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 /* 23 * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 /* Copyright (c) 1990 Mentat Inc. */ 27 28 #ifndef _INET_IP_H 29 #define _INET_IP_H 30 31 #pragma ident "%Z%%M% %I% %E% SMI" 32 33 #ifdef __cplusplus 34 extern "C" { 35 #endif 36 37 #include <sys/isa_defs.h> 38 #include <sys/types.h> 39 #include <inet/mib2.h> 40 #include <inet/nd.h> 41 #include <sys/atomic.h> 42 #include <sys/socket.h> 43 #include <net/if_dl.h> 44 #include <net/if.h> 45 #include <netinet/ip.h> 46 #include <sys/dlpi.h> 47 #include <netinet/igmp.h> 48 49 #ifdef _KERNEL 50 #include <netinet/ip6.h> 51 #include <sys/avl.h> 52 #include <sys/vmem.h> 53 #include <sys/squeue.h> 54 #include <net/route.h> 55 #include <sys/systm.h> 56 #include <sys/multidata.h> 57 #include <net/radix.h> 58 59 #ifdef DEBUG 60 #define ILL_DEBUG 61 #define IRE_DEBUG 62 #define NCE_DEBUG 63 #define CONN_DEBUG 64 #endif 65 66 #define IP_DEBUG 67 /* 68 * The mt-streams(9F) flags for the IP module; put here so that other 69 * "drivers" that are actually IP (e.g., ICMP, UDP) can use the same set 70 * of flags. 71 */ 72 #define IP_DEVMTFLAGS D_MP 73 #endif /* _KERNEL */ 74 75 #define IP_MOD_NAME "ip" 76 #define IP_DEV_NAME "/dev/ip" 77 #define IP6_DEV_NAME "/dev/ip6" 78 79 #define UDP_MOD_NAME "udp" 80 #define UDP_DEV_NAME "/dev/udp" 81 #define UDP6_DEV_NAME "/dev/udp6" 82 83 #define TCP_MOD_NAME "tcp" 84 #define TCP_DEV_NAME "/dev/tcp" 85 #define TCP6_DEV_NAME "/dev/tcp6" 86 87 #define SCTP_MOD_NAME "sctp" 88 89 /* Minor numbers */ 90 #define IPV4_MINOR 0 91 #define IPV6_MINOR 1 92 #define TCP_MINOR 2 93 #define TCP_MINOR6 3 94 95 #ifndef _IPADDR_T 96 #define _IPADDR_T 97 typedef uint32_t ipaddr_t; 98 #endif 99 100 /* Number of bits in an address */ 101 #define IP_ABITS 32 102 #define IPV6_ABITS 128 103 104 #define IP_HOST_MASK (ipaddr_t)0xffffffffU 105 106 #define IP_CSUM(mp, off, sum) (~ip_cksum(mp, off, sum) & 0xFFFF) 107 #define IP_CSUM_PARTIAL(mp, off, sum) ip_cksum(mp, off, sum) 108 #define IP_BCSUM_PARTIAL(bp, len, sum) bcksum(bp, len, sum) 109 #define IP_MD_CSUM(pd, off, sum) (~ip_md_cksum(pd, off, sum) & 0xffff) 110 #define IP_MD_CSUM_PARTIAL(pd, off, sum) ip_md_cksum(pd, off, sum) 111 112 /* 113 * Flag to IP write side to indicate that the appln has sent in a pre-built 114 * IP header. Stored in ipha_ident (which is otherwise zero). 115 */ 116 #define IP_HDR_INCLUDED 0xFFFF 117 118 #define ILL_FRAG_HASH_TBL_COUNT ((unsigned int)64) 119 #define ILL_FRAG_HASH_TBL_SIZE (ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t)) 120 121 #define IPV4_ADDR_LEN 4 122 #define IP_ADDR_LEN IPV4_ADDR_LEN 123 #define IP_ARP_PROTO_TYPE 0x0800 124 125 #define IPV4_VERSION 4 126 #define IP_VERSION IPV4_VERSION 127 #define IP_SIMPLE_HDR_LENGTH_IN_WORDS 5 128 #define IP_SIMPLE_HDR_LENGTH 20 129 #define IP_MAX_HDR_LENGTH 60 130 131 #define IP_MAX_OPT_LENGTH (IP_MAX_HDR_LENGTH-IP_SIMPLE_HDR_LENGTH) 132 133 #define IP_MIN_MTU (IP_MAX_HDR_LENGTH + 8) /* 68 bytes */ 134 135 /* 136 * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the 137 * 2 files should be cleaned up to remove all redundant definitions. 138 */ 139 #define IP_MAXPACKET 65535 140 #define IP_SIMPLE_HDR_VERSION \ 141 ((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS) 142 143 #define UDPH_SIZE 8 144 145 /* Leave room for ip_newroute to tack on the src and target addresses */ 146 #define OK_RESOLVER_MP(mp) \ 147 ((mp) && ((mp)->b_wptr - (mp)->b_rptr) >= (2 * IP_ADDR_LEN)) 148 149 /* 150 * Constants and type definitions to support IP IOCTL commands 151 */ 152 #define IP_IOCTL (('i'<<8)|'p') 153 #define IP_IOC_IRE_DELETE 4 154 #define IP_IOC_IRE_DELETE_NO_REPLY 5 155 #define IP_IOC_IRE_ADVISE_NO_REPLY 6 156 #define IP_IOC_RTS_REQUEST 7 157 158 /* Common definitions used by IP IOCTL data structures */ 159 typedef struct ipllcmd_s { 160 uint_t ipllc_cmd; 161 uint_t ipllc_name_offset; 162 uint_t ipllc_name_length; 163 } ipllc_t; 164 165 /* IP IRE Change Command Structure. */ 166 typedef struct ipic_s { 167 ipllc_t ipic_ipllc; 168 uint_t ipic_ire_type; 169 uint_t ipic_max_frag; 170 uint_t ipic_addr_offset; 171 uint_t ipic_addr_length; 172 uint_t ipic_mask_offset; 173 uint_t ipic_mask_length; 174 uint_t ipic_src_addr_offset; 175 uint_t ipic_src_addr_length; 176 uint_t ipic_ll_hdr_offset; 177 uint_t ipic_ll_hdr_length; 178 uint_t ipic_gateway_addr_offset; 179 uint_t ipic_gateway_addr_length; 180 clock_t ipic_rtt; 181 uint32_t ipic_ssthresh; 182 clock_t ipic_rtt_sd; 183 uchar_t ipic_ire_marks; 184 } ipic_t; 185 186 #define ipic_cmd ipic_ipllc.ipllc_cmd 187 #define ipic_ll_name_length ipic_ipllc.ipllc_name_length 188 #define ipic_ll_name_offset ipic_ipllc.ipllc_name_offset 189 190 /* IP IRE Delete Command Structure. */ 191 typedef struct ipid_s { 192 ipllc_t ipid_ipllc; 193 uint_t ipid_ire_type; 194 uint_t ipid_addr_offset; 195 uint_t ipid_addr_length; 196 uint_t ipid_mask_offset; 197 uint_t ipid_mask_length; 198 } ipid_t; 199 200 #define ipid_cmd ipid_ipllc.ipllc_cmd 201 202 #ifdef _KERNEL 203 /* 204 * Temporary state for ip options parser. 205 */ 206 typedef struct ipoptp_s 207 { 208 uint8_t *ipoptp_next; /* next option to look at */ 209 uint8_t *ipoptp_end; /* end of options */ 210 uint8_t *ipoptp_cur; /* start of current option */ 211 uint8_t ipoptp_len; /* length of current option */ 212 uint32_t ipoptp_flags; 213 } ipoptp_t; 214 215 /* 216 * Flag(s) for ipoptp_flags 217 */ 218 #define IPOPTP_ERROR 0x00000001 219 #endif /* _KERNEL */ 220 221 222 /* Controls forwarding of IP packets, set via ndd */ 223 #define IP_FORWARD_NEVER 0 224 #define IP_FORWARD_ALWAYS 1 225 226 #define WE_ARE_FORWARDING (ip_g_forward == IP_FORWARD_ALWAYS) 227 228 #define IPH_HDR_LENGTH(ipha) \ 229 ((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2) 230 231 #define IPH_HDR_VERSION(ipha) \ 232 ((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4) 233 234 #ifdef _KERNEL 235 /* 236 * IP reassembly macros. We hide starting and ending offsets in b_next and 237 * b_prev of messages on the reassembly queue. The messages are chained using 238 * b_cont. These macros are used in ip_reassemble() so we don't have to see 239 * the ugly casts and assignments. 240 * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent 241 * them. 242 */ 243 #define IP_REASS_START(mp) ((uint_t)(uintptr_t)((mp)->b_next)) 244 #define IP_REASS_SET_START(mp, u) \ 245 ((mp)->b_next = (mblk_t *)(uintptr_t)(u)) 246 #define IP_REASS_END(mp) ((uint_t)(uintptr_t)((mp)->b_prev)) 247 #define IP_REASS_SET_END(mp, u) \ 248 ((mp)->b_prev = (mblk_t *)(uintptr_t)(u)) 249 250 #define IP_REASS_COMPLETE 0x1 251 #define IP_REASS_PARTIAL 0x2 252 #define IP_REASS_FAILED 0x4 253 254 /* 255 * Test to determine whether this is a module instance of IP or a 256 * driver instance of IP. 257 */ 258 #define CONN_Q(q) (WR(q)->q_next == NULL) 259 260 #define Q_TO_CONN(q) ((conn_t *)(q)->q_ptr) 261 #define Q_TO_TCP(q) (Q_TO_CONN((q))->conn_tcp) 262 #define Q_TO_UDP(q) (Q_TO_CONN((q))->conn_udp) 263 264 /* 265 * The following two macros are used by IP to get the appropriate 266 * wq and rq for a conn. If it is a TCP conn, then we need 267 * tcp_wq/tcp_rq else, conn_wq/conn_rq. IP can use conn_wq and conn_rq 268 * from a conn directly if it knows that the conn is not TCP. 269 */ 270 #define CONNP_TO_WQ(connp) \ 271 (IPCL_IS_TCP(connp) ? (connp)->conn_tcp->tcp_wq : (connp)->conn_wq) 272 273 #define CONNP_TO_RQ(connp) RD(CONNP_TO_WQ(connp)) 274 275 #define GRAB_CONN_LOCK(q) { \ 276 if (q != NULL && CONN_Q(q)) \ 277 mutex_enter(&(Q_TO_CONN(q))->conn_lock); \ 278 } 279 280 #define RELEASE_CONN_LOCK(q) { \ 281 if (q != NULL && CONN_Q(q)) \ 282 mutex_exit(&(Q_TO_CONN(q))->conn_lock); \ 283 } 284 285 /* "Congestion controlled" protocol */ 286 #define IP_FLOW_CONTROLLED_ULP(p) ((p) == IPPROTO_TCP || (p) == IPPROTO_SCTP) 287 288 /* 289 * Complete the pending operation. Usually an ioctl. Can also 290 * be a bind or option management request that got enqueued 291 * in an ipsq_t. Called on completion of the operation. 292 */ 293 #define CONN_OPER_PENDING_DONE(connp) { \ 294 mutex_enter(&(connp)->conn_lock); \ 295 (connp)->conn_oper_pending_ill = NULL; \ 296 cv_broadcast(&(connp)->conn_refcv); \ 297 mutex_exit(&(connp)->conn_lock); \ 298 CONN_DEC_REF(connp); \ 299 } 300 301 /* Get the credential of an IP queue of unknown type */ 302 #define GET_QUEUE_CRED(wq) \ 303 ((wq)->q_next ? (((ill_t *)(wq)->q_ptr)->ill_credp) \ 304 : ((Q_TO_CONN((wq)))->conn_cred)) 305 306 /* 307 * Flags for the various ip_fanout_* routines. 308 */ 309 #define IP_FF_SEND_ICMP 0x01 /* Send an ICMP error */ 310 #define IP_FF_HDR_COMPLETE 0x02 /* Call ip_hdr_complete if error */ 311 #define IP_FF_CKSUM 0x04 /* Recompute ipha_cksum if error */ 312 #define IP_FF_RAWIP 0x08 /* Use rawip mib variable */ 313 #define IP_FF_SRC_QUENCH 0x10 /* OK to send ICMP_SOURCE_QUENCH */ 314 #define IP_FF_SYN_ADDIRE 0x20 /* Add IRE if TCP syn packet */ 315 #define IP_FF_IP6INFO 0x80 /* Add ip6i_t if needed */ 316 #define IP_FF_SEND_SLLA 0x100 /* Send source link layer info ? */ 317 #define IPV6_REACHABILITY_CONFIRMATION 0x200 /* Flags for ip_xmit_v6 */ 318 #define IP_FF_NO_MCAST_LOOP 0x400 /* No multicasts for sending zone */ 319 320 /* 321 * Following flags are used by IPQoS to determine if policy processing is 322 * required. 323 */ 324 #define IP6_NO_IPPOLICY 0x800 /* Don't do IPQoS processing */ 325 #define IP6_IN_LLMCAST 0x1000 /* Multicast */ 326 327 #define IP_FF_LOOPBACK 0x2000 /* Loopback fanout */ 328 329 #ifndef IRE_DB_TYPE 330 #define IRE_DB_TYPE M_SIG 331 #endif 332 333 #ifndef IRE_DB_REQ_TYPE 334 #define IRE_DB_REQ_TYPE M_PCSIG 335 #endif 336 337 #ifndef IRE_ARPRESOLVE_TYPE 338 #define IRE_ARPRESOLVE_TYPE M_EVENT 339 #endif 340 341 /* 342 * Values for squeue switch: 343 */ 344 345 #define IP_SQUEUE_ENTER_NODRAIN 1 346 #define IP_SQUEUE_ENTER 2 347 /* 348 * This is part of the interface between Transport provider and 349 * IP which can be used to set policy information. This is usually 350 * accompanied with O_T_BIND_REQ/T_BIND_REQ.ip_bind assumes that 351 * only IPSEC_POLICY_SET is there when it is found in the chain. 352 * The information contained is an struct ipsec_req_t. On success 353 * or failure, either the T_BIND_ACK or the T_ERROR_ACK is returned. 354 * IPSEC_POLICY_SET is never returned. 355 */ 356 #define IPSEC_POLICY_SET M_SETOPTS 357 358 #define IRE_IS_LOCAL(ire) ((ire != NULL) && \ 359 ((ire)->ire_type & (IRE_LOCAL | IRE_LOOPBACK))) 360 361 #define IRE_IS_TARGET(ire) ((ire != NULL) && \ 362 ((ire)->ire_type != IRE_BROADCAST)) 363 364 /* IP Fragmentation Reassembly Header */ 365 typedef struct ipf_s { 366 struct ipf_s *ipf_hash_next; 367 struct ipf_s **ipf_ptphn; /* Pointer to previous hash next. */ 368 uint32_t ipf_ident; /* Ident to match. */ 369 uint8_t ipf_protocol; /* Protocol to match. */ 370 uchar_t ipf_last_frag_seen : 1; /* Last fragment seen ? */ 371 time_t ipf_timestamp; /* Reassembly start time. */ 372 mblk_t *ipf_mp; /* mblk we live in. */ 373 mblk_t *ipf_tail_mp; /* Frag queue tail pointer. */ 374 int ipf_hole_cnt; /* Number of holes (hard-case). */ 375 int ipf_end; /* Tail end offset (0 -> hard-case). */ 376 uint_t ipf_gen; /* Frag queue generation */ 377 size_t ipf_count; /* Count of bytes used by frag */ 378 uint_t ipf_nf_hdr_len; /* Length of nonfragmented header */ 379 in6_addr_t ipf_v6src; /* IPv6 source address */ 380 in6_addr_t ipf_v6dst; /* IPv6 dest address */ 381 uint_t ipf_prev_nexthdr_offset; /* Offset for nexthdr value */ 382 uint8_t ipf_ecn; /* ECN info for the fragments */ 383 uint8_t ipf_num_dups; /* Number of times dup frags recvd */ 384 uint16_t ipf_checksum_flags; /* Hardware checksum flags */ 385 uint32_t ipf_checksum; /* Partial checksum of fragment data */ 386 } ipf_t; 387 388 #define ipf_src V4_PART_OF_V6(ipf_v6src) 389 #define ipf_dst V4_PART_OF_V6(ipf_v6dst) 390 391 typedef enum { 392 IB_PKT = 0x01, 393 OB_PKT = 0x02 394 } ip_pkt_t; 395 396 #define UPDATE_IB_PKT_COUNT(ire)\ 397 { \ 398 (ire)->ire_ib_pkt_count++; \ 399 if ((ire)->ire_ipif != NULL) { \ 400 /* \ 401 * forwarding packet \ 402 */ \ 403 if ((ire)->ire_type & (IRE_LOCAL|IRE_BROADCAST)) \ 404 atomic_add_32(&(ire)->ire_ipif->ipif_ib_pkt_count, 1);\ 405 else \ 406 atomic_add_32(&(ire)->ire_ipif->ipif_fo_pkt_count, 1);\ 407 } \ 408 } 409 #define UPDATE_OB_PKT_COUNT(ire)\ 410 { \ 411 (ire)->ire_ob_pkt_count++;\ 412 if ((ire)->ire_ipif != NULL) { \ 413 atomic_add_32(&(ire)->ire_ipif->ipif_ob_pkt_count, 1); \ 414 } \ 415 } 416 417 418 #define IP_RPUT_LOCAL(q, mp, ipha, ire, recv_ill) \ 419 { \ 420 switch (ipha->ipha_protocol) { \ 421 case IPPROTO_UDP: \ 422 ip_udp_input(q, mp, ipha, ire, recv_ill); \ 423 break; \ 424 default: \ 425 ip_proto_input(q, mp, ipha, ire, recv_ill); \ 426 break; \ 427 } \ 428 } 429 430 /* 431 * NCE_EXPIRED is TRUE when we have a non-permanent nce that was 432 * found to be REACHABLE more than ip_ire_arp_interval ms ago. 433 * This macro is used to trigger re-arp of existing nce_t entries 434 * so that they can be updated with the latest information. 435 * nce's will get cleaned up (or updated with new info) in the following 436 * circumstances: 437 * - ip_ire_trash_reclaim will free nce's using ndp_cache_reclaim 438 * when memory is low, 439 * - ip_arp_news, when updates are received. 440 * - if the nce is NCE_EXPIRED(), it will be re-initialized to ND_INITIAL, 441 * so that a new arp request will be triggered. 442 * nce_last is the timestamp that indicates when the nce_res_mp in the 443 * nce_t was last updated to a valid link-layer address. nce_last gets 444 * modified/updated : 445 * - when the nce is created or reinit-ed 446 * - every time we get a sane arp response for the nce. 447 */ 448 #define NCE_EXPIRED(nce) (nce->nce_last > 0 && \ 449 ((nce->nce_flags & NCE_F_PERMANENT) == 0) && \ 450 ((TICK_TO_MSEC(lbolt64) - nce->nce_last) > ip_ire_arp_interval)) 451 452 #endif /* _KERNEL */ 453 454 /* ICMP types */ 455 #define ICMP_ECHO_REPLY 0 456 #define ICMP_DEST_UNREACHABLE 3 457 #define ICMP_SOURCE_QUENCH 4 458 #define ICMP_REDIRECT 5 459 #define ICMP_ECHO_REQUEST 8 460 #define ICMP_ROUTER_ADVERTISEMENT 9 461 #define ICMP_ROUTER_SOLICITATION 10 462 #define ICMP_TIME_EXCEEDED 11 463 #define ICMP_PARAM_PROBLEM 12 464 #define ICMP_TIME_STAMP_REQUEST 13 465 #define ICMP_TIME_STAMP_REPLY 14 466 #define ICMP_INFO_REQUEST 15 467 #define ICMP_INFO_REPLY 16 468 #define ICMP_ADDRESS_MASK_REQUEST 17 469 #define ICMP_ADDRESS_MASK_REPLY 18 470 471 /* ICMP_TIME_EXCEEDED codes */ 472 #define ICMP_TTL_EXCEEDED 0 473 #define ICMP_REASSEMBLY_TIME_EXCEEDED 1 474 475 /* ICMP_DEST_UNREACHABLE codes */ 476 #define ICMP_NET_UNREACHABLE 0 477 #define ICMP_HOST_UNREACHABLE 1 478 #define ICMP_PROTOCOL_UNREACHABLE 2 479 #define ICMP_PORT_UNREACHABLE 3 480 #define ICMP_FRAGMENTATION_NEEDED 4 481 #define ICMP_SOURCE_ROUTE_FAILED 5 482 #define ICMP_DEST_NET_UNKNOWN 6 483 #define ICMP_DEST_HOST_UNKNOWN 7 484 #define ICMP_SRC_HOST_ISOLATED 8 485 #define ICMP_DEST_NET_UNREACH_ADMIN 9 486 #define ICMP_DEST_HOST_UNREACH_ADMIN 10 487 #define ICMP_DEST_NET_UNREACH_TOS 11 488 #define ICMP_DEST_HOST_UNREACH_TOS 12 489 490 /* ICMP Header Structure */ 491 typedef struct icmph_s { 492 uint8_t icmph_type; 493 uint8_t icmph_code; 494 uint16_t icmph_checksum; 495 union { 496 struct { /* ECHO request/response structure */ 497 uint16_t u_echo_ident; 498 uint16_t u_echo_seqnum; 499 } u_echo; 500 struct { /* Destination unreachable structure */ 501 uint16_t u_du_zero; 502 uint16_t u_du_mtu; 503 } u_du; 504 struct { /* Parameter problem structure */ 505 uint8_t u_pp_ptr; 506 uint8_t u_pp_rsvd[3]; 507 } u_pp; 508 struct { /* Redirect structure */ 509 ipaddr_t u_rd_gateway; 510 } u_rd; 511 } icmph_u; 512 } icmph_t; 513 514 #define icmph_echo_ident icmph_u.u_echo.u_echo_ident 515 #define icmph_echo_seqnum icmph_u.u_echo.u_echo_seqnum 516 #define icmph_du_zero icmph_u.u_du.u_du_zero 517 #define icmph_du_mtu icmph_u.u_du.u_du_mtu 518 #define icmph_pp_ptr icmph_u.u_pp.u_pp_ptr 519 #define icmph_rd_gateway icmph_u.u_rd.u_rd_gateway 520 521 #define ICMPH_SIZE 8 522 523 /* 524 * Minimum length of transport layer header included in an ICMP error 525 * message for it to be considered valid. 526 */ 527 #define ICMP_MIN_TP_HDR_LEN 8 528 529 /* Aligned IP header */ 530 typedef struct ipha_s { 531 uint8_t ipha_version_and_hdr_length; 532 uint8_t ipha_type_of_service; 533 uint16_t ipha_length; 534 uint16_t ipha_ident; 535 uint16_t ipha_fragment_offset_and_flags; 536 uint8_t ipha_ttl; 537 uint8_t ipha_protocol; 538 uint16_t ipha_hdr_checksum; 539 ipaddr_t ipha_src; 540 ipaddr_t ipha_dst; 541 } ipha_t; 542 543 #define IPH_DF 0x4000 /* Don't fragment */ 544 #define IPH_MF 0x2000 /* More fragments to come */ 545 #define IPH_OFFSET 0x1FFF /* Where the offset lives */ 546 #define IPH_FRAG_HDR 0x8000 /* IPv6 don't fragment bit */ 547 548 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */ 549 #define IPH_ECN_NECT 0x0 /* Not ECN-Capabable Transport */ 550 #define IPH_ECN_ECT1 0x1 /* ECN-Capable Transport, ECT(1) */ 551 #define IPH_ECN_ECT0 0x2 /* ECN-Capable Transport, ECT(0) */ 552 #define IPH_ECN_CE 0x3 /* ECN-Congestion Experienced (CE) */ 553 554 /* IP Mac info structure */ 555 typedef struct ip_m_s { 556 t_uscalar_t ip_m_mac_type; /* From <sys/dlpi.h> */ 557 int ip_m_type; /* From <net/if_types.h> */ 558 boolean_t (*ip_m_v4mapinfo)(uint_t, uint8_t *, uint8_t *, 559 uint32_t *, ipaddr_t *); 560 boolean_t (*ip_m_v6mapinfo)(uint_t, uint8_t *, uint8_t *, 561 uint32_t *, in6_addr_t *); 562 boolean_t (*ip_m_v6intfid)(uint_t, uint8_t *, in6_addr_t *); 563 } ip_m_t; 564 565 /* 566 * The following functions attempt to reduce the link layer dependency 567 * of the IP stack. The current set of link specific operations are: 568 * a. map from IPv4 class D (224.0/4) multicast address range to the link 569 * layer multicast address range. 570 * b. map from IPv6 multicast address range (ff00::/8) to the link 571 * layer multicast address range. 572 * c. derive the default IPv6 interface identifier from the link layer 573 * address. 574 */ 575 #define MEDIA_V4MINFO(ip_m, plen, bphys, maddr, hwxp, v4ptr) \ 576 (((ip_m)->ip_m_v4mapinfo != NULL) && \ 577 (*(ip_m)->ip_m_v4mapinfo)(plen, bphys, maddr, hwxp, v4ptr)) 578 #define MEDIA_V6INTFID(ip_m, plen, phys, v6ptr) \ 579 (((ip_m)->ip_m_v6intfid != NULL) && \ 580 (*(ip_m)->ip_m_v6intfid)(plen, phys, v6ptr)) 581 #define MEDIA_V6MINFO(ip_m, plen, bphys, maddr, hwxp, v6ptr) \ 582 (((ip_m)->ip_m_v6mapinfo != NULL) && \ 583 (*(ip_m)->ip_m_v6mapinfo)(plen, bphys, maddr, hwxp, v6ptr)) 584 585 /* Router entry types */ 586 #define IRE_BROADCAST 0x0001 /* Route entry for broadcast address */ 587 #define IRE_DEFAULT 0x0002 /* Route entry for default gateway */ 588 #define IRE_LOCAL 0x0004 /* Route entry for local address */ 589 #define IRE_LOOPBACK 0x0008 /* Route entry for loopback address */ 590 #define IRE_PREFIX 0x0010 /* Route entry for prefix routes */ 591 #define IRE_CACHE 0x0020 /* Cached Route entry */ 592 #define IRE_IF_NORESOLVER 0x0040 /* Route entry for local interface */ 593 /* net without any address mapping. */ 594 #define IRE_IF_RESOLVER 0x0080 /* Route entry for local interface */ 595 /* net with resolver. */ 596 #define IRE_HOST 0x0100 /* Host route entry */ 597 #define IRE_HOST_REDIRECT 0x0200 /* Host route entry from redirects */ 598 599 /* 600 * IRE_MIPRTUN is only set on the ires in the ip_mrtun_table. 601 * This ire_type must not be set for ftable and ctable routing entries. 602 */ 603 #define IRE_MIPRTUN 0x0400 /* Reverse tunnel route entry */ 604 605 #define IRE_INTERFACE (IRE_IF_NORESOLVER | IRE_IF_RESOLVER) 606 #define IRE_OFFSUBNET (IRE_DEFAULT | IRE_PREFIX | IRE_HOST | \ 607 IRE_HOST_REDIRECT) 608 #define IRE_CACHETABLE (IRE_CACHE | IRE_BROADCAST | IRE_LOCAL | \ 609 IRE_LOOPBACK) 610 #define IRE_FORWARDTABLE (IRE_INTERFACE | IRE_OFFSUBNET) 611 612 /* 613 * If an IRE is marked with IRE_MARK_CONDEMNED, the last walker of 614 * the bucket should delete this IRE from this bucket. 615 */ 616 #define IRE_MARK_CONDEMNED 0x0001 617 /* 618 * If a broadcast IRE is marked with IRE_MARK_NORECV, ip_rput will drop the 619 * broadcast packets received on that interface. This is marked only 620 * on broadcast ires. Employed by IPMP, where we have multiple NICs on the 621 * same subnet receiving the same broadcast packet. 622 */ 623 #define IRE_MARK_NORECV 0x0002 624 /* 625 * IRE_CACHE marked this way won't be returned by ire_cache_lookup. Need 626 * to look specifically using MATCH_IRE_MARK_HIDDEN. Used by IPMP. 627 */ 628 #define IRE_MARK_HIDDEN 0x0004 /* Typically Used by in.mpathd */ 629 630 /* 631 * ire with IRE_MARK_NOADD is created in ip_newroute_ipif, when outgoing 632 * interface is specified by IP_XMIT_IF socket option. This ire is not 633 * added in IRE_CACHE. For example, this is used by mipagent to prevent 634 * any entry to be added in the cache table. We do not want to add any 635 * entry for a mobile-node in the routing table for foreign agent originated 636 * packets. Adding routes in cache table in this case, may run the risks of 637 * incorrect routing path in case of private overlapping addresses. 638 */ 639 #define IRE_MARK_NOADD 0x0008 /* Mark not to add ire in cache */ 640 641 /* 642 * IRE marked with IRE_MARK_TEMPORARY means that this IRE has been used 643 * either for forwarding a packet or has not been used for sending 644 * traffic on TCP connections terminated on this system. In both 645 * cases, this IRE is the first to go when IRE is being cleaned up. 646 */ 647 #define IRE_MARK_TEMPORARY 0x0010 648 649 /* 650 * IRE marked with IRE_MARK_USESRC_CHECK means that while adding an IRE with 651 * this mark, additional atomic checks need to be performed. For eg: by the 652 * time an IRE_CACHE is created, sent up to ARP and then comes back to IP; the 653 * usesrc grouping could have changed in which case we want to fail adding 654 * the IRE_CACHE entry 655 */ 656 #define IRE_MARK_USESRC_CHECK 0x0020 657 658 /* 659 * IRE_MARK_PRIVATE_ADDR is used for IP_NEXTHOP. When IP_NEXTHOP is set, the 660 * routing table lookup for the destination is bypassed and the packet is 661 * sent directly to the specified nexthop. The associated IRE_CACHE entries 662 * should be marked with IRE_MARK_PRIVATE_ADDR flag so that they don't show up 663 * in regular ire cache lookups. 664 */ 665 #define IRE_MARK_PRIVATE_ADDR 0x0040 666 667 /* 668 * When we send an ARP resolution query for the nexthop gateway's ire, 669 * we use esballoc to create the ire_t in the AR_ENTRY_QUERY mblk 670 * chain, and mark its ire_marks with IRE_MARK_UNCACHED. This flag 671 * indicates that information from ARP has not been transferred to a 672 * permanent IRE_CACHE entry. The flag is reset only when the 673 * information is successfully transferred to an ire_cache entry (in 674 * ire_add()). Attempting to free the AR_ENTRY_QUERY mblk chain prior 675 * to ire_add (e.g., from arp, or from ip`ip_wput_nondata) will 676 * require that the resources (incomplete ire_cache and/or nce) must 677 * be cleaned up. The free callback routine (ire_freemblk()) checks 678 * for IRE_MARK_UNCACHED to see if any resources that are pinned down 679 * will need to be cleaned up or not. 680 */ 681 682 #define IRE_MARK_UNCACHED 0x0080 683 684 /* Flags with ire_expire routine */ 685 #define FLUSH_ARP_TIME 0x0001 /* ARP info potentially stale timer */ 686 #define FLUSH_REDIRECT_TIME 0x0002 /* Redirects potentially stale */ 687 #define FLUSH_MTU_TIME 0x0004 /* Include path MTU per RFC 1191 */ 688 689 /* Arguments to ire_flush_cache() */ 690 #define IRE_FLUSH_DELETE 0 691 #define IRE_FLUSH_ADD 1 692 693 /* 694 * Open/close synchronization flags. 695 * These are kept in a separate field in the conn and the synchronization 696 * depends on the atomic 32 bit access to that field. 697 */ 698 #define CONN_CLOSING 0x01 /* ip_close waiting for ip_wsrv */ 699 #define CONN_IPSEC_LOAD_WAIT 0x02 /* waiting for load */ 700 #define CONN_CONDEMNED 0x04 /* conn is closing, no more refs */ 701 #define CONN_INCIPIENT 0x08 /* conn not yet visible, no refs */ 702 #define CONN_QUIESCED 0x10 /* conn is now quiescent */ 703 704 /* 705 * Parameter to ip_output giving the identity of the caller. 706 * IP_WSRV means the packet was enqueued in the STREAMS queue 707 * due to flow control and is now being reprocessed in the context of 708 * the STREAMS service procedure, consequent to flow control relief. 709 * IRE_SEND means the packet is being reprocessed consequent to an 710 * ire cache creation and addition and this may or may not be happening 711 * in the service procedure context. Anything other than the above 2 712 * cases is identified as IP_WPUT. Most commonly this is the case of 713 * packets coming down from the application. 714 */ 715 #ifdef _KERNEL 716 #define IP_WSRV 1 /* Called from ip_wsrv */ 717 #define IP_WPUT 2 /* Called from ip_wput */ 718 #define IRE_SEND 3 /* Called from ire_send */ 719 720 /* 721 * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2) 722 */ 723 #define MAX_FILTER_SIZE 64 724 725 typedef struct slist_s { 726 int sl_numsrc; 727 in6_addr_t sl_addr[MAX_FILTER_SIZE]; 728 } slist_t; 729 730 /* 731 * Following struct is used to maintain retransmission state for 732 * a multicast group. One rtx_state_t struct is an in-line field 733 * of the ilm_t struct; the slist_ts in the rtx_state_t struct are 734 * alloc'd as needed. 735 */ 736 typedef struct rtx_state_s { 737 uint_t rtx_timer; /* retrans timer */ 738 int rtx_cnt; /* retrans count */ 739 int rtx_fmode_cnt; /* retrans count for fmode change */ 740 slist_t *rtx_allow; 741 slist_t *rtx_block; 742 } rtx_state_t; 743 744 /* 745 * Used to construct list of multicast address records that will be 746 * sent in a single listener report. 747 */ 748 typedef struct mrec_s { 749 struct mrec_s *mrec_next; 750 uint8_t mrec_type; 751 uint8_t mrec_auxlen; /* currently unused */ 752 in6_addr_t mrec_group; 753 slist_t mrec_srcs; 754 } mrec_t; 755 756 /* Group membership list per upper conn */ 757 /* 758 * XXX add ilg info for ifaddr/ifindex. 759 * XXX can we make ilg survive an ifconfig unplumb + plumb 760 * by setting the ipif/ill to NULL and recover that later? 761 * 762 * ilg_ipif is used by IPv4 as multicast groups are joined using an interface 763 * address (ipif). 764 * ilg_ill is used by IPv6 as multicast groups are joined using an interface 765 * index (phyint->phyint_ifindex). 766 * ilg_ill is NULL for IPv4 and ilg_ipif is NULL for IPv6. 767 * 768 * ilg records the state of multicast memberships of a socket end point. 769 * ilm records the state of multicast memberships with the driver and is 770 * maintained per interface. 771 * 772 * Notes : 773 * 774 * 1) There is no direct link between a given ilg and ilm. If the 775 * application has joined a group G with ifindex I, we will have 776 * an ilg with ilg_v6group and ilg_ill. There will be a corresponding 777 * ilm with ilm_ill/ilm_v6addr recording the multicast membership. 778 * To delete the membership, 779 * 780 * a) Search for ilg matching on G and I with ilg_v6group 781 * and ilg_ill. Delete ilg_ill. 782 * b) Search the corresponding ilm matching on G and I with 783 * ilm_v6addr and ilm_ill. Delete ilm. 784 * 785 * In IPv4, the only difference is, we look using ipifs instead of 786 * ills. 787 * 788 * 2) With IP multipathing, we want to keep receiving even after the 789 * interface has failed. We do this by moving multicast memberships 790 * to a new_ill within the group. This is acheived by sending 791 * DL_DISABMULTI_REQS on ilg_ill/ilm_ill and sending DL_ENABMULTIREQS 792 * on the new_ill and changing ilg_ill/ilm_ill to new_ill. But, we 793 * need to be able to delete memberships which will still come down 794 * with the ifindex of the old ill which is what the application 795 * knows of. Thus we store the ilm_/ilg_orig_ifindex to keep track 796 * of where we joined initially so that we can lookup even after we 797 * moved the membership. It is also used for moving back the membership 798 * when the old ill has been repaired. This is done by looking up for 799 * ilms with ilm_orig_ifindex matching on the old ill's ifindex. Only 800 * ilms actually move from old ill to new ill. ilgs don't move (just 801 * the ilg_ill is changed when it moves) as it just records the state 802 * of the application that has joined a group G where as ilm records 803 * the state joined with the driver. Thus when we send DL_XXXMULTI_REQs 804 * we also need to keep the ilm in the right ill. 805 * 806 * In IPv4, as ipifs move from old ill to new_ill, ilgs and ilms move 807 * implicitly as we use only ipifs in IPv4. Thus, one can always lookup 808 * a given ilm/ilg even after it fails without the support of 809 * orig_ifindex. We move ilms still to record the driver state as 810 * mentioned above. 811 */ 812 813 /* 814 * The ilg_t and ilm_t members are protected by ipsq. They can be changed only 815 * by a thread executing in the ipsq. In other words add/delete of a 816 * multicast group has to execute in the ipsq. 817 */ 818 #define ILG_DELETED 0x1 /* ilg_flags */ 819 typedef struct ilg_s { 820 in6_addr_t ilg_v6group; 821 struct ipif_s *ilg_ipif; /* Logical interface we are member on */ 822 struct ill_s *ilg_ill; /* Used by IPv6 */ 823 int ilg_orig_ifindex; /* Interface originally joined on */ 824 uint_t ilg_flags; 825 mcast_record_t ilg_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */ 826 slist_t *ilg_filter; 827 } ilg_t; 828 829 830 /* 831 * Multicast address list entry for lower ill. 832 * ilm_ipif is used by IPv4 as multicast groups are joined using ipif. 833 * ilm_ill is used by IPv6 as multicast groups are joined using ill. 834 * ilm_ill is NULL for IPv4 and ilm_ipif is NULL for IPv6. 835 */ 836 #define ILM_DELETED 0x1 /* ilm_flags */ 837 typedef struct ilm_s { 838 in6_addr_t ilm_v6addr; 839 int ilm_refcnt; 840 uint_t ilm_timer; /* IGMP/MLD query resp timer, in msec */ 841 struct ipif_s *ilm_ipif; /* Back pointer to ipif for IPv4 */ 842 struct ilm_s *ilm_next; /* Linked list for each ill */ 843 uint_t ilm_state; /* state of the membership */ 844 struct ill_s *ilm_ill; /* Back pointer to ill for IPv6 */ 845 int ilm_orig_ifindex; /* V6_MULTICAST_IF/ilm_ipif index */ 846 uint_t ilm_flags; 847 boolean_t ilm_is_new; /* new ilm */ 848 boolean_t ilm_notify_driver; /* Need to notify the driver */ 849 zoneid_t ilm_zoneid; 850 int ilm_no_ilg_cnt; /* number of joins w/ no ilg */ 851 mcast_record_t ilm_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */ 852 slist_t *ilm_filter; /* source filter list */ 853 slist_t *ilm_pendsrcs; /* relevant src addrs for pending req */ 854 rtx_state_t ilm_rtx; /* SCR retransmission state */ 855 } ilm_t; 856 857 #define ilm_addr V4_PART_OF_V6(ilm_v6addr) 858 859 /* 860 * ilm_walker_cleanup needs to execute when the ilm_walker_cnt goes down to 861 * zero. In addition it needs to block new walkers while it is unlinking ilm's 862 * from the list. Thus simple atomics for the ill_ilm_walker_cnt don't suffice. 863 */ 864 #define ILM_WALKER_HOLD(ill) { \ 865 mutex_enter(&(ill)->ill_lock); \ 866 ill->ill_ilm_walker_cnt++; \ 867 mutex_exit(&(ill)->ill_lock); \ 868 } 869 870 #define ILM_WALKER_RELE(ill) { \ 871 mutex_enter(&(ill)->ill_lock); \ 872 (ill)->ill_ilm_walker_cnt--; \ 873 if ((ill)->ill_ilm_walker_cnt == 0 && (ill)->ill_ilm_cleanup_reqd) \ 874 ilm_walker_cleanup(ill); \ 875 mutex_exit(&(ill)->ill_lock); \ 876 } 877 878 /* 879 * Soft reference to an IPsec SA. 880 * 881 * On relative terms, conn's can be persistant (living as long as the 882 * processes which create them), while SA's are ephemeral (dying when 883 * they hit their time-based or byte-based lifetimes). 884 * 885 * We could hold a hard reference to an SA from an ipsec_latch_t, 886 * but this would cause expired SA's to linger for a potentially 887 * unbounded time. 888 * 889 * Instead, we remember the hash bucket number and bucket generation 890 * in addition to the pointer. The bucket generation is incremented on 891 * each deletion. 892 */ 893 typedef struct ipsa_ref_s 894 { 895 struct ipsa_s *ipsr_sa; 896 struct isaf_s *ipsr_bucket; 897 uint64_t ipsr_gen; 898 } ipsa_ref_t; 899 900 /* 901 * IPsec "latching" state. 902 * 903 * In the presence of IPsec policy, fully-bound conn's bind a connection 904 * to more than just the 5-tuple, but also a specific IPsec action and 905 * identity-pair. 906 * 907 * As an optimization, we also cache soft references to IPsec SA's 908 * here so that we can fast-path around most of the work needed for 909 * outbound IPsec SA selection. 910 * 911 * Were it not for TCP's detached connections, this state would be 912 * in-line in conn_t; instead, this is in a separate structure so it 913 * can be handed off to TCP when a connection is detached. 914 */ 915 typedef struct ipsec_latch_s 916 { 917 kmutex_t ipl_lock; 918 uint32_t ipl_refcnt; 919 920 uint64_t ipl_unique; 921 struct ipsec_policy_s *ipl_in_policy; /* latched policy (in) */ 922 struct ipsec_policy_s *ipl_out_policy; /* latched policy (out) */ 923 struct ipsec_action_s *ipl_in_action; /* latched action (in) */ 924 struct ipsec_action_s *ipl_out_action; /* latched action (out) */ 925 cred_t *ipl_local_id; 926 struct ipsid_s *ipl_local_cid; 927 struct ipsid_s *ipl_remote_cid; 928 unsigned int 929 ipl_out_action_latched : 1, 930 ipl_in_action_latched : 1, 931 ipl_out_policy_latched : 1, 932 ipl_in_policy_latched : 1, 933 934 ipl_ids_latched : 1, 935 936 ipl_pad_to_bit_31 : 27; 937 938 ipsa_ref_t ipl_ref[2]; /* 0: ESP, 1: AH */ 939 940 } ipsec_latch_t; 941 942 #define IPLATCH_REFHOLD(ipl) { \ 943 atomic_add_32(&(ipl)->ipl_refcnt, 1); \ 944 ASSERT((ipl)->ipl_refcnt != 0); \ 945 } 946 947 #define IPLATCH_REFRELE(ipl) { \ 948 ASSERT((ipl)->ipl_refcnt != 0); \ 949 membar_exit(); \ 950 if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0) \ 951 iplatch_free(ipl); \ 952 } 953 954 /* 955 * peer identity structure. 956 */ 957 958 typedef struct conn_s conn_t; 959 960 /* 961 * The old IP client structure "ipc_t" is gone. All the data is stored in the 962 * connection structure "conn_t" now. The mapping of old and new fields looks 963 * like this: 964 * 965 * ipc_ulp conn_ulp 966 * ipc_rq conn_rq 967 * ipc_wq conn_wq 968 * 969 * ipc_laddr conn_src 970 * ipc_faddr conn_rem 971 * ipc_v6laddr conn_srcv6 972 * ipc_v6faddr conn_remv6 973 * 974 * ipc_lport conn_lport 975 * ipc_fport conn_fport 976 * ipc_ports conn_ports 977 * 978 * ipc_policy conn_policy 979 * ipc_latch conn_latch 980 * 981 * ipc_irc_lock conn_lock 982 * ipc_ire_cache conn_ire_cache 983 * 984 * ipc_state_flags conn_state_flags 985 * ipc_outgoing_ill conn_outgoing_ill 986 * 987 * ipc_dontroute conn_dontroute 988 * ipc_loopback conn_loopback 989 * ipc_broadcast conn_broadcast 990 * ipc_reuseaddr conn_reuseaddr 991 * 992 * ipc_multicast_loop conn_multicast_loop 993 * ipc_multi_router conn_multi_router 994 * ipc_priv_stream conn_priv_stream 995 * ipc_draining conn_draining 996 * 997 * ipc_did_putbq conn_did_putbq 998 * ipc_unspec_src conn_unspec_src 999 * ipc_policy_cached conn_policy_cached 1000 * 1001 * ipc_in_enforce_policy conn_in_enforce_policy 1002 * ipc_out_enforce_policy conn_out_enforce_policy 1003 * ipc_af_isv6 conn_af_isv6 1004 * ipc_pkt_isv6 conn_pkt_isv6 1005 * 1006 * ipc_ipv6_recvpktinfo conn_ipv6_recvpktinfo 1007 * 1008 * ipc_ipv6_recvhoplimit conn_ipv6_recvhoplimit 1009 * ipc_ipv6_recvhopopts conn_ipv6_recvhopopts 1010 * ipc_ipv6_recvdstopts conn_ipv6_recvdstopts 1011 * 1012 * ipc_ipv6_recvrthdr conn_ipv6_recvrthdr 1013 * ipc_ipv6_recvrtdstopts conn_ipv6_recvrtdstopts 1014 * ipc_fully_bound conn_fully_bound 1015 * 1016 * ipc_recvif conn_recvif 1017 * 1018 * ipc_recvslla conn_recvslla 1019 * ipc_acking_unbind conn_acking_unbind 1020 * ipc_pad_to_bit_31 conn_pad_to_bit_31 1021 * 1022 * ipc_xmit_if_ill conn_xmit_if_ill 1023 * ipc_nofailover_ill conn_nofailover_ill 1024 * 1025 * ipc_proto conn_proto 1026 * ipc_incoming_ill conn_incoming_ill 1027 * ipc_outgoing_pill conn_outgoing_pill 1028 * ipc_pending_ill conn_pending_ill 1029 * ipc_unbind_mp conn_unbind_mp 1030 * ipc_ilg conn_ilg 1031 * ipc_ilg_allocated conn_ilg_allocated 1032 * ipc_ilg_inuse conn_ilg_inuse 1033 * ipc_ilg_walker_cnt conn_ilg_walker_cnt 1034 * ipc_refcv conn_refcv 1035 * ipc_multicast_ipif conn_multicast_ipif 1036 * ipc_multicast_ill conn_multicast_ill 1037 * ipc_orig_bound_ifindex conn_orig_bound_ifindex 1038 * ipc_orig_multicast_ifindex conn_orig_multicast_ifindex 1039 * ipc_orig_xmit_ifindex conn_orig_xmit_ifindex 1040 * ipc_drain_next conn_drain_next 1041 * ipc_drain_prev conn_drain_prev 1042 * ipc_idl conn_idl 1043 */ 1044 1045 /* 1046 * This is used to match an inbound/outbound datagram with 1047 * policy. 1048 */ 1049 1050 typedef struct ipsec_selector { 1051 in6_addr_t ips_local_addr_v6; 1052 in6_addr_t ips_remote_addr_v6; 1053 uint16_t ips_local_port; 1054 uint16_t ips_remote_port; 1055 uint8_t ips_icmp_type; 1056 uint8_t ips_icmp_code; 1057 uint8_t ips_protocol; 1058 uint8_t ips_isv4 : 1, 1059 ips_is_icmp_inv_acq: 1; 1060 } ipsec_selector_t; 1061 1062 /* 1063 * Note that we put v4 addresses in the *first* 32-bit word of the 1064 * selector rather than the last to simplify the prefix match/mask code 1065 * in spd.c 1066 */ 1067 #define ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0] 1068 #define ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0] 1069 1070 /* Values used in IP by IPSEC Code */ 1071 #define IPSEC_OUTBOUND B_TRUE 1072 #define IPSEC_INBOUND B_FALSE 1073 1074 /* 1075 * There are two variants in policy failures. The packet may come in 1076 * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not 1077 * have the desired level of protection (IPSEC_POLICY_MISMATCH). 1078 */ 1079 #define IPSEC_POLICY_NOT_NEEDED 0 1080 #define IPSEC_POLICY_MISMATCH 1 1081 #define IPSEC_POLICY_AUTH_NOT_NEEDED 2 1082 #define IPSEC_POLICY_ENCR_NOT_NEEDED 3 1083 #define IPSEC_POLICY_SE_NOT_NEEDED 4 1084 #define IPSEC_POLICY_MAX 5 /* Always max + 1. */ 1085 1086 /* 1087 * Folowing macro is used whenever the code does not know whether there 1088 * is a M_CTL present in the front and it needs to examine the actual mp 1089 * i.e the IP header. As a M_CTL message could be in the front, this 1090 * extracts the packet into mp and the M_CTL mp into first_mp. If M_CTL 1091 * mp is not present, both first_mp and mp point to the same message. 1092 */ 1093 #define EXTRACT_PKT_MP(mp, first_mp, mctl_present) \ 1094 (first_mp) = (mp); \ 1095 if ((mp)->b_datap->db_type == M_CTL) { \ 1096 (mp) = (mp)->b_cont; \ 1097 (mctl_present) = B_TRUE; \ 1098 } else { \ 1099 (mctl_present) = B_FALSE; \ 1100 } 1101 1102 /* 1103 * Check with IPSEC inbound policy if 1104 * 1105 * 1) per-socket policy is present - indicated by conn_in_enforce_policy. 1106 * 2) Or if we have not cached policy on the conn and the global policy is 1107 * non-empty. 1108 */ 1109 #define CONN_INBOUND_POLICY_PRESENT(connp) \ 1110 ((connp)->conn_in_enforce_policy || \ 1111 (!((connp)->conn_policy_cached) && \ 1112 ipsec_inbound_v4_policy_present)) 1113 1114 #define CONN_INBOUND_POLICY_PRESENT_V6(connp) \ 1115 ((connp)->conn_in_enforce_policy || \ 1116 (!(connp)->conn_policy_cached && \ 1117 ipsec_inbound_v6_policy_present)) 1118 1119 #define CONN_OUTBOUND_POLICY_PRESENT(connp) \ 1120 ((connp)->conn_out_enforce_policy || \ 1121 (!((connp)->conn_policy_cached) && \ 1122 ipsec_outbound_v4_policy_present)) 1123 1124 #define CONN_OUTBOUND_POLICY_PRESENT_V6(connp) \ 1125 ((connp)->conn_out_enforce_policy || \ 1126 (!(connp)->conn_policy_cached && \ 1127 ipsec_outbound_v6_policy_present)) 1128 1129 /* 1130 * Information cached in IRE for upper layer protocol (ULP). 1131 * 1132 * Notice that ire_max_frag is not included in the iulp_t structure, which 1133 * it may seem that it should. But ire_max_frag cannot really be cached. It 1134 * is fixed for each interface. For MTU found by PMTUd, we may want to cache 1135 * it. But currently, we do not do that. 1136 */ 1137 typedef struct iulp_s { 1138 boolean_t iulp_set; /* Is any metric set? */ 1139 uint32_t iulp_ssthresh; /* Slow start threshold (TCP). */ 1140 clock_t iulp_rtt; /* Guestimate in millisecs. */ 1141 clock_t iulp_rtt_sd; /* Cached value of RTT variance. */ 1142 uint32_t iulp_spipe; /* Send pipe size. */ 1143 uint32_t iulp_rpipe; /* Receive pipe size. */ 1144 uint32_t iulp_rtomax; /* Max round trip timeout. */ 1145 uint32_t iulp_sack; /* Use SACK option (TCP)? */ 1146 uint32_t 1147 iulp_tstamp_ok : 1, /* Use timestamp option (TCP)? */ 1148 iulp_wscale_ok : 1, /* Use window scale option (TCP)? */ 1149 iulp_ecn_ok : 1, /* Enable ECN (for TCP)? */ 1150 iulp_pmtud_ok : 1, /* Enable PMTUd? */ 1151 1152 iulp_not_used : 28; 1153 } iulp_t; 1154 1155 /* Zero iulp_t. */ 1156 extern const iulp_t ire_uinfo_null; 1157 1158 /* 1159 * The conn drain list structure. 1160 * The list is protected by idl_lock. Each conn_t inserted in the list 1161 * points back at this idl_t using conn_idl. IP primes the draining of the 1162 * conns queued in these lists, by qenabling the 1st conn of each list. This 1163 * occurs when STREAMS backenables ip_wsrv on the IP module. Each conn instance 1164 * of ip_wsrv successively qenables the next conn in the list. 1165 * idl_lock protects all other members of idl_t and conn_drain_next 1166 * and conn_drain_prev of conn_t. The conn_lock protects IPCF_DRAIN_DISABLED 1167 * flag of the conn_t and conn_idl. 1168 */ 1169 typedef struct idl_s { 1170 conn_t *idl_conn; /* Head of drain list */ 1171 kmutex_t idl_lock; /* Lock for this list */ 1172 conn_t *idl_conn_draining; /* conn that is draining */ 1173 uint32_t 1174 idl_repeat : 1, /* Last conn must re-enable */ 1175 /* drain list again */ 1176 idl_unused : 31; 1177 } idl_t; 1178 1179 #define CONN_DRAIN_LIST_LOCK(connp) (&((connp)->conn_idl->idl_lock)) 1180 /* 1181 * Interface route structure which holds the necessary information to recreate 1182 * routes that are tied to an interface (namely where ire_ipif != NULL). 1183 * These routes which were initially created via a routing socket or via the 1184 * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be 1185 * traditional interface routes. When an interface comes back up after being 1186 * marked down, this information will be used to recreate the routes. These 1187 * are part of an mblk_t chain that hangs off of the IPIF (ipif_saved_ire_mp). 1188 */ 1189 typedef struct ifrt_s { 1190 ushort_t ifrt_type; /* Type of IRE */ 1191 in6_addr_t ifrt_v6addr; /* Address IRE represents. */ 1192 in6_addr_t ifrt_v6gateway_addr; /* Gateway if IRE_OFFSUBNET */ 1193 in6_addr_t ifrt_v6src_addr; /* Src addr if RTF_SETSRC */ 1194 in6_addr_t ifrt_v6mask; /* Mask for matching IRE. */ 1195 uint32_t ifrt_flags; /* flags related to route */ 1196 uint_t ifrt_max_frag; /* MTU (next hop or path). */ 1197 iulp_t ifrt_iulp_info; /* Cached IRE ULP info. */ 1198 } ifrt_t; 1199 1200 #define ifrt_addr V4_PART_OF_V6(ifrt_v6addr) 1201 #define ifrt_gateway_addr V4_PART_OF_V6(ifrt_v6gateway_addr) 1202 #define ifrt_src_addr V4_PART_OF_V6(ifrt_v6src_addr) 1203 #define ifrt_mask V4_PART_OF_V6(ifrt_v6mask) 1204 1205 /* Number of IP addresses that can be hosted on a physical interface */ 1206 #define MAX_ADDRS_PER_IF 8192 1207 /* 1208 * Number of Source addresses to be considered for source address 1209 * selection. Used by ipif_select_source[_v6]. 1210 */ 1211 #define MAX_IPIF_SELECT_SOURCE 50 1212 1213 #ifdef IP_DEBUG 1214 /* 1215 * Tracing refholds and refreleases for debugging. Existing tracing mechanisms 1216 * do not allow the granularity need to trace refrences to ipif/ill/ire's. This 1217 * mechanism should be revisited once dtrace is available. 1218 */ 1219 #define IP_STACK_DEPTH 15 1220 typedef struct tr_buf_s { 1221 int tr_depth; 1222 pc_t tr_stack[IP_STACK_DEPTH]; 1223 } tr_buf_t; 1224 1225 typedef struct th_trace_s { 1226 struct th_trace_s *th_next; 1227 struct th_trace_s **th_prev; 1228 kthread_t *th_id; 1229 int th_refcnt; 1230 uint_t th_trace_lastref; 1231 #define TR_BUF_MAX 38 1232 tr_buf_t th_trbuf[TR_BUF_MAX]; 1233 } th_trace_t; 1234 #endif 1235 1236 /* The following are ipif_state_flags */ 1237 #define IPIF_CONDEMNED 0x1 /* The ipif is being removed */ 1238 #define IPIF_CHANGING 0x2 /* A critcal ipif field is changing */ 1239 #define IPIF_MOVING 0x8 /* The ipif is being moved */ 1240 #define IPIF_SET_LINKLOCAL 0x10 /* transient flag during bringup */ 1241 #define IPIF_ZERO_SOURCE 0x20 /* transient flag during bringup */ 1242 1243 /* IP interface structure, one per local address */ 1244 typedef struct ipif_s { 1245 struct ipif_s *ipif_next; 1246 struct ill_s *ipif_ill; /* Back pointer to our ill */ 1247 int ipif_id; /* Logical unit number */ 1248 uint_t ipif_mtu; /* Starts at ipif_ill->ill_max_frag */ 1249 uint_t ipif_saved_mtu; /* Save of mtu during ipif_move() */ 1250 in6_addr_t ipif_v6lcl_addr; /* Local IP address for this if. */ 1251 in6_addr_t ipif_v6src_addr; /* Source IP address for this if. */ 1252 in6_addr_t ipif_v6subnet; /* Subnet prefix for this if. */ 1253 in6_addr_t ipif_v6net_mask; /* Net mask for this interface. */ 1254 in6_addr_t ipif_v6brd_addr; /* Broadcast addr for this interface. */ 1255 in6_addr_t ipif_v6pp_dst_addr; /* Point-to-point dest address. */ 1256 uint64_t ipif_flags; /* Interface flags. */ 1257 uint_t ipif_metric; /* BSD if metric, for compatibility. */ 1258 uint_t ipif_ire_type; /* IRE_LOCAL or IRE_LOOPBACK */ 1259 mblk_t *ipif_arp_del_mp; /* Allocated at time arp comes up, to */ 1260 /* prevent awkward out of mem */ 1261 /* condition later */ 1262 mblk_t *ipif_saved_ire_mp; /* Allocated for each extra */ 1263 /* IRE_IF_NORESOLVER/IRE_IF_RESOLVER */ 1264 /* on this interface so that they */ 1265 /* can survive ifconfig down. */ 1266 kmutex_t ipif_saved_ire_lock; /* Protects ipif_saved_ire_mp */ 1267 1268 mrec_t *ipif_igmp_rpt; /* List of group memberships which */ 1269 /* will be reported on. Used when */ 1270 /* handling an igmp timeout. */ 1271 1272 /* 1273 * The packet counts in the ipif contain the sum of the 1274 * packet counts in dead IREs that were affiliated with 1275 * this ipif. 1276 */ 1277 uint_t ipif_fo_pkt_count; /* Forwarded thru our dead IREs */ 1278 uint_t ipif_ib_pkt_count; /* Inbound packets for our dead IREs */ 1279 uint_t ipif_ob_pkt_count; /* Outbound packets to our dead IREs */ 1280 /* Exclusive bit fields, protected by ipsq_t */ 1281 unsigned int 1282 ipif_multicast_up : 1, /* We have joined the allhosts group */ 1283 ipif_solmcast_up : 1, /* We joined solicited node mcast */ 1284 ipif_replace_zero : 1, /* Replacement for zero */ 1285 ipif_was_up : 1, /* ipif was up before */ 1286 1287 ipif_pad_to_31 : 28; 1288 1289 int ipif_orig_ifindex; /* ifindex before SLIFFAILOVER */ 1290 uint_t ipif_seqid; /* unique index across all ills */ 1291 uint_t ipif_orig_ipifid; /* ipif_id before SLIFFAILOVER */ 1292 uint_t ipif_state_flags; /* See IPIF_* flag defs above */ 1293 uint_t ipif_refcnt; /* active consistent reader cnt */ 1294 uint_t ipif_ire_cnt; /* Number of ire's referencing ipif */ 1295 uint_t ipif_saved_ire_cnt; 1296 zoneid_t 1297 ipif_zoneid; /* zone ID number */ 1298 #ifdef ILL_DEBUG 1299 #define IP_TR_HASH_MAX 64 1300 th_trace_t *ipif_trace[IP_TR_HASH_MAX]; 1301 boolean_t ipif_trace_disable; /* True when alloc fails */ 1302 #endif 1303 } ipif_t; 1304 1305 /* 1306 * The following table lists the protection levels of the various members 1307 * of the ipif_t. The following notation is used. 1308 * 1309 * Write once - Written to only once at the time of bringing up 1310 * the interface and can be safely read after the bringup without any lock. 1311 * 1312 * ipsq - Need to execute in the ipsq to perform the indicated access. 1313 * 1314 * ill_lock - Need to hold this mutex to perform the indicated access. 1315 * 1316 * ill_g_lock - Need to hold this rw lock as reader/writer for read access or 1317 * write access respectively. 1318 * 1319 * down ill - Written to only when the ill is down (i.e all ipifs are down) 1320 * up ill - Read only when the ill is up (i.e. at least 1 ipif is up) 1321 * 1322 * Table of ipif_t members and their protection 1323 * 1324 * ipif_next ill_g_lock ill_g_lock 1325 * ipif_ill ipsq + down ipif write once 1326 * ipif_id ipsq + down ipif write once 1327 * ipif_mtu ipsq 1328 * ipif_v6lcl_addr ipsq + down ipif up ipif 1329 * ipif_v6src_addr ipsq + down ipif up ipif 1330 * ipif_v6subnet ipsq + down ipif up ipif 1331 * ipif_v6net_mask ipsq + down ipif up ipif 1332 * 1333 * ipif_v6brd_addr 1334 * ipif_v6pp_dst_addr 1335 * ipif_flags ill_lock ill_lock 1336 * ipif_metric 1337 * ipif_ire_type ipsq + down ill up ill 1338 * 1339 * ipif_arp_del_mp ipsq ipsq 1340 * ipif_saved_ire_mp ipif_saved_ire_lock ipif_saved_ire_lock 1341 * ipif_igmp_rpt ipsq ipsq 1342 * 1343 * ipif_fo_pkt_count Approx 1344 * ipif_ib_pkt_count Approx 1345 * ipif_ob_pkt_count Approx 1346 * 1347 * bit fields ill_lock ill_lock 1348 * 1349 * ipif_orig_ifindex ipsq None 1350 * ipif_orig_ipifid ipsq None 1351 * ipif_seqid ipsq Write once 1352 * 1353 * ipif_state_flags ill_lock ill_lock 1354 * ipif_refcnt ill_lock ill_lock 1355 * ipif_ire_cnt ill_lock ill_lock 1356 * ipif_saved_ire_cnt 1357 */ 1358 1359 #define IP_TR_HASH(tid) ((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1)) 1360 1361 #ifdef ILL_DEBUG 1362 #define IPIF_TRACE_REF(ipif) ipif_trace_ref(ipif) 1363 #define ILL_TRACE_REF(ill) ill_trace_ref(ill) 1364 #define IPIF_UNTRACE_REF(ipif) ipif_untrace_ref(ipif) 1365 #define ILL_UNTRACE_REF(ill) ill_untrace_ref(ill) 1366 #define ILL_TRACE_CLEANUP(ill) ill_trace_cleanup(ill) 1367 #define IPIF_TRACE_CLEANUP(ipif) ipif_trace_cleanup(ipif) 1368 #else 1369 #define IPIF_TRACE_REF(ipif) 1370 #define ILL_TRACE_REF(ill) 1371 #define IPIF_UNTRACE_REF(ipif) 1372 #define ILL_UNTRACE_REF(ill) 1373 #define ILL_TRACE_CLEANUP(ill) 1374 #define IPIF_TRACE_CLEANUP(ipif) 1375 #endif 1376 1377 /* IPv4 compatability macros */ 1378 #define ipif_lcl_addr V4_PART_OF_V6(ipif_v6lcl_addr) 1379 #define ipif_src_addr V4_PART_OF_V6(ipif_v6src_addr) 1380 #define ipif_subnet V4_PART_OF_V6(ipif_v6subnet) 1381 #define ipif_net_mask V4_PART_OF_V6(ipif_v6net_mask) 1382 #define ipif_brd_addr V4_PART_OF_V6(ipif_v6brd_addr) 1383 #define ipif_pp_dst_addr V4_PART_OF_V6(ipif_v6pp_dst_addr) 1384 1385 /* Macros for easy backreferences to the ill. */ 1386 #define ipif_wq ipif_ill->ill_wq 1387 #define ipif_rq ipif_ill->ill_rq 1388 #define ipif_net_type ipif_ill->ill_net_type 1389 #define ipif_resolver_mp ipif_ill->ill_resolver_mp 1390 #define ipif_ipif_up_count ipif_ill->ill_ipif_up_count 1391 #define ipif_bcast_mp ipif_ill->ill_bcast_mp 1392 #define ipif_type ipif_ill->ill_type 1393 #define ipif_isv6 ipif_ill->ill_isv6 1394 1395 #define SIOCLIFADDR_NDX 112 /* ndx of SIOCLIFADDR in the ndx ioctl table */ 1396 1397 /* 1398 * mode value for ip_ioctl_finish for finishing an ioctl 1399 */ 1400 #define CONN_CLOSE 1 /* No mi_copy */ 1401 #define COPYOUT 2 /* do an mi_copyout if needed */ 1402 #define NO_COPYOUT 3 /* do an mi_copy_done */ 1403 1404 /* 1405 * The IP-MT design revolves around the serialization object ipsq_t. 1406 * It is associated with an IPMP group. If IPMP is not enabled, there is 1407 * 1 ipsq_t per phyint. Eg. an ipsq_t would cover both hme0's IPv4 stream 1408 * 1409 * ipsq_lock protects 1410 * ipsq_reentry_cnt, ipsq_writer, ipsq_xopq_mphead, ipsq_xopq_mptail, 1411 * ipsq_mphead, ipsq_mptail, ipsq_split 1412 * 1413 * ipsq_pending_ipif, ipsq_current_ipif, ipsq_pending_mp, ipsq_flags, 1414 * ipsq_waitfor 1415 * 1416 * The fields in the last line above below are set mostly by a writer thread 1417 * But there is an exception in the last call to ipif_ill_refrele_tail which 1418 * could also race with a conn close which could be cleaning up the 1419 * fields. So we choose to protect using ipsq_lock instead of depending on 1420 * the property of the writer. 1421 * ill_g_lock protects 1422 * ipsq_refs, ipsq_phyint_list 1423 */ 1424 typedef struct ipsq_s { 1425 kmutex_t ipsq_lock; 1426 int ipsq_reentry_cnt; 1427 kthread_t *ipsq_writer; /* current owner (thread id) */ 1428 int ipsq_flags; 1429 mblk_t *ipsq_xopq_mphead; /* list of excl ops mostly ioctls */ 1430 mblk_t *ipsq_xopq_mptail; 1431 mblk_t *ipsq_mphead; /* msgs on ipsq linked thru b_next */ 1432 mblk_t *ipsq_mptail; /* msgs on ipsq linked thru b_next */ 1433 ipif_t *ipsq_pending_ipif; /* ipif associated w. ipsq_pending_mp */ 1434 ipif_t *ipsq_current_ipif; /* ipif associated with current ioctl */ 1435 mblk_t *ipsq_pending_mp; /* current ioctl mp while waiting for */ 1436 /* response from another module */ 1437 struct ipsq_s *ipsq_next; /* list of all syncq's (ipsq_g_list) */ 1438 uint_t ipsq_refs; /* Number of phyints on this ipsq */ 1439 struct phyint *ipsq_phyint_list; /* List of phyints on this ipsq */ 1440 boolean_t ipsq_split; /* ipsq may need to be split */ 1441 int ipsq_waitfor; /* Values encoded below */ 1442 char ipsq_name[LIFNAMSIZ+1]; /* same as phyint_groupname */ 1443 int ipsq_last_cmd; /* debugging aid */ 1444 #ifdef ILL_DEBUG 1445 int ipsq_depth; /* debugging aid */ 1446 pc_t ipsq_stack[IP_STACK_DEPTH]; /* debugging aid */ 1447 #endif 1448 } ipsq_t; 1449 1450 /* ipsq_flags */ 1451 #define IPSQ_GROUP 0x1 /* This ipsq belongs to an IPMP group */ 1452 1453 /* 1454 * ipsq_waitfor: 1455 * 1456 * IPIF_DOWN 1 ipif_down waiting for refcnts to drop 1457 * ILL_DOWN 2 ill_down waiting for refcnts to drop 1458 * IPIF_FREE 3 ipif_free waiting for refcnts to drop 1459 * ILL_FREE 4 ill unplumb waiting for refcnts to drop 1460 * ILL_MOVE_OK 5 failover waiting for refcnts to drop 1461 */ 1462 1463 enum { IPIF_DOWN = 1, ILL_DOWN, IPIF_FREE, ILL_FREE, ILL_MOVE_OK }; 1464 1465 /* Flags passed to ipsq_try_enter */ 1466 #define CUR_OP 0 /* Current ioctl continuing again */ 1467 #define NEW_OP 1 /* New ioctl starting afresh */ 1468 1469 /* 1470 * phyint represents state that is common to both IPv4 and IPv6 interfaces. 1471 * There is a separate ill_t representing IPv4 and IPv6 which has a 1472 * backpointer to the phyint structure for acessing common state. 1473 * 1474 * NOTE : It just stores the group name as there is only one name for 1475 * IPv4 and IPv6 i.e it is a underlying link property. Actually 1476 * IPv4 and IPv6 ill are grouped together when their phyints have 1477 * the same name. 1478 */ 1479 typedef struct phyint { 1480 struct ill_s *phyint_illv4; 1481 struct ill_s *phyint_illv6; 1482 uint_t phyint_ifindex; /* SIOCLSLIFINDEX */ 1483 uint_t phyint_notify_delay; /* SIOCSLIFNOTIFYDELAY */ 1484 char *phyint_groupname; /* SIOCSLIFGROUPNAME */ 1485 uint_t phyint_groupname_len; 1486 uint64_t phyint_flags; 1487 avl_node_t phyint_avl_by_index; /* avl tree by index */ 1488 avl_node_t phyint_avl_by_name; /* avl tree by name */ 1489 kmutex_t phyint_lock; 1490 struct ipsq_s *phyint_ipsq; /* back pointer to ipsq */ 1491 struct phyint *phyint_ipsq_next; /* phyint list on this ipsq */ 1492 } phyint_t; 1493 1494 #define CACHE_ALIGN_SIZE 64 1495 1496 #define CACHE_ALIGN(align_struct) P2ROUNDUP(sizeof (struct align_struct),\ 1497 CACHE_ALIGN_SIZE) 1498 struct _phyint_list_s_ { 1499 avl_tree_t phyint_list_avl_by_index; /* avl tree by index */ 1500 avl_tree_t phyint_list_avl_by_name; /* avl tree by name */ 1501 }; 1502 1503 typedef union phyint_list_u { 1504 struct _phyint_list_s_ phyint_list_s; 1505 char phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)]; 1506 } phyint_list_t; 1507 1508 #define phyint_list_avl_by_index phyint_list_s.phyint_list_avl_by_index 1509 #define phyint_list_avl_by_name phyint_list_s.phyint_list_avl_by_name 1510 /* 1511 * ILL groups. We group ills, 1512 * 1513 * - if the ills have the same group name. (New way) 1514 * 1515 * ill_group locking notes: 1516 * 1517 * illgrp_lock protects ill_grp_ill_schednext. 1518 * 1519 * ill_g_lock protects ill_grp_next, illgrp_ill, illgrp_ill_count. 1520 * Holding ill_g_lock freezes the memberships of ills in IPMP groups. 1521 * It also freezes the global list of ills and all ipifs in all ills. 1522 * 1523 * To remove an ipif from the linked list of ipifs of that ill ipif_free_tail 1524 * holds both ill_g_lock, and ill_lock. Similarly to remove an ill from the 1525 * global list of ills, ill_delete_glist holds ill_g_lock as writer. 1526 * This simplifies things for ipif_select_source, illgrp_scheduler etc. 1527 * that need to walk the members of an illgrp. They just hold ill_g_lock 1528 * as reader to do the walk. 1529 * 1530 */ 1531 typedef struct ill_group { 1532 kmutex_t illgrp_lock; 1533 struct ill_group *illgrp_next; /* Next ill_group */ 1534 struct ill_s *illgrp_ill_schednext; /* Next ill to be scheduled */ 1535 struct ill_s *illgrp_ill; /* First ill in the group */ 1536 int illgrp_ill_count; 1537 } ill_group_t; 1538 1539 extern ill_group_t *illgrp_head_v6; 1540 1541 /* 1542 * Fragmentation hash bucket 1543 */ 1544 typedef struct ipfb_s { 1545 struct ipf_s *ipfb_ipf; /* List of ... */ 1546 size_t ipfb_count; /* Count of bytes used by frag(s) */ 1547 kmutex_t ipfb_lock; /* Protect all ipf in list */ 1548 uint_t ipfb_frag_pkts; /* num of distinct fragmented pkts */ 1549 } ipfb_t; 1550 1551 /* 1552 * IRE bucket structure. Usually there is an array of such structures, 1553 * each pointing to a linked list of ires. irb_refcnt counts the number 1554 * of walkers of a given hash bucket. Usually the reference count is 1555 * bumped up if the walker wants no IRES to be DELETED while walking the 1556 * list. Bumping up does not PREVENT ADDITION. This allows walking a given 1557 * hash bucket without stumbling up on a free pointer. 1558 * 1559 * irb_t structures in ip_ftable are dynamically allocated and freed. 1560 * In order to identify the irb_t structures that cna be safely kmem_free'd 1561 * we need to ensure that 1562 * - the irb_refcnt is quiescent, indicating no other walkers, 1563 * - no other threads or ire's are holding references to the irb, 1564 * i.e., irb_nire == 0, 1565 * - there are no active ire's in the bucket, i.e., irb_ire_cnt == 0 1566 */ 1567 typedef struct irb { 1568 struct ire_s *irb_ire; /* First ire in this bucket */ 1569 /* Should be first in this struct */ 1570 krwlock_t irb_lock; /* Protect this bucket */ 1571 uint_t irb_refcnt; /* Protected by irb_lock */ 1572 uchar_t irb_marks; /* CONDEMNED ires in this bucket ? */ 1573 #define IRB_MARK_CONDEMNED 0x0001 1574 #define IRB_MARK_FTABLE 0x0002 1575 uint_t irb_ire_cnt; /* Num of active IRE in this bucket */ 1576 uint_t irb_tmp_ire_cnt; /* Num of temporary IRE */ 1577 struct ire_s *irb_rr_origin; /* origin for round-robin */ 1578 int irb_nire; /* Num of ftable ire's that ref irb */ 1579 } irb_t; 1580 1581 #define IRB2RT(irb) (rt_t *)((caddr_t)(irb) - offsetof(rt_t, rt_irb)) 1582 1583 /* The following are return values of ip_xmit_v4() */ 1584 typedef enum { 1585 SEND_PASSED = 0, /* sent packet out on wire */ 1586 SEND_FAILED, /* sending of packet failed */ 1587 LOOKUP_IN_PROGRESS, /* ire cache found, ARP resolution in progress */ 1588 LLHDR_RESLV_FAILED /* macaddr resl of onlink dst or nexthop failed */ 1589 } ipxmit_state_t; 1590 1591 #define IP_V4_G_HEAD 0 1592 #define IP_V6_G_HEAD 1 1593 1594 #define MAX_G_HEADS 2 1595 1596 /* 1597 * unpadded ill_if structure 1598 */ 1599 struct _ill_if_s_ { 1600 union ill_if_u *illif_next; 1601 union ill_if_u *illif_prev; 1602 avl_tree_t illif_avl_by_ppa; /* AVL tree sorted on ppa */ 1603 vmem_t *illif_ppa_arena; /* ppa index space */ 1604 uint16_t illif_mcast_v1; /* hints for */ 1605 uint16_t illif_mcast_v2; /* [igmp|mld]_slowtimo */ 1606 int illif_name_len; /* name length */ 1607 char illif_name[LIFNAMSIZ]; /* name of interface type */ 1608 }; 1609 1610 /* cache aligned ill_if structure */ 1611 typedef union ill_if_u { 1612 struct _ill_if_s_ ill_if_s; 1613 char illif_filler[CACHE_ALIGN(_ill_if_s_)]; 1614 } ill_if_t; 1615 1616 1617 #define illif_next ill_if_s.illif_next 1618 #define illif_prev ill_if_s.illif_prev 1619 #define illif_avl_by_ppa ill_if_s.illif_avl_by_ppa 1620 #define illif_ppa_arena ill_if_s.illif_ppa_arena 1621 #define illif_mcast_v1 ill_if_s.illif_mcast_v1 1622 #define illif_mcast_v2 ill_if_s.illif_mcast_v2 1623 #define illif_name ill_if_s.illif_name 1624 #define illif_name_len ill_if_s.illif_name_len 1625 1626 typedef struct ill_walk_context_s { 1627 int ctx_current_list; /* current list being searched */ 1628 int ctx_last_list; /* last list to search */ 1629 } ill_walk_context_t; 1630 1631 /* 1632 * ill_gheads structure, one for IPV4 and one for IPV6 1633 */ 1634 struct _ill_g_head_s_ { 1635 ill_if_t *ill_g_list_head; 1636 ill_if_t *ill_g_list_tail; 1637 }; 1638 1639 typedef union ill_g_head_u { 1640 struct _ill_g_head_s_ ill_g_head_s; 1641 char ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)]; 1642 } ill_g_head_t; 1643 1644 #define ill_g_list_head ill_g_head_s.ill_g_list_head 1645 #define ill_g_list_tail ill_g_head_s.ill_g_list_tail 1646 1647 #pragma align CACHE_ALIGN_SIZE(ill_g_heads) 1648 extern ill_g_head_t ill_g_heads[]; /* ILL List Head */ 1649 1650 1651 #define IP_V4_ILL_G_LIST ill_g_heads[IP_V4_G_HEAD].ill_g_list_head 1652 #define IP_V6_ILL_G_LIST ill_g_heads[IP_V6_G_HEAD].ill_g_list_head 1653 #define IP_VX_ILL_G_LIST(i) ill_g_heads[i].ill_g_list_head 1654 1655 #define ILL_START_WALK_V4(ctx_ptr) ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, \ 1656 ctx_ptr) 1657 #define ILL_START_WALK_V6(ctx_ptr) ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, \ 1658 ctx_ptr) 1659 #define ILL_START_WALK_ALL(ctx_ptr) ill_first(MAX_G_HEADS, MAX_G_HEADS, \ 1660 ctx_ptr) 1661 1662 /* 1663 * Capabilities, possible flags for ill_capabilities. 1664 */ 1665 1666 #define ILL_CAPAB_AH 0x01 /* IPsec AH acceleration */ 1667 #define ILL_CAPAB_ESP 0x02 /* IPsec ESP acceleration */ 1668 #define ILL_CAPAB_MDT 0x04 /* Multidata Transmit */ 1669 #define ILL_CAPAB_HCKSUM 0x08 /* Hardware checksumming */ 1670 #define ILL_CAPAB_ZEROCOPY 0x10 /* Zero-copy */ 1671 #define ILL_CAPAB_POLL 0x20 /* Polling Toggle */ 1672 #define ILL_CAPAB_SOFT_RING 0x40 /* Soft_Ring capability */ 1673 1674 /* 1675 * Per-ill Multidata Transmit capabilities. 1676 */ 1677 typedef struct ill_mdt_capab_s ill_mdt_capab_t; 1678 1679 /* 1680 * Per-ill IPsec capabilities. 1681 */ 1682 typedef struct ill_ipsec_capab_s ill_ipsec_capab_t; 1683 1684 /* 1685 * Per-ill Hardware Checksumming capbilities. 1686 */ 1687 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t; 1688 1689 /* 1690 * Per-ill Zero-copy capabilities. 1691 */ 1692 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t; 1693 1694 /* 1695 * Per-ill Polling/soft ring capbilities. 1696 */ 1697 typedef struct ill_dls_capab_s ill_dls_capab_t; 1698 1699 /* 1700 * Per-ill polling resource map. 1701 */ 1702 typedef struct ill_rx_ring ill_rx_ring_t; 1703 1704 /* The following are ill_state_flags */ 1705 #define ILL_LL_SUBNET_PENDING 0x01 /* Waiting for DL_INFO_ACK from drv */ 1706 #define ILL_CONDEMNED 0x02 /* No more new ref's to the ILL */ 1707 #define ILL_CHANGING 0x04 /* ILL not globally visible */ 1708 #define ILL_DL_UNBIND_IN_PROGRESS 0x08 /* UNBIND_REQ is sent */ 1709 #define ILL_SOFT_RING_ASSIGN 0x10 /* Makeing soft ring assigment */ 1710 1711 /* Is this an ILL whose source address is used by other ILL's ? */ 1712 #define IS_USESRC_ILL(ill) \ 1713 (((ill)->ill_usesrc_ifindex == 0) && \ 1714 ((ill)->ill_usesrc_grp_next != NULL)) 1715 1716 /* Is this a client/consumer of the usesrc ILL ? */ 1717 #define IS_USESRC_CLI_ILL(ill) \ 1718 (((ill)->ill_usesrc_ifindex != 0) && \ 1719 ((ill)->ill_usesrc_grp_next != NULL)) 1720 1721 /* Is this an virtual network interface (vni) ILL ? */ 1722 #define IS_VNI(ill) \ 1723 (((ill) != NULL) && \ 1724 (((ill)->ill_phyint->phyint_flags & (PHYI_LOOPBACK|PHYI_VIRTUAL)) == \ 1725 PHYI_VIRTUAL)) 1726 1727 /* 1728 * IP Lower level Structure. 1729 * Instance data structure in ip_open when there is a device below us. 1730 */ 1731 typedef struct ill_s { 1732 ill_if_t *ill_ifptr; /* pointer to interface type */ 1733 queue_t *ill_rq; /* Read queue. */ 1734 queue_t *ill_wq; /* Write queue. */ 1735 1736 int ill_error; /* Error value sent up by device. */ 1737 1738 ipif_t *ill_ipif; /* Interface chain for this ILL. */ 1739 1740 uint_t ill_ipif_up_count; /* Number of IPIFs currently up. */ 1741 uint_t ill_max_frag; /* Max IDU from DLPI. */ 1742 char *ill_name; /* Our name. */ 1743 uint_t ill_name_length; /* Name length, incl. terminator. */ 1744 char *ill_ndd_name; /* Name + ":ip?_forwarding" for NDD. */ 1745 uint_t ill_net_type; /* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */ 1746 /* 1747 * Physical Point of Attachment num. If DLPI style 1 provider 1748 * then this is derived from the devname. 1749 */ 1750 uint_t ill_ppa; 1751 t_uscalar_t ill_sap; 1752 t_scalar_t ill_sap_length; /* Including sign (for position) */ 1753 uint_t ill_phys_addr_length; /* Excluding the sap. */ 1754 uint_t ill_bcast_addr_length; /* Only set when the DL provider */ 1755 /* supports broadcast. */ 1756 t_uscalar_t ill_mactype; 1757 uint8_t *ill_frag_ptr; /* Reassembly state. */ 1758 timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */ 1759 ipfb_t *ill_frag_hash_tbl; /* Fragment hash list head. */ 1760 ipif_t *ill_pending_ipif; /* IPIF waiting for DL operation. */ 1761 1762 ilm_t *ill_ilm; /* Multicast mebership for lower ill */ 1763 uint_t ill_global_timer; /* for IGMPv3/MLDv2 general queries */ 1764 int ill_mcast_type; /* type of router which is querier */ 1765 /* on this interface */ 1766 uint16_t ill_mcast_v1_time; /* # slow timeouts since last v1 qry */ 1767 uint16_t ill_mcast_v2_time; /* # slow timeouts since last v2 qry */ 1768 uint8_t ill_mcast_v1_tset; /* 1 => timer is set; 0 => not set */ 1769 uint8_t ill_mcast_v2_tset; /* 1 => timer is set; 0 => not set */ 1770 1771 uint8_t ill_mcast_rv; /* IGMPv3/MLDv2 robustness variable */ 1772 int ill_mcast_qi; /* IGMPv3/MLDv2 query interval var */ 1773 1774 mblk_t *ill_pending_mp; /* IOCTL/DLPI awaiting completion. */ 1775 /* 1776 * All non-NULL cells between 'ill_first_mp_to_free' and 1777 * 'ill_last_mp_to_free' are freed in ill_delete. 1778 */ 1779 #define ill_first_mp_to_free ill_bcast_mp 1780 mblk_t *ill_bcast_mp; /* DLPI header for broadcasts. */ 1781 mblk_t *ill_resolver_mp; /* Resolver template. */ 1782 mblk_t *ill_detach_mp; /* detach mp, or NULL if style1 */ 1783 mblk_t *ill_unbind_mp; /* unbind mp from ill_dl_up() */ 1784 mblk_t *ill_dlpi_deferred; /* b_next chain of control messages */ 1785 mblk_t *ill_phys_addr_mp; /* mblk which holds ill_phys_addr */ 1786 #define ill_last_mp_to_free ill_phys_addr_mp 1787 1788 cred_t *ill_credp; /* opener's credentials */ 1789 uint8_t *ill_phys_addr; /* ill_phys_addr_mp->b_rptr + off */ 1790 1791 uint_t ill_state_flags; /* see ILL_* flags above */ 1792 1793 /* Following bit fields protected by ipsq_t */ 1794 uint_t 1795 ill_needs_attach : 1, 1796 ill_reserved : 1, 1797 ill_isv6 : 1, 1798 ill_dlpi_style_set : 1, 1799 1800 ill_ifname_pending : 1, 1801 ill_move_in_progress : 1, /* FAILOVER/FAILBACK in progress */ 1802 ill_join_allmulti : 1, 1803 ill_logical_down : 1, 1804 1805 ill_is_6to4tun : 1, /* Interface is a 6to4 tunnel */ 1806 ill_promisc_on_phys : 1, /* phys interface in promisc mode */ 1807 ill_dl_up : 1, 1808 ill_up_ipifs : 1, 1809 1810 ill_pad_to_bit_31 : 20; 1811 1812 /* Following bit fields protected by ill_lock */ 1813 uint_t 1814 ill_fragtimer_executing : 1, 1815 ill_fragtimer_needrestart : 1, 1816 ill_ilm_cleanup_reqd : 1, 1817 ill_arp_closing : 1, 1818 1819 ill_arp_bringup_pending : 1, 1820 ill_mtu_userspecified : 1, /* SIOCSLNKINFO has set the mtu */ 1821 ill_pad_bit_31 : 26; 1822 1823 /* 1824 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'. 1825 */ 1826 int ill_arp_muxid; /* muxid returned from plink for arp */ 1827 int ill_ip_muxid; /* muxid returned from plink for ip */ 1828 1829 /* 1830 * Used for IP frag reassembly throttling on a per ILL basis. 1831 * 1832 * Note: frag_count is approximate, its added to and subtracted from 1833 * without any locking, so simultaneous load/modify/stores can 1834 * collide, also ill_frag_purge() recalculates its value by 1835 * summing all the ipfb_count's without locking out updates 1836 * to the ipfb's. 1837 */ 1838 uint_t ill_ipf_gen; /* Generation of next fragment queue */ 1839 uint_t ill_frag_count; /* Approx count of all mblk bytes */ 1840 uint_t ill_frag_free_num_pkts; /* num of fragmented packets to free */ 1841 clock_t ill_last_frag_clean_time; /* time when frag's were pruned */ 1842 int ill_type; /* From <net/if_types.h> */ 1843 uint_t ill_dlpi_multicast_state; /* See below IDMS_* */ 1844 uint_t ill_dlpi_fastpath_state; /* See below IDMS_* */ 1845 1846 /* 1847 * Capabilities related fields. 1848 */ 1849 uint_t ill_capab_state; /* State of capability query, IDMS_* */ 1850 uint64_t ill_capabilities; /* Enabled capabilities, ILL_CAPAB_* */ 1851 ill_mdt_capab_t *ill_mdt_capab; /* Multidata Transmit capabilities */ 1852 ill_ipsec_capab_t *ill_ipsec_capab_ah; /* IPsec AH capabilities */ 1853 ill_ipsec_capab_t *ill_ipsec_capab_esp; /* IPsec ESP capabilities */ 1854 ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */ 1855 ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */ 1856 ill_dls_capab_t *ill_dls_capab; /* Polling, soft ring capabilities */ 1857 1858 /* 1859 * New fields for IPv6 1860 */ 1861 uint8_t ill_max_hops; /* Maximum hops for any logical interface */ 1862 uint_t ill_max_mtu; /* Maximum MTU for any logical interface */ 1863 uint32_t ill_reachable_time; /* Value for ND algorithm in msec */ 1864 uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */ 1865 uint_t ill_max_buf; /* Max # of req to buffer for ND */ 1866 in6_addr_t ill_token; 1867 uint_t ill_token_length; 1868 uint32_t ill_xmit_count; /* ndp max multicast xmits */ 1869 mib2_ipv6IfStatsEntry_t *ill_ip6_mib; /* Per interface mib */ 1870 mib2_ipv6IfIcmpEntry_t *ill_icmp6_mib; /* Per interface mib */ 1871 /* 1872 * Following two mblks are allocated common to all 1873 * the ipifs when the first interface is coming up. 1874 * It is sent up to arp when the last ipif is coming 1875 * down. 1876 */ 1877 mblk_t *ill_arp_down_mp; 1878 mblk_t *ill_arp_del_mapping_mp; 1879 /* 1880 * Used for implementing IFF_NOARP. As IFF_NOARP is used 1881 * to turn off for all the logicals, it is here instead 1882 * of the ipif. 1883 */ 1884 mblk_t *ill_arp_on_mp; 1885 /* Peer ill of an IPMP move operation */ 1886 struct ill_s *ill_move_peer; 1887 1888 phyint_t *ill_phyint; 1889 uint64_t ill_flags; 1890 ill_group_t *ill_group; 1891 struct ill_s *ill_group_next; 1892 /* 1893 * Reverse tunnel related count. This count 1894 * determines how many mobile nodes are using this 1895 * ill to send packet to reverse tunnel via foreign 1896 * agent. A non-zero count specifies presence of 1897 * mobile node(s) using reverse tunnel through this 1898 * interface. 1899 */ 1900 uint32_t ill_mrtun_refcnt; 1901 1902 /* 1903 * This count is bumped up when a route is added with 1904 * RTA_SRCIFP bit flag using routing socket. 1905 */ 1906 uint32_t ill_srcif_refcnt; 1907 /* 1908 * Pointer to the special interface based routing table. 1909 * This routing table is created dynamically when RTA_SRCIFP 1910 * is set by the routing socket. 1911 */ 1912 irb_t *ill_srcif_table; 1913 kmutex_t ill_lock; /* Please see table below */ 1914 /* 1915 * The ill_nd_lla* fields handle the link layer address option 1916 * from neighbor discovery. This is used for external IPv6 1917 * address resolution. 1918 */ 1919 mblk_t *ill_nd_lla_mp; /* mblk which holds ill_nd_lla */ 1920 uint8_t *ill_nd_lla; /* Link Layer Address */ 1921 uint_t ill_nd_lla_len; /* Link Layer Address length */ 1922 /* 1923 * We now have 3 phys_addr_req's sent down. This field keeps track 1924 * of which one is pending. 1925 */ 1926 t_uscalar_t ill_phys_addr_pend; /* which dl_phys_addr_req pending */ 1927 /* 1928 * Used to save errors that occur during plumbing 1929 */ 1930 uint_t ill_ifname_pending_err; 1931 avl_node_t ill_avl_byppa; /* avl node based on ppa */ 1932 void *ill_fastpath_list; /* both ire and nce hang off this */ 1933 uint_t ill_refcnt; /* active refcnt by threads */ 1934 uint_t ill_ire_cnt; /* ires associated with this ill */ 1935 kcondvar_t ill_cv; 1936 uint_t ill_ilm_walker_cnt; /* snmp ilm walkers */ 1937 uint_t ill_nce_cnt; /* nces associated with this ill */ 1938 uint_t ill_waiters; /* threads waiting in ipsq_enter */ 1939 /* 1940 * Contains the upper read queue pointer of the module immediately 1941 * beneath IP. This field allows IP to validate sub-capability 1942 * acknowledgments coming up from downstream. 1943 */ 1944 queue_t *ill_lmod_rq; /* read queue pointer of module below */ 1945 uint_t ill_lmod_cnt; /* number of modules beneath IP */ 1946 ip_m_t *ill_media; /* media specific params/functions */ 1947 t_uscalar_t ill_dlpi_pending; /* Last DLPI primitive issued */ 1948 uint_t ill_usesrc_ifindex; /* use src addr from this ILL */ 1949 struct ill_s *ill_usesrc_grp_next; /* Next ILL in the usesrc group */ 1950 #ifdef ILL_DEBUG 1951 th_trace_t *ill_trace[IP_TR_HASH_MAX]; 1952 boolean_t ill_trace_disable; /* True when alloc fails */ 1953 #endif 1954 } ill_t; 1955 1956 extern void ill_delete_glist(ill_t *); 1957 1958 /* 1959 * The following table lists the protection levels of the various members 1960 * of the ill_t. Same notation as that used for ipif_t above is used. 1961 * 1962 * Write Read 1963 * 1964 * ill_ifptr ill_g_lock + s Write once 1965 * ill_rq ipsq Write once 1966 * ill_wq ipsq Write once 1967 * 1968 * ill_error ipsq None 1969 * ill_ipif ill_g_lock + ipsq ill_g_lock OR ipsq 1970 * ill_ipif_up_count ill_lock + ipsq ill_lock 1971 * ill_max_frag ipsq Write once 1972 * 1973 * ill_name ill_g_lock + ipsq Write once 1974 * ill_name_length ill_g_lock + ipsq Write once 1975 * ill_ndd_name ipsq Write once 1976 * ill_net_type ipsq Write once 1977 * ill_ppa ill_g_lock + ipsq Write once 1978 * ill_sap ipsq + down ill Write once 1979 * ill_sap_length ipsq + down ill Write once 1980 * ill_phys_addr_length ipsq + down ill Write once 1981 * 1982 * ill_bcast_addr_length ipsq ipsq 1983 * ill_mactype ipsq ipsq 1984 * ill_frag_ptr ipsq ipsq 1985 * 1986 * ill_frag_timer_id ill_lock ill_lock 1987 * ill_frag_hash_tbl ipsq up ill 1988 * ill_ilm ipsq + ill_lock ill_lock 1989 * ill_mcast_type ill_lock ill_lock 1990 * ill_mcast_v1_time ill_lock ill_lock 1991 * ill_mcast_v2_time ill_lock ill_lock 1992 * ill_mcast_v1_tset ill_lock ill_lock 1993 * ill_mcast_v2_tset ill_lock ill_lock 1994 * ill_mcast_rv ill_lock ill_lock 1995 * ill_mcast_qi ill_lock ill_lock 1996 * ill_pending_mp ill_lock ill_lock 1997 * 1998 * ill_bcast_mp ipsq ipsq 1999 * ill_resolver_mp ipsq only when ill is up 2000 * ill_down_mp ipsq ipsq 2001 * ill_dlpi_deferred ipsq ipsq 2002 * ill_phys_addr_mp ipsq ipsq 2003 * ill_phys_addr ipsq up ill 2004 * ill_ick ipsq + down ill only when ill is up 2005 * 2006 * ill_state_flags ill_lock ill_lock 2007 * exclusive bit flags ipsq_t ipsq_t 2008 * shared bit flags ill_lock ill_lock 2009 * 2010 * ill_arp_muxid ipsq Not atomic 2011 * ill_ip_muxid ipsq Not atomic 2012 * 2013 * ill_ipf_gen Not atomic 2014 * ill_frag_count Approx. not protected 2015 * ill_type ipsq + down ill only when ill is up 2016 * ill_dlpi_multicast_state ill_lock ill_lock 2017 * ill_dlpi_fastpath_state ill_lock ill_lock 2018 * ill_max_hops ipsq Not atomic 2019 * 2020 * ill_max_mtu 2021 * 2022 * ill_reachable_time ipsq + ill_lock ill_lock 2023 * ill_reachable_retrans_time ipsq + ill_lock ill_lock 2024 * ill_max_buf ipsq + ill_lock ill_lock 2025 * 2026 * Next 2 fields need ill_lock because of the get ioctls. They should not 2027 * report partially updated results without executing in the ipsq. 2028 * ill_token ipsq + ill_lock ill_lock 2029 * ill_token_length ipsq + ill_lock ill_lock 2030 * ill_xmit_count ipsq + down ill write once 2031 * ill_ip6_mib ipsq + down ill only when ill is up 2032 * ill_icmp6_mib ipsq + down ill only when ill is up 2033 * ill_arp_down_mp ipsq ipsq 2034 * ill_arp_del_mapping_mp ipsq ipsq 2035 * ill_arp_on_mp ipsq ipsq 2036 * ill_move_peer ipsq ipsq 2037 * 2038 * ill_phyint ipsq, ill_g_lock, ill_lock Any of them 2039 * ill_flags ill_lock ill_lock 2040 * ill_group ipsq, ill_g_lock, ill_lock Any of them 2041 * ill_group_next ipsq, ill_g_lock, ill_lock Any of them 2042 * ill_mrtun_refcnt ill_lock ill_lock 2043 * ill_srcif_refcnt ill_lock ill_lock 2044 * ill_srcif_table ill_lock ill_lock 2045 * ill_nd_lla_mp ill_lock ill_lock 2046 * ill_nd_lla ill_lock ill_lock 2047 * ill_nd_lla_len ill_lock ill_lock 2048 * ill_phys_addr_pend ipsq + down ill only when ill is up 2049 * ill_ifname_pending_err ipsq ipsq 2050 * ill_avl_byppa ipsq, ill_g_lock Write once 2051 * 2052 * ill_fastpath_list ill_lock ill_lock 2053 * ill_refcnt ill_lock ill_lock 2054 * ill_ire_cnt ill_lock ill_lock 2055 * ill_cv ill_lock ill_lock 2056 * ill_ilm_walker_cnt ill_lock ill_lock 2057 * ill_nce_cnt ill_lock ill_lock 2058 * ill_trace ill_lock ill_lock 2059 * ill_usesrc_grp_next ill_g_usesrc_lock ill_g_usesrc_lock 2060 */ 2061 2062 /* 2063 * For ioctl restart mechanism see ip_reprocess_ioctl() 2064 */ 2065 struct ip_ioctl_cmd_s; 2066 2067 typedef int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *, 2068 struct ip_ioctl_cmd_s *, void *); 2069 2070 typedef struct ip_ioctl_cmd_s { 2071 int ipi_cmd; 2072 size_t ipi_copyin_size; 2073 uint_t ipi_flags; 2074 uint_t ipi_cmd_type; 2075 ifunc_t ipi_func; 2076 ifunc_t ipi_func_restart; 2077 } ip_ioctl_cmd_t; 2078 2079 /* 2080 * ipi_cmd_type: 2081 * 2082 * IF_CMD 1 old style ifreq cmd 2083 * LIF_CMD 2 new style lifreq cmd 2084 * MISC_CMD 3 Misc. (non [l]ifreq, tun) cmds 2085 * TUN_CMD 4 tunnel related 2086 */ 2087 2088 enum { IF_CMD = 1, LIF_CMD, MISC_CMD, TUN_CMD }; 2089 2090 #define IPI_DONTCARE 0 /* For ioctl encoded values that don't matter */ 2091 2092 /* Flag values in ipi_flags */ 2093 #define IPI_PRIV 0x1 /* Root only command */ 2094 #define IPI_MODOK 0x2 /* Permitted on mod instance of IP */ 2095 #define IPI_WR 0x4 /* Need to grab writer access */ 2096 #define IPI_GET_CMD 0x8 /* branch to mi_copyout on success */ 2097 #define IPI_REPL 0x10 /* valid for replacement ipif created in MOVE */ 2098 #define IPI_NULL_BCONT 0x20 /* ioctl has not data and hence no b_cont */ 2099 #define IPI_PASS_DOWN 0x40 /* pass this ioctl down when a module only */ 2100 2101 extern ip_ioctl_cmd_t ip_ndx_ioctl_table[]; 2102 extern ip_ioctl_cmd_t ip_misc_ioctl_table[]; 2103 extern int ip_ndx_ioctl_count; 2104 extern int ip_misc_ioctl_count; 2105 2106 #define ILL_CLEAR_MOVE(ill) { \ 2107 ill_t *peer_ill; \ 2108 \ 2109 peer_ill = (ill)->ill_move_peer; \ 2110 ASSERT(peer_ill != NULL); \ 2111 (ill)->ill_move_in_progress = B_FALSE; \ 2112 peer_ill->ill_move_in_progress = B_FALSE; \ 2113 (ill)->ill_move_peer = NULL; \ 2114 peer_ill->ill_move_peer = NULL; \ 2115 } 2116 /* The 2 ill's are same or belong to the same IPMP group */ 2117 #define SAME_IPMP_GROUP(ill_1, ill_2) \ 2118 (((ill_1)->ill_group != NULL) && \ 2119 ((ill_1)->ill_group == (ill_2)->ill_group)) 2120 2121 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */ 2122 typedef struct ipmx_s { 2123 char ipmx_name[LIFNAMSIZ]; /* if name */ 2124 uint_t 2125 ipmx_arpdev_stream : 1, /* This is the arp stream */ 2126 ipmx_notused : 31; 2127 } ipmx_t; 2128 2129 /* 2130 * State for detecting if a driver supports certain features. 2131 * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state. 2132 * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state. 2133 */ 2134 #define IDMS_UNKNOWN 0 /* No DL_ENABMULTI_REQ sent */ 2135 #define IDMS_INPROGRESS 1 /* Sent DL_ENABMULTI_REQ */ 2136 #define IDMS_OK 2 /* DL_ENABMULTI_REQ ok */ 2137 #define IDMS_FAILED 3 /* DL_ENABMULTI_REQ failed */ 2138 #define IDMS_RENEG 4 /* Driver asked for a renegotiation */ 2139 2140 /* Named Dispatch Parameter Management Structure */ 2141 typedef struct ipparam_s { 2142 uint_t ip_param_min; 2143 uint_t ip_param_max; 2144 uint_t ip_param_value; 2145 char *ip_param_name; 2146 } ipparam_t; 2147 2148 /* Extended NDP Management Structure */ 2149 typedef struct ipndp_s { 2150 ndgetf_t ip_ndp_getf; 2151 ndsetf_t ip_ndp_setf; 2152 caddr_t ip_ndp_data; 2153 char *ip_ndp_name; 2154 } ipndp_t; 2155 2156 /* 2157 * The kernel stores security attributes of all gateways in a database made 2158 * up of one or more tsol_gcdb_t elements. Each tsol_gcdb_t contains the 2159 * security-related credentials of the gateway. More than one gateways may 2160 * share entries in the database. 2161 * 2162 * The tsol_gc_t structure represents the gateway to credential association, 2163 * and refers to an entry in the database. One or more tsol_gc_t entities are 2164 * grouped together to form one or more tsol_gcgrp_t, each representing the 2165 * list of security attributes specific to the gateway. A gateway may be 2166 * associated with at most one credentials group. 2167 */ 2168 struct tsol_gcgrp_s; 2169 2170 extern uchar_t ip6opt_ls; /* TX IPv6 enabler */ 2171 2172 /* 2173 * Gateway security credential record. 2174 */ 2175 typedef struct tsol_gcdb_s { 2176 uint_t gcdb_refcnt; /* reference count */ 2177 struct rtsa_s gcdb_attr; /* security attributes */ 2178 #define gcdb_mask gcdb_attr.rtsa_mask 2179 #define gcdb_doi gcdb_attr.rtsa_doi 2180 #define gcdb_slrange gcdb_attr.rtsa_slrange 2181 } tsol_gcdb_t; 2182 2183 /* 2184 * Gateway to credential association. 2185 */ 2186 typedef struct tsol_gc_s { 2187 uint_t gc_refcnt; /* reference count */ 2188 struct tsol_gcgrp_s *gc_grp; /* pointer to group */ 2189 struct tsol_gc_s *gc_prev; /* previous in list */ 2190 struct tsol_gc_s *gc_next; /* next in list */ 2191 tsol_gcdb_t *gc_db; /* pointer to actual credentials */ 2192 } tsol_gc_t; 2193 2194 /* 2195 * Gateway credentials group address. 2196 */ 2197 typedef struct tsol_gcgrp_addr_s { 2198 int ga_af; /* address family */ 2199 in6_addr_t ga_addr; /* IPv4 mapped or IPv6 address */ 2200 } tsol_gcgrp_addr_t; 2201 2202 /* 2203 * Gateway credentials group. 2204 */ 2205 typedef struct tsol_gcgrp_s { 2206 uint_t gcgrp_refcnt; /* reference count */ 2207 krwlock_t gcgrp_rwlock; /* lock to protect following */ 2208 uint_t gcgrp_count; /* number of credentials */ 2209 tsol_gc_t *gcgrp_head; /* first credential in list */ 2210 tsol_gc_t *gcgrp_tail; /* last credential in list */ 2211 tsol_gcgrp_addr_t gcgrp_addr; /* next-hop gateway address */ 2212 } tsol_gcgrp_t; 2213 2214 extern kmutex_t gcgrp_lock; 2215 2216 #define GC_REFRELE(p) { \ 2217 ASSERT((p)->gc_grp != NULL); \ 2218 rw_enter(&(p)->gc_grp->gcgrp_rwlock, RW_WRITER); \ 2219 ASSERT((p)->gc_refcnt > 0); \ 2220 if (--((p)->gc_refcnt) == 0) \ 2221 gc_inactive(p); \ 2222 else \ 2223 rw_exit(&(p)->gc_grp->gcgrp_rwlock); \ 2224 } 2225 2226 #define GCGRP_REFHOLD(p) { \ 2227 mutex_enter(&gcgrp_lock); \ 2228 ++((p)->gcgrp_refcnt); \ 2229 ASSERT((p)->gcgrp_refcnt != 0); \ 2230 mutex_exit(&gcgrp_lock); \ 2231 } 2232 2233 #define GCGRP_REFRELE(p) { \ 2234 mutex_enter(&gcgrp_lock); \ 2235 ASSERT((p)->gcgrp_refcnt > 0); \ 2236 if (--((p)->gcgrp_refcnt) == 0) \ 2237 gcgrp_inactive(p); \ 2238 ASSERT(MUTEX_HELD(&gcgrp_lock)); \ 2239 mutex_exit(&gcgrp_lock); \ 2240 } 2241 2242 /* 2243 * IRE gateway security attributes structure, pointed to by tsol_ire_gw_secattr 2244 */ 2245 struct tsol_tnrhc; 2246 2247 typedef struct tsol_ire_gw_secattr_s { 2248 kmutex_t igsa_lock; /* lock to protect following */ 2249 struct tsol_tnrhc *igsa_rhc; /* host entry for gateway */ 2250 tsol_gc_t *igsa_gc; /* for prefix IREs */ 2251 tsol_gcgrp_t *igsa_gcgrp; /* for cache IREs */ 2252 } tsol_ire_gw_secattr_t; 2253 2254 /* 2255 * Following are the macros to increment/decrement the reference 2256 * count of the IREs and IRBs (ire bucket). 2257 * 2258 * 1) We bump up the reference count of an IRE to make sure that 2259 * it does not get deleted and freed while we are using it. 2260 * Typically all the lookup functions hold the bucket lock, 2261 * and look for the IRE. If it finds an IRE, it bumps up the 2262 * reference count before dropping the lock. Sometimes we *may* want 2263 * to bump up the reference count after we *looked* up i.e without 2264 * holding the bucket lock. So, the IRE_REFHOLD macro does not assert 2265 * on the bucket lock being held. Any thread trying to delete from 2266 * the hash bucket can still do so but cannot free the IRE if 2267 * ire_refcnt is not 0. 2268 * 2269 * 2) We bump up the reference count on the bucket where the IRE resides 2270 * (IRB), when we want to prevent the IREs getting deleted from a given 2271 * hash bucket. This makes life easier for ire_walk type functions which 2272 * wants to walk the IRE list, call a function, but needs to drop 2273 * the bucket lock to prevent recursive rw_enters. While the 2274 * lock is dropped, the list could be changed by other threads or 2275 * the same thread could end up deleting the ire or the ire pointed by 2276 * ire_next. IRE_REFHOLDing the ire or ire_next is not sufficient as 2277 * a delete will still remove the ire from the bucket while we have 2278 * dropped the lock and hence the ire_next would be NULL. Thus, we 2279 * need a mechanism to prevent deletions from a given bucket. 2280 * 2281 * To prevent deletions, we bump up the reference count on the 2282 * bucket. If the bucket is held, ire_delete just marks IRE_MARK_CONDEMNED 2283 * both on the ire's ire_marks and the bucket's irb_marks. When the 2284 * reference count on the bucket drops to zero, all the CONDEMNED ires 2285 * are deleted. We don't have to bump up the reference count on the 2286 * bucket if we are walking the bucket and never have to drop the bucket 2287 * lock. Note that IRB_REFHOLD does not prevent addition of new ires 2288 * in the list. It is okay because addition of new ires will not cause 2289 * ire_next to point to freed memory. We do IRB_REFHOLD only when 2290 * all of the 3 conditions are true : 2291 * 2292 * 1) The code needs to walk the IRE bucket from start to end. 2293 * 2) It may have to drop the bucket lock sometimes while doing (1) 2294 * 3) It does not want any ires to be deleted meanwhile. 2295 */ 2296 2297 /* 2298 * Bump up the reference count on the IRE. We cannot assert that the 2299 * bucket lock is being held as it is legal to bump up the reference 2300 * count after the first lookup has returned the IRE without 2301 * holding the lock. Currently ip_wput does this for caching IRE_CACHEs. 2302 */ 2303 2304 #ifndef IRE_DEBUG 2305 2306 #define IRE_REFHOLD_NOTR(ire) IRE_REFHOLD(ire) 2307 #define IRE_UNTRACE_REF(ire) 2308 #define IRE_TRACE_REF(ire) 2309 2310 #else 2311 2312 #define IRE_REFHOLD_NOTR(ire) { \ 2313 atomic_add_32(&(ire)->ire_refcnt, 1); \ 2314 ASSERT((ire)->ire_refcnt != 0); \ 2315 } 2316 2317 #define IRE_UNTRACE_REF(ire) ire_untrace_ref(ire); 2318 #define IRE_TRACE_REF(ire) ire_trace_ref(ire); 2319 #endif 2320 2321 #define IRE_REFHOLD(ire) { \ 2322 atomic_add_32(&(ire)->ire_refcnt, 1); \ 2323 ASSERT((ire)->ire_refcnt != 0); \ 2324 IRE_TRACE_REF(ire); \ 2325 } 2326 2327 #define IRE_REFHOLD_LOCKED(ire) { \ 2328 IRE_TRACE_REF(ire); \ 2329 (ire)->ire_refcnt++; \ 2330 } 2331 2332 /* 2333 * Decrement the reference count on the IRE. 2334 * In architectures e.g sun4u, where atomic_add_32_nv is just 2335 * a cas, we need to maintain the right memory barrier semantics 2336 * as that of mutex_exit i.e all the loads and stores should complete 2337 * before the cas is executed. membar_exit() does that here. 2338 * 2339 * NOTE : This macro is used only in places where we want performance. 2340 * To avoid bloating the code, we use the function "ire_refrele" 2341 * which essentially calls the macro. 2342 */ 2343 #ifndef IRE_DEBUG 2344 #define IRE_REFRELE(ire) { \ 2345 ASSERT((ire)->ire_refcnt != 0); \ 2346 membar_exit(); \ 2347 if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0) \ 2348 ire_inactive(ire); \ 2349 } 2350 #define IRE_REFRELE_NOTR(ire) IRE_REFRELE(ire) 2351 #else 2352 #define IRE_REFRELE(ire) { \ 2353 if (ire->ire_bucket != NULL) \ 2354 ire_untrace_ref(ire); \ 2355 ASSERT((ire)->ire_refcnt != 0); \ 2356 membar_exit(); \ 2357 if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0) \ 2358 ire_inactive(ire); \ 2359 } 2360 #define IRE_REFRELE_NOTR(ire) { \ 2361 ASSERT((ire)->ire_refcnt != 0); \ 2362 membar_exit(); \ 2363 if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0) \ 2364 ire_inactive(ire); \ 2365 } 2366 #endif 2367 2368 /* 2369 * Bump up the reference count on the hash bucket - IRB to 2370 * prevent ires from being deleted in this bucket. 2371 */ 2372 #define IRB_REFHOLD(irb) { \ 2373 rw_enter(&(irb)->irb_lock, RW_WRITER); \ 2374 (irb)->irb_refcnt++; \ 2375 ASSERT((irb)->irb_refcnt != 0); \ 2376 rw_exit(&(irb)->irb_lock); \ 2377 } 2378 #define IRB_REFHOLD_LOCKED(irb) { \ 2379 ASSERT(RW_WRITE_HELD(&(irb)->irb_lock)); \ 2380 (irb)->irb_refcnt++; \ 2381 ASSERT((irb)->irb_refcnt != 0); \ 2382 } 2383 2384 void irb_refrele_ftable(irb_t *); 2385 /* 2386 * Note: when IRB_MARK_FTABLE (i.e., IRE_CACHETABLE entry), the irb_t 2387 * is statically allocated, so that when the irb_refcnt goes to 0, 2388 * we simply clean up the ire list and continue. 2389 */ 2390 #define IRB_REFRELE(irb) { \ 2391 if ((irb)->irb_marks & IRB_MARK_FTABLE) { \ 2392 irb_refrele_ftable((irb)); \ 2393 } else { \ 2394 rw_enter(&(irb)->irb_lock, RW_WRITER); \ 2395 ASSERT((irb)->irb_refcnt != 0); \ 2396 if (--(irb)->irb_refcnt == 0 && \ 2397 ((irb)->irb_marks & IRE_MARK_CONDEMNED)) { \ 2398 ire_t *ire_list; \ 2399 \ 2400 ire_list = ire_unlink(irb); \ 2401 rw_exit(&(irb)->irb_lock); \ 2402 ASSERT(ire_list != NULL); \ 2403 ire_cleanup(ire_list); \ 2404 } else { \ 2405 rw_exit(&(irb)->irb_lock); \ 2406 } \ 2407 } \ 2408 } 2409 2410 extern struct kmem_cache *rt_entry_cache; 2411 2412 /* 2413 * Lock the fast path mp for access, since the fp_mp can be deleted 2414 * due a DL_NOTE_FASTPATH_FLUSH in the case of IRE_BROADCAST and IRE_MIPRTUN 2415 */ 2416 2417 #define LOCK_IRE_FP_MP(ire) { \ 2418 if ((ire)->ire_type == IRE_BROADCAST || \ 2419 (ire)->ire_type == IRE_MIPRTUN) \ 2420 mutex_enter(&ire->ire_nce->nce_lock); \ 2421 } 2422 #define UNLOCK_IRE_FP_MP(ire) { \ 2423 if ((ire)->ire_type == IRE_BROADCAST || \ 2424 (ire)->ire_type == IRE_MIPRTUN) \ 2425 mutex_exit(&ire->ire_nce->nce_lock); \ 2426 } 2427 2428 typedef struct ire4 { 2429 ipaddr_t ire4_src_addr; /* Source address to use. */ 2430 ipaddr_t ire4_mask; /* Mask for matching this IRE. */ 2431 ipaddr_t ire4_addr; /* Address this IRE represents. */ 2432 ipaddr_t ire4_gateway_addr; /* Gateway if IRE_CACHE/IRE_OFFSUBNET */ 2433 ipaddr_t ire4_cmask; /* Mask from parent prefix route */ 2434 } ire4_t; 2435 2436 typedef struct ire6 { 2437 in6_addr_t ire6_src_addr; /* Source address to use. */ 2438 in6_addr_t ire6_mask; /* Mask for matching this IRE. */ 2439 in6_addr_t ire6_addr; /* Address this IRE represents. */ 2440 in6_addr_t ire6_gateway_addr; /* Gateway if IRE_CACHE/IRE_OFFSUBNET */ 2441 in6_addr_t ire6_cmask; /* Mask from parent prefix route */ 2442 } ire6_t; 2443 2444 typedef union ire_addr { 2445 ire6_t ire6_u; 2446 ire4_t ire4_u; 2447 } ire_addr_u_t; 2448 2449 /* Internet Routing Entry */ 2450 typedef struct ire_s { 2451 struct ire_s *ire_next; /* The hash chain must be first. */ 2452 struct ire_s **ire_ptpn; /* Pointer to previous next. */ 2453 uint32_t ire_refcnt; /* Number of references */ 2454 mblk_t *ire_mp; /* Non-null if allocated as mblk */ 2455 queue_t *ire_rfq; /* recv from this queue */ 2456 queue_t *ire_stq; /* send to this queue */ 2457 union { 2458 uint_t *max_fragp; /* Used only during ire creation */ 2459 uint_t max_frag; /* MTU (next hop or path). */ 2460 } imf_u; 2461 #define ire_max_frag imf_u.max_frag 2462 #define ire_max_fragp imf_u.max_fragp 2463 uint32_t ire_frag_flag; /* IPH_DF or zero. */ 2464 uint32_t ire_ident; /* Per IRE IP ident. */ 2465 uint32_t ire_tire_mark; /* Used for reclaim of unused. */ 2466 uchar_t ire_ipversion; /* IPv4/IPv6 version */ 2467 uchar_t ire_marks; /* IRE_MARK_CONDEMNED etc. */ 2468 ushort_t ire_type; /* Type of IRE */ 2469 uint_t ire_ib_pkt_count; /* Inbound packets for ire_addr */ 2470 uint_t ire_ob_pkt_count; /* Outbound packets to ire_addr */ 2471 uint_t ire_ll_hdr_length; /* Non-zero if we do M_DATA prepends */ 2472 time_t ire_create_time; /* Time (in secs) IRE was created. */ 2473 uint32_t ire_phandle; /* Associate prefix IREs to cache */ 2474 uint32_t ire_ihandle; /* Associate interface IREs to cache */ 2475 ipif_t *ire_ipif; /* the interface that this ire uses */ 2476 uint32_t ire_flags; /* flags related to route (RTF_*) */ 2477 uint_t ire_ipsec_overhead; /* IPSEC overhead */ 2478 /* 2479 * Neighbor Cache Entry for IPv6; arp info for IPv4 2480 */ 2481 struct nce_s *ire_nce; 2482 uint_t ire_masklen; /* # bits in ire_mask{,_v6} */ 2483 ire_addr_u_t ire_u; /* IPv4/IPv6 address info. */ 2484 2485 irb_t *ire_bucket; /* Hash bucket when ire_ptphn is set */ 2486 iulp_t ire_uinfo; /* Upper layer protocol info. */ 2487 /* 2488 * Protects ire_uinfo, ire_max_frag, and ire_frag_flag. 2489 */ 2490 kmutex_t ire_lock; 2491 uint_t ire_ipif_seqid; /* ipif_seqid of ire_ipif */ 2492 /* 2493 * For regular routes in forwarding table and cache table the 2494 * the following entries are NULL/zero. Only reverse tunnel 2495 * table and interface based forwarding table use these fields. 2496 * Routes added with RTA_SRCIFP and RTA_SRC respectively have 2497 * non-zero values for the following fields. 2498 */ 2499 ill_t *ire_in_ill; /* Incoming ill interface */ 2500 ipaddr_t ire_in_src_addr; 2501 /* source ip-addr of incoming packet */ 2502 clock_t ire_last_used_time; /* Last used time */ 2503 struct ire_s *ire_fastpath; /* Pointer to next ire in fastpath */ 2504 zoneid_t ire_zoneid; /* for local address discrimination */ 2505 tsol_ire_gw_secattr_t *ire_gw_secattr; /* gateway security attributes */ 2506 #ifdef IRE_DEBUG 2507 th_trace_t *ire_trace[IP_TR_HASH_MAX]; 2508 boolean_t ire_trace_disable; /* True when alloc fails */ 2509 #endif 2510 /* 2511 * ire's that are embedded inside mblk_t and sent to the external 2512 * resolver use the ire_stq_ifindex to track the ifindex of the 2513 * ire_stq, so that the ill (if it exists) can be correctly recovered 2514 * for cleanup in the esbfree routine when arp failure occurs 2515 */ 2516 uint_t ire_stq_ifindex; 2517 } ire_t; 2518 2519 /* IPv4 compatiblity macros */ 2520 #define ire_src_addr ire_u.ire4_u.ire4_src_addr 2521 #define ire_mask ire_u.ire4_u.ire4_mask 2522 #define ire_addr ire_u.ire4_u.ire4_addr 2523 #define ire_gateway_addr ire_u.ire4_u.ire4_gateway_addr 2524 #define ire_cmask ire_u.ire4_u.ire4_cmask 2525 2526 #define ire_src_addr_v6 ire_u.ire6_u.ire6_src_addr 2527 #define ire_mask_v6 ire_u.ire6_u.ire6_mask 2528 #define ire_addr_v6 ire_u.ire6_u.ire6_addr 2529 #define ire_gateway_addr_v6 ire_u.ire6_u.ire6_gateway_addr 2530 #define ire_cmask_v6 ire_u.ire6_u.ire6_cmask 2531 2532 /* Convenient typedefs for sockaddrs */ 2533 typedef struct sockaddr_in sin_t; 2534 typedef struct sockaddr_in6 sin6_t; 2535 2536 /* Address structure used for internal bind with IP */ 2537 typedef struct ipa_conn_s { 2538 ipaddr_t ac_laddr; 2539 ipaddr_t ac_faddr; 2540 uint16_t ac_fport; 2541 uint16_t ac_lport; 2542 } ipa_conn_t; 2543 2544 typedef struct ipa6_conn_s { 2545 in6_addr_t ac6_laddr; 2546 in6_addr_t ac6_faddr; 2547 uint16_t ac6_fport; 2548 uint16_t ac6_lport; 2549 } ipa6_conn_t; 2550 2551 /* 2552 * Using ipa_conn_x_t or ipa6_conn_x_t allows us to modify the behavior of IP's 2553 * bind handler. 2554 */ 2555 typedef struct ipa_conn_extended_s { 2556 uint64_t acx_flags; 2557 ipa_conn_t acx_conn; 2558 } ipa_conn_x_t; 2559 2560 typedef struct ipa6_conn_extended_s { 2561 uint64_t ac6x_flags; 2562 ipa6_conn_t ac6x_conn; 2563 } ipa6_conn_x_t; 2564 2565 /* flag values for ipa_conn_x_t and ipa6_conn_x_t. */ 2566 #define ACX_VERIFY_DST 0x1ULL /* verify destination address is reachable */ 2567 2568 /* Name/Value Descriptor. */ 2569 typedef struct nv_s { 2570 uint64_t nv_value; 2571 char *nv_name; 2572 } nv_t; 2573 2574 /* IP Forwarding Ticket */ 2575 typedef struct ipftk_s { 2576 queue_t *ipftk_queue; 2577 ipaddr_t ipftk_dst; 2578 } ipftk_t; 2579 2580 typedef struct ipt_s { 2581 pfv_t func; /* Routine to call */ 2582 uchar_t *arg; /* ire or nce passed in */ 2583 } ipt_t; 2584 2585 #define ILL_FRAG_HASH(s, i) \ 2586 ((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT) 2587 2588 /* 2589 * The MAX number of allowed fragmented packets per hash bucket 2590 * calculation is based on the most common mtu size of 1500. This limit 2591 * will work well for other mtu sizes as well. 2592 */ 2593 #define COMMON_IP_MTU 1500 2594 #define MAX_FRAG_MIN 10 2595 #define MAX_FRAG_PKTS \ 2596 MAX(MAX_FRAG_MIN, (2 * (ip_reass_queue_bytes / \ 2597 (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT)))) 2598 2599 /* 2600 * Maximum dups allowed per packet. 2601 */ 2602 extern uint_t ip_max_frag_dups; 2603 2604 /* 2605 * Per-packet information for received packets and transmitted. 2606 * Used by the transport protocols when converting between the packet 2607 * and ancillary data and socket options. 2608 * 2609 * Note: This private data structure and related IPPF_* constant 2610 * definitions are exposed to enable compilation of some debugging tools 2611 * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be 2612 * a temporary hack and long term alternate interfaces should be defined 2613 * to support the needs of such tools and private definitions moved to 2614 * private headers. 2615 */ 2616 struct ip6_pkt_s { 2617 uint_t ipp_fields; /* Which fields are valid */ 2618 uint_t ipp_sticky_ignored; /* sticky fields to ignore */ 2619 uint_t ipp_ifindex; /* pktinfo ifindex */ 2620 in6_addr_t ipp_addr; /* pktinfo src/dst addr */ 2621 uint_t ipp_unicast_hops; /* IPV6_UNICAST_HOPS */ 2622 uint_t ipp_multicast_hops; /* IPV6_MULTICAST_HOPS */ 2623 uint_t ipp_hoplimit; /* IPV6_HOPLIMIT */ 2624 uint_t ipp_hopoptslen; 2625 uint_t ipp_rtdstoptslen; 2626 uint_t ipp_rthdrlen; 2627 uint_t ipp_dstoptslen; 2628 uint_t ipp_pathmtulen; 2629 ip6_hbh_t *ipp_hopopts; 2630 ip6_dest_t *ipp_rtdstopts; 2631 ip6_rthdr_t *ipp_rthdr; 2632 ip6_dest_t *ipp_dstopts; 2633 struct ip6_mtuinfo *ipp_pathmtu; 2634 in6_addr_t ipp_nexthop; /* Transmit only */ 2635 uint8_t ipp_tclass; 2636 int8_t ipp_use_min_mtu; 2637 }; 2638 typedef struct ip6_pkt_s ip6_pkt_t; 2639 2640 extern void ip6_pkt_free(ip6_pkt_t *); /* free storage inside ip6_pkt_t */ 2641 2642 /* 2643 * This structure is used to convey information from IP and the ULP. 2644 * Currently used for the IP_RECVSLLA and IP_RECVIF options. The 2645 * type of information field is set to IN_PKTINFO (i.e inbound pkt info) 2646 */ 2647 typedef struct in_pktinfo { 2648 uint32_t in_pkt_ulp_type; /* type of info sent */ 2649 /* to UDP */ 2650 uint32_t in_pkt_flags; /* what is sent up by IP */ 2651 uint32_t in_pkt_ifindex; /* inbound interface index */ 2652 struct sockaddr_dl in_pkt_slla; /* has source link layer addr */ 2653 } in_pktinfo_t; 2654 2655 /* 2656 * flags to tell UDP what IP is sending; in_pkt_flags 2657 */ 2658 #define IPF_RECVIF 0x01 /* inbound interface index */ 2659 #define IPF_RECVSLLA 0x02 /* source link layer address */ 2660 2661 /* ipp_fields values */ 2662 #define IPPF_IFINDEX 0x0001 /* Part of in6_pktinfo: ifindex */ 2663 #define IPPF_ADDR 0x0002 /* Part of in6_pktinfo: src/dst addr */ 2664 #define IPPF_SCOPE_ID 0x0004 /* Add xmit ip6i_t for sin6_scope_id */ 2665 #define IPPF_NO_CKSUM 0x0008 /* Add xmit ip6i_t for IP6I_NO_*_CKSUM */ 2666 2667 #define IPPF_RAW_CKSUM 0x0010 /* Add xmit ip6i_t for IP6I_RAW_CHECKSUM */ 2668 #define IPPF_HOPLIMIT 0x0020 2669 #define IPPF_HOPOPTS 0x0040 2670 #define IPPF_RTHDR 0x0080 2671 2672 #define IPPF_RTDSTOPTS 0x0100 2673 #define IPPF_DSTOPTS 0x0200 2674 #define IPPF_NEXTHOP 0x0400 2675 #define IPPF_PATHMTU 0x0800 2676 2677 #define IPPF_TCLASS 0x1000 2678 #define IPPF_DONTFRAG 0x2000 2679 #define IPPF_USE_MIN_MTU 0x04000 2680 #define IPPF_MULTICAST_HOPS 0x08000 2681 #define IPPF_UNICAST_HOPS 0x10000 2682 2683 #define IPPF_HAS_IP6I \ 2684 (IPPF_IFINDEX|IPPF_ADDR|IPPF_NEXTHOP|IPPF_SCOPE_ID| \ 2685 IPPF_NO_CKSUM|IPPF_RAW_CKSUM|IPPF_HOPLIMIT|IPPF_DONTFRAG| \ 2686 IPPF_USE_MIN_MTU|IPPF_MULTICAST_HOPS|IPPF_UNICAST_HOPS) 2687 2688 #define TCP_PORTS_OFFSET 0 2689 #define UDP_PORTS_OFFSET 0 2690 2691 /* 2692 * lookups return the ill/ipif only if the flags are clear OR Iam writer. 2693 * ill / ipif lookup functions increment the refcnt on the ill / ipif only 2694 * after calling these macros. This ensures that the refcnt on the ipif or 2695 * ill will eventually drop down to zero. 2696 */ 2697 #define ILL_LOOKUP_FAILED 1 /* Used as error code */ 2698 #define IPIF_LOOKUP_FAILED 2 /* Used as error code */ 2699 2700 #define ILL_CAN_LOOKUP(ill) \ 2701 (!((ill)->ill_state_flags & (ILL_CONDEMNED | ILL_CHANGING)) || \ 2702 IAM_WRITER_ILL(ill)) 2703 2704 #define ILL_CAN_WAIT(ill, q) \ 2705 (((q) != NULL) && !((ill)->ill_state_flags & (ILL_CONDEMNED))) 2706 2707 #define ILL_CAN_LOOKUP_WALKER(ill) \ 2708 (!((ill)->ill_state_flags & ILL_CONDEMNED)) 2709 2710 #define IPIF_CAN_LOOKUP(ipif) \ 2711 (!((ipif)->ipif_state_flags & (IPIF_CONDEMNED | IPIF_CHANGING)) || \ 2712 IAM_WRITER_IPIF(ipif)) 2713 2714 /* 2715 * If the parameter 'q' is NULL, the caller is not interested in wait and 2716 * restart of the operation if the ILL or IPIF cannot be looked up when it is 2717 * marked as 'CHANGING'. Typically a thread that tries to send out data will 2718 * end up passing NULLs as the last 4 parameters to ill_lookup_on_ifindex and 2719 * in this case 'q' is NULL 2720 */ 2721 #define IPIF_CAN_WAIT(ipif, q) \ 2722 (((q) != NULL) && !((ipif)->ipif_state_flags & (IPIF_CONDEMNED))) 2723 2724 #define IPIF_CAN_LOOKUP_WALKER(ipif) \ 2725 (!((ipif)->ipif_state_flags & (IPIF_CONDEMNED)) || \ 2726 IAM_WRITER_IPIF(ipif)) 2727 2728 /* 2729 * These macros are used by critical set ioctls and failover ioctls to 2730 * mark the ipif appropriately before starting the operation and to clear the 2731 * marks after completing the operation. 2732 */ 2733 #define IPIF_UNMARK_MOVING(ipif) \ 2734 (ipif)->ipif_state_flags &= ~IPIF_MOVING & ~IPIF_CHANGING; 2735 2736 #define ILL_UNMARK_CHANGING(ill) \ 2737 (ill)->ill_state_flags &= ~ILL_CHANGING; 2738 2739 /* Macros used to assert that this thread is a writer */ 2740 #define IAM_WRITER_IPSQ(ipsq) ((ipsq)->ipsq_writer == curthread) 2741 #define IAM_WRITER_ILL(ill) \ 2742 ((ill)->ill_phyint->phyint_ipsq->ipsq_writer == curthread) 2743 #define IAM_WRITER_IPIF(ipif) \ 2744 ((ipif)->ipif_ill->ill_phyint->phyint_ipsq->ipsq_writer == curthread) 2745 2746 /* 2747 * Grab ill locks in the proper order. The order is highest addressed 2748 * ill is locked first. 2749 */ 2750 #define GRAB_ILL_LOCKS(ill_1, ill_2) \ 2751 { \ 2752 if ((ill_1) > (ill_2)) { \ 2753 if (ill_1 != NULL) \ 2754 mutex_enter(&(ill_1)->ill_lock); \ 2755 if (ill_2 != NULL) \ 2756 mutex_enter(&(ill_2)->ill_lock); \ 2757 } else { \ 2758 if (ill_2 != NULL) \ 2759 mutex_enter(&(ill_2)->ill_lock); \ 2760 if (ill_1 != NULL && ill_1 != ill_2) \ 2761 mutex_enter(&(ill_1)->ill_lock); \ 2762 } \ 2763 } 2764 2765 #define RELEASE_ILL_LOCKS(ill_1, ill_2) \ 2766 { \ 2767 if (ill_1 != NULL) \ 2768 mutex_exit(&(ill_1)->ill_lock); \ 2769 if (ill_2 != NULL && ill_2 != ill_1) \ 2770 mutex_exit(&(ill_2)->ill_lock); \ 2771 } 2772 2773 /* Get the other protocol instance ill */ 2774 #define ILL_OTHER(ill) \ 2775 ((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 : \ 2776 (ill)->ill_phyint->phyint_illv6) 2777 2778 #define MATCH_V4_ONLY 0x1 2779 #define MATCH_V6_ONLY 0x2 2780 #define MATCH_ILL_ONLY 0x4 2781 2782 /* ioctl command info: Ioctl properties extracted and stored in here */ 2783 typedef struct cmd_info_s 2784 { 2785 char ci_groupname[LIFNAMSIZ + 1]; /* SIOCSLIFGROUPNAME */ 2786 ipif_t *ci_ipif; /* ipif associated with [l]ifreq ioctl's */ 2787 sin_t *ci_sin; /* the sin struct passed down */ 2788 sin6_t *ci_sin6; /* the sin6_t struct passed down */ 2789 struct lifreq *ci_lifr; /* the lifreq struct passed down */ 2790 } cmd_info_t; 2791 2792 extern krwlock_t ill_g_lock; 2793 extern kmutex_t ip_addr_avail_lock; 2794 extern ipsq_t *ipsq_g_head; 2795 2796 extern ill_t *ip_timer_ill; /* ILL for IRE expiration timer. */ 2797 extern timeout_id_t ip_ire_expire_id; /* IRE expiration timeout id. */ 2798 extern timeout_id_t ip_ire_reclaim_id; /* IRE recalaim timeout id. */ 2799 2800 extern kmutex_t ip_mi_lock; 2801 extern krwlock_t ip_g_nd_lock; /* For adding/removing nd variables */ 2802 extern kmutex_t ip_trash_timer_lock; /* Protects ip_ire_expire_id */ 2803 2804 extern kmutex_t igmp_timer_lock; /* Protects the igmp timer */ 2805 extern kmutex_t mld_timer_lock; /* Protects the mld timer */ 2806 2807 extern krwlock_t ill_g_usesrc_lock; /* Protects usesrc related fields */ 2808 2809 extern struct kmem_cache *ire_cache; 2810 2811 extern uint_t ip_redirect_cnt; /* Num of redirect routes in ftable */ 2812 2813 extern ipaddr_t ip_g_all_ones; 2814 extern caddr_t ip_g_nd; /* Named Dispatch List Head */ 2815 2816 extern uint_t ip_loopback_mtu; 2817 2818 extern ipparam_t *ip_param_arr; 2819 2820 extern int ip_g_forward; 2821 extern int ipv6_forward; 2822 extern vmem_t *ip_minor_arena; 2823 2824 #define ip_respond_to_address_mask_broadcast ip_param_arr[0].ip_param_value 2825 #define ip_g_send_redirects ip_param_arr[5].ip_param_value 2826 #define ip_debug ip_param_arr[7].ip_param_value 2827 #define ip_mrtdebug ip_param_arr[8].ip_param_value 2828 #define ip_timer_interval ip_param_arr[9].ip_param_value 2829 #define ip_ire_arp_interval ip_param_arr[10].ip_param_value 2830 #define ip_def_ttl ip_param_arr[12].ip_param_value 2831 #define ip_wroff_extra ip_param_arr[14].ip_param_value 2832 #define ip_path_mtu_discovery ip_param_arr[17].ip_param_value 2833 #define ip_ignore_delete_time ip_param_arr[18].ip_param_value 2834 #define ip_output_queue ip_param_arr[20].ip_param_value 2835 #define ip_broadcast_ttl ip_param_arr[21].ip_param_value 2836 #define ip_icmp_err_interval ip_param_arr[22].ip_param_value 2837 #define ip_icmp_err_burst ip_param_arr[23].ip_param_value 2838 #define ip_reass_queue_bytes ip_param_arr[24].ip_param_value 2839 #define ip_addrs_per_if ip_param_arr[26].ip_param_value 2840 #define ipsec_override_persocket_policy ip_param_arr[27].ip_param_value 2841 #define icmp_accept_clear_messages ip_param_arr[28].ip_param_value 2842 #define delay_first_probe_time ip_param_arr[30].ip_param_value 2843 #define max_unicast_solicit ip_param_arr[31].ip_param_value 2844 #define ipv6_def_hops ip_param_arr[32].ip_param_value 2845 #define ipv6_icmp_return ip_param_arr[33].ip_param_value 2846 #define ipv6_forward_src_routed ip_param_arr[34].ip_param_value 2847 #define ipv6_resp_echo_mcast ip_param_arr[35].ip_param_value 2848 #define ipv6_send_redirects ip_param_arr[36].ip_param_value 2849 #define ipv6_ignore_redirect ip_param_arr[37].ip_param_value 2850 #define ipv6_strict_dst_multihoming ip_param_arr[38].ip_param_value 2851 #define ip_ire_reclaim_fraction ip_param_arr[39].ip_param_value 2852 #define ipsec_policy_log_interval ip_param_arr[40].ip_param_value 2853 #define ip_ndp_unsolicit_interval ip_param_arr[42].ip_param_value 2854 #define ip_ndp_unsolicit_count ip_param_arr[43].ip_param_value 2855 #define ipv6_ignore_home_address_opt ip_param_arr[44].ip_param_value 2856 #define ip_policy_mask ip_param_arr[45].ip_param_value 2857 #define ip_multirt_resolution_interval ip_param_arr[46].ip_param_value 2858 #define ip_multirt_ttl ip_param_arr[47].ip_param_value 2859 #define ip_multidata_outbound ip_param_arr[48].ip_param_value 2860 #ifdef DEBUG 2861 #define ipv6_drop_inbound_icmpv6 ip_param_arr[49].ip_param_value 2862 #else 2863 #define ipv6_drop_inbound_icmpv6 0 2864 #endif 2865 2866 extern hrtime_t ipsec_policy_failure_last; 2867 2868 extern int dohwcksum; /* use h/w cksum if supported by the h/w */ 2869 #ifdef ZC_TEST 2870 extern int noswcksum; 2871 #endif 2872 2873 extern char ipif_loopback_name[]; 2874 2875 extern nv_t *ire_nv_tbl; 2876 2877 extern time_t ip_g_frag_timeout; 2878 extern clock_t ip_g_frag_timo_ms; 2879 2880 extern mib2_ip_t ip_mib; /* For tcpInErrs and udpNoPorts */ 2881 2882 extern struct module_info ip_mod_info; 2883 2884 extern timeout_id_t igmp_slowtimeout_id; 2885 extern timeout_id_t mld_slowtimeout_id; 2886 2887 extern uint_t loopback_packets; 2888 2889 /* 2890 * Network byte order macros 2891 */ 2892 #ifdef _BIG_ENDIAN 2893 #define N_IN_CLASSD_NET IN_CLASSD_NET 2894 #define N_INADDR_UNSPEC_GROUP INADDR_UNSPEC_GROUP 2895 #else /* _BIG_ENDIAN */ 2896 #define N_IN_CLASSD_NET (ipaddr_t)0x000000f0U 2897 #define N_INADDR_UNSPEC_GROUP (ipaddr_t)0x000000e0U 2898 #endif /* _BIG_ENDIAN */ 2899 #define CLASSD(addr) (((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP) 2900 2901 #define IP_LOOPBACK_ADDR(addr) \ 2902 ((ntohl(addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) 2903 2904 #ifdef DEBUG 2905 /* IPsec HW acceleration debugging support */ 2906 2907 #define IPSECHW_CAPAB 0x0001 /* capability negotiation */ 2908 #define IPSECHW_SADB 0x0002 /* SADB exchange */ 2909 #define IPSECHW_PKT 0x0004 /* general packet flow */ 2910 #define IPSECHW_PKTIN 0x0008 /* driver in pkt processing details */ 2911 #define IPSECHW_PKTOUT 0x0010 /* driver out pkt processing details */ 2912 2913 #define IPSECHW_DEBUG(f, x) if (ipsechw_debug & (f)) { (void) printf x; } 2914 #define IPSECHW_CALL(f, r, x) if (ipsechw_debug & (f)) { (void) r x; } 2915 2916 extern uint32_t ipsechw_debug; 2917 #else 2918 #define IPSECHW_DEBUG(f, x) {} 2919 #define IPSECHW_CALL(f, r, x) {} 2920 #endif 2921 2922 #ifdef IP_DEBUG 2923 #include <sys/debug.h> 2924 #include <sys/promif.h> 2925 2926 #define ip0dbg(a) printf a 2927 #define ip1dbg(a) if (ip_debug > 2) printf a 2928 #define ip2dbg(a) if (ip_debug > 3) printf a 2929 #define ip3dbg(a) if (ip_debug > 4) printf a 2930 #else 2931 #define ip0dbg(a) /* */ 2932 #define ip1dbg(a) /* */ 2933 #define ip2dbg(a) /* */ 2934 #define ip3dbg(a) /* */ 2935 #endif /* IP_DEBUG */ 2936 2937 struct ipsec_out_s; 2938 2939 extern const char *dlpi_prim_str(int); 2940 extern const char *dlpi_err_str(int); 2941 extern void ill_frag_timer(void *); 2942 extern ill_t *ill_first(int, int, ill_walk_context_t *); 2943 extern ill_t *ill_next(ill_walk_context_t *, ill_t *); 2944 extern void ill_frag_timer_start(ill_t *); 2945 extern mblk_t *ip_carve_mp(mblk_t **, ssize_t); 2946 extern mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t); 2947 extern char *ip_dot_addr(ipaddr_t, char *); 2948 extern void ip_lwput(queue_t *, mblk_t *); 2949 extern boolean_t icmp_err_rate_limit(void); 2950 extern void icmp_time_exceeded(queue_t *, mblk_t *, uint8_t); 2951 extern void icmp_unreachable(queue_t *, mblk_t *, uint8_t); 2952 extern mblk_t *ip_add_info(mblk_t *, ill_t *, uint_t); 2953 extern mblk_t *ip_bind_v4(queue_t *, mblk_t *, conn_t *); 2954 extern int ip_bind_connected(conn_t *, mblk_t *, ipaddr_t *, uint16_t, 2955 ipaddr_t, uint16_t, boolean_t, boolean_t, boolean_t, 2956 boolean_t); 2957 extern boolean_t ip_bind_ipsec_policy_set(conn_t *, mblk_t *); 2958 extern int ip_bind_laddr(conn_t *, mblk_t *, ipaddr_t, uint16_t, 2959 boolean_t, boolean_t, boolean_t); 2960 extern uint_t ip_cksum(mblk_t *, int, uint32_t); 2961 extern int ip_close(queue_t *, int); 2962 extern uint16_t ip_csum_hdr(ipha_t *); 2963 extern void ip_proto_not_sup(queue_t *, mblk_t *, uint_t, zoneid_t); 2964 extern void ip_ire_fini(void); 2965 extern void ip_ire_init(void); 2966 extern int ip_open(queue_t *, dev_t *, int, int, cred_t *); 2967 extern int ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *, 2968 size_t); 2969 extern int ip_opt_set_ill(conn_t *, int, boolean_t, boolean_t, 2970 int, int, mblk_t *); 2971 extern void ip_rput(queue_t *, mblk_t *); 2972 extern void ip_input(ill_t *, ill_rx_ring_t *, mblk_t *, size_t); 2973 extern void ip_rput_dlpi(queue_t *, mblk_t *); 2974 extern void ip_rput_forward(ire_t *, ipha_t *, mblk_t *, ill_t *); 2975 extern void ip_rput_forward_multicast(ipaddr_t, mblk_t *, ipif_t *); 2976 2977 extern int ip_snmpmod_close(queue_t *); 2978 extern void ip_snmpmod_wput(queue_t *, mblk_t *); 2979 extern void ip_udp_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *); 2980 extern void ip_proto_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *); 2981 extern void ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *); 2982 extern void ip_setqinfo(queue_t *, minor_t, boolean_t); 2983 extern void ip_trash_ire_reclaim(void *); 2984 extern void ip_trash_timer_expire(void *); 2985 extern void ip_wput(queue_t *, mblk_t *); 2986 extern void ip_output(void *, mblk_t *, void *, int); 2987 extern void ip_wput_md(queue_t *, mblk_t *, conn_t *); 2988 2989 extern void ip_wput_ire(queue_t *, mblk_t *, ire_t *, conn_t *, int); 2990 extern void ip_wput_local(queue_t *, ill_t *, ipha_t *, mblk_t *, ire_t *, 2991 int, zoneid_t); 2992 extern void ip_wput_multicast(queue_t *, mblk_t *, ipif_t *); 2993 extern void ip_wput_nondata(ipsq_t *, queue_t *, mblk_t *, void *); 2994 extern void ip_wsrv(queue_t *); 2995 extern char *ip_nv_lookup(nv_t *, int); 2996 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *); 2997 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *); 2998 extern ipaddr_t ip_massage_options(ipha_t *); 2999 extern ipaddr_t ip_net_mask(ipaddr_t); 3000 extern void ip_newroute(queue_t *, mblk_t *, ipaddr_t, ill_t *, conn_t *); 3001 extern ipxmit_state_t ip_xmit_v4(mblk_t *, ire_t *, struct ipsec_out_s *, 3002 boolean_t); 3003 extern int ip_hdr_complete(ipha_t *, zoneid_t); 3004 3005 extern struct qinit rinit_ipv6; 3006 extern struct qinit winit_ipv6; 3007 extern struct qinit rinit_tcp; 3008 extern struct qinit rinit_tcp6; 3009 extern struct qinit winit_tcp; 3010 extern struct qinit rinit_acceptor_tcp; 3011 extern struct qinit winit_acceptor_tcp; 3012 3013 extern void conn_drain_insert(conn_t *connp); 3014 extern int conn_ipsec_length(conn_t *connp); 3015 extern void ip_wput_ipsec_out(queue_t *, mblk_t *, ipha_t *, ill_t *, 3016 ire_t *); 3017 extern ipaddr_t ip_get_dst(ipha_t *); 3018 extern int ipsec_out_extra_length(mblk_t *); 3019 extern int ipsec_in_extra_length(mblk_t *); 3020 extern mblk_t *ipsec_in_alloc(); 3021 extern boolean_t ipsec_in_is_secure(mblk_t *); 3022 extern void ipsec_out_process(queue_t *, mblk_t *, ire_t *, uint_t); 3023 extern void ipsec_out_to_in(mblk_t *); 3024 extern int ill_forward_set(queue_t *, mblk_t *, boolean_t, caddr_t); 3025 extern void ip_fanout_proto_again(mblk_t *, ill_t *, ill_t *, ire_t *); 3026 3027 extern void ire_cleanup(ire_t *); 3028 extern void ire_inactive(ire_t *); 3029 extern boolean_t irb_inactive(irb_t *); 3030 extern ire_t *ire_unlink(irb_t *); 3031 #ifdef IRE_DEBUG 3032 extern void ire_trace_ref(ire_t *ire); 3033 extern void ire_untrace_ref(ire_t *ire); 3034 extern void ire_thread_exit(ire_t *ire, caddr_t); 3035 #endif 3036 #ifdef ILL_DEBUG 3037 extern void ill_trace_cleanup(ill_t *); 3038 extern void ipif_trace_cleanup(ipif_t *); 3039 #endif 3040 3041 extern int ip_srcid_insert(const in6_addr_t *, zoneid_t); 3042 extern int ip_srcid_remove(const in6_addr_t *, zoneid_t); 3043 extern void ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t); 3044 extern uint_t ip_srcid_find_addr(const in6_addr_t *, zoneid_t); 3045 extern int ip_srcid_report(queue_t *, mblk_t *, caddr_t, cred_t *); 3046 3047 extern uint8_t ipoptp_next(ipoptp_t *); 3048 extern uint8_t ipoptp_first(ipoptp_t *, ipha_t *); 3049 extern int ip_opt_get_user(const ipha_t *, uchar_t *); 3050 extern ill_t *ip_grab_attach_ill(ill_t *, mblk_t *, int, boolean_t); 3051 extern ire_t *conn_set_outgoing_ill(conn_t *, ire_t *, ill_t **); 3052 extern int ipsec_req_from_conn(conn_t *, ipsec_req_t *, int); 3053 extern int ip_snmp_get(queue_t *q, mblk_t *mctl); 3054 extern int ip_snmp_set(queue_t *q, int, int, uchar_t *, int); 3055 extern void ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *); 3056 extern void ip_quiesce_conn(conn_t *); 3057 extern void ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *); 3058 extern void ip_restart_optmgmt(ipsq_t *, queue_t *, mblk_t *, void *); 3059 extern void ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipif_t *, 3060 ipsq_t *); 3061 3062 extern boolean_t ip_cmpbuf(const void *, uint_t, boolean_t, const void *, 3063 uint_t); 3064 extern boolean_t ip_allocbuf(void **, uint_t *, boolean_t, const void *, 3065 uint_t); 3066 extern void ip_savebuf(void **, uint_t *, boolean_t, const void *, uint_t); 3067 3068 extern boolean_t ipsq_pending_mp_cleanup(ill_t *, conn_t *); 3069 extern void conn_ioctl_cleanup(conn_t *); 3070 extern ill_t *conn_get_held_ill(conn_t *, ill_t **, int *); 3071 extern ill_t *ip_newroute_get_dst_ill(ill_t *); 3072 3073 struct multidata_s; 3074 struct pdesc_s; 3075 3076 extern mblk_t *ip_mdinfo_alloc(ill_mdt_capab_t *); 3077 extern mblk_t *ip_mdinfo_return(ire_t *, conn_t *, char *, ill_mdt_capab_t *); 3078 extern uint_t ip_md_cksum(struct pdesc_s *, int, uint_t); 3079 extern boolean_t ip_md_addr_attr(struct multidata_s *, struct pdesc_s *, 3080 const mblk_t *); 3081 extern boolean_t ip_md_hcksum_attr(struct multidata_s *, struct pdesc_s *, 3082 uint32_t, uint32_t, uint32_t, uint32_t); 3083 extern boolean_t ip_md_zcopy_attr(struct multidata_s *, struct pdesc_s *, 3084 uint_t); 3085 extern mblk_t *ip_unbind(queue_t *, mblk_t *); 3086 3087 extern void tnet_init(void); 3088 extern void tnet_fini(void); 3089 3090 /* Hooks for CGTP (multirt routes) filtering module */ 3091 #define CGTP_FILTER_REV_1 1 3092 #define CGTP_FILTER_REV_2 2 3093 #define CGTP_FILTER_REV CGTP_FILTER_REV_2 3094 3095 /* cfo_filter, cfo_filter_fp, cfo_filter_v6 hooks return values */ 3096 #define CGTP_IP_PKT_NOT_CGTP 0 3097 #define CGTP_IP_PKT_PREMIUM 1 3098 #define CGTP_IP_PKT_DUPLICATE 2 3099 3100 typedef struct cgtp_filter_ops { 3101 int cfo_filter_rev; 3102 int (*cfo_change_state)(int); 3103 int (*cfo_add_dest_v4)(ipaddr_t, ipaddr_t, ipaddr_t, ipaddr_t); 3104 int (*cfo_del_dest_v4)(ipaddr_t, ipaddr_t); 3105 int (*cfo_add_dest_v6)(in6_addr_t *, in6_addr_t *, in6_addr_t *, 3106 in6_addr_t *); 3107 int (*cfo_del_dest_v6)(in6_addr_t *, in6_addr_t *); 3108 int (*cfo_filter)(queue_t *, mblk_t *); 3109 int (*cfo_filter_fp)(queue_t *, mblk_t *); 3110 int (*cfo_filter_v6)(queue_t *, ip6_t *, ip6_frag_t *); 3111 } cgtp_filter_ops_t; 3112 3113 #define CGTP_MCAST_SUCCESS 1 3114 3115 extern cgtp_filter_ops_t *ip_cgtp_filter_ops; 3116 extern boolean_t ip_cgtp_filter; 3117 3118 extern int ip_cgtp_filter_supported(void); 3119 extern int ip_cgtp_filter_register(cgtp_filter_ops_t *); 3120 3121 /* Flags for ire_multirt_lookup() */ 3122 3123 #define MULTIRT_USESTAMP 0x0001 3124 #define MULTIRT_SETSTAMP 0x0002 3125 #define MULTIRT_CACHEGW 0x0004 3126 3127 /* Debug stuff for multirt route resolution. */ 3128 #if defined(DEBUG) && !defined(__lint) 3129 /* Our "don't send, rather drop" flag. */ 3130 #define MULTIRT_DEBUG_FLAG 0x8000 3131 3132 #define MULTIRT_TRACE(x) ip2dbg(x) 3133 3134 #define MULTIRT_DEBUG_TAG(mblk) \ 3135 do { \ 3136 ASSERT(mblk != NULL); \ 3137 MULTIRT_TRACE(("%s[%d]: tagging mblk %p, tag was %d\n", \ 3138 __FILE__, __LINE__, \ 3139 (void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \ 3140 (mblk)->b_flag |= MULTIRT_DEBUG_FLAG; \ 3141 } while (0) 3142 3143 #define MULTIRT_DEBUG_UNTAG(mblk) \ 3144 do { \ 3145 ASSERT(mblk != NULL); \ 3146 MULTIRT_TRACE(("%s[%d]: untagging mblk %p, tag was %d\n", \ 3147 __FILE__, __LINE__, \ 3148 (void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \ 3149 (mblk)->b_flag &= ~MULTIRT_DEBUG_FLAG; \ 3150 } while (0) 3151 3152 #define MULTIRT_DEBUG_TAGGED(mblk) \ 3153 (((mblk)->b_flag & MULTIRT_DEBUG_FLAG) ? B_TRUE : B_FALSE) 3154 #else 3155 #define MULTIRT_DEBUG_TAG(mblk) ASSERT(mblk != NULL) 3156 #define MULTIRT_DEBUG_UNTAG(mblk) ASSERT(mblk != NULL) 3157 #define MULTIRT_DEBUG_TAGGED(mblk) B_FALSE 3158 #endif 3159 3160 /* 3161 * Per-ILL Multidata Transmit capabilities. 3162 */ 3163 struct ill_mdt_capab_s { 3164 uint_t ill_mdt_version; /* interface version */ 3165 uint_t ill_mdt_on; /* on/off switch for MDT on this ILL */ 3166 uint_t ill_mdt_hdr_head; /* leading header fragment extra space */ 3167 uint_t ill_mdt_hdr_tail; /* trailing header fragment extra space */ 3168 uint_t ill_mdt_max_pld; /* maximum payload buffers per Multidata */ 3169 uint_t ill_mdt_span_limit; /* maximum payload span per packet */ 3170 }; 3171 3172 struct ill_hcksum_capab_s { 3173 uint_t ill_hcksum_version; /* interface version */ 3174 uint_t ill_hcksum_txflags; /* capabilities on transmit */ 3175 }; 3176 3177 struct ill_zerocopy_capab_s { 3178 uint_t ill_zerocopy_version; /* interface version */ 3179 uint_t ill_zerocopy_flags; /* capabilities */ 3180 }; 3181 3182 /* Possible ill_states */ 3183 #define ILL_RING_INPROC 3 /* Being assigned to squeue */ 3184 #define ILL_RING_INUSE 2 /* Already Assigned to Rx Ring */ 3185 #define ILL_RING_BEING_FREED 1 /* Being Unassigned */ 3186 #define ILL_RING_FREE 0 /* Available to be assigned to Ring */ 3187 3188 #define ILL_MAX_RINGS 256 /* Max num of rx rings we can manage */ 3189 #define ILL_POLLING 0x01 /* Polling in use */ 3190 3191 /* 3192 * These functions pointer types are exported by the mac/dls layer. 3193 * we need to duplicate the definitions here because we cannot 3194 * include mac/dls header files here. 3195 */ 3196 typedef void (*ip_mac_blank_t)(void *, time_t, uint_t); 3197 typedef void (*ip_dld_tx_t)(void *, mblk_t *); 3198 3199 typedef void (*ip_dls_chg_soft_ring_t)(void *, int); 3200 typedef void (*ip_dls_bind_t)(void *, processorid_t); 3201 typedef void (*ip_dls_unbind_t)(void *); 3202 3203 struct ill_rx_ring { 3204 ip_mac_blank_t rr_blank; /* Driver interrupt blanking func */ 3205 void *rr_handle; /* Handle for Rx ring */ 3206 squeue_t *rr_sqp; /* Squeue the ring is bound to */ 3207 ill_t *rr_ill; /* back pointer to ill */ 3208 clock_t rr_poll_time; /* Last lbolt polling was used */ 3209 uint32_t rr_poll_state; /* polling state flags */ 3210 uint32_t rr_max_blank_time; /* Max interrupt blank */ 3211 uint32_t rr_min_blank_time; /* Min interrupt blank */ 3212 uint32_t rr_max_pkt_cnt; /* Max pkts before interrupt */ 3213 uint32_t rr_min_pkt_cnt; /* Mix pkts before interrupt */ 3214 uint32_t rr_normal_blank_time; /* Normal intr freq */ 3215 uint32_t rr_normal_pkt_cnt; /* Normal intr pkt cnt */ 3216 uint32_t rr_ring_state; /* State of this ring */ 3217 }; 3218 3219 struct ill_dls_capab_s { 3220 ip_dld_tx_t ill_tx; /* Driver Tx routine */ 3221 void *ill_tx_handle; /* Driver Tx handle */ 3222 ip_dls_chg_soft_ring_t ill_dls_change_status; 3223 /* change soft ring fanout */ 3224 ip_dls_bind_t ill_dls_bind; /* to add CPU affinity */ 3225 ip_dls_unbind_t ill_dls_unbind; /* remove CPU affinity */ 3226 ill_rx_ring_t *ill_ring_tbl; /* Ring to Sqp mapping table */ 3227 uint_t ill_dls_soft_ring_cnt; /* Number of soft ring */ 3228 conn_t *ill_unbind_conn; /* Conn used during unplumb */ 3229 }; 3230 3231 /* 3232 * This message is sent by an upper-layer protocol to tell IP that it knows all 3233 * about labels and will construct them itself. IP takes the slow path and 3234 * recomputes the label on every packet when this isn't true. 3235 */ 3236 #define IP_ULP_OUT_LABELED (('O' << 8) + 'L') 3237 typedef struct out_labeled_s { 3238 uint32_t out_labeled_type; /* OUT_LABELED */ 3239 queue_t *out_qnext; /* intermediate detection */ 3240 } out_labeled_t; 3241 3242 /* 3243 * IP squeues exports 3244 */ 3245 extern int ip_squeue_profile; 3246 extern int ip_squeue_bind; 3247 extern boolean_t ip_squeue_fanout; 3248 extern boolean_t ip_squeue_soft_ring; 3249 extern uint_t ip_threads_per_cpu; 3250 extern uint_t ip_squeues_per_cpu; 3251 extern uint_t ip_soft_rings_cnt; 3252 3253 typedef struct squeue_set_s { 3254 kmutex_t sqs_lock; 3255 struct squeue_s **sqs_list; 3256 int sqs_size; 3257 int sqs_max_size; 3258 processorid_t sqs_bind; 3259 } squeue_set_t; 3260 3261 #define IP_SQUEUE_GET(hint) \ 3262 ((!ip_squeue_fanout) ? (CPU->cpu_squeue_set->sqs_list[0]) : \ 3263 ip_squeue_random(hint)) 3264 3265 typedef void (*squeue_func_t)(squeue_t *, mblk_t *, sqproc_t, void *, uint8_t); 3266 3267 extern void ip_squeue_init(void (*)(squeue_t *)); 3268 extern squeue_t *ip_squeue_random(uint_t); 3269 extern squeue_t *ip_squeue_get(ill_rx_ring_t *); 3270 extern void ip_squeue_get_pkts(squeue_t *); 3271 extern int ip_squeue_bind_set(queue_t *, mblk_t *, char *, caddr_t, cred_t *); 3272 extern int ip_squeue_bind_get(queue_t *, mblk_t *, caddr_t, cred_t *); 3273 extern void ip_squeue_clean(void *, mblk_t *, void *); 3274 extern void ip_resume_tcp_bind(void *, mblk_t *, void *); 3275 extern void ip_soft_ring_assignment(ill_t *, ill_rx_ring_t *, 3276 mblk_t *, size_t); 3277 3278 extern void tcp_wput(queue_t *, mblk_t *); 3279 3280 extern int ip_fill_mtuinfo(struct in6_addr *, in_port_t, 3281 struct ip6_mtuinfo *); 3282 extern ipif_t *conn_get_held_ipif(conn_t *, ipif_t **, int *); 3283 3284 typedef void (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *); 3285 3286 /* 3287 * Squeue tags. Tags only need to be unique when the callback function is the 3288 * same to distinguish between different calls, but we use unique tags for 3289 * convenience anyway. 3290 */ 3291 #define SQTAG_IP_INPUT 1 3292 #define SQTAG_TCP_INPUT_ICMP_ERR 2 3293 #define SQTAG_TCP6_INPUT_ICMP_ERR 3 3294 #define SQTAG_IP_TCP_INPUT 4 3295 #define SQTAG_IP6_TCP_INPUT 5 3296 #define SQTAG_IP_TCP_CLOSE 6 3297 #define SQTAG_TCP_OUTPUT 7 3298 #define SQTAG_TCP_TIMER 8 3299 #define SQTAG_TCP_TIMEWAIT 9 3300 #define SQTAG_TCP_ACCEPT_FINISH 10 3301 #define SQTAG_TCP_ACCEPT_FINISH_Q0 11 3302 #define SQTAG_TCP_ACCEPT_PENDING 12 3303 #define SQTAG_TCP_LISTEN_DISCON 13 3304 #define SQTAG_TCP_CONN_REQ 14 3305 #define SQTAG_TCP_EAGER_BLOWOFF 15 3306 #define SQTAG_TCP_EAGER_CLEANUP 16 3307 #define SQTAG_TCP_EAGER_CLEANUP_Q0 17 3308 #define SQTAG_TCP_CONN_IND 18 3309 #define SQTAG_TCP_RSRV 19 3310 #define SQTAG_TCP_ABORT_BUCKET 20 3311 #define SQTAG_TCP_REINPUT 21 3312 #define SQTAG_TCP_REINPUT_EAGER 22 3313 #define SQTAG_TCP_INPUT_MCTL 23 3314 #define SQTAG_TCP_RPUTOTHER 24 3315 #define SQTAG_IP_PROTO_AGAIN 25 3316 #define SQTAG_IP_FANOUT_TCP 26 3317 #define SQTAG_IPSQ_CLEAN_RING 27 3318 #define SQTAG_TCP_WPUT_OTHER 28 3319 #define SQTAG_TCP_CONN_REQ_UNBOUND 29 3320 #define SQTAG_TCP_SEND_PENDING 30 3321 #define SQTAG_BIND_RETRY 31 3322 #define SQTAG_UDP_FANOUT 32 3323 #define SQTAG_UDP_INPUT 33 3324 #define SQTAG_UDP_WPUT 34 3325 #define SQTAG_UDP_OUTPUT 35 3326 #define SQTAG_TCP_KSSL_INPUT 36 3327 3328 #endif /* _KERNEL */ 3329 3330 #ifdef __cplusplus 3331 } 3332 #endif 3333 3334 #endif /* _INET_IP_H */ 3335