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