so that they 1110 * can be viewed as two independent transport protocols. 1111 * However, anonymouns ports are allocated from the same range to avoid 1112 * duplicating the udp_g_next_port_to_try. 1113 */ 1114 static void 1115 udp_bind(queue_t *q, mblk_t *mp) 1116 { 1117 sin_t *sin; 1118 sin6_t *sin6; 1119 mblk_t *mp1; 1120 in_port_t port; /* Host byte order */ 1121 in_port_t requested_port; /* Host byte order */ 1122 struct T_bind_req *tbr; 1123 int count; 1124 in6_addr_t v6src; 1125 boolean_t bind_to_req_port_only; 1126 int loopmax; 1127 udp_fanout_t *udpf; 1128 in_port_t lport; /* Network byte order */ 1129 zoneid_t zoneid; 1130 conn_t *connp; 1131 udp_t *udp; 1132 boolean_t is_inaddr_any; 1133 mlp_type_t addrtype, mlptype; 1134 1135 connp = Q_TO_CONN(q); 1136 udp = connp->conn_udp; 1137 if ((mp->b_wptr - mp->b_rptr) < sizeof (*tbr)) { 1138 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 1139 "udp_bind: bad req, len %u", 1140 (uint_t)(mp->b_wptr - mp->b_rptr)); 1141 udp_err_ack(q, mp, TPROTO, 0); 1142 return; 1143 } 1144 1145 if (udp->udp_state != TS_UNBND) { 1146 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 1147 "udp_bind: bad state, %u", udp->udp_state); 1148 udp_err_ack(q, mp, TOUTSTATE, 0); 1149 return; 1150 } 1151 /* 1152 * Reallocate the message to make sure we have enough room for an 1153 * address and the protocol type. 1154 */ 1155 mp1 = reallocb(mp, sizeof (struct T_bind_ack) + sizeof (sin6_t) + 1, 1); 1156 if (!mp1) { 1157 udp_err_ack(q, mp, TSYSERR, ENOMEM); 1158 return; 1159 } 1160 1161 mp = mp1; 1162 tbr = (struct T_bind_req *)mp->b_rptr; 1163 switch (tbr->ADDR_length) { 1164 case 0: /* Request for a generic port */ 1165 tbr->ADDR_offset = sizeof (struct T_bind_req); 1166 if (udp->udp_family == AF_INET) { 1167 tbr->ADDR_length = sizeof (sin_t); 1168 sin = (sin_t *)&tbr[1]; 1169 *sin = sin_null; 1170 sin->sin_family = AF_INET; 1171 mp->b_wptr = (uchar_t *)&sin[1]; 1172 } else { 1173 ASSERT(udp->udp_family == AF_INET6); 1174 tbr->ADDR_length = sizeof (sin6_t); 1175 sin6 = (sin6_t *)&tbr[1]; 1176 *sin6 = sin6_null; 1177 sin6->sin6_family = AF_INET6; 1178 mp->b_wptr = (uchar_t *)&sin6[1]; 1179 } 1180 port = 0; 1181 break; 1182 1183 case sizeof (sin_t): /* Complete IPv4 address */ 1184 sin = (sin_t *)mi_offset_param(mp, tbr->ADDR_offset, 1185 sizeof (sin_t)); 1186 if (sin == NULL || !OK_32PTR((char *)sin)) { 1187 udp_err_ack(q, mp, TSYSERR, EINVAL); 1188 return; 1189 } 1190 if (udp->udp_family != AF_INET || 1191 sin->sin_family != AF_INET) { 1192 udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 1193 return; 1194 } 1195 port = ntohs(sin->sin_port); 1196 break; 1197 1198 case sizeof (sin6_t): /* complete IPv6 address */ 1199 sin6 = (sin6_t *)mi_offset_param(mp, tbr->ADDR_offset, 1200 sizeof (sin6_t)); 1201 if (sin6 == NULL || !OK_32PTR((char *)sin6)) { 1202 udp_err_ack(q, mp, TSYSERR, EINVAL); 1203 return; 1204 } 1205 if (udp->udp_family != AF_INET6 || 1206 sin6->sin6_family != AF_INET6) { 1207 udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 1208 return; 1209 } 1210 port = ntohs(sin6->sin6_port); 1211 break; 1212 1213 default: /* Invalid request */ 1214 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 1215 "udp_bind: bad ADDR_length length %u", tbr->ADDR_length); 1216 udp_err_ack(q, mp, TBADADDR, 0); 1217 return; 1218 } 1219 1220 requested_port = port; 1221 1222 if (requested_port == 0 || tbr->PRIM_type == O_T_BIND_REQ) 1223 bind_to_req_port_only = B_FALSE; 1224 else /* T_BIND_REQ and requested_port != 0 */ 1225 bind_to_req_port_only = B_TRUE; 1226 1227 if (requested_port == 0) { 1228 /* 1229 * If the application passed in zero for the port number, it 1230 * doesn't care which port number we bind to. Get one in the 1231 * valid range. 1232 */ 1233 if (udp->udp_anon_priv_bind) { 1234 port = udp_get_next_priv_port(udp); 1235 } else { 1236 port = udp_update_next_port(udp, 1237 udp_g_next_port_to_try, B_TRUE); 1238 } 1239 } else { 1240 /* 1241 * If the port is in the well-known privileged range, 1242 * make sure the caller was privileged. 1243 */ 1244 int i; 1245 boolean_t priv = B_FALSE; 1246 1247 if (port < udp_smallest_nonpriv_port) { 1248 priv = B_TRUE; 1249 } else { 1250 for (i = 0; i < udp_g_num_epriv_ports; i++) { 1251 if (port == udp_g_epriv_ports[i]) { 1252 priv = B_TRUE; 1253 break; 1254 } 1255 } 1256 } 1257 1258 if (priv) { 1259 cred_t *cr = DB_CREDDEF(mp, connp->conn_cred); 1260 1261 if (secpolicy_net_privaddr(cr, port) != 0) { 1262 udp_err_ack(q, mp, TACCES, 0); 1263 return; 1264 } 1265 } 1266 } 1267 1268 if (port == 0) { 1269 udp_err_ack(q, mp, TNOADDR, 0); 1270 return; 1271 } 1272 1273 /* 1274 * Copy the source address into our udp structure. This address 1275 * may still be zero; if so, IP will fill in the correct address 1276 * each time an outbound packet is passed to it. 1277 */ 1278 if (udp->udp_family == AF_INET) { 1279 ASSERT(sin != NULL); 1280 ASSERT(udp->udp_ipversion == IPV4_VERSION); 1281 udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + 1282 udp->udp_ip_snd_options_len; 1283 IN6_IPADDR_TO_V4MAPPED(sin->sin_addr.s_addr, &v6src); 1284 } else { 1285 ASSERT(sin6 != NULL); 1286 v6src = sin6->sin6_addr; 1287 if (IN6_IS_ADDR_V4MAPPED(&v6src)) { 1288 udp->udp_ipversion = IPV4_VERSION; 1289 udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + 1290 UDPH_SIZE + udp->udp_ip_snd_options_len; 1291 } else { 1292 udp->udp_ipversion = IPV6_VERSION; 1293 udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len; 1294 } 1295 } 1296 1297 /* 1298 * If udp_reuseaddr is not set, then we have to make sure that 1299 * the IP address and port number the application requested 1300 * (or we selected for the application) is not being used by 1301 * another stream. If another stream is already using the 1302 * requested IP address and port, the behavior depends on 1303 * "bind_to_req_port_only". If set the bind fails; otherwise we 1304 * search for any an unused port to bind to the the stream. 1305 * 1306 * As per the BSD semantics, as modified by the Deering multicast 1307 * changes, if udp_reuseaddr is set, then we allow multiple binds 1308 * to the same port independent of the local IP address. 1309 * 1310 * This is slightly different than in SunOS 4.X which did not 1311 * support IP multicast. Note that the change implemented by the 1312 * Deering multicast code effects all binds - not only binding 1313 * to IP multicast addresses. 1314 * 1315 * Note that when binding to port zero we ignore SO_REUSEADDR in 1316 * order to guarantee a unique port. 1317 */ 1318 1319 count = 0; 1320 if (udp->udp_anon_priv_bind) { 1321 /* loopmax = (IPPORT_RESERVED-1) - udp_min_anonpriv_port + 1 */ 1322 loopmax = IPPORT_RESERVED - udp_min_anonpriv_port; 1323 } else { 1324 loopmax = udp_largest_anon_port - udp_smallest_anon_port + 1; 1325 } 1326 1327 is_inaddr_any = V6_OR_V4_INADDR_ANY(v6src); 1328 zoneid = connp->conn_zoneid; 1329 1330 for (;;) { 1331 udp_t *udp1; 1332 boolean_t found_exclbind = B_FALSE; 1333 1334 /* 1335 * Walk through the list of udp streams bound to 1336 * requested port with the same IP address. 1337 */ 1338 lport = htons(port); 1339 udpf = &udp_bind_fanout[UDP_BIND_HASH(lport)]; 1340 mutex_enter(&udpf->uf_lock); 1341 for (udp1 = udpf->uf_udp; udp1 != NULL; 1342 udp1 = udp1->udp_bind_hash) { 1343 if (lport != udp1->udp_port) 1344 continue; 1345 1346 /* 1347 * On a labeled system, we must treat bindings to ports 1348 * on shared IP addresses by sockets with MAC exemption 1349 * privilege as being in all zones, as there's 1350 * otherwise no way to identify the right receiver. 1351 */ 1352 if (zoneid != udp1->udp_connp->conn_zoneid && 1353 !udp->udp_mac_exempt && !udp1->udp_mac_exempt) 1354 continue; 1355 1356 /* 1357 * If UDP_EXCLBIND is set for either the bound or 1358 * binding endpoint, the semantics of bind 1359 * is changed according to the following chart. 1360 * 1361 * spec = specified address (v4 or v6) 1362 * unspec = unspecified address (v4 or v6) 1363 * A = specified addresses are different for endpoints 1364 * 1365 * bound bind to allowed? 1366 * ------------------------------------- 1367 * unspec unspec no 1368 * unspec spec no 1369 * spec unspec no 1370 * spec spec yes if A 1371 * 1372 * For labeled systems, SO_MAC_EXEMPT behaves the same 1373 * as UDP_EXCLBIND, except that zoneid is ignored. 1374 */ 1375 if (udp1->udp_exclbind || udp->udp_exclbind || 1376 udp1->udp_mac_exempt || udp->udp_mac_exempt) { 1377 if (V6_OR_V4_INADDR_ANY( 1378 udp1->udp_bound_v6src) || 1379 is_inaddr_any || 1380 IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src, 1381 &v6src)) { 1382 found_exclbind = B_TRUE; 1383 break; 1384 } 1385 continue; 1386 } 1387 1388 /* 1389 * Check ipversion to allow IPv4 and IPv6 sockets to 1390 * have disjoint port number spaces. 1391 */ 1392 if (udp->udp_ipversion != udp1->udp_ipversion) { 1393 1394 /* 1395 * On the first time through the loop, if the 1396 * the user intentionally specified a 1397 * particular port number, then ignore any 1398 * bindings of the other protocol that may 1399 * conflict. This allows the user to bind IPv6 1400 * alone and get both v4 and v6, or bind both 1401 * both and get each seperately. On subsequent 1402 * times through the loop, we're checking a 1403 * port that we chose (not the user) and thus 1404 * we do not allow casual duplicate bindings. 1405 */ 1406 if (count == 0 && requested_port != 0) 1407 continue; 1408 } 1409 1410 /* 1411 * No difference depending on SO_REUSEADDR. 1412 * 1413 * If existing port is bound to a 1414 * non-wildcard IP address and 1415 * the requesting stream is bound to 1416 * a distinct different IP addresses 1417 * (non-wildcard, also), keep going. 1418 */ 1419 if (!is_inaddr_any && 1420 !V6_OR_V4_INADDR_ANY(udp1->udp_bound_v6src) && 1421 !IN6_ARE_ADDR_EQUAL(&udp1->udp_bound_v6src, 1422 &v6src)) { 1423 continue; 1424 } 1425 break; 1426 } 1427 1428 if (!found_exclbind && 1429 (udp->udp_reuseaddr && requested_port != 0)) { 1430 break; 1431 } 1432 1433 if (udp1 == NULL) { 1434 /* 1435 * No other stream has this IP address 1436 * and port number. We can use it. 1437 */ 1438 break; 1439 } 1440 mutex_exit(&udpf->uf_lock); 1441 if (bind_to_req_port_only) { 1442 /* 1443 * We get here only when requested port 1444 * is bound (and only first of the for() 1445 * loop iteration). 1446 * 1447 * The semantics of this bind request 1448 * require it to fail so we return from 1449 * the routine (and exit the loop). 1450 * 1451 */ 1452 udp_err_ack(q, mp, TADDRBUSY, 0); 1453 return; 1454 } 1455 1456 if (udp->udp_anon_priv_bind) { 1457 port = udp_get_next_priv_port(udp); 1458 } else { 1459 if ((count == 0) && (requested_port != 0)) { 1460 /* 1461 * If the application wants us to find 1462 * a port, get one to start with. Set 1463 * requested_port to 0, so that we will 1464 * update udp_g_next_port_to_try below. 1465 */ 1466 port = udp_update_next_port(udp, 1467 udp_g_next_port_to_try, B_TRUE); 1468 requested_port = 0; 1469 } else { 1470 port = udp_update_next_port(udp, port + 1, 1471 B_FALSE); 1472 } 1473 } 1474 1475 if (port == 0 || ++count >= loopmax) { 1476 /* 1477 * We've tried every possible port number and 1478 * there are none available, so send an error 1479 * to the user. 1480 */ 1481 udp_err_ack(q, mp, TNOADDR, 0); 1482 return; 1483 } 1484 } 1485 1486 /* 1487 * Copy the source address into our udp structure. This address 1488 * may still be zero; if so, ip will fill in the correct address 1489 * each time an outbound packet is passed to it. 1490 * If we are binding to a broadcast or multicast address udp_rput 1491 * will clear the source address when it receives the T_BIND_ACK. 1492 */ 1493 udp->udp_v6src = udp->udp_bound_v6src = v6src; 1494 udp->udp_port = lport; 1495 /* 1496 * Now reset the the next anonymous port if the application requested 1497 * an anonymous port, or we handed out the next anonymous port. 1498 */ 1499 if ((requested_port == 0) && (!udp->udp_anon_priv_bind)) { 1500 udp_g_next_port_to_try = port + 1; 1501 } 1502 1503 /* Initialize the O_T_BIND_REQ/T_BIND_REQ for ip. */ 1504 if (udp->udp_family == AF_INET) { 1505 sin->sin_port = udp->udp_port; 1506 } else { 1507 int error; 1508 1509 sin6->sin6_port = udp->udp_port; 1510 /* Rebuild the header template */ 1511 error = udp_build_hdrs(q, udp); 1512 if (error != 0) { 1513 mutex_exit(&udpf->uf_lock); 1514 udp_err_ack(q, mp, TSYSERR, error); 1515 return; 1516 } 1517 } 1518 udp->udp_state = TS_IDLE; 1519 udp_bind_hash_insert(udpf, udp); 1520 mutex_exit(&udpf->uf_lock); 1521 1522 if (cl_inet_bind) { 1523 /* 1524 * Running in cluster mode - register bind information 1525 */ 1526 if (udp->udp_ipversion == IPV4_VERSION) { 1527 (*cl_inet_bind)(IPPROTO_UDP, AF_INET, 1528 (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)), 1529 (in_port_t)udp->udp_port); 1530 } else { 1531 (*cl_inet_bind)(IPPROTO_UDP, AF_INET6, 1532 (uint8_t *)&(udp->udp_v6src), 1533 (in_port_t)udp->udp_port); 1534 } 1535 1536 } 1537 1538 connp->conn_anon_port = (is_system_labeled() && requested_port == 0); 1539 if (is_system_labeled() && (!connp->conn_anon_port || 1540 connp->conn_anon_mlp)) { 1541 uint16_t mlpport; 1542 cred_t *cr = connp->conn_cred; 1543 zone_t *zone; 1544 1545 connp->conn_mlp_type = udp->udp_recvucred ? mlptBoth : 1546 mlptSingle; 1547 addrtype = tsol_mlp_addr_type(zoneid, IPV6_VERSION, &v6src); 1548 if (addrtype == mlptSingle) { 1549 udp_err_ack(q, mp, TNOADDR, 0); 1550 connp->conn_anon_port = B_FALSE; 1551 connp->conn_mlp_type = mlptSingle; 1552 return; 1553 } 1554 mlpport = connp->conn_anon_port ? PMAPPORT : port; 1555 zone = crgetzone(cr); 1556 mlptype = tsol_mlp_port_type(zone, IPPROTO_UDP, mlpport, 1557 addrtype); 1558 if (mlptype != mlptSingle && 1559 (connp->conn_mlp_type == mlptSingle || 1560 secpolicy_net_bindmlp(cr) != 0)) { 1561 if (udp->udp_debug) { 1562 (void) strlog(UDP_MOD_ID, 0, 1, 1563 SL_ERROR|SL_TRACE, 1564 "udp_bind: no priv for multilevel port %d", 1565 mlpport); 1566 } 1567 udp_err_ack(q, mp, TACCES, 0); 1568 connp->conn_anon_port = B_FALSE; 1569 connp->conn_mlp_type = mlptSingle; 1570 return; 1571 } 1572 1573 /* 1574 * If we're specifically binding a shared IP address and the 1575 * port is MLP on shared addresses, then check to see if this 1576 * zone actually owns the MLP. Reject if not. 1577 */ 1578 if (mlptype == mlptShared && addrtype == mlptShared) { 1579 zoneid_t mlpzone; 1580 1581 mlpzone = tsol_mlp_findzone(IPPROTO_UDP, 1582 htons(mlpport)); 1583 if (connp->conn_zoneid != mlpzone) { 1584 if (udp->udp_debug) { 1585 (void) strlog(UDP_MOD_ID, 0, 1, 1586 SL_ERROR|SL_TRACE, 1587 "udp_bind: attempt to bind port " 1588 "%d on shared addr in zone %d " 1589 "(should be %d)", 1590 mlpport, connp->conn_zoneid, 1591 mlpzone); 1592 } 1593 udp_err_ack(q, mp, TACCES, 0); 1594 connp->conn_anon_port = B_FALSE; 1595 connp->conn_mlp_type = mlptSingle; 1596 return; 1597 } 1598 } 1599 if (connp->conn_anon_port) { 1600 int error; 1601 1602 error = tsol_mlp_anon(zone, mlptype, connp->conn_ulp, 1603 port, B_TRUE); 1604 if (error != 0) { 1605 if (udp->udp_debug) { 1606 (void) strlog(UDP_MOD_ID, 0, 1, 1607 SL_ERROR|SL_TRACE, 1608 "udp_bind: cannot establish anon " 1609 "MLP for port %d", port); 1610 } 1611 udp_err_ack(q, mp, TACCES, 0); 1612 connp->conn_anon_port = B_FALSE; 1613 connp->conn_mlp_type = mlptSingle; 1614 return; 1615 } 1616 } 1617 connp->conn_mlp_type = mlptype; 1618 } 1619 1620 /* Pass the protocol number in the message following the address. */ 1621 *mp->b_wptr++ = IPPROTO_UDP; 1622 if (!V6_OR_V4_INADDR_ANY(udp->udp_v6src)) { 1623 /* 1624 * Append a request for an IRE if udp_v6src not 1625 * zero (IPv4 - INADDR_ANY, or IPv6 - all-zeroes address). 1626 */ 1627 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 1628 if (!mp->b_cont) { 1629 udp_err_ack(q, mp, TSYSERR, ENOMEM); 1630 return; 1631 } 1632 mp->b_cont->b_wptr += sizeof (ire_t); 1633 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 1634 } 1635 if (udp->udp_family == AF_INET6) 1636 mp = ip_bind_v6(q, mp, connp, NULL); 1637 else 1638 mp = ip_bind_v4(q, mp, connp); 1639 1640 if (mp != NULL) 1641 udp_rput_other(_RD(q), mp); 1642 else 1643 CONN_INC_REF(connp); 1644 } 1645 1646 1647 void 1648 udp_resume_bind(conn_t *connp, mblk_t *mp) 1649 { 1650 udp_enter(connp, mp, udp_resume_bind_cb, SQTAG_BIND_RETRY); 1651 } 1652 1653 /* 1654 * This is called from ip_wput_nondata to resume a deferred UDP bind. 1655 */ 1656 /* ARGSUSED */ 1657 static void 1658 udp_resume_bind_cb(void *arg, mblk_t *mp, void *arg2) 1659 { 1660 conn_t *connp = arg; 1661 1662 ASSERT(connp != NULL && IPCL_IS_UDP(connp)); 1663 1664 udp_rput_other(connp->conn_rq, mp); 1665 1666 CONN_OPER_PENDING_DONE(connp); 1667 udp_exit(connp); 1668 } 1669 1670 /* 1671 * This routine handles each T_CONN_REQ message passed to udp. It 1672 * associates a default destination address with the stream. 1673 * 1674 * This routine sends down a T_BIND_REQ to IP with the following mblks: 1675 * T_BIND_REQ - specifying local and remote address/port 1676 * IRE_DB_REQ_TYPE - to get an IRE back containing ire_type and src 1677 * T_OK_ACK - for the T_CONN_REQ 1678 * T_CONN_CON - to keep the TPI user happy 1679 * 1680 * The connect completes in udp_rput. 1681 * When a T_BIND_ACK is received information is extracted from the IRE 1682 * and the two appended messages are sent to the TPI user. 1683 * Should udp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert 1684 * it to an error ack for the appropriate primitive. 1685 */ 1686 static void 1687 udp_connect(queue_t *q, mblk_t *mp) 1688 { 1689 sin6_t *sin6; 1690 sin_t *sin; 1691 struct T_conn_req *tcr; 1692 in6_addr_t v6dst; 1693 ipaddr_t v4dst; 1694 uint16_t dstport; 1695 uint32_t flowinfo; 1696 mblk_t *mp1, *mp2; 1697 udp_fanout_t *udpf; 1698 udp_t *udp, *udp1; 1699 1700 udp = Q_TO_UDP(q); 1701 1702 tcr = (struct T_conn_req *)mp->b_rptr; 1703 1704 /* A bit of sanity checking */ 1705 if ((mp->b_wptr - mp->b_rptr) < sizeof (struct T_conn_req)) { 1706 udp_err_ack(q, mp, TPROTO, 0); 1707 return; 1708 } 1709 /* 1710 * This UDP must have bound to a port already before doing 1711 * a connect. 1712 */ 1713 if (udp->udp_state == TS_UNBND) { 1714 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 1715 "udp_connect: bad state, %u", udp->udp_state); 1716 udp_err_ack(q, mp, TOUTSTATE, 0); 1717 return; 1718 } 1719 ASSERT(udp->udp_port != 0 && udp->udp_ptpbhn != NULL); 1720 1721 udpf = &udp_bind_fanout[UDP_BIND_HASH(udp->udp_port)]; 1722 1723 if (udp->udp_state == TS_DATA_XFER) { 1724 /* Already connected - clear out state */ 1725 mutex_enter(&udpf->uf_lock); 1726 udp->udp_v6src = udp->udp_bound_v6src; 1727 udp->udp_state = TS_IDLE; 1728 mutex_exit(&udpf->uf_lock); 1729 } 1730 1731 if (tcr->OPT_length != 0) { 1732 udp_err_ack(q, mp, TBADOPT, 0); 1733 return; 1734 } 1735 1736 /* 1737 * Determine packet type based on type of address passed in 1738 * the request should contain an IPv4 or IPv6 address. 1739 * Make sure that address family matches the type of 1740 * family of the the address passed down 1741 */ 1742 switch (tcr->DEST_length) { 1743 default: 1744 udp_err_ack(q, mp, TBADADDR, 0); 1745 return; 1746 1747 case sizeof (sin_t): 1748 sin = (sin_t *)mi_offset_param(mp, tcr->DEST_offset, 1749 sizeof (sin_t)); 1750 if (sin == NULL || !OK_32PTR((char *)sin)) { 1751 udp_err_ack(q, mp, TSYSERR, EINVAL); 1752 return; 1753 } 1754 if (udp->udp_family != AF_INET || 1755 sin->sin_family != AF_INET) { 1756 udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 1757 return; 1758 } 1759 v4dst = sin->sin_addr.s_addr; 1760 dstport = sin->sin_port; 1761 IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst); 1762 ASSERT(udp->udp_ipversion == IPV4_VERSION); 1763 udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + 1764 udp->udp_ip_snd_options_len; 1765 break; 1766 1767 case sizeof (sin6_t): 1768 sin6 = (sin6_t *)mi_offset_param(mp, tcr->DEST_offset, 1769 sizeof (sin6_t)); 1770 if (sin6 == NULL || !OK_32PTR((char *)sin6)) { 1771 udp_err_ack(q, mp, TSYSERR, EINVAL); 1772 return; 1773 } 1774 if (udp->udp_family != AF_INET6 || 1775 sin6->sin6_family != AF_INET6) { 1776 udp_err_ack(q, mp, TSYSERR, EAFNOSUPPORT); 1777 return; 1778 } 1779 v6dst = sin6->sin6_addr; 1780 if (IN6_IS_ADDR_V4MAPPED(&v6dst)) { 1781 IN6_V4MAPPED_TO_IPADDR(&v6dst, v4dst); 1782 udp->udp_ipversion = IPV4_VERSION; 1783 udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + 1784 UDPH_SIZE + udp->udp_ip_snd_options_len; 1785 flowinfo = 0; 1786 } else { 1787 udp->udp_ipversion = IPV6_VERSION; 1788 udp->udp_max_hdr_len = udp->udp_sticky_hdrs_len; 1789 flowinfo = sin6->sin6_flowinfo; 1790 } 1791 dstport = sin6->sin6_port; 1792 break; 1793 } 1794 if (dstport == 0) { 1795 udp_err_ack(q, mp, TBADADDR, 0); 1796 return; 1797 } 1798 1799 /* 1800 * Create a default IP header with no IP options. 1801 */ 1802 udp->udp_dstport = dstport; 1803 if (udp->udp_ipversion == IPV4_VERSION) { 1804 /* 1805 * Interpret a zero destination to mean loopback. 1806 * Update the T_CONN_REQ (sin/sin6) since it is used to 1807 * generate the T_CONN_CON. 1808 */ 1809 if (v4dst == INADDR_ANY) { 1810 v4dst = htonl(INADDR_LOOPBACK); 1811 IN6_IPADDR_TO_V4MAPPED(v4dst, &v6dst); 1812 if (udp->udp_family == AF_INET) { 1813 sin->sin_addr.s_addr = v4dst; 1814 } else { 1815 sin6->sin6_addr = v6dst; 1816 } 1817 } 1818 udp->udp_v6dst = v6dst; 1819 udp->udp_flowinfo = 0; 1820 1821 /* 1822 * If the destination address is multicast and 1823 * an outgoing multicast interface has been set, 1824 * use the address of that interface as our 1825 * source address if no source address has been set. 1826 */ 1827 if (V4_PART_OF_V6(udp->udp_v6src) == INADDR_ANY && 1828 CLASSD(v4dst) && 1829 udp->udp_multicast_if_addr != INADDR_ANY) { 1830 IN6_IPADDR_TO_V4MAPPED(udp->udp_multicast_if_addr, 1831 &udp->udp_v6src); 1832 } 1833 } else { 1834 ASSERT(udp->udp_ipversion == IPV6_VERSION); 1835 /* 1836 * Interpret a zero destination to mean loopback. 1837 * Update the T_CONN_REQ (sin/sin6) since it is used to 1838 * generate the T_CONN_CON. 1839 */ 1840 if (IN6_IS_ADDR_UNSPECIFIED(&v6dst)) { 1841 v6dst = ipv6_loopback; 1842 sin6->sin6_addr = v6dst; 1843 } 1844 udp->udp_v6dst = v6dst; 1845 udp->udp_flowinfo = flowinfo; 1846 /* 1847 * If the destination address is multicast and 1848 * an outgoing multicast interface has been set, 1849 * then the ip bind logic will pick the correct source 1850 * address (i.e. matching the outgoing multicast interface). 1851 */ 1852 } 1853 1854 /* 1855 * Verify that the src/port/dst/port is unique for all 1856 * connections in TS_DATA_XFER 1857 */ 1858 mutex_enter(&udpf->uf_lock); 1859 for (udp1 = udpf->uf_udp; udp1 != NULL; udp1 = udp1->udp_bind_hash) { 1860 if (udp1->udp_state != TS_DATA_XFER) 1861 continue; 1862 if (udp->udp_port != udp1->udp_port || 1863 udp->udp_ipversion != udp1->udp_ipversion || 1864 dstport != udp1->udp_dstport || 1865 !IN6_ARE_ADDR_EQUAL(&udp->udp_v6src, &udp1->udp_v6src) || 1866 !IN6_ARE_ADDR_EQUAL(&v6dst, &udp1->udp_v6dst)) 1867 continue; 1868 mutex_exit(&udpf->uf_lock); 1869 udp_err_ack(q, mp, TBADADDR, 0); 1870 return; 1871 } 1872 udp->udp_state = TS_DATA_XFER; 1873 mutex_exit(&udpf->uf_lock); 1874 1875 /* 1876 * Send down bind to IP to verify that there is a route 1877 * and to determine the source address. 1878 * This will come back as T_BIND_ACK with an IRE_DB_TYPE in rput. 1879 */ 1880 if (udp->udp_family == AF_INET) 1881 mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (ipa_conn_t)); 1882 else 1883 mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (ipa6_conn_t)); 1884 if (mp1 == NULL) { 1885 udp_err_ack(q, mp, TSYSERR, ENOMEM); 1886 bind_failed: 1887 mutex_enter(&udpf->uf_lock); 1888 udp->udp_state = TS_IDLE; 1889 mutex_exit(&udpf->uf_lock); 1890 return; 1891 } 1892 1893 /* 1894 * We also have to send a connection confirmation to 1895 * keep TLI happy. Prepare it for udp_rput. 1896 */ 1897 if (udp->udp_family == AF_INET) 1898 mp2 = mi_tpi_conn_con(NULL, (char *)sin, 1899 sizeof (*sin), NULL, 0); 1900 else 1901 mp2 = mi_tpi_conn_con(NULL, (char *)sin6, 1902 sizeof (*sin6), NULL, 0); 1903 if (mp2 == NULL) { 1904 freemsg(mp1); 1905 udp_err_ack(q, mp, TSYSERR, ENOMEM); 1906 goto bind_failed; 1907 } 1908 1909 mp = mi_tpi_ok_ack_alloc(mp); 1910 if (mp == NULL) { 1911 /* Unable to reuse the T_CONN_REQ for the ack. */ 1912 freemsg(mp2); 1913 udp_err_ack_prim(q, mp1, T_CONN_REQ, TSYSERR, ENOMEM); 1914 goto bind_failed; 1915 } 1916 1917 /* Hang onto the T_OK_ACK and T_CONN_CON for later. */ 1918 linkb(mp1, mp); 1919 linkb(mp1, mp2); 1920 1921 mblk_setcred(mp1, udp->udp_connp->conn_cred); 1922 if (udp->udp_family == AF_INET) 1923 mp1 = ip_bind_v4(q, mp1, udp->udp_connp); 1924 else 1925 mp1 = ip_bind_v6(q, mp1, udp->udp_connp, NULL); 1926 1927 if (mp1 != NULL) 1928 udp_rput_other(_RD(q), mp1); 1929 else 1930 CONN_INC_REF(udp->udp_connp); 1931 } 1932 1933 static int 1934 udp_close(queue_t *q) 1935 { 1936 conn_t *connp = Q_TO_CONN(UDP_WR(q)); 1937 udp_t *udp; 1938 queue_t *ip_rq = RD(UDP_WR(q)); 1939 1940 ASSERT(connp != NULL && IPCL_IS_UDP(connp)); 1941 udp = connp->conn_udp; 1942 1943 ip_quiesce_conn(connp); 1944 /* 1945 * Disable read-side synchronous stream 1946 * interface and drain any queued data. 1947 */ 1948 udp_rcv_drain(q, udp, B_TRUE); 1949 ASSERT(!udp->udp_direct_sockfs); 1950 1951 qprocsoff(q); 1952 1953 /* restore IP module's high and low water marks to default values */ 1954 ip_rq->q_hiwat = ip_rq->q_qinfo->qi_minfo->mi_hiwat; 1955 WR(ip_rq)->q_hiwat = WR(ip_rq)->q_qinfo->qi_minfo->mi_hiwat; 1956 WR(ip_rq)->q_lowat = WR(ip_rq)->q_qinfo->qi_minfo->mi_lowat; 1957 1958 ASSERT(udp->udp_rcv_cnt == 0); 1959 ASSERT(udp->udp_rcv_msgcnt == 0); 1960 ASSERT(udp->udp_rcv_list_head == NULL); 1961 ASSERT(udp->udp_rcv_list_tail == NULL); 1962 1963 /* connp is now single threaded. */ 1964 udp_close_free(connp); 1965 /* 1966 * Restore connp as an IP endpoint. We don't need 1967 * any locks since we are now single threaded 1968 */ 1969 connp->conn_flags &= ~IPCL_UDP; 1970 connp->conn_state_flags &= 1971 ~(CONN_CLOSING | CONN_CONDEMNED | CONN_QUIESCED); 1972 connp->conn_ulp_labeled = B_FALSE; 1973 return (0); 1974 } 1975 1976 /* 1977 * Called in the close path from IP (ip_quiesce_conn) to quiesce the conn 1978 */ 1979 void 1980 udp_quiesce_conn(conn_t *connp) 1981 { 1982 udp_t *udp = connp->conn_udp; 1983 1984 if (cl_inet_unbind != NULL && udp->udp_state == TS_IDLE) { 1985 /* 1986 * Running in cluster mode - register unbind information 1987 */ 1988 if (udp->udp_ipversion == IPV4_VERSION) { 1989 (*cl_inet_unbind)(IPPROTO_UDP, AF_INET, 1990 (uint8_t *)(&(V4_PART_OF_V6(udp->udp_v6src))), 1991 (in_port_t)udp->udp_port); 1992 } else { 1993 (*cl_inet_unbind)(IPPROTO_UDP, AF_INET6, 1994 (uint8_t *)(&(udp->udp_v6src)), 1995 (in_port_t)udp->udp_port); 1996 } 1997 } 1998 1999 udp_bind_hash_remove(udp, B_FALSE); 2000 2001 mutex_enter(&connp->conn_lock); 2002 while (udp->udp_reader_count != 0 || udp->udp_squeue_count != 0 || 2003 udp->udp_mode != UDP_MT_HOT) { 2004 cv_wait(&connp->conn_cv, &connp->conn_lock); 2005 } 2006 mutex_exit(&connp->conn_lock); 2007 } 2008 2009 void 2010 udp_close_free(conn_t *connp) 2011 { 2012 udp_t *udp = connp->conn_udp; 2013 2014 /* If there are any options associated with the stream, free them. */ 2015 if (udp->udp_ip_snd_options) { 2016 mi_free((char *)udp->udp_ip_snd_options); 2017 udp->udp_ip_snd_options = NULL; 2018 } 2019 2020 if (udp->udp_ip_rcv_options) { 2021 mi_free((char *)udp->udp_ip_rcv_options); 2022 udp->udp_ip_rcv_options = NULL; 2023 } 2024 2025 /* Free memory associated with sticky options */ 2026 if (udp->udp_sticky_hdrs_len != 0) { 2027 kmem_free(udp->udp_sticky_hdrs, 2028 udp->udp_sticky_hdrs_len); 2029 udp->udp_sticky_hdrs = NULL; 2030 udp->udp_sticky_hdrs_len = 0; 2031 } 2032 2033 ip6_pkt_free(&udp->udp_sticky_ipp); 2034 2035 udp->udp_connp = NULL; 2036 connp->conn_udp = NULL; 2037 kmem_cache_free(udp_cache, udp); 2038 } 2039 2040 /* 2041 * This routine handles each T_DISCON_REQ message passed to udp 2042 * as an indicating that UDP is no longer connected. This results 2043 * in sending a T_BIND_REQ to IP to restore the binding to just 2044 * the local address/port. 2045 * 2046 * This routine sends down a T_BIND_REQ to IP with the following mblks: 2047 * T_BIND_REQ - specifying just the local address/port 2048 * T_OK_ACK - for the T_DISCON_REQ 2049 * 2050 * The disconnect completes in udp_rput. 2051 * When a T_BIND_ACK is received the appended T_OK_ACK is sent to the TPI user. 2052 * Should udp_rput receive T_ERROR_ACK for the T_BIND_REQ it will convert 2053 * it to an error ack for the appropriate primitive. 2054 */ 2055 static void 2056 udp_disconnect(queue_t *q, mblk_t *mp) 2057 { 2058 udp_t *udp = Q_TO_UDP(q); 2059 mblk_t *mp1; 2060 udp_fanout_t *udpf; 2061 2062 if (udp->udp_state != TS_DATA_XFER) { 2063 (void) mi_strlog(q, 1, SL_ERROR|SL_TRACE, 2064 "udp_disconnect: bad state, %u", udp->udp_state); 2065 udp_err_ack(q, mp, TOUTSTATE, 0); 2066 return; 2067 } 2068 udpf = &udp_bind_fanout[UDP_BIND_HASH(udp->udp_port)]; 2069 mutex_enter(&udpf->uf_lock); 2070 udp->udp_v6src = udp->udp_bound_v6src; 2071 udp->udp_state = TS_IDLE; 2072 mutex_exit(&udpf->uf_lock); 2073 2074 /* 2075 * Send down bind to IP to remove the full binding and revert 2076 * to the local address binding. 2077 */ 2078 if (udp->udp_family == AF_INET) 2079 mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (sin_t)); 2080 else 2081 mp1 = udp_ip_bind_mp(udp, O_T_BIND_REQ, sizeof (sin6_t)); 2082 if (mp1 == NULL) { 2083 udp_err_ack(q, mp, TSYSERR, ENOMEM); 2084 return; 2085 } 2086 mp = mi_tpi_ok_ack_alloc(mp); 2087 if (mp == NULL) { 2088 /* Unable to reuse the T_DISCON_REQ for the ack. */ 2089 udp_err_ack_prim(q, mp1, T_DISCON_REQ, TSYSERR, ENOMEM); 2090 return; 2091 } 2092 2093 if (udp->udp_family == AF_INET6) { 2094 int error; 2095 2096 /* Rebuild the header template */ 2097 error = udp_build_hdrs(q, udp); 2098 if (error != 0) { 2099 udp_err_ack_prim(q, mp, T_DISCON_REQ, TSYSERR, error); 2100 freemsg(mp1); 2101 return; 2102 } 2103 } 2104 mutex_enter(&udpf->uf_lock); 2105 udp->udp_discon_pending = 1; 2106 mutex_exit(&udpf->uf_lock); 2107 2108 /* Append the T_OK_ACK to the T_BIND_REQ for udp_rput */ 2109 linkb(mp1, mp); 2110 2111 if (udp->udp_family == AF_INET6) 2112 mp1 = ip_bind_v6(q, mp1, udp->udp_connp, NULL); 2113 else 2114 mp1 = ip_bind_v4(q, mp1, udp->udp_connp); 2115 2116 if (mp1 != NULL) 2117 udp_rput_other(_RD(q), mp1); 2118 else 2119 CONN_INC_REF(udp->udp_connp); 2120 } 2121 2122 /* This routine creates a T_ERROR_ACK message and passes it upstream. */ 2123 static void 2124 udp_err_ack(queue_t *q, mblk_t *mp, t_scalar_t t_error, int sys_error) 2125 { 2126 if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL) 2127 putnext(UDP_RD(q), mp); 2128 } 2129 2130 /* Shorthand to generate and send TPI error acks to our client */ 2131 static void 2132 udp_err_ack_prim(queue_t *q, mblk_t *mp, int primitive, t_scalar_t t_error, 2133 int sys_error) 2134 { 2135 struct T_error_ack *teackp; 2136 2137 if ((mp = tpi_ack_alloc(mp, sizeof (struct T_error_ack), 2138 M_PCPROTO, T_ERROR_ACK)) != NULL) { 2139 teackp = (struct T_error_ack *)mp->b_rptr; 2140 teackp->ERROR_prim = primitive; 2141 teackp->TLI_error = t_error; 2142 teackp->UNIX_error = sys_error; 2143 putnext(UDP_RD(q), mp); 2144 } 2145 } 2146 2147 /*ARGSUSED*/ 2148 static int 2149 udp_extra_priv_ports_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr) 2150 { 2151 int i; 2152 2153 for (i = 0; i < udp_g_num_epriv_ports; i++) { 2154 if (udp_g_epriv_ports[i] != 0) 2155 (void) mi_mpprintf(mp, "%d ", udp_g_epriv_ports[i]); 2156 } 2157 return (0); 2158 } 2159 2160 /* ARGSUSED */ 2161 static int 2162 udp_extra_priv_ports_add(queue_t *q, mblk_t *mp, char *value, caddr_t cp, 2163 cred_t *cr) 2164 { 2165 long new_value; 2166 int i; 2167 2168 /* 2169 * Fail the request if the new value does not lie within the 2170 * port number limits. 2171 */ 2172 if (ddi_strtol(value, NULL, 10, &new_value) != 0 || 2173 new_value <= 0 || new_value >= 65536) { 2174 return (EINVAL); 2175 } 2176 2177 /* Check if the value is already in the list */ 2178 for (i = 0; i < udp_g_num_epriv_ports; i++) { 2179 if (new_value == udp_g_epriv_ports[i]) { 2180 return (EEXIST); 2181 } 2182 } 2183 /* Find an empty slot */ 2184 for (i = 0; i < udp_g_num_epriv_ports; i++) { 2185 if (udp_g_epriv_ports[i] == 0) 2186 break; 2187 } 2188 if (i == udp_g_num_epriv_ports) { 2189 return (EOVERFLOW); 2190 } 2191 2192 /* Set the new value */ 2193 udp_g_epriv_ports[i] = (in_port_t)new_value; 2194 return (0); 2195 } 2196 2197 /* ARGSUSED */ 2198 static int 2199 udp_extra_priv_ports_del(queue_t *q, mblk_t *mp, char *value, caddr_t cp, 2200 cred_t *cr) 2201 { 2202 long new_value; 2203 int i; 2204 2205 /* 2206 * Fail the request if the new value does not lie within the 2207 * port number limits. 2208 */ 2209 if (ddi_strtol(value, NULL, 10, &new_value) != 0 || 2210 new_value <= 0 || new_value >= 65536) { 2211 return (EINVAL); 2212 } 2213 2214 /* Check that the value is already in the list */ 2215 for (i = 0; i < udp_g_num_epriv_ports; i++) { 2216 if (udp_g_epriv_ports[i] == new_value) 2217 break; 2218 } 2219 if (i == udp_g_num_epriv_ports) { 2220 return (ESRCH); 2221 } 2222 2223 /* Clear the value */ 2224 udp_g_epriv_ports[i] = 0; 2225 return (0); 2226 } 2227 2228 /* At minimum we need 4 bytes of UDP header */ 2229 #define ICMP_MIN_UDP_HDR 4 2230 2231 /* 2232 * udp_icmp_error is called by udp_rput to process ICMP msgs. passed up by IP. 2233 * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors. 2234 * Assumes that IP has pulled up everything up to and including the ICMP header. 2235 * An M_CTL could potentially come here from some other module (i.e. if UDP 2236 * is pushed on some module other than IP). Thus, if we find that the M_CTL 2237 * does not have enough ICMP information , following STREAMS conventions, 2238 * we send it upstream assuming it is an M_CTL we don't understand. 2239 */ 2240 static void 2241 udp_icmp_error(queue_t *q, mblk_t *mp) 2242 { 2243 icmph_t *icmph; 2244 ipha_t *ipha; 2245 int iph_hdr_length; 2246 udpha_t *udpha; 2247 sin_t sin; 2248 sin6_t sin6; 2249 mblk_t *mp1; 2250 int error = 0; 2251 size_t mp_size = MBLKL(mp); 2252 udp_t *udp = Q_TO_UDP(q); 2253 2254 /* 2255 * Assume IP provides aligned packets - otherwise toss 2256 */ 2257 if (!OK_32PTR(mp->b_rptr)) { 2258 freemsg(mp); 2259 return; 2260 } 2261 2262 /* 2263 * Verify that we have a complete IP header and the application has 2264 * asked for errors. If not, send it upstream. 2265 */ 2266 if (!udp->udp_dgram_errind || mp_size < sizeof (ipha_t)) { 2267 noticmpv4: 2268 putnext(UDP_RD(q), mp); 2269 return; 2270 } 2271 2272 ipha = (ipha_t *)mp->b_rptr; 2273 /* 2274 * Verify IP version. Anything other than IPv4 or IPv6 packet is sent 2275 * upstream. ICMPv6 is handled in udp_icmp_error_ipv6. 2276 */ 2277 switch (IPH_HDR_VERSION(ipha)) { 2278 case IPV6_VERSION: 2279 udp_icmp_error_ipv6(q, mp); 2280 return; 2281 case IPV4_VERSION: 2282 break; 2283 default: 2284 goto noticmpv4; 2285 } 2286 2287 /* Skip past the outer IP and ICMP headers */ 2288 iph_hdr_length = IPH_HDR_LENGTH(ipha); 2289 icmph = (icmph_t *)&mp->b_rptr[iph_hdr_length]; 2290 /* 2291 * If we don't have the correct outer IP header length or if the ULP 2292 * is not IPPROTO_ICMP or if we don't have a complete inner IP header 2293 * send the packet upstream. 2294 */ 2295 if (iph_hdr_length < sizeof (ipha_t) || 2296 ipha->ipha_protocol != IPPROTO_ICMP || 2297 (ipha_t *)&icmph[1] + 1 > (ipha_t *)mp->b_wptr) { 2298 goto noticmpv4; 2299 } 2300 ipha = (ipha_t *)&icmph[1]; 2301 2302 /* Skip past the inner IP and find the ULP header */ 2303 iph_hdr_length = IPH_HDR_LENGTH(ipha); 2304 udpha = (udpha_t *)((char *)ipha + iph_hdr_length); 2305 /* 2306 * If we don't have the correct inner IP header length or if the ULP 2307 * is not IPPROTO_UDP or if we don't have at least ICMP_MIN_UDP_HDR 2308 * bytes of UDP header, send it upstream. 2309 */ 2310 if (iph_hdr_length < sizeof (ipha_t) || 2311 ipha->ipha_protocol != IPPROTO_UDP || 2312 (uchar_t *)udpha + ICMP_MIN_UDP_HDR > mp->b_wptr) { 2313 goto noticmpv4; 2314 } 2315 2316 switch (icmph->icmph_type) { 2317 case ICMP_DEST_UNREACHABLE: 2318 switch (icmph->icmph_code) { 2319 case ICMP_FRAGMENTATION_NEEDED: 2320 /* 2321 * IP has already adjusted the path MTU. 2322 * XXX Somehow pass MTU indication to application? 2323 */ 2324 break; 2325 case ICMP_PORT_UNREACHABLE: 2326 case ICMP_PROTOCOL_UNREACHABLE: 2327 error = ECONNREFUSED; 2328 break; 2329 default: 2330 /* Transient errors */ 2331 break; 2332 } 2333 break; 2334 default: 2335 /* Transient errors */ 2336 break; 2337 } 2338 if (error == 0) { 2339 freemsg(mp); 2340 return; 2341 } 2342 2343 switch (udp->udp_family) { 2344 case AF_INET: 2345 sin = sin_null; 2346 sin.sin_family = AF_INET; 2347 sin.sin_addr.s_addr = ipha->ipha_dst; 2348 sin.sin_port = udpha->uha_dst_port; 2349 mp1 = mi_tpi_uderror_ind((char *)&sin, sizeof (sin_t), NULL, 0, 2350 error); 2351 break; 2352 case AF_INET6: 2353 sin6 = sin6_null; 2354 sin6.sin6_family = AF_INET6; 2355 IN6_IPADDR_TO_V4MAPPED(ipha->ipha_dst, &sin6.sin6_addr); 2356 sin6.sin6_port = udpha->uha_dst_port; 2357 2358 mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), 2359 NULL, 0, error); 2360 break; 2361 } 2362 if (mp1) 2363 putnext(UDP_RD(q), mp1); 2364 freemsg(mp); 2365 } 2366 2367 /* 2368 * udp_icmp_error_ipv6 is called by udp_icmp_error to process ICMP for IPv6. 2369 * Generates the appropriate T_UDERROR_IND for permanent (non-transient) errors. 2370 * Assumes that IP has pulled up all the extension headers as well as the 2371 * ICMPv6 header. 2372 * An M_CTL could potentially come here from some other module (i.e. if UDP 2373 * is pushed on some module other than IP). Thus, if we find that the M_CTL 2374 * does not have enough ICMP information , following STREAMS conventions, 2375 * we send it upstream assuming it is an M_CTL we don't understand. The reason 2376 * it might get here is if the non-ICMP M_CTL accidently has 6 in the version 2377 * field (when cast to ipha_t in udp_icmp_error). 2378 */ 2379 static void 2380 udp_icmp_error_ipv6(queue_t *q, mblk_t *mp) 2381 { 2382 icmp6_t *icmp6; 2383 ip6_t *ip6h, *outer_ip6h; 2384 uint16_t hdr_length; 2385 uint8_t *nexthdrp; 2386 udpha_t *udpha; 2387 sin6_t sin6; 2388 mblk_t *mp1; 2389 int error = 0; 2390 size_t mp_size = MBLKL(mp); 2391 udp_t *udp = Q_TO_UDP(q); 2392 2393 /* 2394 * Verify that we have a complete IP header. If not, send it upstream. 2395 */ 2396 if (mp_size < sizeof (ip6_t)) { 2397 noticmpv6: 2398 putnext(UDP_RD(q), mp); 2399 return; 2400 } 2401 2402 outer_ip6h = (ip6_t *)mp->b_rptr; 2403 /* 2404 * Verify this is an ICMPV6 packet, else send it upstream 2405 */ 2406 if (outer_ip6h->ip6_nxt == IPPROTO_ICMPV6) { 2407 hdr_length = IPV6_HDR_LEN; 2408 } else if (!ip_hdr_length_nexthdr_v6(mp, outer_ip6h, &hdr_length, 2409 &nexthdrp) || 2410 *nexthdrp != IPPROTO_ICMPV6) { 2411 goto noticmpv6; 2412 } 2413 icmp6 = (icmp6_t *)&mp->b_rptr[hdr_length]; 2414 ip6h = (ip6_t *)&icmp6[1]; 2415 /* 2416 * Verify we have a complete ICMP and inner IP header. 2417 */ 2418 if ((uchar_t *)&ip6h[1] > mp->b_wptr) 2419 goto noticmpv6; 2420 2421 if (!ip_hdr_length_nexthdr_v6(mp, ip6h, &hdr_length, &nexthdrp)) 2422 goto noticmpv6; 2423 udpha = (udpha_t *)((char *)ip6h + hdr_length); 2424 /* 2425 * Validate inner header. If the ULP is not IPPROTO_UDP or if we don't 2426 * have at least ICMP_MIN_UDP_HDR bytes of UDP header send the 2427 * packet upstream. 2428 */ 2429 if ((*nexthdrp != IPPROTO_UDP) || 2430 ((uchar_t *)udpha + ICMP_MIN_UDP_HDR) > mp->b_wptr) { 2431 goto noticmpv6; 2432 } 2433 2434 switch (icmp6->icmp6_type) { 2435 case ICMP6_DST_UNREACH: 2436 switch (icmp6->icmp6_code) { 2437 case ICMP6_DST_UNREACH_NOPORT: 2438 error = ECONNREFUSED; 2439 break; 2440 case ICMP6_DST_UNREACH_ADMIN: 2441 case ICMP6_DST_UNREACH_NOROUTE: 2442 case ICMP6_DST_UNREACH_BEYONDSCOPE: 2443 case ICMP6_DST_UNREACH_ADDR: 2444 /* Transient errors */ 2445 break; 2446 default: 2447 break; 2448 } 2449 break; 2450 case ICMP6_PACKET_TOO_BIG: { 2451 struct T_unitdata_ind *tudi; 2452 struct T_opthdr *toh; 2453 size_t udi_size; 2454 mblk_t *newmp; 2455 t_scalar_t opt_length = sizeof (struct T_opthdr) + 2456 sizeof (struct ip6_mtuinfo); 2457 sin6_t *sin6; 2458 struct ip6_mtuinfo *mtuinfo; 2459 2460 /* 2461 * If the application has requested to receive path mtu 2462 * information, send up an empty message containing an 2463 * IPV6_PATHMTU ancillary data item. 2464 */ 2465 if (!udp->udp_ipv6_recvpathmtu) 2466 break; 2467 2468 udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t) + 2469 opt_length; 2470 if ((newmp = allocb(udi_size, BPRI_MED)) == NULL) { 2471 BUMP_MIB(&udp_mib, udpInErrors); 2472 break; 2473 } 2474 2475 /* 2476 * newmp->b_cont is left to NULL on purpose. This is an 2477 * empty message containing only ancillary data. 2478 */ 2479 newmp->b_datap->db_type = M_PROTO; 2480 tudi = (struct T_unitdata_ind *)newmp->b_rptr; 2481 newmp->b_wptr = (uchar_t *)tudi + udi_size; 2482 tudi->PRIM_type = T_UNITDATA_IND; 2483 tudi->SRC_length = sizeof (sin6_t); 2484 tudi->SRC_offset = sizeof (struct T_unitdata_ind); 2485 tudi->OPT_offset = tudi->SRC_offset + sizeof (sin6_t); 2486 tudi->OPT_length = opt_length; 2487 2488 sin6 = (sin6_t *)&tudi[1]; 2489 bzero(sin6, sizeof (sin6_t)); 2490 sin6->sin6_family = AF_INET6; 2491 sin6->sin6_addr = udp->udp_v6dst; 2492 2493 toh = (struct T_opthdr *)&sin6[1]; 2494 toh->level = IPPROTO_IPV6; 2495 toh->name = IPV6_PATHMTU; 2496 toh->len = opt_length; 2497 toh->status = 0; 2498 2499 mtuinfo = (struct ip6_mtuinfo *)&toh[1]; 2500 bzero(mtuinfo, sizeof (struct ip6_mtuinfo)); 2501 mtuinfo->ip6m_addr.sin6_family = AF_INET6; 2502 mtuinfo->ip6m_addr.sin6_addr = ip6h->ip6_dst; 2503 mtuinfo->ip6m_mtu = icmp6->icmp6_mtu; 2504 /* 2505 * We've consumed everything we need from the original 2506 * message. Free it, then send our empty message. 2507 */ 2508 freemsg(mp); 2509 putnext(UDP_RD(q), newmp); 2510 return; 2511 } 2512 case ICMP6_TIME_EXCEEDED: 2513 /* Transient errors */ 2514 break; 2515 case ICMP6_PARAM_PROB: 2516 /* If this corresponds to an ICMP_PROTOCOL_UNREACHABLE */ 2517 if (icmp6->icmp6_code == ICMP6_PARAMPROB_NEXTHEADER && 2518 (uchar_t *)ip6h + icmp6->icmp6_pptr == 2519 (uchar_t *)nexthdrp) { 2520 error = ECONNREFUSED; 2521 break; 2522 } 2523 break; 2524 } 2525 if (error == 0) { 2526 freemsg(mp); 2527 return; 2528 } 2529 2530 sin6 = sin6_null; 2531 sin6.sin6_family = AF_INET6; 2532 sin6.sin6_addr = ip6h->ip6_dst; 2533 sin6.sin6_port = udpha->uha_dst_port; 2534 sin6.sin6_flowinfo = ip6h->ip6_vcf & ~IPV6_VERS_AND_FLOW_MASK; 2535 2536 mp1 = mi_tpi_uderror_ind((char *)&sin6, sizeof (sin6_t), NULL, 0, 2537 error); 2538 if (mp1) 2539 putnext(UDP_RD(q), mp1); 2540 freemsg(mp); 2541 } 2542 2543 /* 2544 * This routine responds to T_ADDR_REQ messages. It is called by udp_wput. 2545 * The local address is filled in if endpoint is bound. The remote address 2546 * is filled in if remote address has been precified ("connected endpoint") 2547 * (The concept of connected CLTS sockets is alien to published TPI 2548 * but we support it anyway). 2549 */ 2550 static void 2551 udp_addr_req(queue_t *q, mblk_t *mp) 2552 { 2553 sin_t *sin; 2554 sin6_t *sin6; 2555 mblk_t *ackmp; 2556 struct T_addr_ack *taa; 2557 udp_t *udp = Q_TO_UDP(q); 2558 2559 /* Make it large enough for worst case */ 2560 ackmp = reallocb(mp, sizeof (struct T_addr_ack) + 2561 2 * sizeof (sin6_t), 1); 2562 if (ackmp == NULL) { 2563 udp_err_ack(q, mp, TSYSERR, ENOMEM); 2564 return; 2565 } 2566 taa = (struct T_addr_ack *)ackmp->b_rptr; 2567 2568 bzero(taa, sizeof (struct T_addr_ack)); 2569 ackmp->b_wptr = (uchar_t *)&taa[1]; 2570 2571 taa->PRIM_type = T_ADDR_ACK; 2572 ackmp->b_datap->db_type = M_PCPROTO; 2573 /* 2574 * Note: Following code assumes 32 bit alignment of basic 2575 * data structures like sin_t and struct T_addr_ack. 2576 */ 2577 if (udp->udp_state != TS_UNBND) { 2578 /* 2579 * Fill in local address first 2580 */ 2581 taa->LOCADDR_offset = sizeof (*taa); 2582 if (udp->udp_family == AF_INET) { 2583 taa->LOCADDR_length = sizeof (sin_t); 2584 sin = (sin_t *)&taa[1]; 2585 /* Fill zeroes and then initialize non-zero fields */ 2586 *sin = sin_null; 2587 sin->sin_family = AF_INET; 2588 if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) && 2589 !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) { 2590 IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src, 2591 sin->sin_addr.s_addr); 2592 } else { 2593 /* 2594 * INADDR_ANY 2595 * udp_v6src is not set, we might be bound to 2596 * broadcast/multicast. Use udp_bound_v6src as 2597 * local address instead (that could 2598 * also still be INADDR_ANY) 2599 */ 2600 IN6_V4MAPPED_TO_IPADDR(&udp->udp_bound_v6src, 2601 sin->sin_addr.s_addr); 2602 } 2603 sin->sin_port = udp->udp_port; 2604 ackmp->b_wptr = (uchar_t *)&sin[1]; 2605 if (udp->udp_state == TS_DATA_XFER) { 2606 /* 2607 * connected, fill remote address too 2608 */ 2609 taa->REMADDR_length = sizeof (sin_t); 2610 /* assumed 32-bit alignment */ 2611 taa->REMADDR_offset = taa->LOCADDR_offset + 2612 taa->LOCADDR_length; 2613 2614 sin = (sin_t *)(ackmp->b_rptr + 2615 taa->REMADDR_offset); 2616 /* initialize */ 2617 *sin = sin_null; 2618 sin->sin_family = AF_INET; 2619 sin->sin_addr.s_addr = 2620 V4_PART_OF_V6(udp->udp_v6dst); 2621 sin->sin_port = udp->udp_dstport; 2622 ackmp->b_wptr = (uchar_t *)&sin[1]; 2623 } 2624 } else { 2625 taa->LOCADDR_length = sizeof (sin6_t); 2626 sin6 = (sin6_t *)&taa[1]; 2627 /* Fill zeroes and then initialize non-zero fields */ 2628 *sin6 = sin6_null; 2629 sin6->sin6_family = AF_INET6; 2630 if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) { 2631 sin6->sin6_addr = udp->udp_v6src; 2632 } else { 2633 /* 2634 * UNSPECIFIED 2635 * udp_v6src is not set, we might be bound to 2636 * broadcast/multicast. Use udp_bound_v6src as 2637 * local address instead (that could 2638 * also still be UNSPECIFIED) 2639 */ 2640 sin6->sin6_addr = 2641 udp->udp_bound_v6src; 2642 } 2643 sin6->sin6_port = udp->udp_port; 2644 ackmp->b_wptr = (uchar_t *)&sin6[1]; 2645 if (udp->udp_state == TS_DATA_XFER) { 2646 /* 2647 * connected, fill remote address too 2648 */ 2649 taa->REMADDR_length = sizeof (sin6_t); 2650 /* assumed 32-bit alignment */ 2651 taa->REMADDR_offset = taa->LOCADDR_offset + 2652 taa->LOCADDR_length; 2653 2654 sin6 = (sin6_t *)(ackmp->b_rptr + 2655 taa->REMADDR_offset); 2656 /* initialize */ 2657 *sin6 = sin6_null; 2658 sin6->sin6_family = AF_INET6; 2659 sin6->sin6_addr = udp->udp_v6dst; 2660 sin6->sin6_port = udp->udp_dstport; 2661 ackmp->b_wptr = (uchar_t *)&sin6[1]; 2662 } 2663 ackmp->b_wptr = (uchar_t *)&sin6[1]; 2664 } 2665 } 2666 ASSERT(ackmp->b_wptr <= ackmp->b_datap->db_lim); 2667 putnext(UDP_RD(q), ackmp); 2668 } 2669 2670 static void 2671 udp_copy_info(struct T_info_ack *tap, udp_t *udp) 2672 { 2673 if (udp->udp_family == AF_INET) { 2674 *tap = udp_g_t_info_ack_ipv4; 2675 } else { 2676 *tap = udp_g_t_info_ack_ipv6; 2677 } 2678 tap->CURRENT_state = udp->udp_state; 2679 tap->OPT_size = udp_max_optsize; 2680 } 2681 2682 /* 2683 * This routine responds to T_CAPABILITY_REQ messages. It is called by 2684 * udp_wput. Much of the T_CAPABILITY_ACK information is copied from 2685 * udp_g_t_info_ack. The current state of the stream is copied from 2686 * udp_state. 2687 */ 2688 static void 2689 udp_capability_req(queue_t *q, mblk_t *mp) 2690 { 2691 t_uscalar_t cap_bits1; 2692 struct T_capability_ack *tcap; 2693 udp_t *udp = Q_TO_UDP(q); 2694 2695 cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1; 2696 2697 mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack), 2698 mp->b_datap->db_type, T_CAPABILITY_ACK); 2699 if (!mp) 2700 return; 2701 2702 tcap = (struct T_capability_ack *)mp->b_rptr; 2703 tcap->CAP_bits1 = 0; 2704 2705 if (cap_bits1 & TC1_INFO) { 2706 udp_copy_info(&tcap->INFO_ack, udp); 2707 tcap->CAP_bits1 |= TC1_INFO; 2708 } 2709 2710 putnext(UDP_RD(q), mp); 2711 } 2712 2713 /* 2714 * This routine responds to T_INFO_REQ messages. It is called by udp_wput. 2715 * Most of the T_INFO_ACK information is copied from udp_g_t_info_ack. 2716 * The current state of the stream is copied from udp_state. 2717 */ 2718 static void 2719 udp_info_req(queue_t *q, mblk_t *mp) 2720 { 2721 udp_t *udp = Q_TO_UDP(q); 2722 2723 /* Create a T_INFO_ACK message. */ 2724 mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO, 2725 T_INFO_ACK); 2726 if (!mp) 2727 return; 2728 udp_copy_info((struct T_info_ack *)mp->b_rptr, udp); 2729 putnext(UDP_RD(q), mp); 2730 } 2731 2732 /* 2733 * IP recognizes seven kinds of bind requests: 2734 * 2735 * - A zero-length address binds only to the protocol number. 2736 * 2737 * - A 4-byte address is treated as a request to 2738 * validate that the address is a valid local IPv4 2739 * address, appropriate for an application to bind to. 2740 * IP does the verification, but does not make any note 2741 * of the address at this time. 2742 * 2743 * - A 16-byte address contains is treated as a request 2744 * to validate a local IPv6 address, as the 4-byte 2745 * address case above. 2746 * 2747 * - A 16-byte sockaddr_in to validate the local IPv4 address and also 2748 * use it for the inbound fanout of packets. 2749 * 2750 * - A 24-byte sockaddr_in6 to validate the local IPv6 address and also 2751 * use it for the inbound fanout of packets. 2752 * 2753 * - A 12-byte address (ipa_conn_t) containing complete IPv4 fanout 2754 * information consisting of local and remote addresses 2755 * and ports. In this case, the addresses are both 2756 * validated as appropriate for this operation, and, if 2757 * so, the information is retained for use in the 2758 * inbound fanout. 2759 * 2760 * - A 36-byte address address (ipa6_conn_t) containing complete IPv6 2761 * fanout information, like the 12-byte case above. 2762 * 2763 * IP will also fill in the IRE request mblk with information 2764 * regarding our peer. In all cases, we notify IP of our protocol 2765 * type by appending a single protocol byte to the bind request. 2766 */ 2767 static mblk_t * 2768 udp_ip_bind_mp(udp_t *udp, t_scalar_t bind_prim, t_scalar_t addr_length) 2769 { 2770 char *cp; 2771 mblk_t *mp; 2772 struct T_bind_req *tbr; 2773 ipa_conn_t *ac; 2774 ipa6_conn_t *ac6; 2775 sin_t *sin; 2776 sin6_t *sin6; 2777 2778 ASSERT(bind_prim == O_T_BIND_REQ || bind_prim == T_BIND_REQ); 2779 2780 mp = allocb(sizeof (*tbr) + addr_length + 1, BPRI_HI); 2781 if (!mp) 2782 return (mp); 2783 mp->b_datap->db_type = M_PROTO; 2784 tbr = (struct T_bind_req *)mp->b_rptr; 2785 tbr->PRIM_type = bind_prim; 2786 tbr->ADDR_offset = sizeof (*tbr); 2787 tbr->CONIND_number = 0; 2788 tbr->ADDR_length = addr_length; 2789 cp = (char *)&tbr[1]; 2790 switch (addr_length) { 2791 case sizeof (ipa_conn_t): 2792 ASSERT(udp->udp_family == AF_INET); 2793 /* Append a request for an IRE */ 2794 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 2795 if (!mp->b_cont) { 2796 freemsg(mp); 2797 return (NULL); 2798 } 2799 mp->b_cont->b_wptr += sizeof (ire_t); 2800 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 2801 2802 /* cp known to be 32 bit aligned */ 2803 ac = (ipa_conn_t *)cp; 2804 ac->ac_laddr = V4_PART_OF_V6(udp->udp_v6src); 2805 ac->ac_faddr = V4_PART_OF_V6(udp->udp_v6dst); 2806 ac->ac_fport = udp->udp_dstport; 2807 ac->ac_lport = udp->udp_port; 2808 break; 2809 2810 case sizeof (ipa6_conn_t): 2811 ASSERT(udp->udp_family == AF_INET6); 2812 /* Append a request for an IRE */ 2813 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 2814 if (!mp->b_cont) { 2815 freemsg(mp); 2816 return (NULL); 2817 } 2818 mp->b_cont->b_wptr += sizeof (ire_t); 2819 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 2820 2821 /* cp known to be 32 bit aligned */ 2822 ac6 = (ipa6_conn_t *)cp; 2823 ac6->ac6_laddr = udp->udp_v6src; 2824 ac6->ac6_faddr = udp->udp_v6dst; 2825 ac6->ac6_fport = udp->udp_dstport; 2826 ac6->ac6_lport = udp->udp_port; 2827 break; 2828 2829 case sizeof (sin_t): 2830 ASSERT(udp->udp_family == AF_INET); 2831 /* Append a request for an IRE */ 2832 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 2833 if (!mp->b_cont) { 2834 freemsg(mp); 2835 return (NULL); 2836 } 2837 mp->b_cont->b_wptr += sizeof (ire_t); 2838 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 2839 2840 sin = (sin_t *)cp; 2841 *sin = sin_null; 2842 sin->sin_family = AF_INET; 2843 sin->sin_addr.s_addr = V4_PART_OF_V6(udp->udp_bound_v6src); 2844 sin->sin_port = udp->udp_port; 2845 break; 2846 2847 case sizeof (sin6_t): 2848 ASSERT(udp->udp_family == AF_INET6); 2849 /* Append a request for an IRE */ 2850 mp->b_cont = allocb(sizeof (ire_t), BPRI_HI); 2851 if (!mp->b_cont) { 2852 freemsg(mp); 2853 return (NULL); 2854 } 2855 mp->b_cont->b_wptr += sizeof (ire_t); 2856 mp->b_cont->b_datap->db_type = IRE_DB_REQ_TYPE; 2857 2858 sin6 = (sin6_t *)cp; 2859 *sin6 = sin6_null; 2860 sin6->sin6_family = AF_INET6; 2861 sin6->sin6_addr = udp->udp_bound_v6src; 2862 sin6->sin6_port = udp->udp_port; 2863 break; 2864 } 2865 /* Add protocol number to end */ 2866 cp[addr_length] = (char)IPPROTO_UDP; 2867 mp->b_wptr = (uchar_t *)&cp[addr_length + 1]; 2868 return (mp); 2869 } 2870 2871 /* 2872 * This is the open routine for udp. It allocates a udp_t structure for 2873 * the stream and, on the first open of the module, creates an ND table. 2874 */ 2875 /* ARGSUSED */ 2876 static int 2877 udp_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp) 2878 { 2879 int err; 2880 udp_t *udp; 2881 conn_t *connp; 2882 zoneid_t zoneid = getzoneid(); 2883 queue_t *ip_wq; 2884 char *name; 2885 2886 TRACE_1(TR_FAC_UDP, TR_UDP_OPEN, "udp_open: q %p", q); 2887 2888 /* If the stream is already open, return immediately. */ 2889 if (q->q_ptr != NULL) 2890 return (0); 2891 2892 /* If this is not a push of udp as a module, fail. */ 2893 if (sflag != MODOPEN) 2894 return (EINVAL); 2895 2896 q->q_hiwat = udp_recv_hiwat; 2897 WR(q)->q_hiwat = udp_xmit_hiwat; 2898 WR(q)->q_lowat = udp_xmit_lowat; 2899 2900 /* Insert ourselves in the stream since we're about to walk q_next */ 2901 qprocson(q); 2902 2903 udp = kmem_cache_alloc(udp_cache, KM_SLEEP); 2904 bzero(udp, sizeof (*udp)); 2905 2906 /* 2907 * UDP is supported only as a module and it has to be pushed directly 2908 * above the device instance of IP. If UDP is pushed anywhere else 2909 * on a stream, it will support just T_SVR4_OPTMGMT_REQ for the 2910 * sake of MIB browsers and fail everything else. 2911 */ 2912 ip_wq = WR(q)->q_next; 2913 if (ip_wq->q_next != NULL || 2914 (name = ip_wq->q_qinfo->qi_minfo->mi_idname) == NULL || 2915 strcmp(name, IP_MOD_NAME) != 0 || 2916 ip_wq->q_qinfo->qi_minfo->mi_idnum != IP_MOD_ID) { 2917 /* Support just SNMP for MIB browsers */ 2918 connp = ipcl_conn_create(IPCL_IPCCONN, KM_SLEEP); 2919 connp->conn_rq = q; 2920 connp->conn_wq = WR(q); 2921 connp->conn_flags |= IPCL_UDPMOD; 2922 connp->conn_cred = credp; 2923 connp->conn_zoneid = zoneid; 2924 connp->conn_udp = udp; 2925 udp->udp_connp = connp; 2926 q->q_ptr = WR(q)->q_ptr = connp; 2927 crhold(credp); 2928 q->q_qinfo = &udp_snmp_rinit; 2929 WR(q)->q_qinfo = &udp_snmp_winit; 2930 return (0); 2931 } 2932 2933 /* 2934 * Initialize the udp_t structure for this stream. 2935 */ 2936 q = RD(ip_wq); 2937 connp = Q_TO_CONN(q); 2938 mutex_enter(&connp->conn_lock); 2939 connp->conn_proto = IPPROTO_UDP; 2940 connp->conn_flags |= IPCL_UDP; 2941 connp->conn_sqp = IP_SQUEUE_GET(lbolt); 2942 connp->conn_udp = udp; 2943 2944 /* Set the initial state of the stream and the privilege status. */ 2945 udp->udp_connp = connp; 2946 udp->udp_state = TS_UNBND; 2947 udp->udp_mode = UDP_MT_HOT; 2948 if (getmajor(*devp) == (major_t)UDP6_MAJ) { 2949 udp->udp_family = AF_INET6; 2950 udp->udp_ipversion = IPV6_VERSION; 2951 udp->udp_max_hdr_len = IPV6_HDR_LEN + UDPH_SIZE; 2952 udp->udp_ttl = udp_ipv6_hoplimit; 2953 connp->conn_af_isv6 = B_TRUE; 2954 connp->conn_flags |= IPCL_ISV6; 2955 } else { 2956 udp->udp_family = AF_INET; 2957 udp->udp_ipversion = IPV4_VERSION; 2958 udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE; 2959 udp->udp_ttl = udp_ipv4_ttl; 2960 connp->conn_af_isv6 = B_FALSE; 2961 connp->conn_flags &= ~IPCL_ISV6; 2962 } 2963 2964 udp->udp_multicast_ttl = IP_DEFAULT_MULTICAST_TTL; 2965 connp->conn_multicast_loop = IP_DEFAULT_MULTICAST_LOOP; 2966 connp->conn_zoneid = zoneid; 2967 2968 /* 2969 * If the caller has the process-wide flag set, then default to MAC 2970 * exempt mode. This allows read-down to unlabeled hosts. 2971 */ 2972 if (getpflags(NET_MAC_AWARE, credp) != 0) 2973 udp->udp_mac_exempt = B_TRUE; 2974 2975 if (connp->conn_flags & IPCL_SOCKET) { 2976 udp->udp_issocket = B_TRUE; 2977 udp->udp_direct_sockfs = B_TRUE; 2978 } 2979 2980 connp->conn_ulp_labeled = is_system_labeled(); 2981 2982 mutex_exit(&connp->conn_lock); 2983 2984 /* 2985 * The transmit hiwat/lowat is only looked at on IP's queue. 2986 * Store in q_hiwat in order to return on SO_SNDBUF/SO_RCVBUF 2987 * getsockopts. 2988 */ 2989 q->q_hiwat = udp_recv_hiwat; 2990 WR(q)->q_hiwat = udp_xmit_hiwat; 2991 WR(q)->q_lowat = udp_xmit_lowat; 2992 2993 if (udp->udp_family == AF_INET6) { 2994 /* Build initial header template for transmit */ 2995 if ((err = udp_build_hdrs(q, udp)) != 0) { 2996 error: 2997 qprocsoff(UDP_RD(q)); 2998 udp->udp_connp = NULL; 2999 connp->conn_udp = NULL; 3000 kmem_cache_free(udp_cache, udp); 3001 return (err); 3002 } 3003 } 3004 3005 /* Set the Stream head write offset and high watermark. */ 3006 (void) mi_set_sth_wroff(UDP_RD(q), 3007 udp->udp_max_hdr_len + udp_wroff_extra); 3008 (void) mi_set_sth_hiwat(UDP_RD(q), udp_set_rcv_hiwat(udp, q->q_hiwat)); 3009 3010 WR(UDP_RD(q))->q_qinfo = &udp_winit; 3011 3012 return (0); 3013 } 3014 3015 /* 3016 * Which UDP options OK to set through T_UNITDATA_REQ... 3017 */ 3018 /* ARGSUSED */ 3019 static boolean_t 3020 udp_opt_allow_udr_set(t_scalar_t level, t_scalar_t name) 3021 { 3022 return (B_TRUE); 3023 } 3024 3025 /* 3026 * This routine gets default values of certain options whose default 3027 * values are maintained by protcol specific code 3028 */ 3029 /* ARGSUSED */ 3030 int 3031 udp_opt_default(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr) 3032 { 3033 int *i1 = (int *)ptr; 3034 3035 switch (level) { 3036 case IPPROTO_IP: 3037 switch (name) { 3038 case IP_MULTICAST_TTL: 3039 *ptr = (uchar_t)IP_DEFAULT_MULTICAST_TTL; 3040 return (sizeof (uchar_t)); 3041 case IP_MULTICAST_LOOP: 3042 *ptr = (uchar_t)IP_DEFAULT_MULTICAST_LOOP; 3043 return (sizeof (uchar_t)); 3044 } 3045 break; 3046 case IPPROTO_IPV6: 3047 switch (name) { 3048 case IPV6_MULTICAST_HOPS: 3049 *i1 = IP_DEFAULT_MULTICAST_TTL; 3050 return (sizeof (int)); 3051 case IPV6_MULTICAST_LOOP: 3052 *i1 = IP_DEFAULT_MULTICAST_LOOP; 3053 return (sizeof (int)); 3054 case IPV6_UNICAST_HOPS: 3055 *i1 = udp_ipv6_hoplimit; 3056 return (sizeof (int)); 3057 } 3058 break; 3059 } 3060 return (-1); 3061 } 3062 3063 /* 3064 * This routine retrieves the current status of socket options 3065 * and expects the caller to pass in the queue pointer of the 3066 * upper instance. It returns the size of the option retrieved. 3067 */ 3068 int 3069 udp_opt_get(queue_t *q, t_scalar_t level, t_scalar_t name, uchar_t *ptr) 3070 { 3071 int *i1 = (int *)ptr; 3072 conn_t *connp; 3073 udp_t *udp; 3074 ip6_pkt_t *ipp; 3075 int len; 3076 3077 q = UDP_WR(q); 3078 connp = Q_TO_CONN(q); 3079 udp = connp->conn_udp; 3080 ipp = &udp->udp_sticky_ipp; 3081 3082 switch (level) { 3083 case SOL_SOCKET: 3084 switch (name) { 3085 case SO_DEBUG: 3086 *i1 = udp->udp_debug; 3087 break; /* goto sizeof (int) option return */ 3088 case SO_REUSEADDR: 3089 *i1 = udp->udp_reuseaddr; 3090 break; /* goto sizeof (int) option return */ 3091 case SO_TYPE: 3092 *i1 = SOCK_DGRAM; 3093 break; /* goto sizeof (int) option return */ 3094 3095 /* 3096 * The following three items are available here, 3097 * but are only meaningful to IP. 3098 */ 3099 case SO_DONTROUTE: 3100 *i1 = udp->udp_dontroute; 3101 break; /* goto sizeof (int) option return */ 3102 case SO_USELOOPBACK: 3103 *i1 = udp->udp_useloopback; 3104 break; /* goto sizeof (int) option return */ 3105 case SO_BROADCAST: 3106 *i1 = udp->udp_broadcast; 3107 break; /* goto sizeof (int) option return */ 3108 3109 case SO_SNDBUF: 3110 *i1 = q->q_hiwat; 3111 break; /* goto sizeof (int) option return */ 3112 case SO_RCVBUF: 3113 *i1 = RD(q)->q_hiwat; 3114 break; /* goto sizeof (int) option return */ 3115 case SO_DGRAM_ERRIND: 3116 *i1 = udp->udp_dgram_errind; 3117 break; /* goto sizeof (int) option return */ 3118 case SO_RECVUCRED: 3119 *i1 = udp->udp_recvucred; 3120 break; /* goto sizeof (int) option return */ 3121 case SO_TIMESTAMP: 3122 *i1 = udp->udp_timestamp; 3123 break; /* goto sizeof (int) option return */ 3124 case SO_ANON_MLP: 3125 *i1 = udp->udp_anon_mlp; 3126 break; /* goto sizeof (int) option return */ 3127 case SO_MAC_EXEMPT: 3128 *i1 = udp->udp_mac_exempt; 3129 break; /* goto sizeof (int) option return */ 3130 case SO_ALLZONES: 3131 *i1 = connp->conn_allzones; 3132 break; /* goto sizeof (int) option return */ 3133 case SO_EXCLBIND: 3134 *i1 = udp->udp_exclbind ? SO_EXCLBIND : 0; 3135 break; 3136 default: 3137 return (-1); 3138 } 3139 break; 3140 case IPPROTO_IP: 3141 if (udp->udp_family != AF_INET) 3142 return (-1); 3143 switch (name) { 3144 case IP_OPTIONS: 3145 case T_IP_OPTIONS: 3146 len = udp->udp_ip_rcv_options_len - udp->udp_label_len; 3147 if (len > 0) { 3148 bcopy(udp->udp_ip_rcv_options + 3149 udp->udp_label_len, ptr, len); 3150 } 3151 return (len); 3152 case IP_TOS: 3153 case T_IP_TOS: 3154 *i1 = (int)udp->udp_type_of_service; 3155 break; /* goto sizeof (int) option return */ 3156 case IP_TTL: 3157 *i1 = (int)udp->udp_ttl; 3158 break; /* goto sizeof (int) option return */ 3159 case IP_NEXTHOP: 3160 /* Handled at IP level */ 3161 return (-EINVAL); 3162 case IP_MULTICAST_IF: 3163 /* 0 address if not set */ 3164 *(ipaddr_t *)ptr = udp->udp_multicast_if_addr; 3165 return (sizeof (ipaddr_t)); 3166 case IP_MULTICAST_TTL: 3167 *(uchar_t *)ptr = udp->udp_multicast_ttl; 3168 return (sizeof (uchar_t)); 3169 case IP_MULTICAST_LOOP: 3170 *ptr = connp->conn_multicast_loop; 3171 return (sizeof (uint8_t)); 3172 case IP_RECVOPTS: 3173 *i1 = udp->udp_recvopts; 3174 break; /* goto sizeof (int) option return */ 3175 case IP_RECVDSTADDR: 3176 *i1 = udp->udp_recvdstaddr; 3177 break; /* goto sizeof (int) option return */ 3178 case IP_RECVIF: 3179 *i1 = udp->udp_recvif; 3180 break; /* goto sizeof (int) option return */ 3181 case IP_RECVSLLA: 3182 *i1 = udp->udp_recvslla; 3183 break; /* goto sizeof (int) option return */ 3184 case IP_RECVTTL: 3185 *i1 = udp->udp_recvttl; 3186 break; /* goto sizeof (int) option return */ 3187 case IP_ADD_MEMBERSHIP: 3188 case IP_DROP_MEMBERSHIP: 3189 case IP_BLOCK_SOURCE: 3190 case IP_UNBLOCK_SOURCE: 3191 case IP_ADD_SOURCE_MEMBERSHIP: 3192 case IP_DROP_SOURCE_MEMBERSHIP: 3193 case MCAST_JOIN_GROUP: 3194 case MCAST_LEAVE_GROUP: 3195 case MCAST_BLOCK_SOURCE: 3196 case MCAST_UNBLOCK_SOURCE: 3197 case MCAST_JOIN_SOURCE_GROUP: 3198 case MCAST_LEAVE_SOURCE_GROUP: 3199 case IP_DONTFAILOVER_IF: 3200 /* cannot "get" the value for these */ 3201 return (-1); 3202 case IP_BOUND_IF: 3203 /* Zero if not set */ 3204 *i1 = udp->udp_bound_if; 3205 break; /* goto sizeof (int) option return */ 3206 case IP_UNSPEC_SRC: 3207 *i1 = udp->udp_unspec_source; 3208 break; /* goto sizeof (int) option return */ 3209 case IP_XMIT_IF: 3210 *i1 = udp->udp_xmit_if; 3211 break; /* goto sizeof (int) option return */ 3212 default: 3213 return (-1); 3214 } 3215 break; 3216 case IPPROTO_IPV6: 3217 if (udp->udp_family != AF_INET6) 3218 return (-1); 3219 switch (name) { 3220 case IPV6_UNICAST_HOPS: 3221 *i1 = (unsigned int)udp->udp_ttl; 3222 break; /* goto sizeof (int) option return */ 3223 case IPV6_MULTICAST_IF: 3224 /* 0 index if not set */ 3225 *i1 = udp->udp_multicast_if_index; 3226 break; /* goto sizeof (int) option return */ 3227 case IPV6_MULTICAST_HOPS: 3228 *i1 = udp->udp_multicast_ttl; 3229 break; /* goto sizeof (int) option return */ 3230 case IPV6_MULTICAST_LOOP: 3231 *i1 = connp->conn_multicast_loop; 3232 break; /* goto sizeof (int) option return */ 3233 case IPV6_JOIN_GROUP: 3234 case IPV6_LEAVE_GROUP: 3235 case MCAST_JOIN_GROUP: 3236 case MCAST_LEAVE_GROUP: 3237 case MCAST_BLOCK_SOURCE: 3238 case MCAST_UNBLOCK_SOURCE: 3239 case MCAST_JOIN_SOURCE_GROUP: 3240 case MCAST_LEAVE_SOURCE_GROUP: 3241 /* cannot "get" the value for these */ 3242 return (-1); 3243 case IPV6_BOUND_IF: 3244 /* Zero if not set */ 3245 *i1 = udp->udp_bound_if; 3246 break; /* goto sizeof (int) option return */ 3247 case IPV6_UNSPEC_SRC: 3248 *i1 = udp->udp_unspec_source; 3249 break; /* goto sizeof (int) option return */ 3250 case IPV6_RECVPKTINFO: 3251 *i1 = udp->udp_ipv6_recvpktinfo; 3252 break; /* goto sizeof (int) option return */ 3253 case IPV6_RECVTCLASS: 3254 *i1 = udp->udp_ipv6_recvtclass; 3255 break; /* goto sizeof (int) option return */ 3256 case IPV6_RECVPATHMTU: 3257 *i1 = udp->udp_ipv6_recvpathmtu; 3258 break; /* goto sizeof (int) option return */ 3259 case IPV6_RECVHOPLIMIT: 3260 *i1 = udp->udp_ipv6_recvhoplimit; 3261 break; /* goto sizeof (int) option return */ 3262 case IPV6_RECVHOPOPTS: 3263 *i1 = udp->udp_ipv6_recvhopopts; 3264 break; /* goto sizeof (int) option return */ 3265 case IPV6_RECVDSTOPTS: 3266 *i1 = udp->udp_ipv6_recvdstopts; 3267 break; /* goto sizeof (int) option return */ 3268 case _OLD_IPV6_RECVDSTOPTS: 3269 *i1 = udp->udp_old_ipv6_recvdstopts; 3270 break; /* goto sizeof (int) option return */ 3271 case IPV6_RECVRTHDRDSTOPTS: 3272 *i1 = udp->udp_ipv6_recvrthdrdstopts; 3273 break; /* goto sizeof (int) option return */ 3274 case IPV6_RECVRTHDR: 3275 *i1 = udp->udp_ipv6_recvrthdr; 3276 break; /* goto sizeof (int) option return */ 3277 case IPV6_PKTINFO: { 3278 /* XXX assumes that caller has room for max size! */ 3279 struct in6_pktinfo *pkti; 3280 3281 pkti = (struct in6_pktinfo *)ptr; 3282 if (ipp->ipp_fields & IPPF_IFINDEX) 3283 pkti->ipi6_ifindex = ipp->ipp_ifindex; 3284 else 3285 pkti->ipi6_ifindex = 0; 3286 if (ipp->ipp_fields & IPPF_ADDR) 3287 pkti->ipi6_addr = ipp->ipp_addr; 3288 else 3289 pkti->ipi6_addr = ipv6_all_zeros; 3290 return (sizeof (struct in6_pktinfo)); 3291 } 3292 case IPV6_TCLASS: 3293 if (ipp->ipp_fields & IPPF_TCLASS) 3294 *i1 = ipp->ipp_tclass; 3295 else 3296 *i1 = IPV6_FLOW_TCLASS( 3297 IPV6_DEFAULT_VERS_AND_FLOW); 3298 break; /* goto sizeof (int) option return */ 3299 case IPV6_NEXTHOP: { 3300 sin6_t *sin6 = (sin6_t *)ptr; 3301 3302 if (!(ipp->ipp_fields & IPPF_NEXTHOP)) 3303 return (0); 3304 *sin6 = sin6_null; 3305 sin6->sin6_family = AF_INET6; 3306 sin6->sin6_addr = ipp->ipp_nexthop; 3307 return (sizeof (sin6_t)); 3308 } 3309 case IPV6_HOPOPTS: 3310 if (!(ipp->ipp_fields & IPPF_HOPOPTS)) 3311 return (0); 3312 if (ipp->ipp_hopoptslen <= udp->udp_label_len_v6) 3313 return (0); 3314 /* 3315 * The cipso/label option is added by kernel. 3316 * User is not usually aware of this option. 3317 * We copy out the hbh opt after the label option. 3318 */ 3319 bcopy((char *)ipp->ipp_hopopts + udp->udp_label_len_v6, 3320 ptr, ipp->ipp_hopoptslen - udp->udp_label_len_v6); 3321 if (udp->udp_label_len_v6 > 0) { 3322 ptr[0] = ((char *)ipp->ipp_hopopts)[0]; 3323 ptr[1] = (ipp->ipp_hopoptslen - 3324 udp->udp_label_len_v6 + 7) / 8 - 1; 3325 } 3326 return (ipp->ipp_hopoptslen - udp->udp_label_len_v6); 3327 case IPV6_RTHDRDSTOPTS: 3328 if (!(ipp->ipp_fields & IPPF_RTDSTOPTS)) 3329 return (0); 3330 bcopy(ipp->ipp_rtdstopts, ptr, ipp->ipp_rtdstoptslen); 3331 return (ipp->ipp_rtdstoptslen); 3332 case IPV6_RTHDR: 3333 if (!(ipp->ipp_fields & IPPF_RTHDR)) 3334 return (0); 3335 bcopy(ipp->ipp_rthdr, ptr, ipp->ipp_rthdrlen); 3336 return (ipp->ipp_rthdrlen); 3337 case IPV6_DSTOPTS: 3338 if (!(ipp->ipp_fields & IPPF_DSTOPTS)) 3339 return (0); 3340 bcopy(ipp->ipp_dstopts, ptr, ipp->ipp_dstoptslen); 3341 return (ipp->ipp_dstoptslen); 3342 case IPV6_PATHMTU: 3343 return (ip_fill_mtuinfo(&udp->udp_v6dst, 3344 udp->udp_dstport, (struct ip6_mtuinfo *)ptr)); 3345 default: 3346 return (-1); 3347 } 3348 break; 3349 case IPPROTO_UDP: 3350 switch (name) { 3351 case UDP_ANONPRIVBIND: 3352 *i1 = udp->udp_anon_priv_bind; 3353 break; 3354 case UDP_EXCLBIND: 3355 *i1 = udp->udp_exclbind ? UDP_EXCLBIND : 0; 3356 break; 3357 case UDP_RCVHDR: 3358 *i1 = udp->udp_rcvhdr ? 1 : 0; 3359 break; 3360 default: 3361 return (-1); 3362 } 3363 break; 3364 default: 3365 return (-1); 3366 } 3367 return (sizeof (int)); 3368 } 3369 3370 /* 3371 * This routine sets socket options; it expects the caller 3372 * to pass in the queue pointer of the upper instance. 3373 */ 3374 /* ARGSUSED */ 3375 int 3376 udp_opt_set(queue_t *q, uint_t optset_context, int level, 3377 int name, uint_t inlen, uchar_t *invalp, uint_t *outlenp, 3378 uchar_t *outvalp, void *thisdg_attrs, cred_t *cr, mblk_t *mblk) 3379 { 3380 udpattrs_t *attrs = thisdg_attrs; 3381 int *i1 = (int *)invalp; 3382 boolean_t onoff = (*i1 == 0) ? 0 : 1; 3383 boolean_t checkonly; 3384 int error; 3385 conn_t *connp; 3386 udp_t *udp; 3387 uint_t newlen; 3388 3389 q = UDP_WR(q); 3390 connp = Q_TO_CONN(q); 3391 udp = connp->conn_udp; 3392 3393 switch (optset_context) { 3394 case SETFN_OPTCOM_CHECKONLY: 3395 checkonly = B_TRUE; 3396 /* 3397 * Note: Implies T_CHECK semantics for T_OPTCOM_REQ 3398 * inlen != 0 implies value supplied and 3399 * we have to "pretend" to set it. 3400 * inlen == 0 implies that there is no 3401 * value part in T_CHECK request and just validation 3402 * done elsewhere should be enough, we just return here. 3403 */ 3404 if (inlen == 0) { 3405 *outlenp = 0; 3406 return (0); 3407 } 3408 break; 3409 case SETFN_OPTCOM_NEGOTIATE: 3410 checkonly = B_FALSE; 3411 break; 3412 case SETFN_UD_NEGOTIATE: 3413 case SETFN_CONN_NEGOTIATE: 3414 checkonly = B_FALSE; 3415 /* 3416 * Negotiating local and "association-related" options 3417 * through T_UNITDATA_REQ. 3418 * 3419 * Following routine can filter out ones we do not 3420 * want to be "set" this way. 3421 */ 3422 if (!udp_opt_allow_udr_set(level, name)) { 3423 *outlenp = 0; 3424 return (EINVAL); 3425 } 3426 break; 3427 default: 3428 /* 3429 * We should never get here 3430 */ 3431 *outlenp = 0; 3432 return (EINVAL); 3433 } 3434 3435 ASSERT((optset_context != SETFN_OPTCOM_CHECKONLY) || 3436 (optset_context == SETFN_OPTCOM_CHECKONLY && inlen != 0)); 3437 3438 /* 3439 * For fixed length options, no sanity check 3440 * of passed in length is done. It is assumed *_optcom_req() 3441 * routines do the right thing. 3442 */ 3443 3444 switch (level) { 3445 case SOL_SOCKET: 3446 switch (name) { 3447 case SO_REUSEADDR: 3448 if (!checkonly) 3449 udp->udp_reuseaddr = onoff; 3450 break; 3451 case SO_DEBUG: 3452 if (!checkonly) 3453 udp->udp_debug = onoff; 3454 break; 3455 /* 3456 * The following three items are available here, 3457 * but are only meaningful to IP. 3458 */ 3459 case SO_DONTROUTE: 3460 if (!checkonly) 3461 udp->udp_dontroute = onoff; 3462 break; 3463 case SO_USELOOPBACK: 3464 if (!checkonly) 3465 udp->udp_useloopback = onoff; 3466 break; 3467 case SO_BROADCAST: 3468 if (!checkonly) 3469 udp->udp_broadcast = onoff; 3470 break; 3471 3472 case SO_SNDBUF: 3473 if (*i1 > udp_max_buf) { 3474 *outlenp = 0; 3475 return (ENOBUFS); 3476 } 3477 if (!checkonly) { 3478 q->q_hiwat = *i1; 3479 WR(UDP_RD(q))->q_hiwat = *i1; 3480 } 3481 break; 3482 case SO_RCVBUF: 3483 if (*i1 > udp_max_buf) { 3484 *outlenp = 0; 3485 return (ENOBUFS); 3486 } 3487 if (!checkonly) { 3488 RD(q)->q_hiwat = *i1; 3489 UDP_RD(q)->q_hiwat = *i1; 3490 (void) mi_set_sth_hiwat(UDP_RD(q), 3491 udp_set_rcv_hiwat(udp, *i1)); 3492 } 3493 break; 3494 case SO_DGRAM_ERRIND: 3495 if (!checkonly) 3496 udp->udp_dgram_errind = onoff; 3497 break; 3498 case SO_RECVUCRED: 3499 if (!checkonly) 3500 udp->udp_recvucred = onoff; 3501 break; 3502 case SO_ALLZONES: 3503 /* 3504 * "soft" error (negative) 3505 * option not handled at this level 3506 * Do not modify *outlenp. 3507 */ 3508 return (-EINVAL); 3509 case SO_TIMESTAMP: 3510 if (!checkonly) 3511 udp->udp_timestamp = onoff; 3512 break; 3513 case SO_ANON_MLP: 3514 if (!checkonly) 3515 udp->udp_anon_mlp = onoff; 3516 break; 3517 case SO_MAC_EXEMPT: 3518 if (secpolicy_net_mac_aware(cr) != 0 || 3519 udp->udp_state != TS_UNBND) 3520 return (EACCES); 3521 if (!checkonly) 3522 udp->udp_mac_exempt = onoff; 3523 break; 3524 case SCM_UCRED: { 3525 struct ucred_s *ucr; 3526 cred_t *cr, *newcr; 3527 ts_label_t *tsl; 3528 3529 /* 3530 * Only sockets that have proper privileges and are 3531 * bound to MLPs will have any other value here, so 3532 * this implicitly tests for privilege to set label. 3533 */ 3534 if (connp->conn_mlp_type == mlptSingle) 3535 break; 3536 ucr = (struct ucred_s *)invalp; 3537 if (inlen != ucredsize || 3538 ucr->uc_labeloff < sizeof (*ucr) || 3539 ucr->uc_labeloff + sizeof (bslabel_t) > inlen) 3540 return (EINVAL); 3541 if (!checkonly) { 3542 mblk_t *mb; 3543 3544 if (attrs == NULL || 3545 (mb = attrs->udpattr_mb) == NULL) 3546 return (EINVAL); 3547 if ((cr = DB_CRED(mb)) == NULL) 3548 cr = udp->udp_connp->conn_cred; 3549 ASSERT(cr != NULL); 3550 if ((tsl = crgetlabel(cr)) == NULL) 3551 return (EINVAL); 3552 newcr = copycred_from_bslabel(cr, UCLABEL(ucr), 3553 tsl->tsl_doi, KM_NOSLEEP); 3554 if (newcr == NULL) 3555 return (ENOSR); 3556 mblk_setcred(mb, newcr); 3557 attrs->udpattr_credset = B_TRUE; 3558 crfree(newcr); 3559 } 3560 break; 3561 } 3562 case SO_EXCLBIND: 3563 if (!checkonly) 3564 udp->udp_exclbind = onoff; 3565 break; 3566 default: 3567 *outlenp = 0; 3568 return (EINVAL); 3569 } 3570 break; 3571 case IPPROTO_IP: 3572 if (udp->udp_family != AF_INET) { 3573 *outlenp = 0; 3574 return (ENOPROTOOPT); 3575 } 3576 switch (name) { 3577 case IP_OPTIONS: 3578 case T_IP_OPTIONS: 3579 /* Save options for use by IP. */ 3580 newlen = inlen + udp->udp_label_len; 3581 if ((inlen & 0x3) || newlen > IP_MAX_OPT_LENGTH) { 3582 *outlenp = 0; 3583 return (EINVAL); 3584 } 3585 if (checkonly) 3586 break; 3587 3588 if (!tsol_option_set(&udp->udp_ip_snd_options, 3589 &udp->udp_ip_snd_options_len, 3590 udp->udp_label_len, invalp, inlen)) { 3591 *outlenp = 0; 3592 return (ENOMEM); 3593 } 3594 3595 udp->udp_max_hdr_len = IP_SIMPLE_HDR_LENGTH + 3596 UDPH_SIZE + udp->udp_ip_snd_options_len; 3597 (void) mi_set_sth_wroff(RD(q), udp->udp_max_hdr_len + 3598 udp_wroff_extra); 3599 break; 3600 3601 case IP_TTL: 3602 if (!checkonly) { 3603 udp->udp_ttl = (uchar_t)*i1; 3604 } 3605 break; 3606 case IP_TOS: 3607 case T_IP_TOS: 3608 if (!checkonly) { 3609 udp->udp_type_of_service = (uchar_t)*i1; 3610 } 3611 break; 3612 case IP_MULTICAST_IF: { 3613 /* 3614 * TODO should check OPTMGMT reply and undo this if 3615 * there is an error. 3616 */ 3617 struct in_addr *inap = (struct in_addr *)invalp; 3618 if (!checkonly) { 3619 udp->udp_multicast_if_addr = 3620 inap->s_addr; 3621 } 3622 break; 3623 } 3624 case IP_MULTICAST_TTL: 3625 if (!checkonly) 3626 udp->udp_multicast_ttl = *invalp; 3627 break; 3628 case IP_MULTICAST_LOOP: 3629 if (!checkonly) 3630 connp->conn_multicast_loop = *invalp; 3631 break; 3632 case IP_RECVOPTS: 3633 if (!checkonly) 3634 udp->udp_recvopts = onoff; 3635 break; 3636 case IP_RECVDSTADDR: 3637 if (!checkonly) 3638 udp->udp_recvdstaddr = onoff; 3639 break; 3640 case IP_RECVIF: 3641 if (!checkonly) 3642 udp->udp_recvif = onoff; 3643 break; 3644 case IP_RECVSLLA: 3645 if (!checkonly) 3646 udp->udp_recvslla = onoff; 3647 break; 3648 case IP_RECVTTL: 3649 if (!checkonly) 3650 udp->udp_recvttl = onoff; 3651 break; 3652 case IP_ADD_MEMBERSHIP: 3653 case IP_DROP_MEMBERSHIP: 3654 case IP_BLOCK_SOURCE: 3655 case IP_UNBLOCK_SOURCE: 3656 case IP_ADD_SOURCE_MEMBERSHIP: 3657 case IP_DROP_SOURCE_MEMBERSHIP: 3658 case MCAST_JOIN_GROUP: 3659 case MCAST_LEAVE_GROUP: 3660 case MCAST_BLOCK_SOURCE: 3661 case MCAST_UNBLOCK_SOURCE: 3662 case MCAST_JOIN_SOURCE_GROUP: 3663 case MCAST_LEAVE_SOURCE_GROUP: 3664 case IP_SEC_OPT: 3665 case IP_NEXTHOP: 3666 /* 3667 * "soft" error (negative) 3668 * option not handled at this level 3669 * Do not modify *outlenp. 3670 */ 3671 return (-EINVAL); 3672 case IP_BOUND_IF: 3673 if (!checkonly) 3674 udp->udp_bound_if = *i1; 3675 break; 3676 case IP_UNSPEC_SRC: 3677 if (!checkonly) 3678 udp->udp_unspec_source = onoff; 3679 break; 3680 case IP_XMIT_IF: 3681 if (!checkonly) 3682 udp->udp_xmit_if = *i1; 3683 break; 3684 default: 3685 *outlenp = 0; 3686 return (EINVAL); 3687 } 3688 break; 3689 case IPPROTO_IPV6: { 3690 ip6_pkt_t *ipp; 3691 boolean_t sticky; 3692 3693 if (udp->udp_family != AF_INET6) { 3694 *outlenp = 0; 3695 return (ENOPROTOOPT); 3696 } 3697 /* 3698 * Deal with both sticky options and ancillary data 3699 */ 3700 sticky = B_FALSE; 3701 if (attrs == NULL || (ipp = attrs->udpattr_ipp) == NULL) { 3702 /* sticky options, or none */ 3703 ipp = &udp->udp_sticky_ipp; 3704 sticky = B_TRUE; 3705 } 3706 3707 switch (name) { 3708 case IPV6_MULTICAST_IF: 3709 if (!checkonly) 3710 udp->udp_multicast_if_index = *i1; 3711 break; 3712 case IPV6_UNICAST_HOPS: 3713 /* -1 means use default */ 3714 if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) { 3715 *outlenp = 0; 3716 return (EINVAL); 3717 } 3718 if (!checkonly) { 3719 if (*i1 == -1) { 3720 udp->udp_ttl = ipp->ipp_unicast_hops = 3721 udp_ipv6_hoplimit; 3722 ipp->ipp_fields &= ~IPPF_UNICAST_HOPS; 3723 /* Pass modified value to IP. */ 3724 *i1 = udp->udp_ttl; 3725 } else { 3726 udp->udp_ttl = ipp->ipp_unicast_hops = 3727 (uint8_t)*i1; 3728 ipp->ipp_fields |= IPPF_UNICAST_HOPS; 3729 } 3730 /* Rebuild the header template */ 3731 error = udp_build_hdrs(q, udp); 3732 if (error != 0) { 3733 *outlenp = 0; 3734 return (error); 3735 } 3736 } 3737 break; 3738 case IPV6_MULTICAST_HOPS: 3739 /* -1 means use default */ 3740 if (*i1 < -1 || *i1 > IPV6_MAX_HOPS) { 3741 *outlenp = 0; 3742 return (EINVAL); 3743 } 3744 if (!checkonly) { 3745 if (*i1 == -1) { 3746 udp->udp_multicast_ttl = 3747 ipp->ipp_multicast_hops = 3748 IP_DEFAULT_MULTICAST_TTL; 3749 ipp->ipp_fields &= ~IPPF_MULTICAST_HOPS; 3750 /* Pass modified value to IP. */ 3751 *i1 = udp->udp_multicast_ttl; 3752 } else { 3753 udp->udp_multicast_ttl = 3754 ipp->ipp_multicast_hops = 3755 (uint8_t)*i1; 3756 ipp->ipp_fields |= IPPF_MULTICAST_HOPS; 3757 } 3758 } 3759 break; 3760 case IPV6_MULTICAST_LOOP: 3761 if (*i1 != 0 && *i1 != 1) { 3762 *outlenp = 0; 3763 return (EINVAL); 3764 } 3765 if (!checkonly) 3766 connp->conn_multicast_loop = *i1; 3767 break; 3768 case IPV6_JOIN_GROUP: 3769 case IPV6_LEAVE_GROUP: 3770 case MCAST_JOIN_GROUP: 3771 case MCAST_LEAVE_GROUP: 3772 case MCAST_BLOCK_SOURCE: 3773 case MCAST_UNBLOCK_SOURCE: 3774 case MCAST_JOIN_SOURCE_GROUP: 3775 case MCAST_LEAVE_SOURCE_GROUP: 3776 /* 3777 * "soft" error (negative) 3778 * option not handled at this level 3779 * Note: Do not modify *outlenp 3780 */ 3781 return (-EINVAL); 3782 case IPV6_BOUND_IF: 3783 if (!checkonly) 3784 udp->udp_bound_if = *i1; 3785 break; 3786 case IPV6_UNSPEC_SRC: 3787 if (!checkonly) 3788 udp->udp_unspec_source = onoff; 3789 break; 3790 /* 3791 * Set boolean switches for ancillary data delivery 3792 */ 3793 case IPV6_RECVPKTINFO: 3794 if (!checkonly) 3795 udp->udp_ipv6_recvpktinfo = onoff; 3796 break; 3797 case IPV6_RECVTCLASS: 3798 if (!checkonly) { 3799 udp->udp_ipv6_recvtclass = onoff; 3800 } 3801 break; 3802 case IPV6_RECVPATHMTU: 3803 if (!checkonly) { 3804 udp->udp_ipv6_recvpathmtu = onoff; 3805 } 3806 break; 3807 case IPV6_RECVHOPLIMIT: 3808 if (!checkonly) 3809 udp->udp_ipv6_recvhoplimit = onoff; 3810 break; 3811 case IPV6_RECVHOPOPTS: 3812 if (!checkonly) 3813 udp->udp_ipv6_recvhopopts = onoff; 3814 break; 3815 case IPV6_RECVDSTOPTS: 3816 if (!checkonly) 3817 udp->udp_ipv6_recvdstopts = onoff; 3818 break; 3819 case _OLD_IPV6_RECVDSTOPTS: 3820 if (!checkonly) 3821 udp->udp_old_ipv6_recvdstopts = onoff; 3822 break; 3823 case IPV6_RECVRTHDRDSTOPTS: 3824 if (!checkonly) 3825 udp->udp_ipv6_recvrthdrdstopts = onoff; 3826 break; 3827 case IPV6_RECVRTHDR: 3828 if (!checkonly) 3829 udp->udp_ipv6_recvrthdr = onoff; 3830 break; 3831 /* 3832 * Set sticky options or ancillary data. 3833 * If sticky options, (re)build any extension headers 3834 * that might be needed as a result. 3835 */ 3836 case IPV6_PKTINFO: 3837 /* 3838 * The source address and ifindex are verified 3839 * in ip_opt_set(). For ancillary data the 3840 * source address is checked in ip_wput_v6. 3841 */ 3842 if (inlen != 0 && inlen != sizeof (struct in6_pktinfo)) 3843 return (EINVAL); 3844 if (checkonly) 3845 break; 3846 3847 if (inlen == 0) { 3848 ipp->ipp_fields &= ~(IPPF_IFINDEX|IPPF_ADDR); 3849 ipp->ipp_sticky_ignored |= 3850 (IPPF_IFINDEX|IPPF_ADDR); 3851 } else { 3852 struct in6_pktinfo *pkti; 3853 3854 pkti = (struct in6_pktinfo *)invalp; 3855 ipp->ipp_ifindex = pkti->ipi6_ifindex; 3856 ipp->ipp_addr = pkti->ipi6_addr; 3857 if (ipp->ipp_ifindex != 0) 3858 ipp->ipp_fields |= IPPF_IFINDEX; 3859 else 3860 ipp->ipp_fields &= ~IPPF_IFINDEX; 3861 if (!IN6_IS_ADDR_UNSPECIFIED( 3862 &ipp->ipp_addr)) 3863 ipp->ipp_fields |= IPPF_ADDR; 3864 else 3865 ipp->ipp_fields &= ~IPPF_ADDR; 3866 } 3867 if (sticky) { 3868 error = udp_build_hdrs(q, udp); 3869 if (error != 0) 3870 return (error); 3871 } 3872 break; 3873 case IPV6_HOPLIMIT: 3874 if (sticky) 3875 return (EINVAL); 3876 if (inlen != 0 && inlen != sizeof (int)) 3877 return (EINVAL); 3878 if (checkonly) 3879 break; 3880 3881 if (inlen == 0) { 3882 ipp->ipp_fields &= ~IPPF_HOPLIMIT; 3883 ipp->ipp_sticky_ignored |= IPPF_HOPLIMIT; 3884 } else { 3885 if (*i1 > 255 || *i1 < -1) 3886 return (EINVAL); 3887 if (*i1 == -1) 3888 ipp->ipp_hoplimit = udp_ipv6_hoplimit; 3889 else 3890 ipp->ipp_hoplimit = *i1; 3891 ipp->ipp_fields |= IPPF_HOPLIMIT; 3892 } 3893 break; 3894 case IPV6_TCLASS: 3895 if (inlen != 0 && inlen != sizeof (int)) 3896 return (EINVAL); 3897 if (checkonly) 3898 break; 3899 3900 if (inlen == 0) { 3901 ipp->ipp_fields &= ~IPPF_TCLASS; 3902 ipp->ipp_sticky_ignored |= IPPF_TCLASS; 3903 } else { 3904 if (*i1 > 255 || *i1 < -1) 3905 return (EINVAL); 3906 if (*i1 == -1) 3907 ipp->ipp_tclass = 0; 3908 else 3909 ipp->ipp_tclass = *i1; 3910 ipp->ipp_fields |= IPPF_TCLASS; 3911 } 3912 if (sticky) { 3913 error = udp_build_hdrs(q, udp); 3914 if (error != 0) 3915 return (error); 3916 } 3917 break; 3918 case IPV6_NEXTHOP: 3919 /* 3920 * IP will verify that the nexthop is reachable 3921 * and fail for sticky options. 3922 */ 3923 if (inlen != 0 && inlen != sizeof (sin6_t)) 3924 return (EINVAL); 3925 if (checkonly) 3926 break; 3927 3928 if (inlen == 0) { 3929 ipp->ipp_fields &= ~IPPF_NEXTHOP; 3930 ipp->ipp_sticky_ignored |= IPPF_NEXTHOP; 3931 } else { 3932 sin6_t *sin6 = (sin6_t *)invalp; 3933 3934 if (sin6->sin6_family != AF_INET6) 3935 return (EAFNOSUPPORT); 3936 if (IN6_IS_ADDR_V4MAPPED( 3937 &sin6->sin6_addr)) 3938 return (EADDRNOTAVAIL); 3939 ipp->ipp_nexthop = sin6->sin6_addr; 3940 if (!IN6_IS_ADDR_UNSPECIFIED( 3941 &ipp->ipp_nexthop)) 3942 ipp->ipp_fields |= IPPF_NEXTHOP; 3943 else 3944 ipp->ipp_fields &= ~IPPF_NEXTHOP; 3945 } 3946 if (sticky) { 3947 error = udp_build_hdrs(q, udp); 3948 if (error != 0) 3949 return (error); 3950 } 3951 break; 3952 case IPV6_HOPOPTS: { 3953 ip6_hbh_t *hopts = (ip6_hbh_t *)invalp; 3954 /* 3955 * Sanity checks - minimum size, size a multiple of 3956 * eight bytes, and matching size passed in. 3957 */ 3958 if (inlen != 0 && 3959 inlen != (8 * (hopts->ip6h_len + 1))) 3960 return (EINVAL); 3961 3962 if (checkonly) 3963 break; 3964 3965 error = optcom_pkt_set(invalp, inlen, sticky, 3966 (uchar_t **)&ipp->ipp_hopopts, 3967 &ipp->ipp_hopoptslen, 3968 sticky ? udp->udp_label_len_v6 : 0); 3969 if (error != 0) 3970 return (error); 3971 if (ipp->ipp_hopoptslen == 0) { 3972 ipp->ipp_fields &= ~IPPF_HOPOPTS; 3973 ipp->ipp_sticky_ignored |= IPPF_HOPOPTS; 3974 } else { 3975 ipp->ipp_fields |= IPPF_HOPOPTS; 3976 } 3977 if (sticky) { 3978 error = udp_build_hdrs(q, udp); 3979 if (error != 0) 3980 return (error); 3981 } 3982 break; 3983 } 3984 case IPV6_RTHDRDSTOPTS: { 3985 ip6_dest_t *dopts = (ip6_dest_t *)invalp; 3986 3987 /* 3988 * Sanity checks - minimum size, size a multiple of 3989 * eight bytes, and matching size passed in. 3990 */ 3991 if (inlen != 0 && 3992 inlen != (8 * (dopts->ip6d_len + 1))) 3993 return (EINVAL); 3994 3995 if (checkonly) 3996 break; 3997 3998 if (inlen == 0) { 3999 if (sticky && 4000 (ipp->ipp_fields & IPPF_RTDSTOPTS) != 0) { 4001 kmem_free(ipp->ipp_rtdstopts, 4002 ipp->ipp_rtdstoptslen); 4003 ipp->ipp_rtdstopts = NULL; 4004 ipp->ipp_rtdstoptslen = 0; 4005 } 4006 4007 ipp->ipp_fields &= ~IPPF_RTDSTOPTS; 4008 ipp->ipp_sticky_ignored |= IPPF_RTDSTOPTS; 4009 } else { 4010 error = optcom_pkt_set(invalp, inlen, sticky, 4011 (uchar_t **)&ipp->ipp_rtdstopts, 4012 &ipp->ipp_rtdstoptslen, 0); 4013 if (error != 0) 4014 return (error); 4015 ipp->ipp_fields |= IPPF_RTDSTOPTS; 4016 } 4017 if (sticky) { 4018 error = udp_build_hdrs(q, udp); 4019 if (error != 0) 4020 return (error); 4021 } 4022 break; 4023 } 4024 case IPV6_DSTOPTS: { 4025 ip6_dest_t *dopts = (ip6_dest_t *)invalp; 4026 4027 /* 4028 * Sanity checks - minimum size, size a multiple of 4029 * eight bytes, and matching size passed in. 4030 */ 4031 if (inlen != 0 && 4032 inlen != (8 * (dopts->ip6d_len + 1))) 4033 return (EINVAL); 4034 4035 if (checkonly) 4036 break; 4037 4038 if (inlen == 0) { 4039 if (sticky && 4040 (ipp->ipp_fields & IPPF_DSTOPTS) != 0) { 4041 kmem_free(ipp->ipp_dstopts, 4042 ipp->ipp_dstoptslen); 4043 ipp->ipp_dstopts = NULL; 4044 ipp->ipp_dstoptslen = 0; 4045 } 4046 ipp->ipp_fields &= ~IPPF_DSTOPTS; 4047 ipp->ipp_sticky_ignored |= IPPF_DSTOPTS; 4048 } else { 4049 error = optcom_pkt_set(invalp, inlen, sticky, 4050 (uchar_t **)&ipp->ipp_dstopts, 4051 &ipp->ipp_dstoptslen, 0); 4052 if (error != 0) 4053 return (error); 4054 ipp->ipp_fields |= IPPF_DSTOPTS; 4055 } 4056 if (sticky) { 4057 error = udp_build_hdrs(q, udp); 4058 if (error != 0) 4059 return (error); 4060 } 4061 break; 4062 } 4063 case IPV6_RTHDR: { 4064 ip6_rthdr_t *rt = (ip6_rthdr_t *)invalp; 4065 4066 /* 4067 * Sanity checks - minimum size, size a multiple of 4068 * eight bytes, and matching size passed in. 4069 */ 4070 if (inlen != 0 && 4071 inlen != (8 * (rt->ip6r_len + 1))) 4072 return (EINVAL); 4073 4074 if (checkonly) 4075 break; 4076 4077 if (inlen == 0) { 4078 if (sticky && 4079 (ipp->ipp_fields & IPPF_RTHDR) != 0) { 4080 kmem_free(ipp->ipp_rthdr, 4081 ipp->ipp_rthdrlen); 4082 ipp->ipp_rthdr = NULL; 4083 ipp->ipp_rthdrlen = 0; 4084 } 4085 ipp->ipp_fields &= ~IPPF_RTHDR; 4086 ipp->ipp_sticky_ignored |= IPPF_RTHDR; 4087 } else { 4088 error = optcom_pkt_set(invalp, inlen, sticky, 4089 (uchar_t **)&ipp->ipp_rthdr, 4090 &ipp->ipp_rthdrlen, 0); 4091 if (error != 0) 4092 return (error); 4093 ipp->ipp_fields |= IPPF_RTHDR; 4094 } 4095 if (sticky) { 4096 error = udp_build_hdrs(q, udp); 4097 if (error != 0) 4098 return (error); 4099 } 4100 break; 4101 } 4102 4103 case IPV6_DONTFRAG: 4104 if (checkonly) 4105 break; 4106 4107 if (onoff) { 4108 ipp->ipp_fields |= IPPF_DONTFRAG; 4109 } else { 4110 ipp->ipp_fields &= ~IPPF_DONTFRAG; 4111 } 4112 break; 4113 4114 case IPV6_USE_MIN_MTU: 4115 if (inlen != sizeof (int)) 4116 return (EINVAL); 4117 4118 if (*i1 < -1 || *i1 > 1) 4119 return (EINVAL); 4120 4121 if (checkonly) 4122 break; 4123 4124 ipp->ipp_fields |= IPPF_USE_MIN_MTU; 4125 ipp->ipp_use_min_mtu = *i1; 4126 break; 4127 4128 case IPV6_BOUND_PIF: 4129 case IPV6_SEC_OPT: 4130 case IPV6_DONTFAILOVER_IF: 4131 case IPV6_SRC_PREFERENCES: 4132 case IPV6_V6ONLY: 4133 /* Handled at the IP level */ 4134 return (-EINVAL); 4135 default: 4136 *outlenp = 0; 4137 return (EINVAL); 4138 } 4139 break; 4140 } /* end IPPROTO_IPV6 */ 4141 case IPPROTO_UDP: 4142 switch (name) { 4143 case UDP_ANONPRIVBIND: 4144 if ((error = secpolicy_net_privaddr(cr, 0)) != 0) { 4145 *outlenp = 0; 4146 return (error); 4147 } 4148 if (!checkonly) { 4149 udp->udp_anon_priv_bind = onoff; 4150 } 4151 break; 4152 case UDP_EXCLBIND: 4153 if (!checkonly) 4154 udp->udp_exclbind = onoff; 4155 break; 4156 case UDP_RCVHDR: 4157 if (!checkonly) 4158 udp->udp_rcvhdr = onoff; 4159 break; 4160 default: 4161 *outlenp = 0; 4162 return (EINVAL); 4163 } 4164 break; 4165 default: 4166 *outlenp = 0; 4167 return (EINVAL); 4168 } 4169 /* 4170 * Common case of OK return with outval same as inval. 4171 */ 4172 if (invalp != outvalp) { 4173 /* don't trust bcopy for identical src/dst */ 4174 (void) bcopy(invalp, outvalp, inlen); 4175 } 4176 *outlenp = inlen; 4177 return (0); 4178 } 4179 4180 /* 4181 * Update udp_sticky_hdrs based on udp_sticky_ipp, udp_v6src, and udp_ttl. 4182 * The headers include ip6i_t (if needed), ip6_t, any sticky extension 4183 * headers, and the udp header. 4184 * Returns failure if can't allocate memory. 4185 */ 4186 static int 4187 udp_build_hdrs(queue_t *q, udp_t *udp) 4188 { 4189 uchar_t *hdrs; 4190 uint_t hdrs_len; 4191 ip6_t *ip6h; 4192 ip6i_t *ip6i; 4193 udpha_t *udpha; 4194 ip6_pkt_t *ipp = &udp->udp_sticky_ipp; 4195 4196 hdrs_len = ip_total_hdrs_len_v6(ipp) + UDPH_SIZE; 4197 ASSERT(hdrs_len != 0); 4198 if (hdrs_len != udp->udp_sticky_hdrs_len) { 4199 /* Need to reallocate */ 4200 hdrs = kmem_alloc(hdrs_len, KM_NOSLEEP); 4201 if (hdrs == NULL) 4202 return (ENOMEM); 4203 4204 if (udp->udp_sticky_hdrs_len != 0) { 4205 kmem_free(udp->udp_sticky_hdrs, 4206 udp->udp_sticky_hdrs_len); 4207 } 4208 udp->udp_sticky_hdrs = hdrs; 4209 udp->udp_sticky_hdrs_len = hdrs_len; 4210 } 4211 ip_build_hdrs_v6(udp->udp_sticky_hdrs, 4212 udp->udp_sticky_hdrs_len - UDPH_SIZE, ipp, IPPROTO_UDP); 4213 4214 /* Set header fields not in ipp */ 4215 if (ipp->ipp_fields & IPPF_HAS_IP6I) { 4216 ip6i = (ip6i_t *)udp->udp_sticky_hdrs; 4217 ip6h = (ip6_t *)&ip6i[1]; 4218 } else { 4219 ip6h = (ip6_t *)udp->udp_sticky_hdrs; 4220 } 4221 4222 if (!(ipp->ipp_fields & IPPF_ADDR)) 4223 ip6h->ip6_src = udp->udp_v6src; 4224 4225 udpha = (udpha_t *)(udp->udp_sticky_hdrs + hdrs_len - UDPH_SIZE); 4226 udpha->uha_src_port = udp->udp_port; 4227 4228 /* Try to get everything in a single mblk */ 4229 if (hdrs_len > udp->udp_max_hdr_len) { 4230 udp->udp_max_hdr_len = hdrs_len; 4231 (void) mi_set_sth_wroff(RD(q), udp->udp_max_hdr_len + 4232 udp_wroff_extra); 4233 } 4234 return (0); 4235 } 4236 4237 /* 4238 * This routine retrieves the value of an ND variable in a udpparam_t 4239 * structure. It is called through nd_getset when a user reads the 4240 * variable. 4241 */ 4242 /* ARGSUSED */ 4243 static int 4244 udp_param_get(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr) 4245 { 4246 udpparam_t *udppa = (udpparam_t *)cp; 4247 4248 (void) mi_mpprintf(mp, "%d", udppa->udp_param_value); 4249 return (0); 4250 } 4251 4252 /* 4253 * Walk through the param array specified registering each element with the 4254 * named dispatch (ND) handler. 4255 */ 4256 static boolean_t 4257 udp_param_register(udpparam_t *udppa, int cnt) 4258 { 4259 for (; cnt-- > 0; udppa++) { 4260 if (udppa->udp_param_name && udppa->udp_param_name[0]) { 4261 if (!nd_load(&udp_g_nd, udppa->udp_param_name, 4262 udp_param_get, udp_param_set, 4263 (caddr_t)udppa)) { 4264 nd_free(&udp_g_nd); 4265 return (B_FALSE); 4266 } 4267 } 4268 } 4269 if (!nd_load(&udp_g_nd, "udp_extra_priv_ports", 4270 udp_extra_priv_ports_get, NULL, NULL)) { 4271 nd_free(&udp_g_nd); 4272 return (B_FALSE); 4273 } 4274 if (!nd_load(&udp_g_nd, "udp_extra_priv_ports_add", 4275 NULL, udp_extra_priv_ports_add, NULL)) { 4276 nd_free(&udp_g_nd); 4277 return (B_FALSE); 4278 } 4279 if (!nd_load(&udp_g_nd, "udp_extra_priv_ports_del", 4280 NULL, udp_extra_priv_ports_del, NULL)) { 4281 nd_free(&udp_g_nd); 4282 return (B_FALSE); 4283 } 4284 if (!nd_load(&udp_g_nd, "udp_status", udp_status_report, NULL, 4285 NULL)) { 4286 nd_free(&udp_g_nd); 4287 return (B_FALSE); 4288 } 4289 if (!nd_load(&udp_g_nd, "udp_bind_hash", udp_bind_hash_report, NULL, 4290 NULL)) { 4291 nd_free(&udp_g_nd); 4292 return (B_FALSE); 4293 } 4294 return (B_TRUE); 4295 } 4296 4297 /* This routine sets an ND variable in a udpparam_t structure. */ 4298 /* ARGSUSED */ 4299 static int 4300 udp_param_set(queue_t *q, mblk_t *mp, char *value, caddr_t cp, cred_t *cr) 4301 { 4302 long new_value; 4303 udpparam_t *udppa = (udpparam_t *)cp; 4304 4305 /* 4306 * Fail the request if the new value does not lie within the 4307 * required bounds. 4308 */ 4309 if (ddi_strtol(value, NULL, 10, &new_value) != 0 || 4310 new_value < udppa->udp_param_min || 4311 new_value > udppa->udp_param_max) { 4312 return (EINVAL); 4313 } 4314 4315 /* Set the new value */ 4316 udppa->udp_param_value = new_value; 4317 return (0); 4318 } 4319 4320 /* 4321 * Copy hop-by-hop option from ipp->ipp_hopopts to the buffer provided (with 4322 * T_opthdr) and return the number of bytes copied. 'dbuf' may be NULL to 4323 * just count the length needed for allocation. If 'dbuf' is non-NULL, 4324 * then it's assumed to be allocated to be large enough. 4325 * 4326 * Returns zero if trimming of the security option causes all options to go 4327 * away. 4328 */ 4329 static size_t 4330 copy_hop_opts(const ip6_pkt_t *ipp, uchar_t *dbuf) 4331 { 4332 struct T_opthdr *toh; 4333 size_t hol = ipp->ipp_hopoptslen; 4334 ip6_hbh_t *dstopt = NULL; 4335 const ip6_hbh_t *srcopt = ipp->ipp_hopopts; 4336 size_t tlen, olen, plen; 4337 boolean_t deleting; 4338 const struct ip6_opt *sopt, *lastpad; 4339 struct ip6_opt *dopt; 4340 4341 if ((toh = (struct T_opthdr *)dbuf) != NULL) { 4342 toh->level = IPPROTO_IPV6; 4343 toh->name = IPV6_HOPOPTS; 4344 toh->status = 0; 4345 dstopt = (ip6_hbh_t *)(toh + 1); 4346 } 4347 4348 /* 4349 * If labeling is enabled, then skip the label option 4350 * but get other options if there are any. 4351 */ 4352 if (is_system_labeled()) { 4353 dopt = NULL; 4354 if (dstopt != NULL) { 4355 /* will fill in ip6h_len later */ 4356 dstopt->ip6h_nxt = srcopt->ip6h_nxt; 4357 dopt = (struct ip6_opt *)(dstopt + 1); 4358 } 4359 sopt = (const struct ip6_opt *)(srcopt + 1); 4360 hol -= sizeof (*srcopt); 4361 tlen = sizeof (*dstopt); 4362 lastpad = NULL; 4363 deleting = B_FALSE; 4364 /* 4365 * This loop finds the first (lastpad pointer) of any number of 4366 * pads that preceeds the security option, then treats the 4367 * security option as though it were a pad, and then finds the 4368 * next non-pad option (or end of list). 4369 * 4370 * It then treats the entire block as one big pad. To preserve 4371 * alignment of any options that follow, or just the end of the 4372 * list, it computes a minimal new padding size that keeps the 4373 * same alignment for the next option. 4374 * 4375 * If it encounters just a sequence of pads with no security 4376 * option, those are copied as-is rather than collapsed. 4377 * 4378 * Note that to handle the end of list case, the code makes one 4379 * loop with 'hol' set to zero. 4380 */ 4381 for (;;) { 4382 if (hol > 0) { 4383 if (sopt->ip6o_type == IP6OPT_PAD1) { 4384 if (lastpad == NULL) 4385 lastpad = sopt; 4386 sopt = (const struct ip6_opt *) 4387 &sopt->ip6o_len; 4388 hol--; 4389 continue; 4390 } 4391 olen = sopt->ip6o_len + sizeof (*sopt); 4392 if (olen > hol) 4393 olen = hol; 4394 if (sopt->ip6o_type == IP6OPT_PADN || 4395 sopt->ip6o_type == ip6opt_ls) { 4396 if (sopt->ip6o_type == ip6opt_ls) 4397 deleting = B_TRUE; 4398 if (lastpad == NULL) 4399 lastpad = sopt; 4400 sopt = (const struct ip6_opt *) 4401 ((const char *)sopt + olen); 4402 hol -= olen; 4403 continue; 4404 } 4405 } else { 4406 /* if nothing was copied at all, then delete */ 4407 if (tlen == sizeof (*dstopt)) 4408 return (0); 4409 /* last pass; pick up any trailing padding */ 4410 olen = 0; 4411 } 4412 if (deleting) { 4413 /* 4414 * compute aligning effect of deleted material 4415 * to reproduce with pad. 4416 */ 4417 plen = ((const char *)sopt - 4418 (const char *)lastpad) & 7; 4419 tlen += plen; 4420 if (dopt != NULL) { 4421 if (plen == 1) { 4422 dopt->ip6o_type = IP6OPT_PAD1; 4423 } else if (plen > 1) { 4424 plen -= sizeof (*dopt); 4425 dopt->ip6o_type = IP6OPT_PADN; 4426 dopt->ip6o_len = plen; 4427 if (plen > 0) 4428 bzero(dopt + 1, plen); 4429 } 4430 dopt = (struct ip6_opt *) 4431 ((char *)dopt + plen); 4432 } 4433 deleting = B_FALSE; 4434 lastpad = NULL; 4435 } 4436 /* if there's uncopied padding, then copy that now */ 4437 if (lastpad != NULL) { 4438 olen += (const char *)sopt - 4439 (const char *)lastpad; 4440 sopt = lastpad; 4441 lastpad = NULL; 4442 } 4443 if (dopt != NULL && olen > 0) { 4444 bcopy(sopt, dopt, olen); 4445 dopt = (struct ip6_opt *)((char *)dopt + olen); 4446 } 4447 if (hol == 0) 4448 break; 4449 tlen += olen; 4450 sopt = (const struct ip6_opt *) 4451 ((const char *)sopt + olen); 4452 hol -= olen; 4453 } 4454 /* go back and patch up the length value, rounded upward */ 4455 if (dstopt != NULL) 4456 dstopt->ip6h_len = (tlen - 1) >> 3; 4457 } else { 4458 tlen = hol; 4459 if (dstopt != NULL) 4460 bcopy(srcopt, dstopt, hol); 4461 } 4462 4463 tlen += sizeof (*toh); 4464 if (toh != NULL) 4465 toh->len = tlen; 4466 4467 return (tlen); 4468 } 4469 4470 static void 4471 udp_input(conn_t *connp, mblk_t *mp) 4472 { 4473 struct T_unitdata_ind *tudi; 4474 uchar_t *rptr; /* Pointer to IP header */ 4475 int hdr_length; /* Length of IP+UDP headers */ 4476 int udi_size; /* Size of T_unitdata_ind */ 4477 int mp_len; 4478 udp_t *udp; 4479 udpha_t *udpha; 4480 int ipversion; 4481 ip6_pkt_t ipp; 4482 ip6_t *ip6h; 4483 ip6i_t *ip6i; 4484 mblk_t *mp1; 4485 mblk_t *options_mp = NULL; 4486 in_pktinfo_t *pinfo = NULL; 4487 cred_t *cr = NULL; 4488 queue_t *q = connp->conn_rq; 4489 pid_t cpid; 4490 cred_t *rcr = connp->conn_cred; 4491 4492 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_START, 4493 "udp_rput_start: q %p mp %p", q, mp); 4494 4495 udp = connp->conn_udp; 4496 rptr = mp->b_rptr; 4497 ASSERT(DB_TYPE(mp) == M_DATA || DB_TYPE(mp) == M_CTL); 4498 ASSERT(OK_32PTR(rptr)); 4499 4500 /* 4501 * IP should have prepended the options data in an M_CTL 4502 * Check M_CTL "type" to make sure are not here bcos of 4503 * a valid ICMP message 4504 */ 4505 if (DB_TYPE(mp) == M_CTL) { 4506 if (MBLKL(mp) == sizeof (in_pktinfo_t) && 4507 ((in_pktinfo_t *)mp->b_rptr)->in_pkt_ulp_type == 4508 IN_PKTINFO) { 4509 /* 4510 * IP_RECVIF or IP_RECVSLLA information has been 4511 * appended to the packet by IP. We need to 4512 * extract the mblk and adjust the rptr 4513 */ 4514 pinfo = (in_pktinfo_t *)mp->b_rptr; 4515 options_mp = mp; 4516 mp = mp->b_cont; 4517 rptr = mp->b_rptr; 4518 UDP_STAT(udp_in_pktinfo); 4519 } else { 4520 /* 4521 * ICMP messages. 4522 */ 4523 udp_icmp_error(q, mp); 4524 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 4525 "udp_rput_end: q %p (%S)", q, "m_ctl"); 4526 return; 4527 } 4528 } 4529 4530 mp_len = msgdsize(mp); 4531 /* 4532 * This is the inbound data path. 4533 * First, we check to make sure the IP version number is correct, 4534 * and then pull the IP and UDP headers into the first mblk. 4535 * Assume IP provides aligned packets - otherwise toss. 4536 * Also, check if we have a complete IP header. 4537 */ 4538 4539 /* Initialize regardless if ipversion is IPv4 or IPv6 */ 4540 ipp.ipp_fields = 0; 4541 4542 ipversion = IPH_HDR_VERSION(rptr); 4543 switch (ipversion) { 4544 case IPV4_VERSION: 4545 ASSERT(MBLKL(mp) >= sizeof (ipha_t)); 4546 ASSERT(((ipha_t *)rptr)->ipha_protocol == IPPROTO_UDP); 4547 hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE; 4548 if ((hdr_length > IP_SIMPLE_HDR_LENGTH + UDPH_SIZE) || 4549 (udp->udp_ip_rcv_options_len)) { 4550 /* 4551 * Handle IPv4 packets with options outside of the 4552 * main data path. Not needed for AF_INET6 sockets 4553 * since they don't support a getsockopt of IP_OPTIONS. 4554 */ 4555 if (udp->udp_family == AF_INET6) 4556 break; 4557 /* 4558 * UDP length check performed for IPv4 packets with 4559 * options to check whether UDP length specified in 4560 * the header is the same as the physical length of 4561 * the packet. 4562 */ 4563 udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE)); 4564 if (mp_len != (ntohs(udpha->uha_length) + 4565 hdr_length - UDPH_SIZE)) { 4566 goto tossit; 4567 } 4568 /* 4569 * Handle the case where the packet has IP options 4570 * and the IP_RECVSLLA & IP_RECVIF are set 4571 */ 4572 if (pinfo != NULL) 4573 mp = options_mp; 4574 udp_become_writer(connp, mp, udp_rput_other_wrapper, 4575 SQTAG_UDP_INPUT); 4576 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 4577 "udp_rput_end: q %p (%S)", q, "end"); 4578 return; 4579 } 4580 4581 /* Handle IPV6_RECVHOPLIMIT. */ 4582 if ((udp->udp_family == AF_INET6) && (pinfo != NULL) && 4583 udp->udp_ipv6_recvpktinfo) { 4584 if (pinfo->in_pkt_flags & IPF_RECVIF) { 4585 ipp.ipp_fields |= IPPF_IFINDEX; 4586 ipp.ipp_ifindex = pinfo->in_pkt_ifindex; 4587 } 4588 } 4589 break; 4590 case IPV6_VERSION: 4591 /* 4592 * IPv6 packets can only be received by applications 4593 * that are prepared to receive IPv6 addresses. 4594 * The IP fanout must ensure this. 4595 */ 4596 ASSERT(udp->udp_family == AF_INET6); 4597 4598 ip6h = (ip6_t *)rptr; 4599 ASSERT((uchar_t *)&ip6h[1] <= mp->b_wptr); 4600 4601 if (ip6h->ip6_nxt != IPPROTO_UDP) { 4602 uint8_t nexthdrp; 4603 /* Look for ifindex information */ 4604 if (ip6h->ip6_nxt == IPPROTO_RAW) { 4605 ip6i = (ip6i_t *)ip6h; 4606 if ((uchar_t *)&ip6i[1] > mp->b_wptr) 4607 goto tossit; 4608 4609 if (ip6i->ip6i_flags & IP6I_IFINDEX) { 4610 ASSERT(ip6i->ip6i_ifindex != 0); 4611 ipp.ipp_fields |= IPPF_IFINDEX; 4612 ipp.ipp_ifindex = ip6i->ip6i_ifindex; 4613 } 4614 rptr = (uchar_t *)&ip6i[1]; 4615 mp->b_rptr = rptr; 4616 if (rptr == mp->b_wptr) { 4617 mp1 = mp->b_cont; 4618 freeb(mp); 4619 mp = mp1; 4620 rptr = mp->b_rptr; 4621 } 4622 if (MBLKL(mp) < (IPV6_HDR_LEN + UDPH_SIZE)) 4623 goto tossit; 4624 ip6h = (ip6_t *)rptr; 4625 mp_len = msgdsize(mp); 4626 } 4627 /* 4628 * Find any potentially interesting extension headers 4629 * as well as the length of the IPv6 + extension 4630 * headers. 4631 */ 4632 hdr_length = ip_find_hdr_v6(mp, ip6h, &ipp, &nexthdrp) + 4633 UDPH_SIZE; 4634 ASSERT(nexthdrp == IPPROTO_UDP); 4635 } else { 4636 hdr_length = IPV6_HDR_LEN + UDPH_SIZE; 4637 ip6i = NULL; 4638 } 4639 break; 4640 default: 4641 ASSERT(0); 4642 } 4643 4644 /* 4645 * IP inspected the UDP header thus all of it must be in the mblk. 4646 * UDP length check is performed for IPv6 packets and IPv4 packets 4647 * without options to check if the size of the packet as specified 4648 * by the header is the same as the physical size of the packet. 4649 */ 4650 udpha = (udpha_t *)(rptr + (hdr_length - UDPH_SIZE)); 4651 if ((MBLKL(mp) < hdr_length) || 4652 (mp_len != (ntohs(udpha->uha_length) + hdr_length - UDPH_SIZE))) { 4653 goto tossit; 4654 } 4655 4656 /* Walk past the headers. */ 4657 if (!udp->udp_rcvhdr) { 4658 mp->b_rptr = rptr + hdr_length; 4659 mp_len -= hdr_length; 4660 } 4661 4662 /* 4663 * This is the inbound data path. Packets are passed upstream as 4664 * T_UNITDATA_IND messages with full IP headers still attached. 4665 */ 4666 if (udp->udp_family == AF_INET) { 4667 sin_t *sin; 4668 4669 ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION); 4670 4671 /* 4672 * Normally only send up the address. 4673 * If IP_RECVDSTADDR is set we include the destination IP 4674 * address as an option. With IP_RECVOPTS we include all 4675 * the IP options. Only ip_rput_other() handles packets 4676 * that contain IP options. 4677 */ 4678 udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t); 4679 if (udp->udp_recvdstaddr) { 4680 udi_size += sizeof (struct T_opthdr) + 4681 sizeof (struct in_addr); 4682 UDP_STAT(udp_in_recvdstaddr); 4683 } 4684 4685 /* 4686 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate 4687 * space accordingly 4688 */ 4689 if (udp->udp_recvif && (pinfo != NULL) && 4690 (pinfo->in_pkt_flags & IPF_RECVIF)) { 4691 udi_size += sizeof (struct T_opthdr) + sizeof (uint_t); 4692 UDP_STAT(udp_in_recvif); 4693 } 4694 4695 if (udp->udp_recvslla && (pinfo != NULL) && 4696 (pinfo->in_pkt_flags & IPF_RECVSLLA)) { 4697 udi_size += sizeof (struct T_opthdr) + 4698 sizeof (struct sockaddr_dl); 4699 UDP_STAT(udp_in_recvslla); 4700 } 4701 4702 if (udp->udp_recvucred && (cr = DB_CRED(mp)) != NULL) { 4703 udi_size += sizeof (struct T_opthdr) + ucredsize; 4704 cpid = DB_CPID(mp); 4705 UDP_STAT(udp_in_recvucred); 4706 } 4707 4708 /* 4709 * If SO_TIMESTAMP is set allocate the appropriate sized 4710 * buffer. Since gethrestime() expects a pointer aligned 4711 * argument, we allocate space necessary for extra 4712 * alignment (even though it might not be used). 4713 */ 4714 if (udp->udp_timestamp) { 4715 udi_size += sizeof (struct T_opthdr) + 4716 sizeof (timestruc_t) + _POINTER_ALIGNMENT; 4717 UDP_STAT(udp_in_timestamp); 4718 } 4719 4720 /* 4721 * If IP_RECVTTL is set allocate the appropriate sized buffer 4722 */ 4723 if (udp->udp_recvttl) { 4724 udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t); 4725 UDP_STAT(udp_in_recvttl); 4726 } 4727 ASSERT(IPH_HDR_LENGTH((ipha_t *)rptr) == IP_SIMPLE_HDR_LENGTH); 4728 4729 /* Allocate a message block for the T_UNITDATA_IND structure. */ 4730 mp1 = allocb(udi_size, BPRI_MED); 4731 if (mp1 == NULL) { 4732 freemsg(mp); 4733 if (options_mp != NULL) 4734 freeb(options_mp); 4735 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 4736 "udp_rput_end: q %p (%S)", q, "allocbfail"); 4737 BUMP_MIB(&udp_mib, udpInErrors); 4738 return; 4739 } 4740 mp1->b_cont = mp; 4741 mp = mp1; 4742 mp->b_datap->db_type = M_PROTO; 4743 tudi = (struct T_unitdata_ind *)mp->b_rptr; 4744 mp->b_wptr = (uchar_t *)tudi + udi_size; 4745 tudi->PRIM_type = T_UNITDATA_IND; 4746 tudi->SRC_length = sizeof (sin_t); 4747 tudi->SRC_offset = sizeof (struct T_unitdata_ind); 4748 tudi->OPT_offset = sizeof (struct T_unitdata_ind) + 4749 sizeof (sin_t); 4750 udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t)); 4751 tudi->OPT_length = udi_size; 4752 sin = (sin_t *)&tudi[1]; 4753 sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src; 4754 sin->sin_port = udpha->uha_src_port; 4755 sin->sin_family = udp->udp_family; 4756 *(uint32_t *)&sin->sin_zero[0] = 0; 4757 *(uint32_t *)&sin->sin_zero[4] = 0; 4758 4759 /* 4760 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or 4761 * IP_RECVTTL has been set. 4762 */ 4763 if (udi_size != 0) { 4764 /* 4765 * Copy in destination address before options to avoid 4766 * any padding issues. 4767 */ 4768 char *dstopt; 4769 4770 dstopt = (char *)&sin[1]; 4771 if (udp->udp_recvdstaddr) { 4772 struct T_opthdr *toh; 4773 ipaddr_t *dstptr; 4774 4775 toh = (struct T_opthdr *)dstopt; 4776 toh->level = IPPROTO_IP; 4777 toh->name = IP_RECVDSTADDR; 4778 toh->len = sizeof (struct T_opthdr) + 4779 sizeof (ipaddr_t); 4780 toh->status = 0; 4781 dstopt += sizeof (struct T_opthdr); 4782 dstptr = (ipaddr_t *)dstopt; 4783 *dstptr = ((ipha_t *)rptr)->ipha_dst; 4784 dstopt = (char *)toh + toh->len; 4785 udi_size -= toh->len; 4786 } 4787 4788 if (udp->udp_recvslla && (pinfo != NULL) && 4789 (pinfo->in_pkt_flags & IPF_RECVSLLA)) { 4790 4791 struct T_opthdr *toh; 4792 struct sockaddr_dl *dstptr; 4793 4794 toh = (struct T_opthdr *)dstopt; 4795 toh->level = IPPROTO_IP; 4796 toh->name = IP_RECVSLLA; 4797 toh->len = sizeof (struct T_opthdr) + 4798 sizeof (struct sockaddr_dl); 4799 toh->status = 0; 4800 dstopt += sizeof (struct T_opthdr); 4801 dstptr = (struct sockaddr_dl *)dstopt; 4802 bcopy(&pinfo->in_pkt_slla, dstptr, 4803 sizeof (struct sockaddr_dl)); 4804 dstopt = (char *)toh + toh->len; 4805 udi_size -= toh->len; 4806 } 4807 4808 if (udp->udp_recvif && (pinfo != NULL) && 4809 (pinfo->in_pkt_flags & IPF_RECVIF)) { 4810 4811 struct T_opthdr *toh; 4812 uint_t *dstptr; 4813 4814 toh = (struct T_opthdr *)dstopt; 4815 toh->level = IPPROTO_IP; 4816 toh->name = IP_RECVIF; 4817 toh->len = sizeof (struct T_opthdr) + 4818 sizeof (uint_t); 4819 toh->status = 0; 4820 dstopt += sizeof (struct T_opthdr); 4821 dstptr = (uint_t *)dstopt; 4822 *dstptr = pinfo->in_pkt_ifindex; 4823 dstopt = (char *)toh + toh->len; 4824 udi_size -= toh->len; 4825 } 4826 4827 if (cr != NULL) { 4828 struct T_opthdr *toh; 4829 4830 toh = (struct T_opthdr *)dstopt; 4831 toh->level = SOL_SOCKET; 4832 toh->name = SCM_UCRED; 4833 toh->len = sizeof (struct T_opthdr) + ucredsize; 4834 toh->status = 0; 4835 (void) cred2ucred(cr, cpid, &toh[1], rcr); 4836 dstopt = (char *)toh + toh->len; 4837 udi_size -= toh->len; 4838 } 4839 4840 if (udp->udp_timestamp) { 4841 struct T_opthdr *toh; 4842 4843 toh = (struct T_opthdr *)dstopt; 4844 toh->level = SOL_SOCKET; 4845 toh->name = SCM_TIMESTAMP; 4846 toh->len = sizeof (struct T_opthdr) + 4847 sizeof (timestruc_t) + _POINTER_ALIGNMENT; 4848 toh->status = 0; 4849 dstopt += sizeof (struct T_opthdr); 4850 /* Align for gethrestime() */ 4851 dstopt = (char *)P2ROUNDUP((intptr_t)dstopt, 4852 sizeof (intptr_t)); 4853 gethrestime((timestruc_t *)dstopt); 4854 dstopt = (char *)toh + toh->len; 4855 udi_size -= toh->len; 4856 } 4857 4858 /* 4859 * CAUTION: 4860 * Due to aligment issues 4861 * Processing of IP_RECVTTL option 4862 * should always be the last. Adding 4863 * any option processing after this will 4864 * cause alignment panic. 4865 */ 4866 if (udp->udp_recvttl) { 4867 struct T_opthdr *toh; 4868 uint8_t *dstptr; 4869 4870 toh = (struct T_opthdr *)dstopt; 4871 toh->level = IPPROTO_IP; 4872 toh->name = IP_RECVTTL; 4873 toh->len = sizeof (struct T_opthdr) + 4874 sizeof (uint8_t); 4875 toh->status = 0; 4876 dstopt += sizeof (struct T_opthdr); 4877 dstptr = (uint8_t *)dstopt; 4878 *dstptr = ((ipha_t *)rptr)->ipha_ttl; 4879 dstopt = (char *)toh + toh->len; 4880 udi_size -= toh->len; 4881 } 4882 4883 /* Consumed all of allocated space */ 4884 ASSERT(udi_size == 0); 4885 } 4886 } else { 4887 sin6_t *sin6; 4888 4889 /* 4890 * Handle both IPv4 and IPv6 packets for IPv6 sockets. 4891 * 4892 * Normally we only send up the address. If receiving of any 4893 * optional receive side information is enabled, we also send 4894 * that up as options. 4895 * [ Only udp_rput_other() handles packets that contain IP 4896 * options so code to account for does not appear immediately 4897 * below but elsewhere ] 4898 */ 4899 udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin6_t); 4900 4901 if (ipp.ipp_fields & (IPPF_HOPOPTS|IPPF_DSTOPTS|IPPF_RTDSTOPTS| 4902 IPPF_RTHDR|IPPF_IFINDEX)) { 4903 if (udp->udp_ipv6_recvhopopts && 4904 (ipp.ipp_fields & IPPF_HOPOPTS)) { 4905 size_t hlen; 4906 4907 UDP_STAT(udp_in_recvhopopts); 4908 hlen = copy_hop_opts(&ipp, NULL); 4909 if (hlen == 0) 4910 ipp.ipp_fields &= ~IPPF_HOPOPTS; 4911 udi_size += hlen; 4912 } 4913 if ((udp->udp_ipv6_recvdstopts || 4914 udp->udp_old_ipv6_recvdstopts) && 4915 (ipp.ipp_fields & IPPF_DSTOPTS)) { 4916 udi_size += sizeof (struct T_opthdr) + 4917 ipp.ipp_dstoptslen; 4918 UDP_STAT(udp_in_recvdstopts); 4919 } 4920 if (((udp->udp_ipv6_recvdstopts && 4921 udp->udp_ipv6_recvrthdr && 4922 (ipp.ipp_fields & IPPF_RTHDR)) || 4923 udp->udp_ipv6_recvrthdrdstopts) && 4924 (ipp.ipp_fields & IPPF_RTDSTOPTS)) { 4925 udi_size += sizeof (struct T_opthdr) + 4926 ipp.ipp_rtdstoptslen; 4927 UDP_STAT(udp_in_recvrtdstopts); 4928 } 4929 if (udp->udp_ipv6_recvrthdr && 4930 (ipp.ipp_fields & IPPF_RTHDR)) { 4931 udi_size += sizeof (struct T_opthdr) + 4932 ipp.ipp_rthdrlen; 4933 UDP_STAT(udp_in_recvrthdr); 4934 } 4935 if (udp->udp_ipv6_recvpktinfo && 4936 (ipp.ipp_fields & IPPF_IFINDEX)) { 4937 udi_size += sizeof (struct T_opthdr) + 4938 sizeof (struct in6_pktinfo); 4939 UDP_STAT(udp_in_recvpktinfo); 4940 } 4941 4942 } 4943 if (udp->udp_recvucred && (cr = DB_CRED(mp)) != NULL) { 4944 udi_size += sizeof (struct T_opthdr) + ucredsize; 4945 cpid = DB_CPID(mp); 4946 UDP_STAT(udp_in_recvucred); 4947 } 4948 4949 if (udp->udp_ipv6_recvhoplimit) { 4950 udi_size += sizeof (struct T_opthdr) + sizeof (int); 4951 UDP_STAT(udp_in_recvhoplimit); 4952 } 4953 4954 if (udp->udp_ipv6_recvtclass) { 4955 udi_size += sizeof (struct T_opthdr) + sizeof (int); 4956 UDP_STAT(udp_in_recvtclass); 4957 } 4958 4959 mp1 = allocb(udi_size, BPRI_MED); 4960 if (mp1 == NULL) { 4961 freemsg(mp); 4962 if (options_mp != NULL) 4963 freeb(options_mp); 4964 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 4965 "udp_rput_end: q %p (%S)", q, "allocbfail"); 4966 BUMP_MIB(&udp_mib, udpInErrors); 4967 return; 4968 } 4969 mp1->b_cont = mp; 4970 mp = mp1; 4971 mp->b_datap->db_type = M_PROTO; 4972 tudi = (struct T_unitdata_ind *)mp->b_rptr; 4973 mp->b_wptr = (uchar_t *)tudi + udi_size; 4974 tudi->PRIM_type = T_UNITDATA_IND; 4975 tudi->SRC_length = sizeof (sin6_t); 4976 tudi->SRC_offset = sizeof (struct T_unitdata_ind); 4977 tudi->OPT_offset = sizeof (struct T_unitdata_ind) + 4978 sizeof (sin6_t); 4979 udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin6_t)); 4980 tudi->OPT_length = udi_size; 4981 sin6 = (sin6_t *)&tudi[1]; 4982 if (ipversion == IPV4_VERSION) { 4983 in6_addr_t v6dst; 4984 4985 IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_src, 4986 &sin6->sin6_addr); 4987 IN6_IPADDR_TO_V4MAPPED(((ipha_t *)rptr)->ipha_dst, 4988 &v6dst); 4989 sin6->sin6_flowinfo = 0; 4990 sin6->sin6_scope_id = 0; 4991 sin6->__sin6_src_id = ip_srcid_find_addr(&v6dst, 4992 connp->conn_zoneid); 4993 } else { 4994 sin6->sin6_addr = ip6h->ip6_src; 4995 /* No sin6_flowinfo per API */ 4996 sin6->sin6_flowinfo = 0; 4997 /* For link-scope source pass up scope id */ 4998 if ((ipp.ipp_fields & IPPF_IFINDEX) && 4999 IN6_IS_ADDR_LINKSCOPE(&ip6h->ip6_src)) 5000 sin6->sin6_scope_id = ipp.ipp_ifindex; 5001 else 5002 sin6->sin6_scope_id = 0; 5003 sin6->__sin6_src_id = ip_srcid_find_addr( 5004 &ip6h->ip6_dst, connp->conn_zoneid); 5005 } 5006 sin6->sin6_port = udpha->uha_src_port; 5007 sin6->sin6_family = udp->udp_family; 5008 5009 if (udi_size != 0) { 5010 uchar_t *dstopt; 5011 5012 dstopt = (uchar_t *)&sin6[1]; 5013 if (udp->udp_ipv6_recvpktinfo && 5014 (ipp.ipp_fields & IPPF_IFINDEX)) { 5015 struct T_opthdr *toh; 5016 struct in6_pktinfo *pkti; 5017 5018 toh = (struct T_opthdr *)dstopt; 5019 toh->level = IPPROTO_IPV6; 5020 toh->name = IPV6_PKTINFO; 5021 toh->len = sizeof (struct T_opthdr) + 5022 sizeof (*pkti); 5023 toh->status = 0; 5024 dstopt += sizeof (struct T_opthdr); 5025 pkti = (struct in6_pktinfo *)dstopt; 5026 if (ipversion == IPV6_VERSION) 5027 pkti->ipi6_addr = ip6h->ip6_dst; 5028 else 5029 IN6_IPADDR_TO_V4MAPPED( 5030 ((ipha_t *)rptr)->ipha_dst, 5031 &pkti->ipi6_addr); 5032 pkti->ipi6_ifindex = ipp.ipp_ifindex; 5033 dstopt += sizeof (*pkti); 5034 udi_size -= toh->len; 5035 } 5036 if (udp->udp_ipv6_recvhoplimit) { 5037 struct T_opthdr *toh; 5038 5039 toh = (struct T_opthdr *)dstopt; 5040 toh->level = IPPROTO_IPV6; 5041 toh->name = IPV6_HOPLIMIT; 5042 toh->len = sizeof (struct T_opthdr) + 5043 sizeof (uint_t); 5044 toh->status = 0; 5045 dstopt += sizeof (struct T_opthdr); 5046 if (ipversion == IPV6_VERSION) 5047 *(uint_t *)dstopt = ip6h->ip6_hops; 5048 else 5049 *(uint_t *)dstopt = 5050 ((ipha_t *)rptr)->ipha_ttl; 5051 dstopt += sizeof (uint_t); 5052 udi_size -= toh->len; 5053 } 5054 if (udp->udp_ipv6_recvtclass) { 5055 struct T_opthdr *toh; 5056 5057 toh = (struct T_opthdr *)dstopt; 5058 toh->level = IPPROTO_IPV6; 5059 toh->name = IPV6_TCLASS; 5060 toh->len = sizeof (struct T_opthdr) + 5061 sizeof (uint_t); 5062 toh->status = 0; 5063 dstopt += sizeof (struct T_opthdr); 5064 if (ipversion == IPV6_VERSION) { 5065 *(uint_t *)dstopt = 5066 IPV6_FLOW_TCLASS(ip6h->ip6_flow); 5067 } else { 5068 ipha_t *ipha = (ipha_t *)rptr; 5069 *(uint_t *)dstopt = 5070 ipha->ipha_type_of_service; 5071 } 5072 dstopt += sizeof (uint_t); 5073 udi_size -= toh->len; 5074 } 5075 if (udp->udp_ipv6_recvhopopts && 5076 (ipp.ipp_fields & IPPF_HOPOPTS)) { 5077 size_t hlen; 5078 5079 hlen = copy_hop_opts(&ipp, dstopt); 5080 dstopt += hlen; 5081 udi_size -= hlen; 5082 } 5083 if (udp->udp_ipv6_recvdstopts && 5084 udp->udp_ipv6_recvrthdr && 5085 (ipp.ipp_fields & IPPF_RTHDR) && 5086 (ipp.ipp_fields & IPPF_RTDSTOPTS)) { 5087 struct T_opthdr *toh; 5088 5089 toh = (struct T_opthdr *)dstopt; 5090 toh->level = IPPROTO_IPV6; 5091 toh->name = IPV6_DSTOPTS; 5092 toh->len = sizeof (struct T_opthdr) + 5093 ipp.ipp_rtdstoptslen; 5094 toh->status = 0; 5095 dstopt += sizeof (struct T_opthdr); 5096 bcopy(ipp.ipp_rtdstopts, dstopt, 5097 ipp.ipp_rtdstoptslen); 5098 dstopt += ipp.ipp_rtdstoptslen; 5099 udi_size -= toh->len; 5100 } 5101 if (udp->udp_ipv6_recvrthdr && 5102 (ipp.ipp_fields & IPPF_RTHDR)) { 5103 struct T_opthdr *toh; 5104 5105 toh = (struct T_opthdr *)dstopt; 5106 toh->level = IPPROTO_IPV6; 5107 toh->name = IPV6_RTHDR; 5108 toh->len = sizeof (struct T_opthdr) + 5109 ipp.ipp_rthdrlen; 5110 toh->status = 0; 5111 dstopt += sizeof (struct T_opthdr); 5112 bcopy(ipp.ipp_rthdr, dstopt, ipp.ipp_rthdrlen); 5113 dstopt += ipp.ipp_rthdrlen; 5114 udi_size -= toh->len; 5115 } 5116 if (udp->udp_ipv6_recvdstopts && 5117 (ipp.ipp_fields & IPPF_DSTOPTS)) { 5118 struct T_opthdr *toh; 5119 5120 toh = (struct T_opthdr *)dstopt; 5121 toh->level = IPPROTO_IPV6; 5122 toh->name = IPV6_DSTOPTS; 5123 toh->len = sizeof (struct T_opthdr) + 5124 ipp.ipp_dstoptslen; 5125 toh->status = 0; 5126 dstopt += sizeof (struct T_opthdr); 5127 bcopy(ipp.ipp_dstopts, dstopt, 5128 ipp.ipp_dstoptslen); 5129 dstopt += ipp.ipp_dstoptslen; 5130 udi_size -= toh->len; 5131 } 5132 5133 if (cr != NULL) { 5134 struct T_opthdr *toh; 5135 5136 toh = (struct T_opthdr *)dstopt; 5137 toh->level = SOL_SOCKET; 5138 toh->name = SCM_UCRED; 5139 toh->len = sizeof (struct T_opthdr) + ucredsize; 5140 toh->status = 0; 5141 (void) cred2ucred(cr, cpid, &toh[1], rcr); 5142 dstopt += toh->len; 5143 udi_size -= toh->len; 5144 } 5145 /* Consumed all of allocated space */ 5146 ASSERT(udi_size == 0); 5147 } 5148 #undef sin6 5149 /* No IP_RECVDSTADDR for IPv6. */ 5150 } 5151 5152 BUMP_MIB(&udp_mib, udpInDatagrams); 5153 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 5154 "udp_rput_end: q %p (%S)", q, "end"); 5155 if (options_mp != NULL) 5156 freeb(options_mp); 5157 5158 if (udp->udp_direct_sockfs) { 5159 /* 5160 * There is nothing above us except for the stream head; 5161 * use the read-side synchronous stream interface in 5162 * order to reduce the time spent in interrupt thread. 5163 */ 5164 ASSERT(udp->udp_issocket); 5165 udp_rcv_enqueue(UDP_RD(q), udp, mp, mp_len); 5166 } else { 5167 /* 5168 * Use regular STREAMS interface to pass data upstream 5169 * if this is not a socket endpoint, or if we have 5170 * switched over to the slow mode due to sockmod being 5171 * popped or a module being pushed on top of us. 5172 */ 5173 putnext(UDP_RD(q), mp); 5174 } 5175 return; 5176 5177 tossit: 5178 freemsg(mp); 5179 if (options_mp != NULL) 5180 freeb(options_mp); 5181 BUMP_MIB(&udp_mib, udpInErrors); 5182 } 5183 5184 void 5185 udp_conn_recv(conn_t *connp, mblk_t *mp) 5186 { 5187 _UDP_ENTER(connp, mp, udp_input_wrapper, SQTAG_UDP_FANOUT); 5188 } 5189 5190 /* ARGSUSED */ 5191 static void 5192 udp_input_wrapper(void *arg, mblk_t *mp, void *arg2) 5193 { 5194 udp_input((conn_t *)arg, mp); 5195 _UDP_EXIT((conn_t *)arg); 5196 } 5197 5198 /* 5199 * Process non-M_DATA messages as well as M_DATA messages that requires 5200 * modifications to udp_ip_rcv_options i.e. IPv4 packets with IP options. 5201 */ 5202 static void 5203 udp_rput_other(queue_t *q, mblk_t *mp) 5204 { 5205 struct T_unitdata_ind *tudi; 5206 mblk_t *mp1; 5207 uchar_t *rptr; 5208 uchar_t *new_rptr; 5209 int hdr_length; 5210 int udi_size; /* Size of T_unitdata_ind */ 5211 int opt_len; /* Length of IP options */ 5212 sin_t *sin; 5213 struct T_error_ack *tea; 5214 mblk_t *options_mp = NULL; 5215 in_pktinfo_t *pinfo; 5216 boolean_t recv_on = B_FALSE; 5217 cred_t *cr = NULL; 5218 udp_t *udp = Q_TO_UDP(q); 5219 pid_t cpid; 5220 cred_t *rcr = udp->udp_connp->conn_cred; 5221 5222 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_START, 5223 "udp_rput_other: q %p mp %p", q, mp); 5224 5225 ASSERT(OK_32PTR(mp->b_rptr)); 5226 rptr = mp->b_rptr; 5227 5228 switch (mp->b_datap->db_type) { 5229 case M_CTL: 5230 /* 5231 * We are here only if IP_RECVSLLA and/or IP_RECVIF are set 5232 */ 5233 recv_on = B_TRUE; 5234 options_mp = mp; 5235 pinfo = (in_pktinfo_t *)options_mp->b_rptr; 5236 5237 /* 5238 * The actual data is in mp->b_cont 5239 */ 5240 mp = mp->b_cont; 5241 ASSERT(OK_32PTR(mp->b_rptr)); 5242 rptr = mp->b_rptr; 5243 break; 5244 case M_DATA: 5245 /* 5246 * M_DATA messages contain IPv4 datagrams. They are handled 5247 * after this switch. 5248 */ 5249 break; 5250 case M_PROTO: 5251 case M_PCPROTO: 5252 /* M_PROTO messages contain some type of TPI message. */ 5253 ASSERT((uintptr_t)(mp->b_wptr - rptr) <= (uintptr_t)INT_MAX); 5254 if (mp->b_wptr - rptr < sizeof (t_scalar_t)) { 5255 freemsg(mp); 5256 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 5257 "udp_rput_other_end: q %p (%S)", q, "protoshort"); 5258 return; 5259 } 5260 tea = (struct T_error_ack *)rptr; 5261 5262 switch (tea->PRIM_type) { 5263 case T_ERROR_ACK: 5264 switch (tea->ERROR_prim) { 5265 case O_T_BIND_REQ: 5266 case T_BIND_REQ: { 5267 /* 5268 * If our O_T_BIND_REQ/T_BIND_REQ fails, 5269 * clear out the associated port and source 5270 * address before passing the message 5271 * upstream. If this was caused by a T_CONN_REQ 5272 * revert back to bound state. 5273 */ 5274 udp_fanout_t *udpf; 5275 5276 udpf = &udp_bind_fanout[ 5277 UDP_BIND_HASH(udp->udp_port)]; 5278 mutex_enter(&udpf->uf_lock); 5279 if (udp->udp_state == TS_DATA_XFER) { 5280 /* Connect failed */ 5281 tea->ERROR_prim = T_CONN_REQ; 5282 /* Revert back to the bound source */ 5283 udp->udp_v6src = udp->udp_bound_v6src; 5284 udp->udp_state = TS_IDLE; 5285 mutex_exit(&udpf->uf_lock); 5286 if (udp->udp_family == AF_INET6) 5287 (void) udp_build_hdrs(q, udp); 5288 break; 5289 } 5290 5291 if (udp->udp_discon_pending) { 5292 tea->ERROR_prim = T_DISCON_REQ; 5293 udp->udp_discon_pending = 0; 5294 } 5295 V6_SET_ZERO(udp->udp_v6src); 5296 V6_SET_ZERO(udp->udp_bound_v6src); 5297 udp->udp_state = TS_UNBND; 5298 udp_bind_hash_remove(udp, B_TRUE); 5299 udp->udp_port = 0; 5300 mutex_exit(&udpf->uf_lock); 5301 if (udp->udp_family == AF_INET6) 5302 (void) udp_build_hdrs(q, udp); 5303 break; 5304 } 5305 default: 5306 break; 5307 } 5308 break; 5309 case T_BIND_ACK: 5310 udp_rput_bind_ack(q, mp); 5311 return; 5312 5313 case T_OPTMGMT_ACK: 5314 case T_OK_ACK: 5315 break; 5316 default: 5317 freemsg(mp); 5318 return; 5319 } 5320 putnext(UDP_RD(q), mp); 5321 return; 5322 } 5323 5324 /* 5325 * This is the inbound data path. 5326 * First, we make sure the data contains both IP and UDP headers. 5327 * 5328 * This handle IPv4 packets for only AF_INET sockets. 5329 * AF_INET6 sockets can never access udp_ip_rcv_options thus there 5330 * is no need saving the options. 5331 */ 5332 ASSERT(IPH_HDR_VERSION((ipha_t *)rptr) == IPV4_VERSION); 5333 hdr_length = IPH_HDR_LENGTH(rptr) + UDPH_SIZE; 5334 if (mp->b_wptr - rptr < hdr_length) { 5335 if (!pullupmsg(mp, hdr_length)) { 5336 freemsg(mp); 5337 if (options_mp != NULL) 5338 freeb(options_mp); 5339 BUMP_MIB(&udp_mib, udpInErrors); 5340 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 5341 "udp_rput_other_end: q %p (%S)", q, "hdrshort"); 5342 BUMP_MIB(&udp_mib, udpInErrors); 5343 return; 5344 } 5345 rptr = mp->b_rptr; 5346 } 5347 /* Walk past the headers. */ 5348 new_rptr = rptr + hdr_length; 5349 if (!udp->udp_rcvhdr) 5350 mp->b_rptr = new_rptr; 5351 5352 /* Save the options if any */ 5353 opt_len = hdr_length - (IP_SIMPLE_HDR_LENGTH + UDPH_SIZE); 5354 if (opt_len > 0) { 5355 if (opt_len > udp->udp_ip_rcv_options_len) { 5356 if (udp->udp_ip_rcv_options_len) 5357 mi_free((char *)udp->udp_ip_rcv_options); 5358 udp->udp_ip_rcv_options_len = 0; 5359 udp->udp_ip_rcv_options = 5360 (uchar_t *)mi_alloc(opt_len, BPRI_HI); 5361 if (udp->udp_ip_rcv_options) 5362 udp->udp_ip_rcv_options_len = opt_len; 5363 } 5364 if (udp->udp_ip_rcv_options_len) { 5365 bcopy(rptr + IP_SIMPLE_HDR_LENGTH, 5366 udp->udp_ip_rcv_options, opt_len); 5367 /* Adjust length if we are resusing the space */ 5368 udp->udp_ip_rcv_options_len = opt_len; 5369 } 5370 } else if (udp->udp_ip_rcv_options_len) { 5371 mi_free((char *)udp->udp_ip_rcv_options); 5372 udp->udp_ip_rcv_options = NULL; 5373 udp->udp_ip_rcv_options_len = 0; 5374 } 5375 5376 /* 5377 * Normally only send up the address. 5378 * If IP_RECVDSTADDR is set we include the destination IP 5379 * address as an option. With IP_RECVOPTS we include all 5380 * the IP options. 5381 */ 5382 udi_size = sizeof (struct T_unitdata_ind) + sizeof (sin_t); 5383 if (udp->udp_recvdstaddr) { 5384 udi_size += sizeof (struct T_opthdr) + sizeof (struct in_addr); 5385 UDP_STAT(udp_in_recvdstaddr); 5386 } 5387 if (udp->udp_recvopts && opt_len > 0) { 5388 udi_size += sizeof (struct T_opthdr) + opt_len; 5389 UDP_STAT(udp_in_recvopts); 5390 } 5391 5392 /* 5393 * If the IP_RECVSLLA or the IP_RECVIF is set then allocate 5394 * space accordingly 5395 */ 5396 if (udp->udp_recvif && recv_on && 5397 (pinfo->in_pkt_flags & IPF_RECVIF)) { 5398 udi_size += sizeof (struct T_opthdr) + sizeof (uint_t); 5399 UDP_STAT(udp_in_recvif); 5400 } 5401 5402 if (udp->udp_recvslla && recv_on && 5403 (pinfo->in_pkt_flags & IPF_RECVSLLA)) { 5404 udi_size += sizeof (struct T_opthdr) + 5405 sizeof (struct sockaddr_dl); 5406 UDP_STAT(udp_in_recvslla); 5407 } 5408 5409 if (udp->udp_recvucred && (cr = DB_CRED(mp)) != NULL) { 5410 udi_size += sizeof (struct T_opthdr) + ucredsize; 5411 cpid = DB_CPID(mp); 5412 UDP_STAT(udp_in_recvucred); 5413 } 5414 /* 5415 * If IP_RECVTTL is set allocate the appropriate sized buffer 5416 */ 5417 if (udp->udp_recvttl) { 5418 udi_size += sizeof (struct T_opthdr) + sizeof (uint8_t); 5419 UDP_STAT(udp_in_recvttl); 5420 } 5421 5422 /* Allocate a message block for the T_UNITDATA_IND structure. */ 5423 mp1 = allocb(udi_size, BPRI_MED); 5424 if (mp1 == NULL) { 5425 freemsg(mp); 5426 if (options_mp != NULL) 5427 freeb(options_mp); 5428 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 5429 "udp_rput_other_end: q %p (%S)", q, "allocbfail"); 5430 BUMP_MIB(&udp_mib, udpInErrors); 5431 return; 5432 } 5433 mp1->b_cont = mp; 5434 mp = mp1; 5435 mp->b_datap->db_type = M_PROTO; 5436 tudi = (struct T_unitdata_ind *)mp->b_rptr; 5437 mp->b_wptr = (uchar_t *)tudi + udi_size; 5438 tudi->PRIM_type = T_UNITDATA_IND; 5439 tudi->SRC_length = sizeof (sin_t); 5440 tudi->SRC_offset = sizeof (struct T_unitdata_ind); 5441 tudi->OPT_offset = sizeof (struct T_unitdata_ind) + sizeof (sin_t); 5442 udi_size -= (sizeof (struct T_unitdata_ind) + sizeof (sin_t)); 5443 tudi->OPT_length = udi_size; 5444 5445 sin = (sin_t *)&tudi[1]; 5446 sin->sin_addr.s_addr = ((ipha_t *)rptr)->ipha_src; 5447 sin->sin_port = ((in_port_t *) 5448 new_rptr)[-(UDPH_SIZE/sizeof (in_port_t))]; 5449 sin->sin_family = AF_INET; 5450 *(uint32_t *)&sin->sin_zero[0] = 0; 5451 *(uint32_t *)&sin->sin_zero[4] = 0; 5452 5453 /* 5454 * Add options if IP_RECVDSTADDR, IP_RECVIF, IP_RECVSLLA or 5455 * IP_RECVTTL has been set. 5456 */ 5457 if (udi_size != 0) { 5458 /* 5459 * Copy in destination address before options to avoid any 5460 * padding issues. 5461 */ 5462 char *dstopt; 5463 5464 dstopt = (char *)&sin[1]; 5465 if (udp->udp_recvdstaddr) { 5466 struct T_opthdr *toh; 5467 ipaddr_t *dstptr; 5468 5469 toh = (struct T_opthdr *)dstopt; 5470 toh->level = IPPROTO_IP; 5471 toh->name = IP_RECVDSTADDR; 5472 toh->len = sizeof (struct T_opthdr) + sizeof (ipaddr_t); 5473 toh->status = 0; 5474 dstopt += sizeof (struct T_opthdr); 5475 dstptr = (ipaddr_t *)dstopt; 5476 *dstptr = (((ipaddr_t *)rptr)[4]); 5477 dstopt += sizeof (ipaddr_t); 5478 udi_size -= toh->len; 5479 } 5480 if (udp->udp_recvopts && udi_size != 0) { 5481 struct T_opthdr *toh; 5482 5483 toh = (struct T_opthdr *)dstopt; 5484 toh->level = IPPROTO_IP; 5485 toh->name = IP_RECVOPTS; 5486 toh->len = sizeof (struct T_opthdr) + opt_len; 5487 toh->status = 0; 5488 dstopt += sizeof (struct T_opthdr); 5489 bcopy(rptr + IP_SIMPLE_HDR_LENGTH, dstopt, opt_len); 5490 dstopt += opt_len; 5491 udi_size -= toh->len; 5492 } 5493 5494 if (udp->udp_recvslla && recv_on && 5495 (pinfo->in_pkt_flags & IPF_RECVSLLA)) { 5496 5497 struct T_opthdr *toh; 5498 struct sockaddr_dl *dstptr; 5499 5500 toh = (struct T_opthdr *)dstopt; 5501 toh->level = IPPROTO_IP; 5502 toh->name = IP_RECVSLLA; 5503 toh->len = sizeof (struct T_opthdr) + 5504 sizeof (struct sockaddr_dl); 5505 toh->status = 0; 5506 dstopt += sizeof (struct T_opthdr); 5507 dstptr = (struct sockaddr_dl *)dstopt; 5508 bcopy(&pinfo->in_pkt_slla, dstptr, 5509 sizeof (struct sockaddr_dl)); 5510 dstopt += sizeof (struct sockaddr_dl); 5511 udi_size -= toh->len; 5512 } 5513 5514 if (udp->udp_recvif && recv_on && 5515 (pinfo->in_pkt_flags & IPF_RECVIF)) { 5516 5517 struct T_opthdr *toh; 5518 uint_t *dstptr; 5519 5520 toh = (struct T_opthdr *)dstopt; 5521 toh->level = IPPROTO_IP; 5522 toh->name = IP_RECVIF; 5523 toh->len = sizeof (struct T_opthdr) + 5524 sizeof (uint_t); 5525 toh->status = 0; 5526 dstopt += sizeof (struct T_opthdr); 5527 dstptr = (uint_t *)dstopt; 5528 *dstptr = pinfo->in_pkt_ifindex; 5529 dstopt += sizeof (uint_t); 5530 udi_size -= toh->len; 5531 } 5532 5533 if (cr != NULL) { 5534 struct T_opthdr *toh; 5535 5536 toh = (struct T_opthdr *)dstopt; 5537 toh->level = SOL_SOCKET; 5538 toh->name = SCM_UCRED; 5539 toh->len = sizeof (struct T_opthdr) + ucredsize; 5540 toh->status = 0; 5541 (void) cred2ucred(cr, cpid, &toh[1], rcr); 5542 dstopt += toh->len; 5543 udi_size -= toh->len; 5544 } 5545 5546 if (udp->udp_recvttl) { 5547 struct T_opthdr *toh; 5548 uint8_t *dstptr; 5549 5550 toh = (struct T_opthdr *)dstopt; 5551 toh->level = IPPROTO_IP; 5552 toh->name = IP_RECVTTL; 5553 toh->len = sizeof (struct T_opthdr) + 5554 sizeof (uint8_t); 5555 toh->status = 0; 5556 dstopt += sizeof (struct T_opthdr); 5557 dstptr = (uint8_t *)dstopt; 5558 *dstptr = ((ipha_t *)rptr)->ipha_ttl; 5559 dstopt += sizeof (uint8_t); 5560 udi_size -= toh->len; 5561 } 5562 5563 ASSERT(udi_size == 0); /* "Consumed" all of allocated space */ 5564 } 5565 BUMP_MIB(&udp_mib, udpInDatagrams); 5566 TRACE_2(TR_FAC_UDP, TR_UDP_RPUT_END, 5567 "udp_rput_other_end: q %p (%S)", q, "end"); 5568 if (options_mp != NULL) 5569 freeb(options_mp); 5570 5571 if (udp->udp_direct_sockfs) { 5572 /* 5573 * There is nothing above us except for the stream head; 5574 * use the read-side synchronous stream interface in 5575 * order to reduce the time spent in interrupt thread. 5576 */ 5577 ASSERT(udp->udp_issocket); 5578 udp_rcv_enqueue(UDP_RD(q), udp, mp, msgdsize(mp)); 5579 } else { 5580 /* 5581 * Use regular STREAMS interface to pass data upstream 5582 * if this is not a socket endpoint, or if we have 5583 * switched over to the slow mode due to sockmod being 5584 * popped or a module being pushed on top of us. 5585 */ 5586 putnext(UDP_RD(q), mp); 5587 } 5588 } 5589 5590 /* ARGSUSED */ 5591 static void 5592 udp_rput_other_wrapper(void *arg, mblk_t *mp, void *arg2) 5593 { 5594 conn_t *connp = arg; 5595 5596 udp_rput_other(connp->conn_rq, mp); 5597 udp_exit(connp); 5598 } 5599 5600 /* 5601 * Process a T_BIND_ACK 5602 */ 5603 static void 5604 udp_rput_bind_ack(queue_t *q, mblk_t *mp) 5605 { 5606 udp_t *udp = Q_TO_UDP(q); 5607 mblk_t *mp1; 5608 ire_t *ire; 5609 struct T_bind_ack *tba; 5610 uchar_t *addrp; 5611 ipa_conn_t *ac; 5612 ipa6_conn_t *ac6; 5613 5614 if (udp->udp_discon_pending) 5615 udp->udp_discon_pending = 0; 5616 5617 /* 5618 * If a broadcast/multicast address was bound set 5619 * the source address to 0. 5620 * This ensures no datagrams with broadcast address 5621 * as source address are emitted (which would violate 5622 * RFC1122 - Hosts requirements) 5623 * 5624 * Note that when connecting the returned IRE is 5625 * for the destination address and we only perform 5626 * the broadcast check for the source address (it 5627 * is OK to connect to a broadcast/multicast address.) 5628 */ 5629 mp1 = mp->b_cont; 5630 if (mp1 != NULL && mp1->b_datap->db_type == IRE_DB_TYPE) { 5631 ire = (ire_t *)mp1->b_rptr; 5632 5633 /* 5634 * Note: we get IRE_BROADCAST for IPv6 to "mark" a multicast 5635 * local address. 5636 */ 5637 if (ire->ire_type == IRE_BROADCAST && 5638 udp->udp_state != TS_DATA_XFER) { 5639 /* This was just a local bind to a broadcast addr */ 5640 V6_SET_ZERO(udp->udp_v6src); 5641 if (udp->udp_family == AF_INET6) 5642 (void) udp_build_hdrs(q, udp); 5643 } else if (V6_OR_V4_INADDR_ANY(udp->udp_v6src)) { 5644 /* 5645 * Local address not yet set - pick it from the 5646 * T_bind_ack 5647 */ 5648 tba = (struct T_bind_ack *)mp->b_rptr; 5649 addrp = &mp->b_rptr[tba->ADDR_offset]; 5650 switch (udp->udp_family) { 5651 case AF_INET: 5652 if (tba->ADDR_length == sizeof (ipa_conn_t)) { 5653 ac = (ipa_conn_t *)addrp; 5654 } else { 5655 ASSERT(tba->ADDR_length == 5656 sizeof (ipa_conn_x_t)); 5657 ac = &((ipa_conn_x_t *)addrp)->acx_conn; 5658 } 5659 IN6_IPADDR_TO_V4MAPPED(ac->ac_laddr, 5660 &udp->udp_v6src); 5661 break; 5662 case AF_INET6: 5663 if (tba->ADDR_length == sizeof (ipa6_conn_t)) { 5664 ac6 = (ipa6_conn_t *)addrp; 5665 } else { 5666 ASSERT(tba->ADDR_length == 5667 sizeof (ipa6_conn_x_t)); 5668 ac6 = &((ipa6_conn_x_t *) 5669 addrp)->ac6x_conn; 5670 } 5671 udp->udp_v6src = ac6->ac6_laddr; 5672 (void) udp_build_hdrs(q, udp); 5673 break; 5674 } 5675 } 5676 mp1 = mp1->b_cont; 5677 } 5678 /* 5679 * Look for one or more appended ACK message added by 5680 * udp_connect or udp_disconnect. 5681 * If none found just send up the T_BIND_ACK. 5682 * udp_connect has appended a T_OK_ACK and a T_CONN_CON. 5683 * udp_disconnect has appended a T_OK_ACK. 5684 */ 5685 if (mp1 != NULL) { 5686 if (mp->b_cont == mp1) 5687 mp->b_cont = NULL; 5688 else { 5689 ASSERT(mp->b_cont->b_cont == mp1); 5690 mp->b_cont->b_cont = NULL; 5691 } 5692 freemsg(mp); 5693 mp = mp1; 5694 while (mp != NULL) { 5695 mp1 = mp->b_cont; 5696 mp->b_cont = NULL; 5697 putnext(UDP_RD(q), mp); 5698 mp = mp1; 5699 } 5700 return; 5701 } 5702 freemsg(mp->b_cont); 5703 mp->b_cont = NULL; 5704 putnext(UDP_RD(q), mp); 5705 } 5706 5707 /* 5708 * return SNMP stuff in buffer in mpdata 5709 */ 5710 int 5711 udp_snmp_get(queue_t *q, mblk_t *mpctl) 5712 { 5713 mblk_t *mpdata; 5714 mblk_t *mp_conn_ctl; 5715 mblk_t *mp_attr_ctl; 5716 mblk_t *mp6_conn_ctl; 5717 mblk_t *mp6_attr_ctl; 5718 mblk_t *mp_conn_tail; 5719 mblk_t *mp_attr_tail; 5720 mblk_t *mp6_conn_tail; 5721 mblk_t *mp6_attr_tail; 5722 struct opthdr *optp; 5723 mib2_udpEntry_t ude; 5724 mib2_udp6Entry_t ude6; 5725 mib2_transportMLPEntry_t mlp; 5726 int state; 5727 zoneid_t zoneid; 5728 int i; 5729 connf_t *connfp; 5730 conn_t *connp = Q_TO_CONN(q); 5731 udp_t *udp = connp->conn_udp; 5732 int v4_conn_idx; 5733 int v6_conn_idx; 5734 boolean_t needattr; 5735 5736 mp_conn_ctl = mp_attr_ctl = mp6_conn_ctl = NULL; 5737 if (mpctl == NULL || 5738 (mpdata = mpctl->b_cont) == NULL || 5739 (mp_conn_ctl = copymsg(mpctl)) == NULL || 5740 (mp_attr_ctl = copymsg(mpctl)) == NULL || 5741 (mp6_conn_ctl = copymsg(mpctl)) == NULL || 5742 (mp6_attr_ctl = copymsg(mpctl)) == NULL) { 5743 freemsg(mp_conn_ctl); 5744 freemsg(mp_attr_ctl); 5745 freemsg(mp6_conn_ctl); 5746 return (0); 5747 } 5748 5749 zoneid = connp->conn_zoneid; 5750 5751 /* fixed length structure for IPv4 and IPv6 counters */ 5752 SET_MIB(udp_mib.udpEntrySize, sizeof (mib2_udpEntry_t)); 5753 SET_MIB(udp_mib.udp6EntrySize, sizeof (mib2_udp6Entry_t)); 5754 optp = (struct opthdr *)&mpctl->b_rptr[sizeof (struct T_optmgmt_ack)]; 5755 optp->level = MIB2_UDP; 5756 optp->name = 0; 5757 (void) snmp_append_data(mpdata, (char *)&udp_mib, sizeof (udp_mib)); 5758 optp->len = msgdsize(mpdata); 5759 qreply(q, mpctl); 5760 5761 mp_conn_tail = mp_attr_tail = mp6_conn_tail = mp6_attr_tail = NULL; 5762 v4_conn_idx = v6_conn_idx = 0; 5763 5764 for (i = 0; i < CONN_G_HASH_SIZE; i++) { 5765 connfp = &ipcl_globalhash_fanout[i]; 5766 connp = NULL; 5767 5768 while ((connp = ipcl_get_next_conn(connfp, connp, 5769 IPCL_UDP))) { 5770 udp = connp->conn_udp; 5771 if (zoneid != connp->conn_zoneid) 5772 continue; 5773 5774 /* 5775 * Note that the port numbers are sent in 5776 * host byte order 5777 */ 5778 5779 if (udp->udp_state == TS_UNBND) 5780 state = MIB2_UDP_unbound; 5781 else if (udp->udp_state == TS_IDLE) 5782 state = MIB2_UDP_idle; 5783 else if (udp->udp_state == TS_DATA_XFER) 5784 state = MIB2_UDP_connected; 5785 else 5786 state = MIB2_UDP_unknown; 5787 5788 needattr = B_FALSE; 5789 bzero(&mlp, sizeof (mlp)); 5790 if (connp->conn_mlp_type != mlptSingle) { 5791 if (connp->conn_mlp_type == mlptShared || 5792 connp->conn_mlp_type == mlptBoth) 5793 mlp.tme_flags |= MIB2_TMEF_SHARED; 5794 if (connp->conn_mlp_type == mlptPrivate || 5795 connp->conn_mlp_type == mlptBoth) 5796 mlp.tme_flags |= MIB2_TMEF_PRIVATE; 5797 needattr = B_TRUE; 5798 } 5799 5800 /* 5801 * Create an IPv4 table entry for IPv4 entries and also 5802 * any IPv6 entries which are bound to in6addr_any 5803 * (i.e. anything a IPv4 peer could connect/send to). 5804 */ 5805 if (udp->udp_ipversion == IPV4_VERSION || 5806 (udp->udp_state <= TS_IDLE && 5807 IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src))) { 5808 ude.udpEntryInfo.ue_state = state; 5809 /* 5810 * If in6addr_any this will set it to 5811 * INADDR_ANY 5812 */ 5813 ude.udpLocalAddress = 5814 V4_PART_OF_V6(udp->udp_v6src); 5815 ude.udpLocalPort = ntohs(udp->udp_port); 5816 if (udp->udp_state == TS_DATA_XFER) { 5817 /* 5818 * Can potentially get here for 5819 * v6 socket if another process 5820 * (say, ping) has just done a 5821 * sendto(), changing the state 5822 * from the TS_IDLE above to 5823 * TS_DATA_XFER by the time we hit 5824 * this part of the code. 5825 */ 5826 ude.udpEntryInfo.ue_RemoteAddress = 5827 V4_PART_OF_V6(udp->udp_v6dst); 5828 ude.udpEntryInfo.ue_RemotePort = 5829 ntohs(udp->udp_dstport); 5830 } else { 5831 ude.udpEntryInfo.ue_RemoteAddress = 0; 5832 ude.udpEntryInfo.ue_RemotePort = 0; 5833 } 5834 (void) snmp_append_data2(mp_conn_ctl->b_cont, 5835 &mp_conn_tail, (char *)&ude, sizeof (ude)); 5836 mlp.tme_connidx = v4_conn_idx++; 5837 if (needattr) 5838 (void) snmp_append_data2( 5839 mp_attr_ctl->b_cont, &mp_attr_tail, 5840 (char *)&mlp, sizeof (mlp)); 5841 } 5842 if (udp->udp_ipversion == IPV6_VERSION) { 5843 ude6.udp6EntryInfo.ue_state = state; 5844 ude6.udp6LocalAddress = udp->udp_v6src; 5845 ude6.udp6LocalPort = ntohs(udp->udp_port); 5846 ude6.udp6IfIndex = udp->udp_bound_if; 5847 if (udp->udp_state == TS_DATA_XFER) { 5848 ude6.udp6EntryInfo.ue_RemoteAddress = 5849 udp->udp_v6dst; 5850 ude6.udp6EntryInfo.ue_RemotePort = 5851 ntohs(udp->udp_dstport); 5852 } else { 5853 ude6.udp6EntryInfo.ue_RemoteAddress = 5854 sin6_null.sin6_addr; 5855 ude6.udp6EntryInfo.ue_RemotePort = 0; 5856 } 5857 (void) snmp_append_data2(mp6_conn_ctl->b_cont, 5858 &mp6_conn_tail, (char *)&ude6, 5859 sizeof (ude6)); 5860 mlp.tme_connidx = v6_conn_idx++; 5861 if (needattr) 5862 (void) snmp_append_data2( 5863 mp6_attr_ctl->b_cont, 5864 &mp6_attr_tail, (char *)&mlp, 5865 sizeof (mlp)); 5866 } 5867 } 5868 } 5869 5870 /* IPv4 UDP endpoints */ 5871 optp = (struct opthdr *)&mp_conn_ctl->b_rptr[ 5872 sizeof (struct T_optmgmt_ack)]; 5873 optp->level = MIB2_UDP; 5874 optp->name = MIB2_UDP_ENTRY; 5875 optp->len = msgdsize(mp_conn_ctl->b_cont); 5876 qreply(q, mp_conn_ctl); 5877 5878 /* table of MLP attributes... */ 5879 optp = (struct opthdr *)&mp_attr_ctl->b_rptr[ 5880 sizeof (struct T_optmgmt_ack)]; 5881 optp->level = MIB2_UDP; 5882 optp->name = EXPER_XPORT_MLP; 5883 optp->len = msgdsize(mp_attr_ctl->b_cont); 5884 if (optp->len == 0) 5885 freemsg(mp_attr_ctl); 5886 else 5887 qreply(q, mp_attr_ctl); 5888 5889 /* IPv6 UDP endpoints */ 5890 optp = (struct opthdr *)&mp6_conn_ctl->b_rptr[ 5891 sizeof (struct T_optmgmt_ack)]; 5892 optp->level = MIB2_UDP6; 5893 optp->name = MIB2_UDP6_ENTRY; 5894 optp->len = msgdsize(mp6_conn_ctl->b_cont); 5895 qreply(q, mp6_conn_ctl); 5896 5897 /* table of MLP attributes... */ 5898 optp = (struct opthdr *)&mp6_attr_ctl->b_rptr[ 5899 sizeof (struct T_optmgmt_ack)]; 5900 optp->level = MIB2_UDP6; 5901 optp->name = EXPER_XPORT_MLP; 5902 optp->len = msgdsize(mp6_attr_ctl->b_cont); 5903 if (optp->len == 0) 5904 freemsg(mp6_attr_ctl); 5905 else 5906 qreply(q, mp6_attr_ctl); 5907 5908 return (1); 5909 } 5910 5911 /* 5912 * Return 0 if invalid set request, 1 otherwise, including non-udp requests. 5913 * NOTE: Per MIB-II, UDP has no writable data. 5914 * TODO: If this ever actually tries to set anything, it needs to be 5915 * to do the appropriate locking. 5916 */ 5917 /* ARGSUSED */ 5918 int 5919 udp_snmp_set(queue_t *q, t_scalar_t level, t_scalar_t name, 5920 uchar_t *ptr, int len) 5921 { 5922 switch (level) { 5923 case MIB2_UDP: 5924 return (0); 5925 default: 5926 return (1); 5927 } 5928 } 5929 5930 static void 5931 udp_report_item(mblk_t *mp, udp_t *udp) 5932 { 5933 char *state; 5934 char addrbuf1[INET6_ADDRSTRLEN]; 5935 char addrbuf2[INET6_ADDRSTRLEN]; 5936 uint_t print_len, buf_len; 5937 5938 buf_len = mp->b_datap->db_lim - mp->b_wptr; 5939 ASSERT(buf_len >= 0); 5940 if (buf_len == 0) 5941 return; 5942 5943 if (udp->udp_state == TS_UNBND) 5944 state = "UNBOUND"; 5945 else if (udp->udp_state == TS_IDLE) 5946 state = "IDLE"; 5947 else if (udp->udp_state == TS_DATA_XFER) 5948 state = "CONNECTED"; 5949 else 5950 state = "UnkState"; 5951 print_len = snprintf((char *)mp->b_wptr, buf_len, 5952 MI_COL_PTRFMT_STR "%4d %5u %s %s %5u %s\n", 5953 (void *)udp, udp->udp_connp->conn_zoneid, ntohs(udp->udp_port), 5954 inet_ntop(AF_INET6, &udp->udp_v6src, 5955 addrbuf1, sizeof (addrbuf1)), 5956 inet_ntop(AF_INET6, &udp->udp_v6dst, 5957 addrbuf2, sizeof (addrbuf2)), 5958 ntohs(udp->udp_dstport), state); 5959 if (print_len < buf_len) { 5960 mp->b_wptr += print_len; 5961 } else { 5962 mp->b_wptr += buf_len; 5963 } 5964 } 5965 5966 /* Report for ndd "udp_status" */ 5967 /* ARGSUSED */ 5968 static int 5969 udp_status_report(queue_t *q, mblk_t *mp, caddr_t cp, cred_t *cr) 5970 { 5971 zoneid_t zoneid; 5972 connf_t *connfp; 5973 conn_t *connp = Q_TO_CONN(q); 5974 udp_t *udp = connp->conn_udp; 5975 int i; 5976 5977 /* 5978 * Because of the ndd constraint, at most we can have 64K buffer 5979 * to put in all UDP info. So to be more efficient, just 5980 * allocate a 64K buffer here, assuming we need that large buffer. 5981 * This may be a problem as any user can read udp_status. Therefore 5982 * we limit the rate of doing this using udp_ndd_get_info_interval. 5983 * This should be OK as normal users should not do this too often. 5984 */ 5985 if (cr == NULL || secpolicy_net_config(cr, B_TRUE) != 0) { 5986 if (ddi_get_lbolt() - udp_last_ndd_get_info_time < 5987 drv_usectohz(udp_ndd_get_info_interval * 1000)) { 5988 (void) mi_mpprintf(mp, NDD_TOO_QUICK_MSG); 5989 return (0); 5990 } 5991 } 5992 if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) { 5993 /* The following may work even if we cannot get a large buf. */ 5994 (void) mi_mpprintf(mp, NDD_OUT_OF_BUF_MSG); 5995 return (0); 5996 } 5997 (void) mi_mpprintf(mp, 5998 "UDP " MI_COL_HDRPAD_STR 5999 /* 12345678[89ABCDEF] */ 6000 " zone lport src addr dest addr port state"); 6001 /* 1234 12345 xxx.xxx.xxx.xxx xxx.xxx.xxx.xxx 12345 UNBOUND */ 6002 6003 zoneid = connp->conn_zoneid; 6004 6005 for (i = 0; i < CONN_G_HASH_SIZE; i++) { 6006 connfp = &ipcl_globalhash_fanout[i]; 6007 connp = NULL; 6008 6009 while ((connp = ipcl_get_next_conn(connfp, connp, 6010 IPCL_UDP))) { 6011 udp = connp->conn_udp; 6012 if (zoneid != GLOBAL_ZONEID && 6013 zoneid != connp->conn_zoneid) 6014 continue; 6015 6016 udp_report_item(mp->b_cont, udp); 6017 } 6018 } 6019 udp_last_ndd_get_info_time = ddi_get_lbolt(); 6020 return (0); 6021 } 6022 6023 /* 6024 * This routine creates a T_UDERROR_IND message and passes it upstream. 6025 * The address and options are copied from the T_UNITDATA_REQ message 6026 * passed in mp. This message is freed. 6027 */ 6028 static void 6029 udp_ud_err(queue_t *q, mblk_t *mp, uchar_t *destaddr, t_scalar_t destlen, 6030 t_scalar_t err) 6031 { 6032 struct T_unitdata_req *tudr; 6033 mblk_t *mp1; 6034 uchar_t *optaddr; 6035 t_scalar_t optlen; 6036 6037 if (DB_TYPE(mp) == M_DATA) { 6038 ASSERT(destaddr != NULL && destlen != 0); 6039 optaddr = NULL; 6040 optlen = 0; 6041 } else { 6042 if ((mp->b_wptr < mp->b_rptr) || 6043 (MBLKL(mp)) < sizeof (struct T_unitdata_req)) { 6044 goto done; 6045 } 6046 tudr = (struct T_unitdata_req *)mp->b_rptr; 6047 destaddr = mp->b_rptr + tudr->DEST_offset; 6048 if (destaddr < mp->b_rptr || destaddr >= mp->b_wptr || 6049 destaddr + tudr->DEST_length < mp->b_rptr || 6050 destaddr + tudr->DEST_length > mp->b_wptr) { 6051 goto done; 6052 } 6053 optaddr = mp->b_rptr + tudr->OPT_offset; 6054 if (optaddr < mp->b_rptr || optaddr >= mp->b_wptr || 6055 optaddr + tudr->OPT_length < mp->b_rptr || 6056 optaddr + tudr->OPT_length > mp->b_wptr) { 6057 goto done; 6058 } 6059 destlen = tudr->DEST_length; 6060 optlen = tudr->OPT_length; 6061 } 6062 6063 mp1 = mi_tpi_uderror_ind((char *)destaddr, destlen, 6064 (char *)optaddr, optlen, err); 6065 if (mp1 != NULL) 6066 putnext(UDP_RD(q), mp1); 6067 6068 done: 6069 freemsg(mp); 6070 } 6071 6072 /* 6073 * This routine removes a port number association from a stream. It 6074 * is called by udp_wput to handle T_UNBIND_REQ messages. 6075 */ 6076 static void 6077 udp_unbind(queue_t *q, mblk_t *mp) 6078 { 6079 udp_t *udp = Q_TO_UDP(q); 6080 6081 /* If a bind has not been done, we can't unbind. */ 6082 if (udp->udp_state == TS_UNBND) { 6083 udp_err_ack(q, mp, TOUTSTATE, 0); 6084 return; 6085 } 6086 if (cl_inet_unbind != NULL) { 6087 /* 6088 * Running in cluster mode - register unbind information 6089 */ 6090 if (udp->udp_ipversion == IPV4_VERSION) { 6091 (*cl_inet_unbind)(IPPROTO_UDP, AF_INET, 6092 (uint8_t *)(&V4_PART_OF_V6(udp->udp_v6src)), 6093 (in_port_t)udp->udp_port); 6094 } else { 6095 (*cl_inet_unbind)(IPPROTO_UDP, AF_INET6, 6096 (uint8_t *)&(udp->udp_v6src), 6097 (in_port_t)udp->udp_port); 6098 } 6099 } 6100 6101 udp_bind_hash_remove(udp, B_FALSE); 6102 V6_SET_ZERO(udp->udp_v6src); 6103 V6_SET_ZERO(udp->udp_bound_v6src); 6104 udp->udp_port = 0; 6105 udp->udp_state = TS_UNBND; 6106 6107 if (udp->udp_family == AF_INET6) { 6108 int error; 6109 6110 /* Rebuild the header template */ 6111 error = udp_build_hdrs(q, udp); 6112 if (error != 0) { 6113 udp_err_ack(q, mp, TSYSERR, error); 6114 return; 6115 } 6116 } 6117 /* 6118 * Pass the unbind to IP; T_UNBIND_REQ is larger than T_OK_ACK 6119 * and therefore ip_unbind must never return NULL. 6120 */ 6121 mp = ip_unbind(q, mp); 6122 ASSERT(mp != NULL); 6123 putnext(UDP_RD(q), mp); 6124 } 6125 6126 /* 6127 * Don't let port fall into the privileged range. 6128 * Since the extra privileged ports can be arbitrary we also 6129 * ensure that we exclude those from consideration. 6130 * udp_g_epriv_ports is not sorted thus we loop over it until 6131 * there are no changes. 6132 */ 6133 static in_port_t 6134 udp_update_next_port(udp_t *udp, in_port_t port, boolean_t random) 6135 { 6136 int i; 6137 in_port_t nextport; 6138 boolean_t restart = B_FALSE; 6139 6140 if (random && udp_random_anon_port != 0) { 6141 (void) random_get_pseudo_bytes((uint8_t *)&port, 6142 sizeof (in_port_t)); 6143 /* 6144 * Unless changed by a sys admin, the smallest anon port 6145 * is 32768 and the largest anon port is 65535. It is 6146 * very likely (50%) for the random port to be smaller 6147 * than the smallest anon port. When that happens, 6148 * add port % (anon port range) to the smallest anon 6149 * port to get the random port. It should fall into the 6150 * valid anon port range. 6151 */ 6152 if (port < udp_smallest_anon_port) { 6153 port = udp_smallest_anon_port + 6154 port % (udp_largest_anon_port - 6155 udp_smallest_anon_port); 6156 } 6157 } 6158 6159 retry: 6160 if (port < udp_smallest_anon_port) 6161 port = udp_smallest_anon_port; 6162 6163 if (port > udp_largest_anon_port) { 6164 port = udp_smallest_anon_port; 6165 if (restart) 6166 return (0); 6167 restart = B_TRUE; 6168 } 6169 6170 if (port < udp_smallest_nonpriv_port) 6171 port = udp_smallest_nonpriv_port; 6172 6173 for (i = 0; i < udp_g_num_epriv_ports; i++) { 6174 if (port == udp_g_epriv_ports[i]) { 6175 port++; 6176 /* 6177 * Make sure that the port is in the 6178 * valid range. 6179 */ 6180 goto retry; 6181 } 6182 } 6183 6184 if (is_system_labeled() && 6185 (nextport = tsol_next_port(crgetzone(udp->udp_connp->conn_cred), 6186 port, IPPROTO_UDP, B_TRUE)) != 0) { 6187 port = nextport; 6188 goto retry; 6189 } 6190 6191 return (port); 6192 } 6193 6194 static int 6195 udp_update_label(queue_t *wq, mblk_t *mp, ipaddr_t dst) 6196 { 6197 int err; 6198 uchar_t opt_storage[IP_MAX_OPT_LENGTH]; 6199 udp_t *udp = Q_TO_UDP(wq); 6200 6201 err = tsol_compute_label(DB_CREDDEF(mp, udp->udp_connp->conn_cred), dst, 6202 opt_storage, udp->udp_mac_exempt); 6203 if (err == 0) { 6204 err = tsol_update_options(&udp->udp_ip_snd_options, 6205 &udp->udp_ip_snd_options_len, &udp->udp_label_len, 6206 opt_storage); 6207 } 6208 if (err != 0) { 6209 DTRACE_PROBE4( 6210 tx__ip__log__info__updatelabel__udp, 6211 char *, "queue(1) failed to update options(2) on mp(3)", 6212 queue_t *, wq, char *, opt_storage, mblk_t *, mp); 6213 } else { 6214 IN6_IPADDR_TO_V4MAPPED(dst, &udp->udp_v6lastdst); 6215 } 6216 return (err); 6217 } 6218 6219 static mblk_t * 6220 udp_output_v4(conn_t *connp, mblk_t *mp, ipaddr_t v4dst, uint16_t port, 6221 uint_t srcid, int *error) 6222 { 6223 udp_t *udp = connp->conn_udp; 6224 queue_t *q = connp->conn_wq; 6225 mblk_t *mp1 = mp; 6226 mblk_t *mp2; 6227 ipha_t *ipha; 6228 int ip_hdr_length; 6229 uint32_t ip_len; 6230 udpha_t *udpha; 6231 udpattrs_t attrs; 6232 uchar_t ip_snd_opt[IP_MAX_OPT_LENGTH]; 6233 uint32_t ip_snd_opt_len = 0; 6234 6235 *error = 0; 6236 6237 if (v4dst == INADDR_ANY) 6238 v4dst = htonl(INADDR_LOOPBACK); 6239 6240 /* 6241 * If options passed in, feed it for verification and handling 6242 */ 6243 attrs.udpattr_credset = B_FALSE; 6244 if (DB_TYPE(mp) != M_DATA) { 6245 mp1 = mp->b_cont; 6246 if (((struct T_unitdata_req *)mp->b_rptr)->OPT_length != 0) { 6247 attrs.udpattr_ipp = NULL; 6248 attrs.udpattr_mb = mp; 6249 if (udp_unitdata_opt_process(q, mp, error, &attrs) < 0) 6250 goto done; 6251 /* 6252 * Note: success in processing options. 6253 * mp option buffer represented by 6254 * OPT_length/offset now potentially modified 6255 * and contain option setting results 6256 */ 6257 ASSERT(*error == 0); 6258 } 6259 } 6260 6261 /* mp1 points to the M_DATA mblk carrying the packet */ 6262 ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA); 6263 6264 /* 6265 * Check if our saved options are valid; update if not 6266 * TSOL Note: Since we are not in WRITER mode, UDP packets 6267 * to different destination may require different labels. 6268 * We use conn_lock to ensure that lastdst, ip_snd_options, 6269 * and ip_snd_options_len are consistent for the current 6270 * destination and are updated atomically. 6271 */ 6272 mutex_enter(&connp->conn_lock); 6273 if (is_system_labeled()) { 6274 /* Using UDP MLP requires SCM_UCRED from user */ 6275 if (connp->conn_mlp_type != mlptSingle && 6276 !attrs.udpattr_credset) { 6277 mutex_exit(&connp->conn_lock); 6278 DTRACE_PROBE4( 6279 tx__ip__log__info__output__udp, 6280 char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)", 6281 mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q); 6282 *error = ECONNREFUSED; 6283 goto done; 6284 } 6285 if ((!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6lastdst) || 6286 V4_PART_OF_V6(udp->udp_v6lastdst) != v4dst) && 6287 (*error = udp_update_label(q, mp, v4dst)) != 0) { 6288 mutex_exit(&connp->conn_lock); 6289 goto done; 6290 } 6291 } 6292 if (udp->udp_ip_snd_options_len > 0) { 6293 ip_snd_opt_len = udp->udp_ip_snd_options_len; 6294 bcopy(udp->udp_ip_snd_options, ip_snd_opt, ip_snd_opt_len); 6295 } 6296 mutex_exit(&connp->conn_lock); 6297 6298 /* Add an IP header */ 6299 ip_hdr_length = IP_SIMPLE_HDR_LENGTH + UDPH_SIZE + ip_snd_opt_len; 6300 ipha = (ipha_t *)&mp1->b_rptr[-ip_hdr_length]; 6301 if (DB_REF(mp1) != 1 || (uchar_t *)ipha < DB_BASE(mp1) || 6302 !OK_32PTR(ipha)) { 6303 mp2 = allocb(ip_hdr_length + udp_wroff_extra, BPRI_LO); 6304 if (mp2 == NULL) { 6305 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6306 "udp_wput_end: q %p (%S)", q, "allocbfail2"); 6307 *error = ENOMEM; 6308 goto done; 6309 } 6310 mp2->b_wptr = DB_LIM(mp2); 6311 mp2->b_cont = mp1; 6312 mp1 = mp2; 6313 if (DB_TYPE(mp) != M_DATA) 6314 mp->b_cont = mp1; 6315 else 6316 mp = mp1; 6317 6318 ipha = (ipha_t *)(mp1->b_wptr - ip_hdr_length); 6319 } 6320 ip_hdr_length -= UDPH_SIZE; 6321 #ifdef _BIG_ENDIAN 6322 /* Set version, header length, and tos */ 6323 *(uint16_t *)&ipha->ipha_version_and_hdr_length = 6324 ((((IP_VERSION << 4) | (ip_hdr_length>>2)) << 8) | 6325 udp->udp_type_of_service); 6326 /* Set ttl and protocol */ 6327 *(uint16_t *)&ipha->ipha_ttl = (udp->udp_ttl << 8) | IPPROTO_UDP; 6328 #else 6329 /* Set version, header length, and tos */ 6330 *(uint16_t *)&ipha->ipha_version_and_hdr_length = 6331 ((udp->udp_type_of_service << 8) | 6332 ((IP_VERSION << 4) | (ip_hdr_length>>2))); 6333 /* Set ttl and protocol */ 6334 *(uint16_t *)&ipha->ipha_ttl = (IPPROTO_UDP << 8) | udp->udp_ttl; 6335 #endif 6336 /* 6337 * Copy our address into the packet. If this is zero, 6338 * first look at __sin6_src_id for a hint. If we leave the source 6339 * as INADDR_ANY then ip will fill in the real source address. 6340 */ 6341 IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6src, ipha->ipha_src); 6342 if (srcid != 0 && ipha->ipha_src == INADDR_ANY) { 6343 in6_addr_t v6src; 6344 6345 ip_srcid_find_id(srcid, &v6src, connp->conn_zoneid); 6346 IN6_V4MAPPED_TO_IPADDR(&v6src, ipha->ipha_src); 6347 } 6348 6349 ipha->ipha_fragment_offset_and_flags = 0; 6350 ipha->ipha_ident = 0; 6351 6352 mp1->b_rptr = (uchar_t *)ipha; 6353 6354 ASSERT((uintptr_t)(mp1->b_wptr - (uchar_t *)ipha) <= 6355 (uintptr_t)UINT_MAX); 6356 6357 /* Determine length of packet */ 6358 ip_len = (uint32_t)(mp1->b_wptr - (uchar_t *)ipha); 6359 if ((mp2 = mp1->b_cont) != NULL) { 6360 do { 6361 ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX); 6362 ip_len += (uint32_t)MBLKL(mp2); 6363 } while ((mp2 = mp2->b_cont) != NULL); 6364 } 6365 /* 6366 * If the size of the packet is greater than the maximum allowed by 6367 * ip, return an error. Passing this down could cause panics because 6368 * the size will have wrapped and be inconsistent with the msg size. 6369 */ 6370 if (ip_len > IP_MAXPACKET) { 6371 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6372 "udp_wput_end: q %p (%S)", q, "IP length exceeded"); 6373 *error = EMSGSIZE; 6374 goto done; 6375 } 6376 ipha->ipha_length = htons((uint16_t)ip_len); 6377 ip_len -= ip_hdr_length; 6378 ip_len = htons((uint16_t)ip_len); 6379 udpha = (udpha_t *)(((uchar_t *)ipha) + ip_hdr_length); 6380 6381 /* 6382 * Copy in the destination address 6383 */ 6384 ipha->ipha_dst = v4dst; 6385 6386 /* 6387 * Set ttl based on IP_MULTICAST_TTL to match IPv6 logic. 6388 */ 6389 if (CLASSD(v4dst)) 6390 ipha->ipha_ttl = udp->udp_multicast_ttl; 6391 6392 udpha->uha_dst_port = port; 6393 udpha->uha_src_port = udp->udp_port; 6394 6395 if (ip_hdr_length > IP_SIMPLE_HDR_LENGTH) { 6396 uint32_t cksum; 6397 6398 bcopy(ip_snd_opt, &ipha[1], ip_snd_opt_len); 6399 /* 6400 * Massage source route putting first source route in ipha_dst. 6401 * Ignore the destination in T_unitdata_req. 6402 * Create a checksum adjustment for a source route, if any. 6403 */ 6404 cksum = ip_massage_options(ipha); 6405 cksum = (cksum & 0xFFFF) + (cksum >> 16); 6406 cksum -= ((ipha->ipha_dst >> 16) & 0xFFFF) + 6407 (ipha->ipha_dst & 0xFFFF); 6408 if ((int)cksum < 0) 6409 cksum--; 6410 cksum = (cksum & 0xFFFF) + (cksum >> 16); 6411 /* 6412 * IP does the checksum if uha_checksum is non-zero, 6413 * We make it easy for IP to include our pseudo header 6414 * by putting our length in uha_checksum. 6415 */ 6416 cksum += ip_len; 6417 cksum = (cksum & 0xFFFF) + (cksum >> 16); 6418 /* There might be a carry. */ 6419 cksum = (cksum & 0xFFFF) + (cksum >> 16); 6420 #ifdef _LITTLE_ENDIAN 6421 if (udp_do_checksum) 6422 ip_len = (cksum << 16) | ip_len; 6423 #else 6424 if (udp_do_checksum) 6425 ip_len = (ip_len << 16) | cksum; 6426 else 6427 ip_len <<= 16; 6428 #endif 6429 } else { 6430 /* 6431 * IP does the checksum if uha_checksum is non-zero, 6432 * We make it easy for IP to include our pseudo header 6433 * by putting our length in uha_checksum. 6434 */ 6435 if (udp_do_checksum) 6436 ip_len |= (ip_len << 16); 6437 #ifndef _LITTLE_ENDIAN 6438 else 6439 ip_len <<= 16; 6440 #endif 6441 } 6442 /* Set UDP length and checksum */ 6443 *((uint32_t *)&udpha->uha_length) = ip_len; 6444 if (DB_CRED(mp) != NULL) 6445 mblk_setcred(mp1, DB_CRED(mp)); 6446 6447 if (DB_TYPE(mp) != M_DATA) { 6448 ASSERT(mp != mp1); 6449 freeb(mp); 6450 } 6451 6452 /* mp has been consumed and we'll return success */ 6453 ASSERT(*error == 0); 6454 mp = NULL; 6455 6456 /* We're done. Pass the packet to ip. */ 6457 BUMP_MIB(&udp_mib, udpOutDatagrams); 6458 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6459 "udp_wput_end: q %p (%S)", q, "end"); 6460 6461 if ((connp->conn_flags & IPCL_CHECK_POLICY) != 0 || 6462 CONN_OUTBOUND_POLICY_PRESENT(connp) || 6463 connp->conn_dontroute || connp->conn_xmit_if_ill != NULL || 6464 connp->conn_nofailover_ill != NULL || 6465 connp->conn_outgoing_ill != NULL || 6466 ipha->ipha_version_and_hdr_length != IP_SIMPLE_HDR_VERSION || 6467 IPP_ENABLED(IPP_LOCAL_OUT) || ip_g_mrouter != NULL) { 6468 UDP_STAT(udp_ip_send); 6469 ip_output(connp, mp1, connp->conn_wq, IP_WPUT); 6470 } else { 6471 udp_send_data(udp, connp->conn_wq, mp1, ipha); 6472 } 6473 6474 done: 6475 if (*error != 0) { 6476 ASSERT(mp != NULL); 6477 BUMP_MIB(&udp_mib, udpOutErrors); 6478 } 6479 return (mp); 6480 } 6481 6482 static void 6483 udp_send_data(udp_t *udp, queue_t *q, mblk_t *mp, ipha_t *ipha) 6484 { 6485 conn_t *connp = udp->udp_connp; 6486 ipaddr_t src, dst; 6487 ill_t *ill; 6488 ire_t *ire; 6489 ipif_t *ipif = NULL; 6490 mblk_t *ire_fp_mp; 6491 uint_t ire_fp_mp_len; 6492 uint16_t *up; 6493 uint32_t cksum, hcksum_txflags; 6494 queue_t *dev_q; 6495 boolean_t retry_caching; 6496 6497 dst = ipha->ipha_dst; 6498 src = ipha->ipha_src; 6499 ASSERT(ipha->ipha_ident == 0); 6500 6501 if (CLASSD(dst)) { 6502 int err; 6503 6504 ipif = conn_get_held_ipif(connp, 6505 &connp->conn_multicast_ipif, &err); 6506 6507 if (ipif == NULL || ipif->ipif_isv6 || 6508 (ipif->ipif_ill->ill_phyint->phyint_flags & 6509 PHYI_LOOPBACK)) { 6510 if (ipif != NULL) 6511 ipif_refrele(ipif); 6512 UDP_STAT(udp_ip_send); 6513 ip_output(connp, mp, q, IP_WPUT); 6514 return; 6515 } 6516 } 6517 6518 retry_caching = B_FALSE; 6519 mutex_enter(&connp->conn_lock); 6520 ire = connp->conn_ire_cache; 6521 ASSERT(!(connp->conn_state_flags & CONN_INCIPIENT)); 6522 6523 if (ire == NULL || ire->ire_addr != dst || 6524 (ire->ire_marks & IRE_MARK_CONDEMNED)) { 6525 retry_caching = B_TRUE; 6526 } else if (CLASSD(dst) && (ire->ire_type & IRE_CACHE)) { 6527 ill_t *stq_ill = (ill_t *)ire->ire_stq->q_ptr; 6528 6529 ASSERT(ipif != NULL); 6530 if (stq_ill != ipif->ipif_ill && (stq_ill->ill_group == NULL || 6531 stq_ill->ill_group != ipif->ipif_ill->ill_group)) 6532 retry_caching = B_TRUE; 6533 } 6534 6535 if (!retry_caching) { 6536 ASSERT(ire != NULL); 6537 IRE_REFHOLD(ire); 6538 mutex_exit(&connp->conn_lock); 6539 } else { 6540 boolean_t cached = B_FALSE; 6541 6542 connp->conn_ire_cache = NULL; 6543 mutex_exit(&connp->conn_lock); 6544 6545 /* Release the old ire */ 6546 if (ire != NULL) { 6547 IRE_REFRELE_NOTR(ire); 6548 ire = NULL; 6549 } 6550 6551 if (CLASSD(dst)) { 6552 ASSERT(ipif != NULL); 6553 ire = ire_ctable_lookup(dst, 0, 0, ipif, 6554 connp->conn_zoneid, MBLK_GETLABEL(mp), 6555 MATCH_IRE_ILL_GROUP); 6556 } else { 6557 ASSERT(ipif == NULL); 6558 ire = ire_cache_lookup(dst, connp->conn_zoneid, 6559 MBLK_GETLABEL(mp)); 6560 } 6561 6562 if (ire == NULL) { 6563 if (ipif != NULL) 6564 ipif_refrele(ipif); 6565 UDP_STAT(udp_ire_null); 6566 ip_output(connp, mp, q, IP_WPUT); 6567 return; 6568 } 6569 IRE_REFHOLD_NOTR(ire); 6570 6571 mutex_enter(&connp->conn_lock); 6572 if (!(connp->conn_state_flags & CONN_CLOSING) && 6573 connp->conn_ire_cache == NULL) { 6574 rw_enter(&ire->ire_bucket->irb_lock, RW_READER); 6575 if (!(ire->ire_marks & IRE_MARK_CONDEMNED)) { 6576 connp->conn_ire_cache = ire; 6577 cached = B_TRUE; 6578 } 6579 rw_exit(&ire->ire_bucket->irb_lock); 6580 } 6581 mutex_exit(&connp->conn_lock); 6582 6583 /* 6584 * We can continue to use the ire but since it was not 6585 * cached, we should drop the extra reference. 6586 */ 6587 if (!cached) 6588 IRE_REFRELE_NOTR(ire); 6589 } 6590 ASSERT(ire != NULL && ire->ire_ipversion == IPV4_VERSION); 6591 ASSERT(!CLASSD(dst) || ipif != NULL); 6592 6593 /* 6594 * Check if we can take the fast-path. 6595 * Note that "incomplete" ire's (where the link-layer for next hop 6596 * is not resolved, or where the fast-path header in nce_fp_mp is not 6597 * available yet) are sent down the legacy (slow) path 6598 */ 6599 if ((ire->ire_type & (IRE_BROADCAST|IRE_LOCAL|IRE_LOOPBACK)) || 6600 (ire->ire_flags & RTF_MULTIRT) || ire->ire_stq == NULL || 6601 ire->ire_max_frag < ntohs(ipha->ipha_length) || 6602 (ire->ire_nce != NULL && 6603 (ire_fp_mp = ire->ire_nce->nce_fp_mp) == NULL) || 6604 (connp->conn_nexthop_set) || 6605 (ire_fp_mp_len = MBLKL(ire_fp_mp)) > MBLKHEAD(mp)) { 6606 if (ipif != NULL) 6607 ipif_refrele(ipif); 6608 UDP_STAT(udp_ip_ire_send); 6609 IRE_REFRELE(ire); 6610 ip_output(connp, mp, q, IP_WPUT); 6611 return; 6612 } 6613 6614 BUMP_MIB(&ip_mib, ipOutRequests); 6615 6616 ill = ire_to_ill(ire); 6617 ASSERT(ill != NULL); 6618 6619 dev_q = ire->ire_stq->q_next; 6620 ASSERT(dev_q != NULL); 6621 /* 6622 * If the service thread is already running, or if the driver 6623 * queue is currently flow-controlled, queue this packet. 6624 */ 6625 if ((q->q_first != NULL || connp->conn_draining) || 6626 ((dev_q->q_next || dev_q->q_first) && !canput(dev_q))) { 6627 if (ip_output_queue) { 6628 (void) putq(q, mp); 6629 } else { 6630 BUMP_MIB(&ip_mib, ipOutDiscards); 6631 freemsg(mp); 6632 } 6633 if (ipif != NULL) 6634 ipif_refrele(ipif); 6635 IRE_REFRELE(ire); 6636 return; 6637 } 6638 6639 ipha->ipha_ident = (uint16_t)atomic_add_32_nv(&ire->ire_ident, 1); 6640 #ifndef _BIG_ENDIAN 6641 ipha->ipha_ident = (ipha->ipha_ident << 8) | (ipha->ipha_ident >> 8); 6642 #endif 6643 6644 if (src == INADDR_ANY && !connp->conn_unspec_src) { 6645 if (CLASSD(dst) && !(ire->ire_flags & RTF_SETSRC)) 6646 src = ipha->ipha_src = ipif->ipif_src_addr; 6647 else 6648 src = ipha->ipha_src = ire->ire_src_addr; 6649 } 6650 6651 if (ILL_HCKSUM_CAPABLE(ill) && dohwcksum) { 6652 ASSERT(ill->ill_hcksum_capab != NULL); 6653 hcksum_txflags = ill->ill_hcksum_capab->ill_hcksum_txflags; 6654 } else { 6655 hcksum_txflags = 0; 6656 } 6657 6658 /* pseudo-header checksum (do it in parts for IP header checksum) */ 6659 cksum = (dst >> 16) + (dst & 0xFFFF) + (src >> 16) + (src & 0xFFFF); 6660 6661 ASSERT(ipha->ipha_version_and_hdr_length == IP_SIMPLE_HDR_VERSION); 6662 up = IPH_UDPH_CHECKSUMP(ipha, IP_SIMPLE_HDR_LENGTH); 6663 if (*up != 0) { 6664 IP_CKSUM_XMIT_FAST(ire->ire_ipversion, hcksum_txflags, 6665 mp, ipha, up, IPPROTO_UDP, IP_SIMPLE_HDR_LENGTH, 6666 ntohs(ipha->ipha_length), cksum); 6667 6668 /* Software checksum? */ 6669 if (DB_CKSUMFLAGS(mp) == 0) { 6670 UDP_STAT(udp_out_sw_cksum); 6671 UDP_STAT_UPDATE(udp_out_sw_cksum_bytes, 6672 ntohs(ipha->ipha_length) - IP_SIMPLE_HDR_LENGTH); 6673 } 6674 } 6675 6676 ipha->ipha_fragment_offset_and_flags |= 6677 (uint32_t)htons(ire->ire_frag_flag); 6678 6679 /* Calculate IP header checksum if hardware isn't capable */ 6680 if (!(DB_CKSUMFLAGS(mp) & HCK_IPV4_HDRCKSUM)) { 6681 IP_HDR_CKSUM(ipha, cksum, ((uint32_t *)ipha)[0], 6682 ((uint16_t *)ipha)[4]); 6683 } 6684 6685 if (CLASSD(dst)) { 6686 ilm_t *ilm; 6687 6688 ILM_WALKER_HOLD(ill); 6689 ilm = ilm_lookup_ill(ill, dst, ALL_ZONES); 6690 ILM_WALKER_RELE(ill); 6691 if (ilm != NULL) { 6692 ip_multicast_loopback(q, ill, mp, 6693 connp->conn_multicast_loop ? 0 : 6694 IP_FF_NO_MCAST_LOOP, connp->conn_zoneid); 6695 } 6696 6697 /* If multicast TTL is 0 then we are done */ 6698 if (ipha->ipha_ttl == 0) { 6699 if (ipif != NULL) 6700 ipif_refrele(ipif); 6701 freemsg(mp); 6702 IRE_REFRELE(ire); 6703 return; 6704 } 6705 } 6706 6707 ASSERT(DB_TYPE(ire_fp_mp) == M_DATA); 6708 mp->b_rptr = (uchar_t *)ipha - ire_fp_mp_len; 6709 bcopy(ire_fp_mp->b_rptr, mp->b_rptr, ire_fp_mp_len); 6710 6711 UPDATE_OB_PKT_COUNT(ire); 6712 ire->ire_last_used_time = lbolt; 6713 6714 if (ILL_DLS_CAPABLE(ill)) { 6715 /* 6716 * Send the packet directly to DLD, where it may be queued 6717 * depending on the availability of transmit resources at 6718 * the media layer. 6719 */ 6720 IP_DLS_ILL_TX(ill, mp); 6721 } else { 6722 putnext(ire->ire_stq, mp); 6723 } 6724 6725 if (ipif != NULL) 6726 ipif_refrele(ipif); 6727 IRE_REFRELE(ire); 6728 } 6729 6730 static boolean_t 6731 udp_update_label_v6(queue_t *wq, mblk_t *mp, in6_addr_t *dst) 6732 { 6733 udp_t *udp = Q_TO_UDP(wq); 6734 int err; 6735 uchar_t opt_storage[TSOL_MAX_IPV6_OPTION]; 6736 6737 err = tsol_compute_label_v6(DB_CREDDEF(mp, udp->udp_connp->conn_cred), 6738 dst, opt_storage, udp->udp_mac_exempt); 6739 if (err == 0) { 6740 err = tsol_update_sticky(&udp->udp_sticky_ipp, 6741 &udp->udp_label_len_v6, opt_storage); 6742 } 6743 if (err != 0) { 6744 DTRACE_PROBE4( 6745 tx__ip__log__drop__updatelabel__udp6, 6746 char *, "queue(1) failed to update options(2) on mp(3)", 6747 queue_t *, wq, char *, opt_storage, mblk_t *, mp); 6748 } else { 6749 udp->udp_v6lastdst = *dst; 6750 } 6751 return (err); 6752 } 6753 6754 /* 6755 * This routine handles all messages passed downstream. It either 6756 * consumes the message or passes it downstream; it never queues a 6757 * a message. 6758 */ 6759 static void 6760 udp_output(conn_t *connp, mblk_t *mp, struct sockaddr *addr, socklen_t addrlen) 6761 { 6762 sin6_t *sin6; 6763 sin_t *sin; 6764 ipaddr_t v4dst; 6765 uint16_t port; 6766 uint_t srcid; 6767 queue_t *q = connp->conn_wq; 6768 udp_t *udp = connp->conn_udp; 6769 int error = 0; 6770 struct sockaddr_storage ss; 6771 6772 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_START, 6773 "udp_wput_start: connp %p mp %p", connp, mp); 6774 6775 /* 6776 * We directly handle several cases here: T_UNITDATA_REQ message 6777 * coming down as M_PROTO/M_PCPROTO and M_DATA messages for both 6778 * connected and non-connected socket. The latter carries the 6779 * address structure along when this routine gets called. 6780 */ 6781 switch (DB_TYPE(mp)) { 6782 case M_DATA: 6783 if (!udp->udp_direct_sockfs || udp->udp_state != TS_DATA_XFER) { 6784 if (!udp->udp_direct_sockfs || 6785 addr == NULL || addrlen == 0) { 6786 /* Not connected; address is required */ 6787 BUMP_MIB(&udp_mib, udpOutErrors); 6788 UDP_STAT(udp_out_err_notconn); 6789 freemsg(mp); 6790 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6791 "udp_wput_end: connp %p (%S)", connp, 6792 "not-connected; address required"); 6793 return; 6794 } 6795 ASSERT(udp->udp_issocket); 6796 UDP_DBGSTAT(udp_data_notconn); 6797 /* Not connected; do some more checks below */ 6798 break; 6799 } 6800 /* M_DATA for connected socket */ 6801 UDP_DBGSTAT(udp_data_conn); 6802 IN6_V4MAPPED_TO_IPADDR(&udp->udp_v6dst, v4dst); 6803 6804 /* Initialize addr and addrlen as if they're passed in */ 6805 if (udp->udp_family == AF_INET) { 6806 sin = (sin_t *)&ss; 6807 sin->sin_family = AF_INET; 6808 sin->sin_port = udp->udp_dstport; 6809 sin->sin_addr.s_addr = v4dst; 6810 addr = (struct sockaddr *)sin; 6811 addrlen = sizeof (*sin); 6812 } else { 6813 sin6 = (sin6_t *)&ss; 6814 sin6->sin6_family = AF_INET6; 6815 sin6->sin6_port = udp->udp_dstport; 6816 sin6->sin6_flowinfo = udp->udp_flowinfo; 6817 sin6->sin6_addr = udp->udp_v6dst; 6818 sin6->sin6_scope_id = 0; 6819 sin6->__sin6_src_id = 0; 6820 addr = (struct sockaddr *)sin6; 6821 addrlen = sizeof (*sin6); 6822 } 6823 6824 if (udp->udp_family == AF_INET || 6825 IN6_IS_ADDR_V4MAPPED(&udp->udp_v6dst)) { 6826 /* 6827 * Handle both AF_INET and AF_INET6; the latter 6828 * for IPV4 mapped destination addresses. Note 6829 * here that both addr and addrlen point to the 6830 * corresponding struct depending on the address 6831 * family of the socket. 6832 */ 6833 mp = udp_output_v4(connp, mp, v4dst, 6834 udp->udp_dstport, 0, &error); 6835 } else { 6836 mp = udp_output_v6(connp, mp, sin6, &error); 6837 } 6838 if (error != 0) { 6839 ASSERT(addr != NULL && addrlen != 0); 6840 goto ud_error; 6841 } 6842 return; 6843 case M_PROTO: 6844 case M_PCPROTO: { 6845 struct T_unitdata_req *tudr; 6846 6847 ASSERT((uintptr_t)MBLKL(mp) <= (uintptr_t)INT_MAX); 6848 tudr = (struct T_unitdata_req *)mp->b_rptr; 6849 6850 /* Handle valid T_UNITDATA_REQ here */ 6851 if (MBLKL(mp) >= sizeof (*tudr) && 6852 ((t_primp_t)mp->b_rptr)->type == T_UNITDATA_REQ) { 6853 if (mp->b_cont == NULL) { 6854 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6855 "udp_wput_end: q %p (%S)", q, "badaddr"); 6856 error = EPROTO; 6857 goto ud_error; 6858 } 6859 6860 if (!MBLKIN(mp, 0, tudr->DEST_offset + 6861 tudr->DEST_length)) { 6862 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6863 "udp_wput_end: q %p (%S)", q, "badaddr"); 6864 error = EADDRNOTAVAIL; 6865 goto ud_error; 6866 } 6867 /* 6868 * If a port has not been bound to the stream, fail. 6869 * This is not a problem when sockfs is directly 6870 * above us, because it will ensure that the socket 6871 * is first bound before allowing data to be sent. 6872 */ 6873 if (udp->udp_state == TS_UNBND) { 6874 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6875 "udp_wput_end: q %p (%S)", q, "outstate"); 6876 error = EPROTO; 6877 goto ud_error; 6878 } 6879 addr = (struct sockaddr *) 6880 &mp->b_rptr[tudr->DEST_offset]; 6881 addrlen = tudr->DEST_length; 6882 if (tudr->OPT_length != 0) 6883 UDP_STAT(udp_out_opt); 6884 break; 6885 } 6886 /* FALLTHRU */ 6887 } 6888 default: 6889 udp_become_writer(connp, mp, udp_wput_other_wrapper, 6890 SQTAG_UDP_OUTPUT); 6891 return; 6892 } 6893 ASSERT(addr != NULL); 6894 6895 switch (udp->udp_family) { 6896 case AF_INET6: 6897 sin6 = (sin6_t *)addr; 6898 if (!OK_32PTR((char *)sin6) || addrlen != sizeof (sin6_t) || 6899 sin6->sin6_family != AF_INET6) { 6900 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6901 "udp_wput_end: q %p (%S)", q, "badaddr"); 6902 error = EADDRNOTAVAIL; 6903 goto ud_error; 6904 } 6905 6906 if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { 6907 /* 6908 * Destination is a non-IPv4-compatible IPv6 address. 6909 * Send out an IPv6 format packet. 6910 */ 6911 mp = udp_output_v6(connp, mp, sin6, &error); 6912 if (error != 0) 6913 goto ud_error; 6914 6915 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6916 "udp_wput_end: q %p (%S)", q, "udp_output_v6"); 6917 return; 6918 } 6919 /* 6920 * If the local address is not zero or a mapped address 6921 * return an error. It would be possible to send an IPv4 6922 * packet but the response would never make it back to the 6923 * application since it is bound to a non-mapped address. 6924 */ 6925 if (!IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src) && 6926 !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) { 6927 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6928 "udp_wput_end: q %p (%S)", q, "badaddr"); 6929 error = EADDRNOTAVAIL; 6930 goto ud_error; 6931 } 6932 /* Send IPv4 packet without modifying udp_ipversion */ 6933 /* Extract port and ipaddr */ 6934 port = sin6->sin6_port; 6935 IN6_V4MAPPED_TO_IPADDR(&sin6->sin6_addr, v4dst); 6936 srcid = sin6->__sin6_src_id; 6937 break; 6938 6939 case AF_INET: 6940 sin = (sin_t *)addr; 6941 if (!OK_32PTR((char *)sin) || addrlen != sizeof (sin_t) || 6942 sin->sin_family != AF_INET) { 6943 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_END, 6944 "udp_wput_end: q %p (%S)", q, "badaddr"); 6945 error = EADDRNOTAVAIL; 6946 goto ud_error; 6947 } 6948 /* Extract port and ipaddr */ 6949 port = sin->sin_port; 6950 v4dst = sin->sin_addr.s_addr; 6951 srcid = 0; 6952 break; 6953 } 6954 6955 mp = udp_output_v4(connp, mp, v4dst, port, srcid, &error); 6956 if (error != 0) { 6957 ud_error: 6958 UDP_STAT(udp_out_err_output); 6959 ASSERT(mp != NULL); 6960 /* mp is freed by the following routine */ 6961 udp_ud_err(q, mp, (uchar_t *)addr, (t_scalar_t)addrlen, 6962 (t_scalar_t)error); 6963 } 6964 } 6965 6966 /* ARGSUSED */ 6967 static void 6968 udp_output_wrapper(void *arg, mblk_t *mp, void *arg2) 6969 { 6970 udp_output((conn_t *)arg, mp, NULL, 0); 6971 _UDP_EXIT((conn_t *)arg); 6972 } 6973 6974 static void 6975 udp_wput(queue_t *q, mblk_t *mp) 6976 { 6977 _UDP_ENTER(Q_TO_CONN(UDP_WR(q)), mp, udp_output_wrapper, 6978 SQTAG_UDP_WPUT); 6979 } 6980 6981 /* 6982 * Allocate and prepare a T_UNITDATA_REQ message. 6983 */ 6984 static mblk_t * 6985 udp_tudr_alloc(struct sockaddr *addr, socklen_t addrlen) 6986 { 6987 struct T_unitdata_req *tudr; 6988 mblk_t *mp; 6989 6990 mp = allocb(sizeof (*tudr) + addrlen, BPRI_MED); 6991 if (mp != NULL) { 6992 mp->b_wptr += sizeof (*tudr) + addrlen; 6993 DB_TYPE(mp) = M_PROTO; 6994 6995 tudr = (struct T_unitdata_req *)mp->b_rptr; 6996 tudr->PRIM_type = T_UNITDATA_REQ; 6997 tudr->DEST_length = addrlen; 6998 tudr->DEST_offset = (t_scalar_t)sizeof (*tudr); 6999 tudr->OPT_length = 0; 7000 tudr->OPT_offset = 0; 7001 bcopy(addr, tudr+1, addrlen); 7002 } 7003 return (mp); 7004 } 7005 7006 /* 7007 * Entry point for sockfs when udp is in "direct sockfs" mode. This mode 7008 * is valid when we are directly beneath the stream head, and thus sockfs 7009 * is able to bypass STREAMS and directly call us, passing along the sockaddr 7010 * structure without the cumbersome T_UNITDATA_REQ interface. Note that 7011 * this is done for both connected and non-connected endpoint. 7012 */ 7013 void 7014 udp_wput_data(queue_t *q, mblk_t *mp, struct sockaddr *addr, socklen_t addrlen) 7015 { 7016 conn_t *connp; 7017 udp_t *udp; 7018 7019 q = UDP_WR(q); 7020 connp = Q_TO_CONN(q); 7021 udp = connp->conn_udp; 7022 7023 /* udpsockfs should only send down M_DATA for this entry point */ 7024 ASSERT(DB_TYPE(mp) == M_DATA); 7025 7026 mutex_enter(&connp->conn_lock); 7027 UDP_MODE_ASSERTIONS(udp, UDP_ENTER); 7028 7029 if (udp->udp_mode != UDP_MT_HOT) { 7030 /* 7031 * We can't enter this conn right away because another 7032 * thread is currently executing as writer; therefore we 7033 * need to deposit the message into the squeue to be 7034 * drained later. If a socket address is present, we 7035 * need to create a T_UNITDATA_REQ message as placeholder. 7036 */ 7037 if (addr != NULL && addrlen != 0) { 7038 mblk_t *tudr_mp = udp_tudr_alloc(addr, addrlen); 7039 7040 if (tudr_mp == NULL) { 7041 mutex_exit(&connp->conn_lock); 7042 BUMP_MIB(&udp_mib, udpOutErrors); 7043 UDP_STAT(udp_out_err_tudr); 7044 freemsg(mp); 7045 return; 7046 } 7047 /* Tag the packet with T_UNITDATA_REQ */ 7048 tudr_mp->b_cont = mp; 7049 mp = tudr_mp; 7050 } 7051 mutex_exit(&connp->conn_lock); 7052 udp_enter(connp, mp, udp_output_wrapper, SQTAG_UDP_WPUT); 7053 return; 7054 } 7055 7056 /* We can execute as reader right away. */ 7057 UDP_READERS_INCREF(udp); 7058 mutex_exit(&connp->conn_lock); 7059 7060 udp_output(connp, mp, addr, addrlen); 7061 7062 udp_exit(connp); 7063 } 7064 7065 /* 7066 * udp_output_v6(): 7067 * Assumes that udp_wput did some sanity checking on the destination 7068 * address. 7069 */ 7070 static mblk_t * 7071 udp_output_v6(conn_t *connp, mblk_t *mp, sin6_t *sin6, int *error) 7072 { 7073 ip6_t *ip6h; 7074 ip6i_t *ip6i; /* mp1->b_rptr even if no ip6i_t */ 7075 mblk_t *mp1 = mp; 7076 mblk_t *mp2; 7077 int udp_ip_hdr_len = IPV6_HDR_LEN + UDPH_SIZE; 7078 size_t ip_len; 7079 udpha_t *udph; 7080 udp_t *udp = connp->conn_udp; 7081 queue_t *q = connp->conn_wq; 7082 ip6_pkt_t ipp_s; /* For ancillary data options */ 7083 ip6_pkt_t *ipp = &ipp_s; 7084 ip6_pkt_t *tipp; /* temporary ipp */ 7085 uint32_t csum = 0; 7086 uint_t ignore = 0; 7087 uint_t option_exists = 0, is_sticky = 0; 7088 uint8_t *cp; 7089 uint8_t *nxthdr_ptr; 7090 in6_addr_t ip6_dst; 7091 udpattrs_t attrs; 7092 boolean_t opt_present; 7093 ip6_hbh_t *hopoptsptr = NULL; 7094 uint_t hopoptslen = 0; 7095 boolean_t is_ancillary = B_FALSE; 7096 7097 *error = 0; 7098 7099 /* 7100 * If the local address is a mapped address return 7101 * an error. 7102 * It would be possible to send an IPv6 packet but the 7103 * response would never make it back to the application 7104 * since it is bound to a mapped address. 7105 */ 7106 if (IN6_IS_ADDR_V4MAPPED(&udp->udp_v6src)) { 7107 *error = EADDRNOTAVAIL; 7108 goto done; 7109 } 7110 7111 ipp->ipp_fields = 0; 7112 ipp->ipp_sticky_ignored = 0; 7113 7114 /* 7115 * If TPI options passed in, feed it for verification and handling 7116 */ 7117 attrs.udpattr_credset = B_FALSE; 7118 opt_present = B_FALSE; 7119 if (DB_TYPE(mp) != M_DATA) { 7120 mp1 = mp->b_cont; 7121 if (((struct T_unitdata_req *)mp->b_rptr)->OPT_length != 0) { 7122 attrs.udpattr_ipp = ipp; 7123 attrs.udpattr_mb = mp; 7124 if (udp_unitdata_opt_process(q, mp, error, &attrs) < 0) 7125 goto done; 7126 ASSERT(*error == 0); 7127 opt_present = B_TRUE; 7128 } 7129 } 7130 ignore = ipp->ipp_sticky_ignored; 7131 7132 /* mp1 points to the M_DATA mblk carrying the packet */ 7133 ASSERT(mp1 != NULL && DB_TYPE(mp1) == M_DATA); 7134 7135 if (sin6->sin6_scope_id != 0 && 7136 IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) { 7137 /* 7138 * IPPF_SCOPE_ID is special. It's neither a sticky 7139 * option nor ancillary data. It needs to be 7140 * explicitly set in options_exists. 7141 */ 7142 option_exists |= IPPF_SCOPE_ID; 7143 } 7144 7145 /* 7146 * Compute the destination address 7147 */ 7148 ip6_dst = sin6->sin6_addr; 7149 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) 7150 ip6_dst = ipv6_loopback; 7151 7152 /* 7153 * If we're not going to the same destination as last time, then 7154 * recompute the label required. This is done in a separate routine to 7155 * avoid blowing up our stack here. 7156 * 7157 * TSOL Note: Since we are not in WRITER mode, UDP packets 7158 * to different destination may require different labels. 7159 * We use conn_lock to ensure that lastdst, sticky ipp_hopopts, 7160 * and sticky ipp_hopoptslen are consistent for the current 7161 * destination and are updated atomically. 7162 */ 7163 mutex_enter(&connp->conn_lock); 7164 if (is_system_labeled()) { 7165 /* Using UDP MLP requires SCM_UCRED from user */ 7166 if (connp->conn_mlp_type != mlptSingle && 7167 !attrs.udpattr_credset) { 7168 DTRACE_PROBE4( 7169 tx__ip__log__info__output__udp6, 7170 char *, "MLP mp(1) lacks SCM_UCRED attr(2) on q(3)", 7171 mblk_t *, mp1, udpattrs_t *, &attrs, queue_t *, q); 7172 *error = ECONNREFUSED; 7173 mutex_exit(&connp->conn_lock); 7174 goto done; 7175 } 7176 if ((opt_present || 7177 !IN6_ARE_ADDR_EQUAL(&udp->udp_v6lastdst, &ip6_dst)) && 7178 (*error = udp_update_label_v6(q, mp, &ip6_dst)) != 0) { 7179 mutex_exit(&connp->conn_lock); 7180 goto done; 7181 } 7182 } 7183 7184 /* 7185 * If there's a security label here, then we ignore any options the 7186 * user may try to set. We keep the peer's label as a hidden sticky 7187 * option. We make a private copy of this label before releasing the 7188 * lock so that label is kept consistent with the destination addr. 7189 */ 7190 if (udp->udp_label_len_v6 > 0) { 7191 ignore &= ~IPPF_HOPOPTS; 7192 ipp->ipp_fields &= ~IPPF_HOPOPTS; 7193 } 7194 7195 if ((udp->udp_sticky_ipp.ipp_fields == 0) && (ipp->ipp_fields == 0)) { 7196 /* No sticky options nor ancillary data. */ 7197 mutex_exit(&connp->conn_lock); 7198 goto no_options; 7199 } 7200 7201 /* 7202 * Go through the options figuring out where each is going to 7203 * come from and build two masks. The first mask indicates if 7204 * the option exists at all. The second mask indicates if the 7205 * option is sticky or ancillary. 7206 */ 7207 if (!(ignore & IPPF_HOPOPTS)) { 7208 if (ipp->ipp_fields & IPPF_HOPOPTS) { 7209 option_exists |= IPPF_HOPOPTS; 7210 udp_ip_hdr_len += ipp->ipp_hopoptslen; 7211 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_HOPOPTS) { 7212 option_exists |= IPPF_HOPOPTS; 7213 is_sticky |= IPPF_HOPOPTS; 7214 ASSERT(udp->udp_sticky_ipp.ipp_hopoptslen != 0); 7215 hopoptsptr = kmem_alloc( 7216 udp->udp_sticky_ipp.ipp_hopoptslen, KM_NOSLEEP); 7217 if (hopoptsptr == NULL) { 7218 *error = ENOMEM; 7219 mutex_exit(&connp->conn_lock); 7220 goto done; 7221 } 7222 hopoptslen = udp->udp_sticky_ipp.ipp_hopoptslen; 7223 bcopy(udp->udp_sticky_ipp.ipp_hopopts, hopoptsptr, 7224 hopoptslen); 7225 udp_ip_hdr_len += hopoptslen; 7226 } 7227 } 7228 mutex_exit(&connp->conn_lock); 7229 7230 if (!(ignore & IPPF_RTHDR)) { 7231 if (ipp->ipp_fields & IPPF_RTHDR) { 7232 option_exists |= IPPF_RTHDR; 7233 udp_ip_hdr_len += ipp->ipp_rthdrlen; 7234 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTHDR) { 7235 option_exists |= IPPF_RTHDR; 7236 is_sticky |= IPPF_RTHDR; 7237 udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rthdrlen; 7238 } 7239 } 7240 7241 if (!(ignore & IPPF_RTDSTOPTS) && (option_exists & IPPF_RTHDR)) { 7242 if (ipp->ipp_fields & IPPF_RTDSTOPTS) { 7243 option_exists |= IPPF_RTDSTOPTS; 7244 udp_ip_hdr_len += ipp->ipp_rtdstoptslen; 7245 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_RTDSTOPTS) { 7246 option_exists |= IPPF_RTDSTOPTS; 7247 is_sticky |= IPPF_RTDSTOPTS; 7248 udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_rtdstoptslen; 7249 } 7250 } 7251 7252 if (!(ignore & IPPF_DSTOPTS)) { 7253 if (ipp->ipp_fields & IPPF_DSTOPTS) { 7254 option_exists |= IPPF_DSTOPTS; 7255 udp_ip_hdr_len += ipp->ipp_dstoptslen; 7256 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DSTOPTS) { 7257 option_exists |= IPPF_DSTOPTS; 7258 is_sticky |= IPPF_DSTOPTS; 7259 udp_ip_hdr_len += udp->udp_sticky_ipp.ipp_dstoptslen; 7260 } 7261 } 7262 7263 if (!(ignore & IPPF_IFINDEX)) { 7264 if (ipp->ipp_fields & IPPF_IFINDEX) { 7265 option_exists |= IPPF_IFINDEX; 7266 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_IFINDEX) { 7267 option_exists |= IPPF_IFINDEX; 7268 is_sticky |= IPPF_IFINDEX; 7269 } 7270 } 7271 7272 if (!(ignore & IPPF_ADDR)) { 7273 if (ipp->ipp_fields & IPPF_ADDR) { 7274 option_exists |= IPPF_ADDR; 7275 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_ADDR) { 7276 option_exists |= IPPF_ADDR; 7277 is_sticky |= IPPF_ADDR; 7278 } 7279 } 7280 7281 if (!(ignore & IPPF_DONTFRAG)) { 7282 if (ipp->ipp_fields & IPPF_DONTFRAG) { 7283 option_exists |= IPPF_DONTFRAG; 7284 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_DONTFRAG) { 7285 option_exists |= IPPF_DONTFRAG; 7286 is_sticky |= IPPF_DONTFRAG; 7287 } 7288 } 7289 7290 if (!(ignore & IPPF_USE_MIN_MTU)) { 7291 if (ipp->ipp_fields & IPPF_USE_MIN_MTU) { 7292 option_exists |= IPPF_USE_MIN_MTU; 7293 } else if (udp->udp_sticky_ipp.ipp_fields & 7294 IPPF_USE_MIN_MTU) { 7295 option_exists |= IPPF_USE_MIN_MTU; 7296 is_sticky |= IPPF_USE_MIN_MTU; 7297 } 7298 } 7299 7300 if (!(ignore & IPPF_HOPLIMIT) && (ipp->ipp_fields & IPPF_HOPLIMIT)) 7301 option_exists |= IPPF_HOPLIMIT; 7302 /* IPV6_HOPLIMIT can never be sticky */ 7303 ASSERT(!(udp->udp_sticky_ipp.ipp_fields & IPPF_HOPLIMIT)); 7304 7305 if (!(ignore & IPPF_UNICAST_HOPS) && 7306 (udp->udp_sticky_ipp.ipp_fields & IPPF_UNICAST_HOPS)) { 7307 option_exists |= IPPF_UNICAST_HOPS; 7308 is_sticky |= IPPF_UNICAST_HOPS; 7309 } 7310 7311 if (!(ignore & IPPF_MULTICAST_HOPS) && 7312 (udp->udp_sticky_ipp.ipp_fields & IPPF_MULTICAST_HOPS)) { 7313 option_exists |= IPPF_MULTICAST_HOPS; 7314 is_sticky |= IPPF_MULTICAST_HOPS; 7315 } 7316 7317 if (!(ignore & IPPF_TCLASS)) { 7318 if (ipp->ipp_fields & IPPF_TCLASS) { 7319 option_exists |= IPPF_TCLASS; 7320 } else if (udp->udp_sticky_ipp.ipp_fields & IPPF_TCLASS) { 7321 option_exists |= IPPF_TCLASS; 7322 is_sticky |= IPPF_TCLASS; 7323 } 7324 } 7325 7326 no_options: 7327 7328 /* 7329 * If any options carried in the ip6i_t were specified, we 7330 * need to account for the ip6i_t in the data we'll be sending 7331 * down. 7332 */ 7333 if (option_exists & IPPF_HAS_IP6I) 7334 udp_ip_hdr_len += sizeof (ip6i_t); 7335 7336 /* check/fix buffer config, setup pointers into it */ 7337 ip6h = (ip6_t *)&mp1->b_rptr[-udp_ip_hdr_len]; 7338 if (DB_REF(mp1) != 1 || ((unsigned char *)ip6h < DB_BASE(mp1)) || 7339 !OK_32PTR(ip6h)) { 7340 /* Try to get everything in a single mblk next time */ 7341 if (udp_ip_hdr_len > udp->udp_max_hdr_len) { 7342 udp->udp_max_hdr_len = udp_ip_hdr_len; 7343 (void) mi_set_sth_wroff(UDP_RD(q), 7344 udp->udp_max_hdr_len + udp_wroff_extra); 7345 } 7346 mp2 = allocb(udp_ip_hdr_len + udp_wroff_extra, BPRI_LO); 7347 if (mp2 == NULL) { 7348 *error = ENOMEM; 7349 goto done; 7350 } 7351 mp2->b_wptr = DB_LIM(mp2); 7352 mp2->b_cont = mp1; 7353 mp1 = mp2; 7354 if (DB_TYPE(mp) != M_DATA) 7355 mp->b_cont = mp1; 7356 else 7357 mp = mp1; 7358 7359 ip6h = (ip6_t *)(mp1->b_wptr - udp_ip_hdr_len); 7360 } 7361 mp1->b_rptr = (unsigned char *)ip6h; 7362 ip6i = (ip6i_t *)ip6h; 7363 7364 #define ANCIL_OR_STICKY_PTR(f) ((is_sticky & f) ? &udp->udp_sticky_ipp : ipp) 7365 if (option_exists & IPPF_HAS_IP6I) { 7366 ip6h = (ip6_t *)&ip6i[1]; 7367 ip6i->ip6i_flags = 0; 7368 ip6i->ip6i_vcf = IPV6_DEFAULT_VERS_AND_FLOW; 7369 7370 /* sin6_scope_id takes precendence over IPPF_IFINDEX */ 7371 if (option_exists & IPPF_SCOPE_ID) { 7372 ip6i->ip6i_flags |= IP6I_IFINDEX; 7373 ip6i->ip6i_ifindex = sin6->sin6_scope_id; 7374 } else if (option_exists & IPPF_IFINDEX) { 7375 tipp = ANCIL_OR_STICKY_PTR(IPPF_IFINDEX); 7376 ASSERT(tipp->ipp_ifindex != 0); 7377 ip6i->ip6i_flags |= IP6I_IFINDEX; 7378 ip6i->ip6i_ifindex = tipp->ipp_ifindex; 7379 } 7380 7381 if (option_exists & IPPF_ADDR) { 7382 /* 7383 * Enable per-packet source address verification if 7384 * IPV6_PKTINFO specified the source address. 7385 * ip6_src is set in the transport's _wput function. 7386 */ 7387 ip6i->ip6i_flags |= IP6I_VERIFY_SRC; 7388 } 7389 7390 if (option_exists & IPPF_DONTFRAG) { 7391 ip6i->ip6i_flags |= IP6I_DONTFRAG; 7392 } 7393 7394 if (option_exists & IPPF_USE_MIN_MTU) { 7395 ip6i->ip6i_flags = IP6I_API_USE_MIN_MTU( 7396 ip6i->ip6i_flags, ipp->ipp_use_min_mtu); 7397 } 7398 7399 if (option_exists & IPPF_NEXTHOP) { 7400 tipp = ANCIL_OR_STICKY_PTR(IPPF_NEXTHOP); 7401 ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_nexthop)); 7402 ip6i->ip6i_flags |= IP6I_NEXTHOP; 7403 ip6i->ip6i_nexthop = tipp->ipp_nexthop; 7404 } 7405 7406 /* 7407 * tell IP this is an ip6i_t private header 7408 */ 7409 ip6i->ip6i_nxt = IPPROTO_RAW; 7410 } 7411 7412 /* Initialize IPv6 header */ 7413 ip6h->ip6_vcf = IPV6_DEFAULT_VERS_AND_FLOW; 7414 bzero(&ip6h->ip6_src, sizeof (ip6h->ip6_src)); 7415 7416 /* Set the hoplimit of the outgoing packet. */ 7417 if (option_exists & IPPF_HOPLIMIT) { 7418 /* IPV6_HOPLIMIT ancillary data overrides all other settings. */ 7419 ip6h->ip6_hops = ipp->ipp_hoplimit; 7420 ip6i->ip6i_flags |= IP6I_HOPLIMIT; 7421 } else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { 7422 ip6h->ip6_hops = udp->udp_multicast_ttl; 7423 if (option_exists & IPPF_MULTICAST_HOPS) 7424 ip6i->ip6i_flags |= IP6I_HOPLIMIT; 7425 } else { 7426 ip6h->ip6_hops = udp->udp_ttl; 7427 if (option_exists & IPPF_UNICAST_HOPS) 7428 ip6i->ip6i_flags |= IP6I_HOPLIMIT; 7429 } 7430 7431 if (option_exists & IPPF_ADDR) { 7432 tipp = ANCIL_OR_STICKY_PTR(IPPF_ADDR); 7433 ASSERT(!IN6_IS_ADDR_UNSPECIFIED(&tipp->ipp_addr)); 7434 ip6h->ip6_src = tipp->ipp_addr; 7435 } else { 7436 /* 7437 * The source address was not set using IPV6_PKTINFO. 7438 * First look at the bound source. 7439 * If unspecified fallback to __sin6_src_id. 7440 */ 7441 ip6h->ip6_src = udp->udp_v6src; 7442 if (sin6->__sin6_src_id != 0 && 7443 IN6_IS_ADDR_UNSPECIFIED(&ip6h->ip6_src)) { 7444 ip_srcid_find_id(sin6->__sin6_src_id, 7445 &ip6h->ip6_src, connp->conn_zoneid); 7446 } 7447 } 7448 7449 nxthdr_ptr = (uint8_t *)&ip6h->ip6_nxt; 7450 cp = (uint8_t *)&ip6h[1]; 7451 7452 /* 7453 * Here's where we have to start stringing together 7454 * any extension headers in the right order: 7455 * Hop-by-hop, destination, routing, and final destination opts. 7456 */ 7457 if (option_exists & IPPF_HOPOPTS) { 7458 /* Hop-by-hop options */ 7459 ip6_hbh_t *hbh = (ip6_hbh_t *)cp; 7460 tipp = ANCIL_OR_STICKY_PTR(IPPF_HOPOPTS); 7461 if (hopoptslen == 0) { 7462 hopoptsptr = tipp->ipp_hopopts; 7463 hopoptslen = tipp->ipp_hopoptslen; 7464 is_ancillary = B_TRUE; 7465 } 7466 7467 *nxthdr_ptr = IPPROTO_HOPOPTS; 7468 nxthdr_ptr = &hbh->ip6h_nxt; 7469 7470 bcopy(hopoptsptr, cp, hopoptslen); 7471 cp += hopoptslen; 7472 7473 if (hopoptsptr != NULL && !is_ancillary) { 7474 kmem_free(hopoptsptr, hopoptslen); 7475 hopoptsptr = NULL; 7476 hopoptslen = 0; 7477 } 7478 } 7479 /* 7480 * En-route destination options 7481 * Only do them if there's a routing header as well 7482 */ 7483 if (option_exists & IPPF_RTDSTOPTS) { 7484 ip6_dest_t *dst = (ip6_dest_t *)cp; 7485 tipp = ANCIL_OR_STICKY_PTR(IPPF_RTDSTOPTS); 7486 7487 *nxthdr_ptr = IPPROTO_DSTOPTS; 7488 nxthdr_ptr = &dst->ip6d_nxt; 7489 7490 bcopy(tipp->ipp_rtdstopts, cp, tipp->ipp_rtdstoptslen); 7491 cp += tipp->ipp_rtdstoptslen; 7492 } 7493 /* 7494 * Routing header next 7495 */ 7496 if (option_exists & IPPF_RTHDR) { 7497 ip6_rthdr_t *rt = (ip6_rthdr_t *)cp; 7498 tipp = ANCIL_OR_STICKY_PTR(IPPF_RTHDR); 7499 7500 *nxthdr_ptr = IPPROTO_ROUTING; 7501 nxthdr_ptr = &rt->ip6r_nxt; 7502 7503 bcopy(tipp->ipp_rthdr, cp, tipp->ipp_rthdrlen); 7504 cp += tipp->ipp_rthdrlen; 7505 } 7506 /* 7507 * Do ultimate destination options 7508 */ 7509 if (option_exists & IPPF_DSTOPTS) { 7510 ip6_dest_t *dest = (ip6_dest_t *)cp; 7511 tipp = ANCIL_OR_STICKY_PTR(IPPF_DSTOPTS); 7512 7513 *nxthdr_ptr = IPPROTO_DSTOPTS; 7514 nxthdr_ptr = &dest->ip6d_nxt; 7515 7516 bcopy(tipp->ipp_dstopts, cp, tipp->ipp_dstoptslen); 7517 cp += tipp->ipp_dstoptslen; 7518 } 7519 /* 7520 * Now set the last header pointer to the proto passed in 7521 */ 7522 ASSERT((int)(cp - (uint8_t *)ip6i) == (udp_ip_hdr_len - UDPH_SIZE)); 7523 *nxthdr_ptr = IPPROTO_UDP; 7524 7525 /* Update UDP header */ 7526 udph = (udpha_t *)((uchar_t *)ip6i + udp_ip_hdr_len - UDPH_SIZE); 7527 udph->uha_dst_port = sin6->sin6_port; 7528 udph->uha_src_port = udp->udp_port; 7529 7530 /* 7531 * Copy in the destination address 7532 */ 7533 ip6h->ip6_dst = ip6_dst; 7534 7535 ip6h->ip6_vcf = 7536 (IPV6_DEFAULT_VERS_AND_FLOW & IPV6_VERS_AND_FLOW_MASK) | 7537 (sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK); 7538 7539 if (option_exists & IPPF_TCLASS) { 7540 tipp = ANCIL_OR_STICKY_PTR(IPPF_TCLASS); 7541 ip6h->ip6_vcf = IPV6_TCLASS_FLOW(ip6h->ip6_vcf, 7542 tipp->ipp_tclass); 7543 } 7544 7545 if (option_exists & IPPF_RTHDR) { 7546 ip6_rthdr_t *rth; 7547 7548 /* 7549 * Perform any processing needed for source routing. 7550 * We know that all extension headers will be in the same mblk 7551 * as the IPv6 header. 7552 */ 7553 rth = ip_find_rthdr_v6(ip6h, mp1->b_wptr); 7554 if (rth != NULL && rth->ip6r_segleft != 0) { 7555 if (rth->ip6r_type != IPV6_RTHDR_TYPE_0) { 7556 /* 7557 * Drop packet - only support Type 0 routing. 7558 * Notify the application as well. 7559 */ 7560 *error = EPROTO; 7561 goto done; 7562 } 7563 7564 /* 7565 * rth->ip6r_len is twice the number of 7566 * addresses in the header. Thus it must be even. 7567 */ 7568 if (rth->ip6r_len & 0x1) { 7569 *error = EPROTO; 7570 goto done; 7571 } 7572 /* 7573 * Shuffle the routing header and ip6_dst 7574 * addresses, and get the checksum difference 7575 * between the first hop (in ip6_dst) and 7576 * the destination (in the last routing hdr entry). 7577 */ 7578 csum = ip_massage_options_v6(ip6h, rth); 7579 /* 7580 * Verify that the first hop isn't a mapped address. 7581 * Routers along the path need to do this verification 7582 * for subsequent hops. 7583 */ 7584 if (IN6_IS_ADDR_V4MAPPED(&ip6h->ip6_dst)) { 7585 *error = EADDRNOTAVAIL; 7586 goto done; 7587 } 7588 7589 cp += (rth->ip6r_len + 1)*8; 7590 } 7591 } 7592 7593 /* count up length of UDP packet */ 7594 ip_len = (mp1->b_wptr - (unsigned char *)ip6h) - IPV6_HDR_LEN; 7595 if ((mp2 = mp1->b_cont) != NULL) { 7596 do { 7597 ASSERT((uintptr_t)MBLKL(mp2) <= (uintptr_t)UINT_MAX); 7598 ip_len += (uint32_t)MBLKL(mp2); 7599 } while ((mp2 = mp2->b_cont) != NULL); 7600 } 7601 7602 /* 7603 * If the size of the packet is greater than the maximum allowed by 7604 * ip, return an error. Passing this down could cause panics because 7605 * the size will have wrapped and be inconsistent with the msg size. 7606 */ 7607 if (ip_len > IP_MAXPACKET) { 7608 *error = EMSGSIZE; 7609 goto done; 7610 } 7611 7612 /* Store the UDP length. Subtract length of extension hdrs */ 7613 udph->uha_length = htons(ip_len + IPV6_HDR_LEN - 7614 (int)((uchar_t *)udph - (uchar_t *)ip6h)); 7615 7616 /* 7617 * We make it easy for IP to include our pseudo header 7618 * by putting our length in uh_checksum, modified (if 7619 * we have a routing header) by the checksum difference 7620 * between the ultimate destination and first hop addresses. 7621 * Note: UDP over IPv6 must always checksum the packet. 7622 */ 7623 csum += udph->uha_length; 7624 csum = (csum & 0xFFFF) + (csum >> 16); 7625 udph->uha_checksum = (uint16_t)csum; 7626 7627 #ifdef _LITTLE_ENDIAN 7628 ip_len = htons(ip_len); 7629 #endif 7630 ip6h->ip6_plen = ip_len; 7631 if (DB_CRED(mp) != NULL) 7632 mblk_setcred(mp1, DB_CRED(mp)); 7633 7634 if (DB_TYPE(mp) != M_DATA) { 7635 ASSERT(mp != mp1); 7636 freeb(mp); 7637 } 7638 7639 /* mp has been consumed and we'll return success */ 7640 ASSERT(*error == 0); 7641 mp = NULL; 7642 7643 /* We're done. Pass the packet to IP */ 7644 BUMP_MIB(&udp_mib, udpOutDatagrams); 7645 ip_output_v6(connp, mp1, q, IP_WPUT); 7646 7647 done: 7648 if (hopoptsptr != NULL && !is_ancillary) { 7649 kmem_free(hopoptsptr, hopoptslen); 7650 hopoptsptr = NULL; 7651 } 7652 if (*error != 0) { 7653 ASSERT(mp != NULL); 7654 BUMP_MIB(&udp_mib, udpOutErrors); 7655 } 7656 return (mp); 7657 } 7658 7659 static void 7660 udp_wput_other(queue_t *q, mblk_t *mp) 7661 { 7662 uchar_t *rptr = mp->b_rptr; 7663 struct datab *db; 7664 struct iocblk *iocp; 7665 cred_t *cr; 7666 conn_t *connp = Q_TO_CONN(q); 7667 udp_t *udp = connp->conn_udp; 7668 7669 TRACE_1(TR_FAC_UDP, TR_UDP_WPUT_OTHER_START, 7670 "udp_wput_other_start: q %p", q); 7671 7672 db = mp->b_datap; 7673 7674 cr = DB_CREDDEF(mp, connp->conn_cred); 7675 7676 switch (db->db_type) { 7677 case M_PROTO: 7678 case M_PCPROTO: 7679 if (mp->b_wptr - rptr < sizeof (t_scalar_t)) { 7680 freemsg(mp); 7681 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7682 "udp_wput_other_end: q %p (%S)", 7683 q, "protoshort"); 7684 return; 7685 } 7686 switch (((t_primp_t)rptr)->type) { 7687 case T_ADDR_REQ: 7688 udp_addr_req(q, mp); 7689 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7690 "udp_wput_other_end: q %p (%S)", q, "addrreq"); 7691 return; 7692 case O_T_BIND_REQ: 7693 case T_BIND_REQ: 7694 udp_bind(q, mp); 7695 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7696 "udp_wput_other_end: q %p (%S)", q, "bindreq"); 7697 return; 7698 case T_CONN_REQ: 7699 udp_connect(q, mp); 7700 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7701 "udp_wput_other_end: q %p (%S)", q, "connreq"); 7702 return; 7703 case T_CAPABILITY_REQ: 7704 udp_capability_req(q, mp); 7705 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7706 "udp_wput_other_end: q %p (%S)", q, "capabreq"); 7707 return; 7708 case T_INFO_REQ: 7709 udp_info_req(q, mp); 7710 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7711 "udp_wput_other_end: q %p (%S)", q, "inforeq"); 7712 return; 7713 case T_UNITDATA_REQ: 7714 /* 7715 * If a T_UNITDATA_REQ gets here, the address must 7716 * be bad. Valid T_UNITDATA_REQs are handled 7717 * in udp_wput. 7718 */ 7719 udp_ud_err(q, mp, NULL, 0, EADDRNOTAVAIL); 7720 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7721 "udp_wput_other_end: q %p (%S)", 7722 q, "unitdatareq"); 7723 return; 7724 case T_UNBIND_REQ: 7725 udp_unbind(q, mp); 7726 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7727 "udp_wput_other_end: q %p (%S)", q, "unbindreq"); 7728 return; 7729 case T_SVR4_OPTMGMT_REQ: 7730 if (!snmpcom_req(q, mp, udp_snmp_set, udp_snmp_get, cr)) 7731 /* 7732 * Use upper queue for option processing in 7733 * case the request is not handled at this 7734 * level and needs to be passed down to IP. 7735 */ 7736 (void) svr4_optcom_req(_WR(UDP_RD(q)), 7737 mp, cr, &udp_opt_obj); 7738 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7739 "udp_wput_other_end: q %p (%S)", 7740 q, "optmgmtreq"); 7741 return; 7742 7743 case T_OPTMGMT_REQ: 7744 /* 7745 * Use upper queue for option processing in 7746 * case the request is not handled at this 7747 * level and needs to be passed down to IP. 7748 */ 7749 (void) tpi_optcom_req(_WR(UDP_RD(q)), 7750 mp, cr, &udp_opt_obj); 7751 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7752 "udp_wput_other_end: q %p (%S)", 7753 q, "optmgmtreq"); 7754 return; 7755 7756 case T_DISCON_REQ: 7757 udp_disconnect(q, mp); 7758 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7759 "udp_wput_other_end: q %p (%S)", 7760 q, "disconreq"); 7761 return; 7762 7763 /* The following TPI message is not supported by udp. */ 7764 case O_T_CONN_RES: 7765 case T_CONN_RES: 7766 udp_err_ack(q, mp, TNOTSUPPORT, 0); 7767 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7768 "udp_wput_other_end: q %p (%S)", 7769 q, "connres/disconreq"); 7770 return; 7771 7772 /* The following 3 TPI messages are illegal for udp. */ 7773 case T_DATA_REQ: 7774 case T_EXDATA_REQ: 7775 case T_ORDREL_REQ: 7776 udp_err_ack(q, mp, TNOTSUPPORT, 0); 7777 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7778 "udp_wput_other_end: q %p (%S)", 7779 q, "data/exdata/ordrel"); 7780 return; 7781 default: 7782 break; 7783 } 7784 break; 7785 case M_FLUSH: 7786 if (*rptr & FLUSHW) 7787 flushq(q, FLUSHDATA); 7788 break; 7789 case M_IOCTL: 7790 iocp = (struct iocblk *)mp->b_rptr; 7791 switch (iocp->ioc_cmd) { 7792 case TI_GETPEERNAME: 7793 if (udp->udp_state != TS_DATA_XFER) { 7794 /* 7795 * If a default destination address has not 7796 * been associated with the stream, then we 7797 * don't know the peer's name. 7798 */ 7799 iocp->ioc_error = ENOTCONN; 7800 iocp->ioc_count = 0; 7801 mp->b_datap->db_type = M_IOCACK; 7802 putnext(UDP_RD(q), mp); 7803 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7804 "udp_wput_other_end: q %p (%S)", 7805 q, "getpeername"); 7806 return; 7807 } 7808 /* FALLTHRU */ 7809 case TI_GETMYNAME: { 7810 /* 7811 * For TI_GETPEERNAME and TI_GETMYNAME, we first 7812 * need to copyin the user's strbuf structure. 7813 * Processing will continue in the M_IOCDATA case 7814 * below. 7815 */ 7816 mi_copyin(q, mp, NULL, 7817 SIZEOF_STRUCT(strbuf, iocp->ioc_flag)); 7818 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7819 "udp_wput_other_end: q %p (%S)", 7820 q, "getmyname"); 7821 return; 7822 } 7823 case ND_SET: 7824 /* nd_getset performs the necessary checking */ 7825 case ND_GET: 7826 if (nd_getset(q, udp_g_nd, mp)) { 7827 putnext(UDP_RD(q), mp); 7828 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7829 "udp_wput_other_end: q %p (%S)", 7830 q, "get"); 7831 return; 7832 } 7833 break; 7834 case _SIOCSOCKFALLBACK: 7835 /* 7836 * Either sockmod is about to be popped and the 7837 * socket would now be treated as a plain stream, 7838 * or a module is about to be pushed so we could 7839 * no longer use read-side synchronous stream. 7840 * Drain any queued data and disable direct sockfs 7841 * interface from now on. 7842 */ 7843 if (!udp->udp_issocket) { 7844 DB_TYPE(mp) = M_IOCNAK; 7845 iocp->ioc_error = EINVAL; 7846 } else { 7847 udp->udp_issocket = B_FALSE; 7848 if (udp->udp_direct_sockfs) { 7849 /* 7850 * Disable read-side synchronous 7851 * stream interface and drain any 7852 * queued data. 7853 */ 7854 udp_rcv_drain(UDP_RD(q), udp, 7855 B_FALSE); 7856 ASSERT(!udp->udp_direct_sockfs); 7857 UDP_STAT(udp_sock_fallback); 7858 } 7859 DB_TYPE(mp) = M_IOCACK; 7860 iocp->ioc_error = 0; 7861 } 7862 iocp->ioc_count = 0; 7863 iocp->ioc_rval = 0; 7864 putnext(UDP_RD(q), mp); 7865 return; 7866 default: 7867 break; 7868 } 7869 break; 7870 case M_IOCDATA: 7871 udp_wput_iocdata(q, mp); 7872 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7873 "udp_wput_other_end: q %p (%S)", q, "iocdata"); 7874 return; 7875 default: 7876 /* Unrecognized messages are passed through without change. */ 7877 break; 7878 } 7879 TRACE_2(TR_FAC_UDP, TR_UDP_WPUT_OTHER_END, 7880 "udp_wput_other_end: q %p (%S)", q, "end"); 7881 ip_output(connp, mp, q, IP_WPUT); 7882 } 7883 7884 /* ARGSUSED */ 7885 static void 7886 udp_wput_other_wrapper(void *arg, mblk_t *mp, void *arg2) 7887 { 7888 udp_wput_other(((conn_t *)arg)->conn_wq, mp); 7889 udp_exit((conn_t *)arg); 7890 } 7891 7892 /* 7893 * udp_wput_iocdata is called by udp_wput_other to handle all M_IOCDATA 7894 * messages. 7895 */ 7896 static void 7897 udp_wput_iocdata(queue_t *q, mblk_t *mp) 7898 { 7899 mblk_t *mp1; 7900 STRUCT_HANDLE(strbuf, sb); 7901 uint16_t port; 7902 in6_addr_t v6addr; 7903 ipaddr_t v4addr; 7904 uint32_t flowinfo = 0; 7905 int addrlen; 7906 udp_t *udp = Q_TO_UDP(q); 7907 7908 /* Make sure it is one of ours. */ 7909 switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) { 7910 case TI_GETMYNAME: 7911 case TI_GETPEERNAME: 7912 break; 7913 default: 7914 ip_output(Q_TO_CONN(q), mp, q, IP_WPUT); 7915 return; 7916 } 7917 7918 q = WR(UDP_RD(q)); 7919 switch (mi_copy_state(q, mp, &mp1)) { 7920 case -1: 7921 return; 7922 case MI_COPY_CASE(MI_COPY_IN, 1): 7923 break; 7924 case MI_COPY_CASE(MI_COPY_OUT, 1): 7925 /* 7926 * The address has been copied out, so now 7927 * copyout the strbuf. 7928 */ 7929 mi_copyout(q, mp); 7930 return; 7931 case MI_COPY_CASE(MI_COPY_OUT, 2): 7932 /* 7933 * The address and strbuf have been copied out. 7934 * We're done, so just acknowledge the original 7935 * M_IOCTL. 7936 */ 7937 mi_copy_done(q, mp, 0); 7938 return; 7939 default: 7940 /* 7941 * Something strange has happened, so acknowledge 7942 * the original M_IOCTL with an EPROTO error. 7943 */ 7944 mi_copy_done(q, mp, EPROTO); 7945 return; 7946 } 7947 7948 /* 7949 * Now we have the strbuf structure for TI_GETMYNAME 7950 * and TI_GETPEERNAME. Next we copyout the requested 7951 * address and then we'll copyout the strbuf. 7952 */ 7953 STRUCT_SET_HANDLE(sb, ((struct iocblk *)mp->b_rptr)->ioc_flag, 7954 (void *)mp1->b_rptr); 7955 if (udp->udp_family == AF_INET) 7956 addrlen = sizeof (sin_t); 7957 else 7958 addrlen = sizeof (sin6_t); 7959 7960 if (STRUCT_FGET(sb, maxlen) < addrlen) { 7961 mi_copy_done(q, mp, EINVAL); 7962 return; 7963 } 7964 switch (((struct iocblk *)mp->b_rptr)->ioc_cmd) { 7965 case TI_GETMYNAME: 7966 if (udp->udp_family == AF_INET) { 7967 ASSERT(udp->udp_ipversion == IPV4_VERSION); 7968 if (!IN6_IS_ADDR_V4MAPPED_ANY(&udp->udp_v6src) && 7969 !IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) { 7970 v4addr = V4_PART_OF_V6(udp->udp_v6src); 7971 } else { 7972 /* 7973 * INADDR_ANY 7974 * udp_v6src is not set, we might be bound to 7975 * broadcast/multicast. Use udp_bound_v6src as 7976 * local address instead (that could 7977 * also still be INADDR_ANY) 7978 */ 7979 v4addr = V4_PART_OF_V6(udp->udp_bound_v6src); 7980 } 7981 } else { 7982 /* udp->udp_family == AF_INET6 */ 7983 if (!IN6_IS_ADDR_UNSPECIFIED(&udp->udp_v6src)) { 7984 v6addr = udp->udp_v6src; 7985 } else { 7986 /* 7987 * UNSPECIFIED 7988 * udp_v6src is not set, we might be bound to 7989 * broadcast/multicast. Use udp_bound_v6src as 7990 * local address instead (that could 7991 * also still be UNSPECIFIED) 7992 */ 7993 v6addr = udp->udp_bound_v6src; 7994 } 7995 } 7996 port = udp->udp_port; 7997 break; 7998 case TI_GETPEERNAME: 7999 if (udp->udp_state != TS_DATA_XFER) { 8000 mi_copy_done(q, mp, ENOTCONN); 8001 return; 8002 } 8003 if (udp->udp_family == AF_INET) { 8004 ASSERT(udp->udp_ipversion == IPV4_VERSION); 8005 v4addr = V4_PART_OF_V6(udp->udp_v6dst); 8006 } else { 8007 /* udp->udp_family == AF_INET6) */ 8008 v6addr = udp->udp_v6dst; 8009 flowinfo = udp->udp_flowinfo; 8010 } 8011 port = udp->udp_dstport; 8012 break; 8013 default: 8014 mi_copy_done(q, mp, EPROTO); 8015 return; 8016 } 8017 mp1 = mi_copyout_alloc(q, mp, STRUCT_FGETP(sb, buf), addrlen, B_TRUE); 8018 if (!mp1) 8019 return; 8020 8021 if (udp->udp_family == AF_INET) { 8022 sin_t *sin; 8023 8024 STRUCT_FSET(sb, len, (int)sizeof (sin_t)); 8025 sin = (sin_t *)mp1->b_rptr; 8026 mp1->b_wptr = (uchar_t *)&sin[1]; 8027 *sin = sin_null; 8028 sin->sin_family = AF_INET; 8029 sin->sin_addr.s_addr = v4addr; 8030 sin->sin_port = port; 8031 } else { 8032 /* udp->udp_family == AF_INET6 */ 8033 sin6_t *sin6; 8034 8035 STRUCT_FSET(sb, len, (int)sizeof (sin6_t)); 8036 sin6 = (sin6_t *)mp1->b_rptr; 8037 mp1->b_wptr = (uchar_t *)&sin6[1]; 8038 *sin6 = sin6_null; 8039 sin6->sin6_family = AF_INET6; 8040 sin6->sin6_flowinfo = flowinfo; 8041 sin6->sin6_addr = v6addr; 8042 sin6->sin6_port = port; 8043 } 8044 /* Copy out the address */ 8045 mi_copyout(q, mp); 8046 } 8047 8048 8049 static int 8050 udp_unitdata_opt_process(queue_t *q, mblk_t *mp, int *errorp, 8051 udpattrs_t *udpattrs) 8052 { 8053 struct T_unitdata_req *udreqp; 8054 int is_absreq_failure; 8055 cred_t *cr; 8056 conn_t *connp = Q_TO_CONN(q); 8057 8058 ASSERT(((t_primp_t)mp->b_rptr)->type); 8059 8060 cr = DB_CREDDEF(mp, connp->conn_cred); 8061 8062 udreqp = (struct T_unitdata_req *)mp->b_rptr; 8063 8064 /* 8065 * Use upper queue for option processing since the callback 8066 * routines expect to be called in UDP instance instead of IP. 8067 */ 8068 *errorp = tpi_optcom_buf(_WR(UDP_RD(q)), mp, &udreqp->OPT_length, 8069 udreqp->OPT_offset, cr, &udp_opt_obj, 8070 udpattrs, &is_absreq_failure); 8071 8072 if (*errorp != 0) { 8073 /* 8074 * Note: No special action needed in this 8075 * module for "is_absreq_failure" 8076 */ 8077 return (-1); /* failure */ 8078 } 8079 ASSERT(is_absreq_failure == 0); 8080 return (0); /* success */ 8081 } 8082 8083 void 8084 udp_ddi_init(void) 8085 { 8086 int i; 8087 8088 UDP6_MAJ = ddi_name_to_major(UDP6); 8089 8090 udp_max_optsize = optcom_max_optsize(udp_opt_obj.odb_opt_des_arr, 8091 udp_opt_obj.odb_opt_arr_cnt); 8092 8093 if (udp_bind_fanout_size & (udp_bind_fanout_size - 1)) { 8094 /* Not a power of two. Round up to nearest power of two */ 8095 for (i = 0; i < 31; i++) { 8096 if (udp_bind_fanout_size < (1 << i)) 8097 break; 8098 } 8099 udp_bind_fanout_size = 1 << i; 8100 } 8101 udp_bind_fanout = kmem_zalloc(udp_bind_fanout_size * 8102 sizeof (udp_fanout_t), KM_SLEEP); 8103 for (i = 0; i < udp_bind_fanout_size; i++) { 8104 mutex_init(&udp_bind_fanout[i].uf_lock, NULL, MUTEX_DEFAULT, 8105 NULL); 8106 } 8107 (void) udp_param_register(udp_param_arr, A_CNT(udp_param_arr)); 8108 8109 udp_kstat_init(); 8110 8111 udp_cache = kmem_cache_create("udp_cache", sizeof (udp_t), 8112 CACHE_ALIGN_SIZE, NULL, NULL, NULL, NULL, NULL, 0); 8113 } 8114 8115 void 8116 udp_ddi_destroy(void) 8117 { 8118 int i; 8119 8120 nd_free(&udp_g_nd); 8121 8122 for (i = 0; i < udp_bind_fanout_size; i++) { 8123 mutex_destroy(&udp_bind_fanout[i].uf_lock); 8124 } 8125 8126 kmem_free(udp_bind_fanout, udp_bind_fanout_size * 8127 sizeof (udp_fanout_t)); 8128 8129 udp_kstat_fini(); 8130 8131 kmem_cache_destroy(udp_cache); 8132 } 8133 8134 static void 8135 udp_kstat_init(void) 8136 { 8137 udp_named_kstat_t template = { 8138 { "inDatagrams", KSTAT_DATA_UINT32, 0 }, 8139 { "inErrors", KSTAT_DATA_UINT32, 0 }, 8140 { "outDatagrams", KSTAT_DATA_UINT32, 0 }, 8141 { "entrySize", KSTAT_DATA_INT32, 0 }, 8142 { "entry6Size", KSTAT_DATA_INT32, 0 }, 8143 { "outErrors", KSTAT_DATA_UINT32, 0 }, 8144 }; 8145 8146 udp_mibkp = kstat_create(UDP_MOD_NAME, 0, UDP_MOD_NAME, 8147 "mib2", KSTAT_TYPE_NAMED, NUM_OF_FIELDS(udp_named_kstat_t), 0); 8148 8149 if (udp_mibkp == NULL) 8150 return; 8151 8152 template.entrySize.value.ui32 = sizeof (mib2_udpEntry_t); 8153 template.entry6Size.value.ui32 = sizeof (mib2_udp6Entry_t); 8154 8155 bcopy(&template, udp_mibkp->ks_data, sizeof (template)); 8156 8157 udp_mibkp->ks_update = udp_kstat_update; 8158 8159 kstat_install(udp_mibkp); 8160 8161 if ((udp_ksp = kstat_create(UDP_MOD_NAME, 0, "udpstat", 8162 "net", KSTAT_TYPE_NAMED, 8163 sizeof (udp_statistics) / sizeof (kstat_named_t), 8164 KSTAT_FLAG_VIRTUAL)) != NULL) { 8165 udp_ksp->ks_data = &udp_statistics; 8166 kstat_install(udp_ksp); 8167 } 8168 } 8169 8170 static void 8171 udp_kstat_fini(void) 8172 { 8173 if (udp_ksp != NULL) { 8174 kstat_delete(udp_ksp); 8175 udp_ksp = NULL; 8176 } 8177 if (udp_mibkp != NULL) { 8178 kstat_delete(udp_mibkp); 8179 udp_mibkp = NULL; 8180 } 8181 } 8182 8183 static int 8184 udp_kstat_update(kstat_t *kp, int rw) 8185 { 8186 udp_named_kstat_t *udpkp; 8187 8188 if ((kp == NULL) || (kp->ks_data == NULL)) 8189 return (EIO); 8190 8191 if (rw == KSTAT_WRITE) 8192 return (EACCES); 8193 8194 udpkp = (udp_named_kstat_t *)kp->ks_data; 8195 8196 udpkp->inDatagrams.value.ui32 = udp_mib.udpInDatagrams; 8197 udpkp->inErrors.value.ui32 = udp_mib.udpInErrors; 8198 udpkp->outDatagrams.value.ui32 = udp_mib.udpOutDatagrams; 8199 udpkp->outErrors.value.ui32 = udp_mib.udpOutErrors; 8200 8201 return (0); 8202 } 8203 8204 /* ARGSUSED */ 8205 static void 8206 udp_rput(queue_t *q, mblk_t *mp) 8207 { 8208 /* 8209 * We get here whenever we do qreply() from IP, 8210 * i.e as part of handlings ioctls, etc. 8211 */ 8212 putnext(q, mp); 8213 } 8214 8215 /* 8216 * Read-side synchronous stream info entry point, called as a 8217 * result of handling certain STREAMS ioctl operations. 8218 */ 8219 static int 8220 udp_rinfop(queue_t *q, infod_t *dp) 8221 { 8222 mblk_t *mp; 8223 uint_t cmd = dp->d_cmd; 8224 int res = 0; 8225 int error = 0; 8226 udp_t *udp = Q_TO_UDP(RD(UDP_WR(q))); 8227 struct stdata *stp = STREAM(q); 8228 8229 mutex_enter(&udp->udp_drain_lock); 8230 /* If shutdown on read has happened, return nothing */ 8231 mutex_enter(&stp->sd_lock); 8232 if (stp->sd_flag & STREOF) { 8233 mutex_exit(&stp->sd_lock); 8234 goto done; 8235 } 8236 mutex_exit(&stp->sd_lock); 8237 8238 if ((mp = udp->udp_rcv_list_head) == NULL) 8239 goto done; 8240 8241 ASSERT(DB_TYPE(mp) != M_DATA && mp->b_cont != NULL); 8242 8243 if (cmd & INFOD_COUNT) { 8244 /* 8245 * Return the number of messages. 8246 */ 8247 dp->d_count += udp->udp_rcv_msgcnt; 8248 res |= INFOD_COUNT; 8249 } 8250 if (cmd & INFOD_BYTES) { 8251 /* 8252 * Return size of all data messages. 8253 */ 8254 dp->d_bytes += udp->udp_rcv_cnt; 8255 res |= INFOD_BYTES; 8256 } 8257 if (cmd & INFOD_FIRSTBYTES) { 8258 /* 8259 * Return size of first data message. 8260 */ 8261 dp->d_bytes = msgdsize(mp); 8262 res |= INFOD_FIRSTBYTES; 8263 dp->d_cmd &= ~INFOD_FIRSTBYTES; 8264 } 8265 if (cmd & INFOD_COPYOUT) { 8266 mblk_t *mp1 = mp->b_cont; 8267 int n; 8268 /* 8269 * Return data contents of first message. 8270 */ 8271 ASSERT(DB_TYPE(mp1) == M_DATA); 8272 while (mp1 != NULL && dp->d_uiop->uio_resid > 0) { 8273 n = MIN(dp->d_uiop->uio_resid, MBLKL(mp1)); 8274 if (n != 0 && (error = uiomove((char *)mp1->b_rptr, n, 8275 UIO_READ, dp->d_uiop)) != 0) { 8276 goto done; 8277 } 8278 mp1 = mp1->b_cont; 8279 } 8280 res |= INFOD_COPYOUT; 8281 dp->d_cmd &= ~INFOD_COPYOUT; 8282 } 8283 done: 8284 mutex_exit(&udp->udp_drain_lock); 8285 8286 dp->d_res |= res; 8287 8288 return (error); 8289 } 8290 8291 /* 8292 * Read-side synchronous stream entry point. This is called as a result 8293 * of recv/read operation done at sockfs, and is guaranteed to execute 8294 * outside of the interrupt thread context. It returns a single datagram 8295 * (b_cont chain of T_UNITDATA_IND plus data) to the upper layer. 8296 */ 8297 static int 8298 udp_rrw(queue_t *q, struiod_t *dp) 8299 { 8300 mblk_t *mp; 8301 udp_t *udp = Q_TO_UDP(_RD(UDP_WR(q))); 8302 8303 /* We should never get here when we're in SNMP mode */ 8304 ASSERT(!(udp->udp_connp->conn_flags & IPCL_UDPMOD)); 8305 8306 /* 8307 * Dequeue datagram from the head of the list and return 8308 * it to caller; also ensure that RSLEEP sd_wakeq flag is 8309 * set/cleared depending on whether or not there's data 8310 * remaining in the list. 8311 */ 8312 mutex_enter(&udp->udp_drain_lock); 8313 if (!udp->udp_direct_sockfs) { 8314 mutex_exit(&udp->udp_drain_lock); 8315 UDP_STAT(udp_rrw_busy); 8316 return (EBUSY); 8317 } 8318 if ((mp = udp->udp_rcv_list_head) != NULL) { 8319 uint_t size = msgdsize(mp); 8320 8321 /* Last datagram in the list? */ 8322 if ((udp->udp_rcv_list_head = mp->b_next) == NULL) 8323 udp->udp_rcv_list_tail = NULL; 8324 mp->b_next = NULL; 8325 8326 udp->udp_rcv_cnt -= size; 8327 udp->udp_rcv_msgcnt--; 8328 UDP_STAT(udp_rrw_msgcnt); 8329 8330 /* No longer flow-controlling? */ 8331 if (udp->udp_rcv_cnt < udp->udp_rcv_hiwat && 8332 udp->udp_rcv_msgcnt < udp->udp_rcv_hiwat) 8333 udp->udp_drain_qfull = B_FALSE; 8334 } 8335 if (udp->udp_rcv_list_head == NULL) { 8336 /* 8337 * Either we just dequeued the last datagram or 8338 * we get here from sockfs and have nothing to 8339 * return; in this case clear RSLEEP. 8340 */ 8341 ASSERT(udp->udp_rcv_cnt == 0); 8342 ASSERT(udp->udp_rcv_msgcnt == 0); 8343 ASSERT(udp->udp_rcv_list_tail == NULL); 8344 STR_WAKEUP_CLEAR(STREAM(q)); 8345 } else { 8346 /* 8347 * More data follows; we need udp_rrw() to be 8348 * called in future to pick up the rest. 8349 */ 8350 STR_WAKEUP_SET(STREAM(q)); 8351 } 8352 mutex_exit(&udp->udp_drain_lock); 8353 dp->d_mp = mp; 8354 return (0); 8355 } 8356 8357 /* 8358 * Enqueue a completely-built T_UNITDATA_IND message into the receive 8359 * list; this is typically executed within the interrupt thread context 8360 * and so we do things as quickly as possible. 8361 */ 8362 static void 8363 udp_rcv_enqueue(queue_t *q, udp_t *udp, mblk_t *mp, uint_t pkt_len) 8364 { 8365 ASSERT(q == RD(q)); 8366 ASSERT(pkt_len == msgdsize(mp)); 8367 ASSERT(mp->b_next == NULL && mp->b_cont != NULL); 8368 ASSERT(DB_TYPE(mp) == M_PROTO && DB_TYPE(mp->b_cont) == M_DATA); 8369 ASSERT(MBLKL(mp) >= sizeof (struct T_unitdata_ind)); 8370 8371 mutex_enter(&udp->udp_drain_lock); 8372 /* 8373 * Wake up and signal the receiving app; it is okay to do this 8374 * before enqueueing the mp because we are holding the drain lock. 8375 * One of the advantages of synchronous stream is the ability for 8376 * us to find out when the application performs a read on the 8377 * socket by way of udp_rrw() entry point being called. We need 8378 * to generate SIGPOLL/SIGIO for each received data in the case 8379 * of asynchronous socket just as in the strrput() case. However, 8380 * we only wake the application up when necessary, i.e. during the 8381 * first enqueue. When udp_rrw() is called, we send up a single 8382 * datagram upstream and call STR_WAKEUP_SET() again when there 8383 * are still data remaining in our receive queue. 8384 */ 8385 if (udp->udp_rcv_list_head == NULL) { 8386 STR_WAKEUP_SET(STREAM(q)); 8387 udp->udp_rcv_list_head = mp; 8388 } else { 8389 udp->udp_rcv_list_tail->b_next = mp; 8390 } 8391 udp->udp_rcv_list_tail = mp; 8392 udp->udp_rcv_cnt += pkt_len; 8393 udp->udp_rcv_msgcnt++; 8394 8395 /* Need to flow-control? */ 8396 if (udp->udp_rcv_cnt >= udp->udp_rcv_hiwat || 8397 udp->udp_rcv_msgcnt >= udp->udp_rcv_hiwat) 8398 udp->udp_drain_qfull = B_TRUE; 8399 8400 /* Update poll events and send SIGPOLL/SIGIO if necessary */ 8401 STR_SENDSIG(STREAM(q)); 8402 mutex_exit(&udp->udp_drain_lock); 8403 } 8404 8405 /* 8406 * Drain the contents of receive list to the module upstream; we do 8407 * this during close or when we fallback to the slow mode due to 8408 * sockmod being popped or a module being pushed on top of us. 8409 */ 8410 static void 8411 udp_rcv_drain(queue_t *q, udp_t *udp, boolean_t closing) 8412 { 8413 mblk_t *mp; 8414 8415 ASSERT(q == RD(q)); 8416 8417 mutex_enter(&udp->udp_drain_lock); 8418 /* 8419 * There is no race with a concurrent udp_input() sending 8420 * up packets using putnext() after we have cleared the 8421 * udp_direct_sockfs flag but before we have completed 8422 * sending up the packets in udp_rcv_list, since we are 8423 * either a writer or we have quiesced the conn. 8424 */ 8425 udp->udp_direct_sockfs = B_FALSE; 8426 mutex_exit(&udp->udp_drain_lock); 8427 8428 if (udp->udp_rcv_list_head != NULL) 8429 UDP_STAT(udp_drain); 8430 8431 /* 8432 * Send up everything via putnext(); note here that we 8433 * don't need the udp_drain_lock to protect us since 8434 * nothing can enter udp_rrw() and that we currently 8435 * have exclusive access to this udp. 8436 */ 8437 while ((mp = udp->udp_rcv_list_head) != NULL) { 8438 udp->udp_rcv_list_head = mp->b_next; 8439 mp->b_next = NULL; 8440 udp->udp_rcv_cnt -= msgdsize(mp); 8441 udp->udp_rcv_msgcnt--; 8442 if (closing) { 8443 freemsg(mp); 8444 } else { 8445 putnext(q, mp); 8446 } 8447 } 8448 ASSERT(udp->udp_rcv_cnt == 0); 8449 ASSERT(udp->udp_rcv_msgcnt == 0); 8450 ASSERT(udp->udp_rcv_list_head == NULL); 8451 udp->udp_rcv_list_tail = NULL; 8452 udp->udp_drain_qfull = B_FALSE; 8453 } 8454 8455 static size_t 8456 udp_set_rcv_hiwat(udp_t *udp, size_t size) 8457 { 8458 /* We add a bit of extra buffering */ 8459 size += size >> 1; 8460 if (size > udp_max_buf) 8461 size = udp_max_buf; 8462 8463 udp->udp_rcv_hiwat = size; 8464 return (size); 8465 } 8466 8467 /* 8468 * Little helper for IPsec's NAT-T processing. 8469 */ 8470 boolean_t 8471 udp_compute_checksum(void) 8472 { 8473 return (udp_do_checksum); 8474 } 8475