1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 /* Copyright (c) 1990 Mentat Inc. */ 27 28 #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 <sys/systm.h> 55 #include <sys/multidata.h> 56 57 #ifdef DEBUG 58 #define ILL_DEBUG 59 #define IRE_DEBUG 60 #define NCE_DEBUG 61 #define CONN_DEBUG 62 #endif 63 64 #define IP_DEBUG 65 /* 66 * The mt-streams(9F) flags for the IP module; put here so that other 67 * "drivers" that are actually IP (e.g., ICMP, UDP) can use the same set 68 * of flags. 69 */ 70 #define IP_DEVMTFLAGS D_MP 71 #endif /* _KERNEL */ 72 73 #define IP_MOD_NAME "ip" 74 #define IP_DEV_NAME "/dev/ip" 75 #define IP6_DEV_NAME "/dev/ip6" 76 77 #define UDP_MOD_NAME "udp" 78 #define UDP_DEV_NAME "/dev/udp" 79 #define UDP6_DEV_NAME "/dev/udp6" 80 81 #define TCP_MOD_NAME "tcp" 82 #define TCP_DEV_NAME "/dev/tcp" 83 #define TCP6_DEV_NAME "/dev/tcp6" 84 85 /* Minor numbers */ 86 #define IPV4_MINOR 0 87 #define IPV6_MINOR 1 88 #define TCP_MINOR 2 89 #define TCP_MINOR6 3 90 91 #ifndef _IPADDR_T 92 #define _IPADDR_T 93 typedef uint32_t ipaddr_t; 94 #endif 95 96 /* Number of bits in an address */ 97 #define IP_ABITS 32 98 #define IPV6_ABITS 128 99 100 #define IP_HOST_MASK (ipaddr_t)0xffffffffU 101 102 #define IP_CSUM(mp, off, sum) (~ip_cksum(mp, off, sum) & 0xFFFF) 103 #define IP_CSUM_PARTIAL(mp, off, sum) ip_cksum(mp, off, sum) 104 #define IP_BCSUM_PARTIAL(bp, len, sum) bcksum(bp, len, sum) 105 #define IP_MD_CSUM(pd, off, sum) (~ip_md_cksum(pd, off, sum) & 0xffff) 106 #define IP_MD_CSUM_PARTIAL(pd, off, sum) ip_md_cksum(pd, off, sum) 107 108 /* 109 * Flag to IP write side to indicate that the appln has sent in a pre-built 110 * IP header. Stored in ipha_ident (which is otherwise zero). 111 */ 112 #define IP_HDR_INCLUDED 0xFFFF 113 114 #define ILL_FRAG_HASH_TBL_COUNT ((unsigned int)64) 115 #define ILL_FRAG_HASH_TBL_SIZE (ILL_FRAG_HASH_TBL_COUNT * sizeof (ipfb_t)) 116 117 #define IPV4_ADDR_LEN 4 118 #define IP_ADDR_LEN IPV4_ADDR_LEN 119 #define IP_ARP_PROTO_TYPE 0x0800 120 121 #define IPV4_VERSION 4 122 #define IP_VERSION IPV4_VERSION 123 #define IP_SIMPLE_HDR_LENGTH_IN_WORDS 5 124 #define IP_SIMPLE_HDR_LENGTH 20 125 #define IP_MAX_HDR_LENGTH 60 126 127 #define IP_MIN_MTU (IP_MAX_HDR_LENGTH + 8) /* 68 bytes */ 128 129 /* 130 * XXX IP_MAXPACKET is defined in <netinet/ip.h> as well. At some point the 131 * 2 files should be cleaned up to remove all redundant definitions. 132 */ 133 #define IP_MAXPACKET 65535 134 #define IP_SIMPLE_HDR_VERSION \ 135 ((IP_VERSION << 4) | IP_SIMPLE_HDR_LENGTH_IN_WORDS) 136 137 /* 138 * Constants and type definitions to support IP IOCTL commands 139 */ 140 #define IP_IOCTL (('i'<<8)|'p') 141 #define IP_IOC_IRE_DELETE 4 142 #define IP_IOC_IRE_DELETE_NO_REPLY 5 143 #define IP_IOC_IRE_ADVISE_NO_REPLY 6 144 #define IP_IOC_RTS_REQUEST 7 145 146 /* Common definitions used by IP IOCTL data structures */ 147 typedef struct ipllcmd_s { 148 uint_t ipllc_cmd; 149 uint_t ipllc_name_offset; 150 uint_t ipllc_name_length; 151 } ipllc_t; 152 153 /* IP IRE Change Command Structure. */ 154 typedef struct ipic_s { 155 ipllc_t ipic_ipllc; 156 uint_t ipic_ire_type; 157 uint_t ipic_max_frag; 158 uint_t ipic_addr_offset; 159 uint_t ipic_addr_length; 160 uint_t ipic_mask_offset; 161 uint_t ipic_mask_length; 162 uint_t ipic_src_addr_offset; 163 uint_t ipic_src_addr_length; 164 uint_t ipic_ll_hdr_offset; 165 uint_t ipic_ll_hdr_length; 166 uint_t ipic_gateway_addr_offset; 167 uint_t ipic_gateway_addr_length; 168 clock_t ipic_rtt; 169 uint32_t ipic_ssthresh; 170 clock_t ipic_rtt_sd; 171 uchar_t ipic_ire_marks; 172 } ipic_t; 173 174 #define ipic_cmd ipic_ipllc.ipllc_cmd 175 #define ipic_ll_name_length ipic_ipllc.ipllc_name_length 176 #define ipic_ll_name_offset ipic_ipllc.ipllc_name_offset 177 178 /* IP IRE Delete Command Structure. */ 179 typedef struct ipid_s { 180 ipllc_t ipid_ipllc; 181 uint_t ipid_ire_type; 182 uint_t ipid_addr_offset; 183 uint_t ipid_addr_length; 184 uint_t ipid_mask_offset; 185 uint_t ipid_mask_length; 186 } ipid_t; 187 188 #define ipid_cmd ipid_ipllc.ipllc_cmd 189 190 #ifdef _KERNEL 191 /* 192 * Temporary state for ip options parser. 193 */ 194 typedef struct ipoptp_s 195 { 196 uint8_t *ipoptp_next; /* next option to look at */ 197 uint8_t *ipoptp_end; /* end of options */ 198 uint8_t *ipoptp_cur; /* start of current option */ 199 uint8_t ipoptp_len; /* length of current option */ 200 uint32_t ipoptp_flags; 201 } ipoptp_t; 202 203 /* 204 * Flag(s) for ipoptp_flags 205 */ 206 #define IPOPTP_ERROR 0x00000001 207 #endif /* _KERNEL */ 208 209 210 /* Controls forwarding of IP packets, set via ndd */ 211 #define IP_FORWARD_NEVER 0 212 #define IP_FORWARD_ALWAYS 1 213 214 #define WE_ARE_FORWARDING (ip_g_forward == IP_FORWARD_ALWAYS) 215 216 #define IPH_HDR_LENGTH(ipha) \ 217 ((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length & 0xF) << 2) 218 219 #define IPH_HDR_VERSION(ipha) \ 220 ((int)(((ipha_t *)ipha)->ipha_version_and_hdr_length) >> 4) 221 222 #ifdef _KERNEL 223 /* 224 * IP reassembly macros. We hide starting and ending offsets in b_next and 225 * b_prev of messages on the reassembly queue. The messages are chained using 226 * b_cont. These macros are used in ip_reassemble() so we don't have to see 227 * the ugly casts and assignments. 228 * Note that the offsets are <= 64k i.e. a uint_t is sufficient to represent 229 * them. 230 */ 231 #define IP_REASS_START(mp) ((uint_t)(uintptr_t)((mp)->b_next)) 232 #define IP_REASS_SET_START(mp, u) \ 233 ((mp)->b_next = (mblk_t *)(uintptr_t)(u)) 234 #define IP_REASS_END(mp) ((uint_t)(uintptr_t)((mp)->b_prev)) 235 #define IP_REASS_SET_END(mp, u) \ 236 ((mp)->b_prev = (mblk_t *)(uintptr_t)(u)) 237 238 #define IP_REASS_COMPLETE 0x1 239 #define IP_REASS_PARTIAL 0x2 240 #define IP_REASS_FAILED 0x4 241 242 /* 243 * Test to determine whether this is a module instance of IP or a 244 * driver instance of IP. 245 */ 246 #define CONN_Q(q) (WR(q)->q_next == NULL) 247 248 #define Q_TO_CONN(q) ((conn_t *)(q)->q_ptr) 249 #define Q_TO_TCP(q) (Q_TO_CONN((q))->conn_tcp) 250 #define Q_TO_UDP(q) (Q_TO_CONN((q))->conn_udp) 251 252 /* 253 * The following two macros are used by IP to get the appropriate 254 * wq and rq for a conn. If it is a TCP conn, then we need 255 * tcp_wq/tcp_rq else, conn_wq/conn_rq. IP can use conn_wq and conn_rq 256 * from a conn directly if it knows that the conn is not TCP. 257 */ 258 #define CONNP_TO_WQ(connp) \ 259 (IPCL_IS_TCP(connp) ? (connp)->conn_tcp->tcp_wq : (connp)->conn_wq) 260 261 #define CONNP_TO_RQ(connp) RD(CONNP_TO_WQ(connp)) 262 263 #define GRAB_CONN_LOCK(q) { \ 264 if (q != NULL && CONN_Q(q)) \ 265 mutex_enter(&(Q_TO_CONN(q))->conn_lock); \ 266 } 267 268 #define RELEASE_CONN_LOCK(q) { \ 269 if (q != NULL && CONN_Q(q)) \ 270 mutex_exit(&(Q_TO_CONN(q))->conn_lock); \ 271 } 272 273 /* "Congestion controlled" protocol */ 274 #define IP_FLOW_CONTROLLED_ULP(p) ((p) == IPPROTO_TCP || (p) == IPPROTO_SCTP) 275 276 /* 277 * Complete the pending operation. Usually an ioctl. Can also 278 * be a bind or option management request that got enqueued 279 * in an ipsq_t. Called on completion of the operation. 280 */ 281 #define CONN_OPER_PENDING_DONE(connp) { \ 282 mutex_enter(&(connp)->conn_lock); \ 283 (connp)->conn_oper_pending_ill = NULL; \ 284 cv_broadcast(&(connp)->conn_refcv); \ 285 mutex_exit(&(connp)->conn_lock); \ 286 CONN_DEC_REF(connp); \ 287 } 288 289 /* Get the credential of an IP queue of unknown type */ 290 #define GET_QUEUE_CRED(wq) \ 291 ((wq)->q_next ? (((ill_t *)(wq)->q_ptr)->ill_credp) \ 292 : ((Q_TO_CONN((wq)))->conn_cred)) 293 294 /* 295 * Flags for the various ip_fanout_* routines. 296 */ 297 #define IP_FF_SEND_ICMP 0x01 /* Send an ICMP error */ 298 #define IP_FF_HDR_COMPLETE 0x02 /* Call ip_hdr_complete if error */ 299 #define IP_FF_CKSUM 0x04 /* Recompute ipha_cksum if error */ 300 #define IP_FF_RAWIP 0x08 /* Use rawip mib variable */ 301 #define IP_FF_SRC_QUENCH 0x10 /* OK to send ICMP_SOURCE_QUENCH */ 302 #define IP_FF_SYN_ADDIRE 0x20 /* Add IRE if TCP syn packet */ 303 #define IP_FF_IP6INFO 0x80 /* Add ip6i_t if needed */ 304 #define IP_FF_SEND_SLLA 0x100 /* Send source link layer info ? */ 305 #define IPV6_REACHABILITY_CONFIRMATION 0x200 /* Flags for ip_xmit_v6 */ 306 #define IP_FF_NO_MCAST_LOOP 0x400 /* No multicasts for sending zone */ 307 308 /* 309 * Following flags are used by IPQoS to determine if policy processing is 310 * required. 311 */ 312 #define IP6_NO_IPPOLICY 0x800 /* Don't do IPQoS processing */ 313 #define IP6_IN_LLMCAST 0x1000 /* Multicast */ 314 315 #define IP_FF_LOOPBACK 0x2000 /* Loopback fanout */ 316 317 #ifndef IRE_DB_TYPE 318 #define IRE_DB_TYPE M_SIG 319 #endif 320 321 #ifndef IRE_DB_REQ_TYPE 322 #define IRE_DB_REQ_TYPE M_PCSIG 323 #endif 324 325 /* 326 * Values for squeue switch: 327 */ 328 329 #define IP_SQUEUE_ENTER_NODRAIN 1 330 #define IP_SQUEUE_ENTER 2 331 /* 332 * This is part of the interface between Transport provider and 333 * IP which can be used to set policy information. This is usually 334 * accompanied with O_T_BIND_REQ/T_BIND_REQ.ip_bind assumes that 335 * only IPSEC_POLICY_SET is there when it is found in the chain. 336 * The information contained is an struct ipsec_req_t. On success 337 * or failure, either the T_BIND_ACK or the T_ERROR_ACK is returned. 338 * IPSEC_POLICY_SET is never returned. 339 */ 340 #define IPSEC_POLICY_SET M_SETOPTS 341 342 #define IRE_IS_LOCAL(ire) ((ire != NULL) && \ 343 ((ire)->ire_type & (IRE_LOCAL | IRE_LOOPBACK))) 344 345 #define IRE_IS_TARGET(ire) ((ire != NULL) && \ 346 ((ire)->ire_type != IRE_BROADCAST)) 347 348 /* IP Fragmentation Reassembly Header */ 349 typedef struct ipf_s { 350 struct ipf_s *ipf_hash_next; 351 struct ipf_s **ipf_ptphn; /* Pointer to previous hash next. */ 352 uint32_t ipf_ident; /* Ident to match. */ 353 uint8_t ipf_protocol; /* Protocol to match. */ 354 uchar_t ipf_last_frag_seen : 1; /* Last fragment seen ? */ 355 time_t ipf_timestamp; /* Reassembly start time. */ 356 mblk_t *ipf_mp; /* mblk we live in. */ 357 mblk_t *ipf_tail_mp; /* Frag queue tail pointer. */ 358 int ipf_hole_cnt; /* Number of holes (hard-case). */ 359 int ipf_end; /* Tail end offset (0 -> hard-case). */ 360 uint_t ipf_gen; /* Frag queue generation */ 361 size_t ipf_count; /* Count of bytes used by frag */ 362 uint_t ipf_nf_hdr_len; /* Length of nonfragmented header */ 363 in6_addr_t ipf_v6src; /* IPv6 source address */ 364 in6_addr_t ipf_v6dst; /* IPv6 dest address */ 365 uint_t ipf_prev_nexthdr_offset; /* Offset for nexthdr value */ 366 uint8_t ipf_ecn; /* ECN info for the fragments */ 367 uint8_t ipf_num_dups; /* Number of times dup frags recvd */ 368 uint16_t ipf_checksum_flags; /* Hardware checksum flags */ 369 uint32_t ipf_checksum; /* Partial checksum of fragment data */ 370 } ipf_t; 371 372 #define ipf_src V4_PART_OF_V6(ipf_v6src) 373 #define ipf_dst V4_PART_OF_V6(ipf_v6dst) 374 375 typedef enum { 376 IB_PKT = 0x01, 377 OB_PKT = 0x02 378 } ip_pkt_t; 379 380 #define UPDATE_IB_PKT_COUNT(ire)\ 381 { \ 382 (ire)->ire_ib_pkt_count++; \ 383 if ((ire)->ire_ipif != NULL) { \ 384 /* \ 385 * forwarding packet \ 386 */ \ 387 if ((ire)->ire_type & (IRE_LOCAL|IRE_BROADCAST)) \ 388 atomic_add_32(&(ire)->ire_ipif->ipif_ib_pkt_count, 1);\ 389 else \ 390 atomic_add_32(&(ire)->ire_ipif->ipif_fo_pkt_count, 1);\ 391 } \ 392 } 393 #define UPDATE_OB_PKT_COUNT(ire)\ 394 { \ 395 (ire)->ire_ob_pkt_count++;\ 396 if ((ire)->ire_ipif != NULL) { \ 397 atomic_add_32(&(ire)->ire_ipif->ipif_ob_pkt_count, 1); \ 398 } \ 399 } 400 401 402 #define IP_RPUT_LOCAL(q, mp, ipha, ire, recv_ill) \ 403 { \ 404 switch (ipha->ipha_protocol) { \ 405 case IPPROTO_UDP: \ 406 ip_udp_input(q, mp, ipha, ire, recv_ill); \ 407 break; \ 408 default: \ 409 ip_proto_input(q, mp, ipha, ire, recv_ill); \ 410 break; \ 411 } \ 412 } 413 414 415 #endif /* _KERNEL */ 416 417 /* ICMP types */ 418 #define ICMP_ECHO_REPLY 0 419 #define ICMP_DEST_UNREACHABLE 3 420 #define ICMP_SOURCE_QUENCH 4 421 #define ICMP_REDIRECT 5 422 #define ICMP_ECHO_REQUEST 8 423 #define ICMP_ROUTER_ADVERTISEMENT 9 424 #define ICMP_ROUTER_SOLICITATION 10 425 #define ICMP_TIME_EXCEEDED 11 426 #define ICMP_PARAM_PROBLEM 12 427 #define ICMP_TIME_STAMP_REQUEST 13 428 #define ICMP_TIME_STAMP_REPLY 14 429 #define ICMP_INFO_REQUEST 15 430 #define ICMP_INFO_REPLY 16 431 #define ICMP_ADDRESS_MASK_REQUEST 17 432 #define ICMP_ADDRESS_MASK_REPLY 18 433 434 /* ICMP_TIME_EXCEEDED codes */ 435 #define ICMP_TTL_EXCEEDED 0 436 #define ICMP_REASSEMBLY_TIME_EXCEEDED 1 437 438 /* ICMP_DEST_UNREACHABLE codes */ 439 #define ICMP_NET_UNREACHABLE 0 440 #define ICMP_HOST_UNREACHABLE 1 441 #define ICMP_PROTOCOL_UNREACHABLE 2 442 #define ICMP_PORT_UNREACHABLE 3 443 #define ICMP_FRAGMENTATION_NEEDED 4 444 #define ICMP_SOURCE_ROUTE_FAILED 5 445 #define ICMP_DEST_NET_UNKNOWN 6 446 #define ICMP_DEST_HOST_UNKNOWN 7 447 #define ICMP_SRC_HOST_ISOLATED 8 448 #define ICMP_DEST_NET_UNREACH_ADMIN 9 449 #define ICMP_DEST_HOST_UNREACH_ADMIN 10 450 #define ICMP_DEST_NET_UNREACH_TOS 11 451 #define ICMP_DEST_HOST_UNREACH_TOS 12 452 453 /* ICMP Header Structure */ 454 typedef struct icmph_s { 455 uint8_t icmph_type; 456 uint8_t icmph_code; 457 uint16_t icmph_checksum; 458 union { 459 struct { /* ECHO request/response structure */ 460 uint16_t u_echo_ident; 461 uint16_t u_echo_seqnum; 462 } u_echo; 463 struct { /* Destination unreachable structure */ 464 uint16_t u_du_zero; 465 uint16_t u_du_mtu; 466 } u_du; 467 struct { /* Parameter problem structure */ 468 uint8_t u_pp_ptr; 469 uint8_t u_pp_rsvd[3]; 470 } u_pp; 471 struct { /* Redirect structure */ 472 ipaddr_t u_rd_gateway; 473 } u_rd; 474 } icmph_u; 475 } icmph_t; 476 477 #define icmph_echo_ident icmph_u.u_echo.u_echo_ident 478 #define icmph_echo_seqnum icmph_u.u_echo.u_echo_seqnum 479 #define icmph_du_zero icmph_u.u_du.u_du_zero 480 #define icmph_du_mtu icmph_u.u_du.u_du_mtu 481 #define icmph_pp_ptr icmph_u.u_pp.u_pp_ptr 482 #define icmph_rd_gateway icmph_u.u_rd.u_rd_gateway 483 484 #define ICMPH_SIZE 8 485 486 /* 487 * Minimum length of transport layer header included in an ICMP error 488 * message for it to be considered valid. 489 */ 490 #define ICMP_MIN_TP_HDR_LEN 8 491 492 /* Aligned IP header */ 493 typedef struct ipha_s { 494 uint8_t ipha_version_and_hdr_length; 495 uint8_t ipha_type_of_service; 496 uint16_t ipha_length; 497 uint16_t ipha_ident; 498 uint16_t ipha_fragment_offset_and_flags; 499 uint8_t ipha_ttl; 500 uint8_t ipha_protocol; 501 uint16_t ipha_hdr_checksum; 502 ipaddr_t ipha_src; 503 ipaddr_t ipha_dst; 504 } ipha_t; 505 506 #define IPH_DF 0x4000 /* Don't fragment */ 507 #define IPH_MF 0x2000 /* More fragments to come */ 508 #define IPH_OFFSET 0x1FFF /* Where the offset lives */ 509 #define IPH_FRAG_HDR 0x8000 /* IPv6 don't fragment bit */ 510 511 /* ECN code points for IPv4 TOS byte and IPv6 traffic class octet. */ 512 #define IPH_ECN_NECT 0x0 /* Not ECN-Capabable Transport */ 513 #define IPH_ECN_ECT1 0x1 /* ECN-Capable Transport, ECT(1) */ 514 #define IPH_ECN_ECT0 0x2 /* ECN-Capable Transport, ECT(0) */ 515 #define IPH_ECN_CE 0x3 /* ECN-Congestion Experienced (CE) */ 516 517 /* IP Mac info structure */ 518 typedef struct ip_m_s { 519 t_uscalar_t ip_m_mac_type; /* From <sys/dlpi.h> */ 520 int ip_m_type; /* From <net/if_types.h> */ 521 boolean_t (*ip_m_v4mapinfo)(uint_t, uint8_t *, uint8_t *, 522 uint32_t *, ipaddr_t *); 523 boolean_t (*ip_m_v6mapinfo)(uint_t, uint8_t *, uint8_t *, 524 uint32_t *, in6_addr_t *); 525 boolean_t (*ip_m_v6intfid)(uint_t, uint8_t *, in6_addr_t *); 526 } ip_m_t; 527 528 /* 529 * The following functions attempt to reduce the link layer dependency 530 * of the IP stack. The current set of link specific operations are: 531 * a. map from IPv4 class D (224.0/4) multicast address range to the link 532 * layer multicast address range. 533 * b. map from IPv6 multicast address range (ff00::/8) to the link 534 * layer multicast address range. 535 * c. derive the default IPv6 interface identifier from the link layer 536 * address. 537 */ 538 #define MEDIA_V4MINFO(ip_m, plen, bphys, maddr, hwxp, v4ptr) \ 539 (((ip_m)->ip_m_v4mapinfo != NULL) && \ 540 (*(ip_m)->ip_m_v4mapinfo)(plen, bphys, maddr, hwxp, v4ptr)) 541 #define MEDIA_V6INTFID(ip_m, plen, phys, v6ptr) \ 542 (((ip_m)->ip_m_v6intfid != NULL) && \ 543 (*(ip_m)->ip_m_v6intfid)(plen, phys, v6ptr)) 544 #define MEDIA_V6MINFO(ip_m, plen, bphys, maddr, hwxp, v6ptr) \ 545 (((ip_m)->ip_m_v6mapinfo != NULL) && \ 546 (*(ip_m)->ip_m_v6mapinfo)(plen, bphys, maddr, hwxp, v6ptr)) 547 548 /* Router entry types */ 549 #define IRE_BROADCAST 0x0001 /* Route entry for broadcast address */ 550 #define IRE_DEFAULT 0x0002 /* Route entry for default gateway */ 551 #define IRE_LOCAL 0x0004 /* Route entry for local address */ 552 #define IRE_LOOPBACK 0x0008 /* Route entry for loopback address */ 553 #define IRE_PREFIX 0x0010 /* Route entry for prefix routes */ 554 #define IRE_CACHE 0x0020 /* Cached Route entry */ 555 #define IRE_IF_NORESOLVER 0x0040 /* Route entry for local interface */ 556 /* net without any address mapping. */ 557 #define IRE_IF_RESOLVER 0x0080 /* Route entry for local interface */ 558 /* net with resolver. */ 559 #define IRE_HOST 0x0100 /* Host route entry */ 560 #define IRE_HOST_REDIRECT 0x0200 /* Host route entry from redirects */ 561 562 /* 563 * IRE_MIPRTUN is only set on the ires in the ip_mrtun_table. 564 * This ire_type must not be set for ftable and ctable routing entries. 565 */ 566 #define IRE_MIPRTUN 0x0400 /* Reverse tunnel route entry */ 567 568 #define IRE_INTERFACE (IRE_IF_NORESOLVER | IRE_IF_RESOLVER) 569 #define IRE_OFFSUBNET (IRE_DEFAULT | IRE_PREFIX | IRE_HOST | \ 570 IRE_HOST_REDIRECT) 571 #define IRE_CACHETABLE (IRE_CACHE | IRE_BROADCAST | IRE_LOCAL | \ 572 IRE_LOOPBACK) 573 #define IRE_FORWARDTABLE (IRE_INTERFACE | IRE_OFFSUBNET) 574 575 /* 576 * If an IRE is marked with IRE_MARK_CONDEMNED, the last walker of 577 * the bucket should delete this IRE from this bucket. 578 */ 579 #define IRE_MARK_CONDEMNED 0x0001 580 /* 581 * If a broadcast IRE is marked with IRE_MARK_NORECV, ip_rput will drop the 582 * broadcast packets received on that interface. This is marked only 583 * on broadcast ires. Employed by IPMP, where we have multiple NICs on the 584 * same subnet receiving the same broadcast packet. 585 */ 586 #define IRE_MARK_NORECV 0x0002 587 /* 588 * IRE_CACHE marked this way won't be returned by ire_cache_lookup. Need 589 * to look specifically using MATCH_IRE_MARK_HIDDEN. Used by IPMP. 590 */ 591 #define IRE_MARK_HIDDEN 0x0004 /* Typically Used by in.mpathd */ 592 593 /* 594 * ire with IRE_MARK_NOADD is created in ip_newroute_ipif, when outgoing 595 * interface is specified by IP_XMIT_IF socket option. This ire is not 596 * added in IRE_CACHE. For example, this is used by mipagent to prevent 597 * any entry to be added in the cache table. We do not want to add any 598 * entry for a mobile-node in the routing table for foreign agent originated 599 * packets. Adding routes in cache table in this case, may run the risks of 600 * incorrect routing path in case of private overlapping addresses. 601 */ 602 #define IRE_MARK_NOADD 0x0008 /* Mark not to add ire in cache */ 603 604 /* 605 * IRE marked with IRE_MARK_TEMPORARY means that this IRE has been used 606 * either for forwarding a packet or has not been used for sending 607 * traffic on TCP connections terminated on this system. In both 608 * cases, this IRE is the first to go when IRE is being cleaned up. 609 */ 610 #define IRE_MARK_TEMPORARY 0x0010 611 612 /* 613 * IRE marked with IRE_MARK_USESRC_CHECK means that while adding an IRE with 614 * this mark, additional atomic checks need to be performed. For eg: by the 615 * time an IRE_CACHE is created, sent up to ARP and then comes back to IP; the 616 * usesrc grouping could have changed in which case we want to fail adding 617 * the IRE_CACHE entry 618 */ 619 #define IRE_MARK_USESRC_CHECK 0x0020 620 621 /* 622 * IRE_MARK_PRIVATE_ADDR is used for IP_NEXTHOP. When IP_NEXTHOP is set, the 623 * routing table lookup for the destination is bypassed and the packet is 624 * sent directly to the specified nexthop. The associated IRE_CACHE entries 625 * should be marked with IRE_MARK_PRIVATE_ADDR flag so that they don't show up 626 * in regular ire cache lookups. 627 */ 628 #define IRE_MARK_PRIVATE_ADDR 0x0040 629 630 /* Flags with ire_expire routine */ 631 #define FLUSH_ARP_TIME 0x0001 /* ARP info potentially stale timer */ 632 #define FLUSH_REDIRECT_TIME 0x0002 /* Redirects potentially stale */ 633 #define FLUSH_MTU_TIME 0x0004 /* Include path MTU per RFC 1191 */ 634 635 /* Arguments to ire_flush_cache() */ 636 #define IRE_FLUSH_DELETE 0 637 #define IRE_FLUSH_ADD 1 638 639 /* 640 * Open/close synchronization flags. 641 * These are kept in a separate field in the conn and the synchronization 642 * depends on the atomic 32 bit access to that field. 643 */ 644 #define CONN_CLOSING 0x01 /* ip_close waiting for ip_wsrv */ 645 #define CONN_IPSEC_LOAD_WAIT 0x02 /* waiting for load */ 646 #define CONN_CONDEMNED 0x04 /* conn is closing, no more refs */ 647 #define CONN_INCIPIENT 0x08 /* conn not yet visible, no refs */ 648 #define CONN_QUIESCED 0x10 /* conn is now quiescent */ 649 650 /* 651 * Parameter to ip_output giving the identity of the caller. 652 * IP_WSRV means the packet was enqueued in the STREAMS queue 653 * due to flow control and is now being reprocessed in the context of 654 * the STREAMS service procedure, consequent to flow control relief. 655 * IRE_SEND means the packet is being reprocessed consequent to an 656 * ire cache creation and addition and this may or may not be happening 657 * in the service procedure context. Anything other than the above 2 658 * cases is identified as IP_WPUT. Most commonly this is the case of 659 * packets coming down from the application. 660 */ 661 #ifdef _KERNEL 662 #define IP_WSRV 1 /* Called from ip_wsrv */ 663 #define IP_WPUT 2 /* Called from ip_wput */ 664 #define IRE_SEND 3 /* Called from ire_send */ 665 666 /* 667 * Extra structures need for per-src-addr filtering (IGMPv3/MLDv2) 668 */ 669 #define MAX_FILTER_SIZE 64 670 671 typedef struct slist_s { 672 int sl_numsrc; 673 in6_addr_t sl_addr[MAX_FILTER_SIZE]; 674 } slist_t; 675 676 /* 677 * Following struct is used to maintain retransmission state for 678 * a multicast group. One rtx_state_t struct is an in-line field 679 * of the ilm_t struct; the slist_ts in the rtx_state_t struct are 680 * alloc'd as needed. 681 */ 682 typedef struct rtx_state_s { 683 uint_t rtx_timer; /* retrans timer */ 684 int rtx_cnt; /* retrans count */ 685 int rtx_fmode_cnt; /* retrans count for fmode change */ 686 slist_t *rtx_allow; 687 slist_t *rtx_block; 688 } rtx_state_t; 689 690 /* 691 * Used to construct list of multicast address records that will be 692 * sent in a single listener report. 693 */ 694 typedef struct mrec_s { 695 struct mrec_s *mrec_next; 696 uint8_t mrec_type; 697 uint8_t mrec_auxlen; /* currently unused */ 698 in6_addr_t mrec_group; 699 slist_t mrec_srcs; 700 } mrec_t; 701 702 /* Group membership list per upper conn */ 703 /* 704 * XXX add ilg info for ifaddr/ifindex. 705 * XXX can we make ilg survive an ifconfig unplumb + plumb 706 * by setting the ipif/ill to NULL and recover that later? 707 * 708 * ilg_ipif is used by IPv4 as multicast groups are joined using an interface 709 * address (ipif). 710 * ilg_ill is used by IPv6 as multicast groups are joined using an interface 711 * index (phyint->phyint_ifindex). 712 * ilg_ill is NULL for IPv4 and ilg_ipif is NULL for IPv6. 713 * 714 * ilg records the state of multicast memberships of a socket end point. 715 * ilm records the state of multicast memberships with the driver and is 716 * maintained per interface. 717 * 718 * Notes : 719 * 720 * 1) There is no direct link between a given ilg and ilm. If the 721 * application has joined a group G with ifindex I, we will have 722 * an ilg with ilg_v6group and ilg_ill. There will be a corresponding 723 * ilm with ilm_ill/ilm_v6addr recording the multicast membership. 724 * To delete the membership, 725 * 726 * a) Search for ilg matching on G and I with ilg_v6group 727 * and ilg_ill. Delete ilg_ill. 728 * b) Search the corresponding ilm matching on G and I with 729 * ilm_v6addr and ilm_ill. Delete ilm. 730 * 731 * In IPv4, the only difference is, we look using ipifs instead of 732 * ills. 733 * 734 * 2) With IP multipathing, we want to keep receiving even after the 735 * interface has failed. We do this by moving multicast memberships 736 * to a new_ill within the group. This is acheived by sending 737 * DL_DISABMULTI_REQS on ilg_ill/ilm_ill and sending DL_ENABMULTIREQS 738 * on the new_ill and changing ilg_ill/ilm_ill to new_ill. But, we 739 * need to be able to delete memberships which will still come down 740 * with the ifindex of the old ill which is what the application 741 * knows of. Thus we store the ilm_/ilg_orig_ifindex to keep track 742 * of where we joined initially so that we can lookup even after we 743 * moved the membership. It is also used for moving back the membership 744 * when the old ill has been repaired. This is done by looking up for 745 * ilms with ilm_orig_ifindex matching on the old ill's ifindex. Only 746 * ilms actually move from old ill to new ill. ilgs don't move (just 747 * the ilg_ill is changed when it moves) as it just records the state 748 * of the application that has joined a group G where as ilm records 749 * the state joined with the driver. Thus when we send DL_XXXMULTI_REQs 750 * we also need to keep the ilm in the right ill. 751 * 752 * In IPv4, as ipifs move from old ill to new_ill, ilgs and ilms move 753 * implicitly as we use only ipifs in IPv4. Thus, one can always lookup 754 * a given ilm/ilg even after it fails without the support of 755 * orig_ifindex. We move ilms still to record the driver state as 756 * mentioned above. 757 */ 758 759 /* 760 * The ilg_t and ilm_t members are protected by ipsq. They can be changed only 761 * by a thread executing in the ipsq. In other words add/delete of a 762 * multicast group has to execute in the ipsq. 763 */ 764 #define ILG_DELETED 0x1 /* ilg_flags */ 765 typedef struct ilg_s { 766 in6_addr_t ilg_v6group; 767 struct ipif_s *ilg_ipif; /* Logical interface we are member on */ 768 struct ill_s *ilg_ill; /* Used by IPv6 */ 769 int ilg_orig_ifindex; /* Interface originally joined on */ 770 uint_t ilg_flags; 771 mcast_record_t ilg_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */ 772 slist_t *ilg_filter; 773 } ilg_t; 774 775 776 /* 777 * Multicast address list entry for lower ill. 778 * ilm_ipif is used by IPv4 as multicast groups are joined using ipif. 779 * ilm_ill is used by IPv6 as multicast groups are joined using ill. 780 * ilm_ill is NULL for IPv4 and ilm_ipif is NULL for IPv6. 781 */ 782 #define ILM_DELETED 0x1 /* ilm_flags */ 783 typedef struct ilm_s { 784 in6_addr_t ilm_v6addr; 785 int ilm_refcnt; 786 uint_t ilm_timer; /* IGMP/MLD query resp timer, in msec */ 787 struct ipif_s *ilm_ipif; /* Back pointer to ipif for IPv4 */ 788 struct ilm_s *ilm_next; /* Linked list for each ill */ 789 uint_t ilm_state; /* state of the membership */ 790 struct ill_s *ilm_ill; /* Back pointer to ill for IPv6 */ 791 int ilm_orig_ifindex; /* V6_MULTICAST_IF/ilm_ipif index */ 792 uint_t ilm_flags; 793 boolean_t ilm_is_new; /* new ilm */ 794 boolean_t ilm_notify_driver; /* Need to notify the driver */ 795 zoneid_t ilm_zoneid; 796 int ilm_no_ilg_cnt; /* number of joins w/ no ilg */ 797 mcast_record_t ilm_fmode; /* MODE_IS_INCLUDE/MODE_IS_EXCLUDE */ 798 slist_t *ilm_filter; /* source filter list */ 799 slist_t *ilm_pendsrcs; /* relevant src addrs for pending req */ 800 rtx_state_t ilm_rtx; /* SCR retransmission state */ 801 } ilm_t; 802 803 #define ilm_addr V4_PART_OF_V6(ilm_v6addr) 804 805 /* 806 * ilm_walker_cleanup needs to execute when the ilm_walker_cnt goes down to 807 * zero. In addition it needs to block new walkers while it is unlinking ilm's 808 * from the list. Thus simple atomics for the ill_ilm_walker_cnt don't suffice. 809 */ 810 #define ILM_WALKER_HOLD(ill) { \ 811 mutex_enter(&(ill)->ill_lock); \ 812 ill->ill_ilm_walker_cnt++; \ 813 mutex_exit(&(ill)->ill_lock); \ 814 } 815 816 #define ILM_WALKER_RELE(ill) { \ 817 mutex_enter(&(ill)->ill_lock); \ 818 (ill)->ill_ilm_walker_cnt--; \ 819 if ((ill)->ill_ilm_walker_cnt == 0 && (ill)->ill_ilm_cleanup_reqd) \ 820 ilm_walker_cleanup(ill); \ 821 mutex_exit(&(ill)->ill_lock); \ 822 } 823 824 /* 825 * Soft reference to an IPsec SA. 826 * 827 * On relative terms, conn's can be persistant (living as long as the 828 * processes which create them), while SA's are ephemeral (dying when 829 * they hit their time-based or byte-based lifetimes). 830 * 831 * We could hold a hard reference to an SA from an ipsec_latch_t, 832 * but this would cause expired SA's to linger for a potentially 833 * unbounded time. 834 * 835 * Instead, we remember the hash bucket number and bucket generation 836 * in addition to the pointer. The bucket generation is incremented on 837 * each deletion. 838 */ 839 typedef struct ipsa_ref_s 840 { 841 struct ipsa_s *ipsr_sa; 842 struct isaf_s *ipsr_bucket; 843 uint64_t ipsr_gen; 844 } ipsa_ref_t; 845 846 /* 847 * IPsec "latching" state. 848 * 849 * In the presence of IPsec policy, fully-bound conn's bind a connection 850 * to more than just the 5-tuple, but also a specific IPsec action and 851 * identity-pair. 852 * 853 * As an optimization, we also cache soft references to IPsec SA's 854 * here so that we can fast-path around most of the work needed for 855 * outbound IPsec SA selection. 856 * 857 * Were it not for TCP's detached connections, this state would be 858 * in-line in conn_t; instead, this is in a separate structure so it 859 * can be handed off to TCP when a connection is detached. 860 */ 861 typedef struct ipsec_latch_s 862 { 863 kmutex_t ipl_lock; 864 uint32_t ipl_refcnt; 865 866 uint64_t ipl_unique; 867 struct ipsec_policy_s *ipl_in_policy; /* latched policy (in) */ 868 struct ipsec_policy_s *ipl_out_policy; /* latched policy (out) */ 869 struct ipsec_action_s *ipl_in_action; /* latched action (in) */ 870 struct ipsec_action_s *ipl_out_action; /* latched action (out) */ 871 cred_t *ipl_local_id; 872 struct ipsid_s *ipl_local_cid; 873 struct ipsid_s *ipl_remote_cid; 874 unsigned int 875 ipl_out_action_latched : 1, 876 ipl_in_action_latched : 1, 877 ipl_out_policy_latched : 1, 878 ipl_in_policy_latched : 1, 879 880 ipl_ids_latched : 1, 881 882 ipl_pad_to_bit_31 : 27; 883 884 ipsa_ref_t ipl_ref[2]; /* 0: ESP, 1: AH */ 885 886 } ipsec_latch_t; 887 888 #define IPLATCH_REFHOLD(ipl) { \ 889 atomic_add_32(&(ipl)->ipl_refcnt, 1); \ 890 ASSERT((ipl)->ipl_refcnt != 0); \ 891 } 892 893 #define IPLATCH_REFRELE(ipl) { \ 894 ASSERT((ipl)->ipl_refcnt != 0); \ 895 membar_exit(); \ 896 if (atomic_add_32_nv(&(ipl)->ipl_refcnt, -1) == 0) \ 897 iplatch_free(ipl); \ 898 } 899 900 /* 901 * peer identity structure. 902 */ 903 904 typedef struct conn_s conn_t; 905 906 /* 907 * The old IP client structure "ipc_t" is gone. All the data is stored in the 908 * connection structure "conn_t" now. The mapping of old and new fields looks 909 * like this: 910 * 911 * ipc_ulp conn_ulp 912 * ipc_rq conn_rq 913 * ipc_wq conn_wq 914 * 915 * ipc_laddr conn_src 916 * ipc_faddr conn_rem 917 * ipc_v6laddr conn_srcv6 918 * ipc_v6faddr conn_remv6 919 * 920 * ipc_lport conn_lport 921 * ipc_fport conn_fport 922 * ipc_ports conn_ports 923 * 924 * ipc_policy conn_policy 925 * ipc_latch conn_latch 926 * 927 * ipc_irc_lock conn_lock 928 * ipc_ire_cache conn_ire_cache 929 * 930 * ipc_state_flags conn_state_flags 931 * ipc_outgoing_ill conn_outgoing_ill 932 * 933 * ipc_dontroute conn_dontroute 934 * ipc_loopback conn_loopback 935 * ipc_broadcast conn_broadcast 936 * ipc_reuseaddr conn_reuseaddr 937 * 938 * ipc_multicast_loop conn_multicast_loop 939 * ipc_multi_router conn_multi_router 940 * ipc_priv_stream conn_priv_stream 941 * ipc_draining conn_draining 942 * 943 * ipc_did_putbq conn_did_putbq 944 * ipc_unspec_src conn_unspec_src 945 * ipc_policy_cached conn_policy_cached 946 * 947 * ipc_in_enforce_policy conn_in_enforce_policy 948 * ipc_out_enforce_policy conn_out_enforce_policy 949 * ipc_af_isv6 conn_af_isv6 950 * ipc_pkt_isv6 conn_pkt_isv6 951 * 952 * ipc_ipv6_recvpktinfo conn_ipv6_recvpktinfo 953 * 954 * ipc_ipv6_recvhoplimit conn_ipv6_recvhoplimit 955 * ipc_ipv6_recvhopopts conn_ipv6_recvhopopts 956 * ipc_ipv6_recvdstopts conn_ipv6_recvdstopts 957 * 958 * ipc_ipv6_recvrthdr conn_ipv6_recvrthdr 959 * ipc_ipv6_recvrtdstopts conn_ipv6_recvrtdstopts 960 * ipc_fully_bound conn_fully_bound 961 * 962 * ipc_recvif conn_recvif 963 * 964 * ipc_recvslla conn_recvslla 965 * ipc_acking_unbind conn_acking_unbind 966 * ipc_pad_to_bit_31 conn_pad_to_bit_31 967 * 968 * ipc_xmit_if_ill conn_xmit_if_ill 969 * ipc_nofailover_ill conn_nofailover_ill 970 * 971 * ipc_proto conn_proto 972 * ipc_incoming_ill conn_incoming_ill 973 * ipc_outgoing_pill conn_outgoing_pill 974 * ipc_pending_ill conn_pending_ill 975 * ipc_unbind_mp conn_unbind_mp 976 * ipc_ilg conn_ilg 977 * ipc_ilg_allocated conn_ilg_allocated 978 * ipc_ilg_inuse conn_ilg_inuse 979 * ipc_ilg_walker_cnt conn_ilg_walker_cnt 980 * ipc_refcv conn_refcv 981 * ipc_multicast_ipif conn_multicast_ipif 982 * ipc_multicast_ill conn_multicast_ill 983 * ipc_orig_bound_ifindex conn_orig_bound_ifindex 984 * ipc_orig_multicast_ifindex conn_orig_multicast_ifindex 985 * ipc_orig_xmit_ifindex conn_orig_xmit_ifindex 986 * ipc_drain_next conn_drain_next 987 * ipc_drain_prev conn_drain_prev 988 * ipc_idl conn_idl 989 */ 990 991 /* 992 * This is used to match an inbound/outbound datagram with 993 * policy. 994 */ 995 996 typedef struct ipsec_selector { 997 in6_addr_t ips_local_addr_v6; 998 in6_addr_t ips_remote_addr_v6; 999 uint16_t ips_local_port; 1000 uint16_t ips_remote_port; 1001 uint8_t ips_icmp_type; 1002 uint8_t ips_icmp_code; 1003 uint8_t ips_protocol; 1004 uint8_t ips_isv4 : 1, 1005 ips_is_icmp_inv_acq: 1; 1006 } ipsec_selector_t; 1007 1008 /* 1009 * Note that we put v4 addresses in the *first* 32-bit word of the 1010 * selector rather than the last to simplify the prefix match/mask code 1011 * in spd.c 1012 */ 1013 #define ips_local_addr_v4 ips_local_addr_v6.s6_addr32[0] 1014 #define ips_remote_addr_v4 ips_remote_addr_v6.s6_addr32[0] 1015 1016 /* Values used in IP by IPSEC Code */ 1017 #define IPSEC_OUTBOUND B_TRUE 1018 #define IPSEC_INBOUND B_FALSE 1019 1020 /* 1021 * There are two variants in policy failures. The packet may come in 1022 * secure when not needed (IPSEC_POLICY_???_NOT_NEEDED) or it may not 1023 * have the desired level of protection (IPSEC_POLICY_MISMATCH). 1024 */ 1025 #define IPSEC_POLICY_NOT_NEEDED 0 1026 #define IPSEC_POLICY_MISMATCH 1 1027 #define IPSEC_POLICY_AUTH_NOT_NEEDED 2 1028 #define IPSEC_POLICY_ENCR_NOT_NEEDED 3 1029 #define IPSEC_POLICY_SE_NOT_NEEDED 4 1030 #define IPSEC_POLICY_MAX 5 /* Always max + 1. */ 1031 1032 /* 1033 * Folowing macro is used whenever the code does not know whether there 1034 * is a M_CTL present in the front and it needs to examine the actual mp 1035 * i.e the IP header. As a M_CTL message could be in the front, this 1036 * extracts the packet into mp and the M_CTL mp into first_mp. If M_CTL 1037 * mp is not present, both first_mp and mp point to the same message. 1038 */ 1039 #define EXTRACT_PKT_MP(mp, first_mp, mctl_present) \ 1040 (first_mp) = (mp); \ 1041 if ((mp)->b_datap->db_type == M_CTL) { \ 1042 (mp) = (mp)->b_cont; \ 1043 (mctl_present) = B_TRUE; \ 1044 } else { \ 1045 (mctl_present) = B_FALSE; \ 1046 } 1047 1048 /* 1049 * Check with IPSEC inbound policy if 1050 * 1051 * 1) per-socket policy is present - indicated by conn_in_enforce_policy. 1052 * 2) Or if we have not cached policy on the conn and the global policy is 1053 * non-empty. 1054 */ 1055 #define CONN_INBOUND_POLICY_PRESENT(connp) \ 1056 ((connp)->conn_in_enforce_policy || \ 1057 (!((connp)->conn_policy_cached) && \ 1058 ipsec_inbound_v4_policy_present)) 1059 1060 #define CONN_INBOUND_POLICY_PRESENT_V6(connp) \ 1061 ((connp)->conn_in_enforce_policy || \ 1062 (!(connp)->conn_policy_cached && \ 1063 ipsec_inbound_v6_policy_present)) 1064 1065 #define CONN_OUTBOUND_POLICY_PRESENT(connp) \ 1066 ((connp)->conn_out_enforce_policy || \ 1067 (!((connp)->conn_policy_cached) && \ 1068 ipsec_outbound_v4_policy_present)) 1069 1070 #define CONN_OUTBOUND_POLICY_PRESENT_V6(connp) \ 1071 ((connp)->conn_out_enforce_policy || \ 1072 (!(connp)->conn_policy_cached && \ 1073 ipsec_outbound_v6_policy_present)) 1074 1075 /* 1076 * Information cached in IRE for upper layer protocol (ULP). 1077 * 1078 * Notice that ire_max_frag is not included in the iulp_t structure, which 1079 * it may seem that it should. But ire_max_frag cannot really be cached. It 1080 * is fixed for each interface. For MTU found by PMTUd, we may want to cache 1081 * it. But currently, we do not do that. 1082 */ 1083 typedef struct iulp_s { 1084 boolean_t iulp_set; /* Is any metric set? */ 1085 uint32_t iulp_ssthresh; /* Slow start threshold (TCP). */ 1086 clock_t iulp_rtt; /* Guestimate in millisecs. */ 1087 clock_t iulp_rtt_sd; /* Cached value of RTT variance. */ 1088 uint32_t iulp_spipe; /* Send pipe size. */ 1089 uint32_t iulp_rpipe; /* Receive pipe size. */ 1090 uint32_t iulp_rtomax; /* Max round trip timeout. */ 1091 uint32_t iulp_sack; /* Use SACK option (TCP)? */ 1092 uint32_t 1093 iulp_tstamp_ok : 1, /* Use timestamp option (TCP)? */ 1094 iulp_wscale_ok : 1, /* Use window scale option (TCP)? */ 1095 iulp_ecn_ok : 1, /* Enable ECN (for TCP)? */ 1096 iulp_pmtud_ok : 1, /* Enable PMTUd? */ 1097 1098 iulp_not_used : 28; 1099 } iulp_t; 1100 1101 /* Zero iulp_t. */ 1102 extern const iulp_t ire_uinfo_null; 1103 1104 /* 1105 * The conn drain list structure. 1106 * The list is protected by idl_lock. Each conn_t inserted in the list 1107 * points back at this idl_t using conn_idl. IP primes the draining of the 1108 * conns queued in these lists, by qenabling the 1st conn of each list. This 1109 * occurs when STREAMS backenables ip_wsrv on the IP module. Each conn instance 1110 * of ip_wsrv successively qenables the next conn in the list. 1111 * idl_lock protects all other members of idl_t and conn_drain_next 1112 * and conn_drain_prev of conn_t. The conn_lock protects IPCF_DRAIN_DISABLED 1113 * flag of the conn_t and conn_idl. 1114 */ 1115 typedef struct idl_s { 1116 conn_t *idl_conn; /* Head of drain list */ 1117 kmutex_t idl_lock; /* Lock for this list */ 1118 conn_t *idl_conn_draining; /* conn that is draining */ 1119 uint32_t 1120 idl_repeat : 1, /* Last conn must re-enable */ 1121 /* drain list again */ 1122 idl_unused : 31; 1123 } idl_t; 1124 1125 #define CONN_DRAIN_LIST_LOCK(connp) (&((connp)->conn_idl->idl_lock)) 1126 /* 1127 * Interface route structure which holds the necessary information to recreate 1128 * routes that are tied to an interface (namely where ire_ipif != NULL). 1129 * These routes which were initially created via a routing socket or via the 1130 * SIOCADDRT ioctl may be gateway routes (RTF_GATEWAY being set) or may be 1131 * traditional interface routes. When an interface comes back up after being 1132 * marked down, this information will be used to recreate the routes. These 1133 * are part of an mblk_t chain that hangs off of the IPIF (ipif_saved_ire_mp). 1134 */ 1135 typedef struct ifrt_s { 1136 ushort_t ifrt_type; /* Type of IRE */ 1137 in6_addr_t ifrt_v6addr; /* Address IRE represents. */ 1138 in6_addr_t ifrt_v6gateway_addr; /* Gateway if IRE_OFFSUBNET */ 1139 in6_addr_t ifrt_v6src_addr; /* Src addr if RTF_SETSRC */ 1140 in6_addr_t ifrt_v6mask; /* Mask for matching IRE. */ 1141 uint32_t ifrt_flags; /* flags related to route */ 1142 uint_t ifrt_max_frag; /* MTU (next hop or path). */ 1143 iulp_t ifrt_iulp_info; /* Cached IRE ULP info. */ 1144 } ifrt_t; 1145 1146 #define ifrt_addr V4_PART_OF_V6(ifrt_v6addr) 1147 #define ifrt_gateway_addr V4_PART_OF_V6(ifrt_v6gateway_addr) 1148 #define ifrt_src_addr V4_PART_OF_V6(ifrt_v6src_addr) 1149 #define ifrt_mask V4_PART_OF_V6(ifrt_v6mask) 1150 1151 /* Number of IP addresses that can be hosted on a physical interface */ 1152 #define MAX_ADDRS_PER_IF 8192 1153 /* 1154 * Number of Source addresses to be considered for source address 1155 * selection. Used by ipif_select_source[_v6]. 1156 */ 1157 #define MAX_IPIF_SELECT_SOURCE 50 1158 1159 #ifdef IP_DEBUG 1160 /* 1161 * Tracing refholds and refreleases for debugging. Existing tracing mechanisms 1162 * do not allow the granularity need to trace refrences to ipif/ill/ire's. This 1163 * mechanism should be revisited once dtrace is available. 1164 */ 1165 #define IP_STACK_DEPTH 15 1166 typedef struct tr_buf_s { 1167 int tr_depth; 1168 pc_t tr_stack[IP_STACK_DEPTH]; 1169 } tr_buf_t; 1170 1171 typedef struct th_trace_s { 1172 struct th_trace_s *th_next; 1173 struct th_trace_s **th_prev; 1174 kthread_t *th_id; 1175 int th_refcnt; 1176 uint_t th_trace_lastref; 1177 #define TR_BUF_MAX 38 1178 tr_buf_t th_trbuf[TR_BUF_MAX]; 1179 } th_trace_t; 1180 #endif 1181 1182 /* The following are ipif_state_flags */ 1183 #define IPIF_CONDEMNED 0x1 /* The ipif is being removed */ 1184 #define IPIF_CHANGING 0x2 /* A critcal ipif field is changing */ 1185 #define IPIF_MOVING 0x8 /* The ipif is being moved */ 1186 #define IPIF_SET_LINKLOCAL 0x10 /* transient flag during bringup */ 1187 #define IPIF_ZERO_SOURCE 0x20 /* transient flag during bringup */ 1188 1189 /* IP interface structure, one per local address */ 1190 typedef struct ipif_s { 1191 struct ipif_s *ipif_next; 1192 struct ill_s *ipif_ill; /* Back pointer to our ill */ 1193 int ipif_id; /* Logical unit number */ 1194 uint_t ipif_mtu; /* Starts at ipif_ill->ill_max_frag */ 1195 uint_t ipif_saved_mtu; /* Save of mtu during ipif_move() */ 1196 in6_addr_t ipif_v6lcl_addr; /* Local IP address for this if. */ 1197 in6_addr_t ipif_v6src_addr; /* Source IP address for this if. */ 1198 in6_addr_t ipif_v6subnet; /* Subnet prefix for this if. */ 1199 in6_addr_t ipif_v6net_mask; /* Net mask for this interface. */ 1200 in6_addr_t ipif_v6brd_addr; /* Broadcast addr for this interface. */ 1201 in6_addr_t ipif_v6pp_dst_addr; /* Point-to-point dest address. */ 1202 uint64_t ipif_flags; /* Interface flags. */ 1203 uint_t ipif_metric; /* BSD if metric, for compatibility. */ 1204 uint_t ipif_ire_type; /* IRE_LOCAL or IRE_LOOPBACK */ 1205 mblk_t *ipif_arp_del_mp; /* Allocated at time arp comes up, to */ 1206 /* prevent awkward out of mem */ 1207 /* condition later */ 1208 mblk_t *ipif_saved_ire_mp; /* Allocated for each extra */ 1209 /* IRE_IF_NORESOLVER/IRE_IF_RESOLVER */ 1210 /* on this interface so that they */ 1211 /* can survive ifconfig down. */ 1212 kmutex_t ipif_saved_ire_lock; /* Protects ipif_saved_ire_mp */ 1213 1214 mrec_t *ipif_igmp_rpt; /* List of group memberships which */ 1215 /* will be reported on. Used when */ 1216 /* handling an igmp timeout. */ 1217 1218 /* 1219 * The packet counts in the ipif contain the sum of the 1220 * packet counts in dead IREs that were affiliated with 1221 * this ipif. 1222 */ 1223 uint_t ipif_fo_pkt_count; /* Forwarded thru our dead IREs */ 1224 uint_t ipif_ib_pkt_count; /* Inbound packets for our dead IREs */ 1225 uint_t ipif_ob_pkt_count; /* Outbound packets to our dead IREs */ 1226 /* Exclusive bit fields, protected by ipsq_t */ 1227 unsigned int 1228 ipif_multicast_up : 1, /* We have joined the allhosts group */ 1229 ipif_solmcast_up : 1, /* We joined solicited node mcast */ 1230 ipif_replace_zero : 1, /* Replacement for zero */ 1231 ipif_was_up : 1, /* ipif was up before */ 1232 1233 ipif_pad_to_31 : 28; 1234 1235 int ipif_orig_ifindex; /* ifindex before SLIFFAILOVER */ 1236 uint_t ipif_seqid; /* unique index across all ills */ 1237 uint_t ipif_orig_ipifid; /* ipif_id before SLIFFAILOVER */ 1238 uint_t ipif_state_flags; /* See IPIF_* flag defs above */ 1239 uint_t ipif_refcnt; /* active consistent reader cnt */ 1240 uint_t ipif_ire_cnt; /* Number of ire's referencing ipif */ 1241 uint_t ipif_saved_ire_cnt; 1242 zoneid_t 1243 ipif_zoneid; /* zone ID number */ 1244 #ifdef ILL_DEBUG 1245 #define IP_TR_HASH_MAX 64 1246 th_trace_t *ipif_trace[IP_TR_HASH_MAX]; 1247 boolean_t ipif_trace_disable; /* True when alloc fails */ 1248 #endif 1249 } ipif_t; 1250 1251 /* 1252 * The following table lists the protection levels of the various members 1253 * of the ipif_t. The following notation is used. 1254 * 1255 * Write once - Written to only once at the time of bringing up 1256 * the interface and can be safely read after the bringup without any lock. 1257 * 1258 * ipsq - Need to execute in the ipsq to perform the indicated access. 1259 * 1260 * ill_lock - Need to hold this mutex to perform the indicated access. 1261 * 1262 * ill_g_lock - Need to hold this rw lock as reader/writer for read access or 1263 * write access respectively. 1264 * 1265 * down ill - Written to only when the ill is down (i.e all ipifs are down) 1266 * up ill - Read only when the ill is up (i.e. at least 1 ipif is up) 1267 * 1268 * Table of ipif_t members and their protection 1269 * 1270 * ipif_next ill_g_lock ill_g_lock 1271 * ipif_ill ipsq + down ipif write once 1272 * ipif_id ipsq + down ipif write once 1273 * ipif_mtu ipsq 1274 * ipif_v6lcl_addr ipsq + down ipif up ipif 1275 * ipif_v6src_addr ipsq + down ipif up ipif 1276 * ipif_v6subnet ipsq + down ipif up ipif 1277 * ipif_v6net_mask ipsq + down ipif up ipif 1278 * 1279 * ipif_v6brd_addr 1280 * ipif_v6pp_dst_addr 1281 * ipif_flags ill_lock ill_lock 1282 * ipif_metric 1283 * ipif_ire_type ipsq + down ill up ill 1284 * 1285 * ipif_arp_del_mp ipsq ipsq 1286 * ipif_saved_ire_mp ipif_saved_ire_lock ipif_saved_ire_lock 1287 * ipif_igmp_rpt ipsq ipsq 1288 * 1289 * ipif_fo_pkt_count Approx 1290 * ipif_ib_pkt_count Approx 1291 * ipif_ob_pkt_count Approx 1292 * 1293 * bit fields ill_lock ill_lock 1294 * 1295 * ipif_orig_ifindex ipsq None 1296 * ipif_orig_ipifid ipsq None 1297 * ipif_seqid ipsq Write once 1298 * 1299 * ipif_state_flags ill_lock ill_lock 1300 * ipif_refcnt ill_lock ill_lock 1301 * ipif_ire_cnt ill_lock ill_lock 1302 * ipif_saved_ire_cnt 1303 */ 1304 1305 #define IP_TR_HASH(tid) ((((uintptr_t)tid) >> 6) & (IP_TR_HASH_MAX - 1)) 1306 1307 #ifdef ILL_DEBUG 1308 #define IPIF_TRACE_REF(ipif) ipif_trace_ref(ipif) 1309 #define ILL_TRACE_REF(ill) ill_trace_ref(ill) 1310 #define IPIF_UNTRACE_REF(ipif) ipif_untrace_ref(ipif) 1311 #define ILL_UNTRACE_REF(ill) ill_untrace_ref(ill) 1312 #define ILL_TRACE_CLEANUP(ill) ill_trace_cleanup(ill) 1313 #define IPIF_TRACE_CLEANUP(ipif) ipif_trace_cleanup(ipif) 1314 #else 1315 #define IPIF_TRACE_REF(ipif) 1316 #define ILL_TRACE_REF(ill) 1317 #define IPIF_UNTRACE_REF(ipif) 1318 #define ILL_UNTRACE_REF(ill) 1319 #define ILL_TRACE_CLEANUP(ill) 1320 #define IPIF_TRACE_CLEANUP(ipif) 1321 #endif 1322 1323 /* IPv4 compatability macros */ 1324 #define ipif_lcl_addr V4_PART_OF_V6(ipif_v6lcl_addr) 1325 #define ipif_src_addr V4_PART_OF_V6(ipif_v6src_addr) 1326 #define ipif_subnet V4_PART_OF_V6(ipif_v6subnet) 1327 #define ipif_net_mask V4_PART_OF_V6(ipif_v6net_mask) 1328 #define ipif_brd_addr V4_PART_OF_V6(ipif_v6brd_addr) 1329 #define ipif_pp_dst_addr V4_PART_OF_V6(ipif_v6pp_dst_addr) 1330 1331 /* Macros for easy backreferences to the ill. */ 1332 #define ipif_wq ipif_ill->ill_wq 1333 #define ipif_rq ipif_ill->ill_rq 1334 #define ipif_net_type ipif_ill->ill_net_type 1335 #define ipif_resolver_mp ipif_ill->ill_resolver_mp 1336 #define ipif_ipif_up_count ipif_ill->ill_ipif_up_count 1337 #define ipif_bcast_mp ipif_ill->ill_bcast_mp 1338 #define ipif_type ipif_ill->ill_type 1339 #define ipif_isv6 ipif_ill->ill_isv6 1340 1341 #define SIOCLIFADDR_NDX 112 /* ndx of SIOCLIFADDR in the ndx ioctl table */ 1342 1343 /* 1344 * mode value for ip_ioctl_finish for finishing an ioctl 1345 */ 1346 #define CONN_CLOSE 1 /* No mi_copy */ 1347 #define COPYOUT 2 /* do an mi_copyout if needed */ 1348 #define NO_COPYOUT 3 /* do an mi_copy_done */ 1349 1350 /* 1351 * The IP-MT design revolves around the serialization object ipsq_t. 1352 * It is associated with an IPMP group. If IPMP is not enabled, there is 1353 * 1 ipsq_t per phyint. Eg. an ipsq_t would cover both hme0's IPv4 stream 1354 * 1355 * ipsq_lock protects 1356 * ipsq_reentry_cnt, ipsq_writer, ipsq_xopq_mphead, ipsq_xopq_mptail, 1357 * ipsq_mphead, ipsq_mptail, ipsq_split 1358 * 1359 * ipsq_pending_ipif, ipsq_current_ipif, ipsq_pending_mp, ipsq_flags, 1360 * ipsq_waitfor 1361 * 1362 * The fields in the last line above below are set mostly by a writer thread 1363 * But there is an exception in the last call to ipif_ill_refrele_tail which 1364 * could also race with a conn close which could be cleaning up the 1365 * fields. So we choose to protect using ipsq_lock instead of depending on 1366 * the property of the writer. 1367 * ill_g_lock protects 1368 * ipsq_refs, ipsq_phyint_list 1369 */ 1370 typedef struct ipsq_s { 1371 kmutex_t ipsq_lock; 1372 int ipsq_reentry_cnt; 1373 kthread_t *ipsq_writer; /* current owner (thread id) */ 1374 int ipsq_flags; 1375 mblk_t *ipsq_xopq_mphead; /* list of excl ops mostly ioctls */ 1376 mblk_t *ipsq_xopq_mptail; 1377 mblk_t *ipsq_mphead; /* msgs on ipsq linked thru b_next */ 1378 mblk_t *ipsq_mptail; /* msgs on ipsq linked thru b_next */ 1379 ipif_t *ipsq_pending_ipif; /* ipif associated w. ipsq_pending_mp */ 1380 ipif_t *ipsq_current_ipif; /* ipif associated with current ioctl */ 1381 mblk_t *ipsq_pending_mp; /* current ioctl mp while waiting for */ 1382 /* response from another module */ 1383 struct ipsq_s *ipsq_next; /* list of all syncq's (ipsq_g_list) */ 1384 uint_t ipsq_refs; /* Number of phyints on this ipsq */ 1385 struct phyint *ipsq_phyint_list; /* List of phyints on this ipsq */ 1386 boolean_t ipsq_split; /* ipsq may need to be split */ 1387 int ipsq_waitfor; /* Values encoded below */ 1388 char ipsq_name[LIFNAMSIZ+1]; /* same as phyint_groupname */ 1389 int ipsq_last_cmd; /* debugging aid */ 1390 #ifdef ILL_DEBUG 1391 int ipsq_depth; /* debugging aid */ 1392 pc_t ipsq_stack[IP_STACK_DEPTH]; /* debugging aid */ 1393 #endif 1394 } ipsq_t; 1395 1396 /* ipsq_flags */ 1397 #define IPSQ_GROUP 0x1 /* This ipsq belongs to an IPMP group */ 1398 1399 /* 1400 * ipsq_waitfor: 1401 * 1402 * IPIF_DOWN 1 ipif_down waiting for refcnts to drop 1403 * ILL_DOWN 2 ill_down waiting for refcnts to drop 1404 * IPIF_FREE 3 ipif_free waiting for refcnts to drop 1405 * ILL_FREE 4 ill unplumb waiting for refcnts to drop 1406 * ILL_MOVE_OK 5 failover waiting for refcnts to drop 1407 */ 1408 1409 enum { IPIF_DOWN = 1, ILL_DOWN, IPIF_FREE, ILL_FREE, ILL_MOVE_OK }; 1410 1411 /* Flags passed to ipsq_try_enter */ 1412 #define CUR_OP 0 /* Current ioctl continuing again */ 1413 #define NEW_OP 1 /* New ioctl starting afresh */ 1414 1415 /* 1416 * phyint represents state that is common to both IPv4 and IPv6 interfaces. 1417 * There is a separate ill_t representing IPv4 and IPv6 which has a 1418 * backpointer to the phyint structure for acessing common state. 1419 * 1420 * NOTE : It just stores the group name as there is only one name for 1421 * IPv4 and IPv6 i.e it is a underlying link property. Actually 1422 * IPv4 and IPv6 ill are grouped together when their phyints have 1423 * the same name. 1424 */ 1425 typedef struct phyint { 1426 struct ill_s *phyint_illv4; 1427 struct ill_s *phyint_illv6; 1428 uint_t phyint_ifindex; /* SIOCLSLIFINDEX */ 1429 uint_t phyint_notify_delay; /* SIOCSLIFNOTIFYDELAY */ 1430 char *phyint_groupname; /* SIOCSLIFGROUPNAME */ 1431 uint_t phyint_groupname_len; 1432 uint64_t phyint_flags; 1433 avl_node_t phyint_avl_by_index; /* avl tree by index */ 1434 avl_node_t phyint_avl_by_name; /* avl tree by name */ 1435 kmutex_t phyint_lock; 1436 struct ipsq_s *phyint_ipsq; /* back pointer to ipsq */ 1437 struct phyint *phyint_ipsq_next; /* phyint list on this ipsq */ 1438 } phyint_t; 1439 1440 #define CACHE_ALIGN_SIZE 64 1441 1442 #define CACHE_ALIGN(align_struct) P2ROUNDUP(sizeof (struct align_struct),\ 1443 CACHE_ALIGN_SIZE) 1444 struct _phyint_list_s_ { 1445 avl_tree_t phyint_list_avl_by_index; /* avl tree by index */ 1446 avl_tree_t phyint_list_avl_by_name; /* avl tree by name */ 1447 }; 1448 1449 typedef union phyint_list_u { 1450 struct _phyint_list_s_ phyint_list_s; 1451 char phyint_list_filler[CACHE_ALIGN(_phyint_list_s_)]; 1452 } phyint_list_t; 1453 1454 #define phyint_list_avl_by_index phyint_list_s.phyint_list_avl_by_index 1455 #define phyint_list_avl_by_name phyint_list_s.phyint_list_avl_by_name 1456 /* 1457 * ILL groups. We group ills, 1458 * 1459 * - if the ills have the same group name. (New way) 1460 * 1461 * ill_group locking notes: 1462 * 1463 * illgrp_lock protects ill_grp_ill_schednext. 1464 * 1465 * ill_g_lock protects ill_grp_next, illgrp_ill, illgrp_ill_count. 1466 * Holding ill_g_lock freezes the memberships of ills in IPMP groups. 1467 * It also freezes the global list of ills and all ipifs in all ills. 1468 * 1469 * To remove an ipif from the linked list of ipifs of that ill ipif_free_tail 1470 * holds both ill_g_lock, and ill_lock. Similarly to remove an ill from the 1471 * global list of ills, ill_delete_glist holds ill_g_lock as writer. 1472 * This simplifies things for ipif_select_source, illgrp_scheduler etc. 1473 * that need to walk the members of an illgrp. They just hold ill_g_lock 1474 * as reader to do the walk. 1475 * 1476 */ 1477 typedef struct ill_group { 1478 kmutex_t illgrp_lock; 1479 struct ill_group *illgrp_next; /* Next ill_group */ 1480 struct ill_s *illgrp_ill_schednext; /* Next ill to be scheduled */ 1481 struct ill_s *illgrp_ill; /* First ill in the group */ 1482 int illgrp_ill_count; 1483 } ill_group_t; 1484 1485 extern ill_group_t *illgrp_head_v6; 1486 1487 /* 1488 * Fragmentation hash bucket 1489 */ 1490 typedef struct ipfb_s { 1491 struct ipf_s *ipfb_ipf; /* List of ... */ 1492 size_t ipfb_count; /* Count of bytes used by frag(s) */ 1493 kmutex_t ipfb_lock; /* Protect all ipf in list */ 1494 uint_t ipfb_frag_pkts; /* num of distinct fragmented pkts */ 1495 } ipfb_t; 1496 1497 /* 1498 * IRE bucket structure. Usually there is an array of such structures, 1499 * each pointing to a linked list of ires. irb_refcnt counts the number 1500 * of walkers of a given hash bucket. Usually the reference count is 1501 * bumped up if the walker wants no IRES to be DELETED while walking the 1502 * list. Bumping up does not PREVENT ADDITION. This allows walking a given 1503 * hash bucket without stumbling up on a free pointer. 1504 */ 1505 typedef struct irb { 1506 struct ire_s *irb_ire; /* First ire in this bucket */ 1507 /* Should be first in this struct */ 1508 krwlock_t irb_lock; /* Protect this bucket */ 1509 uint_t irb_refcnt; /* Protected by irb_lock */ 1510 uchar_t irb_marks; /* CONDEMNED ires in this bucket ? */ 1511 uint_t irb_ire_cnt; /* Num of IRE in this bucket */ 1512 uint_t irb_tmp_ire_cnt; /* Num of temporary IRE */ 1513 } irb_t; 1514 1515 #define IP_V4_G_HEAD 0 1516 #define IP_V6_G_HEAD 1 1517 1518 #define MAX_G_HEADS 2 1519 1520 /* 1521 * unpadded ill_if structure 1522 */ 1523 struct _ill_if_s_ { 1524 union ill_if_u *illif_next; 1525 union ill_if_u *illif_prev; 1526 avl_tree_t illif_avl_by_ppa; /* AVL tree sorted on ppa */ 1527 vmem_t *illif_ppa_arena; /* ppa index space */ 1528 uint16_t illif_mcast_v1; /* hints for */ 1529 uint16_t illif_mcast_v2; /* [igmp|mld]_slowtimo */ 1530 int illif_name_len; /* name length */ 1531 char illif_name[LIFNAMSIZ]; /* name of interface type */ 1532 }; 1533 1534 /* cache aligned ill_if structure */ 1535 typedef union ill_if_u { 1536 struct _ill_if_s_ ill_if_s; 1537 char illif_filler[CACHE_ALIGN(_ill_if_s_)]; 1538 } ill_if_t; 1539 1540 1541 #define illif_next ill_if_s.illif_next 1542 #define illif_prev ill_if_s.illif_prev 1543 #define illif_avl_by_ppa ill_if_s.illif_avl_by_ppa 1544 #define illif_ppa_arena ill_if_s.illif_ppa_arena 1545 #define illif_mcast_v1 ill_if_s.illif_mcast_v1 1546 #define illif_mcast_v2 ill_if_s.illif_mcast_v2 1547 #define illif_name ill_if_s.illif_name 1548 #define illif_name_len ill_if_s.illif_name_len 1549 1550 typedef struct ill_walk_context_s { 1551 int ctx_current_list; /* current list being searched */ 1552 int ctx_last_list; /* last list to search */ 1553 } ill_walk_context_t; 1554 1555 /* 1556 * ill_gheads structure, one for IPV4 and one for IPV6 1557 */ 1558 struct _ill_g_head_s_ { 1559 ill_if_t *ill_g_list_head; 1560 ill_if_t *ill_g_list_tail; 1561 }; 1562 1563 typedef union ill_g_head_u { 1564 struct _ill_g_head_s_ ill_g_head_s; 1565 char ill_g_head_filler[CACHE_ALIGN(_ill_g_head_s_)]; 1566 } ill_g_head_t; 1567 1568 #define ill_g_list_head ill_g_head_s.ill_g_list_head 1569 #define ill_g_list_tail ill_g_head_s.ill_g_list_tail 1570 1571 #pragma align CACHE_ALIGN_SIZE(ill_g_heads) 1572 extern ill_g_head_t ill_g_heads[]; /* ILL List Head */ 1573 1574 1575 #define IP_V4_ILL_G_LIST ill_g_heads[IP_V4_G_HEAD].ill_g_list_head 1576 #define IP_V6_ILL_G_LIST ill_g_heads[IP_V6_G_HEAD].ill_g_list_head 1577 #define IP_VX_ILL_G_LIST(i) ill_g_heads[i].ill_g_list_head 1578 1579 #define ILL_START_WALK_V4(ctx_ptr) ill_first(IP_V4_G_HEAD, IP_V4_G_HEAD, \ 1580 ctx_ptr) 1581 #define ILL_START_WALK_V6(ctx_ptr) ill_first(IP_V6_G_HEAD, IP_V6_G_HEAD, \ 1582 ctx_ptr) 1583 #define ILL_START_WALK_ALL(ctx_ptr) ill_first(MAX_G_HEADS, MAX_G_HEADS, \ 1584 ctx_ptr) 1585 1586 /* 1587 * Capabilities, possible flags for ill_capabilities. 1588 */ 1589 1590 #define ILL_CAPAB_AH 0x01 /* IPsec AH acceleration */ 1591 #define ILL_CAPAB_ESP 0x02 /* IPsec ESP acceleration */ 1592 #define ILL_CAPAB_MDT 0x04 /* Multidata Transmit */ 1593 #define ILL_CAPAB_HCKSUM 0x08 /* Hardware checksumming */ 1594 #define ILL_CAPAB_ZEROCOPY 0x10 /* Zero-copy */ 1595 #define ILL_CAPAB_POLL 0x20 /* Polling Toggle */ 1596 1597 /* 1598 * Per-ill Multidata Transmit capabilities. 1599 */ 1600 typedef struct ill_mdt_capab_s ill_mdt_capab_t; 1601 1602 /* 1603 * Per-ill IPsec capabilities. 1604 */ 1605 typedef struct ill_ipsec_capab_s ill_ipsec_capab_t; 1606 1607 /* 1608 * Per-ill Hardware Checksumming capbilities. 1609 */ 1610 typedef struct ill_hcksum_capab_s ill_hcksum_capab_t; 1611 1612 /* 1613 * Per-ill Zero-copy capabilities. 1614 */ 1615 typedef struct ill_zerocopy_capab_s ill_zerocopy_capab_t; 1616 1617 /* 1618 * Per-ill Polling capbilities. 1619 */ 1620 typedef struct ill_poll_capab_s ill_poll_capab_t; 1621 1622 /* 1623 * Per-ill polling resource map. 1624 */ 1625 typedef struct ill_rx_ring ill_rx_ring_t; 1626 1627 /* The following are ill_state_flags */ 1628 #define ILL_LL_SUBNET_PENDING 0x01 /* Waiting for DL_INFO_ACK from drv */ 1629 #define ILL_CONDEMNED 0x02 /* No more new ref's to the ILL */ 1630 #define ILL_CHANGING 0x04 /* ILL not globally visible */ 1631 #define ILL_DL_UNBIND_DONE 0x08 /* UNBIND_REQ has been Acked */ 1632 1633 /* Is this an ILL whose source address is used by other ILL's ? */ 1634 #define IS_USESRC_ILL(ill) \ 1635 (((ill)->ill_usesrc_ifindex == 0) && \ 1636 ((ill)->ill_usesrc_grp_next != NULL)) \ 1637 1638 /* Is this a client/consumer of the usesrc ILL ? */ 1639 #define IS_USESRC_CLI_ILL(ill) \ 1640 (((ill)->ill_usesrc_ifindex != 0) && \ 1641 ((ill)->ill_usesrc_grp_next != NULL)) \ 1642 1643 /* Is this an virtual network interface (vni) ILL ? */ 1644 #define IS_VNI(ill) \ 1645 (((ill) != NULL) && !((ill)->ill_phyint->phyint_flags & \ 1646 PHYI_LOOPBACK) && ((ill)->ill_phyint->phyint_flags & \ 1647 PHYI_VIRTUAL)) \ 1648 1649 /* 1650 * IP Lower level Structure. 1651 * Instance data structure in ip_open when there is a device below us. 1652 */ 1653 typedef struct ill_s { 1654 ill_if_t *ill_ifptr; /* pointer to interface type */ 1655 queue_t *ill_rq; /* Read queue. */ 1656 queue_t *ill_wq; /* Write queue. */ 1657 1658 int ill_error; /* Error value sent up by device. */ 1659 1660 ipif_t *ill_ipif; /* Interface chain for this ILL. */ 1661 1662 uint_t ill_ipif_up_count; /* Number of IPIFs currently up. */ 1663 uint_t ill_max_frag; /* Max IDU from DLPI. */ 1664 char *ill_name; /* Our name. */ 1665 uint_t ill_name_length; /* Name length, incl. terminator. */ 1666 char *ill_ndd_name; /* Name + ":ip?_forwarding" for NDD. */ 1667 uint_t ill_net_type; /* IRE_IF_RESOLVER/IRE_IF_NORESOLVER. */ 1668 /* 1669 * Physical Point of Attachment num. If DLPI style 1 provider 1670 * then this is derived from the devname. 1671 */ 1672 uint_t ill_ppa; 1673 t_uscalar_t ill_sap; 1674 t_scalar_t ill_sap_length; /* Including sign (for position) */ 1675 uint_t ill_phys_addr_length; /* Excluding the sap. */ 1676 uint_t ill_bcast_addr_length; /* Only set when the DL provider */ 1677 /* supports broadcast. */ 1678 t_uscalar_t ill_mactype; 1679 uint8_t *ill_frag_ptr; /* Reassembly state. */ 1680 timeout_id_t ill_frag_timer_id; /* timeout id for the frag timer */ 1681 ipfb_t *ill_frag_hash_tbl; /* Fragment hash list head. */ 1682 ipif_t *ill_pending_ipif; /* IPIF waiting for DL operation. */ 1683 1684 ilm_t *ill_ilm; /* Multicast mebership for lower ill */ 1685 uint_t ill_global_timer; /* for IGMPv3/MLDv2 general queries */ 1686 int ill_mcast_type; /* type of router which is querier */ 1687 /* on this interface */ 1688 uint16_t ill_mcast_v1_time; /* # slow timeouts since last v1 qry */ 1689 uint16_t ill_mcast_v2_time; /* # slow timeouts since last v2 qry */ 1690 uint8_t ill_mcast_v1_tset; /* 1 => timer is set; 0 => not set */ 1691 uint8_t ill_mcast_v2_tset; /* 1 => timer is set; 0 => not set */ 1692 1693 uint8_t ill_mcast_rv; /* IGMPv3/MLDv2 robustness variable */ 1694 int ill_mcast_qi; /* IGMPv3/MLDv2 query interval var */ 1695 1696 mblk_t *ill_pending_mp; /* IOCTL/DLPI awaiting completion. */ 1697 /* 1698 * All non-NULL cells between 'ill_first_mp_to_free' and 1699 * 'ill_last_mp_to_free' are freed in ill_delete. 1700 */ 1701 #define ill_first_mp_to_free ill_bcast_mp 1702 mblk_t *ill_bcast_mp; /* DLPI header for broadcasts. */ 1703 mblk_t *ill_resolver_mp; /* Resolver template. */ 1704 mblk_t *ill_detach_mp; /* detach mp, or NULL if style1 */ 1705 mblk_t *ill_unbind_mp; /* unbind mp from ill_dl_up() */ 1706 mblk_t *ill_dlpi_deferred; /* b_next chain of control messages */ 1707 mblk_t *ill_phys_addr_mp; /* mblk which holds ill_phys_addr */ 1708 #define ill_last_mp_to_free ill_phys_addr_mp 1709 1710 cred_t *ill_credp; /* opener's credentials */ 1711 uint8_t *ill_phys_addr; /* ill_phys_addr_mp->b_rptr + off */ 1712 1713 uint_t ill_state_flags; /* see ILL_* flags above */ 1714 1715 /* Following bit fields protected by ipsq_t */ 1716 uint_t 1717 ill_needs_attach : 1, 1718 ill_reserved : 1, 1719 ill_isv6 : 1, 1720 ill_dlpi_style_set : 1, 1721 1722 ill_ifname_pending : 1, 1723 ill_move_in_progress : 1, /* FAILOVER/FAILBACK in progress */ 1724 ill_join_allmulti : 1, 1725 ill_logical_down : 1, 1726 1727 ill_is_6to4tun : 1, /* Interface is a 6to4 tunnel */ 1728 ill_promisc_on_phys : 1, /* phys interface in promisc mode */ 1729 ill_dl_up : 1, 1730 ill_up_ipifs : 1, 1731 1732 ill_pad_to_bit_31 : 20; 1733 1734 /* Following bit fields protected by ill_lock */ 1735 uint_t 1736 ill_fragtimer_executing : 1, 1737 ill_fragtimer_needrestart : 1, 1738 ill_ilm_cleanup_reqd : 1, 1739 ill_arp_closing : 1, 1740 1741 ill_arp_bringup_pending : 1, 1742 ill_mtu_userspecified : 1, /* SIOCSLNKINFO has set the mtu */ 1743 ill_pad_bit_31 : 26; 1744 1745 /* 1746 * Used in SIOCSIFMUXID and SIOCGIFMUXID for 'ifconfig unplumb'. 1747 */ 1748 int ill_arp_muxid; /* muxid returned from plink for arp */ 1749 int ill_ip_muxid; /* muxid returned from plink for ip */ 1750 1751 /* 1752 * Used for IP frag reassembly throttling on a per ILL basis. 1753 * 1754 * Note: frag_count is approximate, its added to and subtracted from 1755 * without any locking, so simultaneous load/modify/stores can 1756 * collide, also ill_frag_purge() recalculates its value by 1757 * summing all the ipfb_count's without locking out updates 1758 * to the ipfb's. 1759 */ 1760 uint_t ill_ipf_gen; /* Generation of next fragment queue */ 1761 uint_t ill_frag_count; /* Approx count of all mblk bytes */ 1762 uint_t ill_frag_free_num_pkts; /* num of fragmented packets to free */ 1763 clock_t ill_last_frag_clean_time; /* time when frag's were pruned */ 1764 int ill_type; /* From <net/if_types.h> */ 1765 uint_t ill_dlpi_multicast_state; /* See below IDMS_* */ 1766 uint_t ill_dlpi_fastpath_state; /* See below IDMS_* */ 1767 1768 /* 1769 * Capabilities related fields. 1770 */ 1771 uint_t ill_capab_state; /* State of capability query, IDMS_* */ 1772 uint64_t ill_capabilities; /* Enabled capabilities, ILL_CAPAB_* */ 1773 ill_mdt_capab_t *ill_mdt_capab; /* Multidata Transmit capabilities */ 1774 ill_ipsec_capab_t *ill_ipsec_capab_ah; /* IPsec AH capabilities */ 1775 ill_ipsec_capab_t *ill_ipsec_capab_esp; /* IPsec ESP capabilities */ 1776 ill_hcksum_capab_t *ill_hcksum_capab; /* H/W cksumming capabilities */ 1777 ill_zerocopy_capab_t *ill_zerocopy_capab; /* Zero-copy capabilities */ 1778 ill_poll_capab_t *ill_poll_capab; /* Polling capabilities */ 1779 1780 /* 1781 * New fields for IPv6 1782 */ 1783 uint8_t ill_max_hops; /* Maximum hops for any logical interface */ 1784 uint_t ill_max_mtu; /* Maximum MTU for any logical interface */ 1785 uint32_t ill_reachable_time; /* Value for ND algorithm in msec */ 1786 uint32_t ill_reachable_retrans_time; /* Value for ND algorithm msec */ 1787 uint_t ill_max_buf; /* Max # of req to buffer for ND */ 1788 in6_addr_t ill_token; 1789 uint_t ill_token_length; 1790 uint32_t ill_xmit_count; /* ndp max multicast xmits */ 1791 mib2_ipv6IfStatsEntry_t *ill_ip6_mib; /* Per interface mib */ 1792 mib2_ipv6IfIcmpEntry_t *ill_icmp6_mib; /* Per interface mib */ 1793 /* 1794 * Following two mblks are allocated common to all 1795 * the ipifs when the first interface is coming up. 1796 * It is sent up to arp when the last ipif is coming 1797 * down. 1798 */ 1799 mblk_t *ill_arp_down_mp; 1800 mblk_t *ill_arp_del_mapping_mp; 1801 /* 1802 * Used for implementing IFF_NOARP. As IFF_NOARP is used 1803 * to turn off for all the logicals, it is here instead 1804 * of the ipif. 1805 */ 1806 mblk_t *ill_arp_on_mp; 1807 /* Peer ill of an IPMP move operation */ 1808 struct ill_s *ill_move_peer; 1809 1810 phyint_t *ill_phyint; 1811 uint64_t ill_flags; 1812 ill_group_t *ill_group; 1813 struct ill_s *ill_group_next; 1814 /* 1815 * Reverse tunnel related count. This count 1816 * determines how many mobile nodes are using this 1817 * ill to send packet to reverse tunnel via foreign 1818 * agent. A non-zero count specifies presence of 1819 * mobile node(s) using reverse tunnel through this 1820 * interface. 1821 */ 1822 uint32_t ill_mrtun_refcnt; 1823 1824 /* 1825 * This count is bumped up when a route is added with 1826 * RTA_SRCIFP bit flag using routing socket. 1827 */ 1828 uint32_t ill_srcif_refcnt; 1829 /* 1830 * Pointer to the special interface based routing table. 1831 * This routing table is created dynamically when RTA_SRCIFP 1832 * is set by the routing socket. 1833 */ 1834 irb_t *ill_srcif_table; 1835 kmutex_t ill_lock; /* Please see table below */ 1836 /* 1837 * The ill_nd_lla* fields handle the link layer address option 1838 * from neighbor discovery. This is used for external IPv6 1839 * address resolution. 1840 */ 1841 mblk_t *ill_nd_lla_mp; /* mblk which holds ill_nd_lla */ 1842 uint8_t *ill_nd_lla; /* Link Layer Address */ 1843 uint_t ill_nd_lla_len; /* Link Layer Address length */ 1844 /* 1845 * We now have 3 phys_addr_req's sent down. This field keeps track 1846 * of which one is pending. 1847 */ 1848 t_uscalar_t ill_phys_addr_pend; /* which dl_phys_addr_req pending */ 1849 /* 1850 * Used to save errors that occur during plumbing 1851 */ 1852 uint_t ill_ifname_pending_err; 1853 avl_node_t ill_avl_byppa; /* avl node based on ppa */ 1854 void *ill_fastpath_list; /* both ire and nce hang off this */ 1855 uint_t ill_refcnt; /* active refcnt by threads */ 1856 uint_t ill_ire_cnt; /* ires associated with this ill */ 1857 kcondvar_t ill_cv; 1858 uint_t ill_ilm_walker_cnt; /* snmp ilm walkers */ 1859 uint_t ill_nce_cnt; /* nces associated with this ill */ 1860 uint_t ill_waiters; /* threads waiting in ipsq_enter */ 1861 /* 1862 * Contains the upper read queue pointer of the module immediately 1863 * beneath IP. This field allows IP to validate sub-capability 1864 * acknowledgments coming up from downstream. 1865 */ 1866 queue_t *ill_lmod_rq; /* read queue pointer of module below */ 1867 uint_t ill_lmod_cnt; /* number of modules beneath IP */ 1868 ip_m_t *ill_media; /* media specific params/functions */ 1869 t_uscalar_t ill_dlpi_pending; /* Last DLPI primitive issued */ 1870 uint_t ill_usesrc_ifindex; /* use src addr from this ILL */ 1871 struct ill_s *ill_usesrc_grp_next; /* Next ILL in the usesrc group */ 1872 #ifdef ILL_DEBUG 1873 th_trace_t *ill_trace[IP_TR_HASH_MAX]; 1874 boolean_t ill_trace_disable; /* True when alloc fails */ 1875 #endif 1876 } ill_t; 1877 1878 extern void ill_delete_glist(ill_t *); 1879 1880 /* 1881 * The following table lists the protection levels of the various members 1882 * of the ill_t. Same notation as that used for ipif_t above is used. 1883 * 1884 * Write Read 1885 * 1886 * ill_ifptr ill_g_lock + s Write once 1887 * ill_rq ipsq Write once 1888 * ill_wq ipsq Write once 1889 * 1890 * ill_error ipsq None 1891 * ill_ipif ill_g_lock + ipsq ill_g_lock OR ipsq 1892 * ill_ipif_up_count ill_lock + ipsq ill_lock 1893 * ill_max_frag ipsq Write once 1894 * 1895 * ill_name ill_g_lock + ipsq Write once 1896 * ill_name_length ill_g_lock + ipsq Write once 1897 * ill_ndd_name ipsq Write once 1898 * ill_net_type ipsq Write once 1899 * ill_ppa ill_g_lock + ipsq Write once 1900 * ill_sap ipsq + down ill Write once 1901 * ill_sap_length ipsq + down ill Write once 1902 * ill_phys_addr_length ipsq + down ill Write once 1903 * 1904 * ill_bcast_addr_length ipsq ipsq 1905 * ill_mactype ipsq ipsq 1906 * ill_frag_ptr ipsq ipsq 1907 * 1908 * ill_frag_timer_id ill_lock ill_lock 1909 * ill_frag_hash_tbl ipsq up ill 1910 * ill_ilm ipsq + ill_lock ill_lock 1911 * ill_mcast_type ill_lock ill_lock 1912 * ill_mcast_v1_time ill_lock ill_lock 1913 * ill_mcast_v2_time ill_lock ill_lock 1914 * ill_mcast_v1_tset ill_lock ill_lock 1915 * ill_mcast_v2_tset ill_lock ill_lock 1916 * ill_mcast_rv ill_lock ill_lock 1917 * ill_mcast_qi ill_lock ill_lock 1918 * ill_pending_mp ill_lock ill_lock 1919 * 1920 * ill_bcast_mp ipsq ipsq 1921 * ill_resolver_mp ipsq only when ill is up 1922 * ill_down_mp ipsq ipsq 1923 * ill_dlpi_deferred ipsq ipsq 1924 * ill_phys_addr_mp ipsq ipsq 1925 * ill_phys_addr ipsq up ill 1926 * ill_ick ipsq + down ill only when ill is up 1927 * 1928 * ill_state_flags ill_lock ill_lock 1929 * exclusive bit flags ipsq_t ipsq_t 1930 * shared bit flags ill_lock ill_lock 1931 * 1932 * ill_arp_muxid ipsq Not atomic 1933 * ill_ip_muxid ipsq Not atomic 1934 * 1935 * ill_ipf_gen Not atomic 1936 * ill_frag_count Approx. not protected 1937 * ill_type ipsq + down ill only when ill is up 1938 * ill_dlpi_multicast_state ill_lock ill_lock 1939 * ill_dlpi_fastpath_state ill_lock ill_lock 1940 * ill_max_hops ipsq Not atomic 1941 * 1942 * ill_max_mtu 1943 * 1944 * ill_reachable_time ipsq + ill_lock ill_lock 1945 * ill_reachable_retrans_time ipsq + ill_lock ill_lock 1946 * ill_max_buf ipsq + ill_lock ill_lock 1947 * 1948 * Next 2 fields need ill_lock because of the get ioctls. They should not 1949 * report partially updated results without executing in the ipsq. 1950 * ill_token ipsq + ill_lock ill_lock 1951 * ill_token_length ipsq + ill_lock ill_lock 1952 * ill_xmit_count ipsq + down ill write once 1953 * ill_ip6_mib ipsq + down ill only when ill is up 1954 * ill_icmp6_mib ipsq + down ill only when ill is up 1955 * ill_arp_down_mp ipsq ipsq 1956 * ill_arp_del_mapping_mp ipsq ipsq 1957 * ill_arp_on_mp ipsq ipsq 1958 * ill_move_peer ipsq ipsq 1959 * 1960 * ill_phyint ipsq, ill_g_lock, ill_lock Any of them 1961 * ill_flags ill_lock ill_lock 1962 * ill_group ipsq, ill_g_lock, ill_lock Any of them 1963 * ill_group_next ipsq, ill_g_lock, ill_lock Any of them 1964 * ill_mrtun_refcnt ill_lock ill_lock 1965 * ill_srcif_refcnt ill_lock ill_lock 1966 * ill_srcif_table ill_lock ill_lock 1967 * ill_nd_lla_mp ill_lock ill_lock 1968 * ill_nd_lla ill_lock ill_lock 1969 * ill_nd_lla_len ill_lock ill_lock 1970 * ill_phys_addr_pend ipsq + down ill only when ill is up 1971 * ill_ifname_pending_err ipsq ipsq 1972 * ill_avl_byppa ipsq, ill_g_lock Write once 1973 * 1974 * ill_fastpath_list ill_lock ill_lock 1975 * ill_refcnt ill_lock ill_lock 1976 * ill_ire_cnt ill_lock ill_lock 1977 * ill_cv ill_lock ill_lock 1978 * ill_ilm_walker_cnt ill_lock ill_lock 1979 * ill_nce_cnt ill_lock ill_lock 1980 * ill_trace ill_lock ill_lock 1981 * ill_usesrc_grp_next ill_g_usesrc_lock ill_g_usesrc_lock 1982 */ 1983 1984 /* 1985 * For ioctl restart mechanism see ip_reprocess_ioctl() 1986 */ 1987 struct ip_ioctl_cmd_s; 1988 1989 typedef int (*ifunc_t)(ipif_t *, struct sockaddr_in *, queue_t *, mblk_t *, 1990 struct ip_ioctl_cmd_s *, void *); 1991 1992 typedef struct ip_ioctl_cmd_s { 1993 int ipi_cmd; 1994 size_t ipi_copyin_size; 1995 uint_t ipi_flags; 1996 uint_t ipi_cmd_type; 1997 ifunc_t ipi_func; 1998 ifunc_t ipi_func_restart; 1999 } ip_ioctl_cmd_t; 2000 2001 /* 2002 * ipi_cmd_type: 2003 * 2004 * IF_CMD 1 old style ifreq cmd 2005 * LIF_CMD 2 new style lifreq cmd 2006 * MISC_CMD 3 Misc. (non [l]ifreq, tun) cmds 2007 * TUN_CMD 4 tunnel related 2008 */ 2009 2010 enum { IF_CMD = 1, LIF_CMD, MISC_CMD, TUN_CMD }; 2011 2012 #define IPI_DONTCARE 0 /* For ioctl encoded values that don't matter */ 2013 2014 /* Flag values in ipi_flags */ 2015 #define IPI_PRIV 0x1 /* Root only command */ 2016 #define IPI_MODOK 0x2 /* Permitted on mod instance of IP */ 2017 #define IPI_WR 0x4 /* Need to grab writer access */ 2018 #define IPI_GET_CMD 0x8 /* branch to mi_copyout on success */ 2019 #define IPI_REPL 0x10 /* valid for replacement ipif created in MOVE */ 2020 #define IPI_NULL_BCONT 0x20 /* ioctl has not data and hence no b_cont */ 2021 #define IPI_PASS_DOWN 0x40 /* pass this ioctl down when a module only */ 2022 2023 extern ip_ioctl_cmd_t ip_ndx_ioctl_table[]; 2024 extern ip_ioctl_cmd_t ip_misc_ioctl_table[]; 2025 extern int ip_ndx_ioctl_count; 2026 extern int ip_misc_ioctl_count; 2027 2028 #define ILL_CLEAR_MOVE(ill) { \ 2029 ill_t *peer_ill; \ 2030 \ 2031 peer_ill = (ill)->ill_move_peer; \ 2032 ASSERT(peer_ill != NULL); \ 2033 (ill)->ill_move_in_progress = B_FALSE; \ 2034 peer_ill->ill_move_in_progress = B_FALSE; \ 2035 (ill)->ill_move_peer = NULL; \ 2036 peer_ill->ill_move_peer = NULL; \ 2037 } 2038 2039 /* Passed down by ARP to IP during I_PLINK/I_PUNLINK */ 2040 typedef struct ipmx_s { 2041 char ipmx_name[LIFNAMSIZ]; /* if name */ 2042 uint_t 2043 ipmx_arpdev_stream : 1, /* This is the arp stream */ 2044 ipmx_notused : 31; 2045 } ipmx_t; 2046 2047 /* 2048 * State for detecting if a driver supports certain features. 2049 * Support for DL_ENABMULTI_REQ uses ill_dlpi_multicast_state. 2050 * Support for DLPI M_DATA fastpath uses ill_dlpi_fastpath_state. 2051 */ 2052 #define IDMS_UNKNOWN 0 /* No DL_ENABMULTI_REQ sent */ 2053 #define IDMS_INPROGRESS 1 /* Sent DL_ENABMULTI_REQ */ 2054 #define IDMS_OK 2 /* DL_ENABMULTI_REQ ok */ 2055 #define IDMS_FAILED 3 /* DL_ENABMULTI_REQ failed */ 2056 #define IDMS_RENEG 4 /* Driver asked for a renegotiation */ 2057 2058 /* Named Dispatch Parameter Management Structure */ 2059 typedef struct ipparam_s { 2060 uint_t ip_param_min; 2061 uint_t ip_param_max; 2062 uint_t ip_param_value; 2063 char *ip_param_name; 2064 } ipparam_t; 2065 2066 /* Extended NDP Management Structure */ 2067 typedef struct ipndp_s { 2068 ndgetf_t ip_ndp_getf; 2069 ndsetf_t ip_ndp_setf; 2070 caddr_t ip_ndp_data; 2071 char *ip_ndp_name; 2072 } ipndp_t; 2073 2074 /* 2075 * Following are the macros to increment/decrement the reference 2076 * count of the IREs and IRBs (ire bucket). 2077 * 2078 * 1) We bump up the reference count of an IRE to make sure that 2079 * it does not get deleted and freed while we are using it. 2080 * Typically all the lookup functions hold the bucket lock, 2081 * and look for the IRE. If it finds an IRE, it bumps up the 2082 * reference count before dropping the lock. Sometimes we *may* want 2083 * to bump up the reference count after we *looked* up i.e without 2084 * holding the bucket lock. So, the IRE_REFHOLD macro does not assert 2085 * on the bucket lock being held. Any thread trying to delete from 2086 * the hash bucket can still do so but cannot free the IRE if 2087 * ire_refcnt is not 0. 2088 * 2089 * 2) We bump up the reference count on the bucket where the IRE resides 2090 * (IRB), when we want to prevent the IREs getting deleted from a given 2091 * hash bucket. This makes life easier for ire_walk type functions which 2092 * wants to walk the IRE list, call a function, but needs to drop 2093 * the bucket lock to prevent recursive rw_enters. While the 2094 * lock is dropped, the list could be changed by other threads or 2095 * the same thread could end up deleting the ire or the ire pointed by 2096 * ire_next. IRE_REFHOLDing the ire or ire_next is not sufficient as 2097 * a delete will still remove the ire from the bucket while we have 2098 * dropped the lock and hence the ire_next would be NULL. Thus, we 2099 * need a mechanism to prevent deletions from a given bucket. 2100 * 2101 * To prevent deletions, we bump up the reference count on the 2102 * bucket. If the bucket is held, ire_delete just marks IRE_MARK_CONDEMNED 2103 * both on the ire's ire_marks and the bucket's irb_marks. When the 2104 * reference count on the bucket drops to zero, all the CONDEMNED ires 2105 * are deleted. We don't have to bump up the reference count on the 2106 * bucket if we are walking the bucket and never have to drop the bucket 2107 * lock. Note that IRB_REFHOLD does not prevent addition of new ires 2108 * in the list. It is okay because addition of new ires will not cause 2109 * ire_next to point to freed memory. We do IRB_REFHOLD only when 2110 * all of the 3 conditions are true : 2111 * 2112 * 1) The code needs to walk the IRE bucket from start to end. 2113 * 2) It may have to drop the bucket lock sometimes while doing (1) 2114 * 3) It does not want any ires to be deleted meanwhile. 2115 */ 2116 2117 /* 2118 * Bump up the reference count on the IRE. We cannot assert that the 2119 * bucket lock is being held as it is legal to bump up the reference 2120 * count after the first lookup has returned the IRE without 2121 * holding the lock. Currently ip_wput does this for caching IRE_CACHEs. 2122 */ 2123 2124 #ifndef IRE_DEBUG 2125 2126 #define IRE_REFHOLD_NOTR(ire) IRE_REFHOLD(ire) 2127 #define IRE_UNTRACE_REF(ire) 2128 #define IRE_TRACE_REF(ire) 2129 2130 #else 2131 2132 #define IRE_REFHOLD_NOTR(ire) { \ 2133 atomic_add_32(&(ire)->ire_refcnt, 1); \ 2134 ASSERT((ire)->ire_refcnt != 0); \ 2135 } 2136 2137 #define IRE_UNTRACE_REF(ire) ire_untrace_ref(ire); 2138 #define IRE_TRACE_REF(ire) ire_trace_ref(ire); 2139 #endif 2140 2141 #define IRE_REFHOLD(ire) { \ 2142 atomic_add_32(&(ire)->ire_refcnt, 1); \ 2143 ASSERT((ire)->ire_refcnt != 0); \ 2144 IRE_TRACE_REF(ire); \ 2145 } 2146 2147 #define IRE_REFHOLD_LOCKED(ire) { \ 2148 IRE_TRACE_REF(ire); \ 2149 (ire)->ire_refcnt++; \ 2150 } 2151 2152 /* 2153 * Decrement the reference count on the IRE. 2154 * In architectures e.g sun4u, where atomic_add_32_nv is just 2155 * a cas, we need to maintain the right memory barrier semantics 2156 * as that of mutex_exit i.e all the loads and stores should complete 2157 * before the cas is executed. membar_exit() does that here. 2158 * 2159 * NOTE : This macro is used only in places where we want performance. 2160 * To avoid bloating the code, we use the function "ire_refrele" 2161 * which essentially calls the macro. 2162 */ 2163 #ifndef IRE_DEBUG 2164 #define IRE_REFRELE(ire) { \ 2165 ASSERT((ire)->ire_refcnt != 0); \ 2166 membar_exit(); \ 2167 if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0) \ 2168 ire_inactive(ire); \ 2169 } 2170 #define IRE_REFRELE_NOTR(ire) IRE_REFRELE(ire) 2171 #else 2172 #define IRE_REFRELE(ire) { \ 2173 if (ire->ire_bucket != NULL) \ 2174 ire_untrace_ref(ire); \ 2175 ASSERT((ire)->ire_refcnt != 0); \ 2176 membar_exit(); \ 2177 if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0) \ 2178 ire_inactive(ire); \ 2179 } 2180 #define IRE_REFRELE_NOTR(ire) { \ 2181 ASSERT((ire)->ire_refcnt != 0); \ 2182 membar_exit(); \ 2183 if (atomic_add_32_nv(&(ire)->ire_refcnt, -1) == 0) \ 2184 ire_inactive(ire); \ 2185 } 2186 #endif 2187 2188 /* 2189 * Bump up the reference count on the hash bucket - IRB to 2190 * prevent ires from being deleted in this bucket. 2191 */ 2192 #define IRB_REFHOLD(irb) { \ 2193 rw_enter(&(irb)->irb_lock, RW_WRITER); \ 2194 (irb)->irb_refcnt++; \ 2195 ASSERT((irb)->irb_refcnt != 0); \ 2196 rw_exit(&(irb)->irb_lock); \ 2197 } 2198 2199 #define IRB_REFRELE(irb) { \ 2200 rw_enter(&(irb)->irb_lock, RW_WRITER); \ 2201 ASSERT((irb)->irb_refcnt != 0); \ 2202 if (--(irb)->irb_refcnt == 0 && \ 2203 ((irb)->irb_marks & IRE_MARK_CONDEMNED)) { \ 2204 ire_t *ire_list; \ 2205 \ 2206 ire_list = ire_unlink(irb); \ 2207 rw_exit(&(irb)->irb_lock); \ 2208 ASSERT(ire_list != NULL); \ 2209 ire_cleanup(ire_list); \ 2210 } else { \ 2211 rw_exit(&(irb)->irb_lock); \ 2212 } \ 2213 } 2214 2215 /* 2216 * Lock the fast path mp for access, since the ire_fp_mp can be deleted 2217 * due a DL_NOTE_FASTPATH_FLUSH in the case of IRE_BROADCAST and IRE_MIPRTUN 2218 */ 2219 2220 #define LOCK_IRE_FP_MP(ire) { \ 2221 if ((ire)->ire_type == IRE_BROADCAST || \ 2222 (ire)->ire_type == IRE_MIPRTUN) \ 2223 mutex_enter(&ire->ire_lock); \ 2224 } 2225 #define UNLOCK_IRE_FP_MP(ire) { \ 2226 if ((ire)->ire_type == IRE_BROADCAST || \ 2227 (ire)->ire_type == IRE_MIPRTUN) \ 2228 mutex_exit(&ire->ire_lock); \ 2229 } 2230 2231 typedef struct ire4 { 2232 ipaddr_t ire4_src_addr; /* Source address to use. */ 2233 ipaddr_t ire4_mask; /* Mask for matching this IRE. */ 2234 ipaddr_t ire4_addr; /* Address this IRE represents. */ 2235 ipaddr_t ire4_gateway_addr; /* Gateway if IRE_CACHE/IRE_OFFSUBNET */ 2236 ipaddr_t ire4_cmask; /* Mask from parent prefix route */ 2237 } ire4_t; 2238 2239 typedef struct ire6 { 2240 in6_addr_t ire6_src_addr; /* Source address to use. */ 2241 in6_addr_t ire6_mask; /* Mask for matching this IRE. */ 2242 in6_addr_t ire6_addr; /* Address this IRE represents. */ 2243 in6_addr_t ire6_gateway_addr; /* Gateway if IRE_CACHE/IRE_OFFSUBNET */ 2244 in6_addr_t ire6_cmask; /* Mask from parent prefix route */ 2245 } ire6_t; 2246 2247 typedef union ire_addr { 2248 ire6_t ire6_u; 2249 ire4_t ire4_u; 2250 } ire_addr_u_t; 2251 2252 /* Internet Routing Entry */ 2253 typedef struct ire_s { 2254 struct ire_s *ire_next; /* The hash chain must be first. */ 2255 struct ire_s **ire_ptpn; /* Pointer to previous next. */ 2256 uint32_t ire_refcnt; /* Number of references */ 2257 mblk_t *ire_mp; /* Non-null if allocated as mblk */ 2258 mblk_t *ire_fp_mp; /* Fast path header */ 2259 queue_t *ire_rfq; /* recv from this queue */ 2260 queue_t *ire_stq; /* send to this queue */ 2261 union { 2262 uint_t *max_fragp; /* Used only during ire creation */ 2263 uint_t max_frag; /* MTU (next hop or path). */ 2264 } imf_u; 2265 #define ire_max_frag imf_u.max_frag 2266 #define ire_max_fragp imf_u.max_fragp 2267 uint32_t ire_frag_flag; /* IPH_DF or zero. */ 2268 uint32_t ire_ident; /* Per IRE IP ident. */ 2269 uint32_t ire_tire_mark; /* Used for reclaim of unused. */ 2270 uchar_t ire_ipversion; /* IPv4/IPv6 version */ 2271 uchar_t ire_marks; /* IRE_MARK_CONDEMNED etc. */ 2272 ushort_t ire_type; /* Type of IRE */ 2273 uint_t ire_ib_pkt_count; /* Inbound packets for ire_addr */ 2274 uint_t ire_ob_pkt_count; /* Outbound packets to ire_addr */ 2275 uint_t ire_ll_hdr_length; /* Non-zero if we do M_DATA prepends */ 2276 time_t ire_create_time; /* Time (in secs) IRE was created. */ 2277 mblk_t *ire_dlureq_mp; /* DL_UNIT_DATA_REQ/RESOLVER mp */ 2278 uint32_t ire_phandle; /* Associate prefix IREs to cache */ 2279 uint32_t ire_ihandle; /* Associate interface IREs to cache */ 2280 ipif_t *ire_ipif; /* the interface that this ire uses */ 2281 uint32_t ire_flags; /* flags related to route (RTF_*) */ 2282 uint_t ire_ipsec_overhead; /* IPSEC overhead */ 2283 struct nce_s *ire_nce; /* Neighbor Cache Entry for IPv6 */ 2284 uint_t ire_masklen; /* # bits in ire_mask{,_v6} */ 2285 ire_addr_u_t ire_u; /* IPv4/IPv6 address info. */ 2286 2287 irb_t *ire_bucket; /* Hash bucket when ire_ptphn is set */ 2288 iulp_t ire_uinfo; /* Upper layer protocol info. */ 2289 /* 2290 * Protects ire_uinfo, ire_max_frag, and ire_frag_flag. 2291 */ 2292 kmutex_t ire_lock; 2293 uint_t ire_ipif_seqid; /* ipif_seqid of ire_ipif */ 2294 /* 2295 * For regular routes in forwarding table and cache table the 2296 * the following entries are NULL/zero. Only reverse tunnel 2297 * table and interface based forwarding table use these fields. 2298 * Routes added with RTA_SRCIFP and RTA_SRC respectively have 2299 * non-zero values for the following fields. 2300 */ 2301 ill_t *ire_in_ill; /* Incoming ill interface */ 2302 ipaddr_t ire_in_src_addr; 2303 /* source ip-addr of incoming packet */ 2304 clock_t ire_last_used_time; /* Last used time */ 2305 struct ire_s *ire_fastpath; /* Pointer to next ire in fastpath */ 2306 zoneid_t ire_zoneid; /* for local address discrimination */ 2307 #ifdef IRE_DEBUG 2308 th_trace_t *ire_trace[IP_TR_HASH_MAX]; 2309 boolean_t ire_trace_disable; /* True when alloc fails */ 2310 #endif 2311 } ire_t; 2312 2313 /* IPv4 compatiblity macros */ 2314 #define ire_src_addr ire_u.ire4_u.ire4_src_addr 2315 #define ire_mask ire_u.ire4_u.ire4_mask 2316 #define ire_addr ire_u.ire4_u.ire4_addr 2317 #define ire_gateway_addr ire_u.ire4_u.ire4_gateway_addr 2318 #define ire_cmask ire_u.ire4_u.ire4_cmask 2319 2320 #define ire_src_addr_v6 ire_u.ire6_u.ire6_src_addr 2321 #define ire_mask_v6 ire_u.ire6_u.ire6_mask 2322 #define ire_addr_v6 ire_u.ire6_u.ire6_addr 2323 #define ire_gateway_addr_v6 ire_u.ire6_u.ire6_gateway_addr 2324 #define ire_cmask_v6 ire_u.ire6_u.ire6_cmask 2325 2326 /* Convenient typedefs for sockaddrs */ 2327 typedef struct sockaddr_in sin_t; 2328 typedef struct sockaddr_in6 sin6_t; 2329 2330 /* Address structure used for internal bind with IP */ 2331 typedef struct ipa_conn_s { 2332 ipaddr_t ac_laddr; 2333 ipaddr_t ac_faddr; 2334 uint16_t ac_fport; 2335 uint16_t ac_lport; 2336 } ipa_conn_t; 2337 2338 typedef struct ipa6_conn_s { 2339 in6_addr_t ac6_laddr; 2340 in6_addr_t ac6_faddr; 2341 uint16_t ac6_fport; 2342 uint16_t ac6_lport; 2343 } ipa6_conn_t; 2344 2345 /* 2346 * Using ipa_conn_x_t or ipa6_conn_x_t allows us to modify the behavior of IP's 2347 * bind handler. 2348 */ 2349 typedef struct ipa_conn_extended_s { 2350 uint64_t acx_flags; 2351 ipa_conn_t acx_conn; 2352 } ipa_conn_x_t; 2353 2354 typedef struct ipa6_conn_extended_s { 2355 uint64_t ac6x_flags; 2356 ipa6_conn_t ac6x_conn; 2357 } ipa6_conn_x_t; 2358 2359 /* flag values for ipa_conn_x_t and ipa6_conn_x_t. */ 2360 #define ACX_VERIFY_DST 0x1ULL /* verify destination address is reachable */ 2361 2362 /* Name/Value Descriptor. */ 2363 typedef struct nv_s { 2364 uint64_t nv_value; 2365 char *nv_name; 2366 } nv_t; 2367 2368 /* IP Forwarding Ticket */ 2369 typedef struct ipftk_s { 2370 queue_t *ipftk_queue; 2371 ipaddr_t ipftk_dst; 2372 } ipftk_t; 2373 2374 typedef struct ipt_s { 2375 pfv_t func; /* Routine to call */ 2376 uchar_t *arg; /* ire or nce passed in */ 2377 } ipt_t; 2378 2379 #define ILL_FRAG_HASH(s, i) \ 2380 ((ntohl(s) ^ ((i) ^ ((i) >> 8))) % ILL_FRAG_HASH_TBL_COUNT) 2381 2382 /* 2383 * The MAX number of allowed fragmented packets per hash bucket 2384 * calculation is based on the most common mtu size of 1500. This limit 2385 * will work well for other mtu sizes as well. 2386 */ 2387 #define COMMON_IP_MTU 1500 2388 #define MAX_FRAG_MIN 10 2389 #define MAX_FRAG_PKTS \ 2390 MAX(MAX_FRAG_MIN, (2 * (ip_reass_queue_bytes / \ 2391 (COMMON_IP_MTU * ILL_FRAG_HASH_TBL_COUNT)))) 2392 2393 /* 2394 * Maximum dups allowed per packet. 2395 */ 2396 extern uint_t ip_max_frag_dups; 2397 2398 /* 2399 * Per-packet information for received packets and transmitted. 2400 * Used by the transport protocols when converting between the packet 2401 * and ancillary data and socket options. 2402 * 2403 * Note: This private data structure and related IPPF_* constant 2404 * definitions are exposed to enable compilation of some debugging tools 2405 * like lsof which use struct tcp_t in <inet/tcp.h>. This is intended to be 2406 * a temporary hack and long term alternate interfaces should be defined 2407 * to support the needs of such tools and private definitions moved to 2408 * private headers. 2409 */ 2410 struct ip6_pkt_s { 2411 uint_t ipp_fields; /* Which fields are valid */ 2412 uint_t ipp_sticky_ignored; /* sticky fields to ignore */ 2413 uint_t ipp_ifindex; /* pktinfo ifindex */ 2414 in6_addr_t ipp_addr; /* pktinfo src/dst addr */ 2415 uint_t ipp_unicast_hops; /* IPV6_UNICAST_HOPS */ 2416 uint_t ipp_multicast_hops; /* IPV6_MULTICAST_HOPS */ 2417 uint_t ipp_hoplimit; /* IPV6_HOPLIMIT */ 2418 uint_t ipp_hopoptslen; 2419 uint_t ipp_rtdstoptslen; 2420 uint_t ipp_rthdrlen; 2421 uint_t ipp_dstoptslen; 2422 uint_t ipp_pathmtulen; 2423 ip6_hbh_t *ipp_hopopts; 2424 ip6_dest_t *ipp_rtdstopts; 2425 ip6_rthdr_t *ipp_rthdr; 2426 ip6_dest_t *ipp_dstopts; 2427 struct ip6_mtuinfo *ipp_pathmtu; 2428 in6_addr_t ipp_nexthop; /* Transmit only */ 2429 uint8_t ipp_tclass; 2430 int8_t ipp_use_min_mtu; 2431 }; 2432 typedef struct ip6_pkt_s ip6_pkt_t; 2433 2434 /* 2435 * This structure is used to convey information from IP and the ULP. 2436 * Currently used for the IP_RECVSLLA and IP_RECVIF options. The 2437 * type of information field is set to IN_PKTINFO (i.e inbound pkt info) 2438 */ 2439 typedef struct in_pktinfo { 2440 uint32_t in_pkt_ulp_type; /* type of info sent */ 2441 /* to UDP */ 2442 uint32_t in_pkt_flags; /* what is sent up by IP */ 2443 uint32_t in_pkt_ifindex; /* inbound interface index */ 2444 struct sockaddr_dl in_pkt_slla; /* has source link layer addr */ 2445 } in_pktinfo_t; 2446 2447 /* 2448 * flags to tell UDP what IP is sending 2449 */ 2450 #define IPF_RECVIF 0x01 /* inbound interface index */ 2451 #define IPF_RECVSLLA 0x02 /* source link layer address */ 2452 2453 /* ipp_fields values */ 2454 #define IPPF_IFINDEX 0x0001 /* Part of in6_pktinfo: ifindex */ 2455 #define IPPF_ADDR 0x0002 /* Part of in6_pktinfo: src/dst addr */ 2456 #define IPPF_SCOPE_ID 0x0004 /* Add xmit ip6i_t for sin6_scope_id */ 2457 #define IPPF_NO_CKSUM 0x0008 /* Add xmit ip6i_t for IP6I_NO_*_CKSUM */ 2458 2459 #define IPPF_RAW_CKSUM 0x0010 /* Add xmit ip6i_t for IP6I_RAW_CHECKSUM */ 2460 #define IPPF_HOPLIMIT 0x0020 2461 #define IPPF_HOPOPTS 0x0040 2462 #define IPPF_RTHDR 0x0080 2463 2464 #define IPPF_RTDSTOPTS 0x0100 2465 #define IPPF_DSTOPTS 0x0200 2466 #define IPPF_NEXTHOP 0x0400 2467 #define IPPF_PATHMTU 0x0800 2468 2469 #define IPPF_TCLASS 0x1000 2470 #define IPPF_DONTFRAG 0x2000 2471 #define IPPF_USE_MIN_MTU 0x04000 2472 #define IPPF_MULTICAST_HOPS 0x08000 2473 #define IPPF_UNICAST_HOPS 0x10000 2474 2475 #define IPPF_HAS_IP6I \ 2476 (IPPF_IFINDEX|IPPF_ADDR|IPPF_NEXTHOP|IPPF_SCOPE_ID| \ 2477 IPPF_NO_CKSUM|IPPF_RAW_CKSUM|IPPF_HOPLIMIT|IPPF_DONTFRAG| \ 2478 IPPF_USE_MIN_MTU|IPPF_MULTICAST_HOPS|IPPF_UNICAST_HOPS) 2479 2480 #define TCP_PORTS_OFFSET 0 2481 #define UDP_PORTS_OFFSET 0 2482 2483 /* 2484 * lookups return the ill/ipif only if the flags are clear OR Iam writer. 2485 * ill / ipif lookup functions increment the refcnt on the ill / ipif only 2486 * after calling these macros. This ensures that the refcnt on the ipif or 2487 * ill will eventually drop down to zero. 2488 */ 2489 #define ILL_LOOKUP_FAILED 1 /* Used as error code */ 2490 #define IPIF_LOOKUP_FAILED 2 /* Used as error code */ 2491 2492 #define ILL_CAN_LOOKUP(ill) \ 2493 (!((ill)->ill_state_flags & (ILL_CONDEMNED | ILL_CHANGING)) || \ 2494 IAM_WRITER_ILL(ill)) 2495 2496 #define ILL_CAN_WAIT(ill, q) \ 2497 (((q) != NULL) && !((ill)->ill_state_flags & (ILL_CONDEMNED))) 2498 2499 #define ILL_CAN_LOOKUP_WALKER(ill) \ 2500 (!((ill)->ill_state_flags & ILL_CONDEMNED)) 2501 2502 #define IPIF_CAN_LOOKUP(ipif) \ 2503 (!((ipif)->ipif_state_flags & (IPIF_CONDEMNED | IPIF_CHANGING)) || \ 2504 IAM_WRITER_IPIF(ipif)) 2505 2506 /* 2507 * If the parameter 'q' is NULL, the caller is not interested in wait and 2508 * restart of the operation if the ILL or IPIF cannot be looked up when it is 2509 * marked as 'CHANGING'. Typically a thread that tries to send out data will 2510 * end up passing NULLs as the last 4 parameters to ill_lookup_on_ifindex and 2511 * in this case 'q' is NULL 2512 */ 2513 #define IPIF_CAN_WAIT(ipif, q) \ 2514 (((q) != NULL) && !((ipif)->ipif_state_flags & (IPIF_CONDEMNED))) 2515 2516 #define IPIF_CAN_LOOKUP_WALKER(ipif) \ 2517 (!((ipif)->ipif_state_flags & (IPIF_CONDEMNED)) || \ 2518 IAM_WRITER_IPIF(ipif)) 2519 2520 /* 2521 * These macros are used by critical set ioctls and failover ioctls to 2522 * mark the ipif appropriately before starting the operation and to clear the 2523 * marks after completing the operation. 2524 */ 2525 #define IPIF_UNMARK_MOVING(ipif) \ 2526 (ipif)->ipif_state_flags &= ~IPIF_MOVING & ~IPIF_CHANGING; 2527 2528 #define ILL_UNMARK_CHANGING(ill) \ 2529 (ill)->ill_state_flags &= ~ILL_CHANGING; 2530 2531 /* Macros used to assert that this thread is a writer */ 2532 #define IAM_WRITER_IPSQ(ipsq) ((ipsq)->ipsq_writer == curthread) 2533 #define IAM_WRITER_ILL(ill) \ 2534 ((ill)->ill_phyint->phyint_ipsq->ipsq_writer == curthread) 2535 #define IAM_WRITER_IPIF(ipif) \ 2536 ((ipif)->ipif_ill->ill_phyint->phyint_ipsq->ipsq_writer == curthread) 2537 2538 /* 2539 * Grab ill locks in the proper order. The order is highest addressed 2540 * ill is locked first. 2541 */ 2542 #define GRAB_ILL_LOCKS(ill_1, ill_2) \ 2543 { \ 2544 if ((ill_1) > (ill_2)) { \ 2545 if (ill_1 != NULL) \ 2546 mutex_enter(&(ill_1)->ill_lock); \ 2547 if (ill_2 != NULL) \ 2548 mutex_enter(&(ill_2)->ill_lock); \ 2549 } else { \ 2550 if (ill_2 != NULL) \ 2551 mutex_enter(&(ill_2)->ill_lock); \ 2552 if (ill_1 != NULL && ill_1 != ill_2) \ 2553 mutex_enter(&(ill_1)->ill_lock); \ 2554 } \ 2555 } 2556 2557 #define RELEASE_ILL_LOCKS(ill_1, ill_2) \ 2558 { \ 2559 if (ill_1 != NULL) \ 2560 mutex_exit(&(ill_1)->ill_lock); \ 2561 if (ill_2 != NULL && ill_2 != ill_1) \ 2562 mutex_exit(&(ill_2)->ill_lock); \ 2563 } 2564 2565 /* Get the other protocol instance ill */ 2566 #define ILL_OTHER(ill) \ 2567 ((ill)->ill_isv6 ? (ill)->ill_phyint->phyint_illv4 : \ 2568 (ill)->ill_phyint->phyint_illv6) 2569 2570 #define MATCH_V4_ONLY 0x1 2571 #define MATCH_V6_ONLY 0x2 2572 #define MATCH_ILL_ONLY 0x4 2573 2574 /* ioctl command info: Ioctl properties extracted and stored in here */ 2575 typedef struct cmd_info_s 2576 { 2577 char ci_groupname[LIFNAMSIZ + 1]; /* SIOCSLIFGROUPNAME */ 2578 ipif_t *ci_ipif; /* ipif associated with [l]ifreq ioctl's */ 2579 sin_t *ci_sin; /* the sin struct passed down */ 2580 sin6_t *ci_sin6; /* the sin6_t struct passed down */ 2581 struct lifreq *ci_lifr; /* the lifreq struct passed down */ 2582 } cmd_info_t; 2583 2584 extern krwlock_t ill_g_lock; 2585 extern kmutex_t ip_addr_avail_lock; 2586 extern ipsq_t *ipsq_g_head; 2587 2588 extern ill_t *ip_timer_ill; /* ILL for IRE expiration timer. */ 2589 extern timeout_id_t ip_ire_expire_id; /* IRE expiration timeout id. */ 2590 extern timeout_id_t ip_ire_reclaim_id; /* IRE recalaim timeout id. */ 2591 2592 extern kmutex_t ip_mi_lock; 2593 extern krwlock_t ip_g_nd_lock; /* For adding/removing nd variables */ 2594 extern kmutex_t ip_trash_timer_lock; /* Protects ip_ire_expire_id */ 2595 2596 extern kmutex_t igmp_timer_lock; /* Protects the igmp timer */ 2597 extern kmutex_t mld_timer_lock; /* Protects the mld timer */ 2598 2599 extern krwlock_t ill_g_usesrc_lock; /* Protects usesrc related fields */ 2600 2601 extern struct kmem_cache *ire_cache; 2602 2603 extern uint_t ip_ire_default_count; /* Number of IPv4 IRE_DEFAULT entries */ 2604 extern uint_t ip_ire_default_index; /* Walking index used to mod in */ 2605 2606 extern ipaddr_t ip_g_all_ones; 2607 extern caddr_t ip_g_nd; /* Named Dispatch List Head */ 2608 2609 extern uint_t ip_loopback_mtu; 2610 2611 extern ipparam_t *ip_param_arr; 2612 2613 extern int ip_g_forward; 2614 extern int ipv6_forward; 2615 extern vmem_t *ip_minor_arena; 2616 2617 #define ip_respond_to_address_mask_broadcast ip_param_arr[0].ip_param_value 2618 #define ip_g_send_redirects ip_param_arr[5].ip_param_value 2619 #define ip_debug ip_param_arr[7].ip_param_value 2620 #define ip_mrtdebug ip_param_arr[8].ip_param_value 2621 #define ip_timer_interval ip_param_arr[9].ip_param_value 2622 #define ip_ire_arp_interval ip_param_arr[10].ip_param_value 2623 #define ip_def_ttl ip_param_arr[12].ip_param_value 2624 #define ip_wroff_extra ip_param_arr[14].ip_param_value 2625 #define ip_path_mtu_discovery ip_param_arr[17].ip_param_value 2626 #define ip_ignore_delete_time ip_param_arr[18].ip_param_value 2627 #define ip_output_queue ip_param_arr[20].ip_param_value 2628 #define ip_broadcast_ttl ip_param_arr[21].ip_param_value 2629 #define ip_icmp_err_interval ip_param_arr[22].ip_param_value 2630 #define ip_icmp_err_burst ip_param_arr[23].ip_param_value 2631 #define ip_reass_queue_bytes ip_param_arr[24].ip_param_value 2632 #define ip_addrs_per_if ip_param_arr[26].ip_param_value 2633 #define ipsec_override_persocket_policy ip_param_arr[27].ip_param_value 2634 #define icmp_accept_clear_messages ip_param_arr[28].ip_param_value 2635 #define delay_first_probe_time ip_param_arr[30].ip_param_value 2636 #define max_unicast_solicit ip_param_arr[31].ip_param_value 2637 #define ipv6_def_hops ip_param_arr[32].ip_param_value 2638 #define ipv6_icmp_return ip_param_arr[33].ip_param_value 2639 #define ipv6_forward_src_routed ip_param_arr[34].ip_param_value 2640 #define ipv6_resp_echo_mcast ip_param_arr[35].ip_param_value 2641 #define ipv6_send_redirects ip_param_arr[36].ip_param_value 2642 #define ipv6_ignore_redirect ip_param_arr[37].ip_param_value 2643 #define ipv6_strict_dst_multihoming ip_param_arr[38].ip_param_value 2644 #define ip_ire_reclaim_fraction ip_param_arr[39].ip_param_value 2645 #define ipsec_policy_log_interval ip_param_arr[40].ip_param_value 2646 #define ip_ndp_unsolicit_interval ip_param_arr[42].ip_param_value 2647 #define ip_ndp_unsolicit_count ip_param_arr[43].ip_param_value 2648 #define ipv6_ignore_home_address_opt ip_param_arr[44].ip_param_value 2649 #define ip_policy_mask ip_param_arr[45].ip_param_value 2650 #define ip_multirt_resolution_interval ip_param_arr[46].ip_param_value 2651 #define ip_multirt_ttl ip_param_arr[47].ip_param_value 2652 #define ip_multidata_outbound ip_param_arr[48].ip_param_value 2653 #ifdef DEBUG 2654 #define ipv6_drop_inbound_icmpv6 ip_param_arr[49].ip_param_value 2655 #else 2656 #define ipv6_drop_inbound_icmpv6 0 2657 #endif 2658 2659 extern hrtime_t ipsec_policy_failure_last; 2660 2661 extern int dohwcksum; /* use h/w cksum if supported by the h/w */ 2662 #ifdef ZC_TEST 2663 extern int noswcksum; 2664 #endif 2665 2666 extern char ipif_loopback_name[]; 2667 2668 extern nv_t *ire_nv_tbl; 2669 2670 extern time_t ip_g_frag_timeout; 2671 extern clock_t ip_g_frag_timo_ms; 2672 2673 extern mib2_ip_t ip_mib; /* For tcpInErrs and udpNoPorts */ 2674 2675 extern struct module_info ip_mod_info; 2676 2677 extern timeout_id_t igmp_slowtimeout_id; 2678 extern timeout_id_t mld_slowtimeout_id; 2679 2680 extern uint_t loopback_packets; 2681 2682 /* 2683 * Network byte order macros 2684 */ 2685 #ifdef _BIG_ENDIAN 2686 #define N_IN_CLASSD_NET IN_CLASSD_NET 2687 #define N_INADDR_UNSPEC_GROUP INADDR_UNSPEC_GROUP 2688 #else /* _BIG_ENDIAN */ 2689 #define N_IN_CLASSD_NET (ipaddr_t)0x000000f0U 2690 #define N_INADDR_UNSPEC_GROUP (ipaddr_t)0x000000e0U 2691 #endif /* _BIG_ENDIAN */ 2692 #define CLASSD(addr) (((addr) & N_IN_CLASSD_NET) == N_INADDR_UNSPEC_GROUP) 2693 2694 #ifdef DEBUG 2695 /* IPsec HW acceleration debugging support */ 2696 2697 #define IPSECHW_CAPAB 0x0001 /* capability negotiation */ 2698 #define IPSECHW_SADB 0x0002 /* SADB exchange */ 2699 #define IPSECHW_PKT 0x0004 /* general packet flow */ 2700 #define IPSECHW_PKTIN 0x0008 /* driver in pkt processing details */ 2701 #define IPSECHW_PKTOUT 0x0010 /* driver out pkt processing details */ 2702 2703 #define IPSECHW_DEBUG(f, x) if (ipsechw_debug & (f)) { (void) printf x; } 2704 #define IPSECHW_CALL(f, r, x) if (ipsechw_debug & (f)) { (void) r x; } 2705 2706 extern uint32_t ipsechw_debug; 2707 #else 2708 #define IPSECHW_DEBUG(f, x) {} 2709 #define IPSECHW_CALL(f, r, x) {} 2710 #endif 2711 2712 #ifdef IP_DEBUG 2713 #include <sys/debug.h> 2714 #include <sys/promif.h> 2715 2716 #define ip0dbg(a) printf a 2717 #define ip1dbg(a) if (ip_debug > 2) printf a 2718 #define ip2dbg(a) if (ip_debug > 3) printf a 2719 #define ip3dbg(a) if (ip_debug > 4) printf a 2720 #else 2721 #define ip0dbg(a) /* */ 2722 #define ip1dbg(a) /* */ 2723 #define ip2dbg(a) /* */ 2724 #define ip3dbg(a) /* */ 2725 #endif /* IP_DEBUG */ 2726 2727 extern const char *dlpi_prim_str(int); 2728 extern const char *dlpi_err_str(int); 2729 extern void ill_frag_timer(void *); 2730 extern ill_t *ill_first(int, int, ill_walk_context_t *); 2731 extern ill_t *ill_next(ill_walk_context_t *, ill_t *); 2732 extern void ill_frag_timer_start(ill_t *); 2733 extern mblk_t *ip_carve_mp(mblk_t **, ssize_t); 2734 extern mblk_t *ip_dlpi_alloc(size_t, t_uscalar_t); 2735 extern char *ip_dot_addr(ipaddr_t, char *); 2736 extern void ip_lwput(queue_t *, mblk_t *); 2737 extern boolean_t icmp_err_rate_limit(void); 2738 extern void icmp_time_exceeded(queue_t *, mblk_t *, uint8_t); 2739 extern void icmp_unreachable(queue_t *, mblk_t *, uint8_t); 2740 extern mblk_t *ip_add_info(mblk_t *, ill_t *, uint_t); 2741 extern mblk_t *ip_bind_v4(queue_t *, mblk_t *, conn_t *); 2742 extern int ip_bind_connected(conn_t *, mblk_t *, ipaddr_t *, uint16_t, 2743 ipaddr_t, uint16_t, boolean_t, boolean_t, boolean_t, 2744 boolean_t); 2745 extern boolean_t ip_bind_ipsec_policy_set(conn_t *, mblk_t *); 2746 extern int ip_bind_laddr(conn_t *, mblk_t *, ipaddr_t, uint16_t, 2747 boolean_t, boolean_t, boolean_t); 2748 extern uint_t ip_cksum(mblk_t *, int, uint32_t); 2749 extern int ip_close(queue_t *, int); 2750 extern uint16_t ip_csum_hdr(ipha_t *); 2751 extern void ip_proto_not_sup(queue_t *, mblk_t *, uint_t, zoneid_t); 2752 extern void ip_ire_fini(void); 2753 extern void ip_ire_init(void); 2754 extern int ip_open(queue_t *, dev_t *, int, int, cred_t *); 2755 extern int ip_reassemble(mblk_t *, ipf_t *, uint_t, boolean_t, ill_t *, 2756 size_t); 2757 extern int ip_opt_set_ill(conn_t *, int, boolean_t, boolean_t, 2758 int, int, mblk_t *); 2759 extern void ip_rput(queue_t *, mblk_t *); 2760 extern void ip_input(ill_t *, ill_rx_ring_t *, mblk_t *, size_t); 2761 extern void ip_rput_dlpi(queue_t *, mblk_t *); 2762 extern void ip_rput_forward(ire_t *, ipha_t *, mblk_t *, ill_t *); 2763 extern void ip_rput_forward_multicast(ipaddr_t, mblk_t *, ipif_t *); 2764 2765 extern int ip_snmpmod_close(queue_t *); 2766 extern void ip_snmpmod_wput(queue_t *, mblk_t *); 2767 extern void ip_udp_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *); 2768 extern void ip_proto_input(queue_t *, mblk_t *, ipha_t *, ire_t *, ill_t *); 2769 extern void ip_rput_other(ipsq_t *, queue_t *, mblk_t *, void *); 2770 extern void ip_setqinfo(queue_t *, minor_t, boolean_t); 2771 extern void ip_trash_ire_reclaim(void *); 2772 extern void ip_trash_timer_expire(void *); 2773 extern void ip_wput(queue_t *, mblk_t *); 2774 extern void ip_output(void *, mblk_t *, void *, int); 2775 extern void ip_wput_md(queue_t *, mblk_t *, conn_t *); 2776 2777 extern void ip_wput_ire(queue_t *, mblk_t *, ire_t *, conn_t *, int); 2778 extern void ip_wput_local(queue_t *, ill_t *, ipha_t *, mblk_t *, ire_t *, 2779 int, zoneid_t); 2780 extern void ip_wput_multicast(queue_t *, mblk_t *, ipif_t *); 2781 extern void ip_wput_nondata(ipsq_t *, queue_t *, mblk_t *, void *); 2782 extern void ip_wsrv(queue_t *); 2783 extern char *ip_nv_lookup(nv_t *, int); 2784 extern boolean_t ip_local_addr_ok_v6(const in6_addr_t *, const in6_addr_t *); 2785 extern boolean_t ip_remote_addr_ok_v6(const in6_addr_t *, const in6_addr_t *); 2786 extern ipaddr_t ip_massage_options(ipha_t *); 2787 extern ipaddr_t ip_net_mask(ipaddr_t); 2788 extern void ip_newroute(queue_t *, mblk_t *, ipaddr_t, ill_t *, conn_t *); 2789 2790 extern struct qinit rinit_ipv6; 2791 extern struct qinit winit_ipv6; 2792 extern struct qinit rinit_tcp; 2793 extern struct qinit rinit_tcp6; 2794 extern struct qinit winit_tcp; 2795 extern struct qinit rinit_acceptor_tcp; 2796 extern struct qinit winit_acceptor_tcp; 2797 2798 extern void conn_drain_insert(conn_t *connp); 2799 extern int conn_ipsec_length(conn_t *connp); 2800 extern void ip_wput_ipsec_out(queue_t *, mblk_t *, ipha_t *, ill_t *, 2801 ire_t *); 2802 extern ipaddr_t ip_get_dst(ipha_t *); 2803 extern int ipsec_out_extra_length(mblk_t *); 2804 extern int ipsec_in_extra_length(mblk_t *); 2805 extern mblk_t *ipsec_in_alloc(); 2806 extern boolean_t ipsec_in_is_secure(mblk_t *); 2807 extern void ipsec_out_process(queue_t *, mblk_t *, ire_t *, uint_t); 2808 extern void ipsec_out_to_in(mblk_t *); 2809 extern int ill_forward_set(queue_t *, mblk_t *, boolean_t, caddr_t); 2810 extern void ip_fanout_proto_again(mblk_t *, ill_t *, ill_t *, ire_t *); 2811 2812 extern void ire_cleanup(ire_t *); 2813 extern void ire_inactive(ire_t *); 2814 extern ire_t *ire_unlink(irb_t *); 2815 #ifdef IRE_DEBUG 2816 extern void ire_trace_ref(ire_t *ire); 2817 extern void ire_untrace_ref(ire_t *ire); 2818 extern void ire_thread_exit(ire_t *ire, caddr_t); 2819 #endif 2820 #ifdef ILL_DEBUG 2821 extern void ill_trace_cleanup(ill_t *); 2822 extern void ipif_trace_cleanup(ipif_t *); 2823 #endif 2824 2825 extern int ip_srcid_insert(const in6_addr_t *, zoneid_t); 2826 extern int ip_srcid_remove(const in6_addr_t *, zoneid_t); 2827 extern void ip_srcid_find_id(uint_t, in6_addr_t *, zoneid_t); 2828 extern uint_t ip_srcid_find_addr(const in6_addr_t *, zoneid_t); 2829 extern int ip_srcid_report(queue_t *, mblk_t *, caddr_t, cred_t *); 2830 2831 extern uint8_t ipoptp_next(ipoptp_t *); 2832 extern uint8_t ipoptp_first(ipoptp_t *, ipha_t *); 2833 extern ill_t *ip_grab_attach_ill(ill_t *, mblk_t *, int, boolean_t); 2834 extern ire_t *conn_set_outgoing_ill(conn_t *, ire_t *, ill_t **); 2835 extern int ipsec_req_from_conn(conn_t *, ipsec_req_t *, int); 2836 extern int ip_snmp_get(queue_t *q, mblk_t *mctl); 2837 extern int ip_snmp_set(queue_t *q, int, int, uchar_t *, int); 2838 extern void ip_process_ioctl(ipsq_t *, queue_t *, mblk_t *, void *); 2839 extern void ip_quiesce_conn(conn_t *); 2840 extern void ip_reprocess_ioctl(ipsq_t *, queue_t *, mblk_t *, void *); 2841 extern void ip_restart_optmgmt(ipsq_t *, queue_t *, mblk_t *, void *); 2842 extern void ip_ioctl_finish(queue_t *, mblk_t *, int, int, ipif_t *, 2843 ipsq_t *); 2844 2845 extern boolean_t ipsq_pending_mp_cleanup(ill_t *, conn_t *); 2846 extern void conn_ioctl_cleanup(conn_t *); 2847 extern ill_t *conn_get_held_ill(conn_t *, ill_t **, int *); 2848 2849 struct multidata_s; 2850 struct pdesc_s; 2851 2852 extern mblk_t *ip_mdinfo_alloc(ill_mdt_capab_t *); 2853 extern mblk_t *ip_mdinfo_return(ire_t *, conn_t *, char *, ill_mdt_capab_t *); 2854 extern uint_t ip_md_cksum(struct pdesc_s *, int, uint_t); 2855 extern boolean_t ip_md_addr_attr(struct multidata_s *, struct pdesc_s *, 2856 const mblk_t *); 2857 extern boolean_t ip_md_hcksum_attr(struct multidata_s *, struct pdesc_s *, 2858 uint32_t, uint32_t, uint32_t, uint32_t); 2859 extern boolean_t ip_md_zcopy_attr(struct multidata_s *, struct pdesc_s *, 2860 uint_t); 2861 extern mblk_t *ip_unbind(queue_t *, mblk_t *); 2862 2863 /* Hooks for CGTP (multirt routes) filtering module */ 2864 #define CGTP_FILTER_REV_1 1 2865 #define CGTP_FILTER_REV_2 2 2866 #define CGTP_FILTER_REV CGTP_FILTER_REV_2 2867 2868 /* cfo_filter, cfo_filter_fp, cfo_filter_v6 hooks return values */ 2869 #define CGTP_IP_PKT_NOT_CGTP 0 2870 #define CGTP_IP_PKT_PREMIUM 1 2871 #define CGTP_IP_PKT_DUPLICATE 2 2872 2873 typedef struct cgtp_filter_ops { 2874 int cfo_filter_rev; 2875 int (*cfo_change_state)(int); 2876 int (*cfo_add_dest_v4)(ipaddr_t, ipaddr_t, ipaddr_t, ipaddr_t); 2877 int (*cfo_del_dest_v4)(ipaddr_t, ipaddr_t); 2878 int (*cfo_add_dest_v6)(in6_addr_t *, in6_addr_t *, in6_addr_t *, 2879 in6_addr_t *); 2880 int (*cfo_del_dest_v6)(in6_addr_t *, in6_addr_t *); 2881 int (*cfo_filter)(queue_t *, mblk_t *); 2882 int (*cfo_filter_fp)(queue_t *, mblk_t *); 2883 int (*cfo_filter_v6)(queue_t *, ip6_t *, ip6_frag_t *); 2884 } cgtp_filter_ops_t; 2885 2886 #define CGTP_MCAST_SUCCESS 1 2887 2888 extern cgtp_filter_ops_t *ip_cgtp_filter_ops; 2889 extern boolean_t ip_cgtp_filter; 2890 2891 extern int ip_cgtp_filter_supported(void); 2892 extern int ip_cgtp_filter_register(cgtp_filter_ops_t *); 2893 2894 /* Flags for ire_multirt_lookup() */ 2895 2896 #define MULTIRT_USESTAMP 0x0001 2897 #define MULTIRT_SETSTAMP 0x0002 2898 #define MULTIRT_CACHEGW 0x0004 2899 2900 /* Debug stuff for multirt route resolution. */ 2901 #if defined(DEBUG) && !defined(__lint) 2902 /* Our "don't send, rather drop" flag. */ 2903 #define MULTIRT_DEBUG_FLAG 0x8000 2904 2905 #define MULTIRT_TRACE(x) ip2dbg(x) 2906 2907 #define MULTIRT_DEBUG_TAG(mblk) \ 2908 do { \ 2909 ASSERT(mblk != NULL); \ 2910 MULTIRT_TRACE(("%s[%d]: tagging mblk %p, tag was %d\n", \ 2911 __FILE__, __LINE__, \ 2912 (void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \ 2913 (mblk)->b_flag |= MULTIRT_DEBUG_FLAG; \ 2914 } while (0) 2915 2916 #define MULTIRT_DEBUG_UNTAG(mblk) \ 2917 do { \ 2918 ASSERT(mblk != NULL); \ 2919 MULTIRT_TRACE(("%s[%d]: untagging mblk %p, tag was %d\n", \ 2920 __FILE__, __LINE__, \ 2921 (void *)(mblk), (mblk)->b_flag & MULTIRT_DEBUG_FLAG)); \ 2922 (mblk)->b_flag &= ~MULTIRT_DEBUG_FLAG; \ 2923 } while (0) 2924 2925 #define MULTIRT_DEBUG_TAGGED(mblk) \ 2926 (((mblk)->b_flag & MULTIRT_DEBUG_FLAG) ? B_TRUE : B_FALSE) 2927 #else 2928 #define MULTIRT_DEBUG_TAG(mblk) ASSERT(mblk != NULL) 2929 #define MULTIRT_DEBUG_UNTAG(mblk) ASSERT(mblk != NULL) 2930 #define MULTIRT_DEBUG_TAGGED(mblk) B_FALSE 2931 #endif 2932 2933 /* 2934 * Per-ILL Multidata Transmit capabilities. 2935 */ 2936 struct ill_mdt_capab_s { 2937 uint_t ill_mdt_version; /* interface version */ 2938 uint_t ill_mdt_on; /* on/off switch for MDT on this ILL */ 2939 uint_t ill_mdt_hdr_head; /* leading header fragment extra space */ 2940 uint_t ill_mdt_hdr_tail; /* trailing header fragment extra space */ 2941 uint_t ill_mdt_max_pld; /* maximum payload buffers per Multidata */ 2942 uint_t ill_mdt_span_limit; /* maximum payload span per packet */ 2943 }; 2944 2945 struct ill_hcksum_capab_s { 2946 uint_t ill_hcksum_version; /* interface version */ 2947 uint_t ill_hcksum_txflags; /* capabilities on transmit */ 2948 }; 2949 2950 struct ill_zerocopy_capab_s { 2951 uint_t ill_zerocopy_version; /* interface version */ 2952 uint_t ill_zerocopy_flags; /* capabilities */ 2953 }; 2954 2955 /* Possible ill_states */ 2956 #define ILL_RING_INPROC 3 /* Being assigned to squeue */ 2957 #define ILL_RING_INUSE 2 /* Already Assigned to Rx Ring */ 2958 #define ILL_RING_BEING_FREED 1 /* Being Unassigned */ 2959 #define ILL_RING_FREE 0 /* Available to be assigned to Ring */ 2960 2961 #define ILL_MAX_RINGS 256 /* Max num of rx rings we can manage */ 2962 #define ILL_POLLING 0x01 /* Polling in use */ 2963 2964 /* 2965 * This function pointer type is exported by the mac layer. 2966 * we need to duplicate the definition here because we cannot 2967 * include mac.h in this file. 2968 */ 2969 typedef void (*ip_mac_blank_t)(void *, time_t, uint_t); 2970 2971 struct ill_rx_ring { 2972 ip_mac_blank_t rr_blank; /* Driver interrupt blanking func */ 2973 void *rr_handle; /* Handle for Rx ring */ 2974 squeue_t *rr_sqp; /* Squeue the ring is bound to */ 2975 ill_t *rr_ill; /* back pointer to ill */ 2976 clock_t rr_poll_time; /* Last lbolt polling was used */ 2977 uint32_t rr_poll_state; /* polling state flags */ 2978 uint32_t rr_max_blank_time; /* Max interrupt blank */ 2979 uint32_t rr_min_blank_time; /* Min interrupt blank */ 2980 uint32_t rr_max_pkt_cnt; /* Max pkts before interrupt */ 2981 uint32_t rr_min_pkt_cnt; /* Mix pkts before interrupt */ 2982 uint32_t rr_normal_blank_time; /* Normal intr freq */ 2983 uint32_t rr_normal_pkt_cnt; /* Normal intr pkt cnt */ 2984 uint32_t rr_ring_state; /* State of this ring */ 2985 }; 2986 2987 /* 2988 * This is exported by dld and is meant to be invoked from a ULP. 2989 */ 2990 typedef void (*ip_dld_tx_t)(void *, mblk_t *); 2991 2992 struct ill_poll_capab_s { 2993 ip_dld_tx_t ill_tx; /* dld-supplied tx routine */ 2994 void *ill_tx_handle; /* dld-supplied tx handle */ 2995 ill_rx_ring_t *ill_ring_tbl; /* Ring to Sqp mapping table */ 2996 conn_t *ill_unbind_conn; /* Conn used during unplumb */ 2997 }; 2998 2999 /* 3000 * IP squeues exports 3001 */ 3002 extern int ip_squeue_profile; 3003 extern int ip_squeue_bind; 3004 extern boolean_t ip_squeue_fanout; 3005 3006 typedef struct squeue_set_s { 3007 kmutex_t sqs_lock; 3008 struct squeue_s **sqs_list; 3009 int sqs_size; 3010 int sqs_max_size; 3011 processorid_t sqs_bind; 3012 } squeue_set_t; 3013 3014 #define IP_SQUEUE_GET(hint) \ 3015 (!ip_squeue_fanout ? \ 3016 (CPU->cpu_squeue_set->sqs_list[hint % \ 3017 CPU->cpu_squeue_set->sqs_size]) : \ 3018 ip_squeue_random(hint)) 3019 3020 typedef void (*squeue_func_t)(squeue_t *, mblk_t *, sqproc_t, void *, uint8_t); 3021 3022 extern void ip_squeue_init(void (*)(squeue_t *)); 3023 extern squeue_t *ip_squeue_random(uint_t); 3024 extern squeue_t *ip_squeue_get(ill_rx_ring_t *); 3025 extern void ip_squeue_get_pkts(squeue_t *); 3026 extern int ip_squeue_bind_set(queue_t *, mblk_t *, char *, caddr_t, cred_t *); 3027 extern int ip_squeue_bind_get(queue_t *, mblk_t *, caddr_t, cred_t *); 3028 extern void ip_squeue_clean(void *, mblk_t *, void *); 3029 extern void ip_resume_tcp_bind(void *, mblk_t *, void *); 3030 3031 extern void tcp_wput(queue_t *, mblk_t *); 3032 3033 extern int ip_fill_mtuinfo(struct in6_addr *, in_port_t, 3034 struct ip6_mtuinfo *); 3035 extern ipif_t *conn_get_held_ipif(conn_t *, ipif_t **, int *); 3036 3037 typedef void (*ipsq_func_t)(ipsq_t *, queue_t *, mblk_t *, void *); 3038 3039 /* 3040 * Squeue tags. Tags only need to be unique when the callback function is the 3041 * same to distinguish between different calls, but we use unique tags for 3042 * convenience anyway. 3043 */ 3044 #define SQTAG_IP_INPUT 1 3045 #define SQTAG_TCP_INPUT_ICMP_ERR 2 3046 #define SQTAG_TCP6_INPUT_ICMP_ERR 3 3047 #define SQTAG_IP_TCP_INPUT 4 3048 #define SQTAG_IP6_TCP_INPUT 5 3049 #define SQTAG_IP_TCP_CLOSE 6 3050 #define SQTAG_TCP_OUTPUT 7 3051 #define SQTAG_TCP_TIMER 8 3052 #define SQTAG_TCP_TIMEWAIT 9 3053 #define SQTAG_TCP_ACCEPT_FINISH 10 3054 #define SQTAG_TCP_ACCEPT_FINISH_Q0 11 3055 #define SQTAG_TCP_ACCEPT_PENDING 12 3056 #define SQTAG_TCP_LISTEN_DISCON 13 3057 #define SQTAG_TCP_CONN_REQ 14 3058 #define SQTAG_TCP_EAGER_BLOWOFF 15 3059 #define SQTAG_TCP_EAGER_CLEANUP 16 3060 #define SQTAG_TCP_EAGER_CLEANUP_Q0 17 3061 #define SQTAG_TCP_CONN_IND 18 3062 #define SQTAG_TCP_RSRV 19 3063 #define SQTAG_TCP_ABORT_BUCKET 20 3064 #define SQTAG_TCP_REINPUT 21 3065 #define SQTAG_TCP_REINPUT_EAGER 22 3066 #define SQTAG_TCP_INPUT_MCTL 23 3067 #define SQTAG_TCP_RPUTOTHER 24 3068 #define SQTAG_IP_PROTO_AGAIN 25 3069 #define SQTAG_IP_FANOUT_TCP 26 3070 #define SQTAG_IPSQ_CLEAN_RING 27 3071 #define SQTAG_TCP_WPUT_OTHER 28 3072 #define SQTAG_TCP_CONN_REQ_UNBOUND 29 3073 #define SQTAG_TCP_SEND_PENDING 30 3074 #define SQTAG_BIND_RETRY 31 3075 #define SQTAG_UDP_FANOUT 32 3076 #define SQTAG_UDP_INPUT 33 3077 #define SQTAG_UDP_WPUT 34 3078 #define SQTAG_UDP_OUTPUT 35 3079 #define SQTAG_TCP_KSSL_INPUT 36 3080 3081 #endif /* _KERNEL */ 3082 3083 #ifdef __cplusplus 3084 } 3085 #endif 3086 3087 #endif /* _INET_IP_H */ 3088