17c478bd9Sstevel@tonic-gate /* 27c478bd9Sstevel@tonic-gate * CDDL HEADER START 37c478bd9Sstevel@tonic-gate * 47c478bd9Sstevel@tonic-gate * The contents of this file are subject to the terms of the 545916cd2Sjpk * Common Development and Distribution License (the "License"). 645916cd2Sjpk * You may not use this file except in compliance with the License. 77c478bd9Sstevel@tonic-gate * 87c478bd9Sstevel@tonic-gate * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 97c478bd9Sstevel@tonic-gate * or http://www.opensolaris.org/os/licensing. 107c478bd9Sstevel@tonic-gate * See the License for the specific language governing permissions 117c478bd9Sstevel@tonic-gate * and limitations under the License. 127c478bd9Sstevel@tonic-gate * 137c478bd9Sstevel@tonic-gate * When distributing Covered Code, include this CDDL HEADER in each 147c478bd9Sstevel@tonic-gate * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 157c478bd9Sstevel@tonic-gate * If applicable, add the following below this CDDL HEADER, with the 167c478bd9Sstevel@tonic-gate * fields enclosed by brackets "[]" replaced with your own identifying 177c478bd9Sstevel@tonic-gate * information: Portions Copyright [yyyy] [name of copyright owner] 187c478bd9Sstevel@tonic-gate * 197c478bd9Sstevel@tonic-gate * CDDL HEADER END 207c478bd9Sstevel@tonic-gate */ 217c478bd9Sstevel@tonic-gate /* 2242ee8d71Sja97890 * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 237c478bd9Sstevel@tonic-gate * Use is subject to license terms. 247c478bd9Sstevel@tonic-gate */ 257c478bd9Sstevel@tonic-gate /* Copyright (c) 1990 Mentat Inc. */ 267c478bd9Sstevel@tonic-gate 277c478bd9Sstevel@tonic-gate #pragma ident "%Z%%M% %I% %E% SMI" 287c478bd9Sstevel@tonic-gate 297c478bd9Sstevel@tonic-gate 307c478bd9Sstevel@tonic-gate /* 317c478bd9Sstevel@tonic-gate * This file contains routines that manipulate Internet Routing Entries (IREs). 327c478bd9Sstevel@tonic-gate */ 337c478bd9Sstevel@tonic-gate 347c478bd9Sstevel@tonic-gate #include <sys/types.h> 357c478bd9Sstevel@tonic-gate #include <sys/stream.h> 367c478bd9Sstevel@tonic-gate #include <sys/stropts.h> 377c478bd9Sstevel@tonic-gate #include <sys/ddi.h> 387c478bd9Sstevel@tonic-gate #include <sys/cmn_err.h> 397c478bd9Sstevel@tonic-gate #include <sys/policy.h> 407c478bd9Sstevel@tonic-gate 417c478bd9Sstevel@tonic-gate #include <sys/systm.h> 427c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 437c478bd9Sstevel@tonic-gate #include <sys/param.h> 447c478bd9Sstevel@tonic-gate #include <sys/socket.h> 457c478bd9Sstevel@tonic-gate #include <net/if.h> 467c478bd9Sstevel@tonic-gate #include <net/route.h> 477c478bd9Sstevel@tonic-gate #include <netinet/in.h> 487c478bd9Sstevel@tonic-gate #include <net/if_dl.h> 497c478bd9Sstevel@tonic-gate #include <netinet/ip6.h> 507c478bd9Sstevel@tonic-gate #include <netinet/icmp6.h> 517c478bd9Sstevel@tonic-gate 527c478bd9Sstevel@tonic-gate #include <inet/common.h> 537c478bd9Sstevel@tonic-gate #include <inet/mi.h> 547c478bd9Sstevel@tonic-gate #include <inet/ip.h> 557c478bd9Sstevel@tonic-gate #include <inet/ip6.h> 567c478bd9Sstevel@tonic-gate #include <inet/ip_ndp.h> 57c793af95Ssangeeta #include <inet/arp.h> 587c478bd9Sstevel@tonic-gate #include <inet/ip_if.h> 597c478bd9Sstevel@tonic-gate #include <inet/ip_ire.h> 60c793af95Ssangeeta #include <inet/ip_ftable.h> 617c478bd9Sstevel@tonic-gate #include <inet/ip_rts.h> 627c478bd9Sstevel@tonic-gate #include <inet/nd.h> 637c478bd9Sstevel@tonic-gate 647c478bd9Sstevel@tonic-gate #include <net/pfkeyv2.h> 657c478bd9Sstevel@tonic-gate #include <inet/ipsec_info.h> 667c478bd9Sstevel@tonic-gate #include <inet/sadb.h> 677c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 687c478bd9Sstevel@tonic-gate #include <inet/tcp.h> 697c478bd9Sstevel@tonic-gate #include <inet/ipclassifier.h> 707c478bd9Sstevel@tonic-gate #include <sys/zone.h> 71f4b3ec61Sdh155122 #include <sys/cpuvar.h> 72f4b3ec61Sdh155122 7345916cd2Sjpk #include <sys/tsol/label.h> 7445916cd2Sjpk #include <sys/tsol/tnet.h> 7545916cd2Sjpk 76c793af95Ssangeeta struct kmem_cache *rt_entry_cache; 77c793af95Ssangeeta 787c478bd9Sstevel@tonic-gate /* 797c478bd9Sstevel@tonic-gate * Synchronization notes: 807c478bd9Sstevel@tonic-gate * 817c478bd9Sstevel@tonic-gate * The fields of the ire_t struct are protected in the following way : 827c478bd9Sstevel@tonic-gate * 837c478bd9Sstevel@tonic-gate * ire_next/ire_ptpn 847c478bd9Sstevel@tonic-gate * 857c478bd9Sstevel@tonic-gate * - bucket lock of the respective tables (cache or forwarding tables). 867c478bd9Sstevel@tonic-gate * 877c478bd9Sstevel@tonic-gate * ire_mp, ire_rfq, ire_stq, ire_u *except* ire_gateway_addr[v6], ire_mask, 887c478bd9Sstevel@tonic-gate * ire_type, ire_create_time, ire_masklen, ire_ipversion, ire_flags, ire_ipif, 897c478bd9Sstevel@tonic-gate * ire_ihandle, ire_phandle, ire_nce, ire_bucket, ire_in_ill, ire_in_src_addr 907c478bd9Sstevel@tonic-gate * 917c478bd9Sstevel@tonic-gate * - Set in ire_create_v4/v6 and never changes after that. Thus, 927c478bd9Sstevel@tonic-gate * we don't need a lock whenever these fields are accessed. 937c478bd9Sstevel@tonic-gate * 947c478bd9Sstevel@tonic-gate * - ire_bucket and ire_masklen (also set in ire_create) is set in 957c478bd9Sstevel@tonic-gate * ire_add_v4/ire_add_v6 before inserting in the bucket and never 967c478bd9Sstevel@tonic-gate * changes after that. Thus we don't need a lock whenever these 977c478bd9Sstevel@tonic-gate * fields are accessed. 987c478bd9Sstevel@tonic-gate * 997c478bd9Sstevel@tonic-gate * ire_gateway_addr_v4[v6] 1007c478bd9Sstevel@tonic-gate * 1017c478bd9Sstevel@tonic-gate * - ire_gateway_addr_v4[v6] is set during ire_create and later modified 1027c478bd9Sstevel@tonic-gate * by rts_setgwr[v6]. As ire_gateway_addr is a uint32_t, updates to 1037c478bd9Sstevel@tonic-gate * it assumed to be atomic and hence the other parts of the code 1047c478bd9Sstevel@tonic-gate * does not use any locks. ire_gateway_addr_v6 updates are not atomic 1057c478bd9Sstevel@tonic-gate * and hence any access to it uses ire_lock to get/set the right value. 1067c478bd9Sstevel@tonic-gate * 1077c478bd9Sstevel@tonic-gate * ire_ident, ire_refcnt 1087c478bd9Sstevel@tonic-gate * 1097c478bd9Sstevel@tonic-gate * - Updated atomically using atomic_add_32 1107c478bd9Sstevel@tonic-gate * 1117c478bd9Sstevel@tonic-gate * ire_ssthresh, ire_rtt_sd, ire_rtt, ire_ib_pkt_count, ire_ob_pkt_count 1127c478bd9Sstevel@tonic-gate * 1137c478bd9Sstevel@tonic-gate * - Assumes that 32 bit writes are atomic. No locks. ire_lock is 1147c478bd9Sstevel@tonic-gate * used to serialize updates to ire_ssthresh, ire_rtt_sd, ire_rtt. 1157c478bd9Sstevel@tonic-gate * 1167c478bd9Sstevel@tonic-gate * ire_max_frag, ire_frag_flag 1177c478bd9Sstevel@tonic-gate * 1187c478bd9Sstevel@tonic-gate * - ire_lock is used to set/read both of them together. 1197c478bd9Sstevel@tonic-gate * 1207c478bd9Sstevel@tonic-gate * ire_tire_mark 1217c478bd9Sstevel@tonic-gate * 1227c478bd9Sstevel@tonic-gate * - Set in ire_create and updated in ire_expire, which is called 1237c478bd9Sstevel@tonic-gate * by only one function namely ip_trash_timer_expire. Thus only 1247c478bd9Sstevel@tonic-gate * one function updates and examines the value. 1257c478bd9Sstevel@tonic-gate * 1267c478bd9Sstevel@tonic-gate * ire_marks 1277c478bd9Sstevel@tonic-gate * - bucket lock protects this. 1287c478bd9Sstevel@tonic-gate * 1297c478bd9Sstevel@tonic-gate * ire_ipsec_overhead/ire_ll_hdr_length 1307c478bd9Sstevel@tonic-gate * 1317c478bd9Sstevel@tonic-gate * - Place holder for returning the information to the upper layers 1327c478bd9Sstevel@tonic-gate * when IRE_DB_REQ comes down. 1337c478bd9Sstevel@tonic-gate * 1347c478bd9Sstevel@tonic-gate * 1357c478bd9Sstevel@tonic-gate * ipv6_ire_default_count is protected by the bucket lock of 1367c478bd9Sstevel@tonic-gate * ip_forwarding_table_v6[0][0]. 1377c478bd9Sstevel@tonic-gate * 138c793af95Ssangeeta * ipv6_ire_default_index is not protected as it is just a hint 139c793af95Ssangeeta * at which default gateway to use. There is nothing 1407c478bd9Sstevel@tonic-gate * wrong in using the same gateway for two different connections. 1417c478bd9Sstevel@tonic-gate * 1427c478bd9Sstevel@tonic-gate * As we always hold the bucket locks in all the places while accessing 1437c478bd9Sstevel@tonic-gate * the above values, it is natural to use them for protecting them. 1447c478bd9Sstevel@tonic-gate * 1457c478bd9Sstevel@tonic-gate * We have a separate cache table and forwarding table for IPv4 and IPv6. 1467c478bd9Sstevel@tonic-gate * Cache table (ip_cache_table/ip_cache_table_v6) is a pointer to an 147219e6466Ssowmini * array of irb_t structures. The IPv6 forwarding table 148219e6466Ssowmini * (ip_forwarding_table_v6) is an array of pointers to arrays of irb_t 149f4b3ec61Sdh155122 * structure. ip_forwarding_table_v6 is allocated dynamically in 150f4b3ec61Sdh155122 * ire_add_v6. ire_ft_init_lock is used to serialize multiple threads 1517c478bd9Sstevel@tonic-gate * initializing the same bucket. Once a bucket is initialized, it is never 152f4b3ec61Sdh155122 * de-alloacted. This assumption enables us to access 153f4b3ec61Sdh155122 * ip_forwarding_table_v6[i] without any locks. 1547c478bd9Sstevel@tonic-gate * 155219e6466Ssowmini * The forwarding table for IPv4 is a radix tree whose leaves 156219e6466Ssowmini * are rt_entry structures containing the irb_t for the rt_dst. The irb_t 157219e6466Ssowmini * for IPv4 is dynamically allocated and freed. 158219e6466Ssowmini * 1597c478bd9Sstevel@tonic-gate * Each irb_t - ire bucket structure has a lock to protect 1607c478bd9Sstevel@tonic-gate * a bucket and the ires residing in the bucket have a back pointer to 1617c478bd9Sstevel@tonic-gate * the bucket structure. It also has a reference count for the number 1627c478bd9Sstevel@tonic-gate * of threads walking the bucket - irb_refcnt which is bumped up 1637c478bd9Sstevel@tonic-gate * using the macro IRB_REFHOLD macro. The flags irb_flags can be 1647c478bd9Sstevel@tonic-gate * set to IRE_MARK_CONDEMNED indicating that there are some ires 1657c478bd9Sstevel@tonic-gate * in this bucket that are marked with IRE_MARK_CONDEMNED and the 1667c478bd9Sstevel@tonic-gate * last thread to leave the bucket should delete the ires. Usually 1677c478bd9Sstevel@tonic-gate * this is done by the IRB_REFRELE macro which is used to decrement 168219e6466Ssowmini * the reference count on a bucket. See comments above irb_t structure 169219e6466Ssowmini * definition in ip.h for further details. 1707c478bd9Sstevel@tonic-gate * 1717c478bd9Sstevel@tonic-gate * IRE_REFHOLD/IRE_REFRELE macros operate on the ire which increments/ 1727c478bd9Sstevel@tonic-gate * decrements the reference count, ire_refcnt, atomically on the ire. 1737c478bd9Sstevel@tonic-gate * ire_refcnt is modified only using this macro. Operations on the IRE 1747c478bd9Sstevel@tonic-gate * could be described as follows : 1757c478bd9Sstevel@tonic-gate * 1767c478bd9Sstevel@tonic-gate * CREATE an ire with reference count initialized to 1. 1777c478bd9Sstevel@tonic-gate * 1787c478bd9Sstevel@tonic-gate * ADDITION of an ire holds the bucket lock, checks for duplicates 1797c478bd9Sstevel@tonic-gate * and then adds the ire. ire_add_v4/ire_add_v6 returns the ire after 1807c478bd9Sstevel@tonic-gate * bumping up once more i.e the reference count is 2. This is to avoid 1817c478bd9Sstevel@tonic-gate * an extra lookup in the functions calling ire_add which wants to 1827c478bd9Sstevel@tonic-gate * work with the ire after adding. 1837c478bd9Sstevel@tonic-gate * 1847c478bd9Sstevel@tonic-gate * LOOKUP of an ire bumps up the reference count using IRE_REFHOLD 1857c478bd9Sstevel@tonic-gate * macro. It is valid to bump up the referece count of the IRE, 1867c478bd9Sstevel@tonic-gate * after the lookup has returned an ire. Following are the lookup 1877c478bd9Sstevel@tonic-gate * functions that return an HELD ire : 1887c478bd9Sstevel@tonic-gate * 1897c478bd9Sstevel@tonic-gate * ire_lookup_local[_v6], ire_ctable_lookup[_v6], ire_ftable_lookup[_v6], 1907c478bd9Sstevel@tonic-gate * ire_cache_lookup[_v6], ire_lookup_multi[_v6], ire_route_lookup[_v6], 1915c0b7edeSseb * ipif_to_ire[_v6]. 1927c478bd9Sstevel@tonic-gate * 1937c478bd9Sstevel@tonic-gate * DELETION of an ire holds the bucket lock, removes it from the list 1947c478bd9Sstevel@tonic-gate * and then decrements the reference count for having removed from the list 1957c478bd9Sstevel@tonic-gate * by using the IRE_REFRELE macro. If some other thread has looked up 1967c478bd9Sstevel@tonic-gate * the ire, the reference count would have been bumped up and hence 1977c478bd9Sstevel@tonic-gate * this ire will not be freed once deleted. It will be freed once the 1987c478bd9Sstevel@tonic-gate * reference count drops to zero. 1997c478bd9Sstevel@tonic-gate * 2007c478bd9Sstevel@tonic-gate * Add and Delete acquires the bucket lock as RW_WRITER, while all the 2017c478bd9Sstevel@tonic-gate * lookups acquire the bucket lock as RW_READER. 2027c478bd9Sstevel@tonic-gate * 2037c478bd9Sstevel@tonic-gate * NOTE : The only functions that does the IRE_REFRELE when an ire is 2047c478bd9Sstevel@tonic-gate * passed as an argument are : 2057c478bd9Sstevel@tonic-gate * 2067c478bd9Sstevel@tonic-gate * 1) ip_wput_ire : This is because it IRE_REFHOLD/RELEs the 2077c478bd9Sstevel@tonic-gate * broadcast ires it looks up internally within 2087c478bd9Sstevel@tonic-gate * the function. Currently, for simplicity it does 2097c478bd9Sstevel@tonic-gate * not differentiate the one that is passed in and 2107c478bd9Sstevel@tonic-gate * the ones it looks up internally. It always 2117c478bd9Sstevel@tonic-gate * IRE_REFRELEs. 2127c478bd9Sstevel@tonic-gate * 2) ire_send 2137c478bd9Sstevel@tonic-gate * ire_send_v6 : As ire_send calls ip_wput_ire and other functions 2147c478bd9Sstevel@tonic-gate * that take ire as an argument, it has to selectively 2157c478bd9Sstevel@tonic-gate * IRE_REFRELE the ire. To maintain symmetry, 2167c478bd9Sstevel@tonic-gate * ire_send_v6 does the same. 2177c478bd9Sstevel@tonic-gate * 2187c478bd9Sstevel@tonic-gate * Otherwise, the general rule is to do the IRE_REFRELE in the function 2197c478bd9Sstevel@tonic-gate * that is passing the ire as an argument. 2207c478bd9Sstevel@tonic-gate * 2217c478bd9Sstevel@tonic-gate * In trying to locate ires the following points are to be noted. 2227c478bd9Sstevel@tonic-gate * 2237c478bd9Sstevel@tonic-gate * IRE_MARK_CONDEMNED signifies that the ire has been logically deleted and is 2247c478bd9Sstevel@tonic-gate * to be ignored when walking the ires using ire_next. 2257c478bd9Sstevel@tonic-gate * 2267c478bd9Sstevel@tonic-gate * IRE_MARK_HIDDEN signifies that the ire is a special ire typically for the 2277c478bd9Sstevel@tonic-gate * benefit of in.mpathd which needs to probe interfaces for failures. Normal 2287c478bd9Sstevel@tonic-gate * applications should not be seeing this ire and hence this ire is ignored 2297c478bd9Sstevel@tonic-gate * in most cases in the search using ire_next. 2307c478bd9Sstevel@tonic-gate * 2317c478bd9Sstevel@tonic-gate * Zones note: 2327c478bd9Sstevel@tonic-gate * Walking IREs within a given zone also walks certain ires in other 2337c478bd9Sstevel@tonic-gate * zones. This is done intentionally. IRE walks with a specified 2347c478bd9Sstevel@tonic-gate * zoneid are used only when doing informational reports, and 2357c478bd9Sstevel@tonic-gate * zone users want to see things that they can access. See block 2367c478bd9Sstevel@tonic-gate * comment in ire_walk_ill_match(). 2377c478bd9Sstevel@tonic-gate */ 2387c478bd9Sstevel@tonic-gate 2397c478bd9Sstevel@tonic-gate /* 2407c478bd9Sstevel@tonic-gate * The minimum size of IRE cache table. It will be recalcuated in 2417c478bd9Sstevel@tonic-gate * ip_ire_init(). 242f4b3ec61Sdh155122 * Setable in /etc/system 2437c478bd9Sstevel@tonic-gate */ 2447c478bd9Sstevel@tonic-gate uint32_t ip_cache_table_size = IP_CACHE_TABLE_SIZE; 2457c478bd9Sstevel@tonic-gate uint32_t ip6_cache_table_size = IP6_CACHE_TABLE_SIZE; 2467c478bd9Sstevel@tonic-gate 2477c478bd9Sstevel@tonic-gate /* 2487c478bd9Sstevel@tonic-gate * The size of the forwarding table. We will make sure that it is a 2497c478bd9Sstevel@tonic-gate * power of 2 in ip_ire_init(). 250f4b3ec61Sdh155122 * Setable in /etc/system 2517c478bd9Sstevel@tonic-gate */ 2527c478bd9Sstevel@tonic-gate uint32_t ip6_ftable_hash_size = IP6_FTABLE_HASH_SIZE; 2537c478bd9Sstevel@tonic-gate 2547c478bd9Sstevel@tonic-gate struct kmem_cache *ire_cache; 2557c478bd9Sstevel@tonic-gate static ire_t ire_null; 2567c478bd9Sstevel@tonic-gate 2577c478bd9Sstevel@tonic-gate /* 2587c478bd9Sstevel@tonic-gate * The threshold number of IRE in a bucket when the IREs are 2597c478bd9Sstevel@tonic-gate * cleaned up. This threshold is calculated later in ip_open() 2607c478bd9Sstevel@tonic-gate * based on the speed of CPU and available memory. This default 2617c478bd9Sstevel@tonic-gate * value is the maximum. 2627c478bd9Sstevel@tonic-gate * 2637c478bd9Sstevel@tonic-gate * We have two kinds of cached IRE, temporary and 2647c478bd9Sstevel@tonic-gate * non-temporary. Temporary IREs are marked with 2657c478bd9Sstevel@tonic-gate * IRE_MARK_TEMPORARY. They are IREs created for non 2667c478bd9Sstevel@tonic-gate * TCP traffic and for forwarding purposes. All others 2677c478bd9Sstevel@tonic-gate * are non-temporary IREs. We don't mark IRE created for 2687c478bd9Sstevel@tonic-gate * TCP as temporary because TCP is stateful and there are 2697c478bd9Sstevel@tonic-gate * info stored in the IRE which can be shared by other TCP 2707c478bd9Sstevel@tonic-gate * connections to the same destination. For connected 2717c478bd9Sstevel@tonic-gate * endpoint, we also don't want to mark the IRE used as 2727c478bd9Sstevel@tonic-gate * temporary because the same IRE will be used frequently, 2737c478bd9Sstevel@tonic-gate * otherwise, the app should not do a connect(). We change 2747c478bd9Sstevel@tonic-gate * the marking at ip_bind_connected_*() if necessary. 2757c478bd9Sstevel@tonic-gate * 2767c478bd9Sstevel@tonic-gate * We want to keep the cache IRE hash bucket length reasonably 2777c478bd9Sstevel@tonic-gate * short, otherwise IRE lookup functions will take "forever." 2787c478bd9Sstevel@tonic-gate * We use the "crude" function that the IRE bucket 2797c478bd9Sstevel@tonic-gate * length should be based on the CPU speed, which is 1 entry 2807c478bd9Sstevel@tonic-gate * per x MHz, depending on the shift factor ip_ire_cpu_ratio 2817c478bd9Sstevel@tonic-gate * (n). This means that with a 750MHz CPU, the max bucket 2827c478bd9Sstevel@tonic-gate * length can be (750 >> n) entries. 2837c478bd9Sstevel@tonic-gate * 2847c478bd9Sstevel@tonic-gate * Note that this threshold is separate for temp and non-temp 2857c478bd9Sstevel@tonic-gate * IREs. This means that the actual bucket length can be 2867c478bd9Sstevel@tonic-gate * twice as that. And while we try to keep temporary IRE 2877c478bd9Sstevel@tonic-gate * length at most at the threshold value, we do not attempt to 2887c478bd9Sstevel@tonic-gate * make the length for non-temporary IREs fixed, for the 2897c478bd9Sstevel@tonic-gate * reason stated above. Instead, we start trying to find 2907c478bd9Sstevel@tonic-gate * "unused" non-temporary IREs when the bucket length reaches 2917c478bd9Sstevel@tonic-gate * this threshold and clean them up. 2927c478bd9Sstevel@tonic-gate * 2937c478bd9Sstevel@tonic-gate * We also want to limit the amount of memory used by 2947c478bd9Sstevel@tonic-gate * IREs. So if we are allowed to use ~3% of memory (M) 2957c478bd9Sstevel@tonic-gate * for those IREs, each bucket should not have more than 2967c478bd9Sstevel@tonic-gate * 2977c478bd9Sstevel@tonic-gate * M / num of cache bucket / sizeof (ire_t) 2987c478bd9Sstevel@tonic-gate * 2997c478bd9Sstevel@tonic-gate * Again the above memory uses are separate for temp and 3007c478bd9Sstevel@tonic-gate * non-temp cached IREs. 3017c478bd9Sstevel@tonic-gate * 3027c478bd9Sstevel@tonic-gate * We may also want the limit to be a function of the number 3037c478bd9Sstevel@tonic-gate * of interfaces and number of CPUs. Doing the initialization 3047c478bd9Sstevel@tonic-gate * in ip_open() means that every time an interface is plumbed, 3057c478bd9Sstevel@tonic-gate * the max is re-calculated. Right now, we don't do anything 3067c478bd9Sstevel@tonic-gate * different. In future, when we have more experience, we 3077c478bd9Sstevel@tonic-gate * may want to change this behavior. 3087c478bd9Sstevel@tonic-gate */ 309f4b3ec61Sdh155122 uint32_t ip_ire_max_bucket_cnt = 10; /* Setable in /etc/system */ 3107c478bd9Sstevel@tonic-gate uint32_t ip6_ire_max_bucket_cnt = 10; 3119a09d68dSja97890 uint32_t ip_ire_cleanup_cnt = 2; 3127c478bd9Sstevel@tonic-gate 3137c478bd9Sstevel@tonic-gate /* 3147c478bd9Sstevel@tonic-gate * The minimum of the temporary IRE bucket count. We do not want 3157c478bd9Sstevel@tonic-gate * the length of each bucket to be too short. This may hurt 3167c478bd9Sstevel@tonic-gate * performance of some apps as the temporary IREs are removed too 3177c478bd9Sstevel@tonic-gate * often. 3187c478bd9Sstevel@tonic-gate */ 319f4b3ec61Sdh155122 uint32_t ip_ire_min_bucket_cnt = 3; /* /etc/system - not used */ 3207c478bd9Sstevel@tonic-gate uint32_t ip6_ire_min_bucket_cnt = 3; 3217c478bd9Sstevel@tonic-gate 3227c478bd9Sstevel@tonic-gate /* 3237c478bd9Sstevel@tonic-gate * The ratio of memory consumed by IRE used for temporary to available 3247c478bd9Sstevel@tonic-gate * memory. This is a shift factor, so 6 means the ratio 1 to 64. This 3257c478bd9Sstevel@tonic-gate * value can be changed in /etc/system. 6 is a reasonable number. 3267c478bd9Sstevel@tonic-gate */ 327f4b3ec61Sdh155122 uint32_t ip_ire_mem_ratio = 6; /* /etc/system */ 3287c478bd9Sstevel@tonic-gate /* The shift factor for CPU speed to calculate the max IRE bucket length. */ 329f4b3ec61Sdh155122 uint32_t ip_ire_cpu_ratio = 7; /* /etc/system */ 3307c478bd9Sstevel@tonic-gate 331c793af95Ssangeeta typedef struct nce_clookup_s { 332c793af95Ssangeeta ipaddr_t ncecl_addr; 333c793af95Ssangeeta boolean_t ncecl_found; 334c793af95Ssangeeta } nce_clookup_t; 335c793af95Ssangeeta 3367c478bd9Sstevel@tonic-gate /* 3377c478bd9Sstevel@tonic-gate * The maximum number of buckets in IRE cache table. In future, we may 3387c478bd9Sstevel@tonic-gate * want to make it a dynamic hash table. For the moment, we fix the 3397c478bd9Sstevel@tonic-gate * size and allocate the table in ip_ire_init() when IP is first loaded. 3407c478bd9Sstevel@tonic-gate * We take into account the amount of memory a system has. 3417c478bd9Sstevel@tonic-gate */ 3427c478bd9Sstevel@tonic-gate #define IP_MAX_CACHE_TABLE_SIZE 4096 3437c478bd9Sstevel@tonic-gate 344f4b3ec61Sdh155122 /* Setable in /etc/system */ 3457c478bd9Sstevel@tonic-gate static uint32_t ip_max_cache_table_size = IP_MAX_CACHE_TABLE_SIZE; 3467c478bd9Sstevel@tonic-gate static uint32_t ip6_max_cache_table_size = IP_MAX_CACHE_TABLE_SIZE; 3477c478bd9Sstevel@tonic-gate 3485c0b7edeSseb #define NUM_ILLS 2 /* To build the ILL list to unlock */ 3497c478bd9Sstevel@tonic-gate 3507c478bd9Sstevel@tonic-gate /* Zero iulp_t for initialization. */ 3517c478bd9Sstevel@tonic-gate const iulp_t ire_uinfo_null = { 0 }; 3527c478bd9Sstevel@tonic-gate 3537c478bd9Sstevel@tonic-gate static int ire_add_v4(ire_t **ire_p, queue_t *q, mblk_t *mp, 354c793af95Ssangeeta ipsq_func_t func, boolean_t); 3557c478bd9Sstevel@tonic-gate static void ire_delete_v4(ire_t *ire); 35645916cd2Sjpk static void ire_walk_ipvers(pfv_t func, void *arg, uchar_t vers, 357f4b3ec61Sdh155122 zoneid_t zoneid, ip_stack_t *); 3587c478bd9Sstevel@tonic-gate static void ire_walk_ill_ipvers(uint_t match_flags, uint_t ire_type, 35945916cd2Sjpk pfv_t func, void *arg, uchar_t vers, ill_t *ill); 3609a09d68dSja97890 static void ire_cache_cleanup(irb_t *irb, uint32_t threshold, 3619a09d68dSja97890 ire_t *ref_ire); 362c793af95Ssangeeta static void ip_nce_clookup_and_delete(nce_t *nce, void *arg); 3636a8288c7Scarlsonj #ifdef DEBUG 3646a8288c7Scarlsonj static void ire_trace_cleanup(const ire_t *); 3657c478bd9Sstevel@tonic-gate #endif 3667c478bd9Sstevel@tonic-gate 3677c478bd9Sstevel@tonic-gate /* 3687c478bd9Sstevel@tonic-gate * To avoid bloating the code, we call this function instead of 3697c478bd9Sstevel@tonic-gate * using the macro IRE_REFRELE. Use macro only in performance 3707c478bd9Sstevel@tonic-gate * critical paths. 3717c478bd9Sstevel@tonic-gate * 3727c478bd9Sstevel@tonic-gate * Must not be called while holding any locks. Otherwise if this is 3737c478bd9Sstevel@tonic-gate * the last reference to be released there is a chance of recursive mutex 3747c478bd9Sstevel@tonic-gate * panic due to ire_refrele -> ipif_ill_refrele_tail -> qwriter_ip trying 3757c478bd9Sstevel@tonic-gate * to restart an ioctl. The one exception is when the caller is sure that 3767c478bd9Sstevel@tonic-gate * this is not the last reference to be released. Eg. if the caller is 3777c478bd9Sstevel@tonic-gate * sure that the ire has not been deleted and won't be deleted. 3787c478bd9Sstevel@tonic-gate */ 3797c478bd9Sstevel@tonic-gate void 3807c478bd9Sstevel@tonic-gate ire_refrele(ire_t *ire) 3817c478bd9Sstevel@tonic-gate { 3827c478bd9Sstevel@tonic-gate IRE_REFRELE(ire); 3837c478bd9Sstevel@tonic-gate } 3847c478bd9Sstevel@tonic-gate 3857c478bd9Sstevel@tonic-gate void 3867c478bd9Sstevel@tonic-gate ire_refrele_notr(ire_t *ire) 3877c478bd9Sstevel@tonic-gate { 3887c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 3897c478bd9Sstevel@tonic-gate } 3907c478bd9Sstevel@tonic-gate 3917c478bd9Sstevel@tonic-gate /* 3927c478bd9Sstevel@tonic-gate * kmem_cache_alloc constructor for IRE in kma space. 3937c478bd9Sstevel@tonic-gate * Note that when ire_mp is set the IRE is stored in that mblk and 3947c478bd9Sstevel@tonic-gate * not in this cache. 3957c478bd9Sstevel@tonic-gate */ 3967c478bd9Sstevel@tonic-gate /* ARGSUSED */ 3977c478bd9Sstevel@tonic-gate static int 3987c478bd9Sstevel@tonic-gate ip_ire_constructor(void *buf, void *cdrarg, int kmflags) 3997c478bd9Sstevel@tonic-gate { 4007c478bd9Sstevel@tonic-gate ire_t *ire = buf; 4017c478bd9Sstevel@tonic-gate 402c793af95Ssangeeta ire->ire_nce = NULL; 4037c478bd9Sstevel@tonic-gate 4047c478bd9Sstevel@tonic-gate return (0); 4057c478bd9Sstevel@tonic-gate } 4067c478bd9Sstevel@tonic-gate 4077c478bd9Sstevel@tonic-gate /* ARGSUSED1 */ 4087c478bd9Sstevel@tonic-gate static void 4097c478bd9Sstevel@tonic-gate ip_ire_destructor(void *buf, void *cdrarg) 4107c478bd9Sstevel@tonic-gate { 4117c478bd9Sstevel@tonic-gate ire_t *ire = buf; 4127c478bd9Sstevel@tonic-gate 413c793af95Ssangeeta ASSERT(ire->ire_nce == NULL); 4147c478bd9Sstevel@tonic-gate } 4157c478bd9Sstevel@tonic-gate 4167c478bd9Sstevel@tonic-gate /* 4177c478bd9Sstevel@tonic-gate * This function is associated with the IP_IOC_IRE_ADVISE_NO_REPLY 4187c478bd9Sstevel@tonic-gate * IOCTL. It is used by TCP (or other ULPs) to supply revised information 4197c478bd9Sstevel@tonic-gate * for an existing CACHED IRE. 4207c478bd9Sstevel@tonic-gate */ 4217c478bd9Sstevel@tonic-gate /* ARGSUSED */ 4227c478bd9Sstevel@tonic-gate int 4237c478bd9Sstevel@tonic-gate ip_ire_advise(queue_t *q, mblk_t *mp, cred_t *ioc_cr) 4247c478bd9Sstevel@tonic-gate { 4257c478bd9Sstevel@tonic-gate uchar_t *addr_ucp; 4267c478bd9Sstevel@tonic-gate ipic_t *ipic; 4277c478bd9Sstevel@tonic-gate ire_t *ire; 4287c478bd9Sstevel@tonic-gate ipaddr_t addr; 4297c478bd9Sstevel@tonic-gate in6_addr_t v6addr; 4307c478bd9Sstevel@tonic-gate irb_t *irb; 4317c478bd9Sstevel@tonic-gate zoneid_t zoneid; 432f4b3ec61Sdh155122 ip_stack_t *ipst = CONNQ_TO_IPST(q); 4337c478bd9Sstevel@tonic-gate 4347c478bd9Sstevel@tonic-gate ASSERT(q->q_next == NULL); 4357c478bd9Sstevel@tonic-gate zoneid = Q_TO_CONN(q)->conn_zoneid; 4367c478bd9Sstevel@tonic-gate 4377c478bd9Sstevel@tonic-gate /* 4387c478bd9Sstevel@tonic-gate * Check privilege using the ioctl credential; if it is NULL 4397c478bd9Sstevel@tonic-gate * then this is a kernel message and therefor privileged. 4407c478bd9Sstevel@tonic-gate */ 441f4b3ec61Sdh155122 if (ioc_cr != NULL && secpolicy_ip_config(ioc_cr, B_FALSE) != 0) 4427c478bd9Sstevel@tonic-gate return (EPERM); 4437c478bd9Sstevel@tonic-gate 4447c478bd9Sstevel@tonic-gate ipic = (ipic_t *)mp->b_rptr; 4457c478bd9Sstevel@tonic-gate if (!(addr_ucp = mi_offset_param(mp, ipic->ipic_addr_offset, 4467c478bd9Sstevel@tonic-gate ipic->ipic_addr_length))) { 4477c478bd9Sstevel@tonic-gate return (EINVAL); 4487c478bd9Sstevel@tonic-gate } 4497c478bd9Sstevel@tonic-gate if (!OK_32PTR(addr_ucp)) 4507c478bd9Sstevel@tonic-gate return (EINVAL); 4517c478bd9Sstevel@tonic-gate switch (ipic->ipic_addr_length) { 4527c478bd9Sstevel@tonic-gate case IP_ADDR_LEN: { 4537c478bd9Sstevel@tonic-gate /* Extract the destination address. */ 4547c478bd9Sstevel@tonic-gate addr = *(ipaddr_t *)addr_ucp; 4557c478bd9Sstevel@tonic-gate /* Find the corresponding IRE. */ 456f4b3ec61Sdh155122 ire = ire_cache_lookup(addr, zoneid, NULL, ipst); 4577c478bd9Sstevel@tonic-gate break; 4587c478bd9Sstevel@tonic-gate } 4597c478bd9Sstevel@tonic-gate case IPV6_ADDR_LEN: { 4607c478bd9Sstevel@tonic-gate /* Extract the destination address. */ 4617c478bd9Sstevel@tonic-gate v6addr = *(in6_addr_t *)addr_ucp; 4627c478bd9Sstevel@tonic-gate /* Find the corresponding IRE. */ 463f4b3ec61Sdh155122 ire = ire_cache_lookup_v6(&v6addr, zoneid, NULL, ipst); 4647c478bd9Sstevel@tonic-gate break; 4657c478bd9Sstevel@tonic-gate } 4667c478bd9Sstevel@tonic-gate default: 4677c478bd9Sstevel@tonic-gate return (EINVAL); 4687c478bd9Sstevel@tonic-gate } 4697c478bd9Sstevel@tonic-gate 4707c478bd9Sstevel@tonic-gate if (ire == NULL) 4717c478bd9Sstevel@tonic-gate return (ENOENT); 4727c478bd9Sstevel@tonic-gate /* 4737c478bd9Sstevel@tonic-gate * Update the round trip time estimate and/or the max frag size 4747c478bd9Sstevel@tonic-gate * and/or the slow start threshold. 4757c478bd9Sstevel@tonic-gate * 4767c478bd9Sstevel@tonic-gate * We serialize multiple advises using ire_lock. 4777c478bd9Sstevel@tonic-gate */ 4787c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 4797c478bd9Sstevel@tonic-gate if (ipic->ipic_rtt) { 4807c478bd9Sstevel@tonic-gate /* 4817c478bd9Sstevel@tonic-gate * If there is no old cached values, initialize them 4827c478bd9Sstevel@tonic-gate * conservatively. Set them to be (1.5 * new value). 4837c478bd9Sstevel@tonic-gate */ 4847c478bd9Sstevel@tonic-gate if (ire->ire_uinfo.iulp_rtt != 0) { 4857c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt = (ire->ire_uinfo.iulp_rtt + 4867c478bd9Sstevel@tonic-gate ipic->ipic_rtt) >> 1; 4877c478bd9Sstevel@tonic-gate } else { 4887c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt = ipic->ipic_rtt + 4897c478bd9Sstevel@tonic-gate (ipic->ipic_rtt >> 1); 4907c478bd9Sstevel@tonic-gate } 4917c478bd9Sstevel@tonic-gate if (ire->ire_uinfo.iulp_rtt_sd != 0) { 4927c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt_sd = 4937c478bd9Sstevel@tonic-gate (ire->ire_uinfo.iulp_rtt_sd + 4947c478bd9Sstevel@tonic-gate ipic->ipic_rtt_sd) >> 1; 4957c478bd9Sstevel@tonic-gate } else { 4967c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt_sd = ipic->ipic_rtt_sd + 4977c478bd9Sstevel@tonic-gate (ipic->ipic_rtt_sd >> 1); 4987c478bd9Sstevel@tonic-gate } 4997c478bd9Sstevel@tonic-gate } 5007c478bd9Sstevel@tonic-gate if (ipic->ipic_max_frag) 5017c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(ipic->ipic_max_frag, IP_MAXPACKET); 5027c478bd9Sstevel@tonic-gate if (ipic->ipic_ssthresh != 0) { 5037c478bd9Sstevel@tonic-gate if (ire->ire_uinfo.iulp_ssthresh != 0) 5047c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_ssthresh = 5057c478bd9Sstevel@tonic-gate (ipic->ipic_ssthresh + 5067c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_ssthresh) >> 1; 5077c478bd9Sstevel@tonic-gate else 5087c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_ssthresh = ipic->ipic_ssthresh; 5097c478bd9Sstevel@tonic-gate } 5107c478bd9Sstevel@tonic-gate /* 5117c478bd9Sstevel@tonic-gate * Don't need the ire_lock below this. ire_type does not change 5127c478bd9Sstevel@tonic-gate * after initialization. ire_marks is protected by irb_lock. 5137c478bd9Sstevel@tonic-gate */ 5147c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 5157c478bd9Sstevel@tonic-gate 5167c478bd9Sstevel@tonic-gate if (ipic->ipic_ire_marks != 0 && ire->ire_type == IRE_CACHE) { 5177c478bd9Sstevel@tonic-gate /* 5187c478bd9Sstevel@tonic-gate * Only increment the temporary IRE count if the original 5197c478bd9Sstevel@tonic-gate * IRE is not already marked temporary. 5207c478bd9Sstevel@tonic-gate */ 5217c478bd9Sstevel@tonic-gate irb = ire->ire_bucket; 5227c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_WRITER); 5237c478bd9Sstevel@tonic-gate if ((ipic->ipic_ire_marks & IRE_MARK_TEMPORARY) && 5247c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_TEMPORARY)) { 5257c478bd9Sstevel@tonic-gate irb->irb_tmp_ire_cnt++; 5267c478bd9Sstevel@tonic-gate } 5277c478bd9Sstevel@tonic-gate ire->ire_marks |= ipic->ipic_ire_marks; 5287c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 5297c478bd9Sstevel@tonic-gate } 5307c478bd9Sstevel@tonic-gate 5317c478bd9Sstevel@tonic-gate ire_refrele(ire); 5327c478bd9Sstevel@tonic-gate return (0); 5337c478bd9Sstevel@tonic-gate } 5347c478bd9Sstevel@tonic-gate 5357c478bd9Sstevel@tonic-gate /* 5367c478bd9Sstevel@tonic-gate * This function is associated with the IP_IOC_IRE_DELETE[_NO_REPLY] 5377c478bd9Sstevel@tonic-gate * IOCTL[s]. The NO_REPLY form is used by TCP to delete a route IRE 5387c478bd9Sstevel@tonic-gate * for a host that is not responding. This will force an attempt to 539dc041e83Scarlsonj * establish a new route, if available, and flush out the ARP entry so 540dc041e83Scarlsonj * it will re-resolve. Management processes may want to use the 541dc041e83Scarlsonj * version that generates a reply. 5427c478bd9Sstevel@tonic-gate * 5437c478bd9Sstevel@tonic-gate * This function does not support IPv6 since Neighbor Unreachability Detection 5447c478bd9Sstevel@tonic-gate * means that negative advise like this is useless. 5457c478bd9Sstevel@tonic-gate */ 5467c478bd9Sstevel@tonic-gate /* ARGSUSED */ 5477c478bd9Sstevel@tonic-gate int 5487c478bd9Sstevel@tonic-gate ip_ire_delete(queue_t *q, mblk_t *mp, cred_t *ioc_cr) 5497c478bd9Sstevel@tonic-gate { 5507c478bd9Sstevel@tonic-gate uchar_t *addr_ucp; 5517c478bd9Sstevel@tonic-gate ipaddr_t addr; 5527c478bd9Sstevel@tonic-gate ire_t *ire; 5537c478bd9Sstevel@tonic-gate ipid_t *ipid; 5547c478bd9Sstevel@tonic-gate boolean_t routing_sock_info = B_FALSE; /* Sent info? */ 5557c478bd9Sstevel@tonic-gate zoneid_t zoneid; 556c793af95Ssangeeta ire_t *gire = NULL; 557dc041e83Scarlsonj ill_t *ill; 558dc041e83Scarlsonj mblk_t *arp_mp; 559f4b3ec61Sdh155122 ip_stack_t *ipst; 5607c478bd9Sstevel@tonic-gate 5617c478bd9Sstevel@tonic-gate ASSERT(q->q_next == NULL); 5627c478bd9Sstevel@tonic-gate zoneid = Q_TO_CONN(q)->conn_zoneid; 563f4b3ec61Sdh155122 ipst = CONNQ_TO_IPST(q); 5647c478bd9Sstevel@tonic-gate 5657c478bd9Sstevel@tonic-gate /* 5667c478bd9Sstevel@tonic-gate * Check privilege using the ioctl credential; if it is NULL 5677c478bd9Sstevel@tonic-gate * then this is a kernel message and therefor privileged. 5687c478bd9Sstevel@tonic-gate */ 569f4b3ec61Sdh155122 if (ioc_cr != NULL && secpolicy_ip_config(ioc_cr, B_FALSE) != 0) 5707c478bd9Sstevel@tonic-gate return (EPERM); 5717c478bd9Sstevel@tonic-gate 5727c478bd9Sstevel@tonic-gate ipid = (ipid_t *)mp->b_rptr; 5737c478bd9Sstevel@tonic-gate 5747c478bd9Sstevel@tonic-gate /* Only actions on IRE_CACHEs are acceptable at present. */ 5757c478bd9Sstevel@tonic-gate if (ipid->ipid_ire_type != IRE_CACHE) 5767c478bd9Sstevel@tonic-gate return (EINVAL); 5777c478bd9Sstevel@tonic-gate 5787c478bd9Sstevel@tonic-gate addr_ucp = mi_offset_param(mp, ipid->ipid_addr_offset, 5797c478bd9Sstevel@tonic-gate ipid->ipid_addr_length); 5807c478bd9Sstevel@tonic-gate if (addr_ucp == NULL || !OK_32PTR(addr_ucp)) 5817c478bd9Sstevel@tonic-gate return (EINVAL); 5827c478bd9Sstevel@tonic-gate switch (ipid->ipid_addr_length) { 5837c478bd9Sstevel@tonic-gate case IP_ADDR_LEN: 5847c478bd9Sstevel@tonic-gate /* addr_ucp points at IP addr */ 5857c478bd9Sstevel@tonic-gate break; 5867c478bd9Sstevel@tonic-gate case sizeof (sin_t): { 5877c478bd9Sstevel@tonic-gate sin_t *sin; 5887c478bd9Sstevel@tonic-gate /* 5897c478bd9Sstevel@tonic-gate * got complete (sockaddr) address - increment addr_ucp to point 5907c478bd9Sstevel@tonic-gate * at the ip_addr field. 5917c478bd9Sstevel@tonic-gate */ 5927c478bd9Sstevel@tonic-gate sin = (sin_t *)addr_ucp; 5937c478bd9Sstevel@tonic-gate addr_ucp = (uchar_t *)&sin->sin_addr.s_addr; 5947c478bd9Sstevel@tonic-gate break; 5957c478bd9Sstevel@tonic-gate } 5967c478bd9Sstevel@tonic-gate default: 5977c478bd9Sstevel@tonic-gate return (EINVAL); 5987c478bd9Sstevel@tonic-gate } 5997c478bd9Sstevel@tonic-gate /* Extract the destination address. */ 6007c478bd9Sstevel@tonic-gate bcopy(addr_ucp, &addr, IP_ADDR_LEN); 6017c478bd9Sstevel@tonic-gate 6027c478bd9Sstevel@tonic-gate /* Try to find the CACHED IRE. */ 603f4b3ec61Sdh155122 ire = ire_cache_lookup(addr, zoneid, NULL, ipst); 6047c478bd9Sstevel@tonic-gate 6057c478bd9Sstevel@tonic-gate /* Nail it. */ 6067c478bd9Sstevel@tonic-gate if (ire) { 6077c478bd9Sstevel@tonic-gate /* Allow delete only on CACHE entries */ 6087c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) { 6097c478bd9Sstevel@tonic-gate ire_refrele(ire); 6107c478bd9Sstevel@tonic-gate return (EINVAL); 6117c478bd9Sstevel@tonic-gate } 6127c478bd9Sstevel@tonic-gate 6137c478bd9Sstevel@tonic-gate /* 6147c478bd9Sstevel@tonic-gate * Verify that the IRE has been around for a while. 6157c478bd9Sstevel@tonic-gate * This is to protect against transport protocols 6167c478bd9Sstevel@tonic-gate * that are too eager in sending delete messages. 6177c478bd9Sstevel@tonic-gate */ 6187c478bd9Sstevel@tonic-gate if (gethrestime_sec() < 619f4b3ec61Sdh155122 ire->ire_create_time + ipst->ips_ip_ignore_delete_time) { 6207c478bd9Sstevel@tonic-gate ire_refrele(ire); 6217c478bd9Sstevel@tonic-gate return (EINVAL); 6227c478bd9Sstevel@tonic-gate } 6237c478bd9Sstevel@tonic-gate /* 6247c478bd9Sstevel@tonic-gate * Now we have a potentially dead cache entry. We need 6257c478bd9Sstevel@tonic-gate * to remove it. 626c793af95Ssangeeta * If this cache entry is generated from a 627c793af95Ssangeeta * default route (i.e., ire_cmask == 0), 6287c478bd9Sstevel@tonic-gate * search the default list and mark it dead and some 6297c478bd9Sstevel@tonic-gate * background process will try to activate it. 6307c478bd9Sstevel@tonic-gate */ 6317c478bd9Sstevel@tonic-gate if ((ire->ire_gateway_addr != 0) && (ire->ire_cmask == 0)) { 6327c478bd9Sstevel@tonic-gate /* 6337c478bd9Sstevel@tonic-gate * Make sure that we pick a different 6347c478bd9Sstevel@tonic-gate * IRE_DEFAULT next time. 6357c478bd9Sstevel@tonic-gate */ 6367c478bd9Sstevel@tonic-gate ire_t *gw_ire; 637c793af95Ssangeeta irb_t *irb = NULL; 638c793af95Ssangeeta uint_t match_flags; 6397c478bd9Sstevel@tonic-gate 640c793af95Ssangeeta match_flags = (MATCH_IRE_DEFAULT | MATCH_IRE_RJ_BHOLE); 641c793af95Ssangeeta 642c793af95Ssangeeta gire = ire_ftable_lookup(ire->ire_addr, 643c793af95Ssangeeta ire->ire_cmask, 0, 0, 644f4b3ec61Sdh155122 ire->ire_ipif, NULL, zoneid, 0, NULL, match_flags, 645f4b3ec61Sdh155122 ipst); 646c793af95Ssangeeta 647c793af95Ssangeeta ip3dbg(("ire_ftable_lookup() returned gire %p\n", 648c793af95Ssangeeta (void *)gire)); 649c793af95Ssangeeta 650c793af95Ssangeeta if (gire != NULL) { 651c793af95Ssangeeta irb = gire->ire_bucket; 6527c478bd9Sstevel@tonic-gate 6537c478bd9Sstevel@tonic-gate /* 6547c478bd9Sstevel@tonic-gate * We grab it as writer just to serialize 6557c478bd9Sstevel@tonic-gate * multiple threads trying to bump up 656c793af95Ssangeeta * irb_rr_origin 6577c478bd9Sstevel@tonic-gate */ 6587c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_WRITER); 659c793af95Ssangeeta if ((gw_ire = irb->irb_rr_origin) == NULL) { 6607c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 6617c478bd9Sstevel@tonic-gate goto done; 6627c478bd9Sstevel@tonic-gate } 663c793af95Ssangeeta 6640ad1ccddSsowmini DTRACE_PROBE1(ip__ire__del__origin, 6650ad1ccddSsowmini (ire_t *), gw_ire); 6667c478bd9Sstevel@tonic-gate 6677c478bd9Sstevel@tonic-gate /* Skip past the potentially bad gateway */ 6687c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == 6690ad1ccddSsowmini gw_ire->ire_gateway_addr) { 6700ad1ccddSsowmini ire_t *next = gw_ire->ire_next; 6717c478bd9Sstevel@tonic-gate 6720ad1ccddSsowmini DTRACE_PROBE2(ip__ire__del, 6730ad1ccddSsowmini (ire_t *), gw_ire, (irb_t *), irb); 6740ad1ccddSsowmini IRE_FIND_NEXT_ORIGIN(next); 6750ad1ccddSsowmini irb->irb_rr_origin = next; 6760ad1ccddSsowmini } 6777c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 6787c478bd9Sstevel@tonic-gate } 6797c478bd9Sstevel@tonic-gate } 6807c478bd9Sstevel@tonic-gate done: 681c793af95Ssangeeta if (gire != NULL) 682c793af95Ssangeeta IRE_REFRELE(gire); 6837c478bd9Sstevel@tonic-gate /* report the bad route to routing sockets */ 6847c478bd9Sstevel@tonic-gate ip_rts_change(RTM_LOSING, ire->ire_addr, ire->ire_gateway_addr, 6857c478bd9Sstevel@tonic-gate ire->ire_mask, ire->ire_src_addr, 0, 0, 0, 686f4b3ec61Sdh155122 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_IFA), ipst); 6877c478bd9Sstevel@tonic-gate routing_sock_info = B_TRUE; 688dc041e83Scarlsonj 689dc041e83Scarlsonj /* 690dc041e83Scarlsonj * TCP is really telling us to start over completely, and it 691dc041e83Scarlsonj * expects that we'll resend the ARP query. Tell ARP to 692dc041e83Scarlsonj * discard the entry, if this is a local destination. 693dc041e83Scarlsonj */ 694dc041e83Scarlsonj ill = ire->ire_stq->q_ptr; 695dc041e83Scarlsonj if (ire->ire_gateway_addr == 0 && 696dc041e83Scarlsonj (arp_mp = ill_ared_alloc(ill, addr)) != NULL) { 697dc041e83Scarlsonj putnext(ill->ill_rq, arp_mp); 698dc041e83Scarlsonj } 699dc041e83Scarlsonj 7007c478bd9Sstevel@tonic-gate ire_delete(ire); 7017c478bd9Sstevel@tonic-gate ire_refrele(ire); 7027c478bd9Sstevel@tonic-gate } 7036bdb8e66Sdd193516 /* 7046bdb8e66Sdd193516 * Also look for an IRE_HOST type redirect ire and 7056bdb8e66Sdd193516 * remove it if present. 7066bdb8e66Sdd193516 */ 7076bdb8e66Sdd193516 ire = ire_route_lookup(addr, 0, 0, IRE_HOST, NULL, NULL, 708f4b3ec61Sdh155122 ALL_ZONES, NULL, MATCH_IRE_TYPE, ipst); 7097c478bd9Sstevel@tonic-gate 7107c478bd9Sstevel@tonic-gate /* Nail it. */ 7116bdb8e66Sdd193516 if (ire != NULL) { 7126bdb8e66Sdd193516 if (ire->ire_flags & RTF_DYNAMIC) { 7137c478bd9Sstevel@tonic-gate if (!routing_sock_info) { 7147c478bd9Sstevel@tonic-gate ip_rts_change(RTM_LOSING, ire->ire_addr, 7157c478bd9Sstevel@tonic-gate ire->ire_gateway_addr, ire->ire_mask, 7167c478bd9Sstevel@tonic-gate ire->ire_src_addr, 0, 0, 0, 71754da8755Ssowmini (RTA_DST | RTA_GATEWAY | 71854da8755Ssowmini RTA_NETMASK | RTA_IFA), 719f4b3ec61Sdh155122 ipst); 7207c478bd9Sstevel@tonic-gate } 7217c478bd9Sstevel@tonic-gate ire_delete(ire); 7226bdb8e66Sdd193516 } 7237c478bd9Sstevel@tonic-gate ire_refrele(ire); 7247c478bd9Sstevel@tonic-gate } 7257c478bd9Sstevel@tonic-gate return (0); 7267c478bd9Sstevel@tonic-gate } 7277c478bd9Sstevel@tonic-gate 7287c478bd9Sstevel@tonic-gate 7297c478bd9Sstevel@tonic-gate /* 7307c478bd9Sstevel@tonic-gate * ip_ire_req is called by ip_wput when an IRE_DB_REQ_TYPE message is handed 7317c478bd9Sstevel@tonic-gate * down from the Upper Level Protocol to request a copy of the IRE (to check 7327c478bd9Sstevel@tonic-gate * its type or to extract information like round-trip time estimates or the 7337c478bd9Sstevel@tonic-gate * MTU.) 7347c478bd9Sstevel@tonic-gate * The address is assumed to be in the ire_addr field. If no IRE is found 7357c478bd9Sstevel@tonic-gate * an IRE is returned with ire_type being zero. 7367c478bd9Sstevel@tonic-gate * Note that the upper lavel protocol has to check for broadcast 7377c478bd9Sstevel@tonic-gate * (IRE_BROADCAST) and multicast (CLASSD(addr)). 7387c478bd9Sstevel@tonic-gate * If there is a b_cont the resulting IRE_DB_TYPE mblk is placed at the 7397c478bd9Sstevel@tonic-gate * end of the returned message. 7407c478bd9Sstevel@tonic-gate * 7417c478bd9Sstevel@tonic-gate * TCP sends down a message of this type with a connection request packet 7427c478bd9Sstevel@tonic-gate * chained on. UDP and ICMP send it down to verify that a route exists for 7437c478bd9Sstevel@tonic-gate * the destination address when they get connected. 7447c478bd9Sstevel@tonic-gate */ 7457c478bd9Sstevel@tonic-gate void 7467c478bd9Sstevel@tonic-gate ip_ire_req(queue_t *q, mblk_t *mp) 7477c478bd9Sstevel@tonic-gate { 7487c478bd9Sstevel@tonic-gate ire_t *inire; 7497c478bd9Sstevel@tonic-gate ire_t *ire; 7507c478bd9Sstevel@tonic-gate mblk_t *mp1; 7517c478bd9Sstevel@tonic-gate ire_t *sire = NULL; 7527c478bd9Sstevel@tonic-gate zoneid_t zoneid = Q_TO_CONN(q)->conn_zoneid; 753f4b3ec61Sdh155122 ip_stack_t *ipst = CONNQ_TO_IPST(q); 754f4b3ec61Sdh155122 755f4b3ec61Sdh155122 ASSERT(q->q_next == NULL); 7567c478bd9Sstevel@tonic-gate 7577c478bd9Sstevel@tonic-gate if ((mp->b_wptr - mp->b_rptr) < sizeof (ire_t) || 7587c478bd9Sstevel@tonic-gate !OK_32PTR(mp->b_rptr)) { 7597c478bd9Sstevel@tonic-gate freemsg(mp); 7607c478bd9Sstevel@tonic-gate return; 7617c478bd9Sstevel@tonic-gate } 7627c478bd9Sstevel@tonic-gate inire = (ire_t *)mp->b_rptr; 7637c478bd9Sstevel@tonic-gate /* 7647c478bd9Sstevel@tonic-gate * Got it, now take our best shot at an IRE. 7657c478bd9Sstevel@tonic-gate */ 7667c478bd9Sstevel@tonic-gate if (inire->ire_ipversion == IPV6_VERSION) { 7677c478bd9Sstevel@tonic-gate ire = ire_route_lookup_v6(&inire->ire_addr_v6, 0, 0, 0, 76845916cd2Sjpk NULL, &sire, zoneid, NULL, 769f4b3ec61Sdh155122 (MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT), ipst); 7707c478bd9Sstevel@tonic-gate } else { 7717c478bd9Sstevel@tonic-gate ASSERT(inire->ire_ipversion == IPV4_VERSION); 7727c478bd9Sstevel@tonic-gate ire = ire_route_lookup(inire->ire_addr, 0, 0, 0, 77345916cd2Sjpk NULL, &sire, zoneid, NULL, 774f4b3ec61Sdh155122 (MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT), ipst); 7757c478bd9Sstevel@tonic-gate } 7767c478bd9Sstevel@tonic-gate 7777c478bd9Sstevel@tonic-gate /* 7787c478bd9Sstevel@tonic-gate * We prevent returning IRES with source address INADDR_ANY 7797c478bd9Sstevel@tonic-gate * as these were temporarily created for sending packets 7807c478bd9Sstevel@tonic-gate * from endpoints that have conn_unspec_src set. 7817c478bd9Sstevel@tonic-gate */ 7827c478bd9Sstevel@tonic-gate if (ire == NULL || 7837c478bd9Sstevel@tonic-gate (ire->ire_ipversion == IPV4_VERSION && 7847c478bd9Sstevel@tonic-gate ire->ire_src_addr == INADDR_ANY) || 7857c478bd9Sstevel@tonic-gate (ire->ire_ipversion == IPV6_VERSION && 7867c478bd9Sstevel@tonic-gate IN6_IS_ADDR_UNSPECIFIED(&ire->ire_src_addr_v6))) { 7877c478bd9Sstevel@tonic-gate inire->ire_type = 0; 7887c478bd9Sstevel@tonic-gate } else { 7897c478bd9Sstevel@tonic-gate bcopy(ire, inire, sizeof (ire_t)); 7907c478bd9Sstevel@tonic-gate /* Copy the route metrics from the parent. */ 7917c478bd9Sstevel@tonic-gate if (sire != NULL) { 7927c478bd9Sstevel@tonic-gate bcopy(&(sire->ire_uinfo), &(inire->ire_uinfo), 7937c478bd9Sstevel@tonic-gate sizeof (iulp_t)); 7947c478bd9Sstevel@tonic-gate } 7957c478bd9Sstevel@tonic-gate 7967c478bd9Sstevel@tonic-gate /* 7977c478bd9Sstevel@tonic-gate * As we don't lookup global policy here, we may not 7987c478bd9Sstevel@tonic-gate * pass the right size if per-socket policy is not 7997c478bd9Sstevel@tonic-gate * present. For these cases, path mtu discovery will 8007c478bd9Sstevel@tonic-gate * do the right thing. 8017c478bd9Sstevel@tonic-gate */ 8027c478bd9Sstevel@tonic-gate inire->ire_ipsec_overhead = conn_ipsec_length(Q_TO_CONN(q)); 8037c478bd9Sstevel@tonic-gate 8047c478bd9Sstevel@tonic-gate /* Pass the latest setting of the ip_path_mtu_discovery */ 805f4b3ec61Sdh155122 inire->ire_frag_flag |= 806f4b3ec61Sdh155122 (ipst->ips_ip_path_mtu_discovery) ? IPH_DF : 0; 8077c478bd9Sstevel@tonic-gate } 8087c478bd9Sstevel@tonic-gate if (ire != NULL) 8097c478bd9Sstevel@tonic-gate ire_refrele(ire); 8107c478bd9Sstevel@tonic-gate if (sire != NULL) 8117c478bd9Sstevel@tonic-gate ire_refrele(sire); 8127c478bd9Sstevel@tonic-gate mp->b_wptr = &mp->b_rptr[sizeof (ire_t)]; 8137c478bd9Sstevel@tonic-gate mp->b_datap->db_type = IRE_DB_TYPE; 8147c478bd9Sstevel@tonic-gate 8157c478bd9Sstevel@tonic-gate /* Put the IRE_DB_TYPE mblk last in the chain */ 8167c478bd9Sstevel@tonic-gate mp1 = mp->b_cont; 8177c478bd9Sstevel@tonic-gate if (mp1 != NULL) { 8187c478bd9Sstevel@tonic-gate mp->b_cont = NULL; 8197c478bd9Sstevel@tonic-gate linkb(mp1, mp); 8207c478bd9Sstevel@tonic-gate mp = mp1; 8217c478bd9Sstevel@tonic-gate } 8227c478bd9Sstevel@tonic-gate qreply(q, mp); 8237c478bd9Sstevel@tonic-gate } 8247c478bd9Sstevel@tonic-gate 8257c478bd9Sstevel@tonic-gate /* 8267c478bd9Sstevel@tonic-gate * Send a packet using the specified IRE. 8277c478bd9Sstevel@tonic-gate * If ire_src_addr_v6 is all zero then discard the IRE after 8287c478bd9Sstevel@tonic-gate * the packet has been sent. 8297c478bd9Sstevel@tonic-gate */ 8307c478bd9Sstevel@tonic-gate static void 8317c478bd9Sstevel@tonic-gate ire_send(queue_t *q, mblk_t *pkt, ire_t *ire) 8327c478bd9Sstevel@tonic-gate { 8337c478bd9Sstevel@tonic-gate mblk_t *ipsec_mp; 8347c478bd9Sstevel@tonic-gate boolean_t is_secure; 8357c478bd9Sstevel@tonic-gate uint_t ifindex; 8367c478bd9Sstevel@tonic-gate ill_t *ill; 8375597b60aSnordmark zoneid_t zoneid = ire->ire_zoneid; 838f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 8397c478bd9Sstevel@tonic-gate 8407c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 8415597b60aSnordmark ASSERT(!(ire->ire_type & IRE_LOCAL)); /* Has different ire_zoneid */ 8427c478bd9Sstevel@tonic-gate ipsec_mp = pkt; 8437c478bd9Sstevel@tonic-gate is_secure = (pkt->b_datap->db_type == M_CTL); 8445597b60aSnordmark if (is_secure) { 8455597b60aSnordmark ipsec_out_t *io; 8465597b60aSnordmark 8477c478bd9Sstevel@tonic-gate pkt = pkt->b_cont; 8485597b60aSnordmark io = (ipsec_out_t *)ipsec_mp->b_rptr; 8495597b60aSnordmark if (io->ipsec_out_type == IPSEC_OUT) 8505597b60aSnordmark zoneid = io->ipsec_out_zoneid; 8515597b60aSnordmark } 8527c478bd9Sstevel@tonic-gate 8537c478bd9Sstevel@tonic-gate /* If the packet originated externally then */ 8547c478bd9Sstevel@tonic-gate if (pkt->b_prev) { 8557c478bd9Sstevel@tonic-gate ire_refrele(ire); 8567c478bd9Sstevel@tonic-gate /* 8577c478bd9Sstevel@tonic-gate * Extract the ifindex from b_prev (set in ip_rput_noire). 8587c478bd9Sstevel@tonic-gate * Look up interface to see if it still exists (it could have 8597c478bd9Sstevel@tonic-gate * been unplumbed by the time the reply came back from ARP) 8607c478bd9Sstevel@tonic-gate */ 8617c478bd9Sstevel@tonic-gate ifindex = (uint_t)(uintptr_t)pkt->b_prev; 8627c478bd9Sstevel@tonic-gate ill = ill_lookup_on_ifindex(ifindex, B_FALSE, 863f4b3ec61Sdh155122 NULL, NULL, NULL, NULL, ipst); 8647c478bd9Sstevel@tonic-gate if (ill == NULL) { 8657c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 8667c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 8677c478bd9Sstevel@tonic-gate freemsg(ipsec_mp); 8687c478bd9Sstevel@tonic-gate return; 8697c478bd9Sstevel@tonic-gate } 8707c478bd9Sstevel@tonic-gate q = ill->ill_rq; 8717c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 8727c478bd9Sstevel@tonic-gate /* 8737c478bd9Sstevel@tonic-gate * This packet has not gone through IPSEC processing 8747c478bd9Sstevel@tonic-gate * and hence we should not have any IPSEC message 8757c478bd9Sstevel@tonic-gate * prepended. 8767c478bd9Sstevel@tonic-gate */ 8777c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 878c793af95Ssangeeta put(q, pkt); 8797c478bd9Sstevel@tonic-gate ill_refrele(ill); 8807c478bd9Sstevel@tonic-gate } else if (pkt->b_next) { 8817c478bd9Sstevel@tonic-gate /* Packets from multicast router */ 8827c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 8837c478bd9Sstevel@tonic-gate /* 8847c478bd9Sstevel@tonic-gate * We never get the IPSEC_OUT while forwarding the 8857c478bd9Sstevel@tonic-gate * packet for multicast router. 8867c478bd9Sstevel@tonic-gate */ 8877c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 8887c478bd9Sstevel@tonic-gate ip_rput_forward(ire, (ipha_t *)pkt->b_rptr, ipsec_mp, NULL); 8897c478bd9Sstevel@tonic-gate ire_refrele(ire); 8907c478bd9Sstevel@tonic-gate } else { 8917c478bd9Sstevel@tonic-gate /* Locally originated packets */ 892*7924222fSmeem boolean_t delete_ire = B_FALSE; 8937c478bd9Sstevel@tonic-gate ipha_t *ipha = (ipha_t *)pkt->b_rptr; 8947c478bd9Sstevel@tonic-gate 8957c478bd9Sstevel@tonic-gate /* 896*7924222fSmeem * If this IRE shouldn't be kept in the table (because its 897*7924222fSmeem * source address is unspecified), hold a reference to it so 898*7924222fSmeem * we can delete it even after e.g. ip_wput_ire() has dropped 899*7924222fSmeem * its reference. 9007c478bd9Sstevel@tonic-gate */ 901*7924222fSmeem if (!(ire->ire_marks & IRE_MARK_NOADD) && 902*7924222fSmeem ire->ire_src_addr == INADDR_ANY) { 903*7924222fSmeem delete_ire = B_TRUE; 9047c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 9057c478bd9Sstevel@tonic-gate } 906*7924222fSmeem 9077c478bd9Sstevel@tonic-gate /* 9087c478bd9Sstevel@tonic-gate * If we were resolving a router we can not use the 9097c478bd9Sstevel@tonic-gate * routers IRE for sending the packet (since it would 9107c478bd9Sstevel@tonic-gate * violate the uniqness of the IP idents) thus we 9117c478bd9Sstevel@tonic-gate * make another pass through ip_wput to create the IRE_CACHE 9127c478bd9Sstevel@tonic-gate * for the destination. 9137c478bd9Sstevel@tonic-gate * When IRE_MARK_NOADD is set, ire_add() is not called. 9147c478bd9Sstevel@tonic-gate * Thus ip_wput() will never find a ire and result in an 9157c478bd9Sstevel@tonic-gate * infinite loop. Thus we check whether IRE_MARK_NOADD is 9167c478bd9Sstevel@tonic-gate * is set. This also implies that IRE_MARK_NOADD can only be 9177c478bd9Sstevel@tonic-gate * used to send packets to directly connected hosts. 9187c478bd9Sstevel@tonic-gate */ 9197c478bd9Sstevel@tonic-gate if (ipha->ipha_dst != ire->ire_addr && 9207c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_NOADD)) { 9217c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add */ 9225597b60aSnordmark if (CONN_Q(q)) { 9235597b60aSnordmark (void) ip_output(Q_TO_CONN(q), ipsec_mp, q, 9245597b60aSnordmark IRE_SEND); 9255597b60aSnordmark } else { 9265597b60aSnordmark (void) ip_output((void *)(uintptr_t)zoneid, 9275597b60aSnordmark ipsec_mp, q, IRE_SEND); 9285597b60aSnordmark } 9297c478bd9Sstevel@tonic-gate } else { 9307c478bd9Sstevel@tonic-gate if (is_secure) { 9317c478bd9Sstevel@tonic-gate ipsec_out_t *oi; 9327c478bd9Sstevel@tonic-gate ipha_t *ipha; 9337c478bd9Sstevel@tonic-gate 9347c478bd9Sstevel@tonic-gate oi = (ipsec_out_t *)ipsec_mp->b_rptr; 9357c478bd9Sstevel@tonic-gate ipha = (ipha_t *)ipsec_mp->b_cont->b_rptr; 9367c478bd9Sstevel@tonic-gate if (oi->ipsec_out_proc_begin) { 9377c478bd9Sstevel@tonic-gate /* 9387c478bd9Sstevel@tonic-gate * This is the case where 9397c478bd9Sstevel@tonic-gate * ip_wput_ipsec_out could not find 9407c478bd9Sstevel@tonic-gate * the IRE and recreated a new one. 9417c478bd9Sstevel@tonic-gate * As ip_wput_ipsec_out does ire 9427c478bd9Sstevel@tonic-gate * lookups, ire_refrele for the extra 9437c478bd9Sstevel@tonic-gate * bump in ire_add. 9447c478bd9Sstevel@tonic-gate */ 9457c478bd9Sstevel@tonic-gate ire_refrele(ire); 9467c478bd9Sstevel@tonic-gate ip_wput_ipsec_out(q, ipsec_mp, ipha, 9477c478bd9Sstevel@tonic-gate NULL, NULL); 9487c478bd9Sstevel@tonic-gate } else { 9497c478bd9Sstevel@tonic-gate /* 9507c478bd9Sstevel@tonic-gate * IRE_REFRELE will be done in 9517c478bd9Sstevel@tonic-gate * ip_wput_ire. 9527c478bd9Sstevel@tonic-gate */ 9537c478bd9Sstevel@tonic-gate ip_wput_ire(q, ipsec_mp, ire, NULL, 9545597b60aSnordmark IRE_SEND, zoneid); 9557c478bd9Sstevel@tonic-gate } 9567c478bd9Sstevel@tonic-gate } else { 9577c478bd9Sstevel@tonic-gate /* 9587c478bd9Sstevel@tonic-gate * IRE_REFRELE will be done in ip_wput_ire. 9597c478bd9Sstevel@tonic-gate */ 9607c478bd9Sstevel@tonic-gate ip_wput_ire(q, ipsec_mp, ire, NULL, 9615597b60aSnordmark IRE_SEND, zoneid); 9627c478bd9Sstevel@tonic-gate } 9637c478bd9Sstevel@tonic-gate } 9647c478bd9Sstevel@tonic-gate /* 9657c478bd9Sstevel@tonic-gate * Special code to support sending a single packet with 9667c478bd9Sstevel@tonic-gate * conn_unspec_src using an IRE which has no source address. 9677c478bd9Sstevel@tonic-gate * The IRE is deleted here after sending the packet to avoid 9687c478bd9Sstevel@tonic-gate * having other code trip on it. But before we delete the 9697c478bd9Sstevel@tonic-gate * ire, somebody could have looked up this ire. 9707c478bd9Sstevel@tonic-gate * We prevent returning/using this IRE by the upper layers 9717c478bd9Sstevel@tonic-gate * by making checks to NULL source address in other places 9727c478bd9Sstevel@tonic-gate * like e.g ip_ire_append, ip_ire_req and ip_bind_connected. 973*7924222fSmeem * Though this does not completely prevent other threads 9747c478bd9Sstevel@tonic-gate * from using this ire, this should not cause any problems. 9757c478bd9Sstevel@tonic-gate */ 976*7924222fSmeem if (delete_ire) { 9777c478bd9Sstevel@tonic-gate ip1dbg(("ire_send: delete IRE\n")); 9787c478bd9Sstevel@tonic-gate ire_delete(ire); 9797c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held above */ 9807c478bd9Sstevel@tonic-gate } 9817c478bd9Sstevel@tonic-gate } 9827c478bd9Sstevel@tonic-gate } 9837c478bd9Sstevel@tonic-gate 9847c478bd9Sstevel@tonic-gate /* 9857c478bd9Sstevel@tonic-gate * Send a packet using the specified IRE. 9867c478bd9Sstevel@tonic-gate * If ire_src_addr_v6 is all zero then discard the IRE after 9877c478bd9Sstevel@tonic-gate * the packet has been sent. 9887c478bd9Sstevel@tonic-gate */ 9897c478bd9Sstevel@tonic-gate static void 9907c478bd9Sstevel@tonic-gate ire_send_v6(queue_t *q, mblk_t *pkt, ire_t *ire) 9917c478bd9Sstevel@tonic-gate { 9927c478bd9Sstevel@tonic-gate mblk_t *ipsec_mp; 9937c478bd9Sstevel@tonic-gate boolean_t secure; 9947c478bd9Sstevel@tonic-gate uint_t ifindex; 9955597b60aSnordmark zoneid_t zoneid = ire->ire_zoneid; 996f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 9977c478bd9Sstevel@tonic-gate 9987c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV6_VERSION); 9995597b60aSnordmark ASSERT(!(ire->ire_type & IRE_LOCAL)); /* Has different ire_zoneid */ 10007c478bd9Sstevel@tonic-gate if (pkt->b_datap->db_type == M_CTL) { 10015597b60aSnordmark ipsec_out_t *io; 10025597b60aSnordmark 10037c478bd9Sstevel@tonic-gate ipsec_mp = pkt; 10047c478bd9Sstevel@tonic-gate pkt = pkt->b_cont; 10057c478bd9Sstevel@tonic-gate secure = B_TRUE; 10065597b60aSnordmark io = (ipsec_out_t *)ipsec_mp->b_rptr; 10075597b60aSnordmark if (io->ipsec_out_type == IPSEC_OUT) 10085597b60aSnordmark zoneid = io->ipsec_out_zoneid; 10097c478bd9Sstevel@tonic-gate } else { 10107c478bd9Sstevel@tonic-gate ipsec_mp = pkt; 10117c478bd9Sstevel@tonic-gate secure = B_FALSE; 10127c478bd9Sstevel@tonic-gate } 10137c478bd9Sstevel@tonic-gate 10147c478bd9Sstevel@tonic-gate /* If the packet originated externally then */ 10157c478bd9Sstevel@tonic-gate if (pkt->b_prev) { 10167c478bd9Sstevel@tonic-gate ill_t *ill; 10177c478bd9Sstevel@tonic-gate /* 10187c478bd9Sstevel@tonic-gate * Extract the ifindex from b_prev (set in ip_rput_data_v6). 10197c478bd9Sstevel@tonic-gate * Look up interface to see if it still exists (it could have 10207c478bd9Sstevel@tonic-gate * been unplumbed by the time the reply came back from the 1021c793af95Ssangeeta * resolver). 10227c478bd9Sstevel@tonic-gate */ 10237c478bd9Sstevel@tonic-gate ifindex = (uint_t)(uintptr_t)pkt->b_prev; 10247c478bd9Sstevel@tonic-gate ill = ill_lookup_on_ifindex(ifindex, B_TRUE, 1025f4b3ec61Sdh155122 NULL, NULL, NULL, NULL, ipst); 10267c478bd9Sstevel@tonic-gate if (ill == NULL) { 10277c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 10287c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 10297c478bd9Sstevel@tonic-gate freemsg(ipsec_mp); 10307c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add */ 10317c478bd9Sstevel@tonic-gate return; 10327c478bd9Sstevel@tonic-gate } 10337c478bd9Sstevel@tonic-gate q = ill->ill_rq; 10347c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 10357c478bd9Sstevel@tonic-gate /* 10367c478bd9Sstevel@tonic-gate * This packet has not gone through IPSEC processing 10377c478bd9Sstevel@tonic-gate * and hence we should not have any IPSEC message 10387c478bd9Sstevel@tonic-gate * prepended. 10397c478bd9Sstevel@tonic-gate */ 10407c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 10417c478bd9Sstevel@tonic-gate put(q, pkt); 10427c478bd9Sstevel@tonic-gate ill_refrele(ill); 10437c478bd9Sstevel@tonic-gate } else if (pkt->b_next) { 10447c478bd9Sstevel@tonic-gate /* Packets from multicast router */ 10457c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 10467c478bd9Sstevel@tonic-gate /* 10477c478bd9Sstevel@tonic-gate * We never get the IPSEC_OUT while forwarding the 10487c478bd9Sstevel@tonic-gate * packet for multicast router. 10497c478bd9Sstevel@tonic-gate */ 10507c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 10517c478bd9Sstevel@tonic-gate /* 10527c478bd9Sstevel@tonic-gate * XXX TODO IPv6. 10537c478bd9Sstevel@tonic-gate */ 10547c478bd9Sstevel@tonic-gate freemsg(pkt); 10557c478bd9Sstevel@tonic-gate #ifdef XXX 10567c478bd9Sstevel@tonic-gate ip_rput_forward(ire, (ipha_t *)pkt->b_rptr, pkt, NULL); 10577c478bd9Sstevel@tonic-gate #endif 10587c478bd9Sstevel@tonic-gate } else { 10597c478bd9Sstevel@tonic-gate if (secure) { 10607c478bd9Sstevel@tonic-gate ipsec_out_t *oi; 10617c478bd9Sstevel@tonic-gate ip6_t *ip6h; 10627c478bd9Sstevel@tonic-gate 10637c478bd9Sstevel@tonic-gate oi = (ipsec_out_t *)ipsec_mp->b_rptr; 10647c478bd9Sstevel@tonic-gate ip6h = (ip6_t *)ipsec_mp->b_cont->b_rptr; 10657c478bd9Sstevel@tonic-gate if (oi->ipsec_out_proc_begin) { 10667c478bd9Sstevel@tonic-gate /* 10677c478bd9Sstevel@tonic-gate * This is the case where 10687c478bd9Sstevel@tonic-gate * ip_wput_ipsec_out could not find 10697c478bd9Sstevel@tonic-gate * the IRE and recreated a new one. 10707c478bd9Sstevel@tonic-gate */ 10717c478bd9Sstevel@tonic-gate ip_wput_ipsec_out_v6(q, ipsec_mp, ip6h, 10727c478bd9Sstevel@tonic-gate NULL, NULL); 10737c478bd9Sstevel@tonic-gate } else { 10745597b60aSnordmark if (CONN_Q(q)) { 10755597b60aSnordmark (void) ip_output_v6(Q_TO_CONN(q), 10765597b60aSnordmark ipsec_mp, q, IRE_SEND); 10775597b60aSnordmark } else { 10785597b60aSnordmark (void) ip_output_v6( 10795597b60aSnordmark (void *)(uintptr_t)zoneid, 10805597b60aSnordmark ipsec_mp, q, IRE_SEND); 10815597b60aSnordmark } 10827c478bd9Sstevel@tonic-gate } 10837c478bd9Sstevel@tonic-gate } else { 10847c478bd9Sstevel@tonic-gate /* 10857c478bd9Sstevel@tonic-gate * Send packets through ip_output_v6 so that any 10867c478bd9Sstevel@tonic-gate * ip6_info header can be processed again. 10877c478bd9Sstevel@tonic-gate */ 10885597b60aSnordmark if (CONN_Q(q)) { 10897c478bd9Sstevel@tonic-gate (void) ip_output_v6(Q_TO_CONN(q), ipsec_mp, q, 10907c478bd9Sstevel@tonic-gate IRE_SEND); 10915597b60aSnordmark } else { 10925597b60aSnordmark (void) ip_output_v6((void *)(uintptr_t)zoneid, 10935597b60aSnordmark ipsec_mp, q, IRE_SEND); 10945597b60aSnordmark } 10957c478bd9Sstevel@tonic-gate } 10967c478bd9Sstevel@tonic-gate /* 10977c478bd9Sstevel@tonic-gate * Special code to support sending a single packet with 10987c478bd9Sstevel@tonic-gate * conn_unspec_src using an IRE which has no source address. 10997c478bd9Sstevel@tonic-gate * The IRE is deleted here after sending the packet to avoid 11007c478bd9Sstevel@tonic-gate * having other code trip on it. But before we delete the 11017c478bd9Sstevel@tonic-gate * ire, somebody could have looked up this ire. 11027c478bd9Sstevel@tonic-gate * We prevent returning/using this IRE by the upper layers 11037c478bd9Sstevel@tonic-gate * by making checks to NULL source address in other places 11047c478bd9Sstevel@tonic-gate * like e.g ip_ire_append_v6, ip_ire_req and 11057c478bd9Sstevel@tonic-gate * ip_bind_connected_v6. Though, this does not completely 11067c478bd9Sstevel@tonic-gate * prevent other threads from using this ire, this should 11077c478bd9Sstevel@tonic-gate * not cause any problems. 11087c478bd9Sstevel@tonic-gate */ 11097c478bd9Sstevel@tonic-gate if (IN6_IS_ADDR_UNSPECIFIED(&ire->ire_src_addr_v6)) { 11107c478bd9Sstevel@tonic-gate ip1dbg(("ire_send_v6: delete IRE\n")); 11117c478bd9Sstevel@tonic-gate ire_delete(ire); 11127c478bd9Sstevel@tonic-gate } 11137c478bd9Sstevel@tonic-gate } 11147c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add */ 11157c478bd9Sstevel@tonic-gate } 11167c478bd9Sstevel@tonic-gate 11177c478bd9Sstevel@tonic-gate /* 11187c478bd9Sstevel@tonic-gate * Make sure that IRE bucket does not get too long. 11197c478bd9Sstevel@tonic-gate * This can cause lock up because ire_cache_lookup() 11207c478bd9Sstevel@tonic-gate * may take "forever" to finish. 11217c478bd9Sstevel@tonic-gate * 11229a09d68dSja97890 * We only remove a maximum of cnt IREs each time. This 11239a09d68dSja97890 * should keep the bucket length approximately constant, 11247c478bd9Sstevel@tonic-gate * depending on cnt. This should be enough to defend 11257c478bd9Sstevel@tonic-gate * against DoS attack based on creating temporary IREs 11267c478bd9Sstevel@tonic-gate * (for forwarding and non-TCP traffic). 11277c478bd9Sstevel@tonic-gate * 11289a09d68dSja97890 * We also pass in the address of the newly created IRE 11299a09d68dSja97890 * as we do not want to remove this straight after adding 11309a09d68dSja97890 * it. New IREs are normally added at the tail of the 11317c478bd9Sstevel@tonic-gate * bucket. This means that we are removing the "oldest" 11329a09d68dSja97890 * temporary IREs added. Only if there are IREs with 11337c478bd9Sstevel@tonic-gate * the same ire_addr, do we not add it at the tail. Refer 11347c478bd9Sstevel@tonic-gate * to ire_add_v*(). It should be OK for our purpose. 11357c478bd9Sstevel@tonic-gate * 11367c478bd9Sstevel@tonic-gate * For non-temporary cached IREs, we make sure that they 11377c478bd9Sstevel@tonic-gate * have not been used for some time (defined below), they 11387c478bd9Sstevel@tonic-gate * are non-local destinations, and there is no one using 11397c478bd9Sstevel@tonic-gate * them at the moment (refcnt == 1). 11407c478bd9Sstevel@tonic-gate * 11417c478bd9Sstevel@tonic-gate * The above means that the IRE bucket length may become 11427c478bd9Sstevel@tonic-gate * very long, consisting of mostly non-temporary IREs. 11437c478bd9Sstevel@tonic-gate * This can happen when the hash function does a bad job 11447c478bd9Sstevel@tonic-gate * so that most TCP connections cluster to a specific bucket. 11457c478bd9Sstevel@tonic-gate * This "hopefully" should never happen. It can also 11467c478bd9Sstevel@tonic-gate * happen if most TCP connections have very long lives. 11477c478bd9Sstevel@tonic-gate * Even with the minimal hash table size of 256, there 11487c478bd9Sstevel@tonic-gate * has to be a lot of such connections to make the bucket 11497c478bd9Sstevel@tonic-gate * length unreasonably long. This should probably not 11507c478bd9Sstevel@tonic-gate * happen either. The third can when this can happen is 11517c478bd9Sstevel@tonic-gate * when the machine is under attack, such as SYN flooding. 11527c478bd9Sstevel@tonic-gate * TCP should already have the proper mechanism to protect 11537c478bd9Sstevel@tonic-gate * that. So we should be safe. 11547c478bd9Sstevel@tonic-gate * 11557c478bd9Sstevel@tonic-gate * This function is called by ire_add_then_send() after 11567c478bd9Sstevel@tonic-gate * a new IRE is added and the packet is sent. 11577c478bd9Sstevel@tonic-gate * 11587c478bd9Sstevel@tonic-gate * The idle cutoff interval is set to 60s. It can be 11597c478bd9Sstevel@tonic-gate * changed using /etc/system. 11607c478bd9Sstevel@tonic-gate */ 11617c478bd9Sstevel@tonic-gate uint32_t ire_idle_cutoff_interval = 60000; 11627c478bd9Sstevel@tonic-gate 11637c478bd9Sstevel@tonic-gate static void 11649a09d68dSja97890 ire_cache_cleanup(irb_t *irb, uint32_t threshold, ire_t *ref_ire) 11657c478bd9Sstevel@tonic-gate { 11667c478bd9Sstevel@tonic-gate ire_t *ire; 11677c478bd9Sstevel@tonic-gate clock_t cut_off = drv_usectohz(ire_idle_cutoff_interval * 1000); 11689a09d68dSja97890 int cnt = ip_ire_cleanup_cnt; 11697c478bd9Sstevel@tonic-gate 11707c478bd9Sstevel@tonic-gate /* 11719a09d68dSja97890 * Try to remove cnt temporary IREs first. 11727c478bd9Sstevel@tonic-gate */ 11739a09d68dSja97890 for (ire = irb->irb_ire; cnt > 0 && ire != NULL; ire = ire->ire_next) { 11749a09d68dSja97890 if (ire == ref_ire) 11759a09d68dSja97890 continue; 11767c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) 11777c478bd9Sstevel@tonic-gate continue; 11787c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_TEMPORARY) { 11797c478bd9Sstevel@tonic-gate ASSERT(ire->ire_type == IRE_CACHE); 11807c478bd9Sstevel@tonic-gate ire_delete(ire); 11819a09d68dSja97890 cnt--; 11827c478bd9Sstevel@tonic-gate } 11837c478bd9Sstevel@tonic-gate } 11849a09d68dSja97890 if (cnt == 0) 11859a09d68dSja97890 return; 11869a09d68dSja97890 11879a09d68dSja97890 /* 11889a09d68dSja97890 * If we didn't satisfy our removal target from temporary IREs 11899a09d68dSja97890 * we see how many non-temporary IREs are currently in the bucket. 11909a09d68dSja97890 * If this quantity is above the threshold then we see if there are any 11919a09d68dSja97890 * candidates for removal. We are still limited to removing a maximum 11929a09d68dSja97890 * of cnt IREs. 11939a09d68dSja97890 */ 11949a09d68dSja97890 if ((irb->irb_ire_cnt - irb->irb_tmp_ire_cnt) > threshold) { 11959a09d68dSja97890 for (ire = irb->irb_ire; cnt > 0 && ire != NULL; 11967c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 11979a09d68dSja97890 if (ire == ref_ire) 11989a09d68dSja97890 continue; 11999a09d68dSja97890 if (ire->ire_type != IRE_CACHE) 12009a09d68dSja97890 continue; 1201f4b3ec61Sdh155122 if (ire->ire_marks & IRE_MARK_CONDEMNED) 1202f4b3ec61Sdh155122 continue; 12039a09d68dSja97890 if ((ire->ire_refcnt == 1) && 12049a09d68dSja97890 (lbolt - ire->ire_last_used_time > cut_off)) { 12057c478bd9Sstevel@tonic-gate ire_delete(ire); 12067c478bd9Sstevel@tonic-gate cnt--; 12077c478bd9Sstevel@tonic-gate } 12087c478bd9Sstevel@tonic-gate } 12097c478bd9Sstevel@tonic-gate } 12107c478bd9Sstevel@tonic-gate } 12117c478bd9Sstevel@tonic-gate 12127c478bd9Sstevel@tonic-gate /* 12137c478bd9Sstevel@tonic-gate * ire_add_then_send is called when a new IRE has been created in order to 12147c478bd9Sstevel@tonic-gate * route an outgoing packet. Typically, it is called from ip_wput when 12157c478bd9Sstevel@tonic-gate * a response comes back down from a resolver. We add the IRE, and then 12167c478bd9Sstevel@tonic-gate * possibly run the packet through ip_wput or ip_rput, as appropriate. 12177c478bd9Sstevel@tonic-gate * However, we do not add the newly created IRE in the cache when 12187c478bd9Sstevel@tonic-gate * IRE_MARK_NOADD is set in the IRE. IRE_MARK_NOADD is set at 12195c0b7edeSseb * ip_newroute_ipif(). The ires with IRE_MARK_NOADD are ire_refrele'd by 12205c0b7edeSseb * ip_wput_ire() and get deleted. 12217c478bd9Sstevel@tonic-gate * Multirouting support: the packet is silently discarded when the new IRE 12227c478bd9Sstevel@tonic-gate * holds the RTF_MULTIRT flag, but is not the first IRE to be added with the 12237c478bd9Sstevel@tonic-gate * RTF_MULTIRT flag for the same destination address. 12247c478bd9Sstevel@tonic-gate * In this case, we just want to register this additional ire without 12257c478bd9Sstevel@tonic-gate * sending the packet, as it has already been replicated through 12267c478bd9Sstevel@tonic-gate * existing multirt routes in ip_wput(). 12277c478bd9Sstevel@tonic-gate */ 12287c478bd9Sstevel@tonic-gate void 12297c478bd9Sstevel@tonic-gate ire_add_then_send(queue_t *q, ire_t *ire, mblk_t *mp) 12307c478bd9Sstevel@tonic-gate { 12317c478bd9Sstevel@tonic-gate irb_t *irb; 12327c478bd9Sstevel@tonic-gate boolean_t drop = B_FALSE; 12337c478bd9Sstevel@tonic-gate /* LINTED : set but not used in function */ 12347c478bd9Sstevel@tonic-gate boolean_t mctl_present; 12357c478bd9Sstevel@tonic-gate mblk_t *first_mp = NULL; 12367c478bd9Sstevel@tonic-gate mblk_t *save_mp = NULL; 12377c478bd9Sstevel@tonic-gate ire_t *dst_ire; 12387c478bd9Sstevel@tonic-gate ipha_t *ipha; 12397c478bd9Sstevel@tonic-gate ip6_t *ip6h; 1240f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 12419a09d68dSja97890 int ire_limit; 12427c478bd9Sstevel@tonic-gate 12437c478bd9Sstevel@tonic-gate if (mp != NULL) { 12447c478bd9Sstevel@tonic-gate /* 12457c478bd9Sstevel@tonic-gate * We first have to retrieve the destination address carried 12467c478bd9Sstevel@tonic-gate * by the packet. 12477c478bd9Sstevel@tonic-gate * We can't rely on ire as it can be related to a gateway. 12487c478bd9Sstevel@tonic-gate * The destination address will help in determining if 12497c478bd9Sstevel@tonic-gate * other RTF_MULTIRT ires are already registered. 12507c478bd9Sstevel@tonic-gate * 12517c478bd9Sstevel@tonic-gate * We first need to know where we are going : v4 or V6. 12527c478bd9Sstevel@tonic-gate * the ire version is enough, as there is no risk that 12537c478bd9Sstevel@tonic-gate * we resolve an IPv6 address with an IPv4 ire 12547c478bd9Sstevel@tonic-gate * or vice versa. 12557c478bd9Sstevel@tonic-gate */ 12567c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 12577c478bd9Sstevel@tonic-gate EXTRACT_PKT_MP(mp, first_mp, mctl_present); 12587c478bd9Sstevel@tonic-gate ipha = (ipha_t *)mp->b_rptr; 12597c478bd9Sstevel@tonic-gate save_mp = mp; 12607c478bd9Sstevel@tonic-gate mp = first_mp; 12617c478bd9Sstevel@tonic-gate 12627c478bd9Sstevel@tonic-gate dst_ire = ire_cache_lookup(ipha->ipha_dst, 1263f4b3ec61Sdh155122 ire->ire_zoneid, MBLK_GETLABEL(mp), ipst); 12647c478bd9Sstevel@tonic-gate } else { 1265c793af95Ssangeeta ASSERT(ire->ire_ipversion == IPV6_VERSION); 12667c478bd9Sstevel@tonic-gate /* 12677c478bd9Sstevel@tonic-gate * Get a pointer to the beginning of the IPv6 header. 12687c478bd9Sstevel@tonic-gate * Ignore leading IPsec control mblks. 12697c478bd9Sstevel@tonic-gate */ 12707c478bd9Sstevel@tonic-gate first_mp = mp; 12717c478bd9Sstevel@tonic-gate if (mp->b_datap->db_type == M_CTL) { 12727c478bd9Sstevel@tonic-gate mp = mp->b_cont; 12737c478bd9Sstevel@tonic-gate } 12747c478bd9Sstevel@tonic-gate ip6h = (ip6_t *)mp->b_rptr; 12757c478bd9Sstevel@tonic-gate save_mp = mp; 12767c478bd9Sstevel@tonic-gate mp = first_mp; 12777c478bd9Sstevel@tonic-gate dst_ire = ire_cache_lookup_v6(&ip6h->ip6_dst, 1278f4b3ec61Sdh155122 ire->ire_zoneid, MBLK_GETLABEL(mp), ipst); 12797c478bd9Sstevel@tonic-gate } 12807c478bd9Sstevel@tonic-gate if (dst_ire != NULL) { 12817c478bd9Sstevel@tonic-gate if (dst_ire->ire_flags & RTF_MULTIRT) { 12827c478bd9Sstevel@tonic-gate /* 12837c478bd9Sstevel@tonic-gate * At least one resolved multirt route 12847c478bd9Sstevel@tonic-gate * already exists for the destination, 12857c478bd9Sstevel@tonic-gate * don't sent this packet: either drop it 12867c478bd9Sstevel@tonic-gate * or complete the pending resolution, 12877c478bd9Sstevel@tonic-gate * depending on the ire. 12887c478bd9Sstevel@tonic-gate */ 12897c478bd9Sstevel@tonic-gate drop = B_TRUE; 12907c478bd9Sstevel@tonic-gate } 12917c478bd9Sstevel@tonic-gate ip1dbg(("ire_add_then_send: dst_ire %p " 12927c478bd9Sstevel@tonic-gate "[dst %08x, gw %08x], drop %d\n", 12937c478bd9Sstevel@tonic-gate (void *)dst_ire, 12947c478bd9Sstevel@tonic-gate (dst_ire->ire_ipversion == IPV4_VERSION) ? \ 12957c478bd9Sstevel@tonic-gate ntohl(dst_ire->ire_addr) : \ 12967c478bd9Sstevel@tonic-gate ntohl(V4_PART_OF_V6(dst_ire->ire_addr_v6)), 12977c478bd9Sstevel@tonic-gate (dst_ire->ire_ipversion == IPV4_VERSION) ? \ 12987c478bd9Sstevel@tonic-gate ntohl(dst_ire->ire_gateway_addr) : \ 12997c478bd9Sstevel@tonic-gate ntohl(V4_PART_OF_V6( 13007c478bd9Sstevel@tonic-gate dst_ire->ire_gateway_addr_v6)), 13017c478bd9Sstevel@tonic-gate drop)); 13027c478bd9Sstevel@tonic-gate ire_refrele(dst_ire); 13037c478bd9Sstevel@tonic-gate } 13047c478bd9Sstevel@tonic-gate } 13057c478bd9Sstevel@tonic-gate 13067c478bd9Sstevel@tonic-gate if (!(ire->ire_marks & IRE_MARK_NOADD)) { 13075c0b7edeSseb /* Regular packets with cache bound ires are here. */ 1308c793af95Ssangeeta (void) ire_add(&ire, NULL, NULL, NULL, B_FALSE); 13097c478bd9Sstevel@tonic-gate 13107c478bd9Sstevel@tonic-gate if (ire == NULL) { 13117c478bd9Sstevel@tonic-gate mp->b_prev = NULL; 13127c478bd9Sstevel@tonic-gate mp->b_next = NULL; 13137c478bd9Sstevel@tonic-gate MULTIRT_DEBUG_UNTAG(mp); 13147c478bd9Sstevel@tonic-gate freemsg(mp); 13157c478bd9Sstevel@tonic-gate return; 13167c478bd9Sstevel@tonic-gate } 13177c478bd9Sstevel@tonic-gate if (mp == NULL) { 13187c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add_v4/v6 */ 13197c478bd9Sstevel@tonic-gate return; 13207c478bd9Sstevel@tonic-gate } 13217c478bd9Sstevel@tonic-gate } 13227c478bd9Sstevel@tonic-gate if (drop) { 13237c478bd9Sstevel@tonic-gate /* 13247c478bd9Sstevel@tonic-gate * If we're adding an RTF_MULTIRT ire, the resolution 13257c478bd9Sstevel@tonic-gate * is over: we just drop the packet. 13267c478bd9Sstevel@tonic-gate */ 13277c478bd9Sstevel@tonic-gate if (ire->ire_flags & RTF_MULTIRT) { 13287c478bd9Sstevel@tonic-gate if (save_mp) { 13297c478bd9Sstevel@tonic-gate save_mp->b_prev = NULL; 13307c478bd9Sstevel@tonic-gate save_mp->b_next = NULL; 13317c478bd9Sstevel@tonic-gate } 13327c478bd9Sstevel@tonic-gate MULTIRT_DEBUG_UNTAG(mp); 13337c478bd9Sstevel@tonic-gate freemsg(mp); 13347c478bd9Sstevel@tonic-gate } else { 13357c478bd9Sstevel@tonic-gate /* 13367c478bd9Sstevel@tonic-gate * Otherwise, we're adding the ire to a gateway 13377c478bd9Sstevel@tonic-gate * for a multirt route. 13387c478bd9Sstevel@tonic-gate * Invoke ip_newroute() to complete the resolution 13397c478bd9Sstevel@tonic-gate * of the route. We will then come back here and 13407c478bd9Sstevel@tonic-gate * finally drop this packet in the above code. 13417c478bd9Sstevel@tonic-gate */ 13427c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 13437c478bd9Sstevel@tonic-gate /* 13447c478bd9Sstevel@tonic-gate * TODO: in order for CGTP to work in non-global 13457c478bd9Sstevel@tonic-gate * zones, ip_newroute() must create the IRE 13467c478bd9Sstevel@tonic-gate * cache in the zone indicated by 13477c478bd9Sstevel@tonic-gate * ire->ire_zoneid. 13487c478bd9Sstevel@tonic-gate */ 13495c0b7edeSseb ip_newroute(q, mp, ipha->ipha_dst, 13505597b60aSnordmark (CONN_Q(q) ? Q_TO_CONN(q) : NULL), 1351f4b3ec61Sdh155122 ire->ire_zoneid, ipst); 13527c478bd9Sstevel@tonic-gate } else { 1353c793af95Ssangeeta ASSERT(ire->ire_ipversion == IPV6_VERSION); 13547c478bd9Sstevel@tonic-gate ip_newroute_v6(q, mp, &ip6h->ip6_dst, NULL, 1355f4b3ec61Sdh155122 NULL, ire->ire_zoneid, ipst); 13567c478bd9Sstevel@tonic-gate } 13577c478bd9Sstevel@tonic-gate } 13587c478bd9Sstevel@tonic-gate 13597c478bd9Sstevel@tonic-gate ire_refrele(ire); /* As done by ire_send(). */ 13607c478bd9Sstevel@tonic-gate return; 13617c478bd9Sstevel@tonic-gate } 13627c478bd9Sstevel@tonic-gate /* 13637c478bd9Sstevel@tonic-gate * Need to remember ire_bucket here as ire_send*() may delete 13647c478bd9Sstevel@tonic-gate * the ire so we cannot reference it after that. 13657c478bd9Sstevel@tonic-gate */ 13667c478bd9Sstevel@tonic-gate irb = ire->ire_bucket; 13679a09d68dSja97890 if (ire->ire_ipversion == IPV4_VERSION) { 13687c478bd9Sstevel@tonic-gate ire_send(q, mp, ire); 13699a09d68dSja97890 ire_limit = ip_ire_max_bucket_cnt; 13709a09d68dSja97890 } else { 13719a09d68dSja97890 ire_send_v6(q, mp, ire); 13729a09d68dSja97890 ire_limit = ip6_ire_max_bucket_cnt; 13739a09d68dSja97890 } 13749a09d68dSja97890 13759a09d68dSja97890 /* 13769a09d68dSja97890 * irb is NULL if the IRE was not added to the hash. This happens 13779a09d68dSja97890 * when IRE_MARK_NOADD is set and when IREs are returned from 13789a09d68dSja97890 * ire_update_srcif_v4(). 13799a09d68dSja97890 */ 13809a09d68dSja97890 if (irb != NULL) { 13819a09d68dSja97890 IRB_REFHOLD(irb); 13829a09d68dSja97890 if (irb->irb_ire_cnt > ire_limit) 13839a09d68dSja97890 ire_cache_cleanup(irb, ire_limit, ire); 13849a09d68dSja97890 IRB_REFRELE(irb); 13857c478bd9Sstevel@tonic-gate } 13867c478bd9Sstevel@tonic-gate } 13877c478bd9Sstevel@tonic-gate 13887c478bd9Sstevel@tonic-gate /* 13897c478bd9Sstevel@tonic-gate * Initialize the ire that is specific to IPv4 part and call 13907c478bd9Sstevel@tonic-gate * ire_init_common to finish it. 13917c478bd9Sstevel@tonic-gate */ 13927c478bd9Sstevel@tonic-gate ire_t * 13937c478bd9Sstevel@tonic-gate ire_init(ire_t *ire, uchar_t *addr, uchar_t *mask, uchar_t *src_addr, 13945c0b7edeSseb uchar_t *gateway, uint_t *max_fragp, nce_t *src_nce, queue_t *rfq, 13955c0b7edeSseb queue_t *stq, ushort_t type, ipif_t *ipif, ipaddr_t cmask, uint32_t phandle, 13965c0b7edeSseb uint32_t ihandle, uint32_t flags, const iulp_t *ulp_info, tsol_gc_t *gc, 13975c0b7edeSseb tsol_gcgrp_t *gcgrp, ip_stack_t *ipst) 13987c478bd9Sstevel@tonic-gate { 139942ee8d71Sja97890 ASSERT(type != IRE_CACHE || stq != NULL); 140045916cd2Sjpk /* 140145916cd2Sjpk * Reject IRE security attribute creation/initialization 140245916cd2Sjpk * if system is not running in Trusted mode. 140345916cd2Sjpk */ 140445916cd2Sjpk if ((gc != NULL || gcgrp != NULL) && !is_system_labeled()) 140545916cd2Sjpk return (NULL); 140645916cd2Sjpk 14077c478bd9Sstevel@tonic-gate 1408f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_alloced); 14097c478bd9Sstevel@tonic-gate 14107c478bd9Sstevel@tonic-gate if (addr != NULL) 14117c478bd9Sstevel@tonic-gate bcopy(addr, &ire->ire_addr, IP_ADDR_LEN); 14127c478bd9Sstevel@tonic-gate if (src_addr != NULL) 14137c478bd9Sstevel@tonic-gate bcopy(src_addr, &ire->ire_src_addr, IP_ADDR_LEN); 14147c478bd9Sstevel@tonic-gate if (mask != NULL) { 14157c478bd9Sstevel@tonic-gate bcopy(mask, &ire->ire_mask, IP_ADDR_LEN); 14167c478bd9Sstevel@tonic-gate ire->ire_masklen = ip_mask_to_plen(ire->ire_mask); 14177c478bd9Sstevel@tonic-gate } 14187c478bd9Sstevel@tonic-gate if (gateway != NULL) { 14197c478bd9Sstevel@tonic-gate bcopy(gateway, &ire->ire_gateway_addr, IP_ADDR_LEN); 14207c478bd9Sstevel@tonic-gate } 14217c478bd9Sstevel@tonic-gate 14227c478bd9Sstevel@tonic-gate if (type == IRE_CACHE) 14237c478bd9Sstevel@tonic-gate ire->ire_cmask = cmask; 14247c478bd9Sstevel@tonic-gate 142545916cd2Sjpk /* ire_init_common will free the mblks upon encountering any failure */ 14265c0b7edeSseb if (!ire_init_common(ire, max_fragp, src_nce, rfq, stq, type, ipif, 14275c0b7edeSseb phandle, ihandle, flags, IPV4_VERSION, ulp_info, gc, gcgrp, ipst)) 142845916cd2Sjpk return (NULL); 14297c478bd9Sstevel@tonic-gate 14307c478bd9Sstevel@tonic-gate return (ire); 14317c478bd9Sstevel@tonic-gate } 14327c478bd9Sstevel@tonic-gate 14337c478bd9Sstevel@tonic-gate /* 14347c478bd9Sstevel@tonic-gate * Similar to ire_create except that it is called only when 14357c478bd9Sstevel@tonic-gate * we want to allocate ire as an mblk e.g. we have an external 14367c478bd9Sstevel@tonic-gate * resolver ARP. 14377c478bd9Sstevel@tonic-gate */ 14387c478bd9Sstevel@tonic-gate ire_t * 14397c478bd9Sstevel@tonic-gate ire_create_mp(uchar_t *addr, uchar_t *mask, uchar_t *src_addr, uchar_t *gateway, 14405c0b7edeSseb uint_t max_frag, nce_t *src_nce, queue_t *rfq, queue_t *stq, ushort_t type, 14415c0b7edeSseb ipif_t *ipif, ipaddr_t cmask, uint32_t phandle, uint32_t ihandle, 14425c0b7edeSseb uint32_t flags, const iulp_t *ulp_info, tsol_gc_t *gc, tsol_gcgrp_t *gcgrp, 1443f4b3ec61Sdh155122 ip_stack_t *ipst) 14447c478bd9Sstevel@tonic-gate { 1445c793af95Ssangeeta ire_t *ire, *buf; 14467c478bd9Sstevel@tonic-gate ire_t *ret_ire; 14477c478bd9Sstevel@tonic-gate mblk_t *mp; 1448c793af95Ssangeeta size_t bufsize; 1449c793af95Ssangeeta frtn_t *frtnp; 1450c793af95Ssangeeta ill_t *ill; 14517c478bd9Sstevel@tonic-gate 1452c793af95Ssangeeta bufsize = sizeof (ire_t) + sizeof (frtn_t); 1453c793af95Ssangeeta buf = kmem_alloc(bufsize, KM_NOSLEEP); 1454c793af95Ssangeeta if (buf == NULL) { 14557c478bd9Sstevel@tonic-gate ip1dbg(("ire_create_mp: alloc failed\n")); 14567c478bd9Sstevel@tonic-gate return (NULL); 14577c478bd9Sstevel@tonic-gate } 1458c793af95Ssangeeta frtnp = (frtn_t *)(buf + 1); 1459c793af95Ssangeeta frtnp->free_arg = (caddr_t)buf; 1460c793af95Ssangeeta frtnp->free_func = ire_freemblk; 14617c478bd9Sstevel@tonic-gate 1462c793af95Ssangeeta /* 1463c793af95Ssangeeta * Allocate the new IRE. The ire created will hold a ref on 1464c793af95Ssangeeta * an nce_t after ire_nce_init, and this ref must either be 1465c793af95Ssangeeta * (a) transferred to the ire_cache entry created when ire_add_v4 1466c793af95Ssangeeta * is called after successful arp resolution, or, 1467c793af95Ssangeeta * (b) released, when arp resolution fails 1468c793af95Ssangeeta * Case (b) is handled in ire_freemblk() which will be called 1469c793af95Ssangeeta * when mp is freed as a result of failed arp. 1470c793af95Ssangeeta */ 1471c793af95Ssangeeta mp = esballoc((unsigned char *)buf, bufsize, BPRI_MED, frtnp); 1472c793af95Ssangeeta if (mp == NULL) { 1473c793af95Ssangeeta ip1dbg(("ire_create_mp: alloc failed\n")); 1474c793af95Ssangeeta kmem_free(buf, bufsize); 1475c793af95Ssangeeta return (NULL); 1476c793af95Ssangeeta } 14777c478bd9Sstevel@tonic-gate ire = (ire_t *)mp->b_rptr; 14787c478bd9Sstevel@tonic-gate mp->b_wptr = (uchar_t *)&ire[1]; 14797c478bd9Sstevel@tonic-gate 14807c478bd9Sstevel@tonic-gate /* Start clean. */ 14817c478bd9Sstevel@tonic-gate *ire = ire_null; 14827c478bd9Sstevel@tonic-gate ire->ire_mp = mp; 14837c478bd9Sstevel@tonic-gate mp->b_datap->db_type = IRE_DB_TYPE; 1484c793af95Ssangeeta ire->ire_marks |= IRE_MARK_UNCACHED; 14857c478bd9Sstevel@tonic-gate 14865c0b7edeSseb ret_ire = ire_init(ire, addr, mask, src_addr, gateway, NULL, src_nce, 14875c0b7edeSseb rfq, stq, type, ipif, cmask, phandle, ihandle, flags, ulp_info, gc, 14885c0b7edeSseb gcgrp, ipst); 14897c478bd9Sstevel@tonic-gate 149053bec8caSsowmini ill = (ill_t *)(stq->q_ptr); 14917c478bd9Sstevel@tonic-gate if (ret_ire == NULL) { 1492f4b3ec61Sdh155122 /* ire_freemblk needs these set */ 149353bec8caSsowmini ire->ire_stq_ifindex = ill->ill_phyint->phyint_ifindex; 1494f4b3ec61Sdh155122 ire->ire_ipst = ipst; 14957c478bd9Sstevel@tonic-gate freeb(ire->ire_mp); 14967c478bd9Sstevel@tonic-gate return (NULL); 14977c478bd9Sstevel@tonic-gate } 1498c793af95Ssangeeta ret_ire->ire_stq_ifindex = ill->ill_phyint->phyint_ifindex; 14997c478bd9Sstevel@tonic-gate ASSERT(ret_ire == ire); 15007c478bd9Sstevel@tonic-gate /* 15017c478bd9Sstevel@tonic-gate * ire_max_frag is normally zero here and is atomically set 15027c478bd9Sstevel@tonic-gate * under the irebucket lock in ire_add_v[46] except for the 15037c478bd9Sstevel@tonic-gate * case of IRE_MARK_NOADD. In that event the the ire_max_frag 15047c478bd9Sstevel@tonic-gate * is non-zero here. 15057c478bd9Sstevel@tonic-gate */ 15067c478bd9Sstevel@tonic-gate ire->ire_max_frag = max_frag; 15077c478bd9Sstevel@tonic-gate return (ire); 15087c478bd9Sstevel@tonic-gate } 15097c478bd9Sstevel@tonic-gate 15107c478bd9Sstevel@tonic-gate /* 15117c478bd9Sstevel@tonic-gate * ire_create is called to allocate and initialize a new IRE. 15127c478bd9Sstevel@tonic-gate * 15137c478bd9Sstevel@tonic-gate * NOTE : This is called as writer sometimes though not required 15147c478bd9Sstevel@tonic-gate * by this function. 15157c478bd9Sstevel@tonic-gate */ 15167c478bd9Sstevel@tonic-gate ire_t * 15177c478bd9Sstevel@tonic-gate ire_create(uchar_t *addr, uchar_t *mask, uchar_t *src_addr, uchar_t *gateway, 15185c0b7edeSseb uint_t *max_fragp, nce_t *src_nce, queue_t *rfq, queue_t *stq, 15195c0b7edeSseb ushort_t type, ipif_t *ipif, ipaddr_t cmask, uint32_t phandle, 15205c0b7edeSseb uint32_t ihandle, uint32_t flags, const iulp_t *ulp_info, tsol_gc_t *gc, 15215c0b7edeSseb tsol_gcgrp_t *gcgrp, ip_stack_t *ipst) 15227c478bd9Sstevel@tonic-gate { 15237c478bd9Sstevel@tonic-gate ire_t *ire; 15247c478bd9Sstevel@tonic-gate ire_t *ret_ire; 15257c478bd9Sstevel@tonic-gate 15267c478bd9Sstevel@tonic-gate ire = kmem_cache_alloc(ire_cache, KM_NOSLEEP); 15277c478bd9Sstevel@tonic-gate if (ire == NULL) { 15287c478bd9Sstevel@tonic-gate ip1dbg(("ire_create: alloc failed\n")); 15297c478bd9Sstevel@tonic-gate return (NULL); 15307c478bd9Sstevel@tonic-gate } 15317c478bd9Sstevel@tonic-gate *ire = ire_null; 15327c478bd9Sstevel@tonic-gate 15335c0b7edeSseb ret_ire = ire_init(ire, addr, mask, src_addr, gateway, max_fragp, 15345c0b7edeSseb src_nce, rfq, stq, type, ipif, cmask, phandle, ihandle, flags, 15355c0b7edeSseb ulp_info, gc, gcgrp, ipst); 15367c478bd9Sstevel@tonic-gate 15377c478bd9Sstevel@tonic-gate if (ret_ire == NULL) { 15387c478bd9Sstevel@tonic-gate kmem_cache_free(ire_cache, ire); 15397c478bd9Sstevel@tonic-gate return (NULL); 15407c478bd9Sstevel@tonic-gate } 15417c478bd9Sstevel@tonic-gate ASSERT(ret_ire == ire); 15427c478bd9Sstevel@tonic-gate return (ire); 15437c478bd9Sstevel@tonic-gate } 15447c478bd9Sstevel@tonic-gate 15457c478bd9Sstevel@tonic-gate 15467c478bd9Sstevel@tonic-gate /* 15477c478bd9Sstevel@tonic-gate * Common to IPv4 and IPv6 15487c478bd9Sstevel@tonic-gate */ 154945916cd2Sjpk boolean_t 155054da8755Ssowmini ire_init_common(ire_t *ire, uint_t *max_fragp, nce_t *src_nce, queue_t *rfq, 15515c0b7edeSseb queue_t *stq, ushort_t type, ipif_t *ipif, uint32_t phandle, 155254da8755Ssowmini uint32_t ihandle, uint32_t flags, uchar_t ipversion, const iulp_t *ulp_info, 155354da8755Ssowmini tsol_gc_t *gc, tsol_gcgrp_t *gcgrp, ip_stack_t *ipst) 15547c478bd9Sstevel@tonic-gate { 15557c478bd9Sstevel@tonic-gate ire->ire_max_fragp = max_fragp; 1556f4b3ec61Sdh155122 ire->ire_frag_flag |= (ipst->ips_ip_path_mtu_discovery) ? IPH_DF : 0; 15577c478bd9Sstevel@tonic-gate 155845916cd2Sjpk #ifdef DEBUG 155945916cd2Sjpk if (ipif != NULL) { 15607c478bd9Sstevel@tonic-gate if (ipif->ipif_isv6) 15617c478bd9Sstevel@tonic-gate ASSERT(ipversion == IPV6_VERSION); 15627c478bd9Sstevel@tonic-gate else 15637c478bd9Sstevel@tonic-gate ASSERT(ipversion == IPV4_VERSION); 15647c478bd9Sstevel@tonic-gate } 156545916cd2Sjpk #endif /* DEBUG */ 156645916cd2Sjpk 156745916cd2Sjpk /* 156845916cd2Sjpk * Create/initialize IRE security attribute only in Trusted mode; 156945916cd2Sjpk * if the passed in gc/gcgrp is non-NULL, we expect that the caller 157045916cd2Sjpk * has held a reference to it and will release it when this routine 157145916cd2Sjpk * returns a failure, otherwise we own the reference. We do this 157245916cd2Sjpk * prior to initializing the rest IRE fields. 1573dc3879f9Sjarrett * 1574dc3879f9Sjarrett * Don't allocate ire_gw_secattr for the resolver case to prevent 1575dc3879f9Sjarrett * memory leak (in case of external resolution failure). We'll 1576dc3879f9Sjarrett * allocate it after a successful external resolution, in ire_add(). 1577dc3879f9Sjarrett * Note that ire->ire_mp != NULL here means this ire is headed 1578dc3879f9Sjarrett * to an external resolver. 157945916cd2Sjpk */ 158045916cd2Sjpk if (is_system_labeled()) { 158145916cd2Sjpk if ((type & (IRE_LOCAL | IRE_LOOPBACK | IRE_BROADCAST | 158245916cd2Sjpk IRE_INTERFACE)) != 0) { 158345916cd2Sjpk /* release references on behalf of caller */ 158445916cd2Sjpk if (gc != NULL) 158545916cd2Sjpk GC_REFRELE(gc); 158645916cd2Sjpk if (gcgrp != NULL) 158745916cd2Sjpk GCGRP_REFRELE(gcgrp); 1588dc3879f9Sjarrett } else if ((ire->ire_mp == NULL) && 1589dc3879f9Sjarrett tsol_ire_init_gwattr(ire, ipversion, gc, gcgrp) != 0) { 159045916cd2Sjpk return (B_FALSE); 159145916cd2Sjpk } 159245916cd2Sjpk } 15937c478bd9Sstevel@tonic-gate 15947c478bd9Sstevel@tonic-gate ire->ire_stq = stq; 15957c478bd9Sstevel@tonic-gate ire->ire_rfq = rfq; 15967c478bd9Sstevel@tonic-gate ire->ire_type = type; 15977c478bd9Sstevel@tonic-gate ire->ire_flags = RTF_UP | flags; 15987c478bd9Sstevel@tonic-gate ire->ire_ident = TICK_TO_MSEC(lbolt); 15997c478bd9Sstevel@tonic-gate bcopy(ulp_info, &ire->ire_uinfo, sizeof (iulp_t)); 16007c478bd9Sstevel@tonic-gate 16017c478bd9Sstevel@tonic-gate ire->ire_tire_mark = ire->ire_ob_pkt_count + ire->ire_ib_pkt_count; 16027c478bd9Sstevel@tonic-gate ire->ire_last_used_time = lbolt; 16037c478bd9Sstevel@tonic-gate ire->ire_create_time = (uint32_t)gethrestime_sec(); 16047c478bd9Sstevel@tonic-gate 16057c478bd9Sstevel@tonic-gate /* 16067c478bd9Sstevel@tonic-gate * If this IRE is an IRE_CACHE, inherit the handles from the 16077c478bd9Sstevel@tonic-gate * parent IREs. For others in the forwarding table, assign appropriate 16087c478bd9Sstevel@tonic-gate * new ones. 16097c478bd9Sstevel@tonic-gate * 16107c478bd9Sstevel@tonic-gate * The mutex protecting ire_handle is because ire_create is not always 16117c478bd9Sstevel@tonic-gate * called as a writer. 16127c478bd9Sstevel@tonic-gate */ 16137c478bd9Sstevel@tonic-gate if (ire->ire_type & IRE_OFFSUBNET) { 1614f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_handle_lock); 1615f4b3ec61Sdh155122 ire->ire_phandle = (uint32_t)ipst->ips_ire_handle++; 1616f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_handle_lock); 16177c478bd9Sstevel@tonic-gate } else if (ire->ire_type & IRE_INTERFACE) { 1618f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_handle_lock); 1619f4b3ec61Sdh155122 ire->ire_ihandle = (uint32_t)ipst->ips_ire_handle++; 1620f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_handle_lock); 16217c478bd9Sstevel@tonic-gate } else if (ire->ire_type == IRE_CACHE) { 16227c478bd9Sstevel@tonic-gate ire->ire_phandle = phandle; 16237c478bd9Sstevel@tonic-gate ire->ire_ihandle = ihandle; 16247c478bd9Sstevel@tonic-gate } 16257c478bd9Sstevel@tonic-gate ire->ire_ipif = ipif; 16267c478bd9Sstevel@tonic-gate if (ipif != NULL) { 16277c478bd9Sstevel@tonic-gate ire->ire_ipif_seqid = ipif->ipif_seqid; 16287c478bd9Sstevel@tonic-gate ire->ire_zoneid = ipif->ipif_zoneid; 16297c478bd9Sstevel@tonic-gate } else { 16307c478bd9Sstevel@tonic-gate ire->ire_zoneid = GLOBAL_ZONEID; 16317c478bd9Sstevel@tonic-gate } 16327c478bd9Sstevel@tonic-gate ire->ire_ipversion = ipversion; 16337c478bd9Sstevel@tonic-gate mutex_init(&ire->ire_lock, NULL, MUTEX_DEFAULT, NULL); 1634c793af95Ssangeeta if (ipversion == IPV4_VERSION) { 163554da8755Ssowmini /* 163654da8755Ssowmini * IPv6 initializes the ire_nce in ire_add_v6, which expects 163754da8755Ssowmini * to find the ire_nce to be null when it is called. 163854da8755Ssowmini */ 163954da8755Ssowmini if (ire_nce_init(ire, src_nce) != 0) { 1640c793af95Ssangeeta /* some failure occurred. propagate error back */ 1641c793af95Ssangeeta return (B_FALSE); 1642c793af95Ssangeeta } 1643c793af95Ssangeeta } 1644c793af95Ssangeeta ire->ire_refcnt = 1; 1645f4b3ec61Sdh155122 ire->ire_ipst = ipst; /* No netstack_hold */ 16466a8288c7Scarlsonj ire->ire_trace_disable = B_FALSE; 164745916cd2Sjpk 164845916cd2Sjpk return (B_TRUE); 16497c478bd9Sstevel@tonic-gate } 16507c478bd9Sstevel@tonic-gate 16517c478bd9Sstevel@tonic-gate /* 16527c478bd9Sstevel@tonic-gate * This routine is called repeatedly by ipif_up to create broadcast IREs. 16537c478bd9Sstevel@tonic-gate * It is passed a pointer to a slot in an IRE pointer array into which to 16547c478bd9Sstevel@tonic-gate * place the pointer to the new IRE, if indeed we create one. If the 16557c478bd9Sstevel@tonic-gate * IRE corresponding to the address passed in would be a duplicate of an 16567c478bd9Sstevel@tonic-gate * existing one, we don't create the new one. irep is incremented before 16577c478bd9Sstevel@tonic-gate * return only if we do create a new IRE. (Always called as writer.) 16587c478bd9Sstevel@tonic-gate * 16597c478bd9Sstevel@tonic-gate * Note that with the "match_flags" parameter, we can match on either 16607c478bd9Sstevel@tonic-gate * a particular logical interface (MATCH_IRE_IPIF) or for all logical 16617c478bd9Sstevel@tonic-gate * interfaces for a given physical interface (MATCH_IRE_ILL). Currently, 16627c478bd9Sstevel@tonic-gate * we only create broadcast ire's on a per physical interface basis. If 16637c478bd9Sstevel@tonic-gate * someone is going to be mucking with logical interfaces, it is important 16647c478bd9Sstevel@tonic-gate * to call "ipif_check_bcast_ires()" to make sure that any change to a 16657c478bd9Sstevel@tonic-gate * logical interface will not cause critical broadcast IRE's to be deleted. 16667c478bd9Sstevel@tonic-gate */ 16677c478bd9Sstevel@tonic-gate ire_t ** 16687c478bd9Sstevel@tonic-gate ire_check_and_create_bcast(ipif_t *ipif, ipaddr_t addr, ire_t **irep, 16697c478bd9Sstevel@tonic-gate int match_flags) 16707c478bd9Sstevel@tonic-gate { 16717c478bd9Sstevel@tonic-gate ire_t *ire; 16727c478bd9Sstevel@tonic-gate uint64_t check_flags = IPIF_DEPRECATED | IPIF_NOLOCAL | IPIF_ANYCAST; 1673f4b3ec61Sdh155122 ip_stack_t *ipst = ipif->ipif_ill->ill_ipst; 16747c478bd9Sstevel@tonic-gate 16757c478bd9Sstevel@tonic-gate /* 16767c478bd9Sstevel@tonic-gate * No broadcast IREs for the LOOPBACK interface 16777c478bd9Sstevel@tonic-gate * or others such as point to point and IPIF_NOXMIT. 16787c478bd9Sstevel@tonic-gate */ 16797c478bd9Sstevel@tonic-gate if (!(ipif->ipif_flags & IPIF_BROADCAST) || 16807c478bd9Sstevel@tonic-gate (ipif->ipif_flags & IPIF_NOXMIT)) 16817c478bd9Sstevel@tonic-gate return (irep); 16827c478bd9Sstevel@tonic-gate 16837c478bd9Sstevel@tonic-gate /* If this would be a duplicate, don't bother. */ 16847c478bd9Sstevel@tonic-gate if ((ire = ire_ctable_lookup(addr, 0, IRE_BROADCAST, ipif, 1685f4b3ec61Sdh155122 ipif->ipif_zoneid, NULL, match_flags, ipst)) != NULL) { 16867c478bd9Sstevel@tonic-gate /* 16877c478bd9Sstevel@tonic-gate * We look for non-deprecated (and non-anycast, non-nolocal) 16887c478bd9Sstevel@tonic-gate * ipifs as the best choice. ipifs with check_flags matching 16897c478bd9Sstevel@tonic-gate * (deprecated, etc) are used only if non-deprecated ipifs 16907c478bd9Sstevel@tonic-gate * are not available. if the existing ire's ipif is deprecated 16917c478bd9Sstevel@tonic-gate * and the new ipif is non-deprecated, switch to the new ipif 16927c478bd9Sstevel@tonic-gate */ 16937c478bd9Sstevel@tonic-gate if ((!(ire->ire_ipif->ipif_flags & check_flags)) || 16947c478bd9Sstevel@tonic-gate (ipif->ipif_flags & check_flags)) { 16957c478bd9Sstevel@tonic-gate ire_refrele(ire); 16967c478bd9Sstevel@tonic-gate return (irep); 16977c478bd9Sstevel@tonic-gate } 16987c478bd9Sstevel@tonic-gate /* 16997c478bd9Sstevel@tonic-gate * Bcast ires exist in pairs. Both have to be deleted, 17007c478bd9Sstevel@tonic-gate * Since we are exclusive we can make the above assertion. 17017c478bd9Sstevel@tonic-gate * The 1st has to be refrele'd since it was ctable_lookup'd. 17027c478bd9Sstevel@tonic-gate */ 17037c478bd9Sstevel@tonic-gate ASSERT(IAM_WRITER_IPIF(ipif)); 17047c478bd9Sstevel@tonic-gate ASSERT(ire->ire_next->ire_addr == ire->ire_addr); 17057c478bd9Sstevel@tonic-gate ire_delete(ire->ire_next); 17067c478bd9Sstevel@tonic-gate ire_delete(ire); 17077c478bd9Sstevel@tonic-gate ire_refrele(ire); 17087c478bd9Sstevel@tonic-gate } 17097c478bd9Sstevel@tonic-gate 17107c478bd9Sstevel@tonic-gate irep = ire_create_bcast(ipif, addr, irep); 17117c478bd9Sstevel@tonic-gate 17127c478bd9Sstevel@tonic-gate return (irep); 17137c478bd9Sstevel@tonic-gate } 17147c478bd9Sstevel@tonic-gate 17157c478bd9Sstevel@tonic-gate uint_t ip_loopback_mtu = IP_LOOPBACK_MTU; 17167c478bd9Sstevel@tonic-gate 17177c478bd9Sstevel@tonic-gate /* 17187c478bd9Sstevel@tonic-gate * This routine is called from ipif_check_bcast_ires and ire_check_bcast. 17197c478bd9Sstevel@tonic-gate * It leaves all the verifying and deleting to those routines. So it always 17207c478bd9Sstevel@tonic-gate * creates 2 bcast ires and chains them into the ire array passed in. 17217c478bd9Sstevel@tonic-gate */ 17227c478bd9Sstevel@tonic-gate ire_t ** 17237c478bd9Sstevel@tonic-gate ire_create_bcast(ipif_t *ipif, ipaddr_t addr, ire_t **irep) 17247c478bd9Sstevel@tonic-gate { 1725f4b3ec61Sdh155122 ip_stack_t *ipst = ipif->ipif_ill->ill_ipst; 1726f4b3ec61Sdh155122 17277c478bd9Sstevel@tonic-gate *irep++ = ire_create( 17287c478bd9Sstevel@tonic-gate (uchar_t *)&addr, /* dest addr */ 17297c478bd9Sstevel@tonic-gate (uchar_t *)&ip_g_all_ones, /* mask */ 17307c478bd9Sstevel@tonic-gate (uchar_t *)&ipif->ipif_src_addr, /* source addr */ 17317c478bd9Sstevel@tonic-gate NULL, /* no gateway */ 17327c478bd9Sstevel@tonic-gate &ipif->ipif_mtu, /* max frag */ 173354da8755Ssowmini NULL, /* no src nce */ 17347c478bd9Sstevel@tonic-gate ipif->ipif_rq, /* recv-from queue */ 17357c478bd9Sstevel@tonic-gate ipif->ipif_wq, /* send-to queue */ 17367c478bd9Sstevel@tonic-gate IRE_BROADCAST, 17377c478bd9Sstevel@tonic-gate ipif, 17387c478bd9Sstevel@tonic-gate 0, 17397c478bd9Sstevel@tonic-gate 0, 17407c478bd9Sstevel@tonic-gate 0, 17417c478bd9Sstevel@tonic-gate 0, 174245916cd2Sjpk &ire_uinfo_null, 174345916cd2Sjpk NULL, 1744f4b3ec61Sdh155122 NULL, 1745f4b3ec61Sdh155122 ipst); 17467c478bd9Sstevel@tonic-gate 17477c478bd9Sstevel@tonic-gate *irep++ = ire_create( 17487c478bd9Sstevel@tonic-gate (uchar_t *)&addr, /* dest address */ 17497c478bd9Sstevel@tonic-gate (uchar_t *)&ip_g_all_ones, /* mask */ 17507c478bd9Sstevel@tonic-gate (uchar_t *)&ipif->ipif_src_addr, /* source address */ 17517c478bd9Sstevel@tonic-gate NULL, /* no gateway */ 17527c478bd9Sstevel@tonic-gate &ip_loopback_mtu, /* max frag size */ 175354da8755Ssowmini NULL, /* no src_nce */ 17547c478bd9Sstevel@tonic-gate ipif->ipif_rq, /* recv-from queue */ 17557c478bd9Sstevel@tonic-gate NULL, /* no send-to queue */ 17567c478bd9Sstevel@tonic-gate IRE_BROADCAST, /* Needed for fanout in wput */ 17577c478bd9Sstevel@tonic-gate ipif, 17587c478bd9Sstevel@tonic-gate 0, 17597c478bd9Sstevel@tonic-gate 0, 17607c478bd9Sstevel@tonic-gate 0, 17617c478bd9Sstevel@tonic-gate 0, 176245916cd2Sjpk &ire_uinfo_null, 176345916cd2Sjpk NULL, 1764f4b3ec61Sdh155122 NULL, 1765f4b3ec61Sdh155122 ipst); 17667c478bd9Sstevel@tonic-gate 17677c478bd9Sstevel@tonic-gate return (irep); 17687c478bd9Sstevel@tonic-gate } 17697c478bd9Sstevel@tonic-gate 17707c478bd9Sstevel@tonic-gate /* 17717c478bd9Sstevel@tonic-gate * ire_walk routine to delete or update any IRE_CACHE that might contain 17727c478bd9Sstevel@tonic-gate * stale information. 17737c478bd9Sstevel@tonic-gate * The flags state which entries to delete or update. 17747c478bd9Sstevel@tonic-gate * Garbage collection is done separately using kmem alloc callbacks to 17757c478bd9Sstevel@tonic-gate * ip_trash_ire_reclaim. 17767c478bd9Sstevel@tonic-gate * Used for both IPv4 and IPv6. However, IPv6 only uses FLUSH_MTU_TIME 17777c478bd9Sstevel@tonic-gate * since other stale information is cleaned up using NUD. 17787c478bd9Sstevel@tonic-gate */ 17797c478bd9Sstevel@tonic-gate void 17807c478bd9Sstevel@tonic-gate ire_expire(ire_t *ire, char *arg) 17817c478bd9Sstevel@tonic-gate { 1782f4b3ec61Sdh155122 ire_expire_arg_t *ieap = (ire_expire_arg_t *)(uintptr_t)arg; 17837c478bd9Sstevel@tonic-gate ill_t *stq_ill; 1784f4b3ec61Sdh155122 int flush_flags = ieap->iea_flush_flag; 1785f4b3ec61Sdh155122 ip_stack_t *ipst = ieap->iea_ipst; 17867c478bd9Sstevel@tonic-gate 17877c478bd9Sstevel@tonic-gate if ((flush_flags & FLUSH_REDIRECT_TIME) && 17886bdb8e66Sdd193516 (ire->ire_flags & RTF_DYNAMIC)) { 17897c478bd9Sstevel@tonic-gate /* Make sure we delete the corresponding IRE_CACHE */ 17907c478bd9Sstevel@tonic-gate ip1dbg(("ire_expire: all redirects\n")); 1791f4b3ec61Sdh155122 ip_rts_rtmsg(RTM_DELETE, ire, 0, ipst); 17927c478bd9Sstevel@tonic-gate ire_delete(ire); 1793f4b3ec61Sdh155122 atomic_dec_32(&ipst->ips_ip_redirect_cnt); 17947c478bd9Sstevel@tonic-gate return; 17957c478bd9Sstevel@tonic-gate } 17967c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 17977c478bd9Sstevel@tonic-gate return; 17987c478bd9Sstevel@tonic-gate 17997c478bd9Sstevel@tonic-gate if (flush_flags & FLUSH_ARP_TIME) { 18007c478bd9Sstevel@tonic-gate /* 1801beef8517Sjprakash * Remove all IRE_CACHE except IPv4 multicast ires. These 1802beef8517Sjprakash * ires will be deleted by ip_trash_ire_reclaim_stack() 1803beef8517Sjprakash * when system runs low in memory. 1804beef8517Sjprakash * Verify that create time is more than ip_ire_arp_interval 1805beef8517Sjprakash * milliseconds ago. 18067c478bd9Sstevel@tonic-gate */ 1807beef8517Sjprakash 1808beef8517Sjprakash if (!(ire->ire_ipversion == IPV4_VERSION && 1809beef8517Sjprakash CLASSD(ire->ire_addr)) && NCE_EXPIRED(ire->ire_nce, ipst)) { 18107c478bd9Sstevel@tonic-gate ire_delete(ire); 18117c478bd9Sstevel@tonic-gate return; 18127c478bd9Sstevel@tonic-gate } 18137c478bd9Sstevel@tonic-gate } 18147c478bd9Sstevel@tonic-gate 1815f4b3ec61Sdh155122 if (ipst->ips_ip_path_mtu_discovery && (flush_flags & FLUSH_MTU_TIME) && 18167c478bd9Sstevel@tonic-gate (ire->ire_ipif != NULL)) { 18177c478bd9Sstevel@tonic-gate /* Increase pmtu if it is less than the interface mtu */ 18187c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 18197c478bd9Sstevel@tonic-gate /* 18207c478bd9Sstevel@tonic-gate * If the ipif is a vni (whose mtu is 0, since it's virtual) 18217c478bd9Sstevel@tonic-gate * get the mtu from the sending interfaces' ipif 18227c478bd9Sstevel@tonic-gate */ 18237c478bd9Sstevel@tonic-gate if (IS_VNI(ire->ire_ipif->ipif_ill)) { 18247c478bd9Sstevel@tonic-gate stq_ill = ire->ire_stq->q_ptr; 18257c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(stq_ill->ill_ipif->ipif_mtu, 18267c478bd9Sstevel@tonic-gate IP_MAXPACKET); 18277c478bd9Sstevel@tonic-gate } else { 18287c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(ire->ire_ipif->ipif_mtu, 18297c478bd9Sstevel@tonic-gate IP_MAXPACKET); 18307c478bd9Sstevel@tonic-gate } 18317c478bd9Sstevel@tonic-gate ire->ire_frag_flag |= IPH_DF; 18327c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 18337c478bd9Sstevel@tonic-gate } 18347c478bd9Sstevel@tonic-gate } 18357c478bd9Sstevel@tonic-gate 18367c478bd9Sstevel@tonic-gate /* 18377c478bd9Sstevel@tonic-gate * Return any local address. We use this to target ourselves 18387c478bd9Sstevel@tonic-gate * when the src address was specified as 'default'. 18397c478bd9Sstevel@tonic-gate * Preference for IRE_LOCAL entries. 18407c478bd9Sstevel@tonic-gate */ 18417c478bd9Sstevel@tonic-gate ire_t * 1842f4b3ec61Sdh155122 ire_lookup_local(zoneid_t zoneid, ip_stack_t *ipst) 18437c478bd9Sstevel@tonic-gate { 18447c478bd9Sstevel@tonic-gate ire_t *ire; 18457c478bd9Sstevel@tonic-gate irb_t *irb; 18467c478bd9Sstevel@tonic-gate ire_t *maybe = NULL; 18477c478bd9Sstevel@tonic-gate int i; 18487c478bd9Sstevel@tonic-gate 1849f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 1850f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[i]; 18517c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 18527c478bd9Sstevel@tonic-gate continue; 18537c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_READER); 18547c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; ire = ire->ire_next) { 18557c478bd9Sstevel@tonic-gate if ((ire->ire_marks & IRE_MARK_CONDEMNED) || 185645916cd2Sjpk (ire->ire_zoneid != zoneid && 185745916cd2Sjpk ire->ire_zoneid != ALL_ZONES)) 18587c478bd9Sstevel@tonic-gate continue; 18597c478bd9Sstevel@tonic-gate switch (ire->ire_type) { 18607c478bd9Sstevel@tonic-gate case IRE_LOOPBACK: 18617c478bd9Sstevel@tonic-gate if (maybe == NULL) { 18627c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 18637c478bd9Sstevel@tonic-gate maybe = ire; 18647c478bd9Sstevel@tonic-gate } 18657c478bd9Sstevel@tonic-gate break; 18667c478bd9Sstevel@tonic-gate case IRE_LOCAL: 18677c478bd9Sstevel@tonic-gate if (maybe != NULL) { 18687c478bd9Sstevel@tonic-gate ire_refrele(maybe); 18697c478bd9Sstevel@tonic-gate } 18707c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 18717c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 18727c478bd9Sstevel@tonic-gate return (ire); 18737c478bd9Sstevel@tonic-gate } 18747c478bd9Sstevel@tonic-gate } 18757c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 18767c478bd9Sstevel@tonic-gate } 18777c478bd9Sstevel@tonic-gate return (maybe); 18787c478bd9Sstevel@tonic-gate } 18797c478bd9Sstevel@tonic-gate 18807c478bd9Sstevel@tonic-gate /* 18817c478bd9Sstevel@tonic-gate * If the specified IRE is associated with a particular ILL, return 18827c478bd9Sstevel@tonic-gate * that ILL pointer (May be called as writer.). 18837c478bd9Sstevel@tonic-gate * 18847c478bd9Sstevel@tonic-gate * NOTE : This is not a generic function that can be used always. 18857c478bd9Sstevel@tonic-gate * This function always returns the ill of the outgoing packets 18867c478bd9Sstevel@tonic-gate * if this ire is used. 18877c478bd9Sstevel@tonic-gate */ 18887c478bd9Sstevel@tonic-gate ill_t * 188945916cd2Sjpk ire_to_ill(const ire_t *ire) 18907c478bd9Sstevel@tonic-gate { 18917c478bd9Sstevel@tonic-gate ill_t *ill = NULL; 18927c478bd9Sstevel@tonic-gate 18937c478bd9Sstevel@tonic-gate /* 18947c478bd9Sstevel@tonic-gate * 1) For an IRE_CACHE, ire_ipif is the one where it obtained 18957c478bd9Sstevel@tonic-gate * the source address from. ire_stq is the one where the 18967c478bd9Sstevel@tonic-gate * packets will be sent out on. We return that here. 18977c478bd9Sstevel@tonic-gate * 18987c478bd9Sstevel@tonic-gate * 2) IRE_BROADCAST normally has a loopback and a non-loopback 18997c478bd9Sstevel@tonic-gate * copy and they always exist next to each other with loopback 19007c478bd9Sstevel@tonic-gate * copy being the first one. If we are called on the non-loopback 19017c478bd9Sstevel@tonic-gate * copy, return the one pointed by ire_stq. If it was called on 19027c478bd9Sstevel@tonic-gate * a loopback copy, we still return the one pointed by the next 19037c478bd9Sstevel@tonic-gate * ire's ire_stq pointer i.e the one pointed by the non-loopback 19047c478bd9Sstevel@tonic-gate * copy. We don't want use ire_ipif as it might represent the 19057c478bd9Sstevel@tonic-gate * source address (if we borrow source addresses for 19067c478bd9Sstevel@tonic-gate * IRE_BROADCASTS in the future). 19077c478bd9Sstevel@tonic-gate * However if an interface is currently coming up, the above 19087c478bd9Sstevel@tonic-gate * condition may not hold during that period since the ires 19097c478bd9Sstevel@tonic-gate * are added one at a time. Thus one of the pair could have been 19107c478bd9Sstevel@tonic-gate * added and the other not yet added. 19112bc4236aSnordmark * 3) For many other IREs (e.g., IRE_LOCAL), ire_rfq indicates the ill. 19122bc4236aSnordmark * 4) For all others return the ones pointed by ire_ipif->ipif_ill. 19132bc4236aSnordmark * That handles IRE_LOOPBACK. 19147c478bd9Sstevel@tonic-gate */ 19157c478bd9Sstevel@tonic-gate 19167c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE) { 19177c478bd9Sstevel@tonic-gate ill = (ill_t *)ire->ire_stq->q_ptr; 19187c478bd9Sstevel@tonic-gate } else if (ire->ire_type == IRE_BROADCAST) { 19197c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) { 19207c478bd9Sstevel@tonic-gate ill = (ill_t *)ire->ire_stq->q_ptr; 19217c478bd9Sstevel@tonic-gate } else { 19227c478bd9Sstevel@tonic-gate ire_t *ire_next; 19237c478bd9Sstevel@tonic-gate 19247c478bd9Sstevel@tonic-gate ire_next = ire->ire_next; 19257c478bd9Sstevel@tonic-gate if (ire_next != NULL && 19267c478bd9Sstevel@tonic-gate ire_next->ire_type == IRE_BROADCAST && 19277c478bd9Sstevel@tonic-gate ire_next->ire_addr == ire->ire_addr && 19287c478bd9Sstevel@tonic-gate ire_next->ire_ipif == ire->ire_ipif) { 19297c478bd9Sstevel@tonic-gate ill = (ill_t *)ire_next->ire_stq->q_ptr; 19307c478bd9Sstevel@tonic-gate } 19317c478bd9Sstevel@tonic-gate } 19322bc4236aSnordmark } else if (ire->ire_rfq != NULL) { 19332bc4236aSnordmark ill = ire->ire_rfq->q_ptr; 19347c478bd9Sstevel@tonic-gate } else if (ire->ire_ipif != NULL) { 19357c478bd9Sstevel@tonic-gate ill = ire->ire_ipif->ipif_ill; 19367c478bd9Sstevel@tonic-gate } 19377c478bd9Sstevel@tonic-gate return (ill); 19387c478bd9Sstevel@tonic-gate } 19397c478bd9Sstevel@tonic-gate 19407c478bd9Sstevel@tonic-gate /* Arrange to call the specified function for every IRE in the world. */ 19417c478bd9Sstevel@tonic-gate void 1942f4b3ec61Sdh155122 ire_walk(pfv_t func, void *arg, ip_stack_t *ipst) 19437c478bd9Sstevel@tonic-gate { 1944f4b3ec61Sdh155122 ire_walk_ipvers(func, arg, 0, ALL_ZONES, ipst); 19457c478bd9Sstevel@tonic-gate } 19467c478bd9Sstevel@tonic-gate 19477c478bd9Sstevel@tonic-gate void 1948f4b3ec61Sdh155122 ire_walk_v4(pfv_t func, void *arg, zoneid_t zoneid, ip_stack_t *ipst) 19497c478bd9Sstevel@tonic-gate { 1950f4b3ec61Sdh155122 ire_walk_ipvers(func, arg, IPV4_VERSION, zoneid, ipst); 19517c478bd9Sstevel@tonic-gate } 19527c478bd9Sstevel@tonic-gate 19537c478bd9Sstevel@tonic-gate void 1954f4b3ec61Sdh155122 ire_walk_v6(pfv_t func, void *arg, zoneid_t zoneid, ip_stack_t *ipst) 19557c478bd9Sstevel@tonic-gate { 1956f4b3ec61Sdh155122 ire_walk_ipvers(func, arg, IPV6_VERSION, zoneid, ipst); 19577c478bd9Sstevel@tonic-gate } 19587c478bd9Sstevel@tonic-gate 19597c478bd9Sstevel@tonic-gate /* 19607c478bd9Sstevel@tonic-gate * Walk a particular version. version == 0 means both v4 and v6. 19617c478bd9Sstevel@tonic-gate */ 19627c478bd9Sstevel@tonic-gate static void 1963f4b3ec61Sdh155122 ire_walk_ipvers(pfv_t func, void *arg, uchar_t vers, zoneid_t zoneid, 1964f4b3ec61Sdh155122 ip_stack_t *ipst) 19657c478bd9Sstevel@tonic-gate { 19667c478bd9Sstevel@tonic-gate if (vers != IPV6_VERSION) { 1967c793af95Ssangeeta /* 1968c793af95Ssangeeta * ip_forwarding_table variable doesn't matter for IPv4 since 1969f4b3ec61Sdh155122 * ire_walk_ill_tables uses ips_ip_ftable for IPv4. 1970c793af95Ssangeeta */ 19717c478bd9Sstevel@tonic-gate ire_walk_ill_tables(0, 0, func, arg, IP_MASK_TABLE_SIZE, 1972c793af95Ssangeeta 0, NULL, 1973f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size, ipst->ips_ip_cache_table, 1974f4b3ec61Sdh155122 NULL, zoneid, ipst); 19757c478bd9Sstevel@tonic-gate } 19767c478bd9Sstevel@tonic-gate if (vers != IPV4_VERSION) { 19777c478bd9Sstevel@tonic-gate ire_walk_ill_tables(0, 0, func, arg, IP6_MASK_TABLE_SIZE, 1978f4b3ec61Sdh155122 ipst->ips_ip6_ftable_hash_size, 1979f4b3ec61Sdh155122 ipst->ips_ip_forwarding_table_v6, 1980f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size, 1981f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6, NULL, zoneid, ipst); 19827c478bd9Sstevel@tonic-gate } 19837c478bd9Sstevel@tonic-gate } 19847c478bd9Sstevel@tonic-gate 19857c478bd9Sstevel@tonic-gate /* 1986*7924222fSmeem * Arrange to call the specified function for every IRE that matches the ill. 19877c478bd9Sstevel@tonic-gate */ 19887c478bd9Sstevel@tonic-gate void 198945916cd2Sjpk ire_walk_ill(uint_t match_flags, uint_t ire_type, pfv_t func, void *arg, 19907c478bd9Sstevel@tonic-gate ill_t *ill) 19917c478bd9Sstevel@tonic-gate { 1992*7924222fSmeem uchar_t vers = (ill->ill_isv6 ? IPV6_VERSION : IPV4_VERSION); 1993*7924222fSmeem 1994*7924222fSmeem ire_walk_ill_ipvers(match_flags, ire_type, func, arg, vers, ill); 19957c478bd9Sstevel@tonic-gate } 19967c478bd9Sstevel@tonic-gate 19977c478bd9Sstevel@tonic-gate void 199845916cd2Sjpk ire_walk_ill_v4(uint_t match_flags, uint_t ire_type, pfv_t func, void *arg, 19997c478bd9Sstevel@tonic-gate ill_t *ill) 20007c478bd9Sstevel@tonic-gate { 20017c478bd9Sstevel@tonic-gate ire_walk_ill_ipvers(match_flags, ire_type, func, arg, IPV4_VERSION, 20027c478bd9Sstevel@tonic-gate ill); 20037c478bd9Sstevel@tonic-gate } 20047c478bd9Sstevel@tonic-gate 20057c478bd9Sstevel@tonic-gate void 200645916cd2Sjpk ire_walk_ill_v6(uint_t match_flags, uint_t ire_type, pfv_t func, void *arg, 20077c478bd9Sstevel@tonic-gate ill_t *ill) 20087c478bd9Sstevel@tonic-gate { 20097c478bd9Sstevel@tonic-gate ire_walk_ill_ipvers(match_flags, ire_type, func, arg, IPV6_VERSION, 20107c478bd9Sstevel@tonic-gate ill); 20117c478bd9Sstevel@tonic-gate } 20127c478bd9Sstevel@tonic-gate 20137c478bd9Sstevel@tonic-gate /* 2014*7924222fSmeem * Walk a particular ill and version. 20157c478bd9Sstevel@tonic-gate */ 20167c478bd9Sstevel@tonic-gate static void 20177c478bd9Sstevel@tonic-gate ire_walk_ill_ipvers(uint_t match_flags, uint_t ire_type, pfv_t func, 201845916cd2Sjpk void *arg, uchar_t vers, ill_t *ill) 20197c478bd9Sstevel@tonic-gate { 2020f4b3ec61Sdh155122 ip_stack_t *ipst = ill->ill_ipst; 2021f4b3ec61Sdh155122 2022*7924222fSmeem if (vers == IPV4_VERSION) { 20237c478bd9Sstevel@tonic-gate ire_walk_ill_tables(match_flags, ire_type, func, arg, 2024c793af95Ssangeeta IP_MASK_TABLE_SIZE, 0, 2025f4b3ec61Sdh155122 NULL, ipst->ips_ip_cache_table_size, 2026f4b3ec61Sdh155122 ipst->ips_ip_cache_table, ill, ALL_ZONES, ipst); 2027*7924222fSmeem } else if (vers == IPV6_VERSION) { 20287c478bd9Sstevel@tonic-gate ire_walk_ill_tables(match_flags, ire_type, func, arg, 2029f4b3ec61Sdh155122 IP6_MASK_TABLE_SIZE, ipst->ips_ip6_ftable_hash_size, 2030f4b3ec61Sdh155122 ipst->ips_ip_forwarding_table_v6, 2031f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size, 2032f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6, ill, ALL_ZONES, ipst); 20337c478bd9Sstevel@tonic-gate } 20347c478bd9Sstevel@tonic-gate } 20357c478bd9Sstevel@tonic-gate 2036c793af95Ssangeeta boolean_t 20377c478bd9Sstevel@tonic-gate ire_walk_ill_match(uint_t match_flags, uint_t ire_type, ire_t *ire, 2038f4b3ec61Sdh155122 ill_t *ill, zoneid_t zoneid, ip_stack_t *ipst) 20397c478bd9Sstevel@tonic-gate { 20407c478bd9Sstevel@tonic-gate ill_t *ire_stq_ill = NULL; 20417c478bd9Sstevel@tonic-gate ill_t *ire_ipif_ill = NULL; 20427c478bd9Sstevel@tonic-gate ill_group_t *ire_ill_group = NULL; 20437c478bd9Sstevel@tonic-gate 20447c478bd9Sstevel@tonic-gate ASSERT(match_flags != 0 || zoneid != ALL_ZONES); 20457c478bd9Sstevel@tonic-gate /* 204698e93c29Smeem * MATCH_IRE_ILL/MATCH_IRE_ILL_GROUP : We match both on ill 20477c478bd9Sstevel@tonic-gate * pointed by ire_stq and ire_ipif. Only in the case of 20487c478bd9Sstevel@tonic-gate * IRE_CACHEs can ire_stq and ire_ipif be pointing to 20497c478bd9Sstevel@tonic-gate * different ills. But we want to keep this function generic 20507c478bd9Sstevel@tonic-gate * enough for future use. So, we always try to match on both. 20517c478bd9Sstevel@tonic-gate * The only caller of this function ire_walk_ill_tables, will 20527c478bd9Sstevel@tonic-gate * call "func" after we return from this function. We expect 20537c478bd9Sstevel@tonic-gate * "func" to do the right filtering of ires in this case. 20547c478bd9Sstevel@tonic-gate * 20557c478bd9Sstevel@tonic-gate * NOTE : In the case of MATCH_IRE_ILL_GROUP, groups 20567c478bd9Sstevel@tonic-gate * pointed by ire_stq and ire_ipif should always be the same. 20577c478bd9Sstevel@tonic-gate * So, we just match on only one of them. 20587c478bd9Sstevel@tonic-gate */ 20597c478bd9Sstevel@tonic-gate if (match_flags & (MATCH_IRE_ILL|MATCH_IRE_ILL_GROUP)) { 20607c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 20617c478bd9Sstevel@tonic-gate ire_stq_ill = (ill_t *)ire->ire_stq->q_ptr; 20627c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) 20637c478bd9Sstevel@tonic-gate ire_ipif_ill = ire->ire_ipif->ipif_ill; 20647c478bd9Sstevel@tonic-gate if (ire_stq_ill != NULL) 20657c478bd9Sstevel@tonic-gate ire_ill_group = ire_stq_ill->ill_group; 20667c478bd9Sstevel@tonic-gate if ((ire_ill_group == NULL) && (ire_ipif_ill != NULL)) 20677c478bd9Sstevel@tonic-gate ire_ill_group = ire_ipif_ill->ill_group; 20687c478bd9Sstevel@tonic-gate } 20697c478bd9Sstevel@tonic-gate 20707c478bd9Sstevel@tonic-gate if (zoneid != ALL_ZONES) { 20717c478bd9Sstevel@tonic-gate /* 20727c478bd9Sstevel@tonic-gate * We're walking the IREs for a specific zone. The only relevant 20737c478bd9Sstevel@tonic-gate * IREs are: 20747c478bd9Sstevel@tonic-gate * - all IREs with a matching ire_zoneid 20757c478bd9Sstevel@tonic-gate * - all IRE_OFFSUBNETs as they're shared across all zones 20767c478bd9Sstevel@tonic-gate * - IRE_INTERFACE IREs for interfaces with a usable source addr 20777c478bd9Sstevel@tonic-gate * with a matching zone 20787c478bd9Sstevel@tonic-gate * - IRE_DEFAULTs with a gateway reachable from the zone 20797c478bd9Sstevel@tonic-gate * We should really match on IRE_OFFSUBNETs and IRE_DEFAULTs 20807c478bd9Sstevel@tonic-gate * using the same rule; but the above rules are consistent with 20817c478bd9Sstevel@tonic-gate * the behavior of ire_ftable_lookup[_v6]() so that all the 20827c478bd9Sstevel@tonic-gate * routes that can be matched during lookup are also matched 20837c478bd9Sstevel@tonic-gate * here. 20847c478bd9Sstevel@tonic-gate */ 208545916cd2Sjpk if (zoneid != ire->ire_zoneid && ire->ire_zoneid != ALL_ZONES) { 20867c478bd9Sstevel@tonic-gate /* 20877c478bd9Sstevel@tonic-gate * Note, IRE_INTERFACE can have the stq as NULL. For 20887c478bd9Sstevel@tonic-gate * example, if the default multicast route is tied to 20897c478bd9Sstevel@tonic-gate * the loopback address. 20907c478bd9Sstevel@tonic-gate */ 20917c478bd9Sstevel@tonic-gate if ((ire->ire_type & IRE_INTERFACE) && 20927c478bd9Sstevel@tonic-gate (ire->ire_stq != NULL)) { 20937c478bd9Sstevel@tonic-gate ire_stq_ill = (ill_t *)ire->ire_stq->q_ptr; 20947c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 20957c478bd9Sstevel@tonic-gate if (!ipif_usesrc_avail(ire_stq_ill, 20967c478bd9Sstevel@tonic-gate zoneid)) 20977c478bd9Sstevel@tonic-gate /* No usable src addr in zone */ 20987c478bd9Sstevel@tonic-gate return (B_FALSE); 20997c478bd9Sstevel@tonic-gate } else if (ire_stq_ill->ill_usesrc_ifindex 21007c478bd9Sstevel@tonic-gate != 0) { 21017c478bd9Sstevel@tonic-gate /* 21027c478bd9Sstevel@tonic-gate * For IPv6 use ipif_select_source_v6() 21037c478bd9Sstevel@tonic-gate * so the right scope selection is done 21047c478bd9Sstevel@tonic-gate */ 21057c478bd9Sstevel@tonic-gate ipif_t *src_ipif; 21067c478bd9Sstevel@tonic-gate src_ipif = 21077c478bd9Sstevel@tonic-gate ipif_select_source_v6(ire_stq_ill, 2108309fa174Srk129064 &ire->ire_addr_v6, RESTRICT_TO_NONE, 21097c478bd9Sstevel@tonic-gate IPV6_PREFER_SRC_DEFAULT, 21107c478bd9Sstevel@tonic-gate zoneid); 21117c478bd9Sstevel@tonic-gate if (src_ipif != NULL) { 21127c478bd9Sstevel@tonic-gate ipif_refrele(src_ipif); 21137c478bd9Sstevel@tonic-gate } else { 21147c478bd9Sstevel@tonic-gate return (B_FALSE); 21157c478bd9Sstevel@tonic-gate } 21167c478bd9Sstevel@tonic-gate } else { 21177c478bd9Sstevel@tonic-gate return (B_FALSE); 21187c478bd9Sstevel@tonic-gate } 21197c478bd9Sstevel@tonic-gate 21207c478bd9Sstevel@tonic-gate } else if (!(ire->ire_type & IRE_OFFSUBNET)) { 21217c478bd9Sstevel@tonic-gate return (B_FALSE); 21227c478bd9Sstevel@tonic-gate } 21237c478bd9Sstevel@tonic-gate } 21247c478bd9Sstevel@tonic-gate 21257c478bd9Sstevel@tonic-gate /* 21267c478bd9Sstevel@tonic-gate * Match all default routes from the global zone, irrespective 21275597b60aSnordmark * of reachability. For a non-global zone only match those 21285597b60aSnordmark * where ire_gateway_addr has a IRE_INTERFACE for the zoneid. 21297c478bd9Sstevel@tonic-gate */ 21307c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT && zoneid != GLOBAL_ZONEID) { 21317c478bd9Sstevel@tonic-gate int ire_match_flags = 0; 21327c478bd9Sstevel@tonic-gate in6_addr_t gw_addr_v6; 21337c478bd9Sstevel@tonic-gate ire_t *rire; 21347c478bd9Sstevel@tonic-gate 21355597b60aSnordmark ire_match_flags |= MATCH_IRE_TYPE; 21367c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 21377c478bd9Sstevel@tonic-gate ire_match_flags |= MATCH_IRE_ILL_GROUP; 21387c478bd9Sstevel@tonic-gate } 21397c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 21407c478bd9Sstevel@tonic-gate rire = ire_route_lookup(ire->ire_gateway_addr, 21415597b60aSnordmark 0, 0, IRE_INTERFACE, ire->ire_ipif, NULL, 2142f4b3ec61Sdh155122 zoneid, NULL, ire_match_flags, ipst); 21437c478bd9Sstevel@tonic-gate } else { 21447c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV6_VERSION); 21457c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 21467c478bd9Sstevel@tonic-gate gw_addr_v6 = ire->ire_gateway_addr_v6; 21477c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 21487c478bd9Sstevel@tonic-gate rire = ire_route_lookup_v6(&gw_addr_v6, 21495597b60aSnordmark NULL, NULL, IRE_INTERFACE, ire->ire_ipif, 2150f4b3ec61Sdh155122 NULL, zoneid, NULL, ire_match_flags, ipst); 21517c478bd9Sstevel@tonic-gate } 21527c478bd9Sstevel@tonic-gate if (rire == NULL) { 21537c478bd9Sstevel@tonic-gate return (B_FALSE); 21547c478bd9Sstevel@tonic-gate } 21557c478bd9Sstevel@tonic-gate ire_refrele(rire); 21567c478bd9Sstevel@tonic-gate } 21577c478bd9Sstevel@tonic-gate } 21587c478bd9Sstevel@tonic-gate 21597c478bd9Sstevel@tonic-gate if (((!(match_flags & MATCH_IRE_TYPE)) || 21607c478bd9Sstevel@tonic-gate (ire->ire_type & ire_type)) && 21617c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL)) || 21627c478bd9Sstevel@tonic-gate (ire_stq_ill == ill || ire_ipif_ill == ill)) && 21637c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL_GROUP)) || 21647c478bd9Sstevel@tonic-gate (ire_stq_ill == ill) || (ire_ipif_ill == ill) || 21657c478bd9Sstevel@tonic-gate (ire_ill_group != NULL && 21667c478bd9Sstevel@tonic-gate ire_ill_group == ill->ill_group))) { 21677c478bd9Sstevel@tonic-gate return (B_TRUE); 21687c478bd9Sstevel@tonic-gate } 21697c478bd9Sstevel@tonic-gate return (B_FALSE); 21707c478bd9Sstevel@tonic-gate } 21717c478bd9Sstevel@tonic-gate 2172c793af95Ssangeeta int 2173c793af95Ssangeeta rtfunc(struct radix_node *rn, void *arg) 2174c793af95Ssangeeta { 2175c793af95Ssangeeta struct rtfuncarg *rtf = arg; 2176c793af95Ssangeeta struct rt_entry *rt; 2177c793af95Ssangeeta irb_t *irb; 2178c793af95Ssangeeta ire_t *ire; 2179c793af95Ssangeeta boolean_t ret; 2180c793af95Ssangeeta 2181c793af95Ssangeeta rt = (struct rt_entry *)rn; 2182c793af95Ssangeeta ASSERT(rt != NULL); 2183c793af95Ssangeeta irb = &rt->rt_irb; 2184c793af95Ssangeeta for (ire = irb->irb_ire; ire != NULL; ire = ire->ire_next) { 2185c793af95Ssangeeta if ((rtf->rt_match_flags != 0) || 2186c793af95Ssangeeta (rtf->rt_zoneid != ALL_ZONES)) { 2187c793af95Ssangeeta ret = ire_walk_ill_match(rtf->rt_match_flags, 2188c793af95Ssangeeta rtf->rt_ire_type, ire, 2189f4b3ec61Sdh155122 rtf->rt_ill, rtf->rt_zoneid, rtf->rt_ipst); 2190c793af95Ssangeeta } else 2191c793af95Ssangeeta ret = B_TRUE; 2192c793af95Ssangeeta if (ret) 2193c793af95Ssangeeta (*rtf->rt_func)(ire, rtf->rt_arg); 2194c793af95Ssangeeta } 2195c793af95Ssangeeta return (0); 2196c793af95Ssangeeta } 2197c793af95Ssangeeta 21987c478bd9Sstevel@tonic-gate /* 21997c478bd9Sstevel@tonic-gate * Walk the ftable and the ctable entries that match the ill. 22007c478bd9Sstevel@tonic-gate */ 2201c793af95Ssangeeta void 22027c478bd9Sstevel@tonic-gate ire_walk_ill_tables(uint_t match_flags, uint_t ire_type, pfv_t func, 220345916cd2Sjpk void *arg, size_t ftbl_sz, size_t htbl_sz, irb_t **ipftbl, 2204f4b3ec61Sdh155122 size_t ctbl_sz, irb_t *ipctbl, ill_t *ill, zoneid_t zoneid, 2205f4b3ec61Sdh155122 ip_stack_t *ipst) 22067c478bd9Sstevel@tonic-gate { 22077c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 22087c478bd9Sstevel@tonic-gate irb_t *irb; 22097c478bd9Sstevel@tonic-gate ire_t *ire; 22107c478bd9Sstevel@tonic-gate int i, j; 22117c478bd9Sstevel@tonic-gate boolean_t ret; 2212c793af95Ssangeeta struct rtfuncarg rtfarg; 22137c478bd9Sstevel@tonic-gate 221498e93c29Smeem ASSERT((!(match_flags & (MATCH_IRE_ILL | 22157c478bd9Sstevel@tonic-gate MATCH_IRE_ILL_GROUP))) || (ill != NULL)); 22167c478bd9Sstevel@tonic-gate ASSERT(!(match_flags & MATCH_IRE_TYPE) || (ire_type != 0)); 22177c478bd9Sstevel@tonic-gate /* 22187c478bd9Sstevel@tonic-gate * Optimize by not looking at the forwarding table if there 22197c478bd9Sstevel@tonic-gate * is a MATCH_IRE_TYPE specified with no IRE_FORWARDTABLE 22207c478bd9Sstevel@tonic-gate * specified in ire_type. 22217c478bd9Sstevel@tonic-gate */ 22227c478bd9Sstevel@tonic-gate if (!(match_flags & MATCH_IRE_TYPE) || 22237c478bd9Sstevel@tonic-gate ((ire_type & IRE_FORWARDTABLE) != 0)) { 2224c793af95Ssangeeta /* knobs such that routine is called only for v6 case */ 2225f4b3ec61Sdh155122 if (ipftbl == ipst->ips_ip_forwarding_table_v6) { 22267c478bd9Sstevel@tonic-gate for (i = (ftbl_sz - 1); i >= 0; i--) { 22277c478bd9Sstevel@tonic-gate if ((irb_ptr = ipftbl[i]) == NULL) 22287c478bd9Sstevel@tonic-gate continue; 22297c478bd9Sstevel@tonic-gate for (j = 0; j < htbl_sz; j++) { 22307c478bd9Sstevel@tonic-gate irb = &irb_ptr[j]; 22317c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 22327c478bd9Sstevel@tonic-gate continue; 2233c793af95Ssangeeta 22347c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 22357c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; 22367c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 22377c478bd9Sstevel@tonic-gate if (match_flags == 0 && 22387c478bd9Sstevel@tonic-gate zoneid == ALL_ZONES) { 22397c478bd9Sstevel@tonic-gate ret = B_TRUE; 22407c478bd9Sstevel@tonic-gate } else { 2241c793af95Ssangeeta ret = 2242c793af95Ssangeeta ire_walk_ill_match( 2243c793af95Ssangeeta match_flags, 2244c793af95Ssangeeta ire_type, ire, ill, 2245f4b3ec61Sdh155122 zoneid, ipst); 22467c478bd9Sstevel@tonic-gate } 22477c478bd9Sstevel@tonic-gate if (ret) 22487c478bd9Sstevel@tonic-gate (*func)(ire, arg); 22497c478bd9Sstevel@tonic-gate } 22507c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 22517c478bd9Sstevel@tonic-gate } 22527c478bd9Sstevel@tonic-gate } 2253c793af95Ssangeeta } else { 2254c793af95Ssangeeta (void) memset(&rtfarg, 0, sizeof (rtfarg)); 2255c793af95Ssangeeta rtfarg.rt_func = func; 2256c793af95Ssangeeta rtfarg.rt_arg = arg; 2257c793af95Ssangeeta if (match_flags != 0) { 2258c793af95Ssangeeta rtfarg.rt_match_flags = match_flags; 2259c793af95Ssangeeta } 2260c793af95Ssangeeta rtfarg.rt_ire_type = ire_type; 2261c793af95Ssangeeta rtfarg.rt_ill = ill; 2262c793af95Ssangeeta rtfarg.rt_zoneid = zoneid; 2263f4b3ec61Sdh155122 rtfarg.rt_ipst = ipst; /* No netstack_hold */ 2264f4b3ec61Sdh155122 (void) ipst->ips_ip_ftable->rnh_walktree_mt( 2265f4b3ec61Sdh155122 ipst->ips_ip_ftable, 2266f4b3ec61Sdh155122 rtfunc, &rtfarg, irb_refhold_rn, irb_refrele_rn); 2267c793af95Ssangeeta } 22687c478bd9Sstevel@tonic-gate } 22697c478bd9Sstevel@tonic-gate 22707c478bd9Sstevel@tonic-gate /* 22717c478bd9Sstevel@tonic-gate * Optimize by not looking at the cache table if there 22727c478bd9Sstevel@tonic-gate * is a MATCH_IRE_TYPE specified with no IRE_CACHETABLE 22737c478bd9Sstevel@tonic-gate * specified in ire_type. 22747c478bd9Sstevel@tonic-gate */ 22757c478bd9Sstevel@tonic-gate if (!(match_flags & MATCH_IRE_TYPE) || 22767c478bd9Sstevel@tonic-gate ((ire_type & IRE_CACHETABLE) != 0)) { 22777c478bd9Sstevel@tonic-gate for (i = 0; i < ctbl_sz; i++) { 22787c478bd9Sstevel@tonic-gate irb = &ipctbl[i]; 22797c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 22807c478bd9Sstevel@tonic-gate continue; 22817c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 22827c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; 22837c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 22847c478bd9Sstevel@tonic-gate if (match_flags == 0 && zoneid == ALL_ZONES) { 22857c478bd9Sstevel@tonic-gate ret = B_TRUE; 22867c478bd9Sstevel@tonic-gate } else { 22877c478bd9Sstevel@tonic-gate ret = ire_walk_ill_match( 22887c478bd9Sstevel@tonic-gate match_flags, ire_type, 2289f4b3ec61Sdh155122 ire, ill, zoneid, ipst); 22907c478bd9Sstevel@tonic-gate } 22917c478bd9Sstevel@tonic-gate if (ret) 22927c478bd9Sstevel@tonic-gate (*func)(ire, arg); 22937c478bd9Sstevel@tonic-gate } 22947c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 22957c478bd9Sstevel@tonic-gate } 22967c478bd9Sstevel@tonic-gate } 22977c478bd9Sstevel@tonic-gate } 22987c478bd9Sstevel@tonic-gate 22997c478bd9Sstevel@tonic-gate /* 23007c478bd9Sstevel@tonic-gate * This function takes a mask and returns 23017c478bd9Sstevel@tonic-gate * number of bits set in the mask. If no 23027c478bd9Sstevel@tonic-gate * bit is set it returns 0. 23037c478bd9Sstevel@tonic-gate * Assumes a contiguous mask. 23047c478bd9Sstevel@tonic-gate */ 23057c478bd9Sstevel@tonic-gate int 23067c478bd9Sstevel@tonic-gate ip_mask_to_plen(ipaddr_t mask) 23077c478bd9Sstevel@tonic-gate { 23087c478bd9Sstevel@tonic-gate return (mask == 0 ? 0 : IP_ABITS - (ffs(ntohl(mask)) -1)); 23097c478bd9Sstevel@tonic-gate } 23107c478bd9Sstevel@tonic-gate 23117c478bd9Sstevel@tonic-gate /* 23127c478bd9Sstevel@tonic-gate * Convert length for a mask to the mask. 23137c478bd9Sstevel@tonic-gate */ 23147c478bd9Sstevel@tonic-gate ipaddr_t 23157c478bd9Sstevel@tonic-gate ip_plen_to_mask(uint_t masklen) 23167c478bd9Sstevel@tonic-gate { 23177c478bd9Sstevel@tonic-gate return (htonl(IP_HOST_MASK << (IP_ABITS - masklen))); 23187c478bd9Sstevel@tonic-gate } 23197c478bd9Sstevel@tonic-gate 23207c478bd9Sstevel@tonic-gate void 23217c478bd9Sstevel@tonic-gate ire_atomic_end(irb_t *irb_ptr, ire_t *ire) 23227c478bd9Sstevel@tonic-gate { 23237c478bd9Sstevel@tonic-gate ill_t *ill_list[NUM_ILLS]; 2324f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 23257c478bd9Sstevel@tonic-gate 23267c478bd9Sstevel@tonic-gate ill_list[0] = ire->ire_stq != NULL ? ire->ire_stq->q_ptr : NULL; 23277c478bd9Sstevel@tonic-gate ill_list[1] = ire->ire_ipif != NULL ? ire->ire_ipif->ipif_ill : NULL; 23287c478bd9Sstevel@tonic-gate ill_unlock_ills(ill_list, NUM_ILLS); 23297c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 2330f4b3ec61Sdh155122 rw_exit(&ipst->ips_ill_g_usesrc_lock); 23317c478bd9Sstevel@tonic-gate } 23327c478bd9Sstevel@tonic-gate 23337c478bd9Sstevel@tonic-gate /* 23347c478bd9Sstevel@tonic-gate * ire_add_v[46] atomically make sure that the ipif or ill associated 23357c478bd9Sstevel@tonic-gate * with the new ire being added is stable and not IPIF_CHANGING or ILL_CHANGING 23367c478bd9Sstevel@tonic-gate * before adding the ire to the table. This ensures that we don't create 23377c478bd9Sstevel@tonic-gate * new IRE_CACHEs with stale values for parameters that are passed to 23387c478bd9Sstevel@tonic-gate * ire_create such as ire_max_frag. Note that ire_create() is passed a pointer 23397c478bd9Sstevel@tonic-gate * to the ipif_mtu, and not the value. The actual value is derived from the 23407c478bd9Sstevel@tonic-gate * parent ire or ipif under the bucket lock. 23417c478bd9Sstevel@tonic-gate */ 23427c478bd9Sstevel@tonic-gate int 23437c478bd9Sstevel@tonic-gate ire_atomic_start(irb_t *irb_ptr, ire_t *ire, queue_t *q, mblk_t *mp, 23447c478bd9Sstevel@tonic-gate ipsq_func_t func) 23457c478bd9Sstevel@tonic-gate { 23467c478bd9Sstevel@tonic-gate ill_t *stq_ill; 23477c478bd9Sstevel@tonic-gate ill_t *ipif_ill; 23487c478bd9Sstevel@tonic-gate ill_t *ill_list[NUM_ILLS]; 23497c478bd9Sstevel@tonic-gate int cnt = NUM_ILLS; 23507c478bd9Sstevel@tonic-gate int error = 0; 23517c478bd9Sstevel@tonic-gate ill_t *ill = NULL; 2352f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 23537c478bd9Sstevel@tonic-gate 23547c478bd9Sstevel@tonic-gate ill_list[0] = stq_ill = ire->ire_stq != 23557c478bd9Sstevel@tonic-gate NULL ? ire->ire_stq->q_ptr : NULL; 23567c478bd9Sstevel@tonic-gate ill_list[1] = ipif_ill = ire->ire_ipif != 23577c478bd9Sstevel@tonic-gate NULL ? ire->ire_ipif->ipif_ill : NULL; 23587c478bd9Sstevel@tonic-gate 23597c478bd9Sstevel@tonic-gate ASSERT((q != NULL && mp != NULL && func != NULL) || 23607c478bd9Sstevel@tonic-gate (q == NULL && mp == NULL && func == NULL)); 2361f4b3ec61Sdh155122 rw_enter(&ipst->ips_ill_g_usesrc_lock, RW_READER); 23627c478bd9Sstevel@tonic-gate GRAB_CONN_LOCK(q); 23637c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_WRITER); 23647c478bd9Sstevel@tonic-gate ill_lock_ills(ill_list, cnt); 23657c478bd9Sstevel@tonic-gate 23667c478bd9Sstevel@tonic-gate /* 23677c478bd9Sstevel@tonic-gate * While the IRE is in the process of being added, a user may have 23687c478bd9Sstevel@tonic-gate * invoked the ifconfig usesrc option on the stq_ill to make it a 23697c478bd9Sstevel@tonic-gate * usesrc client ILL. Check for this possibility here, if it is true 23707c478bd9Sstevel@tonic-gate * then we fail adding the IRE_CACHE. Another check is to make sure 23717c478bd9Sstevel@tonic-gate * that an ipif_ill of an IRE_CACHE being added is not part of a usesrc 23727c478bd9Sstevel@tonic-gate * group. The ill_g_usesrc_lock is released in ire_atomic_end 23737c478bd9Sstevel@tonic-gate */ 23747c478bd9Sstevel@tonic-gate if ((ire->ire_type & IRE_CACHE) && 23757c478bd9Sstevel@tonic-gate (ire->ire_marks & IRE_MARK_USESRC_CHECK)) { 23767c478bd9Sstevel@tonic-gate if (stq_ill->ill_usesrc_ifindex != 0) { 23777c478bd9Sstevel@tonic-gate ASSERT(stq_ill->ill_usesrc_grp_next != NULL); 23787c478bd9Sstevel@tonic-gate if ((ipif_ill->ill_phyint->phyint_ifindex != 23797c478bd9Sstevel@tonic-gate stq_ill->ill_usesrc_ifindex) || 23807c478bd9Sstevel@tonic-gate (ipif_ill->ill_usesrc_grp_next == NULL) || 23817c478bd9Sstevel@tonic-gate (ipif_ill->ill_usesrc_ifindex != 0)) { 23827c478bd9Sstevel@tonic-gate error = EINVAL; 23837c478bd9Sstevel@tonic-gate goto done; 23847c478bd9Sstevel@tonic-gate } 23857c478bd9Sstevel@tonic-gate } else if (ipif_ill->ill_usesrc_grp_next != NULL) { 23867c478bd9Sstevel@tonic-gate error = EINVAL; 23877c478bd9Sstevel@tonic-gate goto done; 23887c478bd9Sstevel@tonic-gate } 23897c478bd9Sstevel@tonic-gate } 23907c478bd9Sstevel@tonic-gate 23917c478bd9Sstevel@tonic-gate /* 23927c478bd9Sstevel@tonic-gate * IPMP flag settings happen without taking the exclusive route 23937c478bd9Sstevel@tonic-gate * in ip_sioctl_flags. So we need to make an atomic check here 23947c478bd9Sstevel@tonic-gate * for FAILED/OFFLINE/INACTIVE flags or if it has hit the 23957c478bd9Sstevel@tonic-gate * FAILBACK=no case. 23967c478bd9Sstevel@tonic-gate */ 23977c478bd9Sstevel@tonic-gate if ((stq_ill != NULL) && !IAM_WRITER_ILL(stq_ill)) { 23987c478bd9Sstevel@tonic-gate if (stq_ill->ill_state_flags & ILL_CHANGING) { 23997c478bd9Sstevel@tonic-gate ill = stq_ill; 24007c478bd9Sstevel@tonic-gate error = EAGAIN; 24017c478bd9Sstevel@tonic-gate } else if ((stq_ill->ill_phyint->phyint_flags & PHYI_OFFLINE) || 24027c478bd9Sstevel@tonic-gate (ill_is_probeonly(stq_ill) && 24037c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_HIDDEN))) { 24047c478bd9Sstevel@tonic-gate error = EINVAL; 24057c478bd9Sstevel@tonic-gate } 24067c478bd9Sstevel@tonic-gate goto done; 24077c478bd9Sstevel@tonic-gate } 24087c478bd9Sstevel@tonic-gate 24097c478bd9Sstevel@tonic-gate /* 24107c478bd9Sstevel@tonic-gate * We don't check for OFFLINE/FAILED in this case because 24117c478bd9Sstevel@tonic-gate * the source address selection logic (ipif_select_source) 24127c478bd9Sstevel@tonic-gate * may still select a source address from such an ill. The 24137c478bd9Sstevel@tonic-gate * assumption is that these addresses will be moved by in.mpathd 24147c478bd9Sstevel@tonic-gate * soon. (i.e. this is a race). However link local addresses 24157c478bd9Sstevel@tonic-gate * will not move and hence ipif_select_source_v6 tries to avoid 24167c478bd9Sstevel@tonic-gate * FAILED ills. Please see ipif_select_source_v6 for more info 24177c478bd9Sstevel@tonic-gate */ 24187c478bd9Sstevel@tonic-gate if ((ipif_ill != NULL) && !IAM_WRITER_ILL(ipif_ill) && 24197c478bd9Sstevel@tonic-gate (ipif_ill->ill_state_flags & ILL_CHANGING)) { 24207c478bd9Sstevel@tonic-gate ill = ipif_ill; 24217c478bd9Sstevel@tonic-gate error = EAGAIN; 24227c478bd9Sstevel@tonic-gate goto done; 24237c478bd9Sstevel@tonic-gate } 24247c478bd9Sstevel@tonic-gate 24257c478bd9Sstevel@tonic-gate if ((ire->ire_ipif != NULL) && !IAM_WRITER_IPIF(ire->ire_ipif) && 24267c478bd9Sstevel@tonic-gate (ire->ire_ipif->ipif_state_flags & IPIF_CHANGING)) { 24277c478bd9Sstevel@tonic-gate ill = ire->ire_ipif->ipif_ill; 24287c478bd9Sstevel@tonic-gate ASSERT(ill != NULL); 24297c478bd9Sstevel@tonic-gate error = EAGAIN; 24307c478bd9Sstevel@tonic-gate goto done; 24317c478bd9Sstevel@tonic-gate } 24327c478bd9Sstevel@tonic-gate 24337c478bd9Sstevel@tonic-gate done: 24347c478bd9Sstevel@tonic-gate if (error == EAGAIN && ILL_CAN_WAIT(ill, q)) { 24357c478bd9Sstevel@tonic-gate ipsq_t *ipsq = ill->ill_phyint->phyint_ipsq; 24367c478bd9Sstevel@tonic-gate mutex_enter(&ipsq->ipsq_lock); 24377c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 24387c478bd9Sstevel@tonic-gate ipsq_enq(ipsq, q, mp, func, NEW_OP, ill); 24397c478bd9Sstevel@tonic-gate mutex_exit(&ipsq->ipsq_lock); 24407c478bd9Sstevel@tonic-gate error = EINPROGRESS; 24417c478bd9Sstevel@tonic-gate } else if (error != 0) { 24427c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 24437c478bd9Sstevel@tonic-gate } 24447c478bd9Sstevel@tonic-gate 24457c478bd9Sstevel@tonic-gate RELEASE_CONN_LOCK(q); 24467c478bd9Sstevel@tonic-gate return (error); 24477c478bd9Sstevel@tonic-gate } 24487c478bd9Sstevel@tonic-gate 24497c478bd9Sstevel@tonic-gate /* 24507c478bd9Sstevel@tonic-gate * Add a fully initialized IRE to an appropriate table based on 24517c478bd9Sstevel@tonic-gate * ire_type. 2452c793af95Ssangeeta * 2453c793af95Ssangeeta * allow_unresolved == B_FALSE indicates a legacy code-path call 2454c793af95Ssangeeta * that has prohibited the addition of incomplete ire's. If this 2455c793af95Ssangeeta * parameter is set, and we find an nce that is in a state other 2456c793af95Ssangeeta * than ND_REACHABLE, we fail the add. Note that nce_state could be 2457b9c344b3Ssowmini * something other than ND_REACHABLE if the nce had just expired and 2458b9c344b3Ssowmini * the ire_create preceding the ire_add added a new ND_INITIAL nce. 24597c478bd9Sstevel@tonic-gate */ 24607c478bd9Sstevel@tonic-gate int 2461c793af95Ssangeeta ire_add(ire_t **irep, queue_t *q, mblk_t *mp, ipsq_func_t func, 2462c793af95Ssangeeta boolean_t allow_unresolved) 24637c478bd9Sstevel@tonic-gate { 24647c478bd9Sstevel@tonic-gate ire_t *ire1; 24657c478bd9Sstevel@tonic-gate ill_t *stq_ill = NULL; 24667c478bd9Sstevel@tonic-gate ill_t *ill; 24677c478bd9Sstevel@tonic-gate ipif_t *ipif = NULL; 24687c478bd9Sstevel@tonic-gate ill_walk_context_t ctx; 24697c478bd9Sstevel@tonic-gate ire_t *ire = *irep; 24707c478bd9Sstevel@tonic-gate int error; 2471dc3879f9Sjarrett boolean_t ire_is_mblk = B_FALSE; 2472dc3879f9Sjarrett tsol_gcgrp_t *gcgrp = NULL; 2473dc3879f9Sjarrett tsol_gcgrp_addr_t ga; 2474f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 24757c478bd9Sstevel@tonic-gate 24767c478bd9Sstevel@tonic-gate /* get ready for the day when original ire is not created as mblk */ 24777c478bd9Sstevel@tonic-gate if (ire->ire_mp != NULL) { 2478dc3879f9Sjarrett ire_is_mblk = B_TRUE; 24797c478bd9Sstevel@tonic-gate /* Copy the ire to a kmem_alloc'ed area */ 24807c478bd9Sstevel@tonic-gate ire1 = kmem_cache_alloc(ire_cache, KM_NOSLEEP); 24817c478bd9Sstevel@tonic-gate if (ire1 == NULL) { 24827c478bd9Sstevel@tonic-gate ip1dbg(("ire_add: alloc failed\n")); 24837c478bd9Sstevel@tonic-gate ire_delete(ire); 24847c478bd9Sstevel@tonic-gate *irep = NULL; 24857c478bd9Sstevel@tonic-gate return (ENOMEM); 24867c478bd9Sstevel@tonic-gate } 2487c793af95Ssangeeta ire->ire_marks &= ~IRE_MARK_UNCACHED; 24887c478bd9Sstevel@tonic-gate *ire1 = *ire; 24897c478bd9Sstevel@tonic-gate ire1->ire_mp = NULL; 2490c793af95Ssangeeta ire1->ire_stq_ifindex = 0; 24917c478bd9Sstevel@tonic-gate freeb(ire->ire_mp); 24927c478bd9Sstevel@tonic-gate ire = ire1; 24937c478bd9Sstevel@tonic-gate } 24947c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 24957c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 24967c478bd9Sstevel@tonic-gate 24977c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE) { 24987c478bd9Sstevel@tonic-gate /* 24997c478bd9Sstevel@tonic-gate * If this interface is FAILED, or INACTIVE or has hit 25007c478bd9Sstevel@tonic-gate * the FAILBACK=no case, we create IRE_CACHES marked 25017c478bd9Sstevel@tonic-gate * HIDDEN for some special cases e.g. bind to 25027c478bd9Sstevel@tonic-gate * IPIF_NOFAILOVER address etc. So, if this interface 25037c478bd9Sstevel@tonic-gate * is FAILED/INACTIVE/hit FAILBACK=no case, and we are 25047c478bd9Sstevel@tonic-gate * not creating hidden ires, we should not allow that. 25057c478bd9Sstevel@tonic-gate * This happens because the state of the interface 25067c478bd9Sstevel@tonic-gate * changed while we were waiting in ARP. If this is the 25077c478bd9Sstevel@tonic-gate * daemon sending probes, the next probe will create 25087c478bd9Sstevel@tonic-gate * HIDDEN ires and we will create an ire then. This 25097c478bd9Sstevel@tonic-gate * cannot happen with NDP currently because IRE is 25107c478bd9Sstevel@tonic-gate * never queued in NDP. But it can happen in the 25117c478bd9Sstevel@tonic-gate * future when we have external resolvers with IPv6. 25127c478bd9Sstevel@tonic-gate * If the interface gets marked with OFFLINE while we 25137c478bd9Sstevel@tonic-gate * are waiting in ARP, don't add the ire. 25147c478bd9Sstevel@tonic-gate */ 25157c478bd9Sstevel@tonic-gate if ((stq_ill->ill_phyint->phyint_flags & PHYI_OFFLINE) || 25167c478bd9Sstevel@tonic-gate (ill_is_probeonly(stq_ill) && 25177c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_HIDDEN))) { 25187c478bd9Sstevel@tonic-gate /* 25197c478bd9Sstevel@tonic-gate * We don't know whether it is a valid ipif or not. 25207c478bd9Sstevel@tonic-gate * unless we do the check below. So, set it to NULL. 25217c478bd9Sstevel@tonic-gate */ 25227c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 25237c478bd9Sstevel@tonic-gate ire_delete(ire); 25247c478bd9Sstevel@tonic-gate *irep = NULL; 25257c478bd9Sstevel@tonic-gate return (EINVAL); 25267c478bd9Sstevel@tonic-gate } 25277c478bd9Sstevel@tonic-gate } 25287c478bd9Sstevel@tonic-gate 25297c478bd9Sstevel@tonic-gate if (stq_ill != NULL && ire->ire_type == IRE_CACHE && 25307c478bd9Sstevel@tonic-gate stq_ill->ill_net_type == IRE_IF_RESOLVER) { 2531f4b3ec61Sdh155122 rw_enter(&ipst->ips_ill_g_lock, RW_READER); 2532f4b3ec61Sdh155122 ill = ILL_START_WALK_ALL(&ctx, ipst); 25337c478bd9Sstevel@tonic-gate for (; ill != NULL; ill = ill_next(&ctx, ill)) { 25347c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 25357c478bd9Sstevel@tonic-gate if (ill->ill_state_flags & ILL_CONDEMNED) { 25367c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 25377c478bd9Sstevel@tonic-gate continue; 25387c478bd9Sstevel@tonic-gate } 25397c478bd9Sstevel@tonic-gate /* 25407c478bd9Sstevel@tonic-gate * We need to make sure that the ipif is a valid one 25417c478bd9Sstevel@tonic-gate * before adding the IRE_CACHE. This happens only 25427c478bd9Sstevel@tonic-gate * with IRE_CACHE when there is an external resolver. 25437c478bd9Sstevel@tonic-gate * 25447c478bd9Sstevel@tonic-gate * We can unplumb a logical interface while the 25457c478bd9Sstevel@tonic-gate * packet is waiting in ARP with the IRE. Then, 25467c478bd9Sstevel@tonic-gate * later on when we feed the IRE back, the ipif 25477c478bd9Sstevel@tonic-gate * has to be re-checked. This can't happen with 25487c478bd9Sstevel@tonic-gate * NDP currently, as we never queue the IRE with 25497c478bd9Sstevel@tonic-gate * the packet. We always try to recreate the IRE 25507c478bd9Sstevel@tonic-gate * when the resolution is completed. But, we do 25517c478bd9Sstevel@tonic-gate * it for IPv6 also here so that in future if 25527c478bd9Sstevel@tonic-gate * we have external resolvers, it will work without 25537c478bd9Sstevel@tonic-gate * any change. 25547c478bd9Sstevel@tonic-gate */ 25557c478bd9Sstevel@tonic-gate ipif = ipif_lookup_seqid(ill, ire->ire_ipif_seqid); 25567c478bd9Sstevel@tonic-gate if (ipif != NULL) { 25577c478bd9Sstevel@tonic-gate ipif_refhold_locked(ipif); 25587c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 25597c478bd9Sstevel@tonic-gate break; 25607c478bd9Sstevel@tonic-gate } 25617c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 25627c478bd9Sstevel@tonic-gate } 2563f4b3ec61Sdh155122 rw_exit(&ipst->ips_ill_g_lock); 25647c478bd9Sstevel@tonic-gate if (ipif == NULL || 25657c478bd9Sstevel@tonic-gate (ipif->ipif_isv6 && 25667c478bd9Sstevel@tonic-gate !IN6_ARE_ADDR_EQUAL(&ire->ire_src_addr_v6, 25677c478bd9Sstevel@tonic-gate &ipif->ipif_v6src_addr)) || 25687c478bd9Sstevel@tonic-gate (!ipif->ipif_isv6 && 25697c478bd9Sstevel@tonic-gate ire->ire_src_addr != ipif->ipif_src_addr) || 257045916cd2Sjpk ire->ire_zoneid != ipif->ipif_zoneid) { 25717c478bd9Sstevel@tonic-gate 25727c478bd9Sstevel@tonic-gate if (ipif != NULL) 25737c478bd9Sstevel@tonic-gate ipif_refrele(ipif); 25747c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 25757c478bd9Sstevel@tonic-gate ire_delete(ire); 25767c478bd9Sstevel@tonic-gate *irep = NULL; 25777c478bd9Sstevel@tonic-gate return (EINVAL); 25787c478bd9Sstevel@tonic-gate } 25797c478bd9Sstevel@tonic-gate 25807c478bd9Sstevel@tonic-gate 25817c478bd9Sstevel@tonic-gate ASSERT(ill != NULL); 25827c478bd9Sstevel@tonic-gate /* 25837c478bd9Sstevel@tonic-gate * If this group was dismantled while this packets was 25847c478bd9Sstevel@tonic-gate * queued in ARP, don't add it here. 25857c478bd9Sstevel@tonic-gate */ 25867c478bd9Sstevel@tonic-gate if (ire->ire_ipif->ipif_ill->ill_group != ill->ill_group) { 25877c478bd9Sstevel@tonic-gate /* We don't want ire_inactive bump stats for this */ 25887c478bd9Sstevel@tonic-gate ipif_refrele(ipif); 25897c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 25907c478bd9Sstevel@tonic-gate ire_delete(ire); 25917c478bd9Sstevel@tonic-gate *irep = NULL; 25927c478bd9Sstevel@tonic-gate return (EINVAL); 25937c478bd9Sstevel@tonic-gate } 2594dc3879f9Sjarrett 2595dc3879f9Sjarrett /* 2596dc3879f9Sjarrett * Since we didn't attach label security attributes to the 2597dc3879f9Sjarrett * ire for the resolver case, we need to add it now. (only 2598dc3879f9Sjarrett * for v4 resolver and v6 xresolv case). 2599dc3879f9Sjarrett */ 2600dc3879f9Sjarrett if (is_system_labeled() && ire_is_mblk) { 2601dc3879f9Sjarrett if (ire->ire_ipversion == IPV4_VERSION) { 2602dc3879f9Sjarrett ga.ga_af = AF_INET; 2603dc3879f9Sjarrett IN6_IPADDR_TO_V4MAPPED(ire->ire_gateway_addr != 2604dc3879f9Sjarrett INADDR_ANY ? ire->ire_gateway_addr : 2605dc3879f9Sjarrett ire->ire_addr, &ga.ga_addr); 2606dc3879f9Sjarrett } else { 2607dc3879f9Sjarrett ga.ga_af = AF_INET6; 2608dc3879f9Sjarrett ga.ga_addr = IN6_IS_ADDR_UNSPECIFIED( 2609dc3879f9Sjarrett &ire->ire_gateway_addr_v6) ? 2610dc3879f9Sjarrett ire->ire_addr_v6 : 2611dc3879f9Sjarrett ire->ire_gateway_addr_v6; 2612dc3879f9Sjarrett } 2613dc3879f9Sjarrett gcgrp = gcgrp_lookup(&ga, B_FALSE); 2614dc3879f9Sjarrett error = tsol_ire_init_gwattr(ire, ire->ire_ipversion, 2615dc3879f9Sjarrett NULL, gcgrp); 2616dc3879f9Sjarrett if (error != 0) { 2617dc3879f9Sjarrett if (gcgrp != NULL) { 2618dc3879f9Sjarrett GCGRP_REFRELE(gcgrp); 2619dc3879f9Sjarrett gcgrp = NULL; 2620dc3879f9Sjarrett } 2621dc3879f9Sjarrett ipif_refrele(ipif); 2622dc3879f9Sjarrett ire->ire_ipif = NULL; 2623dc3879f9Sjarrett ire_delete(ire); 2624dc3879f9Sjarrett *irep = NULL; 2625dc3879f9Sjarrett return (error); 2626dc3879f9Sjarrett } 2627dc3879f9Sjarrett } 26287c478bd9Sstevel@tonic-gate } 26297c478bd9Sstevel@tonic-gate 26307c478bd9Sstevel@tonic-gate /* 26317c478bd9Sstevel@tonic-gate * In case ire was changed 26327c478bd9Sstevel@tonic-gate */ 26337c478bd9Sstevel@tonic-gate *irep = ire; 26345c0b7edeSseb if (ire->ire_ipversion == IPV6_VERSION) 26357c478bd9Sstevel@tonic-gate error = ire_add_v6(irep, q, mp, func); 26367c478bd9Sstevel@tonic-gate else 26375c0b7edeSseb error = ire_add_v4(irep, q, mp, func, allow_unresolved); 26387c478bd9Sstevel@tonic-gate if (ipif != NULL) 26397c478bd9Sstevel@tonic-gate ipif_refrele(ipif); 26407c478bd9Sstevel@tonic-gate return (error); 26417c478bd9Sstevel@tonic-gate } 26427c478bd9Sstevel@tonic-gate 26437c478bd9Sstevel@tonic-gate /* 2644dc3879f9Sjarrett * Add an initialized IRE to an appropriate table based on ire_type. 26457c478bd9Sstevel@tonic-gate * 26466bdb8e66Sdd193516 * The forward table contains IRE_PREFIX/IRE_HOST and 26477c478bd9Sstevel@tonic-gate * IRE_IF_RESOLVER/IRE_IF_NORESOLVER and IRE_DEFAULT. 26487c478bd9Sstevel@tonic-gate * 26497c478bd9Sstevel@tonic-gate * The cache table contains IRE_BROADCAST/IRE_LOCAL/IRE_LOOPBACK 26507c478bd9Sstevel@tonic-gate * and IRE_CACHE. 26517c478bd9Sstevel@tonic-gate * 26527c478bd9Sstevel@tonic-gate * NOTE : This function is called as writer though not required 26537c478bd9Sstevel@tonic-gate * by this function. 26547c478bd9Sstevel@tonic-gate */ 26557c478bd9Sstevel@tonic-gate static int 2656c793af95Ssangeeta ire_add_v4(ire_t **ire_p, queue_t *q, mblk_t *mp, ipsq_func_t func, 2657c793af95Ssangeeta boolean_t allow_unresolved) 26587c478bd9Sstevel@tonic-gate { 26597c478bd9Sstevel@tonic-gate ire_t *ire1; 26607c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 26617c478bd9Sstevel@tonic-gate ire_t **irep; 26627c478bd9Sstevel@tonic-gate int flags; 26637c478bd9Sstevel@tonic-gate ire_t *pire = NULL; 26647c478bd9Sstevel@tonic-gate ill_t *stq_ill; 26657c478bd9Sstevel@tonic-gate ire_t *ire = *ire_p; 26667c478bd9Sstevel@tonic-gate int error; 2667c793af95Ssangeeta boolean_t need_refrele = B_FALSE; 2668c793af95Ssangeeta nce_t *nce; 2669f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 26707c478bd9Sstevel@tonic-gate 26717c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) 26727c478bd9Sstevel@tonic-gate ASSERT(!MUTEX_HELD(&ire->ire_ipif->ipif_ill->ill_lock)); 26737c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 26747c478bd9Sstevel@tonic-gate ASSERT(!MUTEX_HELD( 26757c478bd9Sstevel@tonic-gate &((ill_t *)(ire->ire_stq->q_ptr))->ill_lock)); 26767c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 26777c478bd9Sstevel@tonic-gate ASSERT(ire->ire_mp == NULL); /* Calls should go through ire_add */ 26787c478bd9Sstevel@tonic-gate 26797c478bd9Sstevel@tonic-gate /* Find the appropriate list head. */ 26807c478bd9Sstevel@tonic-gate switch (ire->ire_type) { 26817c478bd9Sstevel@tonic-gate case IRE_HOST: 26827c478bd9Sstevel@tonic-gate ire->ire_mask = IP_HOST_MASK; 26837c478bd9Sstevel@tonic-gate ire->ire_masklen = IP_ABITS; 26847c478bd9Sstevel@tonic-gate if ((ire->ire_flags & RTF_SETSRC) == 0) 26857c478bd9Sstevel@tonic-gate ire->ire_src_addr = 0; 26867c478bd9Sstevel@tonic-gate break; 26877c478bd9Sstevel@tonic-gate case IRE_CACHE: 26887c478bd9Sstevel@tonic-gate case IRE_BROADCAST: 26897c478bd9Sstevel@tonic-gate case IRE_LOCAL: 26907c478bd9Sstevel@tonic-gate case IRE_LOOPBACK: 26917c478bd9Sstevel@tonic-gate ire->ire_mask = IP_HOST_MASK; 26927c478bd9Sstevel@tonic-gate ire->ire_masklen = IP_ABITS; 26937c478bd9Sstevel@tonic-gate break; 26947c478bd9Sstevel@tonic-gate case IRE_PREFIX: 26957c478bd9Sstevel@tonic-gate if ((ire->ire_flags & RTF_SETSRC) == 0) 26967c478bd9Sstevel@tonic-gate ire->ire_src_addr = 0; 26977c478bd9Sstevel@tonic-gate break; 26987c478bd9Sstevel@tonic-gate case IRE_DEFAULT: 26997c478bd9Sstevel@tonic-gate if ((ire->ire_flags & RTF_SETSRC) == 0) 27007c478bd9Sstevel@tonic-gate ire->ire_src_addr = 0; 27017c478bd9Sstevel@tonic-gate break; 27027c478bd9Sstevel@tonic-gate case IRE_IF_RESOLVER: 27037c478bd9Sstevel@tonic-gate case IRE_IF_NORESOLVER: 27047c478bd9Sstevel@tonic-gate break; 27057c478bd9Sstevel@tonic-gate default: 2706c793af95Ssangeeta ip0dbg(("ire_add_v4: ire %p has unrecognized IRE type (%d)\n", 2707c793af95Ssangeeta (void *)ire, ire->ire_type)); 27087c478bd9Sstevel@tonic-gate ire_delete(ire); 27097c478bd9Sstevel@tonic-gate *ire_p = NULL; 27107c478bd9Sstevel@tonic-gate return (EINVAL); 27117c478bd9Sstevel@tonic-gate } 27127c478bd9Sstevel@tonic-gate 27137c478bd9Sstevel@tonic-gate /* Make sure the address is properly masked. */ 27147c478bd9Sstevel@tonic-gate ire->ire_addr &= ire->ire_mask; 27157c478bd9Sstevel@tonic-gate 27167c478bd9Sstevel@tonic-gate /* 27177c478bd9Sstevel@tonic-gate * ip_newroute/ip_newroute_multi are unable to prevent the deletion 27187c478bd9Sstevel@tonic-gate * of the interface route while adding an IRE_CACHE for an on-link 27197c478bd9Sstevel@tonic-gate * destination in the IRE_IF_RESOLVER case, since the ire has to 27207c478bd9Sstevel@tonic-gate * go to ARP and return. We can't do a REFHOLD on the 27217c478bd9Sstevel@tonic-gate * associated interface ire for fear of ARP freeing the message. 27227c478bd9Sstevel@tonic-gate * Here we look up the interface ire in the forwarding table and 27237c478bd9Sstevel@tonic-gate * make sure that the interface route has not been deleted. 27247c478bd9Sstevel@tonic-gate */ 27257c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE && ire->ire_gateway_addr == 0 && 27267c478bd9Sstevel@tonic-gate ((ill_t *)ire->ire_stq->q_ptr)->ill_net_type == IRE_IF_RESOLVER) { 2727c793af95Ssangeeta 27287c478bd9Sstevel@tonic-gate ASSERT(ire->ire_max_fragp == NULL); 27297c478bd9Sstevel@tonic-gate if (CLASSD(ire->ire_addr) && !(ire->ire_flags & RTF_SETSRC)) { 27307c478bd9Sstevel@tonic-gate /* 27317c478bd9Sstevel@tonic-gate * The ihandle that we used in ip_newroute_multi 27327c478bd9Sstevel@tonic-gate * comes from the interface route corresponding 27337c478bd9Sstevel@tonic-gate * to ire_ipif. Lookup here to see if it exists 27347c478bd9Sstevel@tonic-gate * still. 27357c478bd9Sstevel@tonic-gate * If the ire has a source address assigned using 27367c478bd9Sstevel@tonic-gate * RTF_SETSRC, ire_ipif is the logical interface holding 27377c478bd9Sstevel@tonic-gate * this source address, so we can't use it to check for 27387c478bd9Sstevel@tonic-gate * the existence of the interface route. Instead we rely 27397c478bd9Sstevel@tonic-gate * on the brute force ihandle search in 27407c478bd9Sstevel@tonic-gate * ire_ihandle_lookup_onlink() below. 27417c478bd9Sstevel@tonic-gate */ 27427c478bd9Sstevel@tonic-gate pire = ipif_to_ire(ire->ire_ipif); 27437c478bd9Sstevel@tonic-gate if (pire == NULL) { 27447c478bd9Sstevel@tonic-gate ire_delete(ire); 27457c478bd9Sstevel@tonic-gate *ire_p = NULL; 27467c478bd9Sstevel@tonic-gate return (EINVAL); 27477c478bd9Sstevel@tonic-gate } else if (pire->ire_ihandle != ire->ire_ihandle) { 27487c478bd9Sstevel@tonic-gate ire_refrele(pire); 27497c478bd9Sstevel@tonic-gate ire_delete(ire); 27507c478bd9Sstevel@tonic-gate *ire_p = NULL; 27517c478bd9Sstevel@tonic-gate return (EINVAL); 27527c478bd9Sstevel@tonic-gate } 27537c478bd9Sstevel@tonic-gate } else { 27547c478bd9Sstevel@tonic-gate pire = ire_ihandle_lookup_onlink(ire); 27557c478bd9Sstevel@tonic-gate if (pire == NULL) { 27567c478bd9Sstevel@tonic-gate ire_delete(ire); 27577c478bd9Sstevel@tonic-gate *ire_p = NULL; 27587c478bd9Sstevel@tonic-gate return (EINVAL); 27597c478bd9Sstevel@tonic-gate } 27607c478bd9Sstevel@tonic-gate } 27617c478bd9Sstevel@tonic-gate /* Prevent pire from getting deleted */ 27627c478bd9Sstevel@tonic-gate IRB_REFHOLD(pire->ire_bucket); 27637c478bd9Sstevel@tonic-gate /* Has it been removed already ? */ 27647c478bd9Sstevel@tonic-gate if (pire->ire_marks & IRE_MARK_CONDEMNED) { 27657c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 27667c478bd9Sstevel@tonic-gate ire_refrele(pire); 27677c478bd9Sstevel@tonic-gate ire_delete(ire); 27687c478bd9Sstevel@tonic-gate *ire_p = NULL; 27697c478bd9Sstevel@tonic-gate return (EINVAL); 27707c478bd9Sstevel@tonic-gate } 27717c478bd9Sstevel@tonic-gate } else { 27727c478bd9Sstevel@tonic-gate ASSERT(ire->ire_max_fragp != NULL); 27737c478bd9Sstevel@tonic-gate } 27747c478bd9Sstevel@tonic-gate flags = (MATCH_IRE_MASK | MATCH_IRE_TYPE | MATCH_IRE_GW); 27757c478bd9Sstevel@tonic-gate 27767c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 27777c478bd9Sstevel@tonic-gate /* 27787c478bd9Sstevel@tonic-gate * We use MATCH_IRE_IPIF while adding IRE_CACHES only 27797c478bd9Sstevel@tonic-gate * for historic reasons and to maintain symmetry with 27807c478bd9Sstevel@tonic-gate * IPv6 code path. Historically this was used by 27817c478bd9Sstevel@tonic-gate * multicast code to create multiple IRE_CACHES on 27827c478bd9Sstevel@tonic-gate * a single ill with different ipifs. This was used 27837c478bd9Sstevel@tonic-gate * so that multicast packets leaving the node had the 27847c478bd9Sstevel@tonic-gate * right source address. This is no longer needed as 27857c478bd9Sstevel@tonic-gate * ip_wput initializes the address correctly. 27867c478bd9Sstevel@tonic-gate */ 27877c478bd9Sstevel@tonic-gate flags |= MATCH_IRE_IPIF; 27887c478bd9Sstevel@tonic-gate /* 27897c478bd9Sstevel@tonic-gate * If we are creating hidden ires, make sure we search on 27907c478bd9Sstevel@tonic-gate * this ill (MATCH_IRE_ILL) and a hidden ire, 27917c478bd9Sstevel@tonic-gate * while we are searching for duplicates below. Otherwise we 27927c478bd9Sstevel@tonic-gate * could potentially find an IRE on some other interface 27937c478bd9Sstevel@tonic-gate * and it may not be a IRE marked with IRE_MARK_HIDDEN. We 27947c478bd9Sstevel@tonic-gate * shouldn't do this as this will lead to an infinite loop 27957c478bd9Sstevel@tonic-gate * (if we get to ip_wput again) eventually we need an hidden 27967c478bd9Sstevel@tonic-gate * ire for this packet to go out. MATCH_IRE_ILL is explicitly 27977c478bd9Sstevel@tonic-gate * done below. 27987c478bd9Sstevel@tonic-gate */ 27997c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE && 28007c478bd9Sstevel@tonic-gate (ire->ire_marks & IRE_MARK_HIDDEN)) 28017c478bd9Sstevel@tonic-gate flags |= (MATCH_IRE_MARK_HIDDEN); 28027c478bd9Sstevel@tonic-gate } 2803c793af95Ssangeeta if ((ire->ire_type & IRE_CACHETABLE) == 0) { 2804c793af95Ssangeeta irb_ptr = ire_get_bucket(ire); 2805c793af95Ssangeeta need_refrele = B_TRUE; 2806c793af95Ssangeeta if (irb_ptr == NULL) { 2807c793af95Ssangeeta /* 2808c793af95Ssangeeta * This assumes that the ire has not added 2809c793af95Ssangeeta * a reference to the ipif. 2810c793af95Ssangeeta */ 2811c793af95Ssangeeta ire->ire_ipif = NULL; 2812c793af95Ssangeeta ire_delete(ire); 2813c793af95Ssangeeta if (pire != NULL) { 2814c793af95Ssangeeta IRB_REFRELE(pire->ire_bucket); 2815c793af95Ssangeeta ire_refrele(pire); 2816c793af95Ssangeeta } 2817c793af95Ssangeeta *ire_p = NULL; 2818c793af95Ssangeeta return (EINVAL); 2819c793af95Ssangeeta } 2820c793af95Ssangeeta } else { 2821f4b3ec61Sdh155122 irb_ptr = &(ipst->ips_ip_cache_table[IRE_ADDR_HASH( 2822f4b3ec61Sdh155122 ire->ire_addr, ipst->ips_ip_cache_table_size)]); 2823c793af95Ssangeeta } 28247c478bd9Sstevel@tonic-gate 28257c478bd9Sstevel@tonic-gate /* 28267c478bd9Sstevel@tonic-gate * Start the atomic add of the ire. Grab the ill locks, 28277c478bd9Sstevel@tonic-gate * ill_g_usesrc_lock and the bucket lock. Check for condemned 28287c478bd9Sstevel@tonic-gate * 28297c478bd9Sstevel@tonic-gate * If ipif or ill is changing ire_atomic_start() may queue the 28307c478bd9Sstevel@tonic-gate * request and return EINPROGRESS. 2831f4b3ec61Sdh155122 * To avoid lock order problems, get the ndp4->ndp_g_lock. 28327c478bd9Sstevel@tonic-gate */ 2833f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ndp4->ndp_g_lock); 28347c478bd9Sstevel@tonic-gate error = ire_atomic_start(irb_ptr, ire, q, mp, func); 28357c478bd9Sstevel@tonic-gate if (error != 0) { 2836f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 28377c478bd9Sstevel@tonic-gate /* 28387c478bd9Sstevel@tonic-gate * We don't know whether it is a valid ipif or not. 28397c478bd9Sstevel@tonic-gate * So, set it to NULL. This assumes that the ire has not added 28407c478bd9Sstevel@tonic-gate * a reference to the ipif. 28417c478bd9Sstevel@tonic-gate */ 28427c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 28437c478bd9Sstevel@tonic-gate ire_delete(ire); 28447c478bd9Sstevel@tonic-gate if (pire != NULL) { 28457c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 28467c478bd9Sstevel@tonic-gate ire_refrele(pire); 28477c478bd9Sstevel@tonic-gate } 28487c478bd9Sstevel@tonic-gate *ire_p = NULL; 2849c793af95Ssangeeta if (need_refrele) 2850c793af95Ssangeeta IRB_REFRELE(irb_ptr); 28517c478bd9Sstevel@tonic-gate return (error); 28527c478bd9Sstevel@tonic-gate } 28537c478bd9Sstevel@tonic-gate /* 28547c478bd9Sstevel@tonic-gate * To avoid creating ires having stale values for the ire_max_frag 28557c478bd9Sstevel@tonic-gate * we get the latest value atomically here. For more details 28567c478bd9Sstevel@tonic-gate * see the block comment in ip_sioctl_mtu and in DL_NOTE_SDU_CHANGE 28577c478bd9Sstevel@tonic-gate * in ip_rput_dlpi_writer 28587c478bd9Sstevel@tonic-gate */ 28597c478bd9Sstevel@tonic-gate if (ire->ire_max_fragp == NULL) { 28607c478bd9Sstevel@tonic-gate if (CLASSD(ire->ire_addr)) 28617c478bd9Sstevel@tonic-gate ire->ire_max_frag = ire->ire_ipif->ipif_mtu; 28627c478bd9Sstevel@tonic-gate else 28637c478bd9Sstevel@tonic-gate ire->ire_max_frag = pire->ire_max_frag; 28647c478bd9Sstevel@tonic-gate } else { 28657c478bd9Sstevel@tonic-gate uint_t max_frag; 28667c478bd9Sstevel@tonic-gate 28677c478bd9Sstevel@tonic-gate max_frag = *ire->ire_max_fragp; 28687c478bd9Sstevel@tonic-gate ire->ire_max_fragp = NULL; 28697c478bd9Sstevel@tonic-gate ire->ire_max_frag = max_frag; 28707c478bd9Sstevel@tonic-gate } 28717c478bd9Sstevel@tonic-gate /* 28727c478bd9Sstevel@tonic-gate * Atomically check for duplicate and insert in the table. 28737c478bd9Sstevel@tonic-gate */ 28747c478bd9Sstevel@tonic-gate for (ire1 = irb_ptr->irb_ire; ire1 != NULL; ire1 = ire1->ire_next) { 28757c478bd9Sstevel@tonic-gate if (ire1->ire_marks & IRE_MARK_CONDEMNED) 28767c478bd9Sstevel@tonic-gate continue; 28777c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 28787c478bd9Sstevel@tonic-gate /* 28797c478bd9Sstevel@tonic-gate * We do MATCH_IRE_ILL implicitly here for IREs 28807c478bd9Sstevel@tonic-gate * with a non-null ire_ipif, including IRE_CACHEs. 28817c478bd9Sstevel@tonic-gate * As ire_ipif and ire_stq could point to two 28827c478bd9Sstevel@tonic-gate * different ills, we can't pass just ire_ipif to 28837c478bd9Sstevel@tonic-gate * ire_match_args and get a match on both ills. 28847c478bd9Sstevel@tonic-gate * This is just needed for duplicate checks here and 28857c478bd9Sstevel@tonic-gate * so we don't add an extra argument to 28867c478bd9Sstevel@tonic-gate * ire_match_args for this. Do it locally. 28877c478bd9Sstevel@tonic-gate * 28887c478bd9Sstevel@tonic-gate * NOTE : Currently there is no part of the code 28897c478bd9Sstevel@tonic-gate * that asks for both MATH_IRE_IPIF and MATCH_IRE_ILL 28907c478bd9Sstevel@tonic-gate * match for IRE_CACHEs. Thus we don't want to 28917c478bd9Sstevel@tonic-gate * extend the arguments to ire_match_args. 28927c478bd9Sstevel@tonic-gate */ 28937c478bd9Sstevel@tonic-gate if (ire1->ire_stq != ire->ire_stq) 28947c478bd9Sstevel@tonic-gate continue; 28957c478bd9Sstevel@tonic-gate /* 28967c478bd9Sstevel@tonic-gate * Multiroute IRE_CACHEs for a given destination can 28977c478bd9Sstevel@tonic-gate * have the same ire_ipif, typically if their source 28987c478bd9Sstevel@tonic-gate * address is forced using RTF_SETSRC, and the same 28997c478bd9Sstevel@tonic-gate * send-to queue. We differentiate them using the parent 29007c478bd9Sstevel@tonic-gate * handle. 29017c478bd9Sstevel@tonic-gate */ 29027c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE && 29037c478bd9Sstevel@tonic-gate (ire1->ire_flags & RTF_MULTIRT) && 29047c478bd9Sstevel@tonic-gate (ire->ire_flags & RTF_MULTIRT) && 29057c478bd9Sstevel@tonic-gate (ire1->ire_phandle != ire->ire_phandle)) 29067c478bd9Sstevel@tonic-gate continue; 29077c478bd9Sstevel@tonic-gate } 29087c478bd9Sstevel@tonic-gate if (ire1->ire_zoneid != ire->ire_zoneid) 29097c478bd9Sstevel@tonic-gate continue; 29107c478bd9Sstevel@tonic-gate if (ire_match_args(ire1, ire->ire_addr, ire->ire_mask, 29117c478bd9Sstevel@tonic-gate ire->ire_gateway_addr, ire->ire_type, ire->ire_ipif, 291245916cd2Sjpk ire->ire_zoneid, 0, NULL, flags)) { 29137c478bd9Sstevel@tonic-gate /* 29147c478bd9Sstevel@tonic-gate * Return the old ire after doing a REFHOLD. 29157c478bd9Sstevel@tonic-gate * As most of the callers continue to use the IRE 29167c478bd9Sstevel@tonic-gate * after adding, we return a held ire. This will 29177c478bd9Sstevel@tonic-gate * avoid a lookup in the caller again. If the callers 29187c478bd9Sstevel@tonic-gate * don't want to use it, they need to do a REFRELE. 29197c478bd9Sstevel@tonic-gate */ 29207c478bd9Sstevel@tonic-gate ip1dbg(("found dup ire existing %p new %p", 29217c478bd9Sstevel@tonic-gate (void *)ire1, (void *)ire)); 29227c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire1); 29237c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 2924f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 29257c478bd9Sstevel@tonic-gate ire_delete(ire); 29267c478bd9Sstevel@tonic-gate if (pire != NULL) { 29277c478bd9Sstevel@tonic-gate /* 29287c478bd9Sstevel@tonic-gate * Assert that it is not removed from the 29297c478bd9Sstevel@tonic-gate * list yet. 29307c478bd9Sstevel@tonic-gate */ 29317c478bd9Sstevel@tonic-gate ASSERT(pire->ire_ptpn != NULL); 29327c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 29337c478bd9Sstevel@tonic-gate ire_refrele(pire); 29347c478bd9Sstevel@tonic-gate } 29357c478bd9Sstevel@tonic-gate *ire_p = ire1; 2936c793af95Ssangeeta if (need_refrele) 2937c793af95Ssangeeta IRB_REFRELE(irb_ptr); 29387c478bd9Sstevel@tonic-gate return (0); 29397c478bd9Sstevel@tonic-gate } 29407c478bd9Sstevel@tonic-gate } 2941c793af95Ssangeeta if (ire->ire_type & IRE_CACHE) { 2942c793af95Ssangeeta ASSERT(ire->ire_stq != NULL); 2943c793af95Ssangeeta nce = ndp_lookup_v4(ire_to_ill(ire), 2944c793af95Ssangeeta ((ire->ire_gateway_addr != INADDR_ANY) ? 2945c793af95Ssangeeta &ire->ire_gateway_addr : &ire->ire_addr), 2946c793af95Ssangeeta B_TRUE); 2947c793af95Ssangeeta if (nce != NULL) 2948c793af95Ssangeeta mutex_enter(&nce->nce_lock); 2949c793af95Ssangeeta /* 2950c793af95Ssangeeta * if the nce is NCE_F_CONDEMNED, or if it is not ND_REACHABLE 2951c793af95Ssangeeta * and the caller has prohibited the addition of incomplete 2952c793af95Ssangeeta * ire's, we fail the add. Note that nce_state could be 2953b9c344b3Ssowmini * something other than ND_REACHABLE if the nce had 2954b9c344b3Ssowmini * just expired and the ire_create preceding the 2955b9c344b3Ssowmini * ire_add added a new ND_INITIAL nce. 2956c793af95Ssangeeta */ 2957c793af95Ssangeeta if ((nce == NULL) || 2958c793af95Ssangeeta (nce->nce_flags & NCE_F_CONDEMNED) || 2959c793af95Ssangeeta (!allow_unresolved && 2960107875b0Ssangeeta (nce->nce_state != ND_REACHABLE))) { 2961b9c344b3Ssowmini if (nce != NULL) { 2962b9c344b3Ssowmini DTRACE_PROBE1(ire__bad__nce, nce_t *, nce); 2963c793af95Ssangeeta mutex_exit(&nce->nce_lock); 2964b9c344b3Ssowmini } 2965c793af95Ssangeeta ire_atomic_end(irb_ptr, ire); 2966f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 2967c793af95Ssangeeta if (nce != NULL) 2968c793af95Ssangeeta NCE_REFRELE(nce); 2969c793af95Ssangeeta DTRACE_PROBE1(ire__no__nce, ire_t *, ire); 2970c793af95Ssangeeta ire_delete(ire); 2971c793af95Ssangeeta if (pire != NULL) { 2972c793af95Ssangeeta IRB_REFRELE(pire->ire_bucket); 2973c793af95Ssangeeta ire_refrele(pire); 2974c793af95Ssangeeta } 2975c793af95Ssangeeta *ire_p = NULL; 2976c793af95Ssangeeta if (need_refrele) 2977c793af95Ssangeeta IRB_REFRELE(irb_ptr); 2978c793af95Ssangeeta return (EINVAL); 2979c793af95Ssangeeta } else { 2980c793af95Ssangeeta ire->ire_nce = nce; 2981c793af95Ssangeeta mutex_exit(&nce->nce_lock); 2982c793af95Ssangeeta /* 2983c793af95Ssangeeta * We are associating this nce to the ire, so 2984c793af95Ssangeeta * change the nce ref taken in ndp_lookup_v4() from 2985c793af95Ssangeeta * NCE_REFHOLD to NCE_REFHOLD_NOTR 2986c793af95Ssangeeta */ 2987c793af95Ssangeeta NCE_REFHOLD_TO_REFHOLD_NOTR(ire->ire_nce); 2988c793af95Ssangeeta } 2989c793af95Ssangeeta } 29907c478bd9Sstevel@tonic-gate /* 29917c478bd9Sstevel@tonic-gate * Make it easy for ip_wput_ire() to hit multiple broadcast ires by 29927c478bd9Sstevel@tonic-gate * grouping identical addresses together on the hash chain. We also 29937c478bd9Sstevel@tonic-gate * don't want to send multiple copies out if there are two ills part 29947c478bd9Sstevel@tonic-gate * of the same group. Thus we group the ires with same addr and same 29957c478bd9Sstevel@tonic-gate * ill group together so that ip_wput_ire can easily skip all the 29967c478bd9Sstevel@tonic-gate * ires with same addr and same group after sending the first copy. 29977c478bd9Sstevel@tonic-gate * We do this only for IRE_BROADCASTs as ip_wput_ire is currently 29987c478bd9Sstevel@tonic-gate * interested in such groupings only for broadcasts. 29997c478bd9Sstevel@tonic-gate * 30007c478bd9Sstevel@tonic-gate * NOTE : If the interfaces are brought up first and then grouped, 30017c478bd9Sstevel@tonic-gate * illgrp_insert will handle it. We come here when the interfaces 30027c478bd9Sstevel@tonic-gate * are already in group and we are bringing them UP. 30037c478bd9Sstevel@tonic-gate * 30047c478bd9Sstevel@tonic-gate * Find the first entry that matches ire_addr. *irep will be null 30057c478bd9Sstevel@tonic-gate * if no match. 300601802382Ssowmini * 300701802382Ssowmini * Note: the loopback and non-loopback broadcast entries for an 300801802382Ssowmini * interface MUST be added before any MULTIRT entries. 30097c478bd9Sstevel@tonic-gate */ 30107c478bd9Sstevel@tonic-gate irep = (ire_t **)irb_ptr; 30117c478bd9Sstevel@tonic-gate while ((ire1 = *irep) != NULL && ire->ire_addr != ire1->ire_addr) 30127c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 30137c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_BROADCAST && *irep != NULL) { 30147c478bd9Sstevel@tonic-gate /* 30157c478bd9Sstevel@tonic-gate * We found some ire (i.e *irep) with a matching addr. We 30167c478bd9Sstevel@tonic-gate * want to group ires with same addr and same ill group 30177c478bd9Sstevel@tonic-gate * together. 30187c478bd9Sstevel@tonic-gate * 30197c478bd9Sstevel@tonic-gate * First get to the entry that matches our address and 30207c478bd9Sstevel@tonic-gate * ill group i.e stop as soon as we find the first ire 30217c478bd9Sstevel@tonic-gate * matching the ill group and address. If there is only 30227c478bd9Sstevel@tonic-gate * an address match, we should walk and look for some 30237c478bd9Sstevel@tonic-gate * group match. These are some of the possible scenarios : 30247c478bd9Sstevel@tonic-gate * 30257c478bd9Sstevel@tonic-gate * 1) There are no groups at all i.e all ire's ill_group 30267c478bd9Sstevel@tonic-gate * are NULL. In that case we will essentially group 30277c478bd9Sstevel@tonic-gate * all the ires with the same addr together. Same as 30287c478bd9Sstevel@tonic-gate * the "else" block of this "if". 30297c478bd9Sstevel@tonic-gate * 30307c478bd9Sstevel@tonic-gate * 2) There are some groups and this ire's ill_group is 30317c478bd9Sstevel@tonic-gate * NULL. In this case, we will first find the group 30327c478bd9Sstevel@tonic-gate * that matches the address and a NULL group. Then 30337c478bd9Sstevel@tonic-gate * we will insert the ire at the end of that group. 30347c478bd9Sstevel@tonic-gate * 30357c478bd9Sstevel@tonic-gate * 3) There are some groups and this ires's ill_group is 30367c478bd9Sstevel@tonic-gate * non-NULL. In this case we will first find the group 30377c478bd9Sstevel@tonic-gate * that matches the address and the ill_group. Then 30387c478bd9Sstevel@tonic-gate * we will insert the ire at the end of that group. 30397c478bd9Sstevel@tonic-gate */ 304001802382Ssowmini for (;;) { 30417c478bd9Sstevel@tonic-gate ire1 = *irep; 30427c478bd9Sstevel@tonic-gate if ((ire1->ire_next == NULL) || 30437c478bd9Sstevel@tonic-gate (ire1->ire_next->ire_addr != ire->ire_addr) || 30447c478bd9Sstevel@tonic-gate (ire1->ire_type != IRE_BROADCAST) || 304501802382Ssowmini (ire1->ire_flags & RTF_MULTIRT) || 30467c478bd9Sstevel@tonic-gate (ire1->ire_ipif->ipif_ill->ill_group == 30477c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ill->ill_group)) 30487c478bd9Sstevel@tonic-gate break; 30497c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 30507c478bd9Sstevel@tonic-gate } 30517c478bd9Sstevel@tonic-gate ASSERT(*irep != NULL); 305201802382Ssowmini /* 305301802382Ssowmini * The ire will be added before *irep, so 305401802382Ssowmini * if irep is a MULTIRT ire, just break to 305501802382Ssowmini * ire insertion code. 305601802382Ssowmini */ 305701802382Ssowmini if (((*irep)->ire_flags & RTF_MULTIRT) != 0) 305801802382Ssowmini goto insert_ire; 305901802382Ssowmini 30607c478bd9Sstevel@tonic-gate irep = &((*irep)->ire_next); 30617c478bd9Sstevel@tonic-gate 30627c478bd9Sstevel@tonic-gate /* 30637c478bd9Sstevel@tonic-gate * Either we have hit the end of the list or the address 30647c478bd9Sstevel@tonic-gate * did not match or the group *matched*. If we found 30657c478bd9Sstevel@tonic-gate * a match on the group, skip to the end of the group. 30667c478bd9Sstevel@tonic-gate */ 30677c478bd9Sstevel@tonic-gate while (*irep != NULL) { 30687c478bd9Sstevel@tonic-gate ire1 = *irep; 30697c478bd9Sstevel@tonic-gate if ((ire1->ire_addr != ire->ire_addr) || 30707c478bd9Sstevel@tonic-gate (ire1->ire_type != IRE_BROADCAST) || 30717c478bd9Sstevel@tonic-gate (ire1->ire_ipif->ipif_ill->ill_group != 30727c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ill->ill_group)) 30737c478bd9Sstevel@tonic-gate break; 30747c478bd9Sstevel@tonic-gate if (ire1->ire_ipif->ipif_ill->ill_group == NULL && 30757c478bd9Sstevel@tonic-gate ire1->ire_ipif == ire->ire_ipif) { 30767c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 30777c478bd9Sstevel@tonic-gate break; 30787c478bd9Sstevel@tonic-gate } 30797c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 30807c478bd9Sstevel@tonic-gate } 30817c478bd9Sstevel@tonic-gate } else if (*irep != NULL) { 30827c478bd9Sstevel@tonic-gate /* 30837c478bd9Sstevel@tonic-gate * Find the last ire which matches ire_addr. 30847c478bd9Sstevel@tonic-gate * Needed to do tail insertion among entries with the same 30857c478bd9Sstevel@tonic-gate * ire_addr. 30867c478bd9Sstevel@tonic-gate */ 30877c478bd9Sstevel@tonic-gate while (ire->ire_addr == ire1->ire_addr) { 30887c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 30897c478bd9Sstevel@tonic-gate ire1 = *irep; 30907c478bd9Sstevel@tonic-gate if (ire1 == NULL) 30917c478bd9Sstevel@tonic-gate break; 30927c478bd9Sstevel@tonic-gate } 30937c478bd9Sstevel@tonic-gate } 30947c478bd9Sstevel@tonic-gate 309501802382Ssowmini insert_ire: 30967c478bd9Sstevel@tonic-gate /* Insert at *irep */ 30977c478bd9Sstevel@tonic-gate ire1 = *irep; 30987c478bd9Sstevel@tonic-gate if (ire1 != NULL) 30997c478bd9Sstevel@tonic-gate ire1->ire_ptpn = &ire->ire_next; 31007c478bd9Sstevel@tonic-gate ire->ire_next = ire1; 31017c478bd9Sstevel@tonic-gate /* Link the new one in. */ 31027c478bd9Sstevel@tonic-gate ire->ire_ptpn = irep; 31037c478bd9Sstevel@tonic-gate 31047c478bd9Sstevel@tonic-gate /* 31057c478bd9Sstevel@tonic-gate * ire_walk routines de-reference ire_next without holding 31067c478bd9Sstevel@tonic-gate * a lock. Before we point to the new ire, we want to make 31077c478bd9Sstevel@tonic-gate * sure the store that sets the ire_next of the new ire 31087c478bd9Sstevel@tonic-gate * reaches global visibility, so that ire_walk routines 31097c478bd9Sstevel@tonic-gate * don't see a truncated list of ires i.e if the ire_next 31107c478bd9Sstevel@tonic-gate * of the new ire gets set after we do "*irep = ire" due 31117c478bd9Sstevel@tonic-gate * to re-ordering, the ire_walk thread will see a NULL 31127c478bd9Sstevel@tonic-gate * once it accesses the ire_next of the new ire. 31137c478bd9Sstevel@tonic-gate * membar_producer() makes sure that the following store 31147c478bd9Sstevel@tonic-gate * happens *after* all of the above stores. 31157c478bd9Sstevel@tonic-gate */ 31167c478bd9Sstevel@tonic-gate membar_producer(); 31177c478bd9Sstevel@tonic-gate *irep = ire; 31187c478bd9Sstevel@tonic-gate ire->ire_bucket = irb_ptr; 31197c478bd9Sstevel@tonic-gate /* 31207c478bd9Sstevel@tonic-gate * We return a bumped up IRE above. Keep it symmetrical 31217c478bd9Sstevel@tonic-gate * so that the callers will always have to release. This 31227c478bd9Sstevel@tonic-gate * helps the callers of this function because they continue 31237c478bd9Sstevel@tonic-gate * to use the IRE after adding and hence they don't have to 31247c478bd9Sstevel@tonic-gate * lookup again after we return the IRE. 31257c478bd9Sstevel@tonic-gate * 31267c478bd9Sstevel@tonic-gate * NOTE : We don't have to use atomics as this is appearing 31277c478bd9Sstevel@tonic-gate * in the list for the first time and no one else can bump 31287c478bd9Sstevel@tonic-gate * up the reference count on this yet. 31297c478bd9Sstevel@tonic-gate */ 31307c478bd9Sstevel@tonic-gate IRE_REFHOLD_LOCKED(ire); 3131f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_inserted); 3132c793af95Ssangeeta 31337c478bd9Sstevel@tonic-gate irb_ptr->irb_ire_cnt++; 3134c793af95Ssangeeta if (irb_ptr->irb_marks & IRB_MARK_FTABLE) 3135c793af95Ssangeeta irb_ptr->irb_nire++; 3136c793af95Ssangeeta 31377c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_TEMPORARY) 31387c478bd9Sstevel@tonic-gate irb_ptr->irb_tmp_ire_cnt++; 31397c478bd9Sstevel@tonic-gate 31407c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 3141968d2fd1Ssowmini DTRACE_PROBE3(ipif__incr__cnt, (ipif_t *), ire->ire_ipif, 3142968d2fd1Ssowmini (char *), "ire", (void *), ire); 3143384ad179Ssowmini ire->ire_ipif->ipif_ire_cnt++; 31447c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) { 31457c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 3146968d2fd1Ssowmini DTRACE_PROBE3(ill__incr__cnt, (ill_t *), stq_ill, 3147968d2fd1Ssowmini (char *), "ire", (void *), ire); 3148384ad179Ssowmini stq_ill->ill_ire_cnt++; 31497c478bd9Sstevel@tonic-gate } 31507c478bd9Sstevel@tonic-gate } else { 31517c478bd9Sstevel@tonic-gate ASSERT(ire->ire_stq == NULL); 31527c478bd9Sstevel@tonic-gate } 31537c478bd9Sstevel@tonic-gate 31547c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 3155f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 31567c478bd9Sstevel@tonic-gate 31577c478bd9Sstevel@tonic-gate if (pire != NULL) { 31587c478bd9Sstevel@tonic-gate /* Assert that it is not removed from the list yet */ 31597c478bd9Sstevel@tonic-gate ASSERT(pire->ire_ptpn != NULL); 31607c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 31617c478bd9Sstevel@tonic-gate ire_refrele(pire); 31627c478bd9Sstevel@tonic-gate } 31637c478bd9Sstevel@tonic-gate 31647c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) { 31657c478bd9Sstevel@tonic-gate /* 3166c793af95Ssangeeta * For ire's with host mask see if there is an entry 31677c478bd9Sstevel@tonic-gate * in the cache. If there is one flush the whole cache as 31687c478bd9Sstevel@tonic-gate * there might be multiple entries due to RTF_MULTIRT (CGTP). 31697c478bd9Sstevel@tonic-gate * If no entry is found than there is no need to flush the 31707c478bd9Sstevel@tonic-gate * cache. 31717c478bd9Sstevel@tonic-gate */ 31727c478bd9Sstevel@tonic-gate if (ire->ire_mask == IP_HOST_MASK) { 31737c478bd9Sstevel@tonic-gate ire_t *lire; 31747c478bd9Sstevel@tonic-gate lire = ire_ctable_lookup(ire->ire_addr, NULL, IRE_CACHE, 3175f4b3ec61Sdh155122 NULL, ALL_ZONES, NULL, MATCH_IRE_TYPE, ipst); 31767c478bd9Sstevel@tonic-gate if (lire != NULL) { 31777c478bd9Sstevel@tonic-gate ire_refrele(lire); 31787c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire, IRE_FLUSH_ADD); 31797c478bd9Sstevel@tonic-gate } 31807c478bd9Sstevel@tonic-gate } else { 31817c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire, IRE_FLUSH_ADD); 31827c478bd9Sstevel@tonic-gate } 31837c478bd9Sstevel@tonic-gate } 31847c478bd9Sstevel@tonic-gate /* 31857c478bd9Sstevel@tonic-gate * We had to delay the fast path probe until the ire is inserted 31867c478bd9Sstevel@tonic-gate * in the list. Otherwise the fast path ack won't find the ire in 31877c478bd9Sstevel@tonic-gate * the table. 31887c478bd9Sstevel@tonic-gate */ 3189516bda92Ssowmini if (ire->ire_type == IRE_CACHE || 3190516bda92Ssowmini (ire->ire_type == IRE_BROADCAST && ire->ire_stq != NULL)) { 3191516bda92Ssowmini ASSERT(ire->ire_nce != NULL); 319254da8755Ssowmini if (ire->ire_nce->nce_state == ND_REACHABLE) 3193516bda92Ssowmini nce_fastpath(ire->ire_nce); 3194516bda92Ssowmini } 31957c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) 31967c478bd9Sstevel@tonic-gate ASSERT(!MUTEX_HELD(&ire->ire_ipif->ipif_ill->ill_lock)); 31977c478bd9Sstevel@tonic-gate *ire_p = ire; 3198c793af95Ssangeeta if (need_refrele) { 3199c793af95Ssangeeta IRB_REFRELE(irb_ptr); 3200c793af95Ssangeeta } 32017c478bd9Sstevel@tonic-gate return (0); 32027c478bd9Sstevel@tonic-gate } 32037c478bd9Sstevel@tonic-gate 32047c478bd9Sstevel@tonic-gate /* 32057c478bd9Sstevel@tonic-gate * IRB_REFRELE is the only caller of the function. ire_unlink calls to 32067c478bd9Sstevel@tonic-gate * do the final cleanup for this ire. 32077c478bd9Sstevel@tonic-gate */ 32087c478bd9Sstevel@tonic-gate void 32097c478bd9Sstevel@tonic-gate ire_cleanup(ire_t *ire) 32107c478bd9Sstevel@tonic-gate { 32117c478bd9Sstevel@tonic-gate ire_t *ire_next; 3212f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 32137c478bd9Sstevel@tonic-gate 32147c478bd9Sstevel@tonic-gate ASSERT(ire != NULL); 32157c478bd9Sstevel@tonic-gate 32167c478bd9Sstevel@tonic-gate while (ire != NULL) { 32177c478bd9Sstevel@tonic-gate ire_next = ire->ire_next; 32187c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 32197c478bd9Sstevel@tonic-gate ire_delete_v4(ire); 3220f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, 3221f4b3ec61Sdh155122 ire_stats_deleted); 32227c478bd9Sstevel@tonic-gate } else { 32237c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV6_VERSION); 32247c478bd9Sstevel@tonic-gate ire_delete_v6(ire); 3225f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v6, 3226f4b3ec61Sdh155122 ire_stats_deleted); 32277c478bd9Sstevel@tonic-gate } 32287c478bd9Sstevel@tonic-gate /* 32297c478bd9Sstevel@tonic-gate * Now it's really out of the list. Before doing the 32307c478bd9Sstevel@tonic-gate * REFRELE, set ire_next to NULL as ire_inactive asserts 32317c478bd9Sstevel@tonic-gate * so. 32327c478bd9Sstevel@tonic-gate */ 32337c478bd9Sstevel@tonic-gate ire->ire_next = NULL; 32347c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 32357c478bd9Sstevel@tonic-gate ire = ire_next; 32367c478bd9Sstevel@tonic-gate } 32377c478bd9Sstevel@tonic-gate } 32387c478bd9Sstevel@tonic-gate 32397c478bd9Sstevel@tonic-gate /* 32407c478bd9Sstevel@tonic-gate * IRB_REFRELE is the only caller of the function. It calls to unlink 32417c478bd9Sstevel@tonic-gate * all the CONDEMNED ires from this bucket. 32427c478bd9Sstevel@tonic-gate */ 32437c478bd9Sstevel@tonic-gate ire_t * 32447c478bd9Sstevel@tonic-gate ire_unlink(irb_t *irb) 32457c478bd9Sstevel@tonic-gate { 32467c478bd9Sstevel@tonic-gate ire_t *ire; 32477c478bd9Sstevel@tonic-gate ire_t *ire1; 32487c478bd9Sstevel@tonic-gate ire_t **ptpn; 32497c478bd9Sstevel@tonic-gate ire_t *ire_list = NULL; 32507c478bd9Sstevel@tonic-gate 32517c478bd9Sstevel@tonic-gate ASSERT(RW_WRITE_HELD(&irb->irb_lock)); 3252c793af95Ssangeeta ASSERT(((irb->irb_marks & IRB_MARK_FTABLE) && irb->irb_refcnt == 1) || 3253c793af95Ssangeeta (irb->irb_refcnt == 0)); 3254c793af95Ssangeeta ASSERT(irb->irb_marks & IRB_MARK_CONDEMNED); 32557c478bd9Sstevel@tonic-gate ASSERT(irb->irb_ire != NULL); 32567c478bd9Sstevel@tonic-gate 32577c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; ire = ire1) { 3258f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 3259f4b3ec61Sdh155122 32607c478bd9Sstevel@tonic-gate ire1 = ire->ire_next; 32617c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) { 32627c478bd9Sstevel@tonic-gate ptpn = ire->ire_ptpn; 32637c478bd9Sstevel@tonic-gate ire1 = ire->ire_next; 32647c478bd9Sstevel@tonic-gate if (ire1) 32657c478bd9Sstevel@tonic-gate ire1->ire_ptpn = ptpn; 32667c478bd9Sstevel@tonic-gate *ptpn = ire1; 32677c478bd9Sstevel@tonic-gate ire->ire_ptpn = NULL; 32687c478bd9Sstevel@tonic-gate ire->ire_next = NULL; 32697c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT) { 32707c478bd9Sstevel@tonic-gate /* 32717c478bd9Sstevel@tonic-gate * IRE is out of the list. We need to adjust 32727c478bd9Sstevel@tonic-gate * the accounting before the caller drops 32737c478bd9Sstevel@tonic-gate * the lock. 32747c478bd9Sstevel@tonic-gate */ 32757c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 3276f4b3ec61Sdh155122 ASSERT(ipst-> 3277f4b3ec61Sdh155122 ips_ipv6_ire_default_count != 3278f4b3ec61Sdh155122 0); 3279f4b3ec61Sdh155122 ipst->ips_ipv6_ire_default_count--; 32807c478bd9Sstevel@tonic-gate } 32817c478bd9Sstevel@tonic-gate } 32827c478bd9Sstevel@tonic-gate /* 32837c478bd9Sstevel@tonic-gate * We need to call ire_delete_v4 or ire_delete_v6 32847c478bd9Sstevel@tonic-gate * to clean up the cache or the redirects pointing at 32857c478bd9Sstevel@tonic-gate * the default gateway. We need to drop the lock 32867c478bd9Sstevel@tonic-gate * as ire_flush_cache/ire_delete_host_redircts require 32877c478bd9Sstevel@tonic-gate * so. But we can't drop the lock, as ire_unlink needs 32887c478bd9Sstevel@tonic-gate * to atomically remove the ires from the list. 32897c478bd9Sstevel@tonic-gate * So, create a temporary list of CONDEMNED ires 32907c478bd9Sstevel@tonic-gate * for doing ire_delete_v4/ire_delete_v6 operations 32917c478bd9Sstevel@tonic-gate * later on. 32927c478bd9Sstevel@tonic-gate */ 32937c478bd9Sstevel@tonic-gate ire->ire_next = ire_list; 32947c478bd9Sstevel@tonic-gate ire_list = ire; 32957c478bd9Sstevel@tonic-gate } 32967c478bd9Sstevel@tonic-gate } 3297c793af95Ssangeeta irb->irb_marks &= ~IRB_MARK_CONDEMNED; 32987c478bd9Sstevel@tonic-gate return (ire_list); 32997c478bd9Sstevel@tonic-gate } 33007c478bd9Sstevel@tonic-gate 33017c478bd9Sstevel@tonic-gate /* 33027c478bd9Sstevel@tonic-gate * Delete all the cache entries with this 'addr'. When IP gets a gratuitous 3303c793af95Ssangeeta * ARP message on any of its interface queue, it scans the nce table and 3304c793af95Ssangeeta * deletes and calls ndp_delete() for the appropriate nce. This action 3305c793af95Ssangeeta * also deletes all the neighbor/ire cache entries for that address. 3306c793af95Ssangeeta * This function is called from ip_arp_news in ip.c and also for 3307c793af95Ssangeeta * ARP ioctl processing in ip_if.c. ip_ire_clookup_and_delete returns 3308c793af95Ssangeeta * true if it finds a nce entry which is used by ip_arp_news to determine if 3309c793af95Ssangeeta * it needs to do an ire_walk_v4. The return value is also used for the 3310c793af95Ssangeeta * same purpose by ARP IOCTL processing * in ip_if.c when deleting 3311c793af95Ssangeeta * ARP entries. For SIOC*IFARP ioctls in addition to the address, 3312c793af95Ssangeeta * ip_if->ipif_ill also needs to be matched. 33137c478bd9Sstevel@tonic-gate */ 33147c478bd9Sstevel@tonic-gate boolean_t 3315f4b3ec61Sdh155122 ip_ire_clookup_and_delete(ipaddr_t addr, ipif_t *ipif, ip_stack_t *ipst) 33167c478bd9Sstevel@tonic-gate { 33177c478bd9Sstevel@tonic-gate ill_t *ill; 3318c793af95Ssangeeta nce_t *nce; 33197c478bd9Sstevel@tonic-gate 3320c793af95Ssangeeta ill = (ipif ? ipif->ipif_ill : NULL); 33217c478bd9Sstevel@tonic-gate 3322c793af95Ssangeeta if (ill != NULL) { 3323c793af95Ssangeeta /* 3324c793af95Ssangeeta * clean up the nce (and any relevant ire's) that matches 3325c793af95Ssangeeta * on addr and ill. 3326c793af95Ssangeeta */ 3327c793af95Ssangeeta nce = ndp_lookup_v4(ill, &addr, B_FALSE); 3328c793af95Ssangeeta if (nce != NULL) { 3329c793af95Ssangeeta ndp_delete(nce); 3330c793af95Ssangeeta return (B_TRUE); 33317c478bd9Sstevel@tonic-gate } 3332c793af95Ssangeeta } else { 3333c793af95Ssangeeta /* 3334c793af95Ssangeeta * ill is wildcard. clean up all nce's and 3335c793af95Ssangeeta * ire's that match on addr 3336c793af95Ssangeeta */ 3337c793af95Ssangeeta nce_clookup_t cl; 33387c478bd9Sstevel@tonic-gate 3339c793af95Ssangeeta cl.ncecl_addr = addr; 3340c793af95Ssangeeta cl.ncecl_found = B_FALSE; 33417c478bd9Sstevel@tonic-gate 3342f4b3ec61Sdh155122 ndp_walk_common(ipst->ips_ndp4, NULL, 3343c793af95Ssangeeta (pfi_t)ip_nce_clookup_and_delete, (uchar_t *)&cl, B_TRUE); 3344c793af95Ssangeeta 3345c793af95Ssangeeta /* 3346c793af95Ssangeeta * ncecl_found would be set by ip_nce_clookup_and_delete if 3347c793af95Ssangeeta * we found a matching nce. 3348c793af95Ssangeeta */ 3349c793af95Ssangeeta return (cl.ncecl_found); 3350c793af95Ssangeeta } 3351c793af95Ssangeeta return (B_FALSE); 3352c793af95Ssangeeta 3353c793af95Ssangeeta } 3354c793af95Ssangeeta 3355c793af95Ssangeeta /* Delete the supplied nce if its nce_addr matches the supplied address */ 3356c793af95Ssangeeta static void 3357c793af95Ssangeeta ip_nce_clookup_and_delete(nce_t *nce, void *arg) 3358c793af95Ssangeeta { 3359c793af95Ssangeeta nce_clookup_t *cl = (nce_clookup_t *)arg; 3360c793af95Ssangeeta ipaddr_t nce_addr; 3361c793af95Ssangeeta 3362c793af95Ssangeeta IN6_V4MAPPED_TO_IPADDR(&nce->nce_addr, nce_addr); 3363c793af95Ssangeeta if (nce_addr == cl->ncecl_addr) { 3364c793af95Ssangeeta cl->ncecl_found = B_TRUE; 3365c793af95Ssangeeta /* clean up the nce (and any relevant ire's) */ 3366c793af95Ssangeeta ndp_delete(nce); 3367c793af95Ssangeeta } 3368c793af95Ssangeeta } 3369c793af95Ssangeeta 3370c793af95Ssangeeta /* 3371c793af95Ssangeeta * Clean up the radix node for this ire. Must be called by IRB_REFRELE 3372c793af95Ssangeeta * when there are no ire's left in the bucket. Returns TRUE if the bucket 3373c793af95Ssangeeta * is deleted and freed. 3374c793af95Ssangeeta */ 3375c793af95Ssangeeta boolean_t 3376c793af95Ssangeeta irb_inactive(irb_t *irb) 3377c793af95Ssangeeta { 3378c793af95Ssangeeta struct rt_entry *rt; 3379c793af95Ssangeeta struct radix_node *rn; 3380f4b3ec61Sdh155122 ip_stack_t *ipst = irb->irb_ipst; 3381f4b3ec61Sdh155122 3382f4b3ec61Sdh155122 ASSERT(irb->irb_ipst != NULL); 3383c793af95Ssangeeta 3384c793af95Ssangeeta rt = IRB2RT(irb); 3385c793af95Ssangeeta rn = (struct radix_node *)rt; 3386c793af95Ssangeeta 3387c793af95Ssangeeta /* first remove it from the radix tree. */ 3388f4b3ec61Sdh155122 RADIX_NODE_HEAD_WLOCK(ipst->ips_ip_ftable); 3389c793af95Ssangeeta rw_enter(&irb->irb_lock, RW_WRITER); 3390c793af95Ssangeeta if (irb->irb_refcnt == 1 && irb->irb_nire == 0) { 3391f4b3ec61Sdh155122 rn = ipst->ips_ip_ftable->rnh_deladdr(rn->rn_key, rn->rn_mask, 3392f4b3ec61Sdh155122 ipst->ips_ip_ftable); 3393c793af95Ssangeeta DTRACE_PROBE1(irb__free, rt_t *, rt); 3394c793af95Ssangeeta ASSERT((void *)rn == (void *)rt); 3395c793af95Ssangeeta Free(rt, rt_entry_cache); 3396c793af95Ssangeeta /* irb_lock is freed */ 3397f4b3ec61Sdh155122 RADIX_NODE_HEAD_UNLOCK(ipst->ips_ip_ftable); 3398c793af95Ssangeeta return (B_TRUE); 3399c793af95Ssangeeta } 3400c793af95Ssangeeta rw_exit(&irb->irb_lock); 3401f4b3ec61Sdh155122 RADIX_NODE_HEAD_UNLOCK(ipst->ips_ip_ftable); 3402c793af95Ssangeeta return (B_FALSE); 34037c478bd9Sstevel@tonic-gate } 34047c478bd9Sstevel@tonic-gate 34057c478bd9Sstevel@tonic-gate /* 34067c478bd9Sstevel@tonic-gate * Delete the specified IRE. 34077c478bd9Sstevel@tonic-gate */ 34087c478bd9Sstevel@tonic-gate void 34097c478bd9Sstevel@tonic-gate ire_delete(ire_t *ire) 34107c478bd9Sstevel@tonic-gate { 34117c478bd9Sstevel@tonic-gate ire_t *ire1; 34127c478bd9Sstevel@tonic-gate ire_t **ptpn; 34137c478bd9Sstevel@tonic-gate irb_t *irb; 3414f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 34157c478bd9Sstevel@tonic-gate 3416c793af95Ssangeeta if ((irb = ire->ire_bucket) == NULL) { 34177c478bd9Sstevel@tonic-gate /* 34187c478bd9Sstevel@tonic-gate * It was never inserted in the list. Should call REFRELE 34197c478bd9Sstevel@tonic-gate * to free this IRE. 34207c478bd9Sstevel@tonic-gate */ 34217c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 34227c478bd9Sstevel@tonic-gate return; 34237c478bd9Sstevel@tonic-gate } 34247c478bd9Sstevel@tonic-gate 34257c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_WRITER); 34267c478bd9Sstevel@tonic-gate 3427c793af95Ssangeeta if (irb->irb_rr_origin == ire) { 3428c793af95Ssangeeta irb->irb_rr_origin = NULL; 3429c793af95Ssangeeta } 3430c793af95Ssangeeta 34317c478bd9Sstevel@tonic-gate /* 34327c478bd9Sstevel@tonic-gate * In case of V4 we might still be waiting for fastpath ack. 34337c478bd9Sstevel@tonic-gate */ 3434516bda92Ssowmini if (ire->ire_ipversion == IPV4_VERSION && 3435516bda92Ssowmini (ire->ire_type == IRE_CACHE || 3436516bda92Ssowmini (ire->ire_type == IRE_BROADCAST && ire->ire_stq != NULL))) { 3437516bda92Ssowmini ASSERT(ire->ire_nce != NULL); 3438516bda92Ssowmini nce_fastpath_list_delete(ire->ire_nce); 34397c478bd9Sstevel@tonic-gate } 34407c478bd9Sstevel@tonic-gate 34417c478bd9Sstevel@tonic-gate if (ire->ire_ptpn == NULL) { 34427c478bd9Sstevel@tonic-gate /* 34437c478bd9Sstevel@tonic-gate * Some other thread has removed us from the list. 34447c478bd9Sstevel@tonic-gate * It should have done the REFRELE for us. 34457c478bd9Sstevel@tonic-gate */ 34467c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 34477c478bd9Sstevel@tonic-gate return; 34487c478bd9Sstevel@tonic-gate } 34497c478bd9Sstevel@tonic-gate 34509a09d68dSja97890 if (!(ire->ire_marks & IRE_MARK_CONDEMNED)) { 34519a09d68dSja97890 irb->irb_ire_cnt--; 34529a09d68dSja97890 ire->ire_marks |= IRE_MARK_CONDEMNED; 34539a09d68dSja97890 if (ire->ire_marks & IRE_MARK_TEMPORARY) { 34549a09d68dSja97890 irb->irb_tmp_ire_cnt--; 34559a09d68dSja97890 ire->ire_marks &= ~IRE_MARK_TEMPORARY; 34569a09d68dSja97890 } 34579a09d68dSja97890 } 34589a09d68dSja97890 34597c478bd9Sstevel@tonic-gate if (irb->irb_refcnt != 0) { 34607c478bd9Sstevel@tonic-gate /* 34617c478bd9Sstevel@tonic-gate * The last thread to leave this bucket will 34627c478bd9Sstevel@tonic-gate * delete this ire. 34637c478bd9Sstevel@tonic-gate */ 3464c793af95Ssangeeta irb->irb_marks |= IRB_MARK_CONDEMNED; 34657c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 34667c478bd9Sstevel@tonic-gate return; 34677c478bd9Sstevel@tonic-gate } 34687c478bd9Sstevel@tonic-gate 34697c478bd9Sstevel@tonic-gate /* 34707c478bd9Sstevel@tonic-gate * Normally to delete an ire, we walk the bucket. While we 34717c478bd9Sstevel@tonic-gate * walk the bucket, we normally bump up irb_refcnt and hence 34727c478bd9Sstevel@tonic-gate * we return from above where we mark CONDEMNED and the ire 34737c478bd9Sstevel@tonic-gate * gets deleted from ire_unlink. This case is where somebody 34747c478bd9Sstevel@tonic-gate * knows the ire e.g by doing a lookup, and wants to delete the 34757c478bd9Sstevel@tonic-gate * IRE. irb_refcnt would be 0 in this case if nobody is walking 34767c478bd9Sstevel@tonic-gate * the bucket. 34777c478bd9Sstevel@tonic-gate */ 34787c478bd9Sstevel@tonic-gate ptpn = ire->ire_ptpn; 34797c478bd9Sstevel@tonic-gate ire1 = ire->ire_next; 34807c478bd9Sstevel@tonic-gate if (ire1 != NULL) 34817c478bd9Sstevel@tonic-gate ire1->ire_ptpn = ptpn; 34827c478bd9Sstevel@tonic-gate ASSERT(ptpn != NULL); 34837c478bd9Sstevel@tonic-gate *ptpn = ire1; 34847c478bd9Sstevel@tonic-gate ire->ire_ptpn = NULL; 34857c478bd9Sstevel@tonic-gate ire->ire_next = NULL; 34867c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 3487f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v6, ire_stats_deleted); 34887c478bd9Sstevel@tonic-gate } else { 3489f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_deleted); 34907c478bd9Sstevel@tonic-gate } 34917c478bd9Sstevel@tonic-gate /* 34927c478bd9Sstevel@tonic-gate * ip_wput/ip_wput_v6 checks this flag to see whether 34937c478bd9Sstevel@tonic-gate * it should still use the cached ire or not. 34947c478bd9Sstevel@tonic-gate */ 34957c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT) { 34967c478bd9Sstevel@tonic-gate /* 34977c478bd9Sstevel@tonic-gate * IRE is out of the list. We need to adjust the 34987c478bd9Sstevel@tonic-gate * accounting before we drop the lock. 34997c478bd9Sstevel@tonic-gate */ 35007c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 3501f4b3ec61Sdh155122 ASSERT(ipst->ips_ipv6_ire_default_count != 0); 3502f4b3ec61Sdh155122 ipst->ips_ipv6_ire_default_count--; 35037c478bd9Sstevel@tonic-gate } 35047c478bd9Sstevel@tonic-gate } 35057c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 35067c478bd9Sstevel@tonic-gate 35077c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 35087c478bd9Sstevel@tonic-gate ire_delete_v6(ire); 35097c478bd9Sstevel@tonic-gate } else { 35107c478bd9Sstevel@tonic-gate ire_delete_v4(ire); 35117c478bd9Sstevel@tonic-gate } 35127c478bd9Sstevel@tonic-gate /* 35137c478bd9Sstevel@tonic-gate * We removed it from the list. Decrement the 35147c478bd9Sstevel@tonic-gate * reference count. 35157c478bd9Sstevel@tonic-gate */ 35167c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 35177c478bd9Sstevel@tonic-gate } 35187c478bd9Sstevel@tonic-gate 35197c478bd9Sstevel@tonic-gate /* 35207c478bd9Sstevel@tonic-gate * Delete the specified IRE. 35217c478bd9Sstevel@tonic-gate * All calls should use ire_delete(). 35227c478bd9Sstevel@tonic-gate * Sometimes called as writer though not required by this function. 35237c478bd9Sstevel@tonic-gate * 35247c478bd9Sstevel@tonic-gate * NOTE : This function is called only if the ire was added 35257c478bd9Sstevel@tonic-gate * in the list. 35267c478bd9Sstevel@tonic-gate */ 35277c478bd9Sstevel@tonic-gate static void 35287c478bd9Sstevel@tonic-gate ire_delete_v4(ire_t *ire) 35297c478bd9Sstevel@tonic-gate { 3530f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 3531f4b3ec61Sdh155122 35327c478bd9Sstevel@tonic-gate ASSERT(ire->ire_refcnt >= 1); 35337c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 35347c478bd9Sstevel@tonic-gate 35357c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 35367c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire, IRE_FLUSH_DELETE); 35377c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT) { 35387c478bd9Sstevel@tonic-gate /* 35397c478bd9Sstevel@tonic-gate * when a default gateway is going away 35407c478bd9Sstevel@tonic-gate * delete all the host redirects pointing at that 35417c478bd9Sstevel@tonic-gate * gateway. 35427c478bd9Sstevel@tonic-gate */ 3543f4b3ec61Sdh155122 ire_delete_host_redirects(ire->ire_gateway_addr, ipst); 35447c478bd9Sstevel@tonic-gate } 35457c478bd9Sstevel@tonic-gate } 35467c478bd9Sstevel@tonic-gate 35477c478bd9Sstevel@tonic-gate /* 35487c478bd9Sstevel@tonic-gate * IRE_REFRELE/ire_refrele are the only caller of the function. It calls 35497c478bd9Sstevel@tonic-gate * to free the ire when the reference count goes to zero. 35507c478bd9Sstevel@tonic-gate */ 35517c478bd9Sstevel@tonic-gate void 35527c478bd9Sstevel@tonic-gate ire_inactive(ire_t *ire) 35537c478bd9Sstevel@tonic-gate { 35547c478bd9Sstevel@tonic-gate nce_t *nce; 35557c478bd9Sstevel@tonic-gate ill_t *ill = NULL; 35567c478bd9Sstevel@tonic-gate ill_t *stq_ill = NULL; 35577c478bd9Sstevel@tonic-gate ipif_t *ipif; 35587c478bd9Sstevel@tonic-gate boolean_t need_wakeup = B_FALSE; 3559c793af95Ssangeeta irb_t *irb; 3560f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 35617c478bd9Sstevel@tonic-gate 35627c478bd9Sstevel@tonic-gate ASSERT(ire->ire_refcnt == 0); 35637c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ptpn == NULL); 35647c478bd9Sstevel@tonic-gate ASSERT(ire->ire_next == NULL); 35657c478bd9Sstevel@tonic-gate 3566c793af95Ssangeeta if (ire->ire_gw_secattr != NULL) { 3567c793af95Ssangeeta ire_gw_secattr_free(ire->ire_gw_secattr); 3568c793af95Ssangeeta ire->ire_gw_secattr = NULL; 3569c793af95Ssangeeta } 3570c793af95Ssangeeta 3571c793af95Ssangeeta if (ire->ire_mp != NULL) { 3572c793af95Ssangeeta ASSERT(ire->ire_bucket == NULL); 3573c793af95Ssangeeta mutex_destroy(&ire->ire_lock); 3574f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_freed); 3575c793af95Ssangeeta if (ire->ire_nce != NULL) 3576c793af95Ssangeeta NCE_REFRELE_NOTR(ire->ire_nce); 3577c793af95Ssangeeta freeb(ire->ire_mp); 3578c793af95Ssangeeta return; 3579c793af95Ssangeeta } 3580c793af95Ssangeeta 35817c478bd9Sstevel@tonic-gate if ((nce = ire->ire_nce) != NULL) { 35827c478bd9Sstevel@tonic-gate NCE_REFRELE_NOTR(nce); 35837c478bd9Sstevel@tonic-gate ire->ire_nce = NULL; 35847c478bd9Sstevel@tonic-gate } 3585c793af95Ssangeeta 35867c478bd9Sstevel@tonic-gate if (ire->ire_ipif == NULL) 35877c478bd9Sstevel@tonic-gate goto end; 35887c478bd9Sstevel@tonic-gate 35897c478bd9Sstevel@tonic-gate ipif = ire->ire_ipif; 35907c478bd9Sstevel@tonic-gate ill = ipif->ipif_ill; 35917c478bd9Sstevel@tonic-gate 35927c478bd9Sstevel@tonic-gate if (ire->ire_bucket == NULL) { 35937c478bd9Sstevel@tonic-gate /* The ire was never inserted in the table. */ 35947c478bd9Sstevel@tonic-gate goto end; 35957c478bd9Sstevel@tonic-gate } 35967c478bd9Sstevel@tonic-gate 35977c478bd9Sstevel@tonic-gate /* 3598384ad179Ssowmini * ipif_ire_cnt on this ipif goes down by 1. If the ire_stq is 35995c0b7edeSseb * non-null ill_ire_count also goes down by 1. 36007c478bd9Sstevel@tonic-gate * 36017c478bd9Sstevel@tonic-gate * The ipif that is associated with an ire is ire->ire_ipif and 3602384ad179Ssowmini * hence when the ire->ire_ipif->ipif_ire_cnt drops to zero we call 36037c478bd9Sstevel@tonic-gate * ipif_ill_refrele_tail. Usually stq_ill is null or the same as 36047c478bd9Sstevel@tonic-gate * ire->ire_ipif->ipif_ill. So nothing more needs to be done. Only 36057c478bd9Sstevel@tonic-gate * in the case of IRE_CACHES when IPMP is used, stq_ill can be 36067c478bd9Sstevel@tonic-gate * different. If this is different from ire->ire_ipif->ipif_ill and 3607384ad179Ssowmini * if the ill_ire_cnt on the stq_ill also has dropped to zero, we call 36085c0b7edeSseb * ipif_ill_refrele_tail on the stq_ill. 36097c478bd9Sstevel@tonic-gate */ 36107c478bd9Sstevel@tonic-gate 36117c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 36127c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 36137c478bd9Sstevel@tonic-gate 36145c0b7edeSseb if (stq_ill == NULL || stq_ill == ill) { 36157c478bd9Sstevel@tonic-gate /* Optimize the most common case */ 36167c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 3617384ad179Ssowmini ASSERT(ipif->ipif_ire_cnt != 0); 3618968d2fd1Ssowmini DTRACE_PROBE3(ipif__decr__cnt, (ipif_t *), ipif, 3619968d2fd1Ssowmini (char *), "ire", (void *), ire); 3620384ad179Ssowmini ipif->ipif_ire_cnt--; 3621968d2fd1Ssowmini if (IPIF_DOWN_OK(ipif)) 36227c478bd9Sstevel@tonic-gate need_wakeup = B_TRUE; 36237c478bd9Sstevel@tonic-gate if (stq_ill != NULL) { 3624384ad179Ssowmini ASSERT(stq_ill->ill_ire_cnt != 0); 3625968d2fd1Ssowmini DTRACE_PROBE3(ill__decr__cnt, (ill_t *), stq_ill, 3626968d2fd1Ssowmini (char *), "ire", (void *), ire); 3627384ad179Ssowmini stq_ill->ill_ire_cnt--; 3628968d2fd1Ssowmini if (ILL_DOWN_OK(stq_ill)) 36297c478bd9Sstevel@tonic-gate need_wakeup = B_TRUE; 36307c478bd9Sstevel@tonic-gate } 36317c478bd9Sstevel@tonic-gate if (need_wakeup) { 36327c478bd9Sstevel@tonic-gate /* Drops the ill lock */ 36337c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(ill); 36347c478bd9Sstevel@tonic-gate } else { 36357c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 36367c478bd9Sstevel@tonic-gate } 36377c478bd9Sstevel@tonic-gate } else { 36387c478bd9Sstevel@tonic-gate /* 36397c478bd9Sstevel@tonic-gate * We can't grab all the ill locks at the same time. 36407c478bd9Sstevel@tonic-gate * It can lead to recursive lock enter in the call to 36417c478bd9Sstevel@tonic-gate * ipif_ill_refrele_tail and later. Instead do it 1 at 36427c478bd9Sstevel@tonic-gate * a time. 36437c478bd9Sstevel@tonic-gate */ 36447c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 3645384ad179Ssowmini ASSERT(ipif->ipif_ire_cnt != 0); 3646968d2fd1Ssowmini DTRACE_PROBE3(ipif__decr__cnt, (ipif_t *), ipif, 3647968d2fd1Ssowmini (char *), "ire", (void *), ire); 3648384ad179Ssowmini ipif->ipif_ire_cnt--; 3649968d2fd1Ssowmini if (IPIF_DOWN_OK(ipif)) { 36507c478bd9Sstevel@tonic-gate /* Drops the lock */ 36517c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(ill); 36527c478bd9Sstevel@tonic-gate } else { 36537c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 36547c478bd9Sstevel@tonic-gate } 36557c478bd9Sstevel@tonic-gate if (stq_ill != NULL) { 36567c478bd9Sstevel@tonic-gate mutex_enter(&stq_ill->ill_lock); 3657384ad179Ssowmini ASSERT(stq_ill->ill_ire_cnt != 0); 3658968d2fd1Ssowmini DTRACE_PROBE3(ill__decr__cnt, (ill_t *), stq_ill, 3659968d2fd1Ssowmini (char *), "ire", (void *), ire); 3660384ad179Ssowmini stq_ill->ill_ire_cnt--; 3661968d2fd1Ssowmini if (ILL_DOWN_OK(stq_ill)) { 36627c478bd9Sstevel@tonic-gate /* Drops the ill lock */ 36637c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(stq_ill); 36647c478bd9Sstevel@tonic-gate } else { 36657c478bd9Sstevel@tonic-gate mutex_exit(&stq_ill->ill_lock); 36667c478bd9Sstevel@tonic-gate } 36677c478bd9Sstevel@tonic-gate } 36687c478bd9Sstevel@tonic-gate } 36697c478bd9Sstevel@tonic-gate end: 36707c478bd9Sstevel@tonic-gate /* This should be true for both V4 and V6 */ 36717c478bd9Sstevel@tonic-gate 3672c793af95Ssangeeta if ((ire->ire_type & IRE_FORWARDTABLE) && 3673c793af95Ssangeeta (ire->ire_ipversion == IPV4_VERSION) && 3674c793af95Ssangeeta ((irb = ire->ire_bucket) != NULL)) { 3675c793af95Ssangeeta rw_enter(&irb->irb_lock, RW_WRITER); 3676c793af95Ssangeeta irb->irb_nire--; 3677c793af95Ssangeeta /* 3678c793af95Ssangeeta * Instead of examining the conditions for freeing 3679c793af95Ssangeeta * the radix node here, we do it by calling 3680c793af95Ssangeeta * IRB_REFRELE which is a single point in the code 3681c793af95Ssangeeta * that embeds that logic. Bump up the refcnt to 3682c793af95Ssangeeta * be able to call IRB_REFRELE 3683c793af95Ssangeeta */ 3684c793af95Ssangeeta IRB_REFHOLD_LOCKED(irb); 3685c793af95Ssangeeta rw_exit(&irb->irb_lock); 3686c793af95Ssangeeta IRB_REFRELE(irb); 3687c793af95Ssangeeta } 36887c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 36897c478bd9Sstevel@tonic-gate 36906a8288c7Scarlsonj #ifdef DEBUG 36916a8288c7Scarlsonj ire_trace_cleanup(ire); 36927c478bd9Sstevel@tonic-gate #endif 36937c478bd9Sstevel@tonic-gate mutex_destroy(&ire->ire_lock); 36947c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 3695f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v6, ire_stats_freed); 36967c478bd9Sstevel@tonic-gate } else { 3697f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_freed); 36987c478bd9Sstevel@tonic-gate } 3699c793af95Ssangeeta ASSERT(ire->ire_mp == NULL); 37007c478bd9Sstevel@tonic-gate /* Has been allocated out of the cache */ 37017c478bd9Sstevel@tonic-gate kmem_cache_free(ire_cache, ire); 37027c478bd9Sstevel@tonic-gate } 37037c478bd9Sstevel@tonic-gate 37047c478bd9Sstevel@tonic-gate /* 37056bdb8e66Sdd193516 * ire_walk routine to delete all IRE_CACHE/IRE_HOST types redirect 37066bdb8e66Sdd193516 * entries that have a given gateway address. 37077c478bd9Sstevel@tonic-gate */ 37087c478bd9Sstevel@tonic-gate void 37097c478bd9Sstevel@tonic-gate ire_delete_cache_gw(ire_t *ire, char *cp) 37107c478bd9Sstevel@tonic-gate { 37117c478bd9Sstevel@tonic-gate ipaddr_t gw_addr; 37127c478bd9Sstevel@tonic-gate 37136bdb8e66Sdd193516 if (!(ire->ire_type & IRE_CACHE) && 37146bdb8e66Sdd193516 !(ire->ire_flags & RTF_DYNAMIC)) 37157c478bd9Sstevel@tonic-gate return; 37167c478bd9Sstevel@tonic-gate 37177c478bd9Sstevel@tonic-gate bcopy(cp, &gw_addr, sizeof (gw_addr)); 37187c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == gw_addr) { 37197c478bd9Sstevel@tonic-gate ip1dbg(("ire_delete_cache_gw: deleted 0x%x type %d to 0x%x\n", 37207c478bd9Sstevel@tonic-gate (int)ntohl(ire->ire_addr), ire->ire_type, 37217c478bd9Sstevel@tonic-gate (int)ntohl(ire->ire_gateway_addr))); 37227c478bd9Sstevel@tonic-gate ire_delete(ire); 37237c478bd9Sstevel@tonic-gate } 37247c478bd9Sstevel@tonic-gate } 37257c478bd9Sstevel@tonic-gate 37267c478bd9Sstevel@tonic-gate /* 37277c478bd9Sstevel@tonic-gate * Remove all IRE_CACHE entries that match the ire specified. 37287c478bd9Sstevel@tonic-gate * 37297c478bd9Sstevel@tonic-gate * The flag argument indicates if the flush request is due to addition 37307c478bd9Sstevel@tonic-gate * of new route (IRE_FLUSH_ADD) or deletion of old route (IRE_FLUSH_DELETE). 37317c478bd9Sstevel@tonic-gate * 37327c478bd9Sstevel@tonic-gate * This routine takes only the IREs from the forwarding table and flushes 37337c478bd9Sstevel@tonic-gate * the corresponding entries from the cache table. 37347c478bd9Sstevel@tonic-gate * 37357c478bd9Sstevel@tonic-gate * When flushing due to the deletion of an old route, it 37367c478bd9Sstevel@tonic-gate * just checks the cache handles (ire_phandle and ire_ihandle) and 37377c478bd9Sstevel@tonic-gate * deletes the ones that match. 37387c478bd9Sstevel@tonic-gate * 37397c478bd9Sstevel@tonic-gate * When flushing due to the creation of a new route, it checks 37407c478bd9Sstevel@tonic-gate * if a cache entry's address matches the one in the IRE and 37417c478bd9Sstevel@tonic-gate * that the cache entry's parent has a less specific mask than the 37427c478bd9Sstevel@tonic-gate * one in IRE. The destination of such a cache entry could be the 37437c478bd9Sstevel@tonic-gate * gateway for other cache entries, so we need to flush those as 37447c478bd9Sstevel@tonic-gate * well by looking for gateway addresses matching the IRE's address. 37457c478bd9Sstevel@tonic-gate */ 37467c478bd9Sstevel@tonic-gate void 37477c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire_t *ire, int flag) 37487c478bd9Sstevel@tonic-gate { 37497c478bd9Sstevel@tonic-gate int i; 37507c478bd9Sstevel@tonic-gate ire_t *cire; 37517c478bd9Sstevel@tonic-gate irb_t *irb; 3752f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 37537c478bd9Sstevel@tonic-gate 37547c478bd9Sstevel@tonic-gate if (ire->ire_type & IRE_CACHE) 37557c478bd9Sstevel@tonic-gate return; 37567c478bd9Sstevel@tonic-gate 37577c478bd9Sstevel@tonic-gate /* 37587c478bd9Sstevel@tonic-gate * If a default is just created, there is no point 37597c478bd9Sstevel@tonic-gate * in going through the cache, as there will not be any 37607c478bd9Sstevel@tonic-gate * cached ires. 37617c478bd9Sstevel@tonic-gate */ 37627c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT && flag == IRE_FLUSH_ADD) 37637c478bd9Sstevel@tonic-gate return; 37647c478bd9Sstevel@tonic-gate if (flag == IRE_FLUSH_ADD) { 37657c478bd9Sstevel@tonic-gate /* 37667c478bd9Sstevel@tonic-gate * This selective flush is due to the addition of 37677c478bd9Sstevel@tonic-gate * new IRE. 37687c478bd9Sstevel@tonic-gate */ 3769f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 3770f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[i]; 37717c478bd9Sstevel@tonic-gate if ((cire = irb->irb_ire) == NULL) 37727c478bd9Sstevel@tonic-gate continue; 37737c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 37747c478bd9Sstevel@tonic-gate for (cire = irb->irb_ire; cire != NULL; 37757c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 37767c478bd9Sstevel@tonic-gate if (cire->ire_type != IRE_CACHE) 37777c478bd9Sstevel@tonic-gate continue; 37787c478bd9Sstevel@tonic-gate /* 37797c478bd9Sstevel@tonic-gate * If 'cire' belongs to the same subnet 37807c478bd9Sstevel@tonic-gate * as the new ire being added, and 'cire' 37817c478bd9Sstevel@tonic-gate * is derived from a prefix that is less 37827c478bd9Sstevel@tonic-gate * specific than the new ire being added, 37837c478bd9Sstevel@tonic-gate * we need to flush 'cire'; for instance, 37847c478bd9Sstevel@tonic-gate * when a new interface comes up. 37857c478bd9Sstevel@tonic-gate */ 37867c478bd9Sstevel@tonic-gate if (((cire->ire_addr & ire->ire_mask) == 37877c478bd9Sstevel@tonic-gate (ire->ire_addr & ire->ire_mask)) && 37887c478bd9Sstevel@tonic-gate (ip_mask_to_plen(cire->ire_cmask) <= 37897c478bd9Sstevel@tonic-gate ire->ire_masklen)) { 37907c478bd9Sstevel@tonic-gate ire_delete(cire); 37917c478bd9Sstevel@tonic-gate continue; 37927c478bd9Sstevel@tonic-gate } 37937c478bd9Sstevel@tonic-gate /* 37947c478bd9Sstevel@tonic-gate * This is the case when the ire_gateway_addr 37957c478bd9Sstevel@tonic-gate * of 'cire' belongs to the same subnet as 37967c478bd9Sstevel@tonic-gate * the new ire being added. 37977c478bd9Sstevel@tonic-gate * Flushing such ires is sometimes required to 37987c478bd9Sstevel@tonic-gate * avoid misrouting: say we have a machine with 37997c478bd9Sstevel@tonic-gate * two interfaces (I1 and I2), a default router 38007c478bd9Sstevel@tonic-gate * R on the I1 subnet, and a host route to an 38017c478bd9Sstevel@tonic-gate * off-link destination D with a gateway G on 38027c478bd9Sstevel@tonic-gate * the I2 subnet. 38037c478bd9Sstevel@tonic-gate * Under normal operation, we will have an 38047c478bd9Sstevel@tonic-gate * on-link cache entry for G and an off-link 38057c478bd9Sstevel@tonic-gate * cache entry for D with G as ire_gateway_addr, 38067c478bd9Sstevel@tonic-gate * traffic to D will reach its destination 38077c478bd9Sstevel@tonic-gate * through gateway G. 38087c478bd9Sstevel@tonic-gate * If the administrator does 'ifconfig I2 down', 38097c478bd9Sstevel@tonic-gate * the cache entries for D and G will be 38107c478bd9Sstevel@tonic-gate * flushed. However, G will now be resolved as 38117c478bd9Sstevel@tonic-gate * an off-link destination using R (the default 38127c478bd9Sstevel@tonic-gate * router) as gateway. Then D will also be 38137c478bd9Sstevel@tonic-gate * resolved as an off-link destination using G 38147c478bd9Sstevel@tonic-gate * as gateway - this behavior is due to 38157c478bd9Sstevel@tonic-gate * compatibility reasons, see comment in 38167c478bd9Sstevel@tonic-gate * ire_ihandle_lookup_offlink(). Traffic to D 38177c478bd9Sstevel@tonic-gate * will go to the router R and probably won't 38187c478bd9Sstevel@tonic-gate * reach the destination. 38197c478bd9Sstevel@tonic-gate * The administrator then does 'ifconfig I2 up'. 38207c478bd9Sstevel@tonic-gate * Since G is on the I2 subnet, this routine 38217c478bd9Sstevel@tonic-gate * will flush its cache entry. It must also 38227c478bd9Sstevel@tonic-gate * flush the cache entry for D, otherwise 38237c478bd9Sstevel@tonic-gate * traffic will stay misrouted until the IRE 38247c478bd9Sstevel@tonic-gate * times out. 38257c478bd9Sstevel@tonic-gate */ 38267c478bd9Sstevel@tonic-gate if ((cire->ire_gateway_addr & ire->ire_mask) == 38277c478bd9Sstevel@tonic-gate (ire->ire_addr & ire->ire_mask)) { 38287c478bd9Sstevel@tonic-gate ire_delete(cire); 38297c478bd9Sstevel@tonic-gate continue; 38307c478bd9Sstevel@tonic-gate } 38317c478bd9Sstevel@tonic-gate } 38327c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 38337c478bd9Sstevel@tonic-gate } 38347c478bd9Sstevel@tonic-gate } else { 38357c478bd9Sstevel@tonic-gate /* 38367c478bd9Sstevel@tonic-gate * delete the cache entries based on 38377c478bd9Sstevel@tonic-gate * handle in the IRE as this IRE is 38387c478bd9Sstevel@tonic-gate * being deleted/changed. 38397c478bd9Sstevel@tonic-gate */ 3840f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 3841f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[i]; 38427c478bd9Sstevel@tonic-gate if ((cire = irb->irb_ire) == NULL) 38437c478bd9Sstevel@tonic-gate continue; 38447c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 38457c478bd9Sstevel@tonic-gate for (cire = irb->irb_ire; cire != NULL; 38467c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 38477c478bd9Sstevel@tonic-gate if (cire->ire_type != IRE_CACHE) 38487c478bd9Sstevel@tonic-gate continue; 38497c478bd9Sstevel@tonic-gate if ((cire->ire_phandle == 0 || 38507c478bd9Sstevel@tonic-gate cire->ire_phandle != ire->ire_phandle) && 38517c478bd9Sstevel@tonic-gate (cire->ire_ihandle == 0 || 38527c478bd9Sstevel@tonic-gate cire->ire_ihandle != ire->ire_ihandle)) 38537c478bd9Sstevel@tonic-gate continue; 38547c478bd9Sstevel@tonic-gate ire_delete(cire); 38557c478bd9Sstevel@tonic-gate } 38567c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 38577c478bd9Sstevel@tonic-gate } 38587c478bd9Sstevel@tonic-gate } 38597c478bd9Sstevel@tonic-gate } 38607c478bd9Sstevel@tonic-gate 38617c478bd9Sstevel@tonic-gate /* 38627c478bd9Sstevel@tonic-gate * Matches the arguments passed with the values in the ire. 38637c478bd9Sstevel@tonic-gate * 38647c478bd9Sstevel@tonic-gate * Note: for match types that match using "ipif" passed in, ipif 38657c478bd9Sstevel@tonic-gate * must be checked for non-NULL before calling this routine. 38667c478bd9Sstevel@tonic-gate */ 3867c793af95Ssangeeta boolean_t 38687c478bd9Sstevel@tonic-gate ire_match_args(ire_t *ire, ipaddr_t addr, ipaddr_t mask, ipaddr_t gateway, 386945916cd2Sjpk int type, const ipif_t *ipif, zoneid_t zoneid, uint32_t ihandle, 387045916cd2Sjpk const ts_label_t *tsl, int match_flags) 38717c478bd9Sstevel@tonic-gate { 38727c478bd9Sstevel@tonic-gate ill_t *ire_ill = NULL, *dst_ill; 38737c478bd9Sstevel@tonic-gate ill_t *ipif_ill = NULL; 38747c478bd9Sstevel@tonic-gate ill_group_t *ire_ill_group = NULL; 38757c478bd9Sstevel@tonic-gate ill_group_t *ipif_ill_group = NULL; 38767c478bd9Sstevel@tonic-gate 38777c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 38787c478bd9Sstevel@tonic-gate ASSERT((ire->ire_addr & ~ire->ire_mask) == 0); 38797c478bd9Sstevel@tonic-gate ASSERT((!(match_flags & (MATCH_IRE_ILL|MATCH_IRE_ILL_GROUP))) || 38807c478bd9Sstevel@tonic-gate (ipif != NULL && !ipif->ipif_isv6)); 38817c478bd9Sstevel@tonic-gate 38827c478bd9Sstevel@tonic-gate /* 38837c478bd9Sstevel@tonic-gate * HIDDEN cache entries have to be looked up specifically with 38847c478bd9Sstevel@tonic-gate * MATCH_IRE_MARK_HIDDEN. MATCH_IRE_MARK_HIDDEN is usually set 38857c478bd9Sstevel@tonic-gate * when the interface is FAILED or INACTIVE. In that case, 38867c478bd9Sstevel@tonic-gate * any IRE_CACHES that exists should be marked with 38877c478bd9Sstevel@tonic-gate * IRE_MARK_HIDDEN. So, we don't really need to match below 38887c478bd9Sstevel@tonic-gate * for IRE_MARK_HIDDEN. But we do so for consistency. 38897c478bd9Sstevel@tonic-gate */ 38907c478bd9Sstevel@tonic-gate if (!(match_flags & MATCH_IRE_MARK_HIDDEN) && 38917c478bd9Sstevel@tonic-gate (ire->ire_marks & IRE_MARK_HIDDEN)) 38927c478bd9Sstevel@tonic-gate return (B_FALSE); 38937c478bd9Sstevel@tonic-gate 389443d18f1cSpriyanka /* 389543d18f1cSpriyanka * MATCH_IRE_MARK_PRIVATE_ADDR is set when IP_NEXTHOP option 389643d18f1cSpriyanka * is used. In that case the routing table is bypassed and the 389743d18f1cSpriyanka * packets are sent directly to the specified nexthop. The 389843d18f1cSpriyanka * IRE_CACHE entry representing this route should be marked 389943d18f1cSpriyanka * with IRE_MARK_PRIVATE_ADDR. 390043d18f1cSpriyanka */ 390143d18f1cSpriyanka 390243d18f1cSpriyanka if (!(match_flags & MATCH_IRE_MARK_PRIVATE_ADDR) && 390343d18f1cSpriyanka (ire->ire_marks & IRE_MARK_PRIVATE_ADDR)) 390443d18f1cSpriyanka return (B_FALSE); 390543d18f1cSpriyanka 390645916cd2Sjpk if (zoneid != ALL_ZONES && zoneid != ire->ire_zoneid && 390745916cd2Sjpk ire->ire_zoneid != ALL_ZONES) { 39087c478bd9Sstevel@tonic-gate /* 39097c478bd9Sstevel@tonic-gate * If MATCH_IRE_ZONEONLY has been set and the supplied zoneid is 39107c478bd9Sstevel@tonic-gate * valid and does not match that of ire_zoneid, a failure to 39117c478bd9Sstevel@tonic-gate * match is reported at this point. Otherwise, since some IREs 39127c478bd9Sstevel@tonic-gate * that are available in the global zone can be used in local 39137c478bd9Sstevel@tonic-gate * zones, additional checks need to be performed: 39147c478bd9Sstevel@tonic-gate * 39157c478bd9Sstevel@tonic-gate * IRE_BROADCAST, IRE_CACHE and IRE_LOOPBACK 39167c478bd9Sstevel@tonic-gate * entries should never be matched in this situation. 39177c478bd9Sstevel@tonic-gate * 39187c478bd9Sstevel@tonic-gate * IRE entries that have an interface associated with them 39197c478bd9Sstevel@tonic-gate * should in general not match unless they are an IRE_LOCAL 39207c478bd9Sstevel@tonic-gate * or in the case when MATCH_IRE_DEFAULT has been set in 39217c478bd9Sstevel@tonic-gate * the caller. In the case of the former, checking of the 39227c478bd9Sstevel@tonic-gate * other fields supplied should take place. 39237c478bd9Sstevel@tonic-gate * 39247c478bd9Sstevel@tonic-gate * In the case where MATCH_IRE_DEFAULT has been set, 39257c478bd9Sstevel@tonic-gate * all of the ipif's associated with the IRE's ill are 39267c478bd9Sstevel@tonic-gate * checked to see if there is a matching zoneid. If any 39277c478bd9Sstevel@tonic-gate * one ipif has a matching zoneid, this IRE is a 39287c478bd9Sstevel@tonic-gate * potential candidate so checking of the other fields 39297c478bd9Sstevel@tonic-gate * takes place. 39307c478bd9Sstevel@tonic-gate * 39317c478bd9Sstevel@tonic-gate * In the case where the IRE_INTERFACE has a usable source 39327c478bd9Sstevel@tonic-gate * address (indicated by ill_usesrc_ifindex) in the 39337c478bd9Sstevel@tonic-gate * correct zone then it's permitted to return this IRE 39347c478bd9Sstevel@tonic-gate */ 39357c478bd9Sstevel@tonic-gate if (match_flags & MATCH_IRE_ZONEONLY) 39367c478bd9Sstevel@tonic-gate return (B_FALSE); 39377c478bd9Sstevel@tonic-gate if (ire->ire_type & (IRE_BROADCAST | IRE_CACHE | IRE_LOOPBACK)) 39387c478bd9Sstevel@tonic-gate return (B_FALSE); 39397c478bd9Sstevel@tonic-gate /* 39407c478bd9Sstevel@tonic-gate * Note, IRE_INTERFACE can have the stq as NULL. For 39417c478bd9Sstevel@tonic-gate * example, if the default multicast route is tied to 39427c478bd9Sstevel@tonic-gate * the loopback address. 39437c478bd9Sstevel@tonic-gate */ 39447c478bd9Sstevel@tonic-gate if ((ire->ire_type & IRE_INTERFACE) && 39457c478bd9Sstevel@tonic-gate (ire->ire_stq != NULL)) { 39467c478bd9Sstevel@tonic-gate dst_ill = (ill_t *)ire->ire_stq->q_ptr; 39477c478bd9Sstevel@tonic-gate /* 39487c478bd9Sstevel@tonic-gate * If there is a usable source address in the 39497c478bd9Sstevel@tonic-gate * zone, then it's ok to return an 39507c478bd9Sstevel@tonic-gate * IRE_INTERFACE 39517c478bd9Sstevel@tonic-gate */ 39527c478bd9Sstevel@tonic-gate if (ipif_usesrc_avail(dst_ill, zoneid)) { 39537c478bd9Sstevel@tonic-gate ip3dbg(("ire_match_args: dst_ill %p match %d\n", 39547c478bd9Sstevel@tonic-gate (void *)dst_ill, 39557c478bd9Sstevel@tonic-gate (ire->ire_addr == (addr & mask)))); 39567c478bd9Sstevel@tonic-gate } else { 39577c478bd9Sstevel@tonic-gate ip3dbg(("ire_match_args: src_ipif NULL" 39587c478bd9Sstevel@tonic-gate " dst_ill %p\n", (void *)dst_ill)); 39597c478bd9Sstevel@tonic-gate return (B_FALSE); 39607c478bd9Sstevel@tonic-gate } 39617c478bd9Sstevel@tonic-gate } 39627c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL && ire->ire_type != IRE_LOCAL && 39637c478bd9Sstevel@tonic-gate !(ire->ire_type & IRE_INTERFACE)) { 39647c478bd9Sstevel@tonic-gate ipif_t *tipif; 39657c478bd9Sstevel@tonic-gate 39667c478bd9Sstevel@tonic-gate if ((match_flags & MATCH_IRE_DEFAULT) == 0) { 39677c478bd9Sstevel@tonic-gate return (B_FALSE); 39687c478bd9Sstevel@tonic-gate } 39697c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_ipif->ipif_ill->ill_lock); 39707c478bd9Sstevel@tonic-gate for (tipif = ire->ire_ipif->ipif_ill->ill_ipif; 39717c478bd9Sstevel@tonic-gate tipif != NULL; tipif = tipif->ipif_next) { 39727c478bd9Sstevel@tonic-gate if (IPIF_CAN_LOOKUP(tipif) && 39737c478bd9Sstevel@tonic-gate (tipif->ipif_flags & IPIF_UP) && 397445916cd2Sjpk (tipif->ipif_zoneid == zoneid || 397545916cd2Sjpk tipif->ipif_zoneid == ALL_ZONES)) 39767c478bd9Sstevel@tonic-gate break; 39777c478bd9Sstevel@tonic-gate } 39787c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_ipif->ipif_ill->ill_lock); 39797c478bd9Sstevel@tonic-gate if (tipif == NULL) { 39807c478bd9Sstevel@tonic-gate return (B_FALSE); 39817c478bd9Sstevel@tonic-gate } 39827c478bd9Sstevel@tonic-gate } 39837c478bd9Sstevel@tonic-gate } 39847c478bd9Sstevel@tonic-gate 39857c478bd9Sstevel@tonic-gate /* 39867c478bd9Sstevel@tonic-gate * For IRE_CACHES, MATCH_IRE_ILL/ILL_GROUP really means that 39877c478bd9Sstevel@tonic-gate * somebody wants to send out on a particular interface which 39887c478bd9Sstevel@tonic-gate * is given by ire_stq and hence use ire_stq to derive the ill 39897c478bd9Sstevel@tonic-gate * value. ire_ipif for IRE_CACHES is just the means of getting 39907c478bd9Sstevel@tonic-gate * a source address i.e ire_src_addr = ire->ire_ipif->ipif_src_addr. 39917c478bd9Sstevel@tonic-gate * ire_to_ill does the right thing for this. 39927c478bd9Sstevel@tonic-gate */ 39937c478bd9Sstevel@tonic-gate if (match_flags & (MATCH_IRE_ILL|MATCH_IRE_ILL_GROUP)) { 39947c478bd9Sstevel@tonic-gate ire_ill = ire_to_ill(ire); 39957c478bd9Sstevel@tonic-gate if (ire_ill != NULL) 39967c478bd9Sstevel@tonic-gate ire_ill_group = ire_ill->ill_group; 39977c478bd9Sstevel@tonic-gate ipif_ill = ipif->ipif_ill; 39987c478bd9Sstevel@tonic-gate ipif_ill_group = ipif_ill->ill_group; 39997c478bd9Sstevel@tonic-gate } 40007c478bd9Sstevel@tonic-gate 40017c478bd9Sstevel@tonic-gate if ((ire->ire_addr == (addr & mask)) && 40027c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_GW)) || 40037c478bd9Sstevel@tonic-gate (ire->ire_gateway_addr == gateway)) && 40047c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_TYPE)) || 40057c478bd9Sstevel@tonic-gate (ire->ire_type & type)) && 40067c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_SRC)) || 40077c478bd9Sstevel@tonic-gate (ire->ire_src_addr == ipif->ipif_src_addr)) && 40087c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_IPIF)) || 40097c478bd9Sstevel@tonic-gate (ire->ire_ipif == ipif)) && 40107c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_MARK_HIDDEN)) || 40117c478bd9Sstevel@tonic-gate (ire->ire_type != IRE_CACHE || 40127c478bd9Sstevel@tonic-gate ire->ire_marks & IRE_MARK_HIDDEN)) && 401343d18f1cSpriyanka ((!(match_flags & MATCH_IRE_MARK_PRIVATE_ADDR)) || 401443d18f1cSpriyanka (ire->ire_type != IRE_CACHE || 401543d18f1cSpriyanka ire->ire_marks & IRE_MARK_PRIVATE_ADDR)) && 40167c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL)) || 40177c478bd9Sstevel@tonic-gate (ire_ill == ipif_ill)) && 40187c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_IHANDLE)) || 40197c478bd9Sstevel@tonic-gate (ire->ire_ihandle == ihandle)) && 4020c793af95Ssangeeta ((!(match_flags & MATCH_IRE_MASK)) || 4021c793af95Ssangeeta (ire->ire_mask == mask)) && 40227c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL_GROUP)) || 40237c478bd9Sstevel@tonic-gate (ire_ill == ipif_ill) || 40247c478bd9Sstevel@tonic-gate (ire_ill_group != NULL && 402545916cd2Sjpk ire_ill_group == ipif_ill_group)) && 402645916cd2Sjpk ((!(match_flags & MATCH_IRE_SECATTR)) || 402745916cd2Sjpk (!is_system_labeled()) || 402845916cd2Sjpk (tsol_ire_match_gwattr(ire, tsl) == 0))) { 40297c478bd9Sstevel@tonic-gate /* We found the matched IRE */ 40307c478bd9Sstevel@tonic-gate return (B_TRUE); 40317c478bd9Sstevel@tonic-gate } 40327c478bd9Sstevel@tonic-gate return (B_FALSE); 40337c478bd9Sstevel@tonic-gate } 40347c478bd9Sstevel@tonic-gate 40357c478bd9Sstevel@tonic-gate 40367c478bd9Sstevel@tonic-gate /* 40377c478bd9Sstevel@tonic-gate * Lookup for a route in all the tables 40387c478bd9Sstevel@tonic-gate */ 40397c478bd9Sstevel@tonic-gate ire_t * 40407c478bd9Sstevel@tonic-gate ire_route_lookup(ipaddr_t addr, ipaddr_t mask, ipaddr_t gateway, 404145916cd2Sjpk int type, const ipif_t *ipif, ire_t **pire, zoneid_t zoneid, 4042f4b3ec61Sdh155122 const ts_label_t *tsl, int flags, ip_stack_t *ipst) 40437c478bd9Sstevel@tonic-gate { 40447c478bd9Sstevel@tonic-gate ire_t *ire = NULL; 40457c478bd9Sstevel@tonic-gate 40467c478bd9Sstevel@tonic-gate /* 40477c478bd9Sstevel@tonic-gate * ire_match_args() will dereference ipif MATCH_IRE_SRC or 40487c478bd9Sstevel@tonic-gate * MATCH_IRE_ILL is set. 40497c478bd9Sstevel@tonic-gate */ 40507c478bd9Sstevel@tonic-gate if ((flags & (MATCH_IRE_SRC | MATCH_IRE_ILL | MATCH_IRE_ILL_GROUP)) && 40517c478bd9Sstevel@tonic-gate (ipif == NULL)) 40527c478bd9Sstevel@tonic-gate return (NULL); 40537c478bd9Sstevel@tonic-gate 40547c478bd9Sstevel@tonic-gate /* 40557c478bd9Sstevel@tonic-gate * might be asking for a cache lookup, 40567c478bd9Sstevel@tonic-gate * This is not best way to lookup cache, 40577c478bd9Sstevel@tonic-gate * user should call ire_cache_lookup directly. 40587c478bd9Sstevel@tonic-gate * 40597c478bd9Sstevel@tonic-gate * If MATCH_IRE_TYPE was set, first lookup in the cache table and then 40607c478bd9Sstevel@tonic-gate * in the forwarding table, if the applicable type flags were set. 40617c478bd9Sstevel@tonic-gate */ 40627c478bd9Sstevel@tonic-gate if ((flags & MATCH_IRE_TYPE) == 0 || (type & IRE_CACHETABLE) != 0) { 40637c478bd9Sstevel@tonic-gate ire = ire_ctable_lookup(addr, gateway, type, ipif, zoneid, 4064f4b3ec61Sdh155122 tsl, flags, ipst); 40657c478bd9Sstevel@tonic-gate if (ire != NULL) 40667c478bd9Sstevel@tonic-gate return (ire); 40677c478bd9Sstevel@tonic-gate } 40687c478bd9Sstevel@tonic-gate if ((flags & MATCH_IRE_TYPE) == 0 || (type & IRE_FORWARDTABLE) != 0) { 40697c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(addr, mask, gateway, type, ipif, pire, 4070f4b3ec61Sdh155122 zoneid, 0, tsl, flags, ipst); 40717c478bd9Sstevel@tonic-gate } 40727c478bd9Sstevel@tonic-gate return (ire); 40737c478bd9Sstevel@tonic-gate } 40747c478bd9Sstevel@tonic-gate 40757c478bd9Sstevel@tonic-gate 40767c478bd9Sstevel@tonic-gate /* 407745916cd2Sjpk * Delete the IRE cache for the gateway and all IRE caches whose 407845916cd2Sjpk * ire_gateway_addr points to this gateway, and allow them to 407945916cd2Sjpk * be created on demand by ip_newroute. 408045916cd2Sjpk */ 408145916cd2Sjpk void 4082f4b3ec61Sdh155122 ire_clookup_delete_cache_gw(ipaddr_t addr, zoneid_t zoneid, ip_stack_t *ipst) 408345916cd2Sjpk { 408445916cd2Sjpk irb_t *irb; 408545916cd2Sjpk ire_t *ire; 408645916cd2Sjpk 4087f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[IRE_ADDR_HASH(addr, 4088f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size)]; 408945916cd2Sjpk IRB_REFHOLD(irb); 409045916cd2Sjpk for (ire = irb->irb_ire; ire != NULL; ire = ire->ire_next) { 409145916cd2Sjpk if (ire->ire_marks & IRE_MARK_CONDEMNED) 409245916cd2Sjpk continue; 409345916cd2Sjpk 409445916cd2Sjpk ASSERT(ire->ire_mask == IP_HOST_MASK); 409545916cd2Sjpk if (ire_match_args(ire, addr, ire->ire_mask, 0, IRE_CACHE, 409645916cd2Sjpk NULL, zoneid, 0, NULL, MATCH_IRE_TYPE)) { 409745916cd2Sjpk ire_delete(ire); 409845916cd2Sjpk } 409945916cd2Sjpk } 410045916cd2Sjpk IRB_REFRELE(irb); 410145916cd2Sjpk 4102f4b3ec61Sdh155122 ire_walk_v4(ire_delete_cache_gw, &addr, zoneid, ipst); 410345916cd2Sjpk } 410445916cd2Sjpk 410545916cd2Sjpk /* 41067c478bd9Sstevel@tonic-gate * Looks up cache table for a route. 41077c478bd9Sstevel@tonic-gate * specific lookup can be indicated by 41087c478bd9Sstevel@tonic-gate * passing the MATCH_* flags and the 41097c478bd9Sstevel@tonic-gate * necessary parameters. 41107c478bd9Sstevel@tonic-gate */ 41117c478bd9Sstevel@tonic-gate ire_t * 411245916cd2Sjpk ire_ctable_lookup(ipaddr_t addr, ipaddr_t gateway, int type, const ipif_t *ipif, 4113f4b3ec61Sdh155122 zoneid_t zoneid, const ts_label_t *tsl, int flags, ip_stack_t *ipst) 41147c478bd9Sstevel@tonic-gate { 41157c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 41167c478bd9Sstevel@tonic-gate ire_t *ire; 41177c478bd9Sstevel@tonic-gate 41187c478bd9Sstevel@tonic-gate /* 41197c478bd9Sstevel@tonic-gate * ire_match_args() will dereference ipif MATCH_IRE_SRC or 41207c478bd9Sstevel@tonic-gate * MATCH_IRE_ILL is set. 41217c478bd9Sstevel@tonic-gate */ 41227c478bd9Sstevel@tonic-gate if ((flags & (MATCH_IRE_SRC | MATCH_IRE_ILL | MATCH_IRE_ILL_GROUP)) && 41237c478bd9Sstevel@tonic-gate (ipif == NULL)) 41247c478bd9Sstevel@tonic-gate return (NULL); 41257c478bd9Sstevel@tonic-gate 4126f4b3ec61Sdh155122 irb_ptr = &ipst->ips_ip_cache_table[IRE_ADDR_HASH(addr, 4127f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size)]; 41287c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_READER); 41297c478bd9Sstevel@tonic-gate for (ire = irb_ptr->irb_ire; ire != NULL; ire = ire->ire_next) { 41307c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) 41317c478bd9Sstevel@tonic-gate continue; 41327c478bd9Sstevel@tonic-gate ASSERT(ire->ire_mask == IP_HOST_MASK); 41337c478bd9Sstevel@tonic-gate if (ire_match_args(ire, addr, ire->ire_mask, gateway, type, 413445916cd2Sjpk ipif, zoneid, 0, tsl, flags)) { 41357c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 41367c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 41377c478bd9Sstevel@tonic-gate return (ire); 41387c478bd9Sstevel@tonic-gate } 41397c478bd9Sstevel@tonic-gate } 41407c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 41417c478bd9Sstevel@tonic-gate return (NULL); 41427c478bd9Sstevel@tonic-gate } 41437c478bd9Sstevel@tonic-gate 41447c478bd9Sstevel@tonic-gate /* 41455597b60aSnordmark * Check whether the IRE_LOCAL and the IRE potentially used to transmit 41465597b60aSnordmark * (could be an IRE_CACHE, IRE_BROADCAST, or IRE_INTERFACE) are part of 41475597b60aSnordmark * the same ill group. 41485597b60aSnordmark */ 41495597b60aSnordmark boolean_t 41505597b60aSnordmark ire_local_same_ill_group(ire_t *ire_local, ire_t *xmit_ire) 41515597b60aSnordmark { 41525597b60aSnordmark ill_t *recv_ill, *xmit_ill; 41535597b60aSnordmark ill_group_t *recv_group, *xmit_group; 41545597b60aSnordmark 41552bc4236aSnordmark ASSERT(ire_local->ire_type & (IRE_LOCAL|IRE_LOOPBACK)); 4156aa2eaee6Snordmark ASSERT(xmit_ire->ire_type & (IRE_CACHETABLE|IRE_INTERFACE)); 41575597b60aSnordmark 41582bc4236aSnordmark recv_ill = ire_to_ill(ire_local); 41592bc4236aSnordmark xmit_ill = ire_to_ill(xmit_ire); 41602bc4236aSnordmark 41612bc4236aSnordmark ASSERT(recv_ill != NULL); 41622bc4236aSnordmark ASSERT(xmit_ill != NULL); 41635597b60aSnordmark 41645597b60aSnordmark if (recv_ill == xmit_ill) 41655597b60aSnordmark return (B_TRUE); 41665597b60aSnordmark 41675597b60aSnordmark recv_group = recv_ill->ill_group; 41685597b60aSnordmark xmit_group = xmit_ill->ill_group; 41695597b60aSnordmark 41705597b60aSnordmark if (recv_group != NULL && recv_group == xmit_group) 41715597b60aSnordmark return (B_TRUE); 41725597b60aSnordmark 41735597b60aSnordmark return (B_FALSE); 41745597b60aSnordmark } 41755597b60aSnordmark 41765597b60aSnordmark /* 41775597b60aSnordmark * Check if the IRE_LOCAL uses the same ill (group) as another route would use. 4178aa2eaee6Snordmark * If there is no alternate route, or the alternate is a REJECT or BLACKHOLE, 4179aa2eaee6Snordmark * then we don't allow this IRE_LOCAL to be used. 41805597b60aSnordmark */ 41815597b60aSnordmark boolean_t 41825597b60aSnordmark ire_local_ok_across_zones(ire_t *ire_local, zoneid_t zoneid, void *addr, 4183f4b3ec61Sdh155122 const ts_label_t *tsl, ip_stack_t *ipst) 41845597b60aSnordmark { 41855597b60aSnordmark ire_t *alt_ire; 41865597b60aSnordmark boolean_t rval; 41875597b60aSnordmark 41885597b60aSnordmark if (ire_local->ire_ipversion == IPV4_VERSION) { 41895597b60aSnordmark alt_ire = ire_ftable_lookup(*((ipaddr_t *)addr), 0, 0, 0, NULL, 41905597b60aSnordmark NULL, zoneid, 0, tsl, 41915597b60aSnordmark MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT | 4192f4b3ec61Sdh155122 MATCH_IRE_RJ_BHOLE, ipst); 41935597b60aSnordmark } else { 41945597b60aSnordmark alt_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL, 41955597b60aSnordmark 0, NULL, NULL, zoneid, 0, tsl, 41965597b60aSnordmark MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT | 4197f4b3ec61Sdh155122 MATCH_IRE_RJ_BHOLE, ipst); 41985597b60aSnordmark } 41995597b60aSnordmark 42005597b60aSnordmark if (alt_ire == NULL) 42015597b60aSnordmark return (B_FALSE); 42025597b60aSnordmark 4203aa2eaee6Snordmark if (alt_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 4204aa2eaee6Snordmark ire_refrele(alt_ire); 4205aa2eaee6Snordmark return (B_FALSE); 4206aa2eaee6Snordmark } 42075597b60aSnordmark rval = ire_local_same_ill_group(ire_local, alt_ire); 42085597b60aSnordmark 42095597b60aSnordmark ire_refrele(alt_ire); 42105597b60aSnordmark return (rval); 42115597b60aSnordmark } 42125597b60aSnordmark 42135597b60aSnordmark /* 42147c478bd9Sstevel@tonic-gate * Lookup cache. Don't return IRE_MARK_HIDDEN entries. Callers 42157c478bd9Sstevel@tonic-gate * should use ire_ctable_lookup with MATCH_IRE_MARK_HIDDEN to get 42167c478bd9Sstevel@tonic-gate * to the hidden ones. 42175597b60aSnordmark * 42185597b60aSnordmark * In general the zoneid has to match (where ALL_ZONES match all of them). 42195597b60aSnordmark * But for IRE_LOCAL we also need to handle the case where L2 should 42205597b60aSnordmark * conceptually loop back the packet. This is necessary since neither 42215597b60aSnordmark * Ethernet drivers nor Ethernet hardware loops back packets sent to their 42225597b60aSnordmark * own MAC address. This loopback is needed when the normal 42235597b60aSnordmark * routes (ignoring IREs with different zoneids) would send out the packet on 42245597b60aSnordmark * the same ill (or ill group) as the ill with which this IRE_LOCAL is 42255597b60aSnordmark * associated. 42265597b60aSnordmark * 42275597b60aSnordmark * Earlier versions of this code always matched an IRE_LOCAL independently of 42285597b60aSnordmark * the zoneid. We preserve that earlier behavior when 42295597b60aSnordmark * ip_restrict_interzone_loopback is turned off. 42307c478bd9Sstevel@tonic-gate */ 42317c478bd9Sstevel@tonic-gate ire_t * 4232f4b3ec61Sdh155122 ire_cache_lookup(ipaddr_t addr, zoneid_t zoneid, const ts_label_t *tsl, 4233f4b3ec61Sdh155122 ip_stack_t *ipst) 42347c478bd9Sstevel@tonic-gate { 42357c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 42367c478bd9Sstevel@tonic-gate ire_t *ire; 42377c478bd9Sstevel@tonic-gate 4238f4b3ec61Sdh155122 irb_ptr = &ipst->ips_ip_cache_table[IRE_ADDR_HASH(addr, 4239f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size)]; 42407c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_READER); 42417c478bd9Sstevel@tonic-gate for (ire = irb_ptr->irb_ire; ire != NULL; ire = ire->ire_next) { 424243d18f1cSpriyanka if (ire->ire_marks & (IRE_MARK_CONDEMNED | 424343d18f1cSpriyanka IRE_MARK_HIDDEN | IRE_MARK_PRIVATE_ADDR)) { 42447c478bd9Sstevel@tonic-gate continue; 424543d18f1cSpriyanka } 42467c478bd9Sstevel@tonic-gate if (ire->ire_addr == addr) { 424745916cd2Sjpk /* 424845916cd2Sjpk * Finally, check if the security policy has any 424945916cd2Sjpk * restriction on using this route for the specified 425045916cd2Sjpk * message. 425145916cd2Sjpk */ 425245916cd2Sjpk if (tsl != NULL && 425345916cd2Sjpk ire->ire_gw_secattr != NULL && 425445916cd2Sjpk tsol_ire_match_gwattr(ire, tsl) != 0) { 425545916cd2Sjpk continue; 425645916cd2Sjpk } 425745916cd2Sjpk 42587c478bd9Sstevel@tonic-gate if (zoneid == ALL_ZONES || ire->ire_zoneid == zoneid || 42595597b60aSnordmark ire->ire_zoneid == ALL_ZONES) { 42605597b60aSnordmark IRE_REFHOLD(ire); 42615597b60aSnordmark rw_exit(&irb_ptr->irb_lock); 42625597b60aSnordmark return (ire); 42635597b60aSnordmark } 42645597b60aSnordmark 42655597b60aSnordmark if (ire->ire_type == IRE_LOCAL) { 4266f4b3ec61Sdh155122 if (ipst->ips_ip_restrict_interzone_loopback && 42675597b60aSnordmark !ire_local_ok_across_zones(ire, zoneid, 4268f4b3ec61Sdh155122 &addr, tsl, ipst)) 42695597b60aSnordmark continue; 42705597b60aSnordmark 42717c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 42727c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 42737c478bd9Sstevel@tonic-gate return (ire); 42747c478bd9Sstevel@tonic-gate } 42757c478bd9Sstevel@tonic-gate } 42767c478bd9Sstevel@tonic-gate } 42777c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 42787c478bd9Sstevel@tonic-gate return (NULL); 42797c478bd9Sstevel@tonic-gate } 42807c478bd9Sstevel@tonic-gate 42817c478bd9Sstevel@tonic-gate /* 42827c478bd9Sstevel@tonic-gate * Locate the interface ire that is tied to the cache ire 'cire' via 42837c478bd9Sstevel@tonic-gate * cire->ire_ihandle. 42847c478bd9Sstevel@tonic-gate * 42857c478bd9Sstevel@tonic-gate * We are trying to create the cache ire for an offlink destn based 42867c478bd9Sstevel@tonic-gate * on the cache ire of the gateway in 'cire'. 'pire' is the prefix ire 42877c478bd9Sstevel@tonic-gate * as found by ip_newroute(). We are called from ip_newroute() in 42887c478bd9Sstevel@tonic-gate * the IRE_CACHE case. 42897c478bd9Sstevel@tonic-gate */ 42907c478bd9Sstevel@tonic-gate ire_t * 42917c478bd9Sstevel@tonic-gate ire_ihandle_lookup_offlink(ire_t *cire, ire_t *pire) 42927c478bd9Sstevel@tonic-gate { 42937c478bd9Sstevel@tonic-gate ire_t *ire; 42947c478bd9Sstevel@tonic-gate int match_flags; 42957c478bd9Sstevel@tonic-gate ipaddr_t gw_addr; 42967c478bd9Sstevel@tonic-gate ipif_t *gw_ipif; 4297f4b3ec61Sdh155122 ip_stack_t *ipst = cire->ire_ipst; 42987c478bd9Sstevel@tonic-gate 42997c478bd9Sstevel@tonic-gate ASSERT(cire != NULL && pire != NULL); 43007c478bd9Sstevel@tonic-gate 43017c478bd9Sstevel@tonic-gate /* 43027c478bd9Sstevel@tonic-gate * We don't need to specify the zoneid to ire_ftable_lookup() below 43037c478bd9Sstevel@tonic-gate * because the ihandle refers to an ipif which can be in only one zone. 43047c478bd9Sstevel@tonic-gate */ 43057c478bd9Sstevel@tonic-gate match_flags = MATCH_IRE_TYPE | MATCH_IRE_IHANDLE | MATCH_IRE_MASK; 43067c478bd9Sstevel@tonic-gate /* 43077c478bd9Sstevel@tonic-gate * ip_newroute calls ire_ftable_lookup with MATCH_IRE_ILL only 43087c478bd9Sstevel@tonic-gate * for on-link hosts. We should never be here for onlink. 43097c478bd9Sstevel@tonic-gate * Thus, use MATCH_IRE_ILL_GROUP. 43107c478bd9Sstevel@tonic-gate */ 43117c478bd9Sstevel@tonic-gate if (pire->ire_ipif != NULL) 43127c478bd9Sstevel@tonic-gate match_flags |= MATCH_IRE_ILL_GROUP; 43137c478bd9Sstevel@tonic-gate /* 43147c478bd9Sstevel@tonic-gate * We know that the mask of the interface ire equals cire->ire_cmask. 43157c478bd9Sstevel@tonic-gate * (When ip_newroute() created 'cire' for the gateway it set its 43167c478bd9Sstevel@tonic-gate * cmask from the interface ire's mask) 43177c478bd9Sstevel@tonic-gate */ 43187c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(cire->ire_addr, cire->ire_cmask, 0, 43197c478bd9Sstevel@tonic-gate IRE_INTERFACE, pire->ire_ipif, NULL, ALL_ZONES, cire->ire_ihandle, 4320f4b3ec61Sdh155122 NULL, match_flags, ipst); 43217c478bd9Sstevel@tonic-gate if (ire != NULL) 43227c478bd9Sstevel@tonic-gate return (ire); 43237c478bd9Sstevel@tonic-gate /* 43247c478bd9Sstevel@tonic-gate * If we didn't find an interface ire above, we can't declare failure. 43257c478bd9Sstevel@tonic-gate * For backwards compatibility, we need to support prefix routes 43267c478bd9Sstevel@tonic-gate * pointing to next hop gateways that are not on-link. 43277c478bd9Sstevel@tonic-gate * 43287c478bd9Sstevel@tonic-gate * Assume we are trying to ping some offlink destn, and we have the 43297c478bd9Sstevel@tonic-gate * routing table below. 43307c478bd9Sstevel@tonic-gate * 43317c478bd9Sstevel@tonic-gate * Eg. default - gw1 <--- pire (line 1) 43327c478bd9Sstevel@tonic-gate * gw1 - gw2 (line 2) 43337c478bd9Sstevel@tonic-gate * gw2 - hme0 (line 3) 43347c478bd9Sstevel@tonic-gate * 43357c478bd9Sstevel@tonic-gate * If we already have a cache ire for gw1 in 'cire', the 43367c478bd9Sstevel@tonic-gate * ire_ftable_lookup above would have failed, since there is no 43377c478bd9Sstevel@tonic-gate * interface ire to reach gw1. We will fallthru below. 43387c478bd9Sstevel@tonic-gate * 43397c478bd9Sstevel@tonic-gate * Here we duplicate the steps that ire_ftable_lookup() did in 43407c478bd9Sstevel@tonic-gate * getting 'cire' from 'pire', in the MATCH_IRE_RECURSIVE case. 43417c478bd9Sstevel@tonic-gate * The differences are the following 43427c478bd9Sstevel@tonic-gate * i. We want the interface ire only, so we call ire_ftable_lookup() 43437c478bd9Sstevel@tonic-gate * instead of ire_route_lookup() 43447c478bd9Sstevel@tonic-gate * ii. We look for only prefix routes in the 1st call below. 43457c478bd9Sstevel@tonic-gate * ii. We want to match on the ihandle in the 2nd call below. 43467c478bd9Sstevel@tonic-gate */ 43477c478bd9Sstevel@tonic-gate match_flags = MATCH_IRE_TYPE; 43487c478bd9Sstevel@tonic-gate if (pire->ire_ipif != NULL) 43497c478bd9Sstevel@tonic-gate match_flags |= MATCH_IRE_ILL_GROUP; 43507c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(pire->ire_gateway_addr, 0, 0, IRE_OFFSUBNET, 4351f4b3ec61Sdh155122 pire->ire_ipif, NULL, ALL_ZONES, 0, NULL, match_flags, ipst); 43527c478bd9Sstevel@tonic-gate if (ire == NULL) 43537c478bd9Sstevel@tonic-gate return (NULL); 43547c478bd9Sstevel@tonic-gate /* 43557c478bd9Sstevel@tonic-gate * At this point 'ire' corresponds to the entry shown in line 2. 43567c478bd9Sstevel@tonic-gate * gw_addr is 'gw2' in the example above. 43577c478bd9Sstevel@tonic-gate */ 43587c478bd9Sstevel@tonic-gate gw_addr = ire->ire_gateway_addr; 43597c478bd9Sstevel@tonic-gate gw_ipif = ire->ire_ipif; 43607c478bd9Sstevel@tonic-gate ire_refrele(ire); 43617c478bd9Sstevel@tonic-gate 43627c478bd9Sstevel@tonic-gate match_flags |= MATCH_IRE_IHANDLE; 43637c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(gw_addr, 0, 0, IRE_INTERFACE, 4364f4b3ec61Sdh155122 gw_ipif, NULL, ALL_ZONES, cire->ire_ihandle, NULL, match_flags, 4365f4b3ec61Sdh155122 ipst); 43667c478bd9Sstevel@tonic-gate return (ire); 43677c478bd9Sstevel@tonic-gate } 43687c478bd9Sstevel@tonic-gate 43697c478bd9Sstevel@tonic-gate /* 43707c478bd9Sstevel@tonic-gate * Return the IRE_LOOPBACK, IRE_IF_RESOLVER or IRE_IF_NORESOLVER 43717c478bd9Sstevel@tonic-gate * ire associated with the specified ipif. 43727c478bd9Sstevel@tonic-gate * 43737c478bd9Sstevel@tonic-gate * This might occasionally be called when IPIF_UP is not set since 43747c478bd9Sstevel@tonic-gate * the IP_MULTICAST_IF as well as creating interface routes 43757c478bd9Sstevel@tonic-gate * allows specifying a down ipif (ipif_lookup* match ipifs that are down). 43767c478bd9Sstevel@tonic-gate * 43777c478bd9Sstevel@tonic-gate * Note that if IPIF_NOLOCAL, IPIF_NOXMIT, or IPIF_DEPRECATED is set on 43787c478bd9Sstevel@tonic-gate * the ipif, this routine might return NULL. 43797c478bd9Sstevel@tonic-gate */ 43807c478bd9Sstevel@tonic-gate ire_t * 438145916cd2Sjpk ipif_to_ire(const ipif_t *ipif) 43827c478bd9Sstevel@tonic-gate { 43837c478bd9Sstevel@tonic-gate ire_t *ire; 4384f4b3ec61Sdh155122 ip_stack_t *ipst = ipif->ipif_ill->ill_ipst; 43857c478bd9Sstevel@tonic-gate 43867c478bd9Sstevel@tonic-gate ASSERT(!ipif->ipif_isv6); 43877c478bd9Sstevel@tonic-gate if (ipif->ipif_ire_type == IRE_LOOPBACK) { 43887c478bd9Sstevel@tonic-gate ire = ire_ctable_lookup(ipif->ipif_lcl_addr, 0, IRE_LOOPBACK, 4389f4b3ec61Sdh155122 ipif, ALL_ZONES, NULL, (MATCH_IRE_TYPE | MATCH_IRE_IPIF), 4390f4b3ec61Sdh155122 ipst); 43917c478bd9Sstevel@tonic-gate } else if (ipif->ipif_flags & IPIF_POINTOPOINT) { 43927c478bd9Sstevel@tonic-gate /* In this case we need to lookup destination address. */ 43937c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(ipif->ipif_pp_dst_addr, IP_HOST_MASK, 0, 439445916cd2Sjpk IRE_INTERFACE, ipif, NULL, ALL_ZONES, 0, NULL, 4395f4b3ec61Sdh155122 (MATCH_IRE_TYPE | MATCH_IRE_IPIF | MATCH_IRE_MASK), ipst); 43967c478bd9Sstevel@tonic-gate } else { 43977c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(ipif->ipif_subnet, 43987c478bd9Sstevel@tonic-gate ipif->ipif_net_mask, 0, IRE_INTERFACE, ipif, NULL, 439945916cd2Sjpk ALL_ZONES, 0, NULL, (MATCH_IRE_TYPE | MATCH_IRE_IPIF | 4400f4b3ec61Sdh155122 MATCH_IRE_MASK), ipst); 44017c478bd9Sstevel@tonic-gate } 44027c478bd9Sstevel@tonic-gate return (ire); 44037c478bd9Sstevel@tonic-gate } 44047c478bd9Sstevel@tonic-gate 44057c478bd9Sstevel@tonic-gate /* 44067c478bd9Sstevel@tonic-gate * ire_walk function. 44077c478bd9Sstevel@tonic-gate * Count the number of IRE_CACHE entries in different categories. 44087c478bd9Sstevel@tonic-gate */ 44097c478bd9Sstevel@tonic-gate void 44107c478bd9Sstevel@tonic-gate ire_cache_count(ire_t *ire, char *arg) 44117c478bd9Sstevel@tonic-gate { 44127c478bd9Sstevel@tonic-gate ire_cache_count_t *icc = (ire_cache_count_t *)arg; 44137c478bd9Sstevel@tonic-gate 44147c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 44157c478bd9Sstevel@tonic-gate return; 44167c478bd9Sstevel@tonic-gate 44177c478bd9Sstevel@tonic-gate icc->icc_total++; 44187c478bd9Sstevel@tonic-gate 44197c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 44207c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 44217c478bd9Sstevel@tonic-gate if (IN6_IS_ADDR_UNSPECIFIED(&ire->ire_gateway_addr_v6)) { 44227c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 44237c478bd9Sstevel@tonic-gate icc->icc_onlink++; 44247c478bd9Sstevel@tonic-gate return; 44257c478bd9Sstevel@tonic-gate } 44267c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 44277c478bd9Sstevel@tonic-gate } else { 44287c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == 0) { 44297c478bd9Sstevel@tonic-gate icc->icc_onlink++; 44307c478bd9Sstevel@tonic-gate return; 44317c478bd9Sstevel@tonic-gate } 44327c478bd9Sstevel@tonic-gate } 44337c478bd9Sstevel@tonic-gate 44347c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipif != NULL); 44357c478bd9Sstevel@tonic-gate if (ire->ire_max_frag < ire->ire_ipif->ipif_mtu) 44367c478bd9Sstevel@tonic-gate icc->icc_pmtu++; 44377c478bd9Sstevel@tonic-gate else if (ire->ire_tire_mark != ire->ire_ob_pkt_count + 44387c478bd9Sstevel@tonic-gate ire->ire_ib_pkt_count) 44397c478bd9Sstevel@tonic-gate icc->icc_offlink++; 44407c478bd9Sstevel@tonic-gate else 44417c478bd9Sstevel@tonic-gate icc->icc_unused++; 44427c478bd9Sstevel@tonic-gate } 44437c478bd9Sstevel@tonic-gate 44447c478bd9Sstevel@tonic-gate /* 44457c478bd9Sstevel@tonic-gate * ire_walk function called by ip_trash_ire_reclaim(). 44467c478bd9Sstevel@tonic-gate * Free a fraction of the IRE_CACHE cache entries. The fractions are 44477c478bd9Sstevel@tonic-gate * different for different categories of IRE_CACHE entries. 44487c478bd9Sstevel@tonic-gate * A fraction of zero means to not free any in that category. 44497c478bd9Sstevel@tonic-gate * Use the hash bucket id plus lbolt as a random number. Thus if the fraction 44507c478bd9Sstevel@tonic-gate * is N then every Nth hash bucket chain will be freed. 44517c478bd9Sstevel@tonic-gate */ 44527c478bd9Sstevel@tonic-gate void 44537c478bd9Sstevel@tonic-gate ire_cache_reclaim(ire_t *ire, char *arg) 44547c478bd9Sstevel@tonic-gate { 44557c478bd9Sstevel@tonic-gate ire_cache_reclaim_t *icr = (ire_cache_reclaim_t *)arg; 44567c478bd9Sstevel@tonic-gate uint_t rand; 4457f4b3ec61Sdh155122 ip_stack_t *ipst = icr->icr_ipst; 44587c478bd9Sstevel@tonic-gate 44597c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 44607c478bd9Sstevel@tonic-gate return; 44617c478bd9Sstevel@tonic-gate 44627c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 44637c478bd9Sstevel@tonic-gate rand = (uint_t)lbolt + 4464f4b3ec61Sdh155122 IRE_ADDR_HASH_V6(ire->ire_addr_v6, 4465f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size); 44667c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 44677c478bd9Sstevel@tonic-gate if (IN6_IS_ADDR_UNSPECIFIED(&ire->ire_gateway_addr_v6)) { 44687c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 44697c478bd9Sstevel@tonic-gate if (icr->icr_onlink != 0 && 44707c478bd9Sstevel@tonic-gate (rand/icr->icr_onlink)*icr->icr_onlink == rand) { 44717c478bd9Sstevel@tonic-gate ire_delete(ire); 44727c478bd9Sstevel@tonic-gate return; 44737c478bd9Sstevel@tonic-gate } 44747c478bd9Sstevel@tonic-gate goto done; 44757c478bd9Sstevel@tonic-gate } 44767c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 44777c478bd9Sstevel@tonic-gate } else { 44787c478bd9Sstevel@tonic-gate rand = (uint_t)lbolt + 4479f4b3ec61Sdh155122 IRE_ADDR_HASH(ire->ire_addr, ipst->ips_ip_cache_table_size); 44807c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == 0) { 44817c478bd9Sstevel@tonic-gate if (icr->icr_onlink != 0 && 44827c478bd9Sstevel@tonic-gate (rand/icr->icr_onlink)*icr->icr_onlink == rand) { 44837c478bd9Sstevel@tonic-gate ire_delete(ire); 44847c478bd9Sstevel@tonic-gate return; 44857c478bd9Sstevel@tonic-gate } 44867c478bd9Sstevel@tonic-gate goto done; 44877c478bd9Sstevel@tonic-gate } 44887c478bd9Sstevel@tonic-gate } 44897c478bd9Sstevel@tonic-gate /* Not onlink IRE */ 44907c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipif != NULL); 44917c478bd9Sstevel@tonic-gate if (ire->ire_max_frag < ire->ire_ipif->ipif_mtu) { 44927c478bd9Sstevel@tonic-gate /* Use ptmu fraction */ 44937c478bd9Sstevel@tonic-gate if (icr->icr_pmtu != 0 && 44947c478bd9Sstevel@tonic-gate (rand/icr->icr_pmtu)*icr->icr_pmtu == rand) { 44957c478bd9Sstevel@tonic-gate ire_delete(ire); 44967c478bd9Sstevel@tonic-gate return; 44977c478bd9Sstevel@tonic-gate } 44987c478bd9Sstevel@tonic-gate } else if (ire->ire_tire_mark != ire->ire_ob_pkt_count + 44997c478bd9Sstevel@tonic-gate ire->ire_ib_pkt_count) { 45007c478bd9Sstevel@tonic-gate /* Use offlink fraction */ 45017c478bd9Sstevel@tonic-gate if (icr->icr_offlink != 0 && 45027c478bd9Sstevel@tonic-gate (rand/icr->icr_offlink)*icr->icr_offlink == rand) { 45037c478bd9Sstevel@tonic-gate ire_delete(ire); 45047c478bd9Sstevel@tonic-gate return; 45057c478bd9Sstevel@tonic-gate } 45067c478bd9Sstevel@tonic-gate } else { 45077c478bd9Sstevel@tonic-gate /* Use unused fraction */ 45087c478bd9Sstevel@tonic-gate if (icr->icr_unused != 0 && 45097c478bd9Sstevel@tonic-gate (rand/icr->icr_unused)*icr->icr_unused == rand) { 45107c478bd9Sstevel@tonic-gate ire_delete(ire); 45117c478bd9Sstevel@tonic-gate return; 45127c478bd9Sstevel@tonic-gate } 45137c478bd9Sstevel@tonic-gate } 45147c478bd9Sstevel@tonic-gate done: 45157c478bd9Sstevel@tonic-gate /* 45167c478bd9Sstevel@tonic-gate * Update tire_mark so that those that haven't been used since this 45177c478bd9Sstevel@tonic-gate * reclaim will be considered unused next time we reclaim. 45187c478bd9Sstevel@tonic-gate */ 45197c478bd9Sstevel@tonic-gate ire->ire_tire_mark = ire->ire_ob_pkt_count + ire->ire_ib_pkt_count; 45207c478bd9Sstevel@tonic-gate } 45217c478bd9Sstevel@tonic-gate 45227c478bd9Sstevel@tonic-gate static void 45237c478bd9Sstevel@tonic-gate power2_roundup(uint32_t *value) 45247c478bd9Sstevel@tonic-gate { 45257c478bd9Sstevel@tonic-gate int i; 45267c478bd9Sstevel@tonic-gate 45277c478bd9Sstevel@tonic-gate for (i = 1; i < 31; i++) { 45287c478bd9Sstevel@tonic-gate if (*value <= (1 << i)) 45297c478bd9Sstevel@tonic-gate break; 45307c478bd9Sstevel@tonic-gate } 45317c478bd9Sstevel@tonic-gate *value = (1 << i); 45327c478bd9Sstevel@tonic-gate } 45337c478bd9Sstevel@tonic-gate 4534f4b3ec61Sdh155122 /* Global init for all zones */ 45357c478bd9Sstevel@tonic-gate void 4536f4b3ec61Sdh155122 ip_ire_g_init() 45377c478bd9Sstevel@tonic-gate { 45387c478bd9Sstevel@tonic-gate /* 45397c478bd9Sstevel@tonic-gate * Create ire caches, ire_reclaim() 45407c478bd9Sstevel@tonic-gate * will give IRE_CACHE back to system when needed. 45417c478bd9Sstevel@tonic-gate * This needs to be done here before anything else, since 45427c478bd9Sstevel@tonic-gate * ire_add() expects the cache to be created. 45437c478bd9Sstevel@tonic-gate */ 45447c478bd9Sstevel@tonic-gate ire_cache = kmem_cache_create("ire_cache", 45457c478bd9Sstevel@tonic-gate sizeof (ire_t), 0, ip_ire_constructor, 45467c478bd9Sstevel@tonic-gate ip_ire_destructor, ip_trash_ire_reclaim, NULL, NULL, 0); 45477c478bd9Sstevel@tonic-gate 4548f4b3ec61Sdh155122 rt_entry_cache = kmem_cache_create("rt_entry", 4549f4b3ec61Sdh155122 sizeof (struct rt_entry), 0, NULL, NULL, NULL, NULL, NULL, 0); 4550f4b3ec61Sdh155122 4551f4b3ec61Sdh155122 /* 4552f4b3ec61Sdh155122 * Have radix code setup kmem caches etc. 4553f4b3ec61Sdh155122 */ 4554f4b3ec61Sdh155122 rn_init(); 4555f4b3ec61Sdh155122 } 4556f4b3ec61Sdh155122 4557f4b3ec61Sdh155122 void 4558f4b3ec61Sdh155122 ip_ire_init(ip_stack_t *ipst) 4559f4b3ec61Sdh155122 { 4560f4b3ec61Sdh155122 int i; 4561f4b3ec61Sdh155122 uint32_t mem_cnt; 4562f4b3ec61Sdh155122 uint32_t cpu_cnt; 4563f4b3ec61Sdh155122 uint32_t min_cnt; 4564f4b3ec61Sdh155122 pgcnt_t mem_avail; 4565f4b3ec61Sdh155122 4566f4b3ec61Sdh155122 /* 4567f4b3ec61Sdh155122 * ip_ire_max_bucket_cnt is sized below based on the memory 4568f4b3ec61Sdh155122 * size and the cpu speed of the machine. This is upper 4569f4b3ec61Sdh155122 * bounded by the compile time value of ip_ire_max_bucket_cnt 4570f4b3ec61Sdh155122 * and is lower bounded by the compile time value of 4571f4b3ec61Sdh155122 * ip_ire_min_bucket_cnt. Similar logic applies to 4572f4b3ec61Sdh155122 * ip6_ire_max_bucket_cnt. 4573f4b3ec61Sdh155122 * 4574f4b3ec61Sdh155122 * We calculate this for each IP Instances in order to use 4575f4b3ec61Sdh155122 * the kmem_avail and ip_ire_{min,max}_bucket_cnt that are 4576f4b3ec61Sdh155122 * in effect when the zone is booted. 4577f4b3ec61Sdh155122 */ 4578f4b3ec61Sdh155122 mem_avail = kmem_avail(); 4579f4b3ec61Sdh155122 mem_cnt = (mem_avail >> ip_ire_mem_ratio) / 4580f4b3ec61Sdh155122 ip_cache_table_size / sizeof (ire_t); 4581f4b3ec61Sdh155122 cpu_cnt = CPU->cpu_type_info.pi_clock >> ip_ire_cpu_ratio; 4582f4b3ec61Sdh155122 4583f4b3ec61Sdh155122 min_cnt = MIN(cpu_cnt, mem_cnt); 4584f4b3ec61Sdh155122 if (min_cnt < ip_ire_min_bucket_cnt) 4585f4b3ec61Sdh155122 min_cnt = ip_ire_min_bucket_cnt; 4586f4b3ec61Sdh155122 if (ip_ire_max_bucket_cnt > min_cnt) { 4587f4b3ec61Sdh155122 ip_ire_max_bucket_cnt = min_cnt; 4588f4b3ec61Sdh155122 } 4589f4b3ec61Sdh155122 4590f4b3ec61Sdh155122 mem_cnt = (mem_avail >> ip_ire_mem_ratio) / 4591f4b3ec61Sdh155122 ip6_cache_table_size / sizeof (ire_t); 4592f4b3ec61Sdh155122 min_cnt = MIN(cpu_cnt, mem_cnt); 4593f4b3ec61Sdh155122 if (min_cnt < ip6_ire_min_bucket_cnt) 4594f4b3ec61Sdh155122 min_cnt = ip6_ire_min_bucket_cnt; 4595f4b3ec61Sdh155122 if (ip6_ire_max_bucket_cnt > min_cnt) { 4596f4b3ec61Sdh155122 ip6_ire_max_bucket_cnt = min_cnt; 4597f4b3ec61Sdh155122 } 4598f4b3ec61Sdh155122 4599f4b3ec61Sdh155122 mutex_init(&ipst->ips_ire_ft_init_lock, NULL, MUTEX_DEFAULT, 0); 4600f4b3ec61Sdh155122 mutex_init(&ipst->ips_ire_handle_lock, NULL, MUTEX_DEFAULT, NULL); 4601f4b3ec61Sdh155122 4602f4b3ec61Sdh155122 (void) rn_inithead((void **)&ipst->ips_ip_ftable, 32); 4603f4b3ec61Sdh155122 4604f4b3ec61Sdh155122 4605f4b3ec61Sdh155122 /* Calculate the IPv4 cache table size. */ 4606f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size = MAX(ip_cache_table_size, 4607f4b3ec61Sdh155122 ((mem_avail >> ip_ire_mem_ratio) / sizeof (ire_t) / 4608f4b3ec61Sdh155122 ip_ire_max_bucket_cnt)); 4609f4b3ec61Sdh155122 if (ipst->ips_ip_cache_table_size > ip_max_cache_table_size) 4610f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size = ip_max_cache_table_size; 4611f4b3ec61Sdh155122 /* 4612f4b3ec61Sdh155122 * Make sure that the table size is always a power of 2. The 4613f4b3ec61Sdh155122 * hash macro IRE_ADDR_HASH() depends on that. 4614f4b3ec61Sdh155122 */ 4615f4b3ec61Sdh155122 power2_roundup(&ipst->ips_ip_cache_table_size); 4616f4b3ec61Sdh155122 4617f4b3ec61Sdh155122 ipst->ips_ip_cache_table = kmem_zalloc(ipst->ips_ip_cache_table_size * 4618f4b3ec61Sdh155122 sizeof (irb_t), KM_SLEEP); 4619f4b3ec61Sdh155122 4620f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 4621f4b3ec61Sdh155122 rw_init(&ipst->ips_ip_cache_table[i].irb_lock, NULL, 4622f4b3ec61Sdh155122 RW_DEFAULT, NULL); 4623f4b3ec61Sdh155122 } 4624f4b3ec61Sdh155122 4625f4b3ec61Sdh155122 /* Calculate the IPv6 cache table size. */ 4626f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size = MAX(ip6_cache_table_size, 4627f4b3ec61Sdh155122 ((mem_avail >> ip_ire_mem_ratio) / sizeof (ire_t) / 4628f4b3ec61Sdh155122 ip6_ire_max_bucket_cnt)); 4629f4b3ec61Sdh155122 if (ipst->ips_ip6_cache_table_size > ip6_max_cache_table_size) 4630f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size = ip6_max_cache_table_size; 4631f4b3ec61Sdh155122 /* 4632f4b3ec61Sdh155122 * Make sure that the table size is always a power of 2. The 4633f4b3ec61Sdh155122 * hash macro IRE_ADDR_HASH_V6() depends on that. 4634f4b3ec61Sdh155122 */ 4635f4b3ec61Sdh155122 power2_roundup(&ipst->ips_ip6_cache_table_size); 4636f4b3ec61Sdh155122 4637f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6 = kmem_zalloc( 4638f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size * sizeof (irb_t), KM_SLEEP); 4639f4b3ec61Sdh155122 4640f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip6_cache_table_size; i++) { 4641f4b3ec61Sdh155122 rw_init(&ipst->ips_ip_cache_table_v6[i].irb_lock, NULL, 4642f4b3ec61Sdh155122 RW_DEFAULT, NULL); 4643f4b3ec61Sdh155122 } 4644f4b3ec61Sdh155122 46457c478bd9Sstevel@tonic-gate /* 46467c478bd9Sstevel@tonic-gate * Make sure that the forwarding table size is a power of 2. 46477c478bd9Sstevel@tonic-gate * The IRE*_ADDR_HASH() macroes depend on that. 46487c478bd9Sstevel@tonic-gate */ 4649f4b3ec61Sdh155122 ipst->ips_ip6_ftable_hash_size = ip6_ftable_hash_size; 4650f4b3ec61Sdh155122 power2_roundup(&ipst->ips_ip6_ftable_hash_size); 4651f4b3ec61Sdh155122 4652f4b3ec61Sdh155122 ipst->ips_ire_handle = 1; 46537c478bd9Sstevel@tonic-gate } 46547c478bd9Sstevel@tonic-gate 46557c478bd9Sstevel@tonic-gate void 4656f4b3ec61Sdh155122 ip_ire_g_fini(void) 4657f4b3ec61Sdh155122 { 4658f4b3ec61Sdh155122 kmem_cache_destroy(ire_cache); 4659f4b3ec61Sdh155122 kmem_cache_destroy(rt_entry_cache); 4660f4b3ec61Sdh155122 4661f4b3ec61Sdh155122 rn_fini(); 4662f4b3ec61Sdh155122 } 4663f4b3ec61Sdh155122 4664f4b3ec61Sdh155122 void 4665f4b3ec61Sdh155122 ip_ire_fini(ip_stack_t *ipst) 46667c478bd9Sstevel@tonic-gate { 46677c478bd9Sstevel@tonic-gate int i; 46687c478bd9Sstevel@tonic-gate 4669f4b3ec61Sdh155122 /* 4670f4b3ec61Sdh155122 * Delete all IREs - assumes that the ill/ipifs have 4671f4b3ec61Sdh155122 * been removed so what remains are just the ftable and IRE_CACHE. 4672f4b3ec61Sdh155122 */ 4673f4b3ec61Sdh155122 ire_walk(ire_delete, NULL, ipst); 4674c793af95Ssangeeta 4675f4b3ec61Sdh155122 rn_freehead(ipst->ips_ip_ftable); 4676f4b3ec61Sdh155122 ipst->ips_ip_ftable = NULL; 46777c478bd9Sstevel@tonic-gate 4678f4b3ec61Sdh155122 mutex_destroy(&ipst->ips_ire_ft_init_lock); 4679f4b3ec61Sdh155122 mutex_destroy(&ipst->ips_ire_handle_lock); 4680f4b3ec61Sdh155122 4681f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 4682f4b3ec61Sdh155122 ASSERT(ipst->ips_ip_cache_table[i].irb_ire == NULL); 4683f4b3ec61Sdh155122 rw_destroy(&ipst->ips_ip_cache_table[i].irb_lock); 46847c478bd9Sstevel@tonic-gate } 4685f4b3ec61Sdh155122 kmem_free(ipst->ips_ip_cache_table, 4686f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size * sizeof (irb_t)); 4687f4b3ec61Sdh155122 ipst->ips_ip_cache_table = NULL; 46887c478bd9Sstevel@tonic-gate 4689f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip6_cache_table_size; i++) { 4690f4b3ec61Sdh155122 ASSERT(ipst->ips_ip_cache_table_v6[i].irb_ire == NULL); 4691f4b3ec61Sdh155122 rw_destroy(&ipst->ips_ip_cache_table_v6[i].irb_lock); 46927c478bd9Sstevel@tonic-gate } 4693f4b3ec61Sdh155122 kmem_free(ipst->ips_ip_cache_table_v6, 4694f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size * sizeof (irb_t)); 4695f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6 = NULL; 46967c478bd9Sstevel@tonic-gate 4697f4b3ec61Sdh155122 for (i = 0; i < IP6_MASK_TABLE_SIZE; i++) { 4698f4b3ec61Sdh155122 irb_t *ptr; 4699f4b3ec61Sdh155122 int j; 4700f4b3ec61Sdh155122 4701f4b3ec61Sdh155122 if ((ptr = ipst->ips_ip_forwarding_table_v6[i]) == NULL) 4702f4b3ec61Sdh155122 continue; 4703f4b3ec61Sdh155122 4704f4b3ec61Sdh155122 for (j = 0; j < ipst->ips_ip6_ftable_hash_size; j++) { 4705f4b3ec61Sdh155122 ASSERT(ptr[j].irb_ire == NULL); 4706f4b3ec61Sdh155122 rw_destroy(&ptr[j].irb_lock); 4707f4b3ec61Sdh155122 } 4708f4b3ec61Sdh155122 mi_free(ptr); 4709f4b3ec61Sdh155122 ipst->ips_ip_forwarding_table_v6[i] = NULL; 4710f4b3ec61Sdh155122 } 47117c478bd9Sstevel@tonic-gate } 47127c478bd9Sstevel@tonic-gate 47137c478bd9Sstevel@tonic-gate /* 47147c478bd9Sstevel@tonic-gate * Check if another multirt route resolution is needed. 47157c478bd9Sstevel@tonic-gate * B_TRUE is returned is there remain a resolvable route, 47167c478bd9Sstevel@tonic-gate * or if no route for that dst is resolved yet. 47177c478bd9Sstevel@tonic-gate * B_FALSE is returned if all routes for that dst are resolved 47187c478bd9Sstevel@tonic-gate * or if the remaining unresolved routes are actually not 47197c478bd9Sstevel@tonic-gate * resolvable. 47207c478bd9Sstevel@tonic-gate * This only works in the global zone. 47217c478bd9Sstevel@tonic-gate */ 47227c478bd9Sstevel@tonic-gate boolean_t 4723f4b3ec61Sdh155122 ire_multirt_need_resolve(ipaddr_t dst, const ts_label_t *tsl, ip_stack_t *ipst) 47247c478bd9Sstevel@tonic-gate { 47257c478bd9Sstevel@tonic-gate ire_t *first_fire; 47267c478bd9Sstevel@tonic-gate ire_t *first_cire; 47277c478bd9Sstevel@tonic-gate ire_t *fire; 47287c478bd9Sstevel@tonic-gate ire_t *cire; 47297c478bd9Sstevel@tonic-gate irb_t *firb; 47307c478bd9Sstevel@tonic-gate irb_t *cirb; 47317c478bd9Sstevel@tonic-gate int unres_cnt = 0; 47327c478bd9Sstevel@tonic-gate boolean_t resolvable = B_FALSE; 47337c478bd9Sstevel@tonic-gate 47347c478bd9Sstevel@tonic-gate /* Retrieve the first IRE_HOST that matches the destination */ 47357c478bd9Sstevel@tonic-gate first_fire = ire_ftable_lookup(dst, IP_HOST_MASK, 0, IRE_HOST, NULL, 473645916cd2Sjpk NULL, ALL_ZONES, 0, tsl, 4737f4b3ec61Sdh155122 MATCH_IRE_MASK | MATCH_IRE_TYPE | MATCH_IRE_SECATTR, ipst); 47387c478bd9Sstevel@tonic-gate 47397c478bd9Sstevel@tonic-gate /* No route at all */ 47407c478bd9Sstevel@tonic-gate if (first_fire == NULL) { 47417c478bd9Sstevel@tonic-gate return (B_TRUE); 47427c478bd9Sstevel@tonic-gate } 47437c478bd9Sstevel@tonic-gate 47447c478bd9Sstevel@tonic-gate firb = first_fire->ire_bucket; 47457c478bd9Sstevel@tonic-gate ASSERT(firb != NULL); 47467c478bd9Sstevel@tonic-gate 47477c478bd9Sstevel@tonic-gate /* Retrieve the first IRE_CACHE ire for that destination. */ 4748f4b3ec61Sdh155122 first_cire = ire_cache_lookup(dst, GLOBAL_ZONEID, tsl, ipst); 47497c478bd9Sstevel@tonic-gate 47507c478bd9Sstevel@tonic-gate /* No resolved route. */ 47517c478bd9Sstevel@tonic-gate if (first_cire == NULL) { 47527c478bd9Sstevel@tonic-gate ire_refrele(first_fire); 47537c478bd9Sstevel@tonic-gate return (B_TRUE); 47547c478bd9Sstevel@tonic-gate } 47557c478bd9Sstevel@tonic-gate 47567c478bd9Sstevel@tonic-gate /* 47577c478bd9Sstevel@tonic-gate * At least one route is resolved. Here we look through the forward 47587c478bd9Sstevel@tonic-gate * and cache tables, to compare the number of declared routes 47597c478bd9Sstevel@tonic-gate * with the number of resolved routes. The search for a resolvable 47607c478bd9Sstevel@tonic-gate * route is performed only if at least one route remains 47617c478bd9Sstevel@tonic-gate * unresolved. 47627c478bd9Sstevel@tonic-gate */ 47637c478bd9Sstevel@tonic-gate cirb = first_cire->ire_bucket; 47647c478bd9Sstevel@tonic-gate ASSERT(cirb != NULL); 47657c478bd9Sstevel@tonic-gate 47667c478bd9Sstevel@tonic-gate /* Count the number of routes to that dest that are declared. */ 47677c478bd9Sstevel@tonic-gate IRB_REFHOLD(firb); 47687c478bd9Sstevel@tonic-gate for (fire = first_fire; fire != NULL; fire = fire->ire_next) { 47697c478bd9Sstevel@tonic-gate if (!(fire->ire_flags & RTF_MULTIRT)) 47707c478bd9Sstevel@tonic-gate continue; 47717c478bd9Sstevel@tonic-gate if (fire->ire_addr != dst) 47727c478bd9Sstevel@tonic-gate continue; 47737c478bd9Sstevel@tonic-gate unres_cnt++; 47747c478bd9Sstevel@tonic-gate } 47757c478bd9Sstevel@tonic-gate IRB_REFRELE(firb); 47767c478bd9Sstevel@tonic-gate 47777c478bd9Sstevel@tonic-gate /* Then subtract the number of routes to that dst that are resolved */ 47787c478bd9Sstevel@tonic-gate IRB_REFHOLD(cirb); 47797c478bd9Sstevel@tonic-gate for (cire = first_cire; cire != NULL; cire = cire->ire_next) { 47807c478bd9Sstevel@tonic-gate if (!(cire->ire_flags & RTF_MULTIRT)) 47817c478bd9Sstevel@tonic-gate continue; 47827c478bd9Sstevel@tonic-gate if (cire->ire_addr != dst) 47837c478bd9Sstevel@tonic-gate continue; 47847c478bd9Sstevel@tonic-gate if (cire->ire_marks & (IRE_MARK_CONDEMNED | IRE_MARK_HIDDEN)) 47857c478bd9Sstevel@tonic-gate continue; 47867c478bd9Sstevel@tonic-gate unres_cnt--; 47877c478bd9Sstevel@tonic-gate } 47887c478bd9Sstevel@tonic-gate IRB_REFRELE(cirb); 47897c478bd9Sstevel@tonic-gate 47907c478bd9Sstevel@tonic-gate /* At least one route is unresolved; search for a resolvable route. */ 47917c478bd9Sstevel@tonic-gate if (unres_cnt > 0) 47927c478bd9Sstevel@tonic-gate resolvable = ire_multirt_lookup(&first_cire, &first_fire, 4793f4b3ec61Sdh155122 MULTIRT_USESTAMP | MULTIRT_CACHEGW, tsl, ipst); 47947c478bd9Sstevel@tonic-gate 47957c478bd9Sstevel@tonic-gate if (first_fire != NULL) 47967c478bd9Sstevel@tonic-gate ire_refrele(first_fire); 47977c478bd9Sstevel@tonic-gate 47987c478bd9Sstevel@tonic-gate if (first_cire != NULL) 47997c478bd9Sstevel@tonic-gate ire_refrele(first_cire); 48007c478bd9Sstevel@tonic-gate 48017c478bd9Sstevel@tonic-gate return (resolvable); 48027c478bd9Sstevel@tonic-gate } 48037c478bd9Sstevel@tonic-gate 48047c478bd9Sstevel@tonic-gate 48057c478bd9Sstevel@tonic-gate /* 48067c478bd9Sstevel@tonic-gate * Explore a forward_table bucket, starting from fire_arg. 48077c478bd9Sstevel@tonic-gate * fire_arg MUST be an IRE_HOST entry. 48087c478bd9Sstevel@tonic-gate * 48097c478bd9Sstevel@tonic-gate * Return B_TRUE and update *ire_arg and *fire_arg 48107c478bd9Sstevel@tonic-gate * if at least one resolvable route is found. *ire_arg 48117c478bd9Sstevel@tonic-gate * is the IRE entry for *fire_arg's gateway. 48127c478bd9Sstevel@tonic-gate * 48137c478bd9Sstevel@tonic-gate * Return B_FALSE otherwise (all routes are resolved or 48147c478bd9Sstevel@tonic-gate * the remaining unresolved routes are all unresolvable). 48157c478bd9Sstevel@tonic-gate * 48167c478bd9Sstevel@tonic-gate * The IRE selection relies on a priority mechanism 48177c478bd9Sstevel@tonic-gate * driven by the flags passed in by the caller. 48187c478bd9Sstevel@tonic-gate * The caller, such as ip_newroute_ipif(), can get the most 48197c478bd9Sstevel@tonic-gate * relevant ire at each stage of a multiple route resolution. 48207c478bd9Sstevel@tonic-gate * 48217c478bd9Sstevel@tonic-gate * The rules are: 48227c478bd9Sstevel@tonic-gate * 48237c478bd9Sstevel@tonic-gate * - if MULTIRT_CACHEGW is specified in flags, IRE_CACHETABLE 48247c478bd9Sstevel@tonic-gate * ires are preferred for the gateway. This gives the highest 48257c478bd9Sstevel@tonic-gate * priority to routes that can be resolved without using 48267c478bd9Sstevel@tonic-gate * a resolver. 48277c478bd9Sstevel@tonic-gate * 48287c478bd9Sstevel@tonic-gate * - if MULTIRT_CACHEGW is not specified, or if MULTIRT_CACHEGW 48297c478bd9Sstevel@tonic-gate * is specified but no IRE_CACHETABLE ire entry for the gateway 48307c478bd9Sstevel@tonic-gate * is found, the following rules apply. 48317c478bd9Sstevel@tonic-gate * 48327c478bd9Sstevel@tonic-gate * - if MULTIRT_USESTAMP is specified in flags, IRE_INTERFACE 48337c478bd9Sstevel@tonic-gate * ires for the gateway, that have not been tried since 48347c478bd9Sstevel@tonic-gate * a configurable amount of time, are preferred. 48357c478bd9Sstevel@tonic-gate * This applies when a resolver must be invoked for 48367c478bd9Sstevel@tonic-gate * a missing route, but we don't want to use the resolver 48377c478bd9Sstevel@tonic-gate * upon each packet emission. If no such resolver is found, 48387c478bd9Sstevel@tonic-gate * B_FALSE is returned. 48397c478bd9Sstevel@tonic-gate * The MULTIRT_USESTAMP flag can be combined with 48407c478bd9Sstevel@tonic-gate * MULTIRT_CACHEGW. 48417c478bd9Sstevel@tonic-gate * 48427c478bd9Sstevel@tonic-gate * - if MULTIRT_USESTAMP is not specified in flags, the first 48437c478bd9Sstevel@tonic-gate * unresolved but resolvable route is selected. 48447c478bd9Sstevel@tonic-gate * 48457c478bd9Sstevel@tonic-gate * - Otherwise, there is no resolvalble route, and 48467c478bd9Sstevel@tonic-gate * B_FALSE is returned. 48477c478bd9Sstevel@tonic-gate * 48487c478bd9Sstevel@tonic-gate * At last, MULTIRT_SETSTAMP can be specified in flags to 48497c478bd9Sstevel@tonic-gate * request the timestamp of unresolvable routes to 48507c478bd9Sstevel@tonic-gate * be refreshed. This prevents the useless exploration 48517c478bd9Sstevel@tonic-gate * of those routes for a while, when MULTIRT_USESTAMP is used. 48527c478bd9Sstevel@tonic-gate * 48537c478bd9Sstevel@tonic-gate * This only works in the global zone. 48547c478bd9Sstevel@tonic-gate */ 48557c478bd9Sstevel@tonic-gate boolean_t 485645916cd2Sjpk ire_multirt_lookup(ire_t **ire_arg, ire_t **fire_arg, uint32_t flags, 4857f4b3ec61Sdh155122 const ts_label_t *tsl, ip_stack_t *ipst) 48587c478bd9Sstevel@tonic-gate { 48597c478bd9Sstevel@tonic-gate clock_t delta; 48607c478bd9Sstevel@tonic-gate ire_t *best_fire = NULL; 48617c478bd9Sstevel@tonic-gate ire_t *best_cire = NULL; 48627c478bd9Sstevel@tonic-gate ire_t *first_fire; 48637c478bd9Sstevel@tonic-gate ire_t *first_cire; 48647c478bd9Sstevel@tonic-gate ire_t *fire; 48657c478bd9Sstevel@tonic-gate ire_t *cire; 48667c478bd9Sstevel@tonic-gate irb_t *firb = NULL; 48677c478bd9Sstevel@tonic-gate irb_t *cirb = NULL; 48687c478bd9Sstevel@tonic-gate ire_t *gw_ire; 48697c478bd9Sstevel@tonic-gate boolean_t already_resolved; 48707c478bd9Sstevel@tonic-gate boolean_t res; 48717c478bd9Sstevel@tonic-gate ipaddr_t dst; 48727c478bd9Sstevel@tonic-gate ipaddr_t gw; 48737c478bd9Sstevel@tonic-gate 48747c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: *ire_arg %p, *fire_arg %p, flags %04x\n", 48757c478bd9Sstevel@tonic-gate (void *)*ire_arg, (void *)*fire_arg, flags)); 48767c478bd9Sstevel@tonic-gate 48777c478bd9Sstevel@tonic-gate ASSERT(ire_arg != NULL); 48787c478bd9Sstevel@tonic-gate ASSERT(fire_arg != NULL); 48797c478bd9Sstevel@tonic-gate 48807c478bd9Sstevel@tonic-gate /* Not an IRE_HOST ire; give up. */ 48817c478bd9Sstevel@tonic-gate if ((*fire_arg == NULL) || ((*fire_arg)->ire_type != IRE_HOST)) { 48827c478bd9Sstevel@tonic-gate return (B_FALSE); 48837c478bd9Sstevel@tonic-gate } 48847c478bd9Sstevel@tonic-gate 48857c478bd9Sstevel@tonic-gate /* This is the first IRE_HOST ire for that destination. */ 48867c478bd9Sstevel@tonic-gate first_fire = *fire_arg; 48877c478bd9Sstevel@tonic-gate firb = first_fire->ire_bucket; 48887c478bd9Sstevel@tonic-gate ASSERT(firb != NULL); 48897c478bd9Sstevel@tonic-gate 48907c478bd9Sstevel@tonic-gate dst = first_fire->ire_addr; 48917c478bd9Sstevel@tonic-gate 48927c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: dst %08x\n", ntohl(dst))); 48937c478bd9Sstevel@tonic-gate 48947c478bd9Sstevel@tonic-gate /* 48957c478bd9Sstevel@tonic-gate * Retrieve the first IRE_CACHE ire for that destination; 48967c478bd9Sstevel@tonic-gate * if we don't find one, no route for that dest is 48977c478bd9Sstevel@tonic-gate * resolved yet. 48987c478bd9Sstevel@tonic-gate */ 4899f4b3ec61Sdh155122 first_cire = ire_cache_lookup(dst, GLOBAL_ZONEID, tsl, ipst); 49007c478bd9Sstevel@tonic-gate if (first_cire != NULL) { 49017c478bd9Sstevel@tonic-gate cirb = first_cire->ire_bucket; 49027c478bd9Sstevel@tonic-gate } 49037c478bd9Sstevel@tonic-gate 49047c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: first_cire %p\n", (void *)first_cire)); 49057c478bd9Sstevel@tonic-gate 49067c478bd9Sstevel@tonic-gate /* 49077c478bd9Sstevel@tonic-gate * Search for a resolvable route, giving the top priority 49087c478bd9Sstevel@tonic-gate * to routes that can be resolved without any call to the resolver. 49097c478bd9Sstevel@tonic-gate */ 49107c478bd9Sstevel@tonic-gate IRB_REFHOLD(firb); 49117c478bd9Sstevel@tonic-gate 49127c478bd9Sstevel@tonic-gate if (!CLASSD(dst)) { 49137c478bd9Sstevel@tonic-gate /* 49147c478bd9Sstevel@tonic-gate * For all multiroute IRE_HOST ires for that destination, 49157c478bd9Sstevel@tonic-gate * check if the route via the IRE_HOST's gateway is 49167c478bd9Sstevel@tonic-gate * resolved yet. 49177c478bd9Sstevel@tonic-gate */ 49187c478bd9Sstevel@tonic-gate for (fire = first_fire; fire != NULL; fire = fire->ire_next) { 49197c478bd9Sstevel@tonic-gate 49207c478bd9Sstevel@tonic-gate if (!(fire->ire_flags & RTF_MULTIRT)) 49217c478bd9Sstevel@tonic-gate continue; 49227c478bd9Sstevel@tonic-gate if (fire->ire_addr != dst) 49237c478bd9Sstevel@tonic-gate continue; 49247c478bd9Sstevel@tonic-gate 492545916cd2Sjpk if (fire->ire_gw_secattr != NULL && 492645916cd2Sjpk tsol_ire_match_gwattr(fire, tsl) != 0) { 492745916cd2Sjpk continue; 492845916cd2Sjpk } 492945916cd2Sjpk 49307c478bd9Sstevel@tonic-gate gw = fire->ire_gateway_addr; 49317c478bd9Sstevel@tonic-gate 49327c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: fire %p, " 49337c478bd9Sstevel@tonic-gate "ire_addr %08x, ire_gateway_addr %08x\n", 49347c478bd9Sstevel@tonic-gate (void *)fire, ntohl(fire->ire_addr), ntohl(gw))); 49357c478bd9Sstevel@tonic-gate 49367c478bd9Sstevel@tonic-gate already_resolved = B_FALSE; 49377c478bd9Sstevel@tonic-gate 49387c478bd9Sstevel@tonic-gate if (first_cire != NULL) { 49397c478bd9Sstevel@tonic-gate ASSERT(cirb != NULL); 49407c478bd9Sstevel@tonic-gate 49417c478bd9Sstevel@tonic-gate IRB_REFHOLD(cirb); 49427c478bd9Sstevel@tonic-gate /* 49437c478bd9Sstevel@tonic-gate * For all IRE_CACHE ires for that 49447c478bd9Sstevel@tonic-gate * destination. 49457c478bd9Sstevel@tonic-gate */ 49467c478bd9Sstevel@tonic-gate for (cire = first_cire; 49477c478bd9Sstevel@tonic-gate cire != NULL; 49487c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 49497c478bd9Sstevel@tonic-gate 49507c478bd9Sstevel@tonic-gate if (!(cire->ire_flags & RTF_MULTIRT)) 49517c478bd9Sstevel@tonic-gate continue; 49527c478bd9Sstevel@tonic-gate if (cire->ire_addr != dst) 49537c478bd9Sstevel@tonic-gate continue; 49547c478bd9Sstevel@tonic-gate if (cire->ire_marks & 49557c478bd9Sstevel@tonic-gate (IRE_MARK_CONDEMNED | 49567c478bd9Sstevel@tonic-gate IRE_MARK_HIDDEN)) 49577c478bd9Sstevel@tonic-gate continue; 495845916cd2Sjpk 495945916cd2Sjpk if (cire->ire_gw_secattr != NULL && 496045916cd2Sjpk tsol_ire_match_gwattr(cire, 496145916cd2Sjpk tsl) != 0) { 496245916cd2Sjpk continue; 496345916cd2Sjpk } 496445916cd2Sjpk 49657c478bd9Sstevel@tonic-gate /* 49667c478bd9Sstevel@tonic-gate * Check if the IRE_CACHE's gateway 49677c478bd9Sstevel@tonic-gate * matches the IRE_HOST's gateway. 49687c478bd9Sstevel@tonic-gate */ 49697c478bd9Sstevel@tonic-gate if (cire->ire_gateway_addr == gw) { 49707c478bd9Sstevel@tonic-gate already_resolved = B_TRUE; 49717c478bd9Sstevel@tonic-gate break; 49727c478bd9Sstevel@tonic-gate } 49737c478bd9Sstevel@tonic-gate } 49747c478bd9Sstevel@tonic-gate IRB_REFRELE(cirb); 49757c478bd9Sstevel@tonic-gate } 49767c478bd9Sstevel@tonic-gate 49777c478bd9Sstevel@tonic-gate /* 49787c478bd9Sstevel@tonic-gate * This route is already resolved; 49797c478bd9Sstevel@tonic-gate * proceed with next one. 49807c478bd9Sstevel@tonic-gate */ 49817c478bd9Sstevel@tonic-gate if (already_resolved) { 49827c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: found cire %p, " 49837c478bd9Sstevel@tonic-gate "already resolved\n", (void *)cire)); 49847c478bd9Sstevel@tonic-gate continue; 49857c478bd9Sstevel@tonic-gate } 49867c478bd9Sstevel@tonic-gate 49877c478bd9Sstevel@tonic-gate /* 49887c478bd9Sstevel@tonic-gate * The route is unresolved; is it actually 49897c478bd9Sstevel@tonic-gate * resolvable, i.e. is there a cache or a resolver 49907c478bd9Sstevel@tonic-gate * for the gateway? 49917c478bd9Sstevel@tonic-gate */ 49927c478bd9Sstevel@tonic-gate gw_ire = ire_route_lookup(gw, 0, 0, 0, NULL, NULL, 499345916cd2Sjpk ALL_ZONES, tsl, 4994f4b3ec61Sdh155122 MATCH_IRE_RECURSIVE | MATCH_IRE_SECATTR, ipst); 49957c478bd9Sstevel@tonic-gate 49967c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: looked up gw_ire %p\n", 49977c478bd9Sstevel@tonic-gate (void *)gw_ire)); 49987c478bd9Sstevel@tonic-gate 49997c478bd9Sstevel@tonic-gate /* 50007c478bd9Sstevel@tonic-gate * If gw_ire is typed IRE_CACHETABLE, 50017c478bd9Sstevel@tonic-gate * this route can be resolved without any call to the 50027c478bd9Sstevel@tonic-gate * resolver. If the MULTIRT_CACHEGW flag is set, 50037c478bd9Sstevel@tonic-gate * give the top priority to this ire and exit the 50047c478bd9Sstevel@tonic-gate * loop. 50057c478bd9Sstevel@tonic-gate * This is typically the case when an ARP reply 50067c478bd9Sstevel@tonic-gate * is processed through ip_wput_nondata(). 50077c478bd9Sstevel@tonic-gate */ 50087c478bd9Sstevel@tonic-gate if ((flags & MULTIRT_CACHEGW) && 50097c478bd9Sstevel@tonic-gate (gw_ire != NULL) && 50107c478bd9Sstevel@tonic-gate (gw_ire->ire_type & IRE_CACHETABLE)) { 5011c793af95Ssangeeta ASSERT(gw_ire->ire_nce == NULL || 5012c793af95Ssangeeta gw_ire->ire_nce->nce_state == ND_REACHABLE); 50137c478bd9Sstevel@tonic-gate /* 50147c478bd9Sstevel@tonic-gate * Release the resolver associated to the 50157c478bd9Sstevel@tonic-gate * previous candidate best ire, if any. 50167c478bd9Sstevel@tonic-gate */ 50177c478bd9Sstevel@tonic-gate if (best_cire != NULL) { 50187c478bd9Sstevel@tonic-gate ire_refrele(best_cire); 50197c478bd9Sstevel@tonic-gate ASSERT(best_fire != NULL); 50207c478bd9Sstevel@tonic-gate } 50217c478bd9Sstevel@tonic-gate 50227c478bd9Sstevel@tonic-gate best_fire = fire; 50237c478bd9Sstevel@tonic-gate best_cire = gw_ire; 50247c478bd9Sstevel@tonic-gate 50257c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: found top prio " 50267c478bd9Sstevel@tonic-gate "best_fire %p, best_cire %p\n", 50277c478bd9Sstevel@tonic-gate (void *)best_fire, (void *)best_cire)); 50287c478bd9Sstevel@tonic-gate break; 50297c478bd9Sstevel@tonic-gate } 50307c478bd9Sstevel@tonic-gate 50317c478bd9Sstevel@tonic-gate /* 50327c478bd9Sstevel@tonic-gate * Compute the time elapsed since our preceding 50337c478bd9Sstevel@tonic-gate * attempt to resolve that route. 50347c478bd9Sstevel@tonic-gate * If the MULTIRT_USESTAMP flag is set, we take that 50357c478bd9Sstevel@tonic-gate * route into account only if this time interval 50367c478bd9Sstevel@tonic-gate * exceeds ip_multirt_resolution_interval; 50377c478bd9Sstevel@tonic-gate * this prevents us from attempting to resolve a 50387c478bd9Sstevel@tonic-gate * broken route upon each sending of a packet. 50397c478bd9Sstevel@tonic-gate */ 50407c478bd9Sstevel@tonic-gate delta = lbolt - fire->ire_last_used_time; 50417c478bd9Sstevel@tonic-gate delta = TICK_TO_MSEC(delta); 50427c478bd9Sstevel@tonic-gate 5043f4b3ec61Sdh155122 res = (boolean_t)((delta > 5044f4b3ec61Sdh155122 ipst->ips_ip_multirt_resolution_interval) || 50457c478bd9Sstevel@tonic-gate (!(flags & MULTIRT_USESTAMP))); 50467c478bd9Sstevel@tonic-gate 50477c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: fire %p, delta %lu, " 50487c478bd9Sstevel@tonic-gate "res %d\n", 50497c478bd9Sstevel@tonic-gate (void *)fire, delta, res)); 50507c478bd9Sstevel@tonic-gate 50517c478bd9Sstevel@tonic-gate if (res) { 50527c478bd9Sstevel@tonic-gate /* 50537c478bd9Sstevel@tonic-gate * We are here if MULTIRT_USESTAMP flag is set 50547c478bd9Sstevel@tonic-gate * and the resolver for fire's gateway 50557c478bd9Sstevel@tonic-gate * has not been tried since 50567c478bd9Sstevel@tonic-gate * ip_multirt_resolution_interval, or if 50577c478bd9Sstevel@tonic-gate * MULTIRT_USESTAMP is not set but gw_ire did 50587c478bd9Sstevel@tonic-gate * not fill the conditions for MULTIRT_CACHEGW, 50597c478bd9Sstevel@tonic-gate * or if neither MULTIRT_USESTAMP nor 50607c478bd9Sstevel@tonic-gate * MULTIRT_CACHEGW are set. 50617c478bd9Sstevel@tonic-gate */ 50627c478bd9Sstevel@tonic-gate if (gw_ire != NULL) { 50637c478bd9Sstevel@tonic-gate if (best_fire == NULL) { 50647c478bd9Sstevel@tonic-gate ASSERT(best_cire == NULL); 50657c478bd9Sstevel@tonic-gate 50667c478bd9Sstevel@tonic-gate best_fire = fire; 50677c478bd9Sstevel@tonic-gate best_cire = gw_ire; 50687c478bd9Sstevel@tonic-gate 50697c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup:" 50707c478bd9Sstevel@tonic-gate "found candidate " 50717c478bd9Sstevel@tonic-gate "best_fire %p, " 50727c478bd9Sstevel@tonic-gate "best_cire %p\n", 50737c478bd9Sstevel@tonic-gate (void *)best_fire, 50747c478bd9Sstevel@tonic-gate (void *)best_cire)); 50757c478bd9Sstevel@tonic-gate 50767c478bd9Sstevel@tonic-gate /* 50777c478bd9Sstevel@tonic-gate * If MULTIRT_CACHEGW is not 50787c478bd9Sstevel@tonic-gate * set, we ignore the top 50797c478bd9Sstevel@tonic-gate * priority ires that can 50807c478bd9Sstevel@tonic-gate * be resolved without any 50817c478bd9Sstevel@tonic-gate * call to the resolver; 50827c478bd9Sstevel@tonic-gate * In that case, there is 50837c478bd9Sstevel@tonic-gate * actually no need 50847c478bd9Sstevel@tonic-gate * to continue the loop. 50857c478bd9Sstevel@tonic-gate */ 50867c478bd9Sstevel@tonic-gate if (!(flags & 50877c478bd9Sstevel@tonic-gate MULTIRT_CACHEGW)) { 50887c478bd9Sstevel@tonic-gate break; 50897c478bd9Sstevel@tonic-gate } 50907c478bd9Sstevel@tonic-gate continue; 50917c478bd9Sstevel@tonic-gate } 50927c478bd9Sstevel@tonic-gate } else { 50937c478bd9Sstevel@tonic-gate /* 50947c478bd9Sstevel@tonic-gate * No resolver for the gateway: the 50957c478bd9Sstevel@tonic-gate * route is not resolvable. 50967c478bd9Sstevel@tonic-gate * If the MULTIRT_SETSTAMP flag is 50977c478bd9Sstevel@tonic-gate * set, we stamp the IRE_HOST ire, 50987c478bd9Sstevel@tonic-gate * so we will not select it again 50997c478bd9Sstevel@tonic-gate * during this resolution interval. 51007c478bd9Sstevel@tonic-gate */ 51017c478bd9Sstevel@tonic-gate if (flags & MULTIRT_SETSTAMP) 51027c478bd9Sstevel@tonic-gate fire->ire_last_used_time = 51037c478bd9Sstevel@tonic-gate lbolt; 51047c478bd9Sstevel@tonic-gate } 51057c478bd9Sstevel@tonic-gate } 51067c478bd9Sstevel@tonic-gate 51077c478bd9Sstevel@tonic-gate if (gw_ire != NULL) 51087c478bd9Sstevel@tonic-gate ire_refrele(gw_ire); 51097c478bd9Sstevel@tonic-gate } 51107c478bd9Sstevel@tonic-gate } else { /* CLASSD(dst) */ 51117c478bd9Sstevel@tonic-gate 51127c478bd9Sstevel@tonic-gate for (fire = first_fire; 51137c478bd9Sstevel@tonic-gate fire != NULL; 51147c478bd9Sstevel@tonic-gate fire = fire->ire_next) { 51157c478bd9Sstevel@tonic-gate 51167c478bd9Sstevel@tonic-gate if (!(fire->ire_flags & RTF_MULTIRT)) 51177c478bd9Sstevel@tonic-gate continue; 51187c478bd9Sstevel@tonic-gate if (fire->ire_addr != dst) 51197c478bd9Sstevel@tonic-gate continue; 51207c478bd9Sstevel@tonic-gate 512145916cd2Sjpk if (fire->ire_gw_secattr != NULL && 512245916cd2Sjpk tsol_ire_match_gwattr(fire, tsl) != 0) { 512345916cd2Sjpk continue; 512445916cd2Sjpk } 512545916cd2Sjpk 51267c478bd9Sstevel@tonic-gate already_resolved = B_FALSE; 51277c478bd9Sstevel@tonic-gate 51287c478bd9Sstevel@tonic-gate gw = fire->ire_gateway_addr; 51297c478bd9Sstevel@tonic-gate 51307c478bd9Sstevel@tonic-gate gw_ire = ire_ftable_lookup(gw, 0, 0, IRE_INTERFACE, 513145916cd2Sjpk NULL, NULL, ALL_ZONES, 0, tsl, 513245916cd2Sjpk MATCH_IRE_RECURSIVE | MATCH_IRE_TYPE | 5133f4b3ec61Sdh155122 MATCH_IRE_SECATTR, ipst); 51347c478bd9Sstevel@tonic-gate 51357c478bd9Sstevel@tonic-gate /* No resolver for the gateway; we skip this ire. */ 51367c478bd9Sstevel@tonic-gate if (gw_ire == NULL) { 51377c478bd9Sstevel@tonic-gate continue; 51387c478bd9Sstevel@tonic-gate } 5139c793af95Ssangeeta ASSERT(gw_ire->ire_nce == NULL || 5140c793af95Ssangeeta gw_ire->ire_nce->nce_state == ND_REACHABLE); 51417c478bd9Sstevel@tonic-gate 51427c478bd9Sstevel@tonic-gate if (first_cire != NULL) { 51437c478bd9Sstevel@tonic-gate 51447c478bd9Sstevel@tonic-gate IRB_REFHOLD(cirb); 51457c478bd9Sstevel@tonic-gate /* 51467c478bd9Sstevel@tonic-gate * For all IRE_CACHE ires for that 51477c478bd9Sstevel@tonic-gate * destination. 51487c478bd9Sstevel@tonic-gate */ 51497c478bd9Sstevel@tonic-gate for (cire = first_cire; 51507c478bd9Sstevel@tonic-gate cire != NULL; 51517c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 51527c478bd9Sstevel@tonic-gate 51537c478bd9Sstevel@tonic-gate if (!(cire->ire_flags & RTF_MULTIRT)) 51547c478bd9Sstevel@tonic-gate continue; 51557c478bd9Sstevel@tonic-gate if (cire->ire_addr != dst) 51567c478bd9Sstevel@tonic-gate continue; 51577c478bd9Sstevel@tonic-gate if (cire->ire_marks & 51587c478bd9Sstevel@tonic-gate (IRE_MARK_CONDEMNED | 51597c478bd9Sstevel@tonic-gate IRE_MARK_HIDDEN)) 51607c478bd9Sstevel@tonic-gate continue; 51617c478bd9Sstevel@tonic-gate 516245916cd2Sjpk if (cire->ire_gw_secattr != NULL && 516345916cd2Sjpk tsol_ire_match_gwattr(cire, 516445916cd2Sjpk tsl) != 0) { 516545916cd2Sjpk continue; 516645916cd2Sjpk } 516745916cd2Sjpk 51687c478bd9Sstevel@tonic-gate /* 51697c478bd9Sstevel@tonic-gate * Cache entries are linked to the 51707c478bd9Sstevel@tonic-gate * parent routes using the parent handle 51717c478bd9Sstevel@tonic-gate * (ire_phandle). If no cache entry has 51727c478bd9Sstevel@tonic-gate * the same handle as fire, fire is 51737c478bd9Sstevel@tonic-gate * still unresolved. 51747c478bd9Sstevel@tonic-gate */ 51757c478bd9Sstevel@tonic-gate ASSERT(cire->ire_phandle != 0); 51767c478bd9Sstevel@tonic-gate if (cire->ire_phandle == 51777c478bd9Sstevel@tonic-gate fire->ire_phandle) { 51787c478bd9Sstevel@tonic-gate already_resolved = B_TRUE; 51797c478bd9Sstevel@tonic-gate break; 51807c478bd9Sstevel@tonic-gate } 51817c478bd9Sstevel@tonic-gate } 51827c478bd9Sstevel@tonic-gate IRB_REFRELE(cirb); 51837c478bd9Sstevel@tonic-gate } 51847c478bd9Sstevel@tonic-gate 51857c478bd9Sstevel@tonic-gate /* 51867c478bd9Sstevel@tonic-gate * This route is already resolved; proceed with 51877c478bd9Sstevel@tonic-gate * next one. 51887c478bd9Sstevel@tonic-gate */ 51897c478bd9Sstevel@tonic-gate if (already_resolved) { 51907c478bd9Sstevel@tonic-gate ire_refrele(gw_ire); 51917c478bd9Sstevel@tonic-gate continue; 51927c478bd9Sstevel@tonic-gate } 51937c478bd9Sstevel@tonic-gate 51947c478bd9Sstevel@tonic-gate /* 51957c478bd9Sstevel@tonic-gate * Compute the time elapsed since our preceding 51967c478bd9Sstevel@tonic-gate * attempt to resolve that route. 51977c478bd9Sstevel@tonic-gate * If the MULTIRT_USESTAMP flag is set, we take 51987c478bd9Sstevel@tonic-gate * that route into account only if this time 51997c478bd9Sstevel@tonic-gate * interval exceeds ip_multirt_resolution_interval; 52007c478bd9Sstevel@tonic-gate * this prevents us from attempting to resolve a 52017c478bd9Sstevel@tonic-gate * broken route upon each sending of a packet. 52027c478bd9Sstevel@tonic-gate */ 52037c478bd9Sstevel@tonic-gate delta = lbolt - fire->ire_last_used_time; 52047c478bd9Sstevel@tonic-gate delta = TICK_TO_MSEC(delta); 52057c478bd9Sstevel@tonic-gate 5206f4b3ec61Sdh155122 res = (boolean_t)((delta > 5207f4b3ec61Sdh155122 ipst->ips_ip_multirt_resolution_interval) || 52087c478bd9Sstevel@tonic-gate (!(flags & MULTIRT_USESTAMP))); 52097c478bd9Sstevel@tonic-gate 52107c478bd9Sstevel@tonic-gate ip3dbg(("ire_multirt_lookup: fire %p, delta %lx, " 52117c478bd9Sstevel@tonic-gate "flags %04x, res %d\n", 52127c478bd9Sstevel@tonic-gate (void *)fire, delta, flags, res)); 52137c478bd9Sstevel@tonic-gate 52147c478bd9Sstevel@tonic-gate if (res) { 52157c478bd9Sstevel@tonic-gate if (best_cire != NULL) { 52167c478bd9Sstevel@tonic-gate /* 52177c478bd9Sstevel@tonic-gate * Release the resolver associated 52187c478bd9Sstevel@tonic-gate * to the preceding candidate best 52197c478bd9Sstevel@tonic-gate * ire, if any. 52207c478bd9Sstevel@tonic-gate */ 52217c478bd9Sstevel@tonic-gate ire_refrele(best_cire); 52227c478bd9Sstevel@tonic-gate ASSERT(best_fire != NULL); 52237c478bd9Sstevel@tonic-gate } 52247c478bd9Sstevel@tonic-gate best_fire = fire; 52257c478bd9Sstevel@tonic-gate best_cire = gw_ire; 52267c478bd9Sstevel@tonic-gate continue; 52277c478bd9Sstevel@tonic-gate } 52287c478bd9Sstevel@tonic-gate 52297c478bd9Sstevel@tonic-gate ire_refrele(gw_ire); 52307c478bd9Sstevel@tonic-gate } 52317c478bd9Sstevel@tonic-gate } 52327c478bd9Sstevel@tonic-gate 52337c478bd9Sstevel@tonic-gate if (best_fire != NULL) { 52347c478bd9Sstevel@tonic-gate IRE_REFHOLD(best_fire); 52357c478bd9Sstevel@tonic-gate } 52367c478bd9Sstevel@tonic-gate IRB_REFRELE(firb); 52377c478bd9Sstevel@tonic-gate 52387c478bd9Sstevel@tonic-gate /* Release the first IRE_CACHE we initially looked up, if any. */ 52397c478bd9Sstevel@tonic-gate if (first_cire != NULL) 52407c478bd9Sstevel@tonic-gate ire_refrele(first_cire); 52417c478bd9Sstevel@tonic-gate 52427c478bd9Sstevel@tonic-gate /* Found a resolvable route. */ 52437c478bd9Sstevel@tonic-gate if (best_fire != NULL) { 52447c478bd9Sstevel@tonic-gate ASSERT(best_cire != NULL); 52457c478bd9Sstevel@tonic-gate 52467c478bd9Sstevel@tonic-gate if (*fire_arg != NULL) 52477c478bd9Sstevel@tonic-gate ire_refrele(*fire_arg); 52487c478bd9Sstevel@tonic-gate if (*ire_arg != NULL) 52497c478bd9Sstevel@tonic-gate ire_refrele(*ire_arg); 52507c478bd9Sstevel@tonic-gate 52517c478bd9Sstevel@tonic-gate /* 52527c478bd9Sstevel@tonic-gate * Update the passed-in arguments with the 52537c478bd9Sstevel@tonic-gate * resolvable multirt route we found. 52547c478bd9Sstevel@tonic-gate */ 52557c478bd9Sstevel@tonic-gate *fire_arg = best_fire; 52567c478bd9Sstevel@tonic-gate *ire_arg = best_cire; 52577c478bd9Sstevel@tonic-gate 52587c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: returning B_TRUE, " 52597c478bd9Sstevel@tonic-gate "*fire_arg %p, *ire_arg %p\n", 52607c478bd9Sstevel@tonic-gate (void *)best_fire, (void *)best_cire)); 52617c478bd9Sstevel@tonic-gate 52627c478bd9Sstevel@tonic-gate return (B_TRUE); 52637c478bd9Sstevel@tonic-gate } 52647c478bd9Sstevel@tonic-gate 52657c478bd9Sstevel@tonic-gate ASSERT(best_cire == NULL); 52667c478bd9Sstevel@tonic-gate 52677c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: returning B_FALSE, *fire_arg %p, " 52687c478bd9Sstevel@tonic-gate "*ire_arg %p\n", 52697c478bd9Sstevel@tonic-gate (void *)*fire_arg, (void *)*ire_arg)); 52707c478bd9Sstevel@tonic-gate 52717c478bd9Sstevel@tonic-gate /* No resolvable route. */ 52727c478bd9Sstevel@tonic-gate return (B_FALSE); 52737c478bd9Sstevel@tonic-gate } 52747c478bd9Sstevel@tonic-gate 52757c478bd9Sstevel@tonic-gate /* 52767c478bd9Sstevel@tonic-gate * IRE iterator for inbound and loopback broadcast processing. 52777c478bd9Sstevel@tonic-gate * Given an IRE_BROADCAST ire, walk the ires with the same destination 52787c478bd9Sstevel@tonic-gate * address, but skip over the passed-in ire. Returns the next ire without 52797c478bd9Sstevel@tonic-gate * a hold - assumes that the caller holds a reference on the IRE bucket. 52807c478bd9Sstevel@tonic-gate */ 52817c478bd9Sstevel@tonic-gate ire_t * 52827c478bd9Sstevel@tonic-gate ire_get_next_bcast_ire(ire_t *curr, ire_t *ire) 52837c478bd9Sstevel@tonic-gate { 52847c478bd9Sstevel@tonic-gate ill_t *ill; 52857c478bd9Sstevel@tonic-gate 52867c478bd9Sstevel@tonic-gate if (curr == NULL) { 52877c478bd9Sstevel@tonic-gate for (curr = ire->ire_bucket->irb_ire; curr != NULL; 52887c478bd9Sstevel@tonic-gate curr = curr->ire_next) { 52897c478bd9Sstevel@tonic-gate if (curr->ire_addr == ire->ire_addr) 52907c478bd9Sstevel@tonic-gate break; 52917c478bd9Sstevel@tonic-gate } 52927c478bd9Sstevel@tonic-gate } else { 52937c478bd9Sstevel@tonic-gate curr = curr->ire_next; 52947c478bd9Sstevel@tonic-gate } 52957c478bd9Sstevel@tonic-gate ill = ire_to_ill(ire); 52967c478bd9Sstevel@tonic-gate for (; curr != NULL; curr = curr->ire_next) { 52977c478bd9Sstevel@tonic-gate if (curr->ire_addr != ire->ire_addr) { 52987c478bd9Sstevel@tonic-gate /* 52997c478bd9Sstevel@tonic-gate * All the IREs to a given destination are contiguous; 53007c478bd9Sstevel@tonic-gate * break out once the address doesn't match. 53017c478bd9Sstevel@tonic-gate */ 53027c478bd9Sstevel@tonic-gate break; 53037c478bd9Sstevel@tonic-gate } 53047c478bd9Sstevel@tonic-gate if (curr == ire) { 53057c478bd9Sstevel@tonic-gate /* skip over the passed-in ire */ 53067c478bd9Sstevel@tonic-gate continue; 53077c478bd9Sstevel@tonic-gate } 53087c478bd9Sstevel@tonic-gate if ((curr->ire_stq != NULL && ire->ire_stq == NULL) || 53097c478bd9Sstevel@tonic-gate (curr->ire_stq == NULL && ire->ire_stq != NULL)) { 53107c478bd9Sstevel@tonic-gate /* 53117c478bd9Sstevel@tonic-gate * If the passed-in ire is loopback, skip over 53127c478bd9Sstevel@tonic-gate * non-loopback ires and vice versa. 53137c478bd9Sstevel@tonic-gate */ 53147c478bd9Sstevel@tonic-gate continue; 53157c478bd9Sstevel@tonic-gate } 53167c478bd9Sstevel@tonic-gate if (ire_to_ill(curr) != ill) { 53177c478bd9Sstevel@tonic-gate /* skip over IREs going through a different interface */ 53187c478bd9Sstevel@tonic-gate continue; 53197c478bd9Sstevel@tonic-gate } 53207c478bd9Sstevel@tonic-gate if (curr->ire_marks & IRE_MARK_CONDEMNED) { 53217c478bd9Sstevel@tonic-gate /* skip over deleted IREs */ 53227c478bd9Sstevel@tonic-gate continue; 53237c478bd9Sstevel@tonic-gate } 53247c478bd9Sstevel@tonic-gate return (curr); 53257c478bd9Sstevel@tonic-gate } 53267c478bd9Sstevel@tonic-gate return (NULL); 53277c478bd9Sstevel@tonic-gate } 53287c478bd9Sstevel@tonic-gate 53296a8288c7Scarlsonj #ifdef DEBUG 53307c478bd9Sstevel@tonic-gate void 53317c478bd9Sstevel@tonic-gate ire_trace_ref(ire_t *ire) 53327c478bd9Sstevel@tonic-gate { 53337c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 53346a8288c7Scarlsonj if (ire->ire_trace_disable) { 53357c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 53367c478bd9Sstevel@tonic-gate return; 53377c478bd9Sstevel@tonic-gate } 53386a8288c7Scarlsonj 53396a8288c7Scarlsonj if (th_trace_ref(ire, ire->ire_ipst)) { 53406a8288c7Scarlsonj mutex_exit(&ire->ire_lock); 53416a8288c7Scarlsonj } else { 53427c478bd9Sstevel@tonic-gate ire->ire_trace_disable = B_TRUE; 53437c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 53446a8288c7Scarlsonj ire_trace_cleanup(ire); 53457c478bd9Sstevel@tonic-gate } 53467c478bd9Sstevel@tonic-gate } 53477c478bd9Sstevel@tonic-gate 53487c478bd9Sstevel@tonic-gate void 53497c478bd9Sstevel@tonic-gate ire_untrace_ref(ire_t *ire) 53507c478bd9Sstevel@tonic-gate { 53517c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 53526a8288c7Scarlsonj if (!ire->ire_trace_disable) 53536a8288c7Scarlsonj th_trace_unref(ire); 53547c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 53557c478bd9Sstevel@tonic-gate } 53567c478bd9Sstevel@tonic-gate 53577c478bd9Sstevel@tonic-gate static void 53586a8288c7Scarlsonj ire_trace_cleanup(const ire_t *ire) 53597c478bd9Sstevel@tonic-gate { 53606a8288c7Scarlsonj th_trace_cleanup(ire, ire->ire_trace_disable); 53617c478bd9Sstevel@tonic-gate } 53626a8288c7Scarlsonj #endif /* DEBUG */ 5363c793af95Ssangeeta 5364c793af95Ssangeeta /* 5365c793af95Ssangeeta * Generate a message chain with an arp request to resolve the in_ire. 5366c793af95Ssangeeta * It is assumed that in_ire itself is currently in the ire cache table, 5367c793af95Ssangeeta * so we create a fake_ire filled with enough information about ire_addr etc. 5368c793af95Ssangeeta * to retrieve in_ire when the DL_UNITDATA response from the resolver 5369c793af95Ssangeeta * comes back. The fake_ire itself is created by calling esballoc with 5370c793af95Ssangeeta * the fr_rtnp (free routine) set to ire_freemblk. This routine will be 5371c793af95Ssangeeta * invoked when the mblk containing fake_ire is freed. 5372c793af95Ssangeeta */ 5373c793af95Ssangeeta void 5374c793af95Ssangeeta ire_arpresolve(ire_t *in_ire, ill_t *dst_ill) 5375c793af95Ssangeeta { 5376c793af95Ssangeeta areq_t *areq; 5377c793af95Ssangeeta ipaddr_t *addrp; 537854da8755Ssowmini mblk_t *ire_mp, *areq_mp; 5379c793af95Ssangeeta ire_t *ire, *buf; 5380c793af95Ssangeeta size_t bufsize; 5381c793af95Ssangeeta frtn_t *frtnp; 5382c793af95Ssangeeta ill_t *ill; 5383f4b3ec61Sdh155122 ip_stack_t *ipst = dst_ill->ill_ipst; 5384c793af95Ssangeeta 5385c793af95Ssangeeta /* 5386c793af95Ssangeeta * Construct message chain for the resolver 5387c793af95Ssangeeta * of the form: 5388c793af95Ssangeeta * ARP_REQ_MBLK-->IRE_MBLK 5389c793af95Ssangeeta * 5390c793af95Ssangeeta * NOTE : If the response does not 5391c793af95Ssangeeta * come back, ARP frees the packet. For this reason, 5392c793af95Ssangeeta * we can't REFHOLD the bucket of save_ire to prevent 5393c793af95Ssangeeta * deletions. We may not be able to REFRELE the bucket 5394c793af95Ssangeeta * if the response never comes back. Thus, before 5395c793af95Ssangeeta * adding the ire, ire_add_v4 will make sure that the 5396c793af95Ssangeeta * interface route does not get deleted. This is the 5397c793af95Ssangeeta * only case unlike ip_newroute_v6, ip_newroute_ipif_v6 5398c793af95Ssangeeta * where we can always prevent deletions because of 5399c793af95Ssangeeta * the synchronous nature of adding IRES i.e 5400c793af95Ssangeeta * ire_add_then_send is called after creating the IRE. 5401c793af95Ssangeeta */ 5402c793af95Ssangeeta 5403c793af95Ssangeeta /* 5404c793af95Ssangeeta * We use esballoc to allocate the second part(the ire_t size mblk) 5405c793af95Ssangeeta * of the message chain depicted above. THis mblk will be freed 5406c793af95Ssangeeta * by arp when there is a timeout, and otherwise passed to IP 5407c793af95Ssangeeta * and IP will * free it after processing the ARP response. 5408c793af95Ssangeeta */ 5409c793af95Ssangeeta 5410c793af95Ssangeeta bufsize = sizeof (ire_t) + sizeof (frtn_t); 5411c793af95Ssangeeta buf = kmem_alloc(bufsize, KM_NOSLEEP); 5412c793af95Ssangeeta if (buf == NULL) { 5413c793af95Ssangeeta ip1dbg(("ire_arpresolver:alloc buffer failed\n ")); 5414c793af95Ssangeeta return; 5415c793af95Ssangeeta } 5416c793af95Ssangeeta frtnp = (frtn_t *)(buf + 1); 5417c793af95Ssangeeta frtnp->free_arg = (caddr_t)buf; 5418c793af95Ssangeeta frtnp->free_func = ire_freemblk; 5419c793af95Ssangeeta 5420c793af95Ssangeeta ire_mp = esballoc((unsigned char *)buf, bufsize, BPRI_MED, frtnp); 5421c793af95Ssangeeta 5422c793af95Ssangeeta if (ire_mp == NULL) { 5423c793af95Ssangeeta ip1dbg(("ire_arpresolve: esballoc failed\n")); 5424c793af95Ssangeeta kmem_free(buf, bufsize); 5425c793af95Ssangeeta return; 5426c793af95Ssangeeta } 5427c793af95Ssangeeta ASSERT(in_ire->ire_nce != NULL); 542854da8755Ssowmini areq_mp = copyb(dst_ill->ill_resolver_mp); 542954da8755Ssowmini if (areq_mp == NULL) { 5430c793af95Ssangeeta kmem_free(buf, bufsize); 5431c793af95Ssangeeta return; 5432c793af95Ssangeeta } 5433c793af95Ssangeeta 5434c793af95Ssangeeta ire_mp->b_datap->db_type = IRE_ARPRESOLVE_TYPE; 5435c793af95Ssangeeta ire = (ire_t *)buf; 5436c793af95Ssangeeta /* 5437c793af95Ssangeeta * keep enough info in the fake ire so that we can pull up 5438c793af95Ssangeeta * the incomplete ire (in_ire) after result comes back from 5439c793af95Ssangeeta * arp and make it complete. 5440c793af95Ssangeeta */ 5441c793af95Ssangeeta *ire = ire_null; 5442c793af95Ssangeeta ire->ire_u = in_ire->ire_u; 5443c793af95Ssangeeta ire->ire_ipif_seqid = in_ire->ire_ipif_seqid; 5444c793af95Ssangeeta ire->ire_ipif = in_ire->ire_ipif; 5445c793af95Ssangeeta ire->ire_stq = in_ire->ire_stq; 5446c793af95Ssangeeta ill = ire_to_ill(ire); 5447c793af95Ssangeeta ire->ire_stq_ifindex = ill->ill_phyint->phyint_ifindex; 5448c793af95Ssangeeta ire->ire_zoneid = in_ire->ire_zoneid; 5449f4b3ec61Sdh155122 ire->ire_ipst = ipst; 5450f4b3ec61Sdh155122 5451c793af95Ssangeeta /* 5452c793af95Ssangeeta * ire_freemblk will be called when ire_mp is freed, both for 5453c793af95Ssangeeta * successful and failed arp resolution. IRE_MARK_UNCACHED will be set 5454c793af95Ssangeeta * when the arp resolution failed. 5455c793af95Ssangeeta */ 5456c793af95Ssangeeta ire->ire_marks |= IRE_MARK_UNCACHED; 5457c793af95Ssangeeta ire->ire_mp = ire_mp; 5458c793af95Ssangeeta ire_mp->b_wptr = (uchar_t *)&ire[1]; 5459c793af95Ssangeeta ire_mp->b_cont = NULL; 546054da8755Ssowmini linkb(areq_mp, ire_mp); 5461c793af95Ssangeeta 5462c793af95Ssangeeta /* 5463c793af95Ssangeeta * Fill in the source and dest addrs for the resolver. 5464c793af95Ssangeeta * NOTE: this depends on memory layouts imposed by 5465c793af95Ssangeeta * ill_init(). 5466c793af95Ssangeeta */ 546754da8755Ssowmini areq = (areq_t *)areq_mp->b_rptr; 5468c793af95Ssangeeta addrp = (ipaddr_t *)((char *)areq + areq->areq_sender_addr_offset); 5469c793af95Ssangeeta *addrp = ire->ire_src_addr; 5470c793af95Ssangeeta 5471c793af95Ssangeeta addrp = (ipaddr_t *)((char *)areq + areq->areq_target_addr_offset); 5472c793af95Ssangeeta if (ire->ire_gateway_addr != INADDR_ANY) { 5473c793af95Ssangeeta *addrp = ire->ire_gateway_addr; 5474c793af95Ssangeeta } else { 5475c793af95Ssangeeta *addrp = ire->ire_addr; 5476c793af95Ssangeeta } 5477c793af95Ssangeeta 5478c793af95Ssangeeta /* Up to the resolver. */ 5479c793af95Ssangeeta if (canputnext(dst_ill->ill_rq)) { 548054da8755Ssowmini putnext(dst_ill->ill_rq, areq_mp); 5481c793af95Ssangeeta } else { 548254da8755Ssowmini freemsg(areq_mp); 5483c793af95Ssangeeta } 5484c793af95Ssangeeta } 5485c793af95Ssangeeta 5486c793af95Ssangeeta /* 5487c793af95Ssangeeta * Esballoc free function for AR_ENTRY_QUERY request to clean up any 5488c793af95Ssangeeta * unresolved ire_t and/or nce_t structures when ARP resolution fails. 5489c793af95Ssangeeta * 5490c793af95Ssangeeta * This function can be called by ARP via free routine for ire_mp or 5491c793af95Ssangeeta * by IPv4(both host and forwarding path) via ire_delete 5492c793af95Ssangeeta * in case ARP resolution fails. 5493c793af95Ssangeeta * NOTE: Since IP is MT, ARP can call into IP but not vice versa 5494c793af95Ssangeeta * (for IP to talk to ARP, it still has to send AR* messages). 5495c793af95Ssangeeta * 5496c793af95Ssangeeta * Note that the ARP/IP merge should replace the functioanlity by providing 5497c793af95Ssangeeta * direct function calls to clean up unresolved entries in ire/nce lists. 5498c793af95Ssangeeta */ 5499c793af95Ssangeeta void 5500c793af95Ssangeeta ire_freemblk(ire_t *ire_mp) 5501c793af95Ssangeeta { 5502c793af95Ssangeeta nce_t *nce = NULL; 5503c793af95Ssangeeta ill_t *ill; 5504f4b3ec61Sdh155122 ip_stack_t *ipst; 5505c793af95Ssangeeta 5506c793af95Ssangeeta ASSERT(ire_mp != NULL); 5507c793af95Ssangeeta 5508c793af95Ssangeeta if ((ire_mp->ire_addr == NULL) && (ire_mp->ire_gateway_addr == NULL)) { 5509c793af95Ssangeeta ip1dbg(("ire_freemblk(0x%p) ire_addr is NULL\n", 5510c793af95Ssangeeta (void *)ire_mp)); 5511c793af95Ssangeeta goto cleanup; 5512c793af95Ssangeeta } 5513c793af95Ssangeeta if ((ire_mp->ire_marks & IRE_MARK_UNCACHED) == 0) { 5514c793af95Ssangeeta goto cleanup; /* everything succeeded. just free and return */ 5515c793af95Ssangeeta } 5516c793af95Ssangeeta 5517c793af95Ssangeeta /* 5518c793af95Ssangeeta * the arp information corresponding to this ire_mp was not 5519c793af95Ssangeeta * transferred to a ire_cache entry. Need 5520c793af95Ssangeeta * to clean up incomplete ire's and nce, if necessary. 5521c793af95Ssangeeta */ 5522c793af95Ssangeeta ASSERT(ire_mp->ire_stq != NULL); 5523c793af95Ssangeeta ASSERT(ire_mp->ire_stq_ifindex != 0); 5524f4b3ec61Sdh155122 ASSERT(ire_mp->ire_ipst != NULL); 5525f4b3ec61Sdh155122 5526f4b3ec61Sdh155122 ipst = ire_mp->ire_ipst; 5527f4b3ec61Sdh155122 5528c793af95Ssangeeta /* 5529c793af95Ssangeeta * Get any nce's corresponding to this ire_mp. We first have to 5530c793af95Ssangeeta * make sure that the ill is still around. 5531c793af95Ssangeeta */ 5532f4b3ec61Sdh155122 ill = ill_lookup_on_ifindex(ire_mp->ire_stq_ifindex, 5533f4b3ec61Sdh155122 B_FALSE, NULL, NULL, NULL, NULL, ipst); 5534c793af95Ssangeeta if (ill == NULL || (ire_mp->ire_stq != ill->ill_wq) || 5535c793af95Ssangeeta (ill->ill_state_flags & ILL_CONDEMNED)) { 5536c793af95Ssangeeta /* 5537c793af95Ssangeeta * ill went away. no nce to clean up. 5538c793af95Ssangeeta * Note that the ill_state_flags could be set to 5539c793af95Ssangeeta * ILL_CONDEMNED after this point, but if we know 5540c793af95Ssangeeta * that it is CONDEMNED now, we just bail out quickly. 5541c793af95Ssangeeta */ 5542c793af95Ssangeeta if (ill != NULL) 5543c793af95Ssangeeta ill_refrele(ill); 5544c793af95Ssangeeta goto cleanup; 5545c793af95Ssangeeta } 5546c793af95Ssangeeta nce = ndp_lookup_v4(ill, 5547c793af95Ssangeeta ((ire_mp->ire_gateway_addr != INADDR_ANY) ? 5548c793af95Ssangeeta &ire_mp->ire_gateway_addr : &ire_mp->ire_addr), 5549c793af95Ssangeeta B_FALSE); 5550c793af95Ssangeeta ill_refrele(ill); 5551c793af95Ssangeeta 5552c793af95Ssangeeta if ((nce != NULL) && (nce->nce_state != ND_REACHABLE)) { 5553c793af95Ssangeeta /* 5554c793af95Ssangeeta * some incomplete nce was found. 5555c793af95Ssangeeta */ 5556c793af95Ssangeeta DTRACE_PROBE2(ire__freemblk__arp__resolv__fail, 5557c793af95Ssangeeta nce_t *, nce, ire_t *, ire_mp); 5558c793af95Ssangeeta /* 5559c793af95Ssangeeta * Send the icmp_unreachable messages for the queued mblks in 5560c793af95Ssangeeta * ire->ire_nce->nce_qd_mp, since ARP resolution failed 5561c793af95Ssangeeta * for this ire 5562c793af95Ssangeeta */ 5563c793af95Ssangeeta arp_resolv_failed(nce); 5564c793af95Ssangeeta /* 5565c793af95Ssangeeta * Delete the nce and clean up all ire's pointing at this nce 5566c793af95Ssangeeta * in the cachetable 5567c793af95Ssangeeta */ 5568c793af95Ssangeeta ndp_delete(nce); 5569c793af95Ssangeeta } 5570c793af95Ssangeeta if (nce != NULL) 5571c793af95Ssangeeta NCE_REFRELE(nce); /* release the ref taken by ndp_lookup_v4 */ 5572c793af95Ssangeeta 5573c793af95Ssangeeta cleanup: 5574c793af95Ssangeeta /* 5575c793af95Ssangeeta * Get rid of the ire buffer 5576c793af95Ssangeeta * We call kmem_free here(instead of ire_delete()), since 5577c793af95Ssangeeta * this is the freeb's callback. 5578c793af95Ssangeeta */ 5579c793af95Ssangeeta kmem_free(ire_mp, sizeof (ire_t) + sizeof (frtn_t)); 5580c793af95Ssangeeta } 5581c793af95Ssangeeta 5582c0861880Ssangeeta /* 558354da8755Ssowmini * find, or create if needed, a neighbor cache entry nce_t for IRE_CACHE and 558454da8755Ssowmini * non-loopback IRE_BROADCAST ire's. 558554da8755Ssowmini * 558654da8755Ssowmini * If a neighbor-cache entry has to be created (i.e., one does not already 558754da8755Ssowmini * exist in the nce list) the nce_res_mp and nce_state of the neighbor cache 558854da8755Ssowmini * entry are initialized in ndp_add_v4(). These values are picked from 558954da8755Ssowmini * the src_nce, if one is passed in. Otherwise (if src_nce == NULL) the 559054da8755Ssowmini * ire->ire_type and the outgoing interface (ire_to_ill(ire)) values 559154da8755Ssowmini * determine the {nce_state, nce_res_mp} of the nce_t created. All 559254da8755Ssowmini * IRE_BROADCAST entries have nce_state = ND_REACHABLE, and the nce_res_mp 559354da8755Ssowmini * is set to the ill_bcast_mp of the outgoing inerface. For unicast ire 559454da8755Ssowmini * entries, 559554da8755Ssowmini * - if the outgoing interface is of type IRE_IF_RESOLVER, a newly created 559654da8755Ssowmini * nce_t will have a null nce_res_mp, and will be in the ND_INITIAL state. 559754da8755Ssowmini * - if the outgoing interface is a IRE_IF_NORESOLVER interface, no link 559854da8755Ssowmini * layer resolution is necessary, so that the nce_t will be in the 559954da8755Ssowmini * ND_REACHABLE state and the nce_res_mp will have a copy of the 560054da8755Ssowmini * ill_resolver_mp of the outgoing interface. 560154da8755Ssowmini * 560254da8755Ssowmini * The link layer information needed for broadcast addresses, and for 560354da8755Ssowmini * packets sent on IRE_IF_NORESOLVER interfaces is a constant mapping that 560454da8755Ssowmini * never needs re-verification for the lifetime of the nce_t. These are 560554da8755Ssowmini * therefore marked NCE_F_PERMANENT, and never allowed to expire via 560654da8755Ssowmini * NCE_EXPIRED. 560754da8755Ssowmini * 560854da8755Ssowmini * IRE_CACHE ire's contain the information for the nexthop (ire_gateway_addr) 560954da8755Ssowmini * in the case of indirect routes, and for the dst itself (ire_addr) in the 5610c793af95Ssangeeta * case of direct routes, with the nce_res_mp containing a template 5611c793af95Ssangeeta * DL_UNITDATA request. 5612c793af95Ssangeeta * 5613c793af95Ssangeeta * The actual association of the ire_nce to the nce created here is 5614c793af95Ssangeeta * typically done in ire_add_v4 for IRE_CACHE entries. Exceptions 5615c793af95Ssangeeta * to this rule are SO_DONTROUTE ire's (IRE_MARK_NO_ADD), for which 56165c0b7edeSseb * the ire_nce assignment is done in ire_add_then_send. 5617c793af95Ssangeeta */ 5618c793af95Ssangeeta int 561954da8755Ssowmini ire_nce_init(ire_t *ire, nce_t *src_nce) 5620c793af95Ssangeeta { 562154da8755Ssowmini in_addr_t addr4; 5622c793af95Ssangeeta int err; 562354da8755Ssowmini nce_t *nce = NULL; 5624c793af95Ssangeeta ill_t *ire_ill; 562554da8755Ssowmini uint16_t nce_flags = 0; 5626f4b3ec61Sdh155122 ip_stack_t *ipst; 5627c793af95Ssangeeta 562854da8755Ssowmini if (ire->ire_stq == NULL) 5629c793af95Ssangeeta return (0); /* no need to create nce for local/loopback */ 5630c793af95Ssangeeta 5631c793af95Ssangeeta switch (ire->ire_type) { 5632c793af95Ssangeeta case IRE_CACHE: 5633c793af95Ssangeeta if (ire->ire_gateway_addr != INADDR_ANY) 5634c793af95Ssangeeta addr4 = ire->ire_gateway_addr; /* 'G' route */ 5635c793af95Ssangeeta else 5636c793af95Ssangeeta addr4 = ire->ire_addr; /* direct route */ 5637c793af95Ssangeeta break; 5638c793af95Ssangeeta case IRE_BROADCAST: 5639c793af95Ssangeeta addr4 = ire->ire_addr; 564054da8755Ssowmini nce_flags |= (NCE_F_PERMANENT|NCE_F_BCAST); 5641c793af95Ssangeeta break; 5642c793af95Ssangeeta default: 5643c793af95Ssangeeta return (0); 5644c793af95Ssangeeta } 5645c793af95Ssangeeta 5646c793af95Ssangeeta /* 5647c793af95Ssangeeta * ire_ipif is picked based on RTF_SETSRC, usesrc etc. 5648c793af95Ssangeeta * rules in ire_forward_src_ipif. We want the dlureq_mp 5649c793af95Ssangeeta * for the outgoing interface, which we get from the ire_stq. 5650c793af95Ssangeeta */ 5651c793af95Ssangeeta ire_ill = ire_to_ill(ire); 5652f4b3ec61Sdh155122 ipst = ire_ill->ill_ipst; 5653c793af95Ssangeeta 5654c793af95Ssangeeta /* 565554da8755Ssowmini * IRE_IF_NORESOLVER entries never need re-verification and 565654da8755Ssowmini * do not expire, so we mark them as NCE_F_PERMANENT. 5657c793af95Ssangeeta */ 565854da8755Ssowmini if (ire_ill->ill_net_type == IRE_IF_NORESOLVER) 565954da8755Ssowmini nce_flags |= NCE_F_PERMANENT; 5660c793af95Ssangeeta 5661b9c344b3Ssowmini retry_nce: 566254da8755Ssowmini err = ndp_lookup_then_add_v4(ire_ill, &addr4, nce_flags, 566354da8755Ssowmini &nce, src_nce); 5664c793af95Ssangeeta 566554da8755Ssowmini if (err == EEXIST && NCE_EXPIRED(nce, ipst)) { 5666b9c344b3Ssowmini /* 5667b9c344b3Ssowmini * We looked up an expired nce. 5668b9c344b3Ssowmini * Go back and try to create one again. 5669b9c344b3Ssowmini */ 567054da8755Ssowmini ndp_delete(nce); 567154da8755Ssowmini NCE_REFRELE(nce); 567254da8755Ssowmini nce = NULL; 5673b9c344b3Ssowmini goto retry_nce; 5674b9c344b3Ssowmini } 5675b9c344b3Ssowmini 567654da8755Ssowmini ip1dbg(("ire 0x%p addr 0x%lx type 0x%x; found nce 0x%p err %d\n", 567754da8755Ssowmini (void *)ire, (ulong_t)addr4, ire->ire_type, (void *)nce, err)); 5678c793af95Ssangeeta 5679c793af95Ssangeeta switch (err) { 5680c793af95Ssangeeta case 0: 5681c793af95Ssangeeta case EEXIST: 5682c793af95Ssangeeta /* 568354da8755Ssowmini * return a pointer to a newly created or existing nce_t; 5684c793af95Ssangeeta * note that the ire-nce mapping is many-one, i.e., 568554da8755Ssowmini * multiple ire's could point to the same nce_t. 5686c793af95Ssangeeta */ 5687c793af95Ssangeeta break; 5688c793af95Ssangeeta default: 5689c793af95Ssangeeta DTRACE_PROBE2(nce__init__fail, ill_t *, ire_ill, int, err); 5690c793af95Ssangeeta return (EINVAL); 5691c793af95Ssangeeta } 5692c793af95Ssangeeta if (ire->ire_type == IRE_BROADCAST) { 5693c793af95Ssangeeta /* 5694c793af95Ssangeeta * Two bcast ires are created for each interface; 5695c793af95Ssangeeta * 1. loopback copy (which does not have an 5696c793af95Ssangeeta * ire_stq, and therefore has no ire_nce), and, 5697c793af95Ssangeeta * 2. the non-loopback copy, which has the nce_res_mp 5698c793af95Ssangeeta * initialized to a copy of the ill_bcast_mp, and 5699c793af95Ssangeeta * is marked as ND_REACHABLE at this point. 5700c793af95Ssangeeta * This nce does not undergo any further state changes, 5701c793af95Ssangeeta * and exists as long as the interface is plumbed. 5702c793af95Ssangeeta * Note: we do the ire_nce assignment here for IRE_BROADCAST 5703c793af95Ssangeeta * because some functions like ill_mark_bcast() inline the 570454da8755Ssowmini * ire_add functionality. 5705c793af95Ssangeeta */ 570654da8755Ssowmini ire->ire_nce = nce; 5707c793af95Ssangeeta /* 5708c793af95Ssangeeta * We are associating this nce to the ire, 5709c793af95Ssangeeta * so change the nce ref taken in 5710c793af95Ssangeeta * ndp_lookup_then_add_v4() from 5711c793af95Ssangeeta * NCE_REFHOLD to NCE_REFHOLD_NOTR 5712c793af95Ssangeeta */ 5713c793af95Ssangeeta NCE_REFHOLD_TO_REFHOLD_NOTR(ire->ire_nce); 5714c793af95Ssangeeta } else { 5715c793af95Ssangeeta /* 5716c793af95Ssangeeta * We are not using this nce_t just yet so release 5717c793af95Ssangeeta * the ref taken in ndp_lookup_then_add_v4() 5718c793af95Ssangeeta */ 571954da8755Ssowmini NCE_REFRELE(nce); 5720c793af95Ssangeeta } 5721c793af95Ssangeeta return (0); 5722c793af95Ssangeeta } 5723