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 /* 22107875b0Ssangeeta * Copyright 2007 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> 3745916cd2Sjpk #include <sys/strsun.h> 387c478bd9Sstevel@tonic-gate #include <sys/ddi.h> 397c478bd9Sstevel@tonic-gate #include <sys/cmn_err.h> 407c478bd9Sstevel@tonic-gate #include <sys/policy.h> 417c478bd9Sstevel@tonic-gate 427c478bd9Sstevel@tonic-gate #include <sys/systm.h> 437c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 447c478bd9Sstevel@tonic-gate #include <sys/param.h> 457c478bd9Sstevel@tonic-gate #include <sys/socket.h> 467c478bd9Sstevel@tonic-gate #include <net/if.h> 477c478bd9Sstevel@tonic-gate #include <net/route.h> 487c478bd9Sstevel@tonic-gate #include <netinet/in.h> 497c478bd9Sstevel@tonic-gate #include <net/if_dl.h> 507c478bd9Sstevel@tonic-gate #include <netinet/ip6.h> 517c478bd9Sstevel@tonic-gate #include <netinet/icmp6.h> 527c478bd9Sstevel@tonic-gate 537c478bd9Sstevel@tonic-gate #include <inet/common.h> 547c478bd9Sstevel@tonic-gate #include <inet/mi.h> 557c478bd9Sstevel@tonic-gate #include <inet/ip.h> 567c478bd9Sstevel@tonic-gate #include <inet/ip6.h> 577c478bd9Sstevel@tonic-gate #include <inet/ip_ndp.h> 58c793af95Ssangeeta #include <inet/arp.h> 597c478bd9Sstevel@tonic-gate #include <inet/ip_if.h> 607c478bd9Sstevel@tonic-gate #include <inet/ip_ire.h> 61c793af95Ssangeeta #include <inet/ip_ftable.h> 627c478bd9Sstevel@tonic-gate #include <inet/ip_rts.h> 637c478bd9Sstevel@tonic-gate #include <inet/nd.h> 647c478bd9Sstevel@tonic-gate 657c478bd9Sstevel@tonic-gate #include <net/pfkeyv2.h> 667c478bd9Sstevel@tonic-gate #include <inet/ipsec_info.h> 677c478bd9Sstevel@tonic-gate #include <inet/sadb.h> 687c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 697c478bd9Sstevel@tonic-gate #include <inet/tcp.h> 707c478bd9Sstevel@tonic-gate #include <inet/ipclassifier.h> 717c478bd9Sstevel@tonic-gate #include <sys/zone.h> 72*f4b3ec61Sdh155122 #include <sys/cpuvar.h> 73*f4b3ec61Sdh155122 7445916cd2Sjpk #include <sys/tsol/label.h> 7545916cd2Sjpk #include <sys/tsol/tnet.h> 7645916cd2Sjpk 77c793af95Ssangeeta struct kmem_cache *rt_entry_cache; 78c793af95Ssangeeta 79c793af95Ssangeeta 807c478bd9Sstevel@tonic-gate /* 817c478bd9Sstevel@tonic-gate * Synchronization notes: 827c478bd9Sstevel@tonic-gate * 837c478bd9Sstevel@tonic-gate * The fields of the ire_t struct are protected in the following way : 847c478bd9Sstevel@tonic-gate * 857c478bd9Sstevel@tonic-gate * ire_next/ire_ptpn 867c478bd9Sstevel@tonic-gate * 877c478bd9Sstevel@tonic-gate * - bucket lock of the respective tables (cache or forwarding tables). 887c478bd9Sstevel@tonic-gate * 897c478bd9Sstevel@tonic-gate * ire_mp, ire_rfq, ire_stq, ire_u *except* ire_gateway_addr[v6], ire_mask, 907c478bd9Sstevel@tonic-gate * ire_type, ire_create_time, ire_masklen, ire_ipversion, ire_flags, ire_ipif, 917c478bd9Sstevel@tonic-gate * ire_ihandle, ire_phandle, ire_nce, ire_bucket, ire_in_ill, ire_in_src_addr 927c478bd9Sstevel@tonic-gate * 937c478bd9Sstevel@tonic-gate * - Set in ire_create_v4/v6 and never changes after that. Thus, 947c478bd9Sstevel@tonic-gate * we don't need a lock whenever these fields are accessed. 957c478bd9Sstevel@tonic-gate * 967c478bd9Sstevel@tonic-gate * - ire_bucket and ire_masklen (also set in ire_create) is set in 977c478bd9Sstevel@tonic-gate * ire_add_v4/ire_add_v6 before inserting in the bucket and never 987c478bd9Sstevel@tonic-gate * changes after that. Thus we don't need a lock whenever these 997c478bd9Sstevel@tonic-gate * fields are accessed. 1007c478bd9Sstevel@tonic-gate * 1017c478bd9Sstevel@tonic-gate * ire_gateway_addr_v4[v6] 1027c478bd9Sstevel@tonic-gate * 1037c478bd9Sstevel@tonic-gate * - ire_gateway_addr_v4[v6] is set during ire_create and later modified 1047c478bd9Sstevel@tonic-gate * by rts_setgwr[v6]. As ire_gateway_addr is a uint32_t, updates to 1057c478bd9Sstevel@tonic-gate * it assumed to be atomic and hence the other parts of the code 1067c478bd9Sstevel@tonic-gate * does not use any locks. ire_gateway_addr_v6 updates are not atomic 1077c478bd9Sstevel@tonic-gate * and hence any access to it uses ire_lock to get/set the right value. 1087c478bd9Sstevel@tonic-gate * 1097c478bd9Sstevel@tonic-gate * ire_ident, ire_refcnt 1107c478bd9Sstevel@tonic-gate * 1117c478bd9Sstevel@tonic-gate * - Updated atomically using atomic_add_32 1127c478bd9Sstevel@tonic-gate * 1137c478bd9Sstevel@tonic-gate * ire_ssthresh, ire_rtt_sd, ire_rtt, ire_ib_pkt_count, ire_ob_pkt_count 1147c478bd9Sstevel@tonic-gate * 1157c478bd9Sstevel@tonic-gate * - Assumes that 32 bit writes are atomic. No locks. ire_lock is 1167c478bd9Sstevel@tonic-gate * used to serialize updates to ire_ssthresh, ire_rtt_sd, ire_rtt. 1177c478bd9Sstevel@tonic-gate * 1187c478bd9Sstevel@tonic-gate * ire_max_frag, ire_frag_flag 1197c478bd9Sstevel@tonic-gate * 1207c478bd9Sstevel@tonic-gate * - ire_lock is used to set/read both of them together. 1217c478bd9Sstevel@tonic-gate * 1227c478bd9Sstevel@tonic-gate * ire_tire_mark 1237c478bd9Sstevel@tonic-gate * 1247c478bd9Sstevel@tonic-gate * - Set in ire_create and updated in ire_expire, which is called 1257c478bd9Sstevel@tonic-gate * by only one function namely ip_trash_timer_expire. Thus only 1267c478bd9Sstevel@tonic-gate * one function updates and examines the value. 1277c478bd9Sstevel@tonic-gate * 1287c478bd9Sstevel@tonic-gate * ire_marks 1297c478bd9Sstevel@tonic-gate * - bucket lock protects this. 1307c478bd9Sstevel@tonic-gate * 1317c478bd9Sstevel@tonic-gate * ire_ipsec_overhead/ire_ll_hdr_length 1327c478bd9Sstevel@tonic-gate * 1337c478bd9Sstevel@tonic-gate * - Place holder for returning the information to the upper layers 1347c478bd9Sstevel@tonic-gate * when IRE_DB_REQ comes down. 1357c478bd9Sstevel@tonic-gate * 1367c478bd9Sstevel@tonic-gate * 1377c478bd9Sstevel@tonic-gate * ipv6_ire_default_count is protected by the bucket lock of 1387c478bd9Sstevel@tonic-gate * ip_forwarding_table_v6[0][0]. 1397c478bd9Sstevel@tonic-gate * 140c793af95Ssangeeta * ipv6_ire_default_index is not protected as it is just a hint 141c793af95Ssangeeta * at which default gateway to use. There is nothing 1427c478bd9Sstevel@tonic-gate * wrong in using the same gateway for two different connections. 1437c478bd9Sstevel@tonic-gate * 1447c478bd9Sstevel@tonic-gate * As we always hold the bucket locks in all the places while accessing 1457c478bd9Sstevel@tonic-gate * the above values, it is natural to use them for protecting them. 1467c478bd9Sstevel@tonic-gate * 1477c478bd9Sstevel@tonic-gate * We have a separate cache table and forwarding table for IPv4 and IPv6. 1487c478bd9Sstevel@tonic-gate * Cache table (ip_cache_table/ip_cache_table_v6) is a pointer to an 1497c478bd9Sstevel@tonic-gate * array of irb_t structure and forwarding table (ip_forwarding_table/ 1507c478bd9Sstevel@tonic-gate * ip_forwarding_table_v6) is an array of pointers to array of irb_t 151*f4b3ec61Sdh155122 * structure. ip_forwarding_table_v6 is allocated dynamically in 152*f4b3ec61Sdh155122 * ire_add_v6. ire_ft_init_lock is used to serialize multiple threads 1537c478bd9Sstevel@tonic-gate * initializing the same bucket. Once a bucket is initialized, it is never 154*f4b3ec61Sdh155122 * de-alloacted. This assumption enables us to access 155*f4b3ec61Sdh155122 * ip_forwarding_table_v6[i] without any locks. 1567c478bd9Sstevel@tonic-gate * 1577c478bd9Sstevel@tonic-gate * Each irb_t - ire bucket structure has a lock to protect 1587c478bd9Sstevel@tonic-gate * a bucket and the ires residing in the bucket have a back pointer to 1597c478bd9Sstevel@tonic-gate * the bucket structure. It also has a reference count for the number 1607c478bd9Sstevel@tonic-gate * of threads walking the bucket - irb_refcnt which is bumped up 1617c478bd9Sstevel@tonic-gate * using the macro IRB_REFHOLD macro. The flags irb_flags can be 1627c478bd9Sstevel@tonic-gate * set to IRE_MARK_CONDEMNED indicating that there are some ires 1637c478bd9Sstevel@tonic-gate * in this bucket that are marked with IRE_MARK_CONDEMNED and the 1647c478bd9Sstevel@tonic-gate * last thread to leave the bucket should delete the ires. Usually 1657c478bd9Sstevel@tonic-gate * this is done by the IRB_REFRELE macro which is used to decrement 1667c478bd9Sstevel@tonic-gate * the reference count on a bucket. 1677c478bd9Sstevel@tonic-gate * 1687c478bd9Sstevel@tonic-gate * IRE_REFHOLD/IRE_REFRELE macros operate on the ire which increments/ 1697c478bd9Sstevel@tonic-gate * decrements the reference count, ire_refcnt, atomically on the ire. 1707c478bd9Sstevel@tonic-gate * ire_refcnt is modified only using this macro. Operations on the IRE 1717c478bd9Sstevel@tonic-gate * could be described as follows : 1727c478bd9Sstevel@tonic-gate * 1737c478bd9Sstevel@tonic-gate * CREATE an ire with reference count initialized to 1. 1747c478bd9Sstevel@tonic-gate * 1757c478bd9Sstevel@tonic-gate * ADDITION of an ire holds the bucket lock, checks for duplicates 1767c478bd9Sstevel@tonic-gate * and then adds the ire. ire_add_v4/ire_add_v6 returns the ire after 1777c478bd9Sstevel@tonic-gate * bumping up once more i.e the reference count is 2. This is to avoid 1787c478bd9Sstevel@tonic-gate * an extra lookup in the functions calling ire_add which wants to 1797c478bd9Sstevel@tonic-gate * work with the ire after adding. 1807c478bd9Sstevel@tonic-gate * 1817c478bd9Sstevel@tonic-gate * LOOKUP of an ire bumps up the reference count using IRE_REFHOLD 1827c478bd9Sstevel@tonic-gate * macro. It is valid to bump up the referece count of the IRE, 1837c478bd9Sstevel@tonic-gate * after the lookup has returned an ire. Following are the lookup 1847c478bd9Sstevel@tonic-gate * functions that return an HELD ire : 1857c478bd9Sstevel@tonic-gate * 1867c478bd9Sstevel@tonic-gate * ire_lookup_local[_v6], ire_ctable_lookup[_v6], ire_ftable_lookup[_v6], 1877c478bd9Sstevel@tonic-gate * ire_cache_lookup[_v6], ire_lookup_multi[_v6], ire_route_lookup[_v6], 1887c478bd9Sstevel@tonic-gate * ipif_to_ire[_v6], ire_mrtun_lookup, ire_srcif_table_lookup. 1897c478bd9Sstevel@tonic-gate * 1907c478bd9Sstevel@tonic-gate * DELETION of an ire holds the bucket lock, removes it from the list 1917c478bd9Sstevel@tonic-gate * and then decrements the reference count for having removed from the list 1927c478bd9Sstevel@tonic-gate * by using the IRE_REFRELE macro. If some other thread has looked up 1937c478bd9Sstevel@tonic-gate * the ire, the reference count would have been bumped up and hence 1947c478bd9Sstevel@tonic-gate * this ire will not be freed once deleted. It will be freed once the 1957c478bd9Sstevel@tonic-gate * reference count drops to zero. 1967c478bd9Sstevel@tonic-gate * 1977c478bd9Sstevel@tonic-gate * Add and Delete acquires the bucket lock as RW_WRITER, while all the 1987c478bd9Sstevel@tonic-gate * lookups acquire the bucket lock as RW_READER. 1997c478bd9Sstevel@tonic-gate * 2007c478bd9Sstevel@tonic-gate * NOTE : The only functions that does the IRE_REFRELE when an ire is 2017c478bd9Sstevel@tonic-gate * passed as an argument are : 2027c478bd9Sstevel@tonic-gate * 2037c478bd9Sstevel@tonic-gate * 1) ip_wput_ire : This is because it IRE_REFHOLD/RELEs the 2047c478bd9Sstevel@tonic-gate * broadcast ires it looks up internally within 2057c478bd9Sstevel@tonic-gate * the function. Currently, for simplicity it does 2067c478bd9Sstevel@tonic-gate * not differentiate the one that is passed in and 2077c478bd9Sstevel@tonic-gate * the ones it looks up internally. It always 2087c478bd9Sstevel@tonic-gate * IRE_REFRELEs. 2097c478bd9Sstevel@tonic-gate * 2) ire_send 2107c478bd9Sstevel@tonic-gate * ire_send_v6 : As ire_send calls ip_wput_ire and other functions 2117c478bd9Sstevel@tonic-gate * that take ire as an argument, it has to selectively 2127c478bd9Sstevel@tonic-gate * IRE_REFRELE the ire. To maintain symmetry, 2137c478bd9Sstevel@tonic-gate * ire_send_v6 does the same. 2147c478bd9Sstevel@tonic-gate * 2157c478bd9Sstevel@tonic-gate * Otherwise, the general rule is to do the IRE_REFRELE in the function 2167c478bd9Sstevel@tonic-gate * that is passing the ire as an argument. 2177c478bd9Sstevel@tonic-gate * 2187c478bd9Sstevel@tonic-gate * In trying to locate ires the following points are to be noted. 2197c478bd9Sstevel@tonic-gate * 2207c478bd9Sstevel@tonic-gate * IRE_MARK_CONDEMNED signifies that the ire has been logically deleted and is 2217c478bd9Sstevel@tonic-gate * to be ignored when walking the ires using ire_next. 2227c478bd9Sstevel@tonic-gate * 2237c478bd9Sstevel@tonic-gate * IRE_MARK_HIDDEN signifies that the ire is a special ire typically for the 2247c478bd9Sstevel@tonic-gate * benefit of in.mpathd which needs to probe interfaces for failures. Normal 2257c478bd9Sstevel@tonic-gate * applications should not be seeing this ire and hence this ire is ignored 2267c478bd9Sstevel@tonic-gate * in most cases in the search using ire_next. 2277c478bd9Sstevel@tonic-gate * 2287c478bd9Sstevel@tonic-gate * Zones note: 2297c478bd9Sstevel@tonic-gate * Walking IREs within a given zone also walks certain ires in other 2307c478bd9Sstevel@tonic-gate * zones. This is done intentionally. IRE walks with a specified 2317c478bd9Sstevel@tonic-gate * zoneid are used only when doing informational reports, and 2327c478bd9Sstevel@tonic-gate * zone users want to see things that they can access. See block 2337c478bd9Sstevel@tonic-gate * comment in ire_walk_ill_match(). 2347c478bd9Sstevel@tonic-gate */ 2357c478bd9Sstevel@tonic-gate 2367c478bd9Sstevel@tonic-gate /* 2377c478bd9Sstevel@tonic-gate * A per-interface routing table is created ( if not present) 2387c478bd9Sstevel@tonic-gate * when the first entry is added to this special routing table. 2397c478bd9Sstevel@tonic-gate * This special routing table is accessed through the ill data structure. 2407c478bd9Sstevel@tonic-gate * The routing table looks like cache table. For example, currently it 2417c478bd9Sstevel@tonic-gate * is used by mobile-ip foreign agent to forward data that only comes from 2427c478bd9Sstevel@tonic-gate * the home agent tunnel for a mobile node. Thus if the outgoing interface 2437c478bd9Sstevel@tonic-gate * is a RESOLVER interface, IP may need to resolve the hardware address for 2447c478bd9Sstevel@tonic-gate * the outgoing interface. The routing entries in this table are not updated 2457c478bd9Sstevel@tonic-gate * in IRE_CACHE. When MCTL msg comes back from ARP, the incoming ill informa- 2467c478bd9Sstevel@tonic-gate * tion is lost as the write queue is passed to ip_wput. 2477c478bd9Sstevel@tonic-gate * But, before sending the packet out, the hardware information must be updated 2487c478bd9Sstevel@tonic-gate * in the special forwarding table. ire_srcif_table_count keeps track of total 2497c478bd9Sstevel@tonic-gate * number of ires that are in interface based tables. Each interface based 2507c478bd9Sstevel@tonic-gate * table hangs off of the incoming ill and each ill_t also keeps a refcnt 2517c478bd9Sstevel@tonic-gate * of ires in that table. 2527c478bd9Sstevel@tonic-gate */ 2537c478bd9Sstevel@tonic-gate 2547c478bd9Sstevel@tonic-gate /* 2557c478bd9Sstevel@tonic-gate * The minimum size of IRE cache table. It will be recalcuated in 2567c478bd9Sstevel@tonic-gate * ip_ire_init(). 257*f4b3ec61Sdh155122 * Setable in /etc/system 2587c478bd9Sstevel@tonic-gate */ 2597c478bd9Sstevel@tonic-gate uint32_t ip_cache_table_size = IP_CACHE_TABLE_SIZE; 2607c478bd9Sstevel@tonic-gate uint32_t ip6_cache_table_size = IP6_CACHE_TABLE_SIZE; 2617c478bd9Sstevel@tonic-gate 2627c478bd9Sstevel@tonic-gate /* 2637c478bd9Sstevel@tonic-gate * The size of the forwarding table. We will make sure that it is a 2647c478bd9Sstevel@tonic-gate * power of 2 in ip_ire_init(). 265*f4b3ec61Sdh155122 * Setable in /etc/system 2667c478bd9Sstevel@tonic-gate */ 2677c478bd9Sstevel@tonic-gate uint32_t ip6_ftable_hash_size = IP6_FTABLE_HASH_SIZE; 2687c478bd9Sstevel@tonic-gate 2697c478bd9Sstevel@tonic-gate struct kmem_cache *ire_cache; 2707c478bd9Sstevel@tonic-gate static ire_t ire_null; 2717c478bd9Sstevel@tonic-gate 2727c478bd9Sstevel@tonic-gate /* 2737c478bd9Sstevel@tonic-gate * The threshold number of IRE in a bucket when the IREs are 2747c478bd9Sstevel@tonic-gate * cleaned up. This threshold is calculated later in ip_open() 2757c478bd9Sstevel@tonic-gate * based on the speed of CPU and available memory. This default 2767c478bd9Sstevel@tonic-gate * value is the maximum. 2777c478bd9Sstevel@tonic-gate * 2787c478bd9Sstevel@tonic-gate * We have two kinds of cached IRE, temporary and 2797c478bd9Sstevel@tonic-gate * non-temporary. Temporary IREs are marked with 2807c478bd9Sstevel@tonic-gate * IRE_MARK_TEMPORARY. They are IREs created for non 2817c478bd9Sstevel@tonic-gate * TCP traffic and for forwarding purposes. All others 2827c478bd9Sstevel@tonic-gate * are non-temporary IREs. We don't mark IRE created for 2837c478bd9Sstevel@tonic-gate * TCP as temporary because TCP is stateful and there are 2847c478bd9Sstevel@tonic-gate * info stored in the IRE which can be shared by other TCP 2857c478bd9Sstevel@tonic-gate * connections to the same destination. For connected 2867c478bd9Sstevel@tonic-gate * endpoint, we also don't want to mark the IRE used as 2877c478bd9Sstevel@tonic-gate * temporary because the same IRE will be used frequently, 2887c478bd9Sstevel@tonic-gate * otherwise, the app should not do a connect(). We change 2897c478bd9Sstevel@tonic-gate * the marking at ip_bind_connected_*() if necessary. 2907c478bd9Sstevel@tonic-gate * 2917c478bd9Sstevel@tonic-gate * We want to keep the cache IRE hash bucket length reasonably 2927c478bd9Sstevel@tonic-gate * short, otherwise IRE lookup functions will take "forever." 2937c478bd9Sstevel@tonic-gate * We use the "crude" function that the IRE bucket 2947c478bd9Sstevel@tonic-gate * length should be based on the CPU speed, which is 1 entry 2957c478bd9Sstevel@tonic-gate * per x MHz, depending on the shift factor ip_ire_cpu_ratio 2967c478bd9Sstevel@tonic-gate * (n). This means that with a 750MHz CPU, the max bucket 2977c478bd9Sstevel@tonic-gate * length can be (750 >> n) entries. 2987c478bd9Sstevel@tonic-gate * 2997c478bd9Sstevel@tonic-gate * Note that this threshold is separate for temp and non-temp 3007c478bd9Sstevel@tonic-gate * IREs. This means that the actual bucket length can be 3017c478bd9Sstevel@tonic-gate * twice as that. And while we try to keep temporary IRE 3027c478bd9Sstevel@tonic-gate * length at most at the threshold value, we do not attempt to 3037c478bd9Sstevel@tonic-gate * make the length for non-temporary IREs fixed, for the 3047c478bd9Sstevel@tonic-gate * reason stated above. Instead, we start trying to find 3057c478bd9Sstevel@tonic-gate * "unused" non-temporary IREs when the bucket length reaches 3067c478bd9Sstevel@tonic-gate * this threshold and clean them up. 3077c478bd9Sstevel@tonic-gate * 3087c478bd9Sstevel@tonic-gate * We also want to limit the amount of memory used by 3097c478bd9Sstevel@tonic-gate * IREs. So if we are allowed to use ~3% of memory (M) 3107c478bd9Sstevel@tonic-gate * for those IREs, each bucket should not have more than 3117c478bd9Sstevel@tonic-gate * 3127c478bd9Sstevel@tonic-gate * M / num of cache bucket / sizeof (ire_t) 3137c478bd9Sstevel@tonic-gate * 3147c478bd9Sstevel@tonic-gate * Again the above memory uses are separate for temp and 3157c478bd9Sstevel@tonic-gate * non-temp cached IREs. 3167c478bd9Sstevel@tonic-gate * 3177c478bd9Sstevel@tonic-gate * We may also want the limit to be a function of the number 3187c478bd9Sstevel@tonic-gate * of interfaces and number of CPUs. Doing the initialization 3197c478bd9Sstevel@tonic-gate * in ip_open() means that every time an interface is plumbed, 3207c478bd9Sstevel@tonic-gate * the max is re-calculated. Right now, we don't do anything 3217c478bd9Sstevel@tonic-gate * different. In future, when we have more experience, we 3227c478bd9Sstevel@tonic-gate * may want to change this behavior. 3237c478bd9Sstevel@tonic-gate */ 324*f4b3ec61Sdh155122 uint32_t ip_ire_max_bucket_cnt = 10; /* Setable in /etc/system */ 3257c478bd9Sstevel@tonic-gate uint32_t ip6_ire_max_bucket_cnt = 10; 3267c478bd9Sstevel@tonic-gate 3277c478bd9Sstevel@tonic-gate /* 3287c478bd9Sstevel@tonic-gate * The minimum of the temporary IRE bucket count. We do not want 3297c478bd9Sstevel@tonic-gate * the length of each bucket to be too short. This may hurt 3307c478bd9Sstevel@tonic-gate * performance of some apps as the temporary IREs are removed too 3317c478bd9Sstevel@tonic-gate * often. 3327c478bd9Sstevel@tonic-gate */ 333*f4b3ec61Sdh155122 uint32_t ip_ire_min_bucket_cnt = 3; /* /etc/system - not used */ 3347c478bd9Sstevel@tonic-gate uint32_t ip6_ire_min_bucket_cnt = 3; 3357c478bd9Sstevel@tonic-gate 3367c478bd9Sstevel@tonic-gate /* 3377c478bd9Sstevel@tonic-gate * The ratio of memory consumed by IRE used for temporary to available 3387c478bd9Sstevel@tonic-gate * memory. This is a shift factor, so 6 means the ratio 1 to 64. This 3397c478bd9Sstevel@tonic-gate * value can be changed in /etc/system. 6 is a reasonable number. 3407c478bd9Sstevel@tonic-gate */ 341*f4b3ec61Sdh155122 uint32_t ip_ire_mem_ratio = 6; /* /etc/system */ 3427c478bd9Sstevel@tonic-gate /* The shift factor for CPU speed to calculate the max IRE bucket length. */ 343*f4b3ec61Sdh155122 uint32_t ip_ire_cpu_ratio = 7; /* /etc/system */ 3447c478bd9Sstevel@tonic-gate 345c793af95Ssangeeta typedef struct nce_clookup_s { 346c793af95Ssangeeta ipaddr_t ncecl_addr; 347c793af95Ssangeeta boolean_t ncecl_found; 348c793af95Ssangeeta } nce_clookup_t; 349c793af95Ssangeeta 3507c478bd9Sstevel@tonic-gate /* 3517c478bd9Sstevel@tonic-gate * The maximum number of buckets in IRE cache table. In future, we may 3527c478bd9Sstevel@tonic-gate * want to make it a dynamic hash table. For the moment, we fix the 3537c478bd9Sstevel@tonic-gate * size and allocate the table in ip_ire_init() when IP is first loaded. 3547c478bd9Sstevel@tonic-gate * We take into account the amount of memory a system has. 3557c478bd9Sstevel@tonic-gate */ 3567c478bd9Sstevel@tonic-gate #define IP_MAX_CACHE_TABLE_SIZE 4096 3577c478bd9Sstevel@tonic-gate 358*f4b3ec61Sdh155122 /* Setable in /etc/system */ 3597c478bd9Sstevel@tonic-gate static uint32_t ip_max_cache_table_size = IP_MAX_CACHE_TABLE_SIZE; 3607c478bd9Sstevel@tonic-gate static uint32_t ip6_max_cache_table_size = IP_MAX_CACHE_TABLE_SIZE; 3617c478bd9Sstevel@tonic-gate 3627c478bd9Sstevel@tonic-gate #define NUM_ILLS 3 /* To build the ILL list to unlock */ 3637c478bd9Sstevel@tonic-gate 3647c478bd9Sstevel@tonic-gate /* Zero iulp_t for initialization. */ 3657c478bd9Sstevel@tonic-gate const iulp_t ire_uinfo_null = { 0 }; 3667c478bd9Sstevel@tonic-gate 3677c478bd9Sstevel@tonic-gate static int ire_add_v4(ire_t **ire_p, queue_t *q, mblk_t *mp, 368c793af95Ssangeeta ipsq_func_t func, boolean_t); 3697c478bd9Sstevel@tonic-gate static int ire_add_srcif_v4(ire_t **ire_p, queue_t *q, mblk_t *mp, 3707c478bd9Sstevel@tonic-gate ipsq_func_t func); 3717c478bd9Sstevel@tonic-gate static ire_t *ire_update_srcif_v4(ire_t *ire); 3727c478bd9Sstevel@tonic-gate static void ire_delete_v4(ire_t *ire); 3737c478bd9Sstevel@tonic-gate static void ire_report_ctable(ire_t *ire, char *mp); 3747c478bd9Sstevel@tonic-gate static void ire_report_mrtun_table(ire_t *ire, char *mp); 375*f4b3ec61Sdh155122 static void ire_report_srcif_table(ire_t *ire, char *mp, ip_stack_t *ipst); 37645916cd2Sjpk static void ire_walk_ipvers(pfv_t func, void *arg, uchar_t vers, 377*f4b3ec61Sdh155122 zoneid_t zoneid, ip_stack_t *); 3787c478bd9Sstevel@tonic-gate static void ire_walk_ill_ipvers(uint_t match_flags, uint_t ire_type, 37945916cd2Sjpk pfv_t func, void *arg, uchar_t vers, ill_t *ill); 3807c478bd9Sstevel@tonic-gate static void ire_cache_cleanup(irb_t *irb, uint32_t threshold, int cnt); 3817c478bd9Sstevel@tonic-gate extern void ill_unlock_ills(ill_t **list, int cnt); 382c793af95Ssangeeta static void ip_nce_clookup_and_delete(nce_t *nce, void *arg); 3837c478bd9Sstevel@tonic-gate extern void th_trace_rrecord(th_trace_t *); 3847c478bd9Sstevel@tonic-gate #ifdef IRE_DEBUG 3857c478bd9Sstevel@tonic-gate static void ire_trace_inactive(ire_t *); 3867c478bd9Sstevel@tonic-gate #endif 3877c478bd9Sstevel@tonic-gate 3887c478bd9Sstevel@tonic-gate /* 3897c478bd9Sstevel@tonic-gate * To avoid bloating the code, we call this function instead of 3907c478bd9Sstevel@tonic-gate * using the macro IRE_REFRELE. Use macro only in performance 3917c478bd9Sstevel@tonic-gate * critical paths. 3927c478bd9Sstevel@tonic-gate * 3937c478bd9Sstevel@tonic-gate * Must not be called while holding any locks. Otherwise if this is 3947c478bd9Sstevel@tonic-gate * the last reference to be released there is a chance of recursive mutex 3957c478bd9Sstevel@tonic-gate * panic due to ire_refrele -> ipif_ill_refrele_tail -> qwriter_ip trying 3967c478bd9Sstevel@tonic-gate * to restart an ioctl. The one exception is when the caller is sure that 3977c478bd9Sstevel@tonic-gate * this is not the last reference to be released. Eg. if the caller is 3987c478bd9Sstevel@tonic-gate * sure that the ire has not been deleted and won't be deleted. 3997c478bd9Sstevel@tonic-gate */ 4007c478bd9Sstevel@tonic-gate void 4017c478bd9Sstevel@tonic-gate ire_refrele(ire_t *ire) 4027c478bd9Sstevel@tonic-gate { 4037c478bd9Sstevel@tonic-gate IRE_REFRELE(ire); 4047c478bd9Sstevel@tonic-gate } 4057c478bd9Sstevel@tonic-gate 4067c478bd9Sstevel@tonic-gate void 4077c478bd9Sstevel@tonic-gate ire_refrele_notr(ire_t *ire) 4087c478bd9Sstevel@tonic-gate { 4097c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 4107c478bd9Sstevel@tonic-gate } 4117c478bd9Sstevel@tonic-gate 4127c478bd9Sstevel@tonic-gate /* 4137c478bd9Sstevel@tonic-gate * kmem_cache_alloc constructor for IRE in kma space. 4147c478bd9Sstevel@tonic-gate * Note that when ire_mp is set the IRE is stored in that mblk and 4157c478bd9Sstevel@tonic-gate * not in this cache. 4167c478bd9Sstevel@tonic-gate */ 4177c478bd9Sstevel@tonic-gate /* ARGSUSED */ 4187c478bd9Sstevel@tonic-gate static int 4197c478bd9Sstevel@tonic-gate ip_ire_constructor(void *buf, void *cdrarg, int kmflags) 4207c478bd9Sstevel@tonic-gate { 4217c478bd9Sstevel@tonic-gate ire_t *ire = buf; 4227c478bd9Sstevel@tonic-gate 423c793af95Ssangeeta ire->ire_nce = NULL; 4247c478bd9Sstevel@tonic-gate 4257c478bd9Sstevel@tonic-gate return (0); 4267c478bd9Sstevel@tonic-gate } 4277c478bd9Sstevel@tonic-gate 4287c478bd9Sstevel@tonic-gate /* ARGSUSED1 */ 4297c478bd9Sstevel@tonic-gate static void 4307c478bd9Sstevel@tonic-gate ip_ire_destructor(void *buf, void *cdrarg) 4317c478bd9Sstevel@tonic-gate { 4327c478bd9Sstevel@tonic-gate ire_t *ire = buf; 4337c478bd9Sstevel@tonic-gate 434c793af95Ssangeeta ASSERT(ire->ire_nce == NULL); 4357c478bd9Sstevel@tonic-gate } 4367c478bd9Sstevel@tonic-gate 4377c478bd9Sstevel@tonic-gate /* 4387c478bd9Sstevel@tonic-gate * This function is associated with the IP_IOC_IRE_ADVISE_NO_REPLY 4397c478bd9Sstevel@tonic-gate * IOCTL. It is used by TCP (or other ULPs) to supply revised information 4407c478bd9Sstevel@tonic-gate * for an existing CACHED IRE. 4417c478bd9Sstevel@tonic-gate */ 4427c478bd9Sstevel@tonic-gate /* ARGSUSED */ 4437c478bd9Sstevel@tonic-gate int 4447c478bd9Sstevel@tonic-gate ip_ire_advise(queue_t *q, mblk_t *mp, cred_t *ioc_cr) 4457c478bd9Sstevel@tonic-gate { 4467c478bd9Sstevel@tonic-gate uchar_t *addr_ucp; 4477c478bd9Sstevel@tonic-gate ipic_t *ipic; 4487c478bd9Sstevel@tonic-gate ire_t *ire; 4497c478bd9Sstevel@tonic-gate ipaddr_t addr; 4507c478bd9Sstevel@tonic-gate in6_addr_t v6addr; 4517c478bd9Sstevel@tonic-gate irb_t *irb; 4527c478bd9Sstevel@tonic-gate zoneid_t zoneid; 453*f4b3ec61Sdh155122 ip_stack_t *ipst = CONNQ_TO_IPST(q); 4547c478bd9Sstevel@tonic-gate 4557c478bd9Sstevel@tonic-gate ASSERT(q->q_next == NULL); 4567c478bd9Sstevel@tonic-gate zoneid = Q_TO_CONN(q)->conn_zoneid; 4577c478bd9Sstevel@tonic-gate 4587c478bd9Sstevel@tonic-gate /* 4597c478bd9Sstevel@tonic-gate * Check privilege using the ioctl credential; if it is NULL 4607c478bd9Sstevel@tonic-gate * then this is a kernel message and therefor privileged. 4617c478bd9Sstevel@tonic-gate */ 462*f4b3ec61Sdh155122 if (ioc_cr != NULL && secpolicy_ip_config(ioc_cr, B_FALSE) != 0) 4637c478bd9Sstevel@tonic-gate return (EPERM); 4647c478bd9Sstevel@tonic-gate 4657c478bd9Sstevel@tonic-gate ipic = (ipic_t *)mp->b_rptr; 4667c478bd9Sstevel@tonic-gate if (!(addr_ucp = mi_offset_param(mp, ipic->ipic_addr_offset, 4677c478bd9Sstevel@tonic-gate ipic->ipic_addr_length))) { 4687c478bd9Sstevel@tonic-gate return (EINVAL); 4697c478bd9Sstevel@tonic-gate } 4707c478bd9Sstevel@tonic-gate if (!OK_32PTR(addr_ucp)) 4717c478bd9Sstevel@tonic-gate return (EINVAL); 4727c478bd9Sstevel@tonic-gate switch (ipic->ipic_addr_length) { 4737c478bd9Sstevel@tonic-gate case IP_ADDR_LEN: { 4747c478bd9Sstevel@tonic-gate /* Extract the destination address. */ 4757c478bd9Sstevel@tonic-gate addr = *(ipaddr_t *)addr_ucp; 4767c478bd9Sstevel@tonic-gate /* Find the corresponding IRE. */ 477*f4b3ec61Sdh155122 ire = ire_cache_lookup(addr, zoneid, NULL, ipst); 4787c478bd9Sstevel@tonic-gate break; 4797c478bd9Sstevel@tonic-gate } 4807c478bd9Sstevel@tonic-gate case IPV6_ADDR_LEN: { 4817c478bd9Sstevel@tonic-gate /* Extract the destination address. */ 4827c478bd9Sstevel@tonic-gate v6addr = *(in6_addr_t *)addr_ucp; 4837c478bd9Sstevel@tonic-gate /* Find the corresponding IRE. */ 484*f4b3ec61Sdh155122 ire = ire_cache_lookup_v6(&v6addr, zoneid, NULL, ipst); 4857c478bd9Sstevel@tonic-gate break; 4867c478bd9Sstevel@tonic-gate } 4877c478bd9Sstevel@tonic-gate default: 4887c478bd9Sstevel@tonic-gate return (EINVAL); 4897c478bd9Sstevel@tonic-gate } 4907c478bd9Sstevel@tonic-gate 4917c478bd9Sstevel@tonic-gate if (ire == NULL) 4927c478bd9Sstevel@tonic-gate return (ENOENT); 4937c478bd9Sstevel@tonic-gate /* 4947c478bd9Sstevel@tonic-gate * Update the round trip time estimate and/or the max frag size 4957c478bd9Sstevel@tonic-gate * and/or the slow start threshold. 4967c478bd9Sstevel@tonic-gate * 4977c478bd9Sstevel@tonic-gate * We serialize multiple advises using ire_lock. 4987c478bd9Sstevel@tonic-gate */ 4997c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 5007c478bd9Sstevel@tonic-gate if (ipic->ipic_rtt) { 5017c478bd9Sstevel@tonic-gate /* 5027c478bd9Sstevel@tonic-gate * If there is no old cached values, initialize them 5037c478bd9Sstevel@tonic-gate * conservatively. Set them to be (1.5 * new value). 5047c478bd9Sstevel@tonic-gate */ 5057c478bd9Sstevel@tonic-gate if (ire->ire_uinfo.iulp_rtt != 0) { 5067c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt = (ire->ire_uinfo.iulp_rtt + 5077c478bd9Sstevel@tonic-gate ipic->ipic_rtt) >> 1; 5087c478bd9Sstevel@tonic-gate } else { 5097c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt = ipic->ipic_rtt + 5107c478bd9Sstevel@tonic-gate (ipic->ipic_rtt >> 1); 5117c478bd9Sstevel@tonic-gate } 5127c478bd9Sstevel@tonic-gate if (ire->ire_uinfo.iulp_rtt_sd != 0) { 5137c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt_sd = 5147c478bd9Sstevel@tonic-gate (ire->ire_uinfo.iulp_rtt_sd + 5157c478bd9Sstevel@tonic-gate ipic->ipic_rtt_sd) >> 1; 5167c478bd9Sstevel@tonic-gate } else { 5177c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt_sd = ipic->ipic_rtt_sd + 5187c478bd9Sstevel@tonic-gate (ipic->ipic_rtt_sd >> 1); 5197c478bd9Sstevel@tonic-gate } 5207c478bd9Sstevel@tonic-gate } 5217c478bd9Sstevel@tonic-gate if (ipic->ipic_max_frag) 5227c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(ipic->ipic_max_frag, IP_MAXPACKET); 5237c478bd9Sstevel@tonic-gate if (ipic->ipic_ssthresh != 0) { 5247c478bd9Sstevel@tonic-gate if (ire->ire_uinfo.iulp_ssthresh != 0) 5257c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_ssthresh = 5267c478bd9Sstevel@tonic-gate (ipic->ipic_ssthresh + 5277c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_ssthresh) >> 1; 5287c478bd9Sstevel@tonic-gate else 5297c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_ssthresh = ipic->ipic_ssthresh; 5307c478bd9Sstevel@tonic-gate } 5317c478bd9Sstevel@tonic-gate /* 5327c478bd9Sstevel@tonic-gate * Don't need the ire_lock below this. ire_type does not change 5337c478bd9Sstevel@tonic-gate * after initialization. ire_marks is protected by irb_lock. 5347c478bd9Sstevel@tonic-gate */ 5357c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 5367c478bd9Sstevel@tonic-gate 5377c478bd9Sstevel@tonic-gate if (ipic->ipic_ire_marks != 0 && ire->ire_type == IRE_CACHE) { 5387c478bd9Sstevel@tonic-gate /* 5397c478bd9Sstevel@tonic-gate * Only increment the temporary IRE count if the original 5407c478bd9Sstevel@tonic-gate * IRE is not already marked temporary. 5417c478bd9Sstevel@tonic-gate */ 5427c478bd9Sstevel@tonic-gate irb = ire->ire_bucket; 5437c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_WRITER); 5447c478bd9Sstevel@tonic-gate if ((ipic->ipic_ire_marks & IRE_MARK_TEMPORARY) && 5457c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_TEMPORARY)) { 5467c478bd9Sstevel@tonic-gate irb->irb_tmp_ire_cnt++; 5477c478bd9Sstevel@tonic-gate } 5487c478bd9Sstevel@tonic-gate ire->ire_marks |= ipic->ipic_ire_marks; 5497c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 5507c478bd9Sstevel@tonic-gate } 5517c478bd9Sstevel@tonic-gate 5527c478bd9Sstevel@tonic-gate ire_refrele(ire); 5537c478bd9Sstevel@tonic-gate return (0); 5547c478bd9Sstevel@tonic-gate } 5557c478bd9Sstevel@tonic-gate 5567c478bd9Sstevel@tonic-gate /* 5577c478bd9Sstevel@tonic-gate * This function is associated with the IP_IOC_IRE_DELETE[_NO_REPLY] 5587c478bd9Sstevel@tonic-gate * IOCTL[s]. The NO_REPLY form is used by TCP to delete a route IRE 5597c478bd9Sstevel@tonic-gate * for a host that is not responding. This will force an attempt to 560dc041e83Scarlsonj * establish a new route, if available, and flush out the ARP entry so 561dc041e83Scarlsonj * it will re-resolve. Management processes may want to use the 562dc041e83Scarlsonj * version that generates a reply. 5637c478bd9Sstevel@tonic-gate * 5647c478bd9Sstevel@tonic-gate * This function does not support IPv6 since Neighbor Unreachability Detection 5657c478bd9Sstevel@tonic-gate * means that negative advise like this is useless. 5667c478bd9Sstevel@tonic-gate */ 5677c478bd9Sstevel@tonic-gate /* ARGSUSED */ 5687c478bd9Sstevel@tonic-gate int 5697c478bd9Sstevel@tonic-gate ip_ire_delete(queue_t *q, mblk_t *mp, cred_t *ioc_cr) 5707c478bd9Sstevel@tonic-gate { 5717c478bd9Sstevel@tonic-gate uchar_t *addr_ucp; 5727c478bd9Sstevel@tonic-gate ipaddr_t addr; 5737c478bd9Sstevel@tonic-gate ire_t *ire; 5747c478bd9Sstevel@tonic-gate ipid_t *ipid; 5757c478bd9Sstevel@tonic-gate boolean_t routing_sock_info = B_FALSE; /* Sent info? */ 5767c478bd9Sstevel@tonic-gate zoneid_t zoneid; 577c793af95Ssangeeta ire_t *gire = NULL; 578dc041e83Scarlsonj ill_t *ill; 579dc041e83Scarlsonj mblk_t *arp_mp; 580*f4b3ec61Sdh155122 ip_stack_t *ipst; 5817c478bd9Sstevel@tonic-gate 5827c478bd9Sstevel@tonic-gate ASSERT(q->q_next == NULL); 5837c478bd9Sstevel@tonic-gate zoneid = Q_TO_CONN(q)->conn_zoneid; 584*f4b3ec61Sdh155122 ipst = CONNQ_TO_IPST(q); 5857c478bd9Sstevel@tonic-gate 5867c478bd9Sstevel@tonic-gate /* 5877c478bd9Sstevel@tonic-gate * Check privilege using the ioctl credential; if it is NULL 5887c478bd9Sstevel@tonic-gate * then this is a kernel message and therefor privileged. 5897c478bd9Sstevel@tonic-gate */ 590*f4b3ec61Sdh155122 if (ioc_cr != NULL && secpolicy_ip_config(ioc_cr, B_FALSE) != 0) 5917c478bd9Sstevel@tonic-gate return (EPERM); 5927c478bd9Sstevel@tonic-gate 5937c478bd9Sstevel@tonic-gate ipid = (ipid_t *)mp->b_rptr; 5947c478bd9Sstevel@tonic-gate 5957c478bd9Sstevel@tonic-gate /* Only actions on IRE_CACHEs are acceptable at present. */ 5967c478bd9Sstevel@tonic-gate if (ipid->ipid_ire_type != IRE_CACHE) 5977c478bd9Sstevel@tonic-gate return (EINVAL); 5987c478bd9Sstevel@tonic-gate 5997c478bd9Sstevel@tonic-gate addr_ucp = mi_offset_param(mp, ipid->ipid_addr_offset, 6007c478bd9Sstevel@tonic-gate ipid->ipid_addr_length); 6017c478bd9Sstevel@tonic-gate if (addr_ucp == NULL || !OK_32PTR(addr_ucp)) 6027c478bd9Sstevel@tonic-gate return (EINVAL); 6037c478bd9Sstevel@tonic-gate switch (ipid->ipid_addr_length) { 6047c478bd9Sstevel@tonic-gate case IP_ADDR_LEN: 6057c478bd9Sstevel@tonic-gate /* addr_ucp points at IP addr */ 6067c478bd9Sstevel@tonic-gate break; 6077c478bd9Sstevel@tonic-gate case sizeof (sin_t): { 6087c478bd9Sstevel@tonic-gate sin_t *sin; 6097c478bd9Sstevel@tonic-gate /* 6107c478bd9Sstevel@tonic-gate * got complete (sockaddr) address - increment addr_ucp to point 6117c478bd9Sstevel@tonic-gate * at the ip_addr field. 6127c478bd9Sstevel@tonic-gate */ 6137c478bd9Sstevel@tonic-gate sin = (sin_t *)addr_ucp; 6147c478bd9Sstevel@tonic-gate addr_ucp = (uchar_t *)&sin->sin_addr.s_addr; 6157c478bd9Sstevel@tonic-gate break; 6167c478bd9Sstevel@tonic-gate } 6177c478bd9Sstevel@tonic-gate default: 6187c478bd9Sstevel@tonic-gate return (EINVAL); 6197c478bd9Sstevel@tonic-gate } 6207c478bd9Sstevel@tonic-gate /* Extract the destination address. */ 6217c478bd9Sstevel@tonic-gate bcopy(addr_ucp, &addr, IP_ADDR_LEN); 6227c478bd9Sstevel@tonic-gate 6237c478bd9Sstevel@tonic-gate /* Try to find the CACHED IRE. */ 624*f4b3ec61Sdh155122 ire = ire_cache_lookup(addr, zoneid, NULL, ipst); 6257c478bd9Sstevel@tonic-gate 6267c478bd9Sstevel@tonic-gate /* Nail it. */ 6277c478bd9Sstevel@tonic-gate if (ire) { 6287c478bd9Sstevel@tonic-gate /* Allow delete only on CACHE entries */ 6297c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) { 6307c478bd9Sstevel@tonic-gate ire_refrele(ire); 6317c478bd9Sstevel@tonic-gate return (EINVAL); 6327c478bd9Sstevel@tonic-gate } 6337c478bd9Sstevel@tonic-gate 6347c478bd9Sstevel@tonic-gate /* 6357c478bd9Sstevel@tonic-gate * Verify that the IRE has been around for a while. 6367c478bd9Sstevel@tonic-gate * This is to protect against transport protocols 6377c478bd9Sstevel@tonic-gate * that are too eager in sending delete messages. 6387c478bd9Sstevel@tonic-gate */ 6397c478bd9Sstevel@tonic-gate if (gethrestime_sec() < 640*f4b3ec61Sdh155122 ire->ire_create_time + ipst->ips_ip_ignore_delete_time) { 6417c478bd9Sstevel@tonic-gate ire_refrele(ire); 6427c478bd9Sstevel@tonic-gate return (EINVAL); 6437c478bd9Sstevel@tonic-gate } 6447c478bd9Sstevel@tonic-gate /* 6457c478bd9Sstevel@tonic-gate * Now we have a potentially dead cache entry. We need 6467c478bd9Sstevel@tonic-gate * to remove it. 647c793af95Ssangeeta * If this cache entry is generated from a 648c793af95Ssangeeta * default route (i.e., ire_cmask == 0), 6497c478bd9Sstevel@tonic-gate * search the default list and mark it dead and some 6507c478bd9Sstevel@tonic-gate * background process will try to activate it. 6517c478bd9Sstevel@tonic-gate */ 6527c478bd9Sstevel@tonic-gate if ((ire->ire_gateway_addr != 0) && (ire->ire_cmask == 0)) { 6537c478bd9Sstevel@tonic-gate /* 6547c478bd9Sstevel@tonic-gate * Make sure that we pick a different 6557c478bd9Sstevel@tonic-gate * IRE_DEFAULT next time. 6567c478bd9Sstevel@tonic-gate */ 6577c478bd9Sstevel@tonic-gate ire_t *gw_ire; 658c793af95Ssangeeta irb_t *irb = NULL; 659c793af95Ssangeeta uint_t match_flags; 6607c478bd9Sstevel@tonic-gate 661c793af95Ssangeeta match_flags = (MATCH_IRE_DEFAULT | MATCH_IRE_RJ_BHOLE); 662c793af95Ssangeeta 663c793af95Ssangeeta gire = ire_ftable_lookup(ire->ire_addr, 664c793af95Ssangeeta ire->ire_cmask, 0, 0, 665*f4b3ec61Sdh155122 ire->ire_ipif, NULL, zoneid, 0, NULL, match_flags, 666*f4b3ec61Sdh155122 ipst); 667c793af95Ssangeeta 668c793af95Ssangeeta ip3dbg(("ire_ftable_lookup() returned gire %p\n", 669c793af95Ssangeeta (void *)gire)); 670c793af95Ssangeeta 671c793af95Ssangeeta if (gire != NULL) { 672c793af95Ssangeeta irb = gire->ire_bucket; 6737c478bd9Sstevel@tonic-gate 6747c478bd9Sstevel@tonic-gate /* 6757c478bd9Sstevel@tonic-gate * We grab it as writer just to serialize 6767c478bd9Sstevel@tonic-gate * multiple threads trying to bump up 677c793af95Ssangeeta * irb_rr_origin 6787c478bd9Sstevel@tonic-gate */ 6797c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_WRITER); 680c793af95Ssangeeta if ((gw_ire = irb->irb_rr_origin) == NULL) { 6817c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 6827c478bd9Sstevel@tonic-gate goto done; 6837c478bd9Sstevel@tonic-gate } 684c793af95Ssangeeta 6850ad1ccddSsowmini DTRACE_PROBE1(ip__ire__del__origin, 6860ad1ccddSsowmini (ire_t *), gw_ire); 6877c478bd9Sstevel@tonic-gate 6887c478bd9Sstevel@tonic-gate /* Skip past the potentially bad gateway */ 6897c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == 6900ad1ccddSsowmini gw_ire->ire_gateway_addr) { 6910ad1ccddSsowmini ire_t *next = gw_ire->ire_next; 6927c478bd9Sstevel@tonic-gate 6930ad1ccddSsowmini DTRACE_PROBE2(ip__ire__del, 6940ad1ccddSsowmini (ire_t *), gw_ire, (irb_t *), irb); 6950ad1ccddSsowmini IRE_FIND_NEXT_ORIGIN(next); 6960ad1ccddSsowmini irb->irb_rr_origin = next; 6970ad1ccddSsowmini } 6987c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 6997c478bd9Sstevel@tonic-gate } 7007c478bd9Sstevel@tonic-gate } 7017c478bd9Sstevel@tonic-gate done: 702c793af95Ssangeeta if (gire != NULL) 703c793af95Ssangeeta IRE_REFRELE(gire); 7047c478bd9Sstevel@tonic-gate /* report the bad route to routing sockets */ 7057c478bd9Sstevel@tonic-gate ip_rts_change(RTM_LOSING, ire->ire_addr, ire->ire_gateway_addr, 7067c478bd9Sstevel@tonic-gate ire->ire_mask, ire->ire_src_addr, 0, 0, 0, 707*f4b3ec61Sdh155122 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_IFA), ipst); 7087c478bd9Sstevel@tonic-gate routing_sock_info = B_TRUE; 709dc041e83Scarlsonj 710dc041e83Scarlsonj /* 711dc041e83Scarlsonj * TCP is really telling us to start over completely, and it 712dc041e83Scarlsonj * expects that we'll resend the ARP query. Tell ARP to 713dc041e83Scarlsonj * discard the entry, if this is a local destination. 714dc041e83Scarlsonj */ 715dc041e83Scarlsonj ill = ire->ire_stq->q_ptr; 716dc041e83Scarlsonj if (ire->ire_gateway_addr == 0 && 717dc041e83Scarlsonj (arp_mp = ill_ared_alloc(ill, addr)) != NULL) { 718dc041e83Scarlsonj putnext(ill->ill_rq, arp_mp); 719dc041e83Scarlsonj } 720dc041e83Scarlsonj 7217c478bd9Sstevel@tonic-gate ire_delete(ire); 7227c478bd9Sstevel@tonic-gate ire_refrele(ire); 7237c478bd9Sstevel@tonic-gate } 7246bdb8e66Sdd193516 /* 7256bdb8e66Sdd193516 * Also look for an IRE_HOST type redirect ire and 7266bdb8e66Sdd193516 * remove it if present. 7276bdb8e66Sdd193516 */ 7286bdb8e66Sdd193516 ire = ire_route_lookup(addr, 0, 0, IRE_HOST, NULL, NULL, 729*f4b3ec61Sdh155122 ALL_ZONES, NULL, MATCH_IRE_TYPE, ipst); 7307c478bd9Sstevel@tonic-gate 7317c478bd9Sstevel@tonic-gate /* Nail it. */ 7326bdb8e66Sdd193516 if (ire != NULL) { 7336bdb8e66Sdd193516 if (ire->ire_flags & RTF_DYNAMIC) { 7347c478bd9Sstevel@tonic-gate if (!routing_sock_info) { 7357c478bd9Sstevel@tonic-gate ip_rts_change(RTM_LOSING, ire->ire_addr, 7367c478bd9Sstevel@tonic-gate ire->ire_gateway_addr, ire->ire_mask, 7377c478bd9Sstevel@tonic-gate ire->ire_src_addr, 0, 0, 0, 738*f4b3ec61Sdh155122 (RTA_DST | RTA_GATEWAY | RTA_NETMASK | RTA_IFA), 739*f4b3ec61Sdh155122 ipst); 7407c478bd9Sstevel@tonic-gate } 7417c478bd9Sstevel@tonic-gate ire_delete(ire); 7426bdb8e66Sdd193516 } 7437c478bd9Sstevel@tonic-gate ire_refrele(ire); 7447c478bd9Sstevel@tonic-gate } 7457c478bd9Sstevel@tonic-gate return (0); 7467c478bd9Sstevel@tonic-gate } 7477c478bd9Sstevel@tonic-gate 7487c478bd9Sstevel@tonic-gate /* 7497c478bd9Sstevel@tonic-gate * Named Dispatch routine to produce a formatted report on all IREs. 7507c478bd9Sstevel@tonic-gate * This report is accessed by using the ndd utility to "get" ND variable 7517c478bd9Sstevel@tonic-gate * "ipv4_ire_status". 7527c478bd9Sstevel@tonic-gate */ 7537c478bd9Sstevel@tonic-gate /* ARGSUSED */ 7547c478bd9Sstevel@tonic-gate int 7557c478bd9Sstevel@tonic-gate ip_ire_report(queue_t *q, mblk_t *mp, caddr_t arg, cred_t *ioc_cr) 7567c478bd9Sstevel@tonic-gate { 7577c478bd9Sstevel@tonic-gate zoneid_t zoneid; 758*f4b3ec61Sdh155122 ip_stack_t *ipst; 759*f4b3ec61Sdh155122 760*f4b3ec61Sdh155122 if (CONN_Q(q)) 761*f4b3ec61Sdh155122 ipst = CONNQ_TO_IPST(q); 762*f4b3ec61Sdh155122 else 763*f4b3ec61Sdh155122 ipst = ILLQ_TO_IPST(q); 7647c478bd9Sstevel@tonic-gate 7657c478bd9Sstevel@tonic-gate (void) mi_mpprintf(mp, 7667c478bd9Sstevel@tonic-gate "IRE " MI_COL_HDRPAD_STR 7677c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 7687c478bd9Sstevel@tonic-gate "rfq " MI_COL_HDRPAD_STR 7697c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 7707c478bd9Sstevel@tonic-gate "stq " MI_COL_HDRPAD_STR 7717c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 7727c478bd9Sstevel@tonic-gate " zone " 7737c478bd9Sstevel@tonic-gate /* 12345 */ 7747c478bd9Sstevel@tonic-gate "addr mask " 7757c478bd9Sstevel@tonic-gate /* 123.123.123.123 123.123.123.123 */ 7767c478bd9Sstevel@tonic-gate "src gateway mxfrg rtt rtt_sd ssthresh ref " 7777c478bd9Sstevel@tonic-gate /* 123.123.123.123 123.123.123.123 12345 12345 123456 12345678 123 */ 7787c478bd9Sstevel@tonic-gate "rtomax tstamp_ok wscale_ok ecn_ok pmtud_ok sack sendpipe " 7797c478bd9Sstevel@tonic-gate /* 123456 123456789 123456789 123456 12345678 1234 12345678 */ 7807c478bd9Sstevel@tonic-gate "recvpipe in/out/forward type"); 7817c478bd9Sstevel@tonic-gate /* 12345678 in/out/forward xxxxxxxxxx */ 7827c478bd9Sstevel@tonic-gate 7837c478bd9Sstevel@tonic-gate /* 7847c478bd9Sstevel@tonic-gate * Because of the ndd constraint, at most we can have 64K buffer 7857c478bd9Sstevel@tonic-gate * to put in all IRE info. So to be more efficient, just 7867c478bd9Sstevel@tonic-gate * allocate a 64K buffer here, assuming we need that large buffer. 7877c478bd9Sstevel@tonic-gate * This should be OK as only root can do ndd /dev/ip. 7887c478bd9Sstevel@tonic-gate */ 7897c478bd9Sstevel@tonic-gate if ((mp->b_cont = allocb(ND_MAX_BUF_LEN, BPRI_HI)) == NULL) { 7907c478bd9Sstevel@tonic-gate /* The following may work even if we cannot get a large buf. */ 7917c478bd9Sstevel@tonic-gate (void) mi_mpprintf(mp, "<< Out of buffer >>\n"); 7927c478bd9Sstevel@tonic-gate return (0); 7937c478bd9Sstevel@tonic-gate } 7947c478bd9Sstevel@tonic-gate 7957c478bd9Sstevel@tonic-gate zoneid = Q_TO_CONN(q)->conn_zoneid; 7967c478bd9Sstevel@tonic-gate if (zoneid == GLOBAL_ZONEID) 7977c478bd9Sstevel@tonic-gate zoneid = ALL_ZONES; 7987c478bd9Sstevel@tonic-gate 799*f4b3ec61Sdh155122 ire_walk_v4(ire_report_ftable, mp->b_cont, zoneid, ipst); 800*f4b3ec61Sdh155122 ire_walk_v4(ire_report_ctable, mp->b_cont, zoneid, ipst); 8017c478bd9Sstevel@tonic-gate 8027c478bd9Sstevel@tonic-gate return (0); 8037c478bd9Sstevel@tonic-gate } 8047c478bd9Sstevel@tonic-gate 8057c478bd9Sstevel@tonic-gate 8067c478bd9Sstevel@tonic-gate /* ire_walk routine invoked for ip_ire_report for each cached IRE. */ 8077c478bd9Sstevel@tonic-gate static void 8087c478bd9Sstevel@tonic-gate ire_report_ctable(ire_t *ire, char *mp) 8097c478bd9Sstevel@tonic-gate { 8107c478bd9Sstevel@tonic-gate char buf1[16]; 8117c478bd9Sstevel@tonic-gate char buf2[16]; 8127c478bd9Sstevel@tonic-gate char buf3[16]; 8137c478bd9Sstevel@tonic-gate char buf4[16]; 8147c478bd9Sstevel@tonic-gate uint_t fo_pkt_count; 8157c478bd9Sstevel@tonic-gate uint_t ib_pkt_count; 8167c478bd9Sstevel@tonic-gate int ref; 8177c478bd9Sstevel@tonic-gate uint_t print_len, buf_len; 8187c478bd9Sstevel@tonic-gate 8197c478bd9Sstevel@tonic-gate if ((ire->ire_type & IRE_CACHETABLE) == 0) 8207c478bd9Sstevel@tonic-gate return; 8217c478bd9Sstevel@tonic-gate buf_len = ((mblk_t *)mp)->b_datap->db_lim - ((mblk_t *)mp)->b_wptr; 8227c478bd9Sstevel@tonic-gate if (buf_len <= 0) 8237c478bd9Sstevel@tonic-gate return; 8247c478bd9Sstevel@tonic-gate 8257c478bd9Sstevel@tonic-gate /* Number of active references of this ire */ 8267c478bd9Sstevel@tonic-gate ref = ire->ire_refcnt; 8277c478bd9Sstevel@tonic-gate /* "inbound" to a non local address is a forward */ 8287c478bd9Sstevel@tonic-gate ib_pkt_count = ire->ire_ib_pkt_count; 8297c478bd9Sstevel@tonic-gate fo_pkt_count = 0; 8307c478bd9Sstevel@tonic-gate if (!(ire->ire_type & (IRE_LOCAL|IRE_BROADCAST))) { 8317c478bd9Sstevel@tonic-gate fo_pkt_count = ib_pkt_count; 8327c478bd9Sstevel@tonic-gate ib_pkt_count = 0; 8337c478bd9Sstevel@tonic-gate } 8347c478bd9Sstevel@tonic-gate print_len = snprintf((char *)((mblk_t *)mp)->b_wptr, buf_len, 8357c478bd9Sstevel@tonic-gate MI_COL_PTRFMT_STR MI_COL_PTRFMT_STR MI_COL_PTRFMT_STR "%5d " 8367c478bd9Sstevel@tonic-gate "%s %s %s %s %05d %05ld %06ld %08d %03d %06d %09d %09d %06d %08d " 8377c478bd9Sstevel@tonic-gate "%04d %08d %08d %d/%d/%d %s\n", 8387c478bd9Sstevel@tonic-gate (void *)ire, (void *)ire->ire_rfq, (void *)ire->ire_stq, 8397c478bd9Sstevel@tonic-gate (int)ire->ire_zoneid, 8407c478bd9Sstevel@tonic-gate ip_dot_addr(ire->ire_addr, buf1), ip_dot_addr(ire->ire_mask, buf2), 8417c478bd9Sstevel@tonic-gate ip_dot_addr(ire->ire_src_addr, buf3), 8427c478bd9Sstevel@tonic-gate ip_dot_addr(ire->ire_gateway_addr, buf4), 8437c478bd9Sstevel@tonic-gate ire->ire_max_frag, ire->ire_uinfo.iulp_rtt, 8447c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtt_sd, ire->ire_uinfo.iulp_ssthresh, ref, 8457c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_rtomax, 8467c478bd9Sstevel@tonic-gate (ire->ire_uinfo.iulp_tstamp_ok ? 1: 0), 8477c478bd9Sstevel@tonic-gate (ire->ire_uinfo.iulp_wscale_ok ? 1: 0), 8487c478bd9Sstevel@tonic-gate (ire->ire_uinfo.iulp_ecn_ok ? 1: 0), 8497c478bd9Sstevel@tonic-gate (ire->ire_uinfo.iulp_pmtud_ok ? 1: 0), 8507c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_sack, 8517c478bd9Sstevel@tonic-gate ire->ire_uinfo.iulp_spipe, ire->ire_uinfo.iulp_rpipe, 8527c478bd9Sstevel@tonic-gate ib_pkt_count, ire->ire_ob_pkt_count, fo_pkt_count, 8537c478bd9Sstevel@tonic-gate ip_nv_lookup(ire_nv_tbl, (int)ire->ire_type)); 8547c478bd9Sstevel@tonic-gate if (print_len < buf_len) { 8557c478bd9Sstevel@tonic-gate ((mblk_t *)mp)->b_wptr += print_len; 8567c478bd9Sstevel@tonic-gate } else { 8577c478bd9Sstevel@tonic-gate ((mblk_t *)mp)->b_wptr += buf_len; 8587c478bd9Sstevel@tonic-gate } 8597c478bd9Sstevel@tonic-gate } 8607c478bd9Sstevel@tonic-gate 8617c478bd9Sstevel@tonic-gate /* ARGSUSED */ 8627c478bd9Sstevel@tonic-gate int 8637c478bd9Sstevel@tonic-gate ip_ire_report_mrtun(queue_t *q, mblk_t *mp, caddr_t arg, cred_t *ioc_cr) 8647c478bd9Sstevel@tonic-gate { 865*f4b3ec61Sdh155122 ip_stack_t *ipst; 866*f4b3ec61Sdh155122 867*f4b3ec61Sdh155122 if (CONN_Q(q)) 868*f4b3ec61Sdh155122 ipst = CONNQ_TO_IPST(q); 869*f4b3ec61Sdh155122 else 870*f4b3ec61Sdh155122 ipst = ILLQ_TO_IPST(q); 871*f4b3ec61Sdh155122 8727c478bd9Sstevel@tonic-gate (void) mi_mpprintf(mp, 8737c478bd9Sstevel@tonic-gate "IRE " MI_COL_HDRPAD_STR 8747c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 8757c478bd9Sstevel@tonic-gate "stq " MI_COL_HDRPAD_STR 8767c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 8777c478bd9Sstevel@tonic-gate "in_ill " MI_COL_HDRPAD_STR 8787c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 8797c478bd9Sstevel@tonic-gate "in_src_addr " 8807c478bd9Sstevel@tonic-gate /* 123.123.123.123 */ 8817c478bd9Sstevel@tonic-gate "max_frag " 8827c478bd9Sstevel@tonic-gate /* 12345 */ 8837c478bd9Sstevel@tonic-gate "ref "); 8847c478bd9Sstevel@tonic-gate /* 123 */ 8857c478bd9Sstevel@tonic-gate 886*f4b3ec61Sdh155122 ire_walk_ill_mrtun(0, 0, ire_report_mrtun_table, mp, NULL, 887*f4b3ec61Sdh155122 ipst); 8887c478bd9Sstevel@tonic-gate return (0); 8897c478bd9Sstevel@tonic-gate } 8907c478bd9Sstevel@tonic-gate 8917c478bd9Sstevel@tonic-gate /* mrtun report table - supports ipv4_mrtun_ire_status ndd variable */ 8927c478bd9Sstevel@tonic-gate 8937c478bd9Sstevel@tonic-gate static void 8947c478bd9Sstevel@tonic-gate ire_report_mrtun_table(ire_t *ire, char *mp) 8957c478bd9Sstevel@tonic-gate { 8967c478bd9Sstevel@tonic-gate char buf1[INET_ADDRSTRLEN]; 8977c478bd9Sstevel@tonic-gate int ref; 8987c478bd9Sstevel@tonic-gate 8997c478bd9Sstevel@tonic-gate /* Number of active references of this ire */ 9007c478bd9Sstevel@tonic-gate ref = ire->ire_refcnt; 9017c478bd9Sstevel@tonic-gate ASSERT(ire->ire_type == IRE_MIPRTUN); 9027c478bd9Sstevel@tonic-gate (void) mi_mpprintf((mblk_t *)mp, 9037c478bd9Sstevel@tonic-gate MI_COL_PTRFMT_STR MI_COL_PTRFMT_STR MI_COL_PTRFMT_STR 9047c478bd9Sstevel@tonic-gate "%s %05d %03d", 9057c478bd9Sstevel@tonic-gate (void *)ire, (void *)ire->ire_stq, 9067c478bd9Sstevel@tonic-gate (void *)ire->ire_in_ill, 9077c478bd9Sstevel@tonic-gate ip_dot_addr(ire->ire_in_src_addr, buf1), 9087c478bd9Sstevel@tonic-gate ire->ire_max_frag, ref); 9097c478bd9Sstevel@tonic-gate } 9107c478bd9Sstevel@tonic-gate 9117c478bd9Sstevel@tonic-gate /* 9127c478bd9Sstevel@tonic-gate * Dispatch routine to format ires in interface based routine 9137c478bd9Sstevel@tonic-gate */ 9147c478bd9Sstevel@tonic-gate /* ARGSUSED */ 9157c478bd9Sstevel@tonic-gate int 9167c478bd9Sstevel@tonic-gate ip_ire_report_srcif(queue_t *q, mblk_t *mp, caddr_t arg, cred_t *ioc_cr) 9177c478bd9Sstevel@tonic-gate { 918*f4b3ec61Sdh155122 ip_stack_t *ipst; 919*f4b3ec61Sdh155122 920*f4b3ec61Sdh155122 if (CONN_Q(q)) 921*f4b3ec61Sdh155122 ipst = CONNQ_TO_IPST(q); 922*f4b3ec61Sdh155122 else 923*f4b3ec61Sdh155122 ipst = ILLQ_TO_IPST(q); 9247c478bd9Sstevel@tonic-gate 9257c478bd9Sstevel@tonic-gate /* Report all interface based ires */ 9267c478bd9Sstevel@tonic-gate 9277c478bd9Sstevel@tonic-gate (void) mi_mpprintf(mp, 9287c478bd9Sstevel@tonic-gate "IRE " MI_COL_HDRPAD_STR 9297c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 9307c478bd9Sstevel@tonic-gate "stq " MI_COL_HDRPAD_STR 9317c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 9327c478bd9Sstevel@tonic-gate "in_ill " MI_COL_HDRPAD_STR 9337c478bd9Sstevel@tonic-gate /* 01234567[89ABCDEF] */ 9347c478bd9Sstevel@tonic-gate "addr " 9357c478bd9Sstevel@tonic-gate /* 123.123.123.123 */ 9367c478bd9Sstevel@tonic-gate "gateway " 9377c478bd9Sstevel@tonic-gate /* 123.123.123.123 */ 9387c478bd9Sstevel@tonic-gate "max_frag " 9397c478bd9Sstevel@tonic-gate /* 12345 */ 9407c478bd9Sstevel@tonic-gate "ref " 9417c478bd9Sstevel@tonic-gate /* 123 */ 9427c478bd9Sstevel@tonic-gate "type " 9437c478bd9Sstevel@tonic-gate /* ABCDEFGH */ 9447c478bd9Sstevel@tonic-gate "in/out/forward"); 945*f4b3ec61Sdh155122 ire_walk_srcif_table_v4(ire_report_srcif_table, mp, ipst); 9467c478bd9Sstevel@tonic-gate return (0); 9477c478bd9Sstevel@tonic-gate } 9487c478bd9Sstevel@tonic-gate 9497c478bd9Sstevel@tonic-gate /* Reports the interface table ires */ 950*f4b3ec61Sdh155122 /* ARGSUSED2 */ 9517c478bd9Sstevel@tonic-gate static void 952*f4b3ec61Sdh155122 ire_report_srcif_table(ire_t *ire, char *mp, ip_stack_t *ipst) 9537c478bd9Sstevel@tonic-gate { 9547c478bd9Sstevel@tonic-gate char buf1[INET_ADDRSTRLEN]; 9557c478bd9Sstevel@tonic-gate char buf2[INET_ADDRSTRLEN]; 9567c478bd9Sstevel@tonic-gate int ref; 9577c478bd9Sstevel@tonic-gate 9587c478bd9Sstevel@tonic-gate ref = ire->ire_refcnt; 9597c478bd9Sstevel@tonic-gate (void) mi_mpprintf((mblk_t *)mp, 9607c478bd9Sstevel@tonic-gate MI_COL_PTRFMT_STR MI_COL_PTRFMT_STR MI_COL_PTRFMT_STR 9617c478bd9Sstevel@tonic-gate "%s %s %05d %03d %s %d", 9627c478bd9Sstevel@tonic-gate (void *)ire, (void *)ire->ire_stq, 9637c478bd9Sstevel@tonic-gate (void *)ire->ire_in_ill, 9647c478bd9Sstevel@tonic-gate ip_dot_addr(ire->ire_addr, buf1), 9657c478bd9Sstevel@tonic-gate ip_dot_addr(ire->ire_gateway_addr, buf2), 9667c478bd9Sstevel@tonic-gate ire->ire_max_frag, ref, 9677c478bd9Sstevel@tonic-gate ip_nv_lookup(ire_nv_tbl, (int)ire->ire_type), 9687c478bd9Sstevel@tonic-gate ire->ire_ib_pkt_count); 9697c478bd9Sstevel@tonic-gate 9707c478bd9Sstevel@tonic-gate } 9717c478bd9Sstevel@tonic-gate /* 9727c478bd9Sstevel@tonic-gate * ip_ire_req is called by ip_wput when an IRE_DB_REQ_TYPE message is handed 9737c478bd9Sstevel@tonic-gate * down from the Upper Level Protocol to request a copy of the IRE (to check 9747c478bd9Sstevel@tonic-gate * its type or to extract information like round-trip time estimates or the 9757c478bd9Sstevel@tonic-gate * MTU.) 9767c478bd9Sstevel@tonic-gate * The address is assumed to be in the ire_addr field. If no IRE is found 9777c478bd9Sstevel@tonic-gate * an IRE is returned with ire_type being zero. 9787c478bd9Sstevel@tonic-gate * Note that the upper lavel protocol has to check for broadcast 9797c478bd9Sstevel@tonic-gate * (IRE_BROADCAST) and multicast (CLASSD(addr)). 9807c478bd9Sstevel@tonic-gate * If there is a b_cont the resulting IRE_DB_TYPE mblk is placed at the 9817c478bd9Sstevel@tonic-gate * end of the returned message. 9827c478bd9Sstevel@tonic-gate * 9837c478bd9Sstevel@tonic-gate * TCP sends down a message of this type with a connection request packet 9847c478bd9Sstevel@tonic-gate * chained on. UDP and ICMP send it down to verify that a route exists for 9857c478bd9Sstevel@tonic-gate * the destination address when they get connected. 9867c478bd9Sstevel@tonic-gate */ 9877c478bd9Sstevel@tonic-gate void 9887c478bd9Sstevel@tonic-gate ip_ire_req(queue_t *q, mblk_t *mp) 9897c478bd9Sstevel@tonic-gate { 9907c478bd9Sstevel@tonic-gate ire_t *inire; 9917c478bd9Sstevel@tonic-gate ire_t *ire; 9927c478bd9Sstevel@tonic-gate mblk_t *mp1; 9937c478bd9Sstevel@tonic-gate ire_t *sire = NULL; 9947c478bd9Sstevel@tonic-gate zoneid_t zoneid = Q_TO_CONN(q)->conn_zoneid; 995*f4b3ec61Sdh155122 ip_stack_t *ipst = CONNQ_TO_IPST(q); 996*f4b3ec61Sdh155122 997*f4b3ec61Sdh155122 ASSERT(q->q_next == NULL); 9987c478bd9Sstevel@tonic-gate 9997c478bd9Sstevel@tonic-gate if ((mp->b_wptr - mp->b_rptr) < sizeof (ire_t) || 10007c478bd9Sstevel@tonic-gate !OK_32PTR(mp->b_rptr)) { 10017c478bd9Sstevel@tonic-gate freemsg(mp); 10027c478bd9Sstevel@tonic-gate return; 10037c478bd9Sstevel@tonic-gate } 10047c478bd9Sstevel@tonic-gate inire = (ire_t *)mp->b_rptr; 10057c478bd9Sstevel@tonic-gate /* 10067c478bd9Sstevel@tonic-gate * Got it, now take our best shot at an IRE. 10077c478bd9Sstevel@tonic-gate */ 10087c478bd9Sstevel@tonic-gate if (inire->ire_ipversion == IPV6_VERSION) { 10097c478bd9Sstevel@tonic-gate ire = ire_route_lookup_v6(&inire->ire_addr_v6, 0, 0, 0, 101045916cd2Sjpk NULL, &sire, zoneid, NULL, 1011*f4b3ec61Sdh155122 (MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT), ipst); 10127c478bd9Sstevel@tonic-gate } else { 10137c478bd9Sstevel@tonic-gate ASSERT(inire->ire_ipversion == IPV4_VERSION); 10147c478bd9Sstevel@tonic-gate ire = ire_route_lookup(inire->ire_addr, 0, 0, 0, 101545916cd2Sjpk NULL, &sire, zoneid, NULL, 1016*f4b3ec61Sdh155122 (MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT), ipst); 10177c478bd9Sstevel@tonic-gate } 10187c478bd9Sstevel@tonic-gate 10197c478bd9Sstevel@tonic-gate /* 10207c478bd9Sstevel@tonic-gate * We prevent returning IRES with source address INADDR_ANY 10217c478bd9Sstevel@tonic-gate * as these were temporarily created for sending packets 10227c478bd9Sstevel@tonic-gate * from endpoints that have conn_unspec_src set. 10237c478bd9Sstevel@tonic-gate */ 10247c478bd9Sstevel@tonic-gate if (ire == NULL || 10257c478bd9Sstevel@tonic-gate (ire->ire_ipversion == IPV4_VERSION && 10267c478bd9Sstevel@tonic-gate ire->ire_src_addr == INADDR_ANY) || 10277c478bd9Sstevel@tonic-gate (ire->ire_ipversion == IPV6_VERSION && 10287c478bd9Sstevel@tonic-gate IN6_IS_ADDR_UNSPECIFIED(&ire->ire_src_addr_v6))) { 10297c478bd9Sstevel@tonic-gate inire->ire_type = 0; 10307c478bd9Sstevel@tonic-gate } else { 10317c478bd9Sstevel@tonic-gate bcopy(ire, inire, sizeof (ire_t)); 10327c478bd9Sstevel@tonic-gate /* Copy the route metrics from the parent. */ 10337c478bd9Sstevel@tonic-gate if (sire != NULL) { 10347c478bd9Sstevel@tonic-gate bcopy(&(sire->ire_uinfo), &(inire->ire_uinfo), 10357c478bd9Sstevel@tonic-gate sizeof (iulp_t)); 10367c478bd9Sstevel@tonic-gate } 10377c478bd9Sstevel@tonic-gate 10387c478bd9Sstevel@tonic-gate /* 10397c478bd9Sstevel@tonic-gate * As we don't lookup global policy here, we may not 10407c478bd9Sstevel@tonic-gate * pass the right size if per-socket policy is not 10417c478bd9Sstevel@tonic-gate * present. For these cases, path mtu discovery will 10427c478bd9Sstevel@tonic-gate * do the right thing. 10437c478bd9Sstevel@tonic-gate */ 10447c478bd9Sstevel@tonic-gate inire->ire_ipsec_overhead = conn_ipsec_length(Q_TO_CONN(q)); 10457c478bd9Sstevel@tonic-gate 10467c478bd9Sstevel@tonic-gate /* Pass the latest setting of the ip_path_mtu_discovery */ 1047*f4b3ec61Sdh155122 inire->ire_frag_flag |= 1048*f4b3ec61Sdh155122 (ipst->ips_ip_path_mtu_discovery) ? IPH_DF : 0; 10497c478bd9Sstevel@tonic-gate } 10507c478bd9Sstevel@tonic-gate if (ire != NULL) 10517c478bd9Sstevel@tonic-gate ire_refrele(ire); 10527c478bd9Sstevel@tonic-gate if (sire != NULL) 10537c478bd9Sstevel@tonic-gate ire_refrele(sire); 10547c478bd9Sstevel@tonic-gate mp->b_wptr = &mp->b_rptr[sizeof (ire_t)]; 10557c478bd9Sstevel@tonic-gate mp->b_datap->db_type = IRE_DB_TYPE; 10567c478bd9Sstevel@tonic-gate 10577c478bd9Sstevel@tonic-gate /* Put the IRE_DB_TYPE mblk last in the chain */ 10587c478bd9Sstevel@tonic-gate mp1 = mp->b_cont; 10597c478bd9Sstevel@tonic-gate if (mp1 != NULL) { 10607c478bd9Sstevel@tonic-gate mp->b_cont = NULL; 10617c478bd9Sstevel@tonic-gate linkb(mp1, mp); 10627c478bd9Sstevel@tonic-gate mp = mp1; 10637c478bd9Sstevel@tonic-gate } 10647c478bd9Sstevel@tonic-gate qreply(q, mp); 10657c478bd9Sstevel@tonic-gate } 10667c478bd9Sstevel@tonic-gate 10677c478bd9Sstevel@tonic-gate /* 10687c478bd9Sstevel@tonic-gate * Send a packet using the specified IRE. 10697c478bd9Sstevel@tonic-gate * If ire_src_addr_v6 is all zero then discard the IRE after 10707c478bd9Sstevel@tonic-gate * the packet has been sent. 10717c478bd9Sstevel@tonic-gate */ 10727c478bd9Sstevel@tonic-gate static void 10737c478bd9Sstevel@tonic-gate ire_send(queue_t *q, mblk_t *pkt, ire_t *ire) 10747c478bd9Sstevel@tonic-gate { 10757c478bd9Sstevel@tonic-gate mblk_t *ipsec_mp; 10767c478bd9Sstevel@tonic-gate boolean_t is_secure; 10777c478bd9Sstevel@tonic-gate uint_t ifindex; 10787c478bd9Sstevel@tonic-gate ill_t *ill; 10795597b60aSnordmark zoneid_t zoneid = ire->ire_zoneid; 1080*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 10817c478bd9Sstevel@tonic-gate 10827c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 10835597b60aSnordmark ASSERT(!(ire->ire_type & IRE_LOCAL)); /* Has different ire_zoneid */ 10847c478bd9Sstevel@tonic-gate ipsec_mp = pkt; 10857c478bd9Sstevel@tonic-gate is_secure = (pkt->b_datap->db_type == M_CTL); 10865597b60aSnordmark if (is_secure) { 10875597b60aSnordmark ipsec_out_t *io; 10885597b60aSnordmark 10897c478bd9Sstevel@tonic-gate pkt = pkt->b_cont; 10905597b60aSnordmark io = (ipsec_out_t *)ipsec_mp->b_rptr; 10915597b60aSnordmark if (io->ipsec_out_type == IPSEC_OUT) 10925597b60aSnordmark zoneid = io->ipsec_out_zoneid; 10935597b60aSnordmark } 10947c478bd9Sstevel@tonic-gate 10957c478bd9Sstevel@tonic-gate /* If the packet originated externally then */ 10967c478bd9Sstevel@tonic-gate if (pkt->b_prev) { 10977c478bd9Sstevel@tonic-gate ire_refrele(ire); 10987c478bd9Sstevel@tonic-gate /* 10997c478bd9Sstevel@tonic-gate * Extract the ifindex from b_prev (set in ip_rput_noire). 11007c478bd9Sstevel@tonic-gate * Look up interface to see if it still exists (it could have 11017c478bd9Sstevel@tonic-gate * been unplumbed by the time the reply came back from ARP) 11027c478bd9Sstevel@tonic-gate */ 11037c478bd9Sstevel@tonic-gate ifindex = (uint_t)(uintptr_t)pkt->b_prev; 11047c478bd9Sstevel@tonic-gate ill = ill_lookup_on_ifindex(ifindex, B_FALSE, 1105*f4b3ec61Sdh155122 NULL, NULL, NULL, NULL, ipst); 11067c478bd9Sstevel@tonic-gate if (ill == NULL) { 11077c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 11087c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 11097c478bd9Sstevel@tonic-gate freemsg(ipsec_mp); 11107c478bd9Sstevel@tonic-gate return; 11117c478bd9Sstevel@tonic-gate } 11127c478bd9Sstevel@tonic-gate q = ill->ill_rq; 11137c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 11147c478bd9Sstevel@tonic-gate /* 11157c478bd9Sstevel@tonic-gate * This packet has not gone through IPSEC processing 11167c478bd9Sstevel@tonic-gate * and hence we should not have any IPSEC message 11177c478bd9Sstevel@tonic-gate * prepended. 11187c478bd9Sstevel@tonic-gate */ 11197c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 1120c793af95Ssangeeta put(q, pkt); 11217c478bd9Sstevel@tonic-gate ill_refrele(ill); 11227c478bd9Sstevel@tonic-gate } else if (pkt->b_next) { 11237c478bd9Sstevel@tonic-gate /* Packets from multicast router */ 11247c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 11257c478bd9Sstevel@tonic-gate /* 11267c478bd9Sstevel@tonic-gate * We never get the IPSEC_OUT while forwarding the 11277c478bd9Sstevel@tonic-gate * packet for multicast router. 11287c478bd9Sstevel@tonic-gate */ 11297c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 11307c478bd9Sstevel@tonic-gate ip_rput_forward(ire, (ipha_t *)pkt->b_rptr, ipsec_mp, NULL); 11317c478bd9Sstevel@tonic-gate ire_refrele(ire); 11327c478bd9Sstevel@tonic-gate } else { 11337c478bd9Sstevel@tonic-gate /* Locally originated packets */ 11347c478bd9Sstevel@tonic-gate boolean_t is_inaddr_any; 11357c478bd9Sstevel@tonic-gate ipha_t *ipha = (ipha_t *)pkt->b_rptr; 11367c478bd9Sstevel@tonic-gate 11377c478bd9Sstevel@tonic-gate /* 11387c478bd9Sstevel@tonic-gate * We need to do an ire_delete below for which 11397c478bd9Sstevel@tonic-gate * we need to make sure that the IRE will be 11407c478bd9Sstevel@tonic-gate * around even after calling ip_wput_ire - 11417c478bd9Sstevel@tonic-gate * which does ire_refrele. Otherwise somebody 11427c478bd9Sstevel@tonic-gate * could potentially delete this ire and hence 11437c478bd9Sstevel@tonic-gate * free this ire and we will be calling ire_delete 11447c478bd9Sstevel@tonic-gate * on a freed ire below. 11457c478bd9Sstevel@tonic-gate */ 11467c478bd9Sstevel@tonic-gate is_inaddr_any = (ire->ire_src_addr == INADDR_ANY); 11477c478bd9Sstevel@tonic-gate if (is_inaddr_any) { 11487c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 11497c478bd9Sstevel@tonic-gate } 11507c478bd9Sstevel@tonic-gate /* 11517c478bd9Sstevel@tonic-gate * If we were resolving a router we can not use the 11527c478bd9Sstevel@tonic-gate * routers IRE for sending the packet (since it would 11537c478bd9Sstevel@tonic-gate * violate the uniqness of the IP idents) thus we 11547c478bd9Sstevel@tonic-gate * make another pass through ip_wput to create the IRE_CACHE 11557c478bd9Sstevel@tonic-gate * for the destination. 11567c478bd9Sstevel@tonic-gate * When IRE_MARK_NOADD is set, ire_add() is not called. 11577c478bd9Sstevel@tonic-gate * Thus ip_wput() will never find a ire and result in an 11587c478bd9Sstevel@tonic-gate * infinite loop. Thus we check whether IRE_MARK_NOADD is 11597c478bd9Sstevel@tonic-gate * is set. This also implies that IRE_MARK_NOADD can only be 11607c478bd9Sstevel@tonic-gate * used to send packets to directly connected hosts. 11617c478bd9Sstevel@tonic-gate */ 11627c478bd9Sstevel@tonic-gate if (ipha->ipha_dst != ire->ire_addr && 11637c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_NOADD)) { 11647c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add */ 11655597b60aSnordmark if (CONN_Q(q)) { 11665597b60aSnordmark (void) ip_output(Q_TO_CONN(q), ipsec_mp, q, 11675597b60aSnordmark IRE_SEND); 11685597b60aSnordmark } else { 11695597b60aSnordmark (void) ip_output((void *)(uintptr_t)zoneid, 11705597b60aSnordmark ipsec_mp, q, IRE_SEND); 11715597b60aSnordmark } 11727c478bd9Sstevel@tonic-gate } else { 11737c478bd9Sstevel@tonic-gate if (is_secure) { 11747c478bd9Sstevel@tonic-gate ipsec_out_t *oi; 11757c478bd9Sstevel@tonic-gate ipha_t *ipha; 11767c478bd9Sstevel@tonic-gate 11777c478bd9Sstevel@tonic-gate oi = (ipsec_out_t *)ipsec_mp->b_rptr; 11787c478bd9Sstevel@tonic-gate ipha = (ipha_t *)ipsec_mp->b_cont->b_rptr; 11797c478bd9Sstevel@tonic-gate if (oi->ipsec_out_proc_begin) { 11807c478bd9Sstevel@tonic-gate /* 11817c478bd9Sstevel@tonic-gate * This is the case where 11827c478bd9Sstevel@tonic-gate * ip_wput_ipsec_out could not find 11837c478bd9Sstevel@tonic-gate * the IRE and recreated a new one. 11847c478bd9Sstevel@tonic-gate * As ip_wput_ipsec_out does ire 11857c478bd9Sstevel@tonic-gate * lookups, ire_refrele for the extra 11867c478bd9Sstevel@tonic-gate * bump in ire_add. 11877c478bd9Sstevel@tonic-gate */ 11887c478bd9Sstevel@tonic-gate ire_refrele(ire); 11897c478bd9Sstevel@tonic-gate ip_wput_ipsec_out(q, ipsec_mp, ipha, 11907c478bd9Sstevel@tonic-gate NULL, NULL); 11917c478bd9Sstevel@tonic-gate } else { 11927c478bd9Sstevel@tonic-gate /* 11937c478bd9Sstevel@tonic-gate * IRE_REFRELE will be done in 11947c478bd9Sstevel@tonic-gate * ip_wput_ire. 11957c478bd9Sstevel@tonic-gate */ 11967c478bd9Sstevel@tonic-gate ip_wput_ire(q, ipsec_mp, ire, NULL, 11975597b60aSnordmark IRE_SEND, zoneid); 11987c478bd9Sstevel@tonic-gate } 11997c478bd9Sstevel@tonic-gate } else { 12007c478bd9Sstevel@tonic-gate /* 12017c478bd9Sstevel@tonic-gate * IRE_REFRELE will be done in ip_wput_ire. 12027c478bd9Sstevel@tonic-gate */ 12037c478bd9Sstevel@tonic-gate ip_wput_ire(q, ipsec_mp, ire, NULL, 12045597b60aSnordmark IRE_SEND, zoneid); 12057c478bd9Sstevel@tonic-gate } 12067c478bd9Sstevel@tonic-gate } 12077c478bd9Sstevel@tonic-gate /* 12087c478bd9Sstevel@tonic-gate * Special code to support sending a single packet with 12097c478bd9Sstevel@tonic-gate * conn_unspec_src using an IRE which has no source address. 12107c478bd9Sstevel@tonic-gate * The IRE is deleted here after sending the packet to avoid 12117c478bd9Sstevel@tonic-gate * having other code trip on it. But before we delete the 12127c478bd9Sstevel@tonic-gate * ire, somebody could have looked up this ire. 12137c478bd9Sstevel@tonic-gate * We prevent returning/using this IRE by the upper layers 12147c478bd9Sstevel@tonic-gate * by making checks to NULL source address in other places 12157c478bd9Sstevel@tonic-gate * like e.g ip_ire_append, ip_ire_req and ip_bind_connected. 12167c478bd9Sstevel@tonic-gate * Though, this does not completely prevent other threads 12177c478bd9Sstevel@tonic-gate * from using this ire, this should not cause any problems. 12187c478bd9Sstevel@tonic-gate * 12197c478bd9Sstevel@tonic-gate * NOTE : We use is_inaddr_any instead of using ire_src_addr 12207c478bd9Sstevel@tonic-gate * because for the normal case i.e !is_inaddr_any, ire_refrele 12217c478bd9Sstevel@tonic-gate * above could have potentially freed the ire. 12227c478bd9Sstevel@tonic-gate */ 12237c478bd9Sstevel@tonic-gate if (is_inaddr_any) { 12247c478bd9Sstevel@tonic-gate /* 12257c478bd9Sstevel@tonic-gate * If this IRE has been deleted by another thread, then 12267c478bd9Sstevel@tonic-gate * ire_bucket won't be NULL, but ire_ptpn will be NULL. 12277c478bd9Sstevel@tonic-gate * Thus, ire_delete will do nothing. This check 12287c478bd9Sstevel@tonic-gate * guards against calling ire_delete when the IRE was 12297c478bd9Sstevel@tonic-gate * never inserted in the table, which is handled by 12307c478bd9Sstevel@tonic-gate * ire_delete as dropping another reference. 12317c478bd9Sstevel@tonic-gate */ 12327c478bd9Sstevel@tonic-gate if (ire->ire_bucket != NULL) { 12337c478bd9Sstevel@tonic-gate ip1dbg(("ire_send: delete IRE\n")); 12347c478bd9Sstevel@tonic-gate ire_delete(ire); 12357c478bd9Sstevel@tonic-gate } 12367c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held above */ 12377c478bd9Sstevel@tonic-gate } 12387c478bd9Sstevel@tonic-gate } 12397c478bd9Sstevel@tonic-gate } 12407c478bd9Sstevel@tonic-gate 12417c478bd9Sstevel@tonic-gate /* 12427c478bd9Sstevel@tonic-gate * Send a packet using the specified IRE. 12437c478bd9Sstevel@tonic-gate * If ire_src_addr_v6 is all zero then discard the IRE after 12447c478bd9Sstevel@tonic-gate * the packet has been sent. 12457c478bd9Sstevel@tonic-gate */ 12467c478bd9Sstevel@tonic-gate static void 12477c478bd9Sstevel@tonic-gate ire_send_v6(queue_t *q, mblk_t *pkt, ire_t *ire) 12487c478bd9Sstevel@tonic-gate { 12497c478bd9Sstevel@tonic-gate mblk_t *ipsec_mp; 12507c478bd9Sstevel@tonic-gate boolean_t secure; 12517c478bd9Sstevel@tonic-gate uint_t ifindex; 12525597b60aSnordmark zoneid_t zoneid = ire->ire_zoneid; 1253*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 12547c478bd9Sstevel@tonic-gate 12557c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV6_VERSION); 12565597b60aSnordmark ASSERT(!(ire->ire_type & IRE_LOCAL)); /* Has different ire_zoneid */ 12577c478bd9Sstevel@tonic-gate if (pkt->b_datap->db_type == M_CTL) { 12585597b60aSnordmark ipsec_out_t *io; 12595597b60aSnordmark 12607c478bd9Sstevel@tonic-gate ipsec_mp = pkt; 12617c478bd9Sstevel@tonic-gate pkt = pkt->b_cont; 12627c478bd9Sstevel@tonic-gate secure = B_TRUE; 12635597b60aSnordmark io = (ipsec_out_t *)ipsec_mp->b_rptr; 12645597b60aSnordmark if (io->ipsec_out_type == IPSEC_OUT) 12655597b60aSnordmark zoneid = io->ipsec_out_zoneid; 12667c478bd9Sstevel@tonic-gate } else { 12677c478bd9Sstevel@tonic-gate ipsec_mp = pkt; 12687c478bd9Sstevel@tonic-gate secure = B_FALSE; 12697c478bd9Sstevel@tonic-gate } 12707c478bd9Sstevel@tonic-gate 12717c478bd9Sstevel@tonic-gate /* If the packet originated externally then */ 12727c478bd9Sstevel@tonic-gate if (pkt->b_prev) { 12737c478bd9Sstevel@tonic-gate ill_t *ill; 12747c478bd9Sstevel@tonic-gate /* 12757c478bd9Sstevel@tonic-gate * Extract the ifindex from b_prev (set in ip_rput_data_v6). 12767c478bd9Sstevel@tonic-gate * Look up interface to see if it still exists (it could have 12777c478bd9Sstevel@tonic-gate * been unplumbed by the time the reply came back from the 1278c793af95Ssangeeta * resolver). 12797c478bd9Sstevel@tonic-gate */ 12807c478bd9Sstevel@tonic-gate ifindex = (uint_t)(uintptr_t)pkt->b_prev; 12817c478bd9Sstevel@tonic-gate ill = ill_lookup_on_ifindex(ifindex, B_TRUE, 1282*f4b3ec61Sdh155122 NULL, NULL, NULL, NULL, ipst); 12837c478bd9Sstevel@tonic-gate if (ill == NULL) { 12847c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 12857c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 12867c478bd9Sstevel@tonic-gate freemsg(ipsec_mp); 12877c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add */ 12887c478bd9Sstevel@tonic-gate return; 12897c478bd9Sstevel@tonic-gate } 12907c478bd9Sstevel@tonic-gate q = ill->ill_rq; 12917c478bd9Sstevel@tonic-gate pkt->b_prev = NULL; 12927c478bd9Sstevel@tonic-gate /* 12937c478bd9Sstevel@tonic-gate * This packet has not gone through IPSEC processing 12947c478bd9Sstevel@tonic-gate * and hence we should not have any IPSEC message 12957c478bd9Sstevel@tonic-gate * prepended. 12967c478bd9Sstevel@tonic-gate */ 12977c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 12987c478bd9Sstevel@tonic-gate put(q, pkt); 12997c478bd9Sstevel@tonic-gate ill_refrele(ill); 13007c478bd9Sstevel@tonic-gate } else if (pkt->b_next) { 13017c478bd9Sstevel@tonic-gate /* Packets from multicast router */ 13027c478bd9Sstevel@tonic-gate pkt->b_next = NULL; 13037c478bd9Sstevel@tonic-gate /* 13047c478bd9Sstevel@tonic-gate * We never get the IPSEC_OUT while forwarding the 13057c478bd9Sstevel@tonic-gate * packet for multicast router. 13067c478bd9Sstevel@tonic-gate */ 13077c478bd9Sstevel@tonic-gate ASSERT(ipsec_mp == pkt); 13087c478bd9Sstevel@tonic-gate /* 13097c478bd9Sstevel@tonic-gate * XXX TODO IPv6. 13107c478bd9Sstevel@tonic-gate */ 13117c478bd9Sstevel@tonic-gate freemsg(pkt); 13127c478bd9Sstevel@tonic-gate #ifdef XXX 13137c478bd9Sstevel@tonic-gate ip_rput_forward(ire, (ipha_t *)pkt->b_rptr, pkt, NULL); 13147c478bd9Sstevel@tonic-gate #endif 13157c478bd9Sstevel@tonic-gate } else { 13167c478bd9Sstevel@tonic-gate if (secure) { 13177c478bd9Sstevel@tonic-gate ipsec_out_t *oi; 13187c478bd9Sstevel@tonic-gate ip6_t *ip6h; 13197c478bd9Sstevel@tonic-gate 13207c478bd9Sstevel@tonic-gate oi = (ipsec_out_t *)ipsec_mp->b_rptr; 13217c478bd9Sstevel@tonic-gate ip6h = (ip6_t *)ipsec_mp->b_cont->b_rptr; 13227c478bd9Sstevel@tonic-gate if (oi->ipsec_out_proc_begin) { 13237c478bd9Sstevel@tonic-gate /* 13247c478bd9Sstevel@tonic-gate * This is the case where 13257c478bd9Sstevel@tonic-gate * ip_wput_ipsec_out could not find 13267c478bd9Sstevel@tonic-gate * the IRE and recreated a new one. 13277c478bd9Sstevel@tonic-gate */ 13287c478bd9Sstevel@tonic-gate ip_wput_ipsec_out_v6(q, ipsec_mp, ip6h, 13297c478bd9Sstevel@tonic-gate NULL, NULL); 13307c478bd9Sstevel@tonic-gate } else { 13315597b60aSnordmark if (CONN_Q(q)) { 13325597b60aSnordmark (void) ip_output_v6(Q_TO_CONN(q), 13335597b60aSnordmark ipsec_mp, q, IRE_SEND); 13345597b60aSnordmark } else { 13355597b60aSnordmark (void) ip_output_v6( 13365597b60aSnordmark (void *)(uintptr_t)zoneid, 13375597b60aSnordmark ipsec_mp, q, IRE_SEND); 13385597b60aSnordmark } 13397c478bd9Sstevel@tonic-gate } 13407c478bd9Sstevel@tonic-gate } else { 13417c478bd9Sstevel@tonic-gate /* 13427c478bd9Sstevel@tonic-gate * Send packets through ip_output_v6 so that any 13437c478bd9Sstevel@tonic-gate * ip6_info header can be processed again. 13447c478bd9Sstevel@tonic-gate */ 13455597b60aSnordmark if (CONN_Q(q)) { 13467c478bd9Sstevel@tonic-gate (void) ip_output_v6(Q_TO_CONN(q), ipsec_mp, q, 13477c478bd9Sstevel@tonic-gate IRE_SEND); 13485597b60aSnordmark } else { 13495597b60aSnordmark (void) ip_output_v6((void *)(uintptr_t)zoneid, 13505597b60aSnordmark ipsec_mp, q, IRE_SEND); 13515597b60aSnordmark } 13527c478bd9Sstevel@tonic-gate } 13537c478bd9Sstevel@tonic-gate /* 13547c478bd9Sstevel@tonic-gate * Special code to support sending a single packet with 13557c478bd9Sstevel@tonic-gate * conn_unspec_src using an IRE which has no source address. 13567c478bd9Sstevel@tonic-gate * The IRE is deleted here after sending the packet to avoid 13577c478bd9Sstevel@tonic-gate * having other code trip on it. But before we delete the 13587c478bd9Sstevel@tonic-gate * ire, somebody could have looked up this ire. 13597c478bd9Sstevel@tonic-gate * We prevent returning/using this IRE by the upper layers 13607c478bd9Sstevel@tonic-gate * by making checks to NULL source address in other places 13617c478bd9Sstevel@tonic-gate * like e.g ip_ire_append_v6, ip_ire_req and 13627c478bd9Sstevel@tonic-gate * ip_bind_connected_v6. Though, this does not completely 13637c478bd9Sstevel@tonic-gate * prevent other threads from using this ire, this should 13647c478bd9Sstevel@tonic-gate * not cause any problems. 13657c478bd9Sstevel@tonic-gate */ 13667c478bd9Sstevel@tonic-gate if (IN6_IS_ADDR_UNSPECIFIED(&ire->ire_src_addr_v6)) { 13677c478bd9Sstevel@tonic-gate ip1dbg(("ire_send_v6: delete IRE\n")); 13687c478bd9Sstevel@tonic-gate ire_delete(ire); 13697c478bd9Sstevel@tonic-gate } 13707c478bd9Sstevel@tonic-gate } 13717c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add */ 13727c478bd9Sstevel@tonic-gate } 13737c478bd9Sstevel@tonic-gate 13747c478bd9Sstevel@tonic-gate /* 13757c478bd9Sstevel@tonic-gate * Make sure that IRE bucket does not get too long. 13767c478bd9Sstevel@tonic-gate * This can cause lock up because ire_cache_lookup() 13777c478bd9Sstevel@tonic-gate * may take "forever" to finish. 13787c478bd9Sstevel@tonic-gate * 13797c478bd9Sstevel@tonic-gate * We just remove cnt IREs each time. This means that 13807c478bd9Sstevel@tonic-gate * the bucket length will stay approximately constant, 13817c478bd9Sstevel@tonic-gate * depending on cnt. This should be enough to defend 13827c478bd9Sstevel@tonic-gate * against DoS attack based on creating temporary IREs 13837c478bd9Sstevel@tonic-gate * (for forwarding and non-TCP traffic). 13847c478bd9Sstevel@tonic-gate * 13857c478bd9Sstevel@tonic-gate * Note that new IRE is normally added at the tail of the 13867c478bd9Sstevel@tonic-gate * bucket. This means that we are removing the "oldest" 13877c478bd9Sstevel@tonic-gate * temporary IRE added. Only if there are IREs with 13887c478bd9Sstevel@tonic-gate * the same ire_addr, do we not add it at the tail. Refer 13897c478bd9Sstevel@tonic-gate * to ire_add_v*(). It should be OK for our purpose. 13907c478bd9Sstevel@tonic-gate * 13917c478bd9Sstevel@tonic-gate * For non-temporary cached IREs, we make sure that they 13927c478bd9Sstevel@tonic-gate * have not been used for some time (defined below), they 13937c478bd9Sstevel@tonic-gate * are non-local destinations, and there is no one using 13947c478bd9Sstevel@tonic-gate * them at the moment (refcnt == 1). 13957c478bd9Sstevel@tonic-gate * 13967c478bd9Sstevel@tonic-gate * The above means that the IRE bucket length may become 13977c478bd9Sstevel@tonic-gate * very long, consisting of mostly non-temporary IREs. 13987c478bd9Sstevel@tonic-gate * This can happen when the hash function does a bad job 13997c478bd9Sstevel@tonic-gate * so that most TCP connections cluster to a specific bucket. 14007c478bd9Sstevel@tonic-gate * This "hopefully" should never happen. It can also 14017c478bd9Sstevel@tonic-gate * happen if most TCP connections have very long lives. 14027c478bd9Sstevel@tonic-gate * Even with the minimal hash table size of 256, there 14037c478bd9Sstevel@tonic-gate * has to be a lot of such connections to make the bucket 14047c478bd9Sstevel@tonic-gate * length unreasonably long. This should probably not 14057c478bd9Sstevel@tonic-gate * happen either. The third can when this can happen is 14067c478bd9Sstevel@tonic-gate * when the machine is under attack, such as SYN flooding. 14077c478bd9Sstevel@tonic-gate * TCP should already have the proper mechanism to protect 14087c478bd9Sstevel@tonic-gate * that. So we should be safe. 14097c478bd9Sstevel@tonic-gate * 14107c478bd9Sstevel@tonic-gate * This function is called by ire_add_then_send() after 14117c478bd9Sstevel@tonic-gate * a new IRE is added and the packet is sent. 14127c478bd9Sstevel@tonic-gate * 14137c478bd9Sstevel@tonic-gate * The idle cutoff interval is set to 60s. It can be 14147c478bd9Sstevel@tonic-gate * changed using /etc/system. 14157c478bd9Sstevel@tonic-gate */ 14167c478bd9Sstevel@tonic-gate uint32_t ire_idle_cutoff_interval = 60000; 14177c478bd9Sstevel@tonic-gate 14187c478bd9Sstevel@tonic-gate static void 14197c478bd9Sstevel@tonic-gate ire_cache_cleanup(irb_t *irb, uint32_t threshold, int cnt) 14207c478bd9Sstevel@tonic-gate { 14217c478bd9Sstevel@tonic-gate ire_t *ire; 14227c478bd9Sstevel@tonic-gate int tmp_cnt = cnt; 14237c478bd9Sstevel@tonic-gate clock_t cut_off = drv_usectohz(ire_idle_cutoff_interval * 1000); 14247c478bd9Sstevel@tonic-gate 14257c478bd9Sstevel@tonic-gate /* 14267c478bd9Sstevel@tonic-gate * irb is NULL if the IRE is not added to the hash. This 14277c478bd9Sstevel@tonic-gate * happens when IRE_MARK_NOADD is set in ire_add_then_send() 14287c478bd9Sstevel@tonic-gate * and when ires are returned from ire_update_srcif_v4() routine. 14297c478bd9Sstevel@tonic-gate */ 14307c478bd9Sstevel@tonic-gate if (irb == NULL) 14317c478bd9Sstevel@tonic-gate return; 14327c478bd9Sstevel@tonic-gate 14337c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 14347c478bd9Sstevel@tonic-gate if (irb->irb_tmp_ire_cnt > threshold) { 14357c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL && tmp_cnt > 0; 14367c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 14377c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) 14387c478bd9Sstevel@tonic-gate continue; 14397c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_TEMPORARY) { 14407c478bd9Sstevel@tonic-gate ASSERT(ire->ire_type == IRE_CACHE); 14417c478bd9Sstevel@tonic-gate ire_delete(ire); 14427c478bd9Sstevel@tonic-gate tmp_cnt--; 14437c478bd9Sstevel@tonic-gate } 14447c478bd9Sstevel@tonic-gate } 14457c478bd9Sstevel@tonic-gate } 14467c478bd9Sstevel@tonic-gate if (irb->irb_ire_cnt - irb->irb_tmp_ire_cnt > threshold) { 14477c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL && cnt > 0; 14487c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 1449*f4b3ec61Sdh155122 if (ire->ire_marks & IRE_MARK_CONDEMNED) 1450*f4b3ec61Sdh155122 continue; 1451*f4b3ec61Sdh155122 if (ire->ire_ipversion == IPV4_VERSION) { 1452*f4b3ec61Sdh155122 if (ire->ire_gateway_addr == 0) 1453*f4b3ec61Sdh155122 continue; 1454*f4b3ec61Sdh155122 } else { 1455*f4b3ec61Sdh155122 if (IN6_IS_ADDR_UNSPECIFIED( 1456*f4b3ec61Sdh155122 &ire->ire_gateway_addr_v6)) 14577c478bd9Sstevel@tonic-gate continue; 14587c478bd9Sstevel@tonic-gate } 14597c478bd9Sstevel@tonic-gate if ((ire->ire_type == IRE_CACHE) && 14607c478bd9Sstevel@tonic-gate (lbolt - ire->ire_last_used_time > cut_off) && 14617c478bd9Sstevel@tonic-gate (ire->ire_refcnt == 1)) { 14627c478bd9Sstevel@tonic-gate ire_delete(ire); 14637c478bd9Sstevel@tonic-gate cnt--; 14647c478bd9Sstevel@tonic-gate } 14657c478bd9Sstevel@tonic-gate } 14667c478bd9Sstevel@tonic-gate } 14677c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 14687c478bd9Sstevel@tonic-gate } 14697c478bd9Sstevel@tonic-gate 14707c478bd9Sstevel@tonic-gate /* 14717c478bd9Sstevel@tonic-gate * ire_add_then_send is called when a new IRE has been created in order to 14727c478bd9Sstevel@tonic-gate * route an outgoing packet. Typically, it is called from ip_wput when 14737c478bd9Sstevel@tonic-gate * a response comes back down from a resolver. We add the IRE, and then 14747c478bd9Sstevel@tonic-gate * possibly run the packet through ip_wput or ip_rput, as appropriate. 14757c478bd9Sstevel@tonic-gate * However, we do not add the newly created IRE in the cache when 14767c478bd9Sstevel@tonic-gate * IRE_MARK_NOADD is set in the IRE. IRE_MARK_NOADD is set at 14777c478bd9Sstevel@tonic-gate * ip_newroute_ipif(). The ires with IRE_MARK_NOADD and ires returned 14787c478bd9Sstevel@tonic-gate * by ire_update_srcif_v4() are ire_refrele'd by ip_wput_ire() and get 14797c478bd9Sstevel@tonic-gate * deleted. 14807c478bd9Sstevel@tonic-gate * Multirouting support: the packet is silently discarded when the new IRE 14817c478bd9Sstevel@tonic-gate * holds the RTF_MULTIRT flag, but is not the first IRE to be added with the 14827c478bd9Sstevel@tonic-gate * RTF_MULTIRT flag for the same destination address. 14837c478bd9Sstevel@tonic-gate * In this case, we just want to register this additional ire without 14847c478bd9Sstevel@tonic-gate * sending the packet, as it has already been replicated through 14857c478bd9Sstevel@tonic-gate * existing multirt routes in ip_wput(). 14867c478bd9Sstevel@tonic-gate */ 14877c478bd9Sstevel@tonic-gate void 14887c478bd9Sstevel@tonic-gate ire_add_then_send(queue_t *q, ire_t *ire, mblk_t *mp) 14897c478bd9Sstevel@tonic-gate { 14907c478bd9Sstevel@tonic-gate irb_t *irb; 14917c478bd9Sstevel@tonic-gate boolean_t drop = B_FALSE; 14927c478bd9Sstevel@tonic-gate /* LINTED : set but not used in function */ 14937c478bd9Sstevel@tonic-gate boolean_t mctl_present; 14947c478bd9Sstevel@tonic-gate mblk_t *first_mp = NULL; 14957c478bd9Sstevel@tonic-gate mblk_t *save_mp = NULL; 14967c478bd9Sstevel@tonic-gate ire_t *dst_ire; 14977c478bd9Sstevel@tonic-gate ipha_t *ipha; 14987c478bd9Sstevel@tonic-gate ip6_t *ip6h; 1499*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 15007c478bd9Sstevel@tonic-gate 15017c478bd9Sstevel@tonic-gate if (mp != NULL) { 15027c478bd9Sstevel@tonic-gate /* 15037c478bd9Sstevel@tonic-gate * We first have to retrieve the destination address carried 15047c478bd9Sstevel@tonic-gate * by the packet. 15057c478bd9Sstevel@tonic-gate * We can't rely on ire as it can be related to a gateway. 15067c478bd9Sstevel@tonic-gate * The destination address will help in determining if 15077c478bd9Sstevel@tonic-gate * other RTF_MULTIRT ires are already registered. 15087c478bd9Sstevel@tonic-gate * 15097c478bd9Sstevel@tonic-gate * We first need to know where we are going : v4 or V6. 15107c478bd9Sstevel@tonic-gate * the ire version is enough, as there is no risk that 15117c478bd9Sstevel@tonic-gate * we resolve an IPv6 address with an IPv4 ire 15127c478bd9Sstevel@tonic-gate * or vice versa. 15137c478bd9Sstevel@tonic-gate */ 15147c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 15157c478bd9Sstevel@tonic-gate EXTRACT_PKT_MP(mp, first_mp, mctl_present); 15167c478bd9Sstevel@tonic-gate ipha = (ipha_t *)mp->b_rptr; 15177c478bd9Sstevel@tonic-gate save_mp = mp; 15187c478bd9Sstevel@tonic-gate mp = first_mp; 15197c478bd9Sstevel@tonic-gate 15207c478bd9Sstevel@tonic-gate dst_ire = ire_cache_lookup(ipha->ipha_dst, 1521*f4b3ec61Sdh155122 ire->ire_zoneid, MBLK_GETLABEL(mp), ipst); 15227c478bd9Sstevel@tonic-gate } else { 1523c793af95Ssangeeta ASSERT(ire->ire_ipversion == IPV6_VERSION); 15247c478bd9Sstevel@tonic-gate /* 15257c478bd9Sstevel@tonic-gate * Get a pointer to the beginning of the IPv6 header. 15267c478bd9Sstevel@tonic-gate * Ignore leading IPsec control mblks. 15277c478bd9Sstevel@tonic-gate */ 15287c478bd9Sstevel@tonic-gate first_mp = mp; 15297c478bd9Sstevel@tonic-gate if (mp->b_datap->db_type == M_CTL) { 15307c478bd9Sstevel@tonic-gate mp = mp->b_cont; 15317c478bd9Sstevel@tonic-gate } 15327c478bd9Sstevel@tonic-gate ip6h = (ip6_t *)mp->b_rptr; 15337c478bd9Sstevel@tonic-gate save_mp = mp; 15347c478bd9Sstevel@tonic-gate mp = first_mp; 15357c478bd9Sstevel@tonic-gate dst_ire = ire_cache_lookup_v6(&ip6h->ip6_dst, 1536*f4b3ec61Sdh155122 ire->ire_zoneid, MBLK_GETLABEL(mp), ipst); 15377c478bd9Sstevel@tonic-gate } 15387c478bd9Sstevel@tonic-gate if (dst_ire != NULL) { 15397c478bd9Sstevel@tonic-gate if (dst_ire->ire_flags & RTF_MULTIRT) { 15407c478bd9Sstevel@tonic-gate /* 15417c478bd9Sstevel@tonic-gate * At least one resolved multirt route 15427c478bd9Sstevel@tonic-gate * already exists for the destination, 15437c478bd9Sstevel@tonic-gate * don't sent this packet: either drop it 15447c478bd9Sstevel@tonic-gate * or complete the pending resolution, 15457c478bd9Sstevel@tonic-gate * depending on the ire. 15467c478bd9Sstevel@tonic-gate */ 15477c478bd9Sstevel@tonic-gate drop = B_TRUE; 15487c478bd9Sstevel@tonic-gate } 15497c478bd9Sstevel@tonic-gate ip1dbg(("ire_add_then_send: dst_ire %p " 15507c478bd9Sstevel@tonic-gate "[dst %08x, gw %08x], drop %d\n", 15517c478bd9Sstevel@tonic-gate (void *)dst_ire, 15527c478bd9Sstevel@tonic-gate (dst_ire->ire_ipversion == IPV4_VERSION) ? \ 15537c478bd9Sstevel@tonic-gate ntohl(dst_ire->ire_addr) : \ 15547c478bd9Sstevel@tonic-gate ntohl(V4_PART_OF_V6(dst_ire->ire_addr_v6)), 15557c478bd9Sstevel@tonic-gate (dst_ire->ire_ipversion == IPV4_VERSION) ? \ 15567c478bd9Sstevel@tonic-gate ntohl(dst_ire->ire_gateway_addr) : \ 15577c478bd9Sstevel@tonic-gate ntohl(V4_PART_OF_V6( 15587c478bd9Sstevel@tonic-gate dst_ire->ire_gateway_addr_v6)), 15597c478bd9Sstevel@tonic-gate drop)); 15607c478bd9Sstevel@tonic-gate ire_refrele(dst_ire); 15617c478bd9Sstevel@tonic-gate } 15627c478bd9Sstevel@tonic-gate } 15637c478bd9Sstevel@tonic-gate 15647c478bd9Sstevel@tonic-gate if (!(ire->ire_marks & IRE_MARK_NOADD)) { 15657c478bd9Sstevel@tonic-gate /* 15667c478bd9Sstevel@tonic-gate * Regular packets with cache bound ires and 15677c478bd9Sstevel@tonic-gate * the packets from ARP response for ires which 15687c478bd9Sstevel@tonic-gate * belong to the ire_srcif_v4 table, are here. 15697c478bd9Sstevel@tonic-gate */ 15707c478bd9Sstevel@tonic-gate if (ire->ire_in_ill == NULL) { 15717c478bd9Sstevel@tonic-gate /* Add the ire */ 1572c793af95Ssangeeta (void) ire_add(&ire, NULL, NULL, NULL, B_FALSE); 15737c478bd9Sstevel@tonic-gate } else { 15747c478bd9Sstevel@tonic-gate /* 15757c478bd9Sstevel@tonic-gate * This must be ARP response for ire in interface based 15767c478bd9Sstevel@tonic-gate * table. Note that we don't add them in cache table, 15777c478bd9Sstevel@tonic-gate * instead we update the existing table with dlureq_mp 15787c478bd9Sstevel@tonic-gate * information. The reverse tunnel ires do not come 15797c478bd9Sstevel@tonic-gate * here, as reverse tunnel is non-resolver interface. 15807c478bd9Sstevel@tonic-gate * XXX- another design alternative was to mark the 15817c478bd9Sstevel@tonic-gate * ires in interface based table with a special mark to 15827c478bd9Sstevel@tonic-gate * make absolutely sure that we operate in right ires. 15837c478bd9Sstevel@tonic-gate * This idea was not implemented as part of code review 15847c478bd9Sstevel@tonic-gate * suggestion, as ire_in_ill suffice to distinguish 15857c478bd9Sstevel@tonic-gate * between the regular ires and interface based 15867c478bd9Sstevel@tonic-gate * ires now and thus we save a bit in the ire_marks. 15877c478bd9Sstevel@tonic-gate */ 15887c478bd9Sstevel@tonic-gate ire = ire_update_srcif_v4(ire); 15897c478bd9Sstevel@tonic-gate } 15907c478bd9Sstevel@tonic-gate 15917c478bd9Sstevel@tonic-gate if (ire == NULL) { 15927c478bd9Sstevel@tonic-gate mp->b_prev = NULL; 15937c478bd9Sstevel@tonic-gate mp->b_next = NULL; 15947c478bd9Sstevel@tonic-gate MULTIRT_DEBUG_UNTAG(mp); 15957c478bd9Sstevel@tonic-gate freemsg(mp); 15967c478bd9Sstevel@tonic-gate return; 15977c478bd9Sstevel@tonic-gate } 15987c478bd9Sstevel@tonic-gate if (mp == NULL) { 15997c478bd9Sstevel@tonic-gate ire_refrele(ire); /* Held in ire_add_v4/v6 */ 16007c478bd9Sstevel@tonic-gate return; 16017c478bd9Sstevel@tonic-gate } 16027c478bd9Sstevel@tonic-gate } 16037c478bd9Sstevel@tonic-gate if (drop) { 16047c478bd9Sstevel@tonic-gate /* 16057c478bd9Sstevel@tonic-gate * If we're adding an RTF_MULTIRT ire, the resolution 16067c478bd9Sstevel@tonic-gate * is over: we just drop the packet. 16077c478bd9Sstevel@tonic-gate */ 16087c478bd9Sstevel@tonic-gate if (ire->ire_flags & RTF_MULTIRT) { 16097c478bd9Sstevel@tonic-gate if (save_mp) { 16107c478bd9Sstevel@tonic-gate save_mp->b_prev = NULL; 16117c478bd9Sstevel@tonic-gate save_mp->b_next = NULL; 16127c478bd9Sstevel@tonic-gate } 16137c478bd9Sstevel@tonic-gate MULTIRT_DEBUG_UNTAG(mp); 16147c478bd9Sstevel@tonic-gate freemsg(mp); 16157c478bd9Sstevel@tonic-gate } else { 16167c478bd9Sstevel@tonic-gate /* 16177c478bd9Sstevel@tonic-gate * Otherwise, we're adding the ire to a gateway 16187c478bd9Sstevel@tonic-gate * for a multirt route. 16197c478bd9Sstevel@tonic-gate * Invoke ip_newroute() to complete the resolution 16207c478bd9Sstevel@tonic-gate * of the route. We will then come back here and 16217c478bd9Sstevel@tonic-gate * finally drop this packet in the above code. 16227c478bd9Sstevel@tonic-gate */ 16237c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 16247c478bd9Sstevel@tonic-gate /* 16257c478bd9Sstevel@tonic-gate * TODO: in order for CGTP to work in non-global 16267c478bd9Sstevel@tonic-gate * zones, ip_newroute() must create the IRE 16277c478bd9Sstevel@tonic-gate * cache in the zone indicated by 16287c478bd9Sstevel@tonic-gate * ire->ire_zoneid. 16297c478bd9Sstevel@tonic-gate */ 16307c478bd9Sstevel@tonic-gate ip_newroute(q, mp, ipha->ipha_dst, 0, 16315597b60aSnordmark (CONN_Q(q) ? Q_TO_CONN(q) : NULL), 1632*f4b3ec61Sdh155122 ire->ire_zoneid, ipst); 16337c478bd9Sstevel@tonic-gate } else { 1634c793af95Ssangeeta ASSERT(ire->ire_ipversion == IPV6_VERSION); 16357c478bd9Sstevel@tonic-gate ip_newroute_v6(q, mp, &ip6h->ip6_dst, NULL, 1636*f4b3ec61Sdh155122 NULL, ire->ire_zoneid, ipst); 16377c478bd9Sstevel@tonic-gate } 16387c478bd9Sstevel@tonic-gate } 16397c478bd9Sstevel@tonic-gate 16407c478bd9Sstevel@tonic-gate ire_refrele(ire); /* As done by ire_send(). */ 16417c478bd9Sstevel@tonic-gate return; 16427c478bd9Sstevel@tonic-gate } 16437c478bd9Sstevel@tonic-gate /* 16447c478bd9Sstevel@tonic-gate * Need to remember ire_bucket here as ire_send*() may delete 16457c478bd9Sstevel@tonic-gate * the ire so we cannot reference it after that. 16467c478bd9Sstevel@tonic-gate */ 16477c478bd9Sstevel@tonic-gate irb = ire->ire_bucket; 16487c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 16497c478bd9Sstevel@tonic-gate ire_send_v6(q, mp, ire); 16507c478bd9Sstevel@tonic-gate /* 16517c478bd9Sstevel@tonic-gate * Clean up more than 1 IRE so that the clean up does not 16527c478bd9Sstevel@tonic-gate * need to be done every time when a new IRE is added and 16537c478bd9Sstevel@tonic-gate * the threshold is reached. 16547c478bd9Sstevel@tonic-gate */ 16557c478bd9Sstevel@tonic-gate ire_cache_cleanup(irb, ip6_ire_max_bucket_cnt, 2); 16567c478bd9Sstevel@tonic-gate } else { 16577c478bd9Sstevel@tonic-gate ire_send(q, mp, ire); 16587c478bd9Sstevel@tonic-gate ire_cache_cleanup(irb, ip_ire_max_bucket_cnt, 2); 16597c478bd9Sstevel@tonic-gate } 16607c478bd9Sstevel@tonic-gate } 16617c478bd9Sstevel@tonic-gate 16627c478bd9Sstevel@tonic-gate /* 16637c478bd9Sstevel@tonic-gate * Initialize the ire that is specific to IPv4 part and call 16647c478bd9Sstevel@tonic-gate * ire_init_common to finish it. 16657c478bd9Sstevel@tonic-gate */ 16667c478bd9Sstevel@tonic-gate ire_t * 16677c478bd9Sstevel@tonic-gate ire_init(ire_t *ire, uchar_t *addr, uchar_t *mask, uchar_t *src_addr, 16687c478bd9Sstevel@tonic-gate uchar_t *gateway, uchar_t *in_src_addr, uint_t *max_fragp, mblk_t *fp_mp, 16697c478bd9Sstevel@tonic-gate queue_t *rfq, queue_t *stq, ushort_t type, mblk_t *dlureq_mp, ipif_t *ipif, 16707c478bd9Sstevel@tonic-gate ill_t *in_ill, ipaddr_t cmask, uint32_t phandle, uint32_t ihandle, 1671*f4b3ec61Sdh155122 uint32_t flags, const iulp_t *ulp_info, tsol_gc_t *gc, tsol_gcgrp_t *gcgrp, 1672*f4b3ec61Sdh155122 ip_stack_t *ipst) 16737c478bd9Sstevel@tonic-gate { 167445916cd2Sjpk /* 167545916cd2Sjpk * Reject IRE security attribute creation/initialization 167645916cd2Sjpk * if system is not running in Trusted mode. 167745916cd2Sjpk */ 167845916cd2Sjpk if ((gc != NULL || gcgrp != NULL) && !is_system_labeled()) 167945916cd2Sjpk return (NULL); 168045916cd2Sjpk 16817c478bd9Sstevel@tonic-gate if (fp_mp != NULL) { 16827c478bd9Sstevel@tonic-gate /* 16837c478bd9Sstevel@tonic-gate * We can't dupb() here as multiple threads could be 16847c478bd9Sstevel@tonic-gate * calling dupb on the same mp which is incorrect. 16857c478bd9Sstevel@tonic-gate * First dupb() should be called only by one thread. 16867c478bd9Sstevel@tonic-gate */ 16877c478bd9Sstevel@tonic-gate fp_mp = copyb(fp_mp); 16887c478bd9Sstevel@tonic-gate if (fp_mp == NULL) 16897c478bd9Sstevel@tonic-gate return (NULL); 16907c478bd9Sstevel@tonic-gate } 16917c478bd9Sstevel@tonic-gate 16927c478bd9Sstevel@tonic-gate if (dlureq_mp != NULL) { 16937c478bd9Sstevel@tonic-gate /* 16947c478bd9Sstevel@tonic-gate * We can't dupb() here as multiple threads could be 16957c478bd9Sstevel@tonic-gate * calling dupb on the same mp which is incorrect. 16967c478bd9Sstevel@tonic-gate * First dupb() should be called only by one thread. 16977c478bd9Sstevel@tonic-gate */ 16987c478bd9Sstevel@tonic-gate dlureq_mp = copyb(dlureq_mp); 16997c478bd9Sstevel@tonic-gate if (dlureq_mp == NULL) { 17007c478bd9Sstevel@tonic-gate if (fp_mp != NULL) 17017c478bd9Sstevel@tonic-gate freeb(fp_mp); 17027c478bd9Sstevel@tonic-gate return (NULL); 17037c478bd9Sstevel@tonic-gate } 17047c478bd9Sstevel@tonic-gate } 17057c478bd9Sstevel@tonic-gate 17067c478bd9Sstevel@tonic-gate /* 17077c478bd9Sstevel@tonic-gate * Check that IRE_IF_RESOLVER and IRE_IF_NORESOLVER have a 17087c478bd9Sstevel@tonic-gate * dlureq_mp which is the ill_resolver_mp for IRE_IF_RESOLVER 17097c478bd9Sstevel@tonic-gate * and DL_UNITDATA_REQ for IRE_IF_NORESOLVER. 17107c478bd9Sstevel@tonic-gate */ 17117c478bd9Sstevel@tonic-gate if ((type & IRE_INTERFACE) && 17127c478bd9Sstevel@tonic-gate dlureq_mp == NULL) { 17137c478bd9Sstevel@tonic-gate ASSERT(fp_mp == NULL); 17147c478bd9Sstevel@tonic-gate ip0dbg(("ire_init: no dlureq_mp\n")); 17157c478bd9Sstevel@tonic-gate return (NULL); 17167c478bd9Sstevel@tonic-gate } 17177c478bd9Sstevel@tonic-gate 1718*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_alloced); 17197c478bd9Sstevel@tonic-gate 17207c478bd9Sstevel@tonic-gate if (addr != NULL) 17217c478bd9Sstevel@tonic-gate bcopy(addr, &ire->ire_addr, IP_ADDR_LEN); 17227c478bd9Sstevel@tonic-gate if (src_addr != NULL) 17237c478bd9Sstevel@tonic-gate bcopy(src_addr, &ire->ire_src_addr, IP_ADDR_LEN); 17247c478bd9Sstevel@tonic-gate if (mask != NULL) { 17257c478bd9Sstevel@tonic-gate bcopy(mask, &ire->ire_mask, IP_ADDR_LEN); 17267c478bd9Sstevel@tonic-gate ire->ire_masklen = ip_mask_to_plen(ire->ire_mask); 17277c478bd9Sstevel@tonic-gate } 17287c478bd9Sstevel@tonic-gate if (gateway != NULL) { 17297c478bd9Sstevel@tonic-gate bcopy(gateway, &ire->ire_gateway_addr, IP_ADDR_LEN); 17307c478bd9Sstevel@tonic-gate } 17317c478bd9Sstevel@tonic-gate if (in_src_addr != NULL) { 17327c478bd9Sstevel@tonic-gate bcopy(in_src_addr, &ire->ire_in_src_addr, IP_ADDR_LEN); 17337c478bd9Sstevel@tonic-gate } 17347c478bd9Sstevel@tonic-gate 17357c478bd9Sstevel@tonic-gate if (type == IRE_CACHE) 17367c478bd9Sstevel@tonic-gate ire->ire_cmask = cmask; 17377c478bd9Sstevel@tonic-gate 173845916cd2Sjpk /* ire_init_common will free the mblks upon encountering any failure */ 173945916cd2Sjpk if (!ire_init_common(ire, max_fragp, fp_mp, rfq, stq, type, dlureq_mp, 174045916cd2Sjpk ipif, in_ill, phandle, ihandle, flags, IPV4_VERSION, ulp_info, 1741*f4b3ec61Sdh155122 gc, gcgrp, ipst)) 174245916cd2Sjpk return (NULL); 17437c478bd9Sstevel@tonic-gate 17447c478bd9Sstevel@tonic-gate return (ire); 17457c478bd9Sstevel@tonic-gate } 17467c478bd9Sstevel@tonic-gate 17477c478bd9Sstevel@tonic-gate /* 17487c478bd9Sstevel@tonic-gate * Similar to ire_create except that it is called only when 17497c478bd9Sstevel@tonic-gate * we want to allocate ire as an mblk e.g. we have an external 17507c478bd9Sstevel@tonic-gate * resolver ARP. 17517c478bd9Sstevel@tonic-gate */ 17527c478bd9Sstevel@tonic-gate ire_t * 17537c478bd9Sstevel@tonic-gate ire_create_mp(uchar_t *addr, uchar_t *mask, uchar_t *src_addr, uchar_t *gateway, 17547c478bd9Sstevel@tonic-gate uchar_t *in_src_addr, uint_t max_frag, mblk_t *fp_mp, queue_t *rfq, 17557c478bd9Sstevel@tonic-gate queue_t *stq, ushort_t type, mblk_t *dlureq_mp, ipif_t *ipif, ill_t *in_ill, 17567c478bd9Sstevel@tonic-gate ipaddr_t cmask, uint32_t phandle, uint32_t ihandle, uint32_t flags, 1757*f4b3ec61Sdh155122 const iulp_t *ulp_info, tsol_gc_t *gc, tsol_gcgrp_t *gcgrp, 1758*f4b3ec61Sdh155122 ip_stack_t *ipst) 17597c478bd9Sstevel@tonic-gate { 1760c793af95Ssangeeta ire_t *ire, *buf; 17617c478bd9Sstevel@tonic-gate ire_t *ret_ire; 17627c478bd9Sstevel@tonic-gate mblk_t *mp; 1763c793af95Ssangeeta size_t bufsize; 1764c793af95Ssangeeta frtn_t *frtnp; 1765c793af95Ssangeeta ill_t *ill; 17667c478bd9Sstevel@tonic-gate 1767c793af95Ssangeeta bufsize = sizeof (ire_t) + sizeof (frtn_t); 1768c793af95Ssangeeta buf = kmem_alloc(bufsize, KM_NOSLEEP); 1769c793af95Ssangeeta if (buf == NULL) { 17707c478bd9Sstevel@tonic-gate ip1dbg(("ire_create_mp: alloc failed\n")); 17717c478bd9Sstevel@tonic-gate return (NULL); 17727c478bd9Sstevel@tonic-gate } 1773c793af95Ssangeeta frtnp = (frtn_t *)(buf + 1); 1774c793af95Ssangeeta frtnp->free_arg = (caddr_t)buf; 1775c793af95Ssangeeta frtnp->free_func = ire_freemblk; 17767c478bd9Sstevel@tonic-gate 1777c793af95Ssangeeta /* 1778c793af95Ssangeeta * Allocate the new IRE. The ire created will hold a ref on 1779c793af95Ssangeeta * an nce_t after ire_nce_init, and this ref must either be 1780c793af95Ssangeeta * (a) transferred to the ire_cache entry created when ire_add_v4 1781c793af95Ssangeeta * is called after successful arp resolution, or, 1782c793af95Ssangeeta * (b) released, when arp resolution fails 1783c793af95Ssangeeta * Case (b) is handled in ire_freemblk() which will be called 1784c793af95Ssangeeta * when mp is freed as a result of failed arp. 1785c793af95Ssangeeta */ 1786c793af95Ssangeeta mp = esballoc((unsigned char *)buf, bufsize, BPRI_MED, frtnp); 1787c793af95Ssangeeta if (mp == NULL) { 1788c793af95Ssangeeta ip1dbg(("ire_create_mp: alloc failed\n")); 1789c793af95Ssangeeta kmem_free(buf, bufsize); 1790c793af95Ssangeeta return (NULL); 1791c793af95Ssangeeta } 17927c478bd9Sstevel@tonic-gate ire = (ire_t *)mp->b_rptr; 17937c478bd9Sstevel@tonic-gate mp->b_wptr = (uchar_t *)&ire[1]; 17947c478bd9Sstevel@tonic-gate 17957c478bd9Sstevel@tonic-gate /* Start clean. */ 17967c478bd9Sstevel@tonic-gate *ire = ire_null; 17977c478bd9Sstevel@tonic-gate ire->ire_mp = mp; 17987c478bd9Sstevel@tonic-gate mp->b_datap->db_type = IRE_DB_TYPE; 1799c793af95Ssangeeta ire->ire_marks |= IRE_MARK_UNCACHED; 18007c478bd9Sstevel@tonic-gate 18017c478bd9Sstevel@tonic-gate ret_ire = ire_init(ire, addr, mask, src_addr, gateway, in_src_addr, 18027c478bd9Sstevel@tonic-gate NULL, fp_mp, rfq, stq, type, dlureq_mp, ipif, in_ill, cmask, 1803*f4b3ec61Sdh155122 phandle, ihandle, flags, ulp_info, gc, gcgrp, ipst); 18047c478bd9Sstevel@tonic-gate 180553bec8caSsowmini ill = (ill_t *)(stq->q_ptr); 18067c478bd9Sstevel@tonic-gate if (ret_ire == NULL) { 1807*f4b3ec61Sdh155122 /* ire_freemblk needs these set */ 180853bec8caSsowmini ire->ire_stq_ifindex = ill->ill_phyint->phyint_ifindex; 1809*f4b3ec61Sdh155122 ire->ire_ipst = ipst; 18107c478bd9Sstevel@tonic-gate freeb(ire->ire_mp); 18117c478bd9Sstevel@tonic-gate return (NULL); 18127c478bd9Sstevel@tonic-gate } 1813c793af95Ssangeeta ret_ire->ire_stq_ifindex = ill->ill_phyint->phyint_ifindex; 18147c478bd9Sstevel@tonic-gate ASSERT(ret_ire == ire); 18157c478bd9Sstevel@tonic-gate /* 18167c478bd9Sstevel@tonic-gate * ire_max_frag is normally zero here and is atomically set 18177c478bd9Sstevel@tonic-gate * under the irebucket lock in ire_add_v[46] except for the 18187c478bd9Sstevel@tonic-gate * case of IRE_MARK_NOADD. In that event the the ire_max_frag 18197c478bd9Sstevel@tonic-gate * is non-zero here. 18207c478bd9Sstevel@tonic-gate */ 18217c478bd9Sstevel@tonic-gate ire->ire_max_frag = max_frag; 18227c478bd9Sstevel@tonic-gate return (ire); 18237c478bd9Sstevel@tonic-gate } 18247c478bd9Sstevel@tonic-gate 18257c478bd9Sstevel@tonic-gate /* 18267c478bd9Sstevel@tonic-gate * ire_create is called to allocate and initialize a new IRE. 18277c478bd9Sstevel@tonic-gate * 18287c478bd9Sstevel@tonic-gate * NOTE : This is called as writer sometimes though not required 18297c478bd9Sstevel@tonic-gate * by this function. 18307c478bd9Sstevel@tonic-gate */ 18317c478bd9Sstevel@tonic-gate ire_t * 18327c478bd9Sstevel@tonic-gate ire_create(uchar_t *addr, uchar_t *mask, uchar_t *src_addr, uchar_t *gateway, 18337c478bd9Sstevel@tonic-gate uchar_t *in_src_addr, uint_t *max_fragp, mblk_t *fp_mp, queue_t *rfq, 18347c478bd9Sstevel@tonic-gate queue_t *stq, ushort_t type, mblk_t *dlureq_mp, ipif_t *ipif, ill_t *in_ill, 18357c478bd9Sstevel@tonic-gate ipaddr_t cmask, uint32_t phandle, uint32_t ihandle, uint32_t flags, 1836*f4b3ec61Sdh155122 const iulp_t *ulp_info, tsol_gc_t *gc, tsol_gcgrp_t *gcgrp, 1837*f4b3ec61Sdh155122 ip_stack_t *ipst) 18387c478bd9Sstevel@tonic-gate { 18397c478bd9Sstevel@tonic-gate ire_t *ire; 18407c478bd9Sstevel@tonic-gate ire_t *ret_ire; 18417c478bd9Sstevel@tonic-gate 18427c478bd9Sstevel@tonic-gate ire = kmem_cache_alloc(ire_cache, KM_NOSLEEP); 18437c478bd9Sstevel@tonic-gate if (ire == NULL) { 18447c478bd9Sstevel@tonic-gate ip1dbg(("ire_create: alloc failed\n")); 18457c478bd9Sstevel@tonic-gate return (NULL); 18467c478bd9Sstevel@tonic-gate } 18477c478bd9Sstevel@tonic-gate *ire = ire_null; 18487c478bd9Sstevel@tonic-gate 18497c478bd9Sstevel@tonic-gate ret_ire = ire_init(ire, addr, mask, src_addr, gateway, in_src_addr, 18507c478bd9Sstevel@tonic-gate max_fragp, fp_mp, rfq, stq, type, dlureq_mp, ipif, in_ill, cmask, 1851*f4b3ec61Sdh155122 phandle, ihandle, flags, ulp_info, gc, gcgrp, ipst); 18527c478bd9Sstevel@tonic-gate 18537c478bd9Sstevel@tonic-gate if (ret_ire == NULL) { 18547c478bd9Sstevel@tonic-gate kmem_cache_free(ire_cache, ire); 18557c478bd9Sstevel@tonic-gate return (NULL); 18567c478bd9Sstevel@tonic-gate } 18577c478bd9Sstevel@tonic-gate ASSERT(ret_ire == ire); 18587c478bd9Sstevel@tonic-gate return (ire); 18597c478bd9Sstevel@tonic-gate } 18607c478bd9Sstevel@tonic-gate 18617c478bd9Sstevel@tonic-gate 18627c478bd9Sstevel@tonic-gate /* 18637c478bd9Sstevel@tonic-gate * Common to IPv4 and IPv6 18647c478bd9Sstevel@tonic-gate */ 186545916cd2Sjpk boolean_t 18667c478bd9Sstevel@tonic-gate ire_init_common(ire_t *ire, uint_t *max_fragp, mblk_t *fp_mp, 18677c478bd9Sstevel@tonic-gate queue_t *rfq, queue_t *stq, ushort_t type, 18687c478bd9Sstevel@tonic-gate mblk_t *dlureq_mp, ipif_t *ipif, ill_t *in_ill, uint32_t phandle, 18697c478bd9Sstevel@tonic-gate uint32_t ihandle, uint32_t flags, uchar_t ipversion, 1870*f4b3ec61Sdh155122 const iulp_t *ulp_info, tsol_gc_t *gc, tsol_gcgrp_t *gcgrp, 1871*f4b3ec61Sdh155122 ip_stack_t *ipst) 18727c478bd9Sstevel@tonic-gate { 18737c478bd9Sstevel@tonic-gate ire->ire_max_fragp = max_fragp; 1874*f4b3ec61Sdh155122 ire->ire_frag_flag |= (ipst->ips_ip_path_mtu_discovery) ? IPH_DF : 0; 18757c478bd9Sstevel@tonic-gate 18767c478bd9Sstevel@tonic-gate ASSERT(fp_mp == NULL || fp_mp->b_datap->db_type == M_DATA); 187745916cd2Sjpk #ifdef DEBUG 187845916cd2Sjpk if (ipif != NULL) { 18797c478bd9Sstevel@tonic-gate if (ipif->ipif_isv6) 18807c478bd9Sstevel@tonic-gate ASSERT(ipversion == IPV6_VERSION); 18817c478bd9Sstevel@tonic-gate else 18827c478bd9Sstevel@tonic-gate ASSERT(ipversion == IPV4_VERSION); 18837c478bd9Sstevel@tonic-gate } 188445916cd2Sjpk #endif /* DEBUG */ 188545916cd2Sjpk 188645916cd2Sjpk /* 188745916cd2Sjpk * Create/initialize IRE security attribute only in Trusted mode; 188845916cd2Sjpk * if the passed in gc/gcgrp is non-NULL, we expect that the caller 188945916cd2Sjpk * has held a reference to it and will release it when this routine 189045916cd2Sjpk * returns a failure, otherwise we own the reference. We do this 189145916cd2Sjpk * prior to initializing the rest IRE fields. 1892dc3879f9Sjarrett * 1893dc3879f9Sjarrett * Don't allocate ire_gw_secattr for the resolver case to prevent 1894dc3879f9Sjarrett * memory leak (in case of external resolution failure). We'll 1895dc3879f9Sjarrett * allocate it after a successful external resolution, in ire_add(). 1896dc3879f9Sjarrett * Note that ire->ire_mp != NULL here means this ire is headed 1897dc3879f9Sjarrett * to an external resolver. 189845916cd2Sjpk */ 189945916cd2Sjpk if (is_system_labeled()) { 190045916cd2Sjpk if ((type & (IRE_LOCAL | IRE_LOOPBACK | IRE_BROADCAST | 190145916cd2Sjpk IRE_INTERFACE)) != 0) { 190245916cd2Sjpk /* release references on behalf of caller */ 190345916cd2Sjpk if (gc != NULL) 190445916cd2Sjpk GC_REFRELE(gc); 190545916cd2Sjpk if (gcgrp != NULL) 190645916cd2Sjpk GCGRP_REFRELE(gcgrp); 1907dc3879f9Sjarrett } else if ((ire->ire_mp == NULL) && 1908dc3879f9Sjarrett tsol_ire_init_gwattr(ire, ipversion, gc, gcgrp) != 0) { 190945916cd2Sjpk /* free any caller-allocated mblks upon failure */ 191045916cd2Sjpk if (fp_mp != NULL) 191145916cd2Sjpk freeb(fp_mp); 191245916cd2Sjpk if (dlureq_mp != NULL) 191345916cd2Sjpk freeb(dlureq_mp); 191445916cd2Sjpk return (B_FALSE); 191545916cd2Sjpk } 191645916cd2Sjpk } 19177c478bd9Sstevel@tonic-gate 19187c478bd9Sstevel@tonic-gate ire->ire_stq = stq; 19197c478bd9Sstevel@tonic-gate ire->ire_rfq = rfq; 19207c478bd9Sstevel@tonic-gate ire->ire_type = type; 19217c478bd9Sstevel@tonic-gate ire->ire_flags = RTF_UP | flags; 19227c478bd9Sstevel@tonic-gate ire->ire_ident = TICK_TO_MSEC(lbolt); 19237c478bd9Sstevel@tonic-gate bcopy(ulp_info, &ire->ire_uinfo, sizeof (iulp_t)); 19247c478bd9Sstevel@tonic-gate 19257c478bd9Sstevel@tonic-gate ire->ire_tire_mark = ire->ire_ob_pkt_count + ire->ire_ib_pkt_count; 19267c478bd9Sstevel@tonic-gate ire->ire_last_used_time = lbolt; 19277c478bd9Sstevel@tonic-gate ire->ire_create_time = (uint32_t)gethrestime_sec(); 19287c478bd9Sstevel@tonic-gate 19297c478bd9Sstevel@tonic-gate /* 19307c478bd9Sstevel@tonic-gate * If this IRE is an IRE_CACHE, inherit the handles from the 19317c478bd9Sstevel@tonic-gate * parent IREs. For others in the forwarding table, assign appropriate 19327c478bd9Sstevel@tonic-gate * new ones. 19337c478bd9Sstevel@tonic-gate * 19347c478bd9Sstevel@tonic-gate * The mutex protecting ire_handle is because ire_create is not always 19357c478bd9Sstevel@tonic-gate * called as a writer. 19367c478bd9Sstevel@tonic-gate */ 19377c478bd9Sstevel@tonic-gate if (ire->ire_type & IRE_OFFSUBNET) { 1938*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_handle_lock); 1939*f4b3ec61Sdh155122 ire->ire_phandle = (uint32_t)ipst->ips_ire_handle++; 1940*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_handle_lock); 19417c478bd9Sstevel@tonic-gate } else if (ire->ire_type & IRE_INTERFACE) { 1942*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_handle_lock); 1943*f4b3ec61Sdh155122 ire->ire_ihandle = (uint32_t)ipst->ips_ire_handle++; 1944*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_handle_lock); 19457c478bd9Sstevel@tonic-gate } else if (ire->ire_type == IRE_CACHE) { 19467c478bd9Sstevel@tonic-gate ire->ire_phandle = phandle; 19477c478bd9Sstevel@tonic-gate ire->ire_ihandle = ihandle; 19487c478bd9Sstevel@tonic-gate } 19497c478bd9Sstevel@tonic-gate ire->ire_in_ill = in_ill; 19507c478bd9Sstevel@tonic-gate ire->ire_ipif = ipif; 19517c478bd9Sstevel@tonic-gate if (ipif != NULL) { 19527c478bd9Sstevel@tonic-gate ire->ire_ipif_seqid = ipif->ipif_seqid; 19537c478bd9Sstevel@tonic-gate ire->ire_zoneid = ipif->ipif_zoneid; 19547c478bd9Sstevel@tonic-gate } else { 19557c478bd9Sstevel@tonic-gate ire->ire_zoneid = GLOBAL_ZONEID; 19567c478bd9Sstevel@tonic-gate } 19577c478bd9Sstevel@tonic-gate ire->ire_ipversion = ipversion; 19587c478bd9Sstevel@tonic-gate mutex_init(&ire->ire_lock, NULL, MUTEX_DEFAULT, NULL); 1959c793af95Ssangeeta if (ipversion == IPV4_VERSION) { 1960c793af95Ssangeeta if (ire_nce_init(ire, fp_mp, dlureq_mp) != 0) { 1961c793af95Ssangeeta /* some failure occurred. propagate error back */ 1962c793af95Ssangeeta return (B_FALSE); 1963c793af95Ssangeeta } 1964c793af95Ssangeeta } else { 1965c793af95Ssangeeta ASSERT(ipversion == IPV6_VERSION); 1966c793af95Ssangeeta /* 1967c793af95Ssangeeta * IPv6 initializes the ire_nce in ire_add_v6, 1968c793af95Ssangeeta * which expects to find the ire_nce to be null when 1969c793af95Ssangeeta * when it is called. 1970c793af95Ssangeeta */ 1971c793af95Ssangeeta if (dlureq_mp) 1972c793af95Ssangeeta freemsg(dlureq_mp); 1973c793af95Ssangeeta if (fp_mp) 1974c793af95Ssangeeta freemsg(fp_mp); 1975c793af95Ssangeeta } 1976c793af95Ssangeeta ire->ire_refcnt = 1; 1977*f4b3ec61Sdh155122 ire->ire_ipst = ipst; /* No netstack_hold */ 19787c478bd9Sstevel@tonic-gate 19797c478bd9Sstevel@tonic-gate #ifdef IRE_DEBUG 19807c478bd9Sstevel@tonic-gate bzero(ire->ire_trace, sizeof (th_trace_t *) * IP_TR_HASH_MAX); 19817c478bd9Sstevel@tonic-gate #endif 198245916cd2Sjpk 198345916cd2Sjpk return (B_TRUE); 19847c478bd9Sstevel@tonic-gate } 19857c478bd9Sstevel@tonic-gate 19867c478bd9Sstevel@tonic-gate /* 19877c478bd9Sstevel@tonic-gate * This routine is called repeatedly by ipif_up to create broadcast IREs. 19887c478bd9Sstevel@tonic-gate * It is passed a pointer to a slot in an IRE pointer array into which to 19897c478bd9Sstevel@tonic-gate * place the pointer to the new IRE, if indeed we create one. If the 19907c478bd9Sstevel@tonic-gate * IRE corresponding to the address passed in would be a duplicate of an 19917c478bd9Sstevel@tonic-gate * existing one, we don't create the new one. irep is incremented before 19927c478bd9Sstevel@tonic-gate * return only if we do create a new IRE. (Always called as writer.) 19937c478bd9Sstevel@tonic-gate * 19947c478bd9Sstevel@tonic-gate * Note that with the "match_flags" parameter, we can match on either 19957c478bd9Sstevel@tonic-gate * a particular logical interface (MATCH_IRE_IPIF) or for all logical 19967c478bd9Sstevel@tonic-gate * interfaces for a given physical interface (MATCH_IRE_ILL). Currently, 19977c478bd9Sstevel@tonic-gate * we only create broadcast ire's on a per physical interface basis. If 19987c478bd9Sstevel@tonic-gate * someone is going to be mucking with logical interfaces, it is important 19997c478bd9Sstevel@tonic-gate * to call "ipif_check_bcast_ires()" to make sure that any change to a 20007c478bd9Sstevel@tonic-gate * logical interface will not cause critical broadcast IRE's to be deleted. 20017c478bd9Sstevel@tonic-gate */ 20027c478bd9Sstevel@tonic-gate ire_t ** 20037c478bd9Sstevel@tonic-gate ire_check_and_create_bcast(ipif_t *ipif, ipaddr_t addr, ire_t **irep, 20047c478bd9Sstevel@tonic-gate int match_flags) 20057c478bd9Sstevel@tonic-gate { 20067c478bd9Sstevel@tonic-gate ire_t *ire; 20077c478bd9Sstevel@tonic-gate uint64_t check_flags = IPIF_DEPRECATED | IPIF_NOLOCAL | IPIF_ANYCAST; 2008*f4b3ec61Sdh155122 ip_stack_t *ipst = ipif->ipif_ill->ill_ipst; 20097c478bd9Sstevel@tonic-gate 20107c478bd9Sstevel@tonic-gate /* 20117c478bd9Sstevel@tonic-gate * No broadcast IREs for the LOOPBACK interface 20127c478bd9Sstevel@tonic-gate * or others such as point to point and IPIF_NOXMIT. 20137c478bd9Sstevel@tonic-gate */ 20147c478bd9Sstevel@tonic-gate if (!(ipif->ipif_flags & IPIF_BROADCAST) || 20157c478bd9Sstevel@tonic-gate (ipif->ipif_flags & IPIF_NOXMIT)) 20167c478bd9Sstevel@tonic-gate return (irep); 20177c478bd9Sstevel@tonic-gate 20187c478bd9Sstevel@tonic-gate /* If this would be a duplicate, don't bother. */ 20197c478bd9Sstevel@tonic-gate if ((ire = ire_ctable_lookup(addr, 0, IRE_BROADCAST, ipif, 2020*f4b3ec61Sdh155122 ipif->ipif_zoneid, NULL, match_flags, ipst)) != NULL) { 20217c478bd9Sstevel@tonic-gate /* 20227c478bd9Sstevel@tonic-gate * We look for non-deprecated (and non-anycast, non-nolocal) 20237c478bd9Sstevel@tonic-gate * ipifs as the best choice. ipifs with check_flags matching 20247c478bd9Sstevel@tonic-gate * (deprecated, etc) are used only if non-deprecated ipifs 20257c478bd9Sstevel@tonic-gate * are not available. if the existing ire's ipif is deprecated 20267c478bd9Sstevel@tonic-gate * and the new ipif is non-deprecated, switch to the new ipif 20277c478bd9Sstevel@tonic-gate */ 20287c478bd9Sstevel@tonic-gate if ((!(ire->ire_ipif->ipif_flags & check_flags)) || 20297c478bd9Sstevel@tonic-gate (ipif->ipif_flags & check_flags)) { 20307c478bd9Sstevel@tonic-gate ire_refrele(ire); 20317c478bd9Sstevel@tonic-gate return (irep); 20327c478bd9Sstevel@tonic-gate } 20337c478bd9Sstevel@tonic-gate /* 20347c478bd9Sstevel@tonic-gate * Bcast ires exist in pairs. Both have to be deleted, 20357c478bd9Sstevel@tonic-gate * Since we are exclusive we can make the above assertion. 20367c478bd9Sstevel@tonic-gate * The 1st has to be refrele'd since it was ctable_lookup'd. 20377c478bd9Sstevel@tonic-gate */ 20387c478bd9Sstevel@tonic-gate ASSERT(IAM_WRITER_IPIF(ipif)); 20397c478bd9Sstevel@tonic-gate ASSERT(ire->ire_next->ire_addr == ire->ire_addr); 20407c478bd9Sstevel@tonic-gate ire_delete(ire->ire_next); 20417c478bd9Sstevel@tonic-gate ire_delete(ire); 20427c478bd9Sstevel@tonic-gate ire_refrele(ire); 20437c478bd9Sstevel@tonic-gate } 20447c478bd9Sstevel@tonic-gate 20457c478bd9Sstevel@tonic-gate irep = ire_create_bcast(ipif, addr, irep); 20467c478bd9Sstevel@tonic-gate 20477c478bd9Sstevel@tonic-gate return (irep); 20487c478bd9Sstevel@tonic-gate } 20497c478bd9Sstevel@tonic-gate 20507c478bd9Sstevel@tonic-gate uint_t ip_loopback_mtu = IP_LOOPBACK_MTU; 20517c478bd9Sstevel@tonic-gate 20527c478bd9Sstevel@tonic-gate /* 20537c478bd9Sstevel@tonic-gate * This routine is called from ipif_check_bcast_ires and ire_check_bcast. 20547c478bd9Sstevel@tonic-gate * It leaves all the verifying and deleting to those routines. So it always 20557c478bd9Sstevel@tonic-gate * creates 2 bcast ires and chains them into the ire array passed in. 20567c478bd9Sstevel@tonic-gate */ 20577c478bd9Sstevel@tonic-gate ire_t ** 20587c478bd9Sstevel@tonic-gate ire_create_bcast(ipif_t *ipif, ipaddr_t addr, ire_t **irep) 20597c478bd9Sstevel@tonic-gate { 2060*f4b3ec61Sdh155122 ip_stack_t *ipst = ipif->ipif_ill->ill_ipst; 2061*f4b3ec61Sdh155122 20627c478bd9Sstevel@tonic-gate *irep++ = ire_create( 20637c478bd9Sstevel@tonic-gate (uchar_t *)&addr, /* dest addr */ 20647c478bd9Sstevel@tonic-gate (uchar_t *)&ip_g_all_ones, /* mask */ 20657c478bd9Sstevel@tonic-gate (uchar_t *)&ipif->ipif_src_addr, /* source addr */ 20667c478bd9Sstevel@tonic-gate NULL, /* no gateway */ 20677c478bd9Sstevel@tonic-gate NULL, /* no in_src_addr */ 20687c478bd9Sstevel@tonic-gate &ipif->ipif_mtu, /* max frag */ 20697c478bd9Sstevel@tonic-gate NULL, /* fast path header */ 20707c478bd9Sstevel@tonic-gate ipif->ipif_rq, /* recv-from queue */ 20717c478bd9Sstevel@tonic-gate ipif->ipif_wq, /* send-to queue */ 20727c478bd9Sstevel@tonic-gate IRE_BROADCAST, 20737c478bd9Sstevel@tonic-gate ipif->ipif_bcast_mp, /* xmit header */ 20747c478bd9Sstevel@tonic-gate ipif, 20757c478bd9Sstevel@tonic-gate NULL, 20767c478bd9Sstevel@tonic-gate 0, 20777c478bd9Sstevel@tonic-gate 0, 20787c478bd9Sstevel@tonic-gate 0, 20797c478bd9Sstevel@tonic-gate 0, 208045916cd2Sjpk &ire_uinfo_null, 208145916cd2Sjpk NULL, 2082*f4b3ec61Sdh155122 NULL, 2083*f4b3ec61Sdh155122 ipst); 20847c478bd9Sstevel@tonic-gate 20857c478bd9Sstevel@tonic-gate *irep++ = ire_create( 20867c478bd9Sstevel@tonic-gate (uchar_t *)&addr, /* dest address */ 20877c478bd9Sstevel@tonic-gate (uchar_t *)&ip_g_all_ones, /* mask */ 20887c478bd9Sstevel@tonic-gate (uchar_t *)&ipif->ipif_src_addr, /* source address */ 20897c478bd9Sstevel@tonic-gate NULL, /* no gateway */ 20907c478bd9Sstevel@tonic-gate NULL, /* no in_src_addr */ 20917c478bd9Sstevel@tonic-gate &ip_loopback_mtu, /* max frag size */ 20927c478bd9Sstevel@tonic-gate NULL, /* Fast Path header */ 20937c478bd9Sstevel@tonic-gate ipif->ipif_rq, /* recv-from queue */ 20947c478bd9Sstevel@tonic-gate NULL, /* no send-to queue */ 20957c478bd9Sstevel@tonic-gate IRE_BROADCAST, /* Needed for fanout in wput */ 20967c478bd9Sstevel@tonic-gate NULL, 20977c478bd9Sstevel@tonic-gate ipif, 20987c478bd9Sstevel@tonic-gate NULL, 20997c478bd9Sstevel@tonic-gate 0, 21007c478bd9Sstevel@tonic-gate 0, 21017c478bd9Sstevel@tonic-gate 0, 21027c478bd9Sstevel@tonic-gate 0, 210345916cd2Sjpk &ire_uinfo_null, 210445916cd2Sjpk NULL, 2105*f4b3ec61Sdh155122 NULL, 2106*f4b3ec61Sdh155122 ipst); 21077c478bd9Sstevel@tonic-gate 21087c478bd9Sstevel@tonic-gate return (irep); 21097c478bd9Sstevel@tonic-gate } 21107c478bd9Sstevel@tonic-gate 21117c478bd9Sstevel@tonic-gate /* 21127c478bd9Sstevel@tonic-gate * ire_walk routine to delete or update any IRE_CACHE that might contain 21137c478bd9Sstevel@tonic-gate * stale information. 21147c478bd9Sstevel@tonic-gate * The flags state which entries to delete or update. 21157c478bd9Sstevel@tonic-gate * Garbage collection is done separately using kmem alloc callbacks to 21167c478bd9Sstevel@tonic-gate * ip_trash_ire_reclaim. 21177c478bd9Sstevel@tonic-gate * Used for both IPv4 and IPv6. However, IPv6 only uses FLUSH_MTU_TIME 21187c478bd9Sstevel@tonic-gate * since other stale information is cleaned up using NUD. 21197c478bd9Sstevel@tonic-gate */ 21207c478bd9Sstevel@tonic-gate void 21217c478bd9Sstevel@tonic-gate ire_expire(ire_t *ire, char *arg) 21227c478bd9Sstevel@tonic-gate { 2123*f4b3ec61Sdh155122 ire_expire_arg_t *ieap = (ire_expire_arg_t *)(uintptr_t)arg; 21247c478bd9Sstevel@tonic-gate ill_t *stq_ill; 2125*f4b3ec61Sdh155122 int flush_flags = ieap->iea_flush_flag; 2126*f4b3ec61Sdh155122 ip_stack_t *ipst = ieap->iea_ipst; 21277c478bd9Sstevel@tonic-gate 21287c478bd9Sstevel@tonic-gate if ((flush_flags & FLUSH_REDIRECT_TIME) && 21296bdb8e66Sdd193516 (ire->ire_flags & RTF_DYNAMIC)) { 21307c478bd9Sstevel@tonic-gate /* Make sure we delete the corresponding IRE_CACHE */ 21317c478bd9Sstevel@tonic-gate ip1dbg(("ire_expire: all redirects\n")); 2132*f4b3ec61Sdh155122 ip_rts_rtmsg(RTM_DELETE, ire, 0, ipst); 21337c478bd9Sstevel@tonic-gate ire_delete(ire); 2134*f4b3ec61Sdh155122 atomic_dec_32(&ipst->ips_ip_redirect_cnt); 21357c478bd9Sstevel@tonic-gate return; 21367c478bd9Sstevel@tonic-gate } 21377c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 21387c478bd9Sstevel@tonic-gate return; 21397c478bd9Sstevel@tonic-gate 21407c478bd9Sstevel@tonic-gate if (flush_flags & FLUSH_ARP_TIME) { 21417c478bd9Sstevel@tonic-gate /* 21427c478bd9Sstevel@tonic-gate * Remove all IRE_CACHE. 21437c478bd9Sstevel@tonic-gate * Verify that create time is more than 21447c478bd9Sstevel@tonic-gate * ip_ire_arp_interval milliseconds ago. 21457c478bd9Sstevel@tonic-gate */ 2146*f4b3ec61Sdh155122 if (NCE_EXPIRED(ire->ire_nce, ipst)) { 21477c478bd9Sstevel@tonic-gate ire_delete(ire); 21487c478bd9Sstevel@tonic-gate return; 21497c478bd9Sstevel@tonic-gate } 21507c478bd9Sstevel@tonic-gate } 21517c478bd9Sstevel@tonic-gate 2152*f4b3ec61Sdh155122 if (ipst->ips_ip_path_mtu_discovery && (flush_flags & FLUSH_MTU_TIME) && 21537c478bd9Sstevel@tonic-gate (ire->ire_ipif != NULL)) { 21547c478bd9Sstevel@tonic-gate /* Increase pmtu if it is less than the interface mtu */ 21557c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 21567c478bd9Sstevel@tonic-gate /* 21577c478bd9Sstevel@tonic-gate * If the ipif is a vni (whose mtu is 0, since it's virtual) 21587c478bd9Sstevel@tonic-gate * get the mtu from the sending interfaces' ipif 21597c478bd9Sstevel@tonic-gate */ 21607c478bd9Sstevel@tonic-gate if (IS_VNI(ire->ire_ipif->ipif_ill)) { 21617c478bd9Sstevel@tonic-gate stq_ill = ire->ire_stq->q_ptr; 21627c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(stq_ill->ill_ipif->ipif_mtu, 21637c478bd9Sstevel@tonic-gate IP_MAXPACKET); 21647c478bd9Sstevel@tonic-gate } else { 21657c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(ire->ire_ipif->ipif_mtu, 21667c478bd9Sstevel@tonic-gate IP_MAXPACKET); 21677c478bd9Sstevel@tonic-gate } 21687c478bd9Sstevel@tonic-gate ire->ire_frag_flag |= IPH_DF; 21697c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 21707c478bd9Sstevel@tonic-gate } 21717c478bd9Sstevel@tonic-gate } 21727c478bd9Sstevel@tonic-gate 21737c478bd9Sstevel@tonic-gate /* 21747c478bd9Sstevel@tonic-gate * Return any local address. We use this to target ourselves 21757c478bd9Sstevel@tonic-gate * when the src address was specified as 'default'. 21767c478bd9Sstevel@tonic-gate * Preference for IRE_LOCAL entries. 21777c478bd9Sstevel@tonic-gate */ 21787c478bd9Sstevel@tonic-gate ire_t * 2179*f4b3ec61Sdh155122 ire_lookup_local(zoneid_t zoneid, ip_stack_t *ipst) 21807c478bd9Sstevel@tonic-gate { 21817c478bd9Sstevel@tonic-gate ire_t *ire; 21827c478bd9Sstevel@tonic-gate irb_t *irb; 21837c478bd9Sstevel@tonic-gate ire_t *maybe = NULL; 21847c478bd9Sstevel@tonic-gate int i; 21857c478bd9Sstevel@tonic-gate 2186*f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 2187*f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[i]; 21887c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 21897c478bd9Sstevel@tonic-gate continue; 21907c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_READER); 21917c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; ire = ire->ire_next) { 21927c478bd9Sstevel@tonic-gate if ((ire->ire_marks & IRE_MARK_CONDEMNED) || 219345916cd2Sjpk (ire->ire_zoneid != zoneid && 219445916cd2Sjpk ire->ire_zoneid != ALL_ZONES)) 21957c478bd9Sstevel@tonic-gate continue; 21967c478bd9Sstevel@tonic-gate switch (ire->ire_type) { 21977c478bd9Sstevel@tonic-gate case IRE_LOOPBACK: 21987c478bd9Sstevel@tonic-gate if (maybe == NULL) { 21997c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 22007c478bd9Sstevel@tonic-gate maybe = ire; 22017c478bd9Sstevel@tonic-gate } 22027c478bd9Sstevel@tonic-gate break; 22037c478bd9Sstevel@tonic-gate case IRE_LOCAL: 22047c478bd9Sstevel@tonic-gate if (maybe != NULL) { 22057c478bd9Sstevel@tonic-gate ire_refrele(maybe); 22067c478bd9Sstevel@tonic-gate } 22077c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 22087c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 22097c478bd9Sstevel@tonic-gate return (ire); 22107c478bd9Sstevel@tonic-gate } 22117c478bd9Sstevel@tonic-gate } 22127c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 22137c478bd9Sstevel@tonic-gate } 22147c478bd9Sstevel@tonic-gate return (maybe); 22157c478bd9Sstevel@tonic-gate } 22167c478bd9Sstevel@tonic-gate 22177c478bd9Sstevel@tonic-gate /* 22187c478bd9Sstevel@tonic-gate * If the specified IRE is associated with a particular ILL, return 22197c478bd9Sstevel@tonic-gate * that ILL pointer (May be called as writer.). 22207c478bd9Sstevel@tonic-gate * 22217c478bd9Sstevel@tonic-gate * NOTE : This is not a generic function that can be used always. 22227c478bd9Sstevel@tonic-gate * This function always returns the ill of the outgoing packets 22237c478bd9Sstevel@tonic-gate * if this ire is used. 22247c478bd9Sstevel@tonic-gate */ 22257c478bd9Sstevel@tonic-gate ill_t * 222645916cd2Sjpk ire_to_ill(const ire_t *ire) 22277c478bd9Sstevel@tonic-gate { 22287c478bd9Sstevel@tonic-gate ill_t *ill = NULL; 22297c478bd9Sstevel@tonic-gate 22307c478bd9Sstevel@tonic-gate /* 22317c478bd9Sstevel@tonic-gate * 1) For an IRE_CACHE, ire_ipif is the one where it obtained 22327c478bd9Sstevel@tonic-gate * the source address from. ire_stq is the one where the 22337c478bd9Sstevel@tonic-gate * packets will be sent out on. We return that here. 22347c478bd9Sstevel@tonic-gate * 22357c478bd9Sstevel@tonic-gate * 2) IRE_BROADCAST normally has a loopback and a non-loopback 22367c478bd9Sstevel@tonic-gate * copy and they always exist next to each other with loopback 22377c478bd9Sstevel@tonic-gate * copy being the first one. If we are called on the non-loopback 22387c478bd9Sstevel@tonic-gate * copy, return the one pointed by ire_stq. If it was called on 22397c478bd9Sstevel@tonic-gate * a loopback copy, we still return the one pointed by the next 22407c478bd9Sstevel@tonic-gate * ire's ire_stq pointer i.e the one pointed by the non-loopback 22417c478bd9Sstevel@tonic-gate * copy. We don't want use ire_ipif as it might represent the 22427c478bd9Sstevel@tonic-gate * source address (if we borrow source addresses for 22437c478bd9Sstevel@tonic-gate * IRE_BROADCASTS in the future). 22447c478bd9Sstevel@tonic-gate * However if an interface is currently coming up, the above 22457c478bd9Sstevel@tonic-gate * condition may not hold during that period since the ires 22467c478bd9Sstevel@tonic-gate * are added one at a time. Thus one of the pair could have been 22477c478bd9Sstevel@tonic-gate * added and the other not yet added. 22482bc4236aSnordmark * 3) For many other IREs (e.g., IRE_LOCAL), ire_rfq indicates the ill. 22492bc4236aSnordmark * 4) For all others return the ones pointed by ire_ipif->ipif_ill. 22502bc4236aSnordmark * That handles IRE_LOOPBACK. 22517c478bd9Sstevel@tonic-gate */ 22527c478bd9Sstevel@tonic-gate 22537c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE) { 22547c478bd9Sstevel@tonic-gate ill = (ill_t *)ire->ire_stq->q_ptr; 22557c478bd9Sstevel@tonic-gate } else if (ire->ire_type == IRE_BROADCAST) { 22567c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) { 22577c478bd9Sstevel@tonic-gate ill = (ill_t *)ire->ire_stq->q_ptr; 22587c478bd9Sstevel@tonic-gate } else { 22597c478bd9Sstevel@tonic-gate ire_t *ire_next; 22607c478bd9Sstevel@tonic-gate 22617c478bd9Sstevel@tonic-gate ire_next = ire->ire_next; 22627c478bd9Sstevel@tonic-gate if (ire_next != NULL && 22637c478bd9Sstevel@tonic-gate ire_next->ire_type == IRE_BROADCAST && 22647c478bd9Sstevel@tonic-gate ire_next->ire_addr == ire->ire_addr && 22657c478bd9Sstevel@tonic-gate ire_next->ire_ipif == ire->ire_ipif) { 22667c478bd9Sstevel@tonic-gate ill = (ill_t *)ire_next->ire_stq->q_ptr; 22677c478bd9Sstevel@tonic-gate } 22687c478bd9Sstevel@tonic-gate } 22692bc4236aSnordmark } else if (ire->ire_rfq != NULL) { 22702bc4236aSnordmark ill = ire->ire_rfq->q_ptr; 22717c478bd9Sstevel@tonic-gate } else if (ire->ire_ipif != NULL) { 22727c478bd9Sstevel@tonic-gate ill = ire->ire_ipif->ipif_ill; 22737c478bd9Sstevel@tonic-gate } 22747c478bd9Sstevel@tonic-gate return (ill); 22757c478bd9Sstevel@tonic-gate } 22767c478bd9Sstevel@tonic-gate 22777c478bd9Sstevel@tonic-gate /* Arrange to call the specified function for every IRE in the world. */ 22787c478bd9Sstevel@tonic-gate void 2279*f4b3ec61Sdh155122 ire_walk(pfv_t func, void *arg, ip_stack_t *ipst) 22807c478bd9Sstevel@tonic-gate { 2281*f4b3ec61Sdh155122 ire_walk_ipvers(func, arg, 0, ALL_ZONES, ipst); 22827c478bd9Sstevel@tonic-gate } 22837c478bd9Sstevel@tonic-gate 22847c478bd9Sstevel@tonic-gate void 2285*f4b3ec61Sdh155122 ire_walk_v4(pfv_t func, void *arg, zoneid_t zoneid, ip_stack_t *ipst) 22867c478bd9Sstevel@tonic-gate { 2287*f4b3ec61Sdh155122 ire_walk_ipvers(func, arg, IPV4_VERSION, zoneid, ipst); 22887c478bd9Sstevel@tonic-gate } 22897c478bd9Sstevel@tonic-gate 22907c478bd9Sstevel@tonic-gate void 2291*f4b3ec61Sdh155122 ire_walk_v6(pfv_t func, void *arg, zoneid_t zoneid, ip_stack_t *ipst) 22927c478bd9Sstevel@tonic-gate { 2293*f4b3ec61Sdh155122 ire_walk_ipvers(func, arg, IPV6_VERSION, zoneid, ipst); 22947c478bd9Sstevel@tonic-gate } 22957c478bd9Sstevel@tonic-gate 22967c478bd9Sstevel@tonic-gate /* 22977c478bd9Sstevel@tonic-gate * Walk a particular version. version == 0 means both v4 and v6. 22987c478bd9Sstevel@tonic-gate */ 22997c478bd9Sstevel@tonic-gate static void 2300*f4b3ec61Sdh155122 ire_walk_ipvers(pfv_t func, void *arg, uchar_t vers, zoneid_t zoneid, 2301*f4b3ec61Sdh155122 ip_stack_t *ipst) 23027c478bd9Sstevel@tonic-gate { 23037c478bd9Sstevel@tonic-gate if (vers != IPV6_VERSION) { 2304c793af95Ssangeeta /* 2305c793af95Ssangeeta * ip_forwarding_table variable doesn't matter for IPv4 since 2306*f4b3ec61Sdh155122 * ire_walk_ill_tables uses ips_ip_ftable for IPv4. 2307c793af95Ssangeeta */ 23087c478bd9Sstevel@tonic-gate ire_walk_ill_tables(0, 0, func, arg, IP_MASK_TABLE_SIZE, 2309c793af95Ssangeeta 0, NULL, 2310*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size, ipst->ips_ip_cache_table, 2311*f4b3ec61Sdh155122 NULL, zoneid, ipst); 23127c478bd9Sstevel@tonic-gate } 23137c478bd9Sstevel@tonic-gate if (vers != IPV4_VERSION) { 23147c478bd9Sstevel@tonic-gate ire_walk_ill_tables(0, 0, func, arg, IP6_MASK_TABLE_SIZE, 2315*f4b3ec61Sdh155122 ipst->ips_ip6_ftable_hash_size, 2316*f4b3ec61Sdh155122 ipst->ips_ip_forwarding_table_v6, 2317*f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size, 2318*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6, NULL, zoneid, ipst); 23197c478bd9Sstevel@tonic-gate } 23207c478bd9Sstevel@tonic-gate } 23217c478bd9Sstevel@tonic-gate 23227c478bd9Sstevel@tonic-gate /* 23237c478bd9Sstevel@tonic-gate * Arrange to call the specified 23247c478bd9Sstevel@tonic-gate * function for every IRE that matches the ill. 23257c478bd9Sstevel@tonic-gate */ 23267c478bd9Sstevel@tonic-gate void 232745916cd2Sjpk ire_walk_ill(uint_t match_flags, uint_t ire_type, pfv_t func, void *arg, 23287c478bd9Sstevel@tonic-gate ill_t *ill) 23297c478bd9Sstevel@tonic-gate { 23307c478bd9Sstevel@tonic-gate ire_walk_ill_ipvers(match_flags, ire_type, func, arg, 0, ill); 23317c478bd9Sstevel@tonic-gate } 23327c478bd9Sstevel@tonic-gate 23337c478bd9Sstevel@tonic-gate void 233445916cd2Sjpk ire_walk_ill_v4(uint_t match_flags, uint_t ire_type, pfv_t func, void *arg, 23357c478bd9Sstevel@tonic-gate ill_t *ill) 23367c478bd9Sstevel@tonic-gate { 23377c478bd9Sstevel@tonic-gate ire_walk_ill_ipvers(match_flags, ire_type, func, arg, IPV4_VERSION, 23387c478bd9Sstevel@tonic-gate ill); 23397c478bd9Sstevel@tonic-gate } 23407c478bd9Sstevel@tonic-gate 23417c478bd9Sstevel@tonic-gate void 234245916cd2Sjpk ire_walk_ill_v6(uint_t match_flags, uint_t ire_type, pfv_t func, void *arg, 23437c478bd9Sstevel@tonic-gate ill_t *ill) 23447c478bd9Sstevel@tonic-gate { 23457c478bd9Sstevel@tonic-gate ire_walk_ill_ipvers(match_flags, ire_type, func, arg, IPV6_VERSION, 23467c478bd9Sstevel@tonic-gate ill); 23477c478bd9Sstevel@tonic-gate } 23487c478bd9Sstevel@tonic-gate 23497c478bd9Sstevel@tonic-gate /* 23507c478bd9Sstevel@tonic-gate * Walk a particular ill and version. version == 0 means both v4 and v6. 23517c478bd9Sstevel@tonic-gate */ 23527c478bd9Sstevel@tonic-gate static void 23537c478bd9Sstevel@tonic-gate ire_walk_ill_ipvers(uint_t match_flags, uint_t ire_type, pfv_t func, 235445916cd2Sjpk void *arg, uchar_t vers, ill_t *ill) 23557c478bd9Sstevel@tonic-gate { 2356*f4b3ec61Sdh155122 ip_stack_t *ipst = ill->ill_ipst; 2357*f4b3ec61Sdh155122 23587c478bd9Sstevel@tonic-gate if (vers != IPV6_VERSION) { 23597c478bd9Sstevel@tonic-gate ire_walk_ill_tables(match_flags, ire_type, func, arg, 2360c793af95Ssangeeta IP_MASK_TABLE_SIZE, 0, 2361*f4b3ec61Sdh155122 NULL, ipst->ips_ip_cache_table_size, 2362*f4b3ec61Sdh155122 ipst->ips_ip_cache_table, ill, ALL_ZONES, ipst); 23637c478bd9Sstevel@tonic-gate } 23647c478bd9Sstevel@tonic-gate if (vers != IPV4_VERSION) { 23657c478bd9Sstevel@tonic-gate ire_walk_ill_tables(match_flags, ire_type, func, arg, 2366*f4b3ec61Sdh155122 IP6_MASK_TABLE_SIZE, ipst->ips_ip6_ftable_hash_size, 2367*f4b3ec61Sdh155122 ipst->ips_ip_forwarding_table_v6, 2368*f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size, 2369*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6, ill, ALL_ZONES, ipst); 23707c478bd9Sstevel@tonic-gate } 23717c478bd9Sstevel@tonic-gate } 23727c478bd9Sstevel@tonic-gate 2373c793af95Ssangeeta boolean_t 23747c478bd9Sstevel@tonic-gate ire_walk_ill_match(uint_t match_flags, uint_t ire_type, ire_t *ire, 2375*f4b3ec61Sdh155122 ill_t *ill, zoneid_t zoneid, ip_stack_t *ipst) 23767c478bd9Sstevel@tonic-gate { 23777c478bd9Sstevel@tonic-gate ill_t *ire_stq_ill = NULL; 23787c478bd9Sstevel@tonic-gate ill_t *ire_ipif_ill = NULL; 23797c478bd9Sstevel@tonic-gate ill_group_t *ire_ill_group = NULL; 23807c478bd9Sstevel@tonic-gate 23817c478bd9Sstevel@tonic-gate ASSERT(match_flags != 0 || zoneid != ALL_ZONES); 23827c478bd9Sstevel@tonic-gate /* 23837c478bd9Sstevel@tonic-gate * 1) MATCH_IRE_WQ : Used specifically to match on ire_stq. 23847c478bd9Sstevel@tonic-gate * The fast path update uses this to make sure it does not 23857c478bd9Sstevel@tonic-gate * update the fast path header of interface X with the fast 23867c478bd9Sstevel@tonic-gate * path updates it recieved on interface Y. It is similar 23877c478bd9Sstevel@tonic-gate * in handling DL_NOTE_FASTPATH_FLUSH. 23887c478bd9Sstevel@tonic-gate * 23897c478bd9Sstevel@tonic-gate * 2) MATCH_IRE_ILL/MATCH_IRE_ILL_GROUP : We match both on ill 23907c478bd9Sstevel@tonic-gate * pointed by ire_stq and ire_ipif. Only in the case of 23917c478bd9Sstevel@tonic-gate * IRE_CACHEs can ire_stq and ire_ipif be pointing to 23927c478bd9Sstevel@tonic-gate * different ills. But we want to keep this function generic 23937c478bd9Sstevel@tonic-gate * enough for future use. So, we always try to match on both. 23947c478bd9Sstevel@tonic-gate * The only caller of this function ire_walk_ill_tables, will 23957c478bd9Sstevel@tonic-gate * call "func" after we return from this function. We expect 23967c478bd9Sstevel@tonic-gate * "func" to do the right filtering of ires in this case. 23977c478bd9Sstevel@tonic-gate * 23987c478bd9Sstevel@tonic-gate * NOTE : In the case of MATCH_IRE_ILL_GROUP, groups 23997c478bd9Sstevel@tonic-gate * pointed by ire_stq and ire_ipif should always be the same. 24007c478bd9Sstevel@tonic-gate * So, we just match on only one of them. 24017c478bd9Sstevel@tonic-gate */ 24027c478bd9Sstevel@tonic-gate if (match_flags & (MATCH_IRE_ILL|MATCH_IRE_ILL_GROUP)) { 24037c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 24047c478bd9Sstevel@tonic-gate ire_stq_ill = (ill_t *)ire->ire_stq->q_ptr; 24057c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) 24067c478bd9Sstevel@tonic-gate ire_ipif_ill = ire->ire_ipif->ipif_ill; 24077c478bd9Sstevel@tonic-gate if (ire_stq_ill != NULL) 24087c478bd9Sstevel@tonic-gate ire_ill_group = ire_stq_ill->ill_group; 24097c478bd9Sstevel@tonic-gate if ((ire_ill_group == NULL) && (ire_ipif_ill != NULL)) 24107c478bd9Sstevel@tonic-gate ire_ill_group = ire_ipif_ill->ill_group; 24117c478bd9Sstevel@tonic-gate } 24127c478bd9Sstevel@tonic-gate 24137c478bd9Sstevel@tonic-gate if (zoneid != ALL_ZONES) { 24147c478bd9Sstevel@tonic-gate /* 24157c478bd9Sstevel@tonic-gate * We're walking the IREs for a specific zone. The only relevant 24167c478bd9Sstevel@tonic-gate * IREs are: 24177c478bd9Sstevel@tonic-gate * - all IREs with a matching ire_zoneid 24187c478bd9Sstevel@tonic-gate * - all IRE_OFFSUBNETs as they're shared across all zones 24197c478bd9Sstevel@tonic-gate * - IRE_INTERFACE IREs for interfaces with a usable source addr 24207c478bd9Sstevel@tonic-gate * with a matching zone 24217c478bd9Sstevel@tonic-gate * - IRE_DEFAULTs with a gateway reachable from the zone 24227c478bd9Sstevel@tonic-gate * We should really match on IRE_OFFSUBNETs and IRE_DEFAULTs 24237c478bd9Sstevel@tonic-gate * using the same rule; but the above rules are consistent with 24247c478bd9Sstevel@tonic-gate * the behavior of ire_ftable_lookup[_v6]() so that all the 24257c478bd9Sstevel@tonic-gate * routes that can be matched during lookup are also matched 24267c478bd9Sstevel@tonic-gate * here. 24277c478bd9Sstevel@tonic-gate */ 242845916cd2Sjpk if (zoneid != ire->ire_zoneid && ire->ire_zoneid != ALL_ZONES) { 24297c478bd9Sstevel@tonic-gate /* 24307c478bd9Sstevel@tonic-gate * Note, IRE_INTERFACE can have the stq as NULL. For 24317c478bd9Sstevel@tonic-gate * example, if the default multicast route is tied to 24327c478bd9Sstevel@tonic-gate * the loopback address. 24337c478bd9Sstevel@tonic-gate */ 24347c478bd9Sstevel@tonic-gate if ((ire->ire_type & IRE_INTERFACE) && 24357c478bd9Sstevel@tonic-gate (ire->ire_stq != NULL)) { 24367c478bd9Sstevel@tonic-gate ire_stq_ill = (ill_t *)ire->ire_stq->q_ptr; 24377c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 24387c478bd9Sstevel@tonic-gate if (!ipif_usesrc_avail(ire_stq_ill, 24397c478bd9Sstevel@tonic-gate zoneid)) 24407c478bd9Sstevel@tonic-gate /* No usable src addr in zone */ 24417c478bd9Sstevel@tonic-gate return (B_FALSE); 24427c478bd9Sstevel@tonic-gate } else if (ire_stq_ill->ill_usesrc_ifindex 24437c478bd9Sstevel@tonic-gate != 0) { 24447c478bd9Sstevel@tonic-gate /* 24457c478bd9Sstevel@tonic-gate * For IPv6 use ipif_select_source_v6() 24467c478bd9Sstevel@tonic-gate * so the right scope selection is done 24477c478bd9Sstevel@tonic-gate */ 24487c478bd9Sstevel@tonic-gate ipif_t *src_ipif; 24497c478bd9Sstevel@tonic-gate src_ipif = 24507c478bd9Sstevel@tonic-gate ipif_select_source_v6(ire_stq_ill, 2451309fa174Srk129064 &ire->ire_addr_v6, RESTRICT_TO_NONE, 24527c478bd9Sstevel@tonic-gate IPV6_PREFER_SRC_DEFAULT, 24537c478bd9Sstevel@tonic-gate zoneid); 24547c478bd9Sstevel@tonic-gate if (src_ipif != NULL) { 24557c478bd9Sstevel@tonic-gate ipif_refrele(src_ipif); 24567c478bd9Sstevel@tonic-gate } else { 24577c478bd9Sstevel@tonic-gate return (B_FALSE); 24587c478bd9Sstevel@tonic-gate } 24597c478bd9Sstevel@tonic-gate } else { 24607c478bd9Sstevel@tonic-gate return (B_FALSE); 24617c478bd9Sstevel@tonic-gate } 24627c478bd9Sstevel@tonic-gate 24637c478bd9Sstevel@tonic-gate } else if (!(ire->ire_type & IRE_OFFSUBNET)) { 24647c478bd9Sstevel@tonic-gate return (B_FALSE); 24657c478bd9Sstevel@tonic-gate } 24667c478bd9Sstevel@tonic-gate } 24677c478bd9Sstevel@tonic-gate 24687c478bd9Sstevel@tonic-gate /* 24697c478bd9Sstevel@tonic-gate * Match all default routes from the global zone, irrespective 24705597b60aSnordmark * of reachability. For a non-global zone only match those 24715597b60aSnordmark * where ire_gateway_addr has a IRE_INTERFACE for the zoneid. 24727c478bd9Sstevel@tonic-gate */ 24737c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT && zoneid != GLOBAL_ZONEID) { 24747c478bd9Sstevel@tonic-gate int ire_match_flags = 0; 24757c478bd9Sstevel@tonic-gate in6_addr_t gw_addr_v6; 24767c478bd9Sstevel@tonic-gate ire_t *rire; 24777c478bd9Sstevel@tonic-gate 24785597b60aSnordmark ire_match_flags |= MATCH_IRE_TYPE; 24797c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 24807c478bd9Sstevel@tonic-gate ire_match_flags |= MATCH_IRE_ILL_GROUP; 24817c478bd9Sstevel@tonic-gate } 24827c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 24837c478bd9Sstevel@tonic-gate rire = ire_route_lookup(ire->ire_gateway_addr, 24845597b60aSnordmark 0, 0, IRE_INTERFACE, ire->ire_ipif, NULL, 2485*f4b3ec61Sdh155122 zoneid, NULL, ire_match_flags, ipst); 24867c478bd9Sstevel@tonic-gate } else { 24877c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV6_VERSION); 24887c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 24897c478bd9Sstevel@tonic-gate gw_addr_v6 = ire->ire_gateway_addr_v6; 24907c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 24917c478bd9Sstevel@tonic-gate rire = ire_route_lookup_v6(&gw_addr_v6, 24925597b60aSnordmark NULL, NULL, IRE_INTERFACE, ire->ire_ipif, 2493*f4b3ec61Sdh155122 NULL, zoneid, NULL, ire_match_flags, ipst); 24947c478bd9Sstevel@tonic-gate } 24957c478bd9Sstevel@tonic-gate if (rire == NULL) { 24967c478bd9Sstevel@tonic-gate return (B_FALSE); 24977c478bd9Sstevel@tonic-gate } 24987c478bd9Sstevel@tonic-gate ire_refrele(rire); 24997c478bd9Sstevel@tonic-gate } 25007c478bd9Sstevel@tonic-gate } 25017c478bd9Sstevel@tonic-gate 25027c478bd9Sstevel@tonic-gate if (((!(match_flags & MATCH_IRE_TYPE)) || 25037c478bd9Sstevel@tonic-gate (ire->ire_type & ire_type)) && 25047c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_WQ)) || 25057c478bd9Sstevel@tonic-gate (ire->ire_stq == ill->ill_wq)) && 25067c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL)) || 25077c478bd9Sstevel@tonic-gate (ire_stq_ill == ill || ire_ipif_ill == ill)) && 25087c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL_GROUP)) || 25097c478bd9Sstevel@tonic-gate (ire_stq_ill == ill) || (ire_ipif_ill == ill) || 25107c478bd9Sstevel@tonic-gate (ire_ill_group != NULL && 25117c478bd9Sstevel@tonic-gate ire_ill_group == ill->ill_group))) { 25127c478bd9Sstevel@tonic-gate return (B_TRUE); 25137c478bd9Sstevel@tonic-gate } 25147c478bd9Sstevel@tonic-gate return (B_FALSE); 25157c478bd9Sstevel@tonic-gate } 25167c478bd9Sstevel@tonic-gate 2517c793af95Ssangeeta int 2518c793af95Ssangeeta rtfunc(struct radix_node *rn, void *arg) 2519c793af95Ssangeeta { 2520c793af95Ssangeeta struct rtfuncarg *rtf = arg; 2521c793af95Ssangeeta struct rt_entry *rt; 2522c793af95Ssangeeta irb_t *irb; 2523c793af95Ssangeeta ire_t *ire; 2524c793af95Ssangeeta boolean_t ret; 2525c793af95Ssangeeta 2526c793af95Ssangeeta rt = (struct rt_entry *)rn; 2527c793af95Ssangeeta ASSERT(rt != NULL); 2528c793af95Ssangeeta irb = &rt->rt_irb; 2529c793af95Ssangeeta for (ire = irb->irb_ire; ire != NULL; ire = ire->ire_next) { 2530c793af95Ssangeeta if ((rtf->rt_match_flags != 0) || 2531c793af95Ssangeeta (rtf->rt_zoneid != ALL_ZONES)) { 2532c793af95Ssangeeta ret = ire_walk_ill_match(rtf->rt_match_flags, 2533c793af95Ssangeeta rtf->rt_ire_type, ire, 2534*f4b3ec61Sdh155122 rtf->rt_ill, rtf->rt_zoneid, rtf->rt_ipst); 2535c793af95Ssangeeta } else 2536c793af95Ssangeeta ret = B_TRUE; 2537c793af95Ssangeeta if (ret) 2538c793af95Ssangeeta (*rtf->rt_func)(ire, rtf->rt_arg); 2539c793af95Ssangeeta } 2540c793af95Ssangeeta return (0); 2541c793af95Ssangeeta } 2542c793af95Ssangeeta 25437c478bd9Sstevel@tonic-gate /* 25447c478bd9Sstevel@tonic-gate * Walk the ftable and the ctable entries that match the ill. 25457c478bd9Sstevel@tonic-gate */ 2546c793af95Ssangeeta void 25477c478bd9Sstevel@tonic-gate ire_walk_ill_tables(uint_t match_flags, uint_t ire_type, pfv_t func, 254845916cd2Sjpk void *arg, size_t ftbl_sz, size_t htbl_sz, irb_t **ipftbl, 2549*f4b3ec61Sdh155122 size_t ctbl_sz, irb_t *ipctbl, ill_t *ill, zoneid_t zoneid, 2550*f4b3ec61Sdh155122 ip_stack_t *ipst) 25517c478bd9Sstevel@tonic-gate { 25527c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 25537c478bd9Sstevel@tonic-gate irb_t *irb; 25547c478bd9Sstevel@tonic-gate ire_t *ire; 25557c478bd9Sstevel@tonic-gate int i, j; 25567c478bd9Sstevel@tonic-gate boolean_t ret; 2557c793af95Ssangeeta struct rtfuncarg rtfarg; 25587c478bd9Sstevel@tonic-gate 25597c478bd9Sstevel@tonic-gate ASSERT((!(match_flags & (MATCH_IRE_WQ | MATCH_IRE_ILL | 25607c478bd9Sstevel@tonic-gate MATCH_IRE_ILL_GROUP))) || (ill != NULL)); 25617c478bd9Sstevel@tonic-gate ASSERT(!(match_flags & MATCH_IRE_TYPE) || (ire_type != 0)); 25627c478bd9Sstevel@tonic-gate /* 25637c478bd9Sstevel@tonic-gate * Optimize by not looking at the forwarding table if there 25647c478bd9Sstevel@tonic-gate * is a MATCH_IRE_TYPE specified with no IRE_FORWARDTABLE 25657c478bd9Sstevel@tonic-gate * specified in ire_type. 25667c478bd9Sstevel@tonic-gate */ 25677c478bd9Sstevel@tonic-gate if (!(match_flags & MATCH_IRE_TYPE) || 25687c478bd9Sstevel@tonic-gate ((ire_type & IRE_FORWARDTABLE) != 0)) { 2569c793af95Ssangeeta /* knobs such that routine is called only for v6 case */ 2570*f4b3ec61Sdh155122 if (ipftbl == ipst->ips_ip_forwarding_table_v6) { 25717c478bd9Sstevel@tonic-gate for (i = (ftbl_sz - 1); i >= 0; i--) { 25727c478bd9Sstevel@tonic-gate if ((irb_ptr = ipftbl[i]) == NULL) 25737c478bd9Sstevel@tonic-gate continue; 25747c478bd9Sstevel@tonic-gate for (j = 0; j < htbl_sz; j++) { 25757c478bd9Sstevel@tonic-gate irb = &irb_ptr[j]; 25767c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 25777c478bd9Sstevel@tonic-gate continue; 2578c793af95Ssangeeta 25797c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 25807c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; 25817c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 25827c478bd9Sstevel@tonic-gate if (match_flags == 0 && 25837c478bd9Sstevel@tonic-gate zoneid == ALL_ZONES) { 25847c478bd9Sstevel@tonic-gate ret = B_TRUE; 25857c478bd9Sstevel@tonic-gate } else { 2586c793af95Ssangeeta ret = 2587c793af95Ssangeeta ire_walk_ill_match( 2588c793af95Ssangeeta match_flags, 2589c793af95Ssangeeta ire_type, ire, ill, 2590*f4b3ec61Sdh155122 zoneid, ipst); 25917c478bd9Sstevel@tonic-gate } 25927c478bd9Sstevel@tonic-gate if (ret) 25937c478bd9Sstevel@tonic-gate (*func)(ire, arg); 25947c478bd9Sstevel@tonic-gate } 25957c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 25967c478bd9Sstevel@tonic-gate } 25977c478bd9Sstevel@tonic-gate } 2598c793af95Ssangeeta } else { 2599c793af95Ssangeeta (void) memset(&rtfarg, 0, sizeof (rtfarg)); 2600c793af95Ssangeeta rtfarg.rt_func = func; 2601c793af95Ssangeeta rtfarg.rt_arg = arg; 2602c793af95Ssangeeta if (match_flags != 0) { 2603c793af95Ssangeeta rtfarg.rt_match_flags = match_flags; 2604c793af95Ssangeeta } 2605c793af95Ssangeeta rtfarg.rt_ire_type = ire_type; 2606c793af95Ssangeeta rtfarg.rt_ill = ill; 2607c793af95Ssangeeta rtfarg.rt_zoneid = zoneid; 2608*f4b3ec61Sdh155122 rtfarg.rt_ipst = ipst; /* No netstack_hold */ 2609*f4b3ec61Sdh155122 (void) ipst->ips_ip_ftable->rnh_walktree_mt( 2610*f4b3ec61Sdh155122 ipst->ips_ip_ftable, 2611*f4b3ec61Sdh155122 rtfunc, &rtfarg, irb_refhold_rn, irb_refrele_rn); 2612c793af95Ssangeeta } 26137c478bd9Sstevel@tonic-gate } 26147c478bd9Sstevel@tonic-gate 26157c478bd9Sstevel@tonic-gate /* 26167c478bd9Sstevel@tonic-gate * Optimize by not looking at the cache table if there 26177c478bd9Sstevel@tonic-gate * is a MATCH_IRE_TYPE specified with no IRE_CACHETABLE 26187c478bd9Sstevel@tonic-gate * specified in ire_type. 26197c478bd9Sstevel@tonic-gate */ 26207c478bd9Sstevel@tonic-gate if (!(match_flags & MATCH_IRE_TYPE) || 26217c478bd9Sstevel@tonic-gate ((ire_type & IRE_CACHETABLE) != 0)) { 26227c478bd9Sstevel@tonic-gate for (i = 0; i < ctbl_sz; i++) { 26237c478bd9Sstevel@tonic-gate irb = &ipctbl[i]; 26247c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 26257c478bd9Sstevel@tonic-gate continue; 26267c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 26277c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; 26287c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 26297c478bd9Sstevel@tonic-gate if (match_flags == 0 && zoneid == ALL_ZONES) { 26307c478bd9Sstevel@tonic-gate ret = B_TRUE; 26317c478bd9Sstevel@tonic-gate } else { 26327c478bd9Sstevel@tonic-gate ret = ire_walk_ill_match( 26337c478bd9Sstevel@tonic-gate match_flags, ire_type, 2634*f4b3ec61Sdh155122 ire, ill, zoneid, ipst); 26357c478bd9Sstevel@tonic-gate } 26367c478bd9Sstevel@tonic-gate if (ret) 26377c478bd9Sstevel@tonic-gate (*func)(ire, arg); 26387c478bd9Sstevel@tonic-gate } 26397c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 26407c478bd9Sstevel@tonic-gate } 26417c478bd9Sstevel@tonic-gate } 26427c478bd9Sstevel@tonic-gate } 26437c478bd9Sstevel@tonic-gate 26447c478bd9Sstevel@tonic-gate /* 26457c478bd9Sstevel@tonic-gate * This routine walks through the ill chain to find if there is any 26467c478bd9Sstevel@tonic-gate * ire linked to the ill's interface based forwarding table 26477c478bd9Sstevel@tonic-gate * The arg could be ill or mp. This routine is called when a ill goes 26487c478bd9Sstevel@tonic-gate * down/deleted or the 'ipv4_ire_srcif_status' report is printed. 26497c478bd9Sstevel@tonic-gate */ 26507c478bd9Sstevel@tonic-gate void 2651*f4b3ec61Sdh155122 ire_walk_srcif_table_v4(pfv_t func, void *arg, ip_stack_t *ipst) 26527c478bd9Sstevel@tonic-gate { 26537c478bd9Sstevel@tonic-gate irb_t *irb; 26547c478bd9Sstevel@tonic-gate ire_t *ire; 26557c478bd9Sstevel@tonic-gate ill_t *ill, *next_ill; 26567c478bd9Sstevel@tonic-gate int i; 26577c478bd9Sstevel@tonic-gate int total_count; 26587c478bd9Sstevel@tonic-gate ill_walk_context_t ctx; 26597c478bd9Sstevel@tonic-gate 26607c478bd9Sstevel@tonic-gate /* 26617c478bd9Sstevel@tonic-gate * Take care of ire's in other ill's per-interface forwarding 26627c478bd9Sstevel@tonic-gate * table. Check if any ire in any of the ill's ill_srcif_table 26637c478bd9Sstevel@tonic-gate * is pointing to this ill. 26647c478bd9Sstevel@tonic-gate */ 2665*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_srcif_table_lock); 2666*f4b3ec61Sdh155122 if (ipst->ips_ire_srcif_table_count == 0) { 2667*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_srcif_table_lock); 26687c478bd9Sstevel@tonic-gate return; 26697c478bd9Sstevel@tonic-gate } 2670*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_srcif_table_lock); 26717c478bd9Sstevel@tonic-gate 26727c478bd9Sstevel@tonic-gate #ifdef DEBUG 26737c478bd9Sstevel@tonic-gate /* Keep accounting of all interface based table ires */ 26747c478bd9Sstevel@tonic-gate total_count = 0; 2675*f4b3ec61Sdh155122 rw_enter(&ipst->ips_ill_g_lock, RW_READER); 2676*f4b3ec61Sdh155122 ill = ILL_START_WALK_V4(&ctx, ipst); 26777c478bd9Sstevel@tonic-gate while (ill != NULL) { 26787c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 26797c478bd9Sstevel@tonic-gate total_count += ill->ill_srcif_refcnt; 26807c478bd9Sstevel@tonic-gate next_ill = ill_next(&ctx, ill); 26817c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 26827c478bd9Sstevel@tonic-gate ill = next_ill; 26837c478bd9Sstevel@tonic-gate } 2684*f4b3ec61Sdh155122 rw_exit(&ipst->ips_ill_g_lock); 26857c478bd9Sstevel@tonic-gate 26867c478bd9Sstevel@tonic-gate /* Hold lock here to make sure ire_srcif_table_count is stable */ 2687*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_srcif_table_lock); 2688*f4b3ec61Sdh155122 i = ipst->ips_ire_srcif_table_count; 2689*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_srcif_table_lock); 26907c478bd9Sstevel@tonic-gate ip1dbg(("ire_walk_srcif_v4: ire_srcif_table_count %d " 26917c478bd9Sstevel@tonic-gate "total ill_srcif_refcnt %d\n", i, total_count)); 26927c478bd9Sstevel@tonic-gate #endif 2693*f4b3ec61Sdh155122 rw_enter(&ipst->ips_ill_g_lock, RW_READER); 2694*f4b3ec61Sdh155122 ill = ILL_START_WALK_V4(&ctx, ipst); 26957c478bd9Sstevel@tonic-gate while (ill != NULL) { 26967c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 26977c478bd9Sstevel@tonic-gate if ((ill->ill_srcif_refcnt == 0) || !ILL_CAN_LOOKUP(ill)) { 26987c478bd9Sstevel@tonic-gate next_ill = ill_next(&ctx, ill); 26997c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 27007c478bd9Sstevel@tonic-gate ill = next_ill; 27017c478bd9Sstevel@tonic-gate continue; 27027c478bd9Sstevel@tonic-gate } 27037c478bd9Sstevel@tonic-gate ill_refhold_locked(ill); 27047c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 2705*f4b3ec61Sdh155122 rw_exit(&ipst->ips_ill_g_lock); 27067c478bd9Sstevel@tonic-gate if (ill->ill_srcif_table != NULL) { 27077c478bd9Sstevel@tonic-gate for (i = 0; i < IP_SRCIF_TABLE_SIZE; i++) { 27087c478bd9Sstevel@tonic-gate irb = &(ill->ill_srcif_table[i]); 27097c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 27107c478bd9Sstevel@tonic-gate continue; 27117c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 27127c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; 27137c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 27147c478bd9Sstevel@tonic-gate (*func)(ire, arg); 27157c478bd9Sstevel@tonic-gate } 27167c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 27177c478bd9Sstevel@tonic-gate } 27187c478bd9Sstevel@tonic-gate } 2719*f4b3ec61Sdh155122 rw_enter(&ipst->ips_ill_g_lock, RW_READER); 27207c478bd9Sstevel@tonic-gate next_ill = ill_next(&ctx, ill); 27217c478bd9Sstevel@tonic-gate ill_refrele(ill); 27227c478bd9Sstevel@tonic-gate ill = next_ill; 27237c478bd9Sstevel@tonic-gate } 2724*f4b3ec61Sdh155122 rw_exit(&ipst->ips_ill_g_lock); 27257c478bd9Sstevel@tonic-gate } 27267c478bd9Sstevel@tonic-gate 27277c478bd9Sstevel@tonic-gate /* 27287c478bd9Sstevel@tonic-gate * This function takes a mask and returns 27297c478bd9Sstevel@tonic-gate * number of bits set in the mask. If no 27307c478bd9Sstevel@tonic-gate * bit is set it returns 0. 27317c478bd9Sstevel@tonic-gate * Assumes a contiguous mask. 27327c478bd9Sstevel@tonic-gate */ 27337c478bd9Sstevel@tonic-gate int 27347c478bd9Sstevel@tonic-gate ip_mask_to_plen(ipaddr_t mask) 27357c478bd9Sstevel@tonic-gate { 27367c478bd9Sstevel@tonic-gate return (mask == 0 ? 0 : IP_ABITS - (ffs(ntohl(mask)) -1)); 27377c478bd9Sstevel@tonic-gate } 27387c478bd9Sstevel@tonic-gate 27397c478bd9Sstevel@tonic-gate /* 27407c478bd9Sstevel@tonic-gate * Convert length for a mask to the mask. 27417c478bd9Sstevel@tonic-gate */ 27427c478bd9Sstevel@tonic-gate ipaddr_t 27437c478bd9Sstevel@tonic-gate ip_plen_to_mask(uint_t masklen) 27447c478bd9Sstevel@tonic-gate { 27457c478bd9Sstevel@tonic-gate return (htonl(IP_HOST_MASK << (IP_ABITS - masklen))); 27467c478bd9Sstevel@tonic-gate } 27477c478bd9Sstevel@tonic-gate 27487c478bd9Sstevel@tonic-gate void 27497c478bd9Sstevel@tonic-gate ire_atomic_end(irb_t *irb_ptr, ire_t *ire) 27507c478bd9Sstevel@tonic-gate { 27517c478bd9Sstevel@tonic-gate ill_t *ill_list[NUM_ILLS]; 2752*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 27537c478bd9Sstevel@tonic-gate 27547c478bd9Sstevel@tonic-gate ill_list[0] = ire->ire_stq != NULL ? ire->ire_stq->q_ptr : NULL; 27557c478bd9Sstevel@tonic-gate ill_list[1] = ire->ire_ipif != NULL ? ire->ire_ipif->ipif_ill : NULL; 27567c478bd9Sstevel@tonic-gate ill_list[2] = ire->ire_in_ill; 27577c478bd9Sstevel@tonic-gate ill_unlock_ills(ill_list, NUM_ILLS); 27587c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 2759*f4b3ec61Sdh155122 rw_exit(&ipst->ips_ill_g_usesrc_lock); 27607c478bd9Sstevel@tonic-gate } 27617c478bd9Sstevel@tonic-gate 27627c478bd9Sstevel@tonic-gate /* 27637c478bd9Sstevel@tonic-gate * ire_add_v[46] atomically make sure that the ipif or ill associated 27647c478bd9Sstevel@tonic-gate * with the new ire being added is stable and not IPIF_CHANGING or ILL_CHANGING 27657c478bd9Sstevel@tonic-gate * before adding the ire to the table. This ensures that we don't create 27667c478bd9Sstevel@tonic-gate * new IRE_CACHEs with stale values for parameters that are passed to 27677c478bd9Sstevel@tonic-gate * ire_create such as ire_max_frag. Note that ire_create() is passed a pointer 27687c478bd9Sstevel@tonic-gate * to the ipif_mtu, and not the value. The actual value is derived from the 27697c478bd9Sstevel@tonic-gate * parent ire or ipif under the bucket lock. 27707c478bd9Sstevel@tonic-gate */ 27717c478bd9Sstevel@tonic-gate int 27727c478bd9Sstevel@tonic-gate ire_atomic_start(irb_t *irb_ptr, ire_t *ire, queue_t *q, mblk_t *mp, 27737c478bd9Sstevel@tonic-gate ipsq_func_t func) 27747c478bd9Sstevel@tonic-gate { 27757c478bd9Sstevel@tonic-gate ill_t *stq_ill; 27767c478bd9Sstevel@tonic-gate ill_t *ipif_ill; 27777c478bd9Sstevel@tonic-gate ill_t *in_ill; 27787c478bd9Sstevel@tonic-gate ill_t *ill_list[NUM_ILLS]; 27797c478bd9Sstevel@tonic-gate int cnt = NUM_ILLS; 27807c478bd9Sstevel@tonic-gate int error = 0; 27817c478bd9Sstevel@tonic-gate ill_t *ill = NULL; 2782*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 27837c478bd9Sstevel@tonic-gate 27847c478bd9Sstevel@tonic-gate ill_list[0] = stq_ill = ire->ire_stq != 27857c478bd9Sstevel@tonic-gate NULL ? ire->ire_stq->q_ptr : NULL; 27867c478bd9Sstevel@tonic-gate ill_list[1] = ipif_ill = ire->ire_ipif != 27877c478bd9Sstevel@tonic-gate NULL ? ire->ire_ipif->ipif_ill : NULL; 27887c478bd9Sstevel@tonic-gate ill_list[2] = in_ill = ire->ire_in_ill; 27897c478bd9Sstevel@tonic-gate 27907c478bd9Sstevel@tonic-gate ASSERT((q != NULL && mp != NULL && func != NULL) || 27917c478bd9Sstevel@tonic-gate (q == NULL && mp == NULL && func == NULL)); 2792*f4b3ec61Sdh155122 rw_enter(&ipst->ips_ill_g_usesrc_lock, RW_READER); 27937c478bd9Sstevel@tonic-gate GRAB_CONN_LOCK(q); 27947c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_WRITER); 27957c478bd9Sstevel@tonic-gate ill_lock_ills(ill_list, cnt); 27967c478bd9Sstevel@tonic-gate 27977c478bd9Sstevel@tonic-gate /* 27987c478bd9Sstevel@tonic-gate * While the IRE is in the process of being added, a user may have 27997c478bd9Sstevel@tonic-gate * invoked the ifconfig usesrc option on the stq_ill to make it a 28007c478bd9Sstevel@tonic-gate * usesrc client ILL. Check for this possibility here, if it is true 28017c478bd9Sstevel@tonic-gate * then we fail adding the IRE_CACHE. Another check is to make sure 28027c478bd9Sstevel@tonic-gate * that an ipif_ill of an IRE_CACHE being added is not part of a usesrc 28037c478bd9Sstevel@tonic-gate * group. The ill_g_usesrc_lock is released in ire_atomic_end 28047c478bd9Sstevel@tonic-gate */ 28057c478bd9Sstevel@tonic-gate if ((ire->ire_type & IRE_CACHE) && 28067c478bd9Sstevel@tonic-gate (ire->ire_marks & IRE_MARK_USESRC_CHECK)) { 28077c478bd9Sstevel@tonic-gate if (stq_ill->ill_usesrc_ifindex != 0) { 28087c478bd9Sstevel@tonic-gate ASSERT(stq_ill->ill_usesrc_grp_next != NULL); 28097c478bd9Sstevel@tonic-gate if ((ipif_ill->ill_phyint->phyint_ifindex != 28107c478bd9Sstevel@tonic-gate stq_ill->ill_usesrc_ifindex) || 28117c478bd9Sstevel@tonic-gate (ipif_ill->ill_usesrc_grp_next == NULL) || 28127c478bd9Sstevel@tonic-gate (ipif_ill->ill_usesrc_ifindex != 0)) { 28137c478bd9Sstevel@tonic-gate error = EINVAL; 28147c478bd9Sstevel@tonic-gate goto done; 28157c478bd9Sstevel@tonic-gate } 28167c478bd9Sstevel@tonic-gate } else if (ipif_ill->ill_usesrc_grp_next != NULL) { 28177c478bd9Sstevel@tonic-gate error = EINVAL; 28187c478bd9Sstevel@tonic-gate goto done; 28197c478bd9Sstevel@tonic-gate } 28207c478bd9Sstevel@tonic-gate } 28217c478bd9Sstevel@tonic-gate 28227c478bd9Sstevel@tonic-gate /* 28237c478bd9Sstevel@tonic-gate * IPMP flag settings happen without taking the exclusive route 28247c478bd9Sstevel@tonic-gate * in ip_sioctl_flags. So we need to make an atomic check here 28257c478bd9Sstevel@tonic-gate * for FAILED/OFFLINE/INACTIVE flags or if it has hit the 28267c478bd9Sstevel@tonic-gate * FAILBACK=no case. 28277c478bd9Sstevel@tonic-gate */ 28287c478bd9Sstevel@tonic-gate if ((stq_ill != NULL) && !IAM_WRITER_ILL(stq_ill)) { 28297c478bd9Sstevel@tonic-gate if (stq_ill->ill_state_flags & ILL_CHANGING) { 28307c478bd9Sstevel@tonic-gate ill = stq_ill; 28317c478bd9Sstevel@tonic-gate error = EAGAIN; 28327c478bd9Sstevel@tonic-gate } else if ((stq_ill->ill_phyint->phyint_flags & PHYI_OFFLINE) || 28337c478bd9Sstevel@tonic-gate (ill_is_probeonly(stq_ill) && 28347c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_HIDDEN))) { 28357c478bd9Sstevel@tonic-gate error = EINVAL; 28367c478bd9Sstevel@tonic-gate } 28377c478bd9Sstevel@tonic-gate goto done; 28387c478bd9Sstevel@tonic-gate } 28397c478bd9Sstevel@tonic-gate 28407c478bd9Sstevel@tonic-gate /* 28417c478bd9Sstevel@tonic-gate * We don't check for OFFLINE/FAILED in this case because 28427c478bd9Sstevel@tonic-gate * the source address selection logic (ipif_select_source) 28437c478bd9Sstevel@tonic-gate * may still select a source address from such an ill. The 28447c478bd9Sstevel@tonic-gate * assumption is that these addresses will be moved by in.mpathd 28457c478bd9Sstevel@tonic-gate * soon. (i.e. this is a race). However link local addresses 28467c478bd9Sstevel@tonic-gate * will not move and hence ipif_select_source_v6 tries to avoid 28477c478bd9Sstevel@tonic-gate * FAILED ills. Please see ipif_select_source_v6 for more info 28487c478bd9Sstevel@tonic-gate */ 28497c478bd9Sstevel@tonic-gate if ((ipif_ill != NULL) && !IAM_WRITER_ILL(ipif_ill) && 28507c478bd9Sstevel@tonic-gate (ipif_ill->ill_state_flags & ILL_CHANGING)) { 28517c478bd9Sstevel@tonic-gate ill = ipif_ill; 28527c478bd9Sstevel@tonic-gate error = EAGAIN; 28537c478bd9Sstevel@tonic-gate goto done; 28547c478bd9Sstevel@tonic-gate } 28557c478bd9Sstevel@tonic-gate 28567c478bd9Sstevel@tonic-gate if ((in_ill != NULL) && !IAM_WRITER_ILL(in_ill) && 28577c478bd9Sstevel@tonic-gate (in_ill->ill_state_flags & ILL_CHANGING)) { 28587c478bd9Sstevel@tonic-gate ill = in_ill; 28597c478bd9Sstevel@tonic-gate error = EAGAIN; 28607c478bd9Sstevel@tonic-gate goto done; 28617c478bd9Sstevel@tonic-gate } 28627c478bd9Sstevel@tonic-gate 28637c478bd9Sstevel@tonic-gate if ((ire->ire_ipif != NULL) && !IAM_WRITER_IPIF(ire->ire_ipif) && 28647c478bd9Sstevel@tonic-gate (ire->ire_ipif->ipif_state_flags & IPIF_CHANGING)) { 28657c478bd9Sstevel@tonic-gate ill = ire->ire_ipif->ipif_ill; 28667c478bd9Sstevel@tonic-gate ASSERT(ill != NULL); 28677c478bd9Sstevel@tonic-gate error = EAGAIN; 28687c478bd9Sstevel@tonic-gate goto done; 28697c478bd9Sstevel@tonic-gate } 28707c478bd9Sstevel@tonic-gate 28717c478bd9Sstevel@tonic-gate done: 28727c478bd9Sstevel@tonic-gate if (error == EAGAIN && ILL_CAN_WAIT(ill, q)) { 28737c478bd9Sstevel@tonic-gate ipsq_t *ipsq = ill->ill_phyint->phyint_ipsq; 28747c478bd9Sstevel@tonic-gate mutex_enter(&ipsq->ipsq_lock); 28757c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 28767c478bd9Sstevel@tonic-gate ipsq_enq(ipsq, q, mp, func, NEW_OP, ill); 28777c478bd9Sstevel@tonic-gate mutex_exit(&ipsq->ipsq_lock); 28787c478bd9Sstevel@tonic-gate error = EINPROGRESS; 28797c478bd9Sstevel@tonic-gate } else if (error != 0) { 28807c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 28817c478bd9Sstevel@tonic-gate } 28827c478bd9Sstevel@tonic-gate 28837c478bd9Sstevel@tonic-gate RELEASE_CONN_LOCK(q); 28847c478bd9Sstevel@tonic-gate return (error); 28857c478bd9Sstevel@tonic-gate } 28867c478bd9Sstevel@tonic-gate 28877c478bd9Sstevel@tonic-gate /* 28887c478bd9Sstevel@tonic-gate * Add a fully initialized IRE to an appropriate table based on 28897c478bd9Sstevel@tonic-gate * ire_type. 2890c793af95Ssangeeta * 2891c793af95Ssangeeta * allow_unresolved == B_FALSE indicates a legacy code-path call 2892c793af95Ssangeeta * that has prohibited the addition of incomplete ire's. If this 2893c793af95Ssangeeta * parameter is set, and we find an nce that is in a state other 2894c793af95Ssangeeta * than ND_REACHABLE, we fail the add. Note that nce_state could be 2895c793af95Ssangeeta * something other than ND_REACHABLE if nce_reinit has just 2896c793af95Ssangeeta * kicked in and reset the nce. 28977c478bd9Sstevel@tonic-gate */ 28987c478bd9Sstevel@tonic-gate int 2899c793af95Ssangeeta ire_add(ire_t **irep, queue_t *q, mblk_t *mp, ipsq_func_t func, 2900c793af95Ssangeeta boolean_t allow_unresolved) 29017c478bd9Sstevel@tonic-gate { 29027c478bd9Sstevel@tonic-gate ire_t *ire1; 29037c478bd9Sstevel@tonic-gate ill_t *stq_ill = NULL; 29047c478bd9Sstevel@tonic-gate ill_t *ill; 29057c478bd9Sstevel@tonic-gate ipif_t *ipif = NULL; 29067c478bd9Sstevel@tonic-gate ill_walk_context_t ctx; 29077c478bd9Sstevel@tonic-gate ire_t *ire = *irep; 29087c478bd9Sstevel@tonic-gate int error; 2909dc3879f9Sjarrett boolean_t ire_is_mblk = B_FALSE; 2910dc3879f9Sjarrett tsol_gcgrp_t *gcgrp = NULL; 2911dc3879f9Sjarrett tsol_gcgrp_addr_t ga; 2912*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 29137c478bd9Sstevel@tonic-gate 29147c478bd9Sstevel@tonic-gate ASSERT(ire->ire_type != IRE_MIPRTUN); 29157c478bd9Sstevel@tonic-gate 29167c478bd9Sstevel@tonic-gate /* get ready for the day when original ire is not created as mblk */ 29177c478bd9Sstevel@tonic-gate if (ire->ire_mp != NULL) { 2918dc3879f9Sjarrett ire_is_mblk = B_TRUE; 29197c478bd9Sstevel@tonic-gate /* Copy the ire to a kmem_alloc'ed area */ 29207c478bd9Sstevel@tonic-gate ire1 = kmem_cache_alloc(ire_cache, KM_NOSLEEP); 29217c478bd9Sstevel@tonic-gate if (ire1 == NULL) { 29227c478bd9Sstevel@tonic-gate ip1dbg(("ire_add: alloc failed\n")); 29237c478bd9Sstevel@tonic-gate ire_delete(ire); 29247c478bd9Sstevel@tonic-gate *irep = NULL; 29257c478bd9Sstevel@tonic-gate return (ENOMEM); 29267c478bd9Sstevel@tonic-gate } 2927c793af95Ssangeeta ire->ire_marks &= ~IRE_MARK_UNCACHED; 29287c478bd9Sstevel@tonic-gate *ire1 = *ire; 29297c478bd9Sstevel@tonic-gate ire1->ire_mp = NULL; 2930c793af95Ssangeeta ire1->ire_stq_ifindex = 0; 29317c478bd9Sstevel@tonic-gate freeb(ire->ire_mp); 29327c478bd9Sstevel@tonic-gate ire = ire1; 29337c478bd9Sstevel@tonic-gate } 29347c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 29357c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 29367c478bd9Sstevel@tonic-gate 29377c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE) { 29387c478bd9Sstevel@tonic-gate /* 29397c478bd9Sstevel@tonic-gate * If this interface is FAILED, or INACTIVE or has hit 29407c478bd9Sstevel@tonic-gate * the FAILBACK=no case, we create IRE_CACHES marked 29417c478bd9Sstevel@tonic-gate * HIDDEN for some special cases e.g. bind to 29427c478bd9Sstevel@tonic-gate * IPIF_NOFAILOVER address etc. So, if this interface 29437c478bd9Sstevel@tonic-gate * is FAILED/INACTIVE/hit FAILBACK=no case, and we are 29447c478bd9Sstevel@tonic-gate * not creating hidden ires, we should not allow that. 29457c478bd9Sstevel@tonic-gate * This happens because the state of the interface 29467c478bd9Sstevel@tonic-gate * changed while we were waiting in ARP. If this is the 29477c478bd9Sstevel@tonic-gate * daemon sending probes, the next probe will create 29487c478bd9Sstevel@tonic-gate * HIDDEN ires and we will create an ire then. This 29497c478bd9Sstevel@tonic-gate * cannot happen with NDP currently because IRE is 29507c478bd9Sstevel@tonic-gate * never queued in NDP. But it can happen in the 29517c478bd9Sstevel@tonic-gate * future when we have external resolvers with IPv6. 29527c478bd9Sstevel@tonic-gate * If the interface gets marked with OFFLINE while we 29537c478bd9Sstevel@tonic-gate * are waiting in ARP, don't add the ire. 29547c478bd9Sstevel@tonic-gate */ 29557c478bd9Sstevel@tonic-gate if ((stq_ill->ill_phyint->phyint_flags & PHYI_OFFLINE) || 29567c478bd9Sstevel@tonic-gate (ill_is_probeonly(stq_ill) && 29577c478bd9Sstevel@tonic-gate !(ire->ire_marks & IRE_MARK_HIDDEN))) { 29587c478bd9Sstevel@tonic-gate /* 29597c478bd9Sstevel@tonic-gate * We don't know whether it is a valid ipif or not. 29607c478bd9Sstevel@tonic-gate * unless we do the check below. So, set it to NULL. 29617c478bd9Sstevel@tonic-gate */ 29627c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 29637c478bd9Sstevel@tonic-gate ire_delete(ire); 29647c478bd9Sstevel@tonic-gate *irep = NULL; 29657c478bd9Sstevel@tonic-gate return (EINVAL); 29667c478bd9Sstevel@tonic-gate } 29677c478bd9Sstevel@tonic-gate } 29687c478bd9Sstevel@tonic-gate 29697c478bd9Sstevel@tonic-gate if (stq_ill != NULL && ire->ire_type == IRE_CACHE && 29707c478bd9Sstevel@tonic-gate stq_ill->ill_net_type == IRE_IF_RESOLVER) { 2971*f4b3ec61Sdh155122 rw_enter(&ipst->ips_ill_g_lock, RW_READER); 2972*f4b3ec61Sdh155122 ill = ILL_START_WALK_ALL(&ctx, ipst); 29737c478bd9Sstevel@tonic-gate for (; ill != NULL; ill = ill_next(&ctx, ill)) { 29747c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 29757c478bd9Sstevel@tonic-gate if (ill->ill_state_flags & ILL_CONDEMNED) { 29767c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 29777c478bd9Sstevel@tonic-gate continue; 29787c478bd9Sstevel@tonic-gate } 29797c478bd9Sstevel@tonic-gate /* 29807c478bd9Sstevel@tonic-gate * We need to make sure that the ipif is a valid one 29817c478bd9Sstevel@tonic-gate * before adding the IRE_CACHE. This happens only 29827c478bd9Sstevel@tonic-gate * with IRE_CACHE when there is an external resolver. 29837c478bd9Sstevel@tonic-gate * 29847c478bd9Sstevel@tonic-gate * We can unplumb a logical interface while the 29857c478bd9Sstevel@tonic-gate * packet is waiting in ARP with the IRE. Then, 29867c478bd9Sstevel@tonic-gate * later on when we feed the IRE back, the ipif 29877c478bd9Sstevel@tonic-gate * has to be re-checked. This can't happen with 29887c478bd9Sstevel@tonic-gate * NDP currently, as we never queue the IRE with 29897c478bd9Sstevel@tonic-gate * the packet. We always try to recreate the IRE 29907c478bd9Sstevel@tonic-gate * when the resolution is completed. But, we do 29917c478bd9Sstevel@tonic-gate * it for IPv6 also here so that in future if 29927c478bd9Sstevel@tonic-gate * we have external resolvers, it will work without 29937c478bd9Sstevel@tonic-gate * any change. 29947c478bd9Sstevel@tonic-gate */ 29957c478bd9Sstevel@tonic-gate ipif = ipif_lookup_seqid(ill, ire->ire_ipif_seqid); 29967c478bd9Sstevel@tonic-gate if (ipif != NULL) { 29977c478bd9Sstevel@tonic-gate ipif_refhold_locked(ipif); 29987c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 29997c478bd9Sstevel@tonic-gate break; 30007c478bd9Sstevel@tonic-gate } 30017c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 30027c478bd9Sstevel@tonic-gate } 3003*f4b3ec61Sdh155122 rw_exit(&ipst->ips_ill_g_lock); 30047c478bd9Sstevel@tonic-gate if (ipif == NULL || 30057c478bd9Sstevel@tonic-gate (ipif->ipif_isv6 && 30067c478bd9Sstevel@tonic-gate !IN6_ARE_ADDR_EQUAL(&ire->ire_src_addr_v6, 30077c478bd9Sstevel@tonic-gate &ipif->ipif_v6src_addr)) || 30087c478bd9Sstevel@tonic-gate (!ipif->ipif_isv6 && 30097c478bd9Sstevel@tonic-gate ire->ire_src_addr != ipif->ipif_src_addr) || 301045916cd2Sjpk ire->ire_zoneid != ipif->ipif_zoneid) { 30117c478bd9Sstevel@tonic-gate 30127c478bd9Sstevel@tonic-gate if (ipif != NULL) 30137c478bd9Sstevel@tonic-gate ipif_refrele(ipif); 30147c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 30157c478bd9Sstevel@tonic-gate ire_delete(ire); 30167c478bd9Sstevel@tonic-gate *irep = NULL; 30177c478bd9Sstevel@tonic-gate return (EINVAL); 30187c478bd9Sstevel@tonic-gate } 30197c478bd9Sstevel@tonic-gate 30207c478bd9Sstevel@tonic-gate 30217c478bd9Sstevel@tonic-gate ASSERT(ill != NULL); 30227c478bd9Sstevel@tonic-gate /* 30237c478bd9Sstevel@tonic-gate * If this group was dismantled while this packets was 30247c478bd9Sstevel@tonic-gate * queued in ARP, don't add it here. 30257c478bd9Sstevel@tonic-gate */ 30267c478bd9Sstevel@tonic-gate if (ire->ire_ipif->ipif_ill->ill_group != ill->ill_group) { 30277c478bd9Sstevel@tonic-gate /* We don't want ire_inactive bump stats for this */ 30287c478bd9Sstevel@tonic-gate ipif_refrele(ipif); 30297c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 30307c478bd9Sstevel@tonic-gate ire_delete(ire); 30317c478bd9Sstevel@tonic-gate *irep = NULL; 30327c478bd9Sstevel@tonic-gate return (EINVAL); 30337c478bd9Sstevel@tonic-gate } 3034dc3879f9Sjarrett 3035dc3879f9Sjarrett /* 3036dc3879f9Sjarrett * Since we didn't attach label security attributes to the 3037dc3879f9Sjarrett * ire for the resolver case, we need to add it now. (only 3038dc3879f9Sjarrett * for v4 resolver and v6 xresolv case). 3039dc3879f9Sjarrett */ 3040dc3879f9Sjarrett if (is_system_labeled() && ire_is_mblk) { 3041dc3879f9Sjarrett if (ire->ire_ipversion == IPV4_VERSION) { 3042dc3879f9Sjarrett ga.ga_af = AF_INET; 3043dc3879f9Sjarrett IN6_IPADDR_TO_V4MAPPED(ire->ire_gateway_addr != 3044dc3879f9Sjarrett INADDR_ANY ? ire->ire_gateway_addr : 3045dc3879f9Sjarrett ire->ire_addr, &ga.ga_addr); 3046dc3879f9Sjarrett } else { 3047dc3879f9Sjarrett ga.ga_af = AF_INET6; 3048dc3879f9Sjarrett ga.ga_addr = IN6_IS_ADDR_UNSPECIFIED( 3049dc3879f9Sjarrett &ire->ire_gateway_addr_v6) ? 3050dc3879f9Sjarrett ire->ire_addr_v6 : 3051dc3879f9Sjarrett ire->ire_gateway_addr_v6; 3052dc3879f9Sjarrett } 3053dc3879f9Sjarrett gcgrp = gcgrp_lookup(&ga, B_FALSE); 3054dc3879f9Sjarrett error = tsol_ire_init_gwattr(ire, ire->ire_ipversion, 3055dc3879f9Sjarrett NULL, gcgrp); 3056dc3879f9Sjarrett if (error != 0) { 3057dc3879f9Sjarrett if (gcgrp != NULL) { 3058dc3879f9Sjarrett GCGRP_REFRELE(gcgrp); 3059dc3879f9Sjarrett gcgrp = NULL; 3060dc3879f9Sjarrett } 3061dc3879f9Sjarrett ipif_refrele(ipif); 3062dc3879f9Sjarrett ire->ire_ipif = NULL; 3063dc3879f9Sjarrett ire_delete(ire); 3064dc3879f9Sjarrett *irep = NULL; 3065dc3879f9Sjarrett return (error); 3066dc3879f9Sjarrett } 3067dc3879f9Sjarrett } 30687c478bd9Sstevel@tonic-gate } 30697c478bd9Sstevel@tonic-gate 30707c478bd9Sstevel@tonic-gate /* 30717c478bd9Sstevel@tonic-gate * In case ire was changed 30727c478bd9Sstevel@tonic-gate */ 30737c478bd9Sstevel@tonic-gate *irep = ire; 30747c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 30757c478bd9Sstevel@tonic-gate error = ire_add_v6(irep, q, mp, func); 30767c478bd9Sstevel@tonic-gate } else { 30777c478bd9Sstevel@tonic-gate if (ire->ire_in_ill == NULL) 3078c793af95Ssangeeta error = ire_add_v4(irep, q, mp, func, allow_unresolved); 30797c478bd9Sstevel@tonic-gate else 30807c478bd9Sstevel@tonic-gate error = ire_add_srcif_v4(irep, q, mp, func); 30817c478bd9Sstevel@tonic-gate } 30827c478bd9Sstevel@tonic-gate if (ipif != NULL) 30837c478bd9Sstevel@tonic-gate ipif_refrele(ipif); 30847c478bd9Sstevel@tonic-gate return (error); 30857c478bd9Sstevel@tonic-gate } 30867c478bd9Sstevel@tonic-gate 30877c478bd9Sstevel@tonic-gate /* 3088dc3879f9Sjarrett * Add an initialized IRE to an appropriate table based on ire_type. 30897c478bd9Sstevel@tonic-gate * 30906bdb8e66Sdd193516 * The forward table contains IRE_PREFIX/IRE_HOST and 30917c478bd9Sstevel@tonic-gate * IRE_IF_RESOLVER/IRE_IF_NORESOLVER and IRE_DEFAULT. 30927c478bd9Sstevel@tonic-gate * 30937c478bd9Sstevel@tonic-gate * The cache table contains IRE_BROADCAST/IRE_LOCAL/IRE_LOOPBACK 30947c478bd9Sstevel@tonic-gate * and IRE_CACHE. 30957c478bd9Sstevel@tonic-gate * 30967c478bd9Sstevel@tonic-gate * NOTE : This function is called as writer though not required 30977c478bd9Sstevel@tonic-gate * by this function. 30987c478bd9Sstevel@tonic-gate */ 30997c478bd9Sstevel@tonic-gate static int 3100c793af95Ssangeeta ire_add_v4(ire_t **ire_p, queue_t *q, mblk_t *mp, ipsq_func_t func, 3101c793af95Ssangeeta boolean_t allow_unresolved) 31027c478bd9Sstevel@tonic-gate { 31037c478bd9Sstevel@tonic-gate ire_t *ire1; 31047c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 31057c478bd9Sstevel@tonic-gate ire_t **irep; 31067c478bd9Sstevel@tonic-gate int flags; 31077c478bd9Sstevel@tonic-gate ire_t *pire = NULL; 31087c478bd9Sstevel@tonic-gate ill_t *stq_ill; 31097c478bd9Sstevel@tonic-gate ire_t *ire = *ire_p; 31107c478bd9Sstevel@tonic-gate int error; 3111c793af95Ssangeeta boolean_t need_refrele = B_FALSE; 3112c793af95Ssangeeta nce_t *nce; 3113*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 31147c478bd9Sstevel@tonic-gate 31157c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) 31167c478bd9Sstevel@tonic-gate ASSERT(!MUTEX_HELD(&ire->ire_ipif->ipif_ill->ill_lock)); 31177c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 31187c478bd9Sstevel@tonic-gate ASSERT(!MUTEX_HELD( 31197c478bd9Sstevel@tonic-gate &((ill_t *)(ire->ire_stq->q_ptr))->ill_lock)); 31207c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 31217c478bd9Sstevel@tonic-gate ASSERT(ire->ire_mp == NULL); /* Calls should go through ire_add */ 31227c478bd9Sstevel@tonic-gate ASSERT(ire->ire_in_ill == NULL); /* No srcif entries */ 31237c478bd9Sstevel@tonic-gate 31247c478bd9Sstevel@tonic-gate /* Find the appropriate list head. */ 31257c478bd9Sstevel@tonic-gate switch (ire->ire_type) { 31267c478bd9Sstevel@tonic-gate case IRE_HOST: 31277c478bd9Sstevel@tonic-gate ire->ire_mask = IP_HOST_MASK; 31287c478bd9Sstevel@tonic-gate ire->ire_masklen = IP_ABITS; 31297c478bd9Sstevel@tonic-gate if ((ire->ire_flags & RTF_SETSRC) == 0) 31307c478bd9Sstevel@tonic-gate ire->ire_src_addr = 0; 31317c478bd9Sstevel@tonic-gate break; 31327c478bd9Sstevel@tonic-gate case IRE_CACHE: 31337c478bd9Sstevel@tonic-gate case IRE_BROADCAST: 31347c478bd9Sstevel@tonic-gate case IRE_LOCAL: 31357c478bd9Sstevel@tonic-gate case IRE_LOOPBACK: 31367c478bd9Sstevel@tonic-gate ire->ire_mask = IP_HOST_MASK; 31377c478bd9Sstevel@tonic-gate ire->ire_masklen = IP_ABITS; 31387c478bd9Sstevel@tonic-gate break; 31397c478bd9Sstevel@tonic-gate case IRE_PREFIX: 31407c478bd9Sstevel@tonic-gate if ((ire->ire_flags & RTF_SETSRC) == 0) 31417c478bd9Sstevel@tonic-gate ire->ire_src_addr = 0; 31427c478bd9Sstevel@tonic-gate break; 31437c478bd9Sstevel@tonic-gate case IRE_DEFAULT: 31447c478bd9Sstevel@tonic-gate if ((ire->ire_flags & RTF_SETSRC) == 0) 31457c478bd9Sstevel@tonic-gate ire->ire_src_addr = 0; 31467c478bd9Sstevel@tonic-gate break; 31477c478bd9Sstevel@tonic-gate case IRE_IF_RESOLVER: 31487c478bd9Sstevel@tonic-gate case IRE_IF_NORESOLVER: 31497c478bd9Sstevel@tonic-gate break; 31507c478bd9Sstevel@tonic-gate default: 3151c793af95Ssangeeta ip0dbg(("ire_add_v4: ire %p has unrecognized IRE type (%d)\n", 3152c793af95Ssangeeta (void *)ire, ire->ire_type)); 31537c478bd9Sstevel@tonic-gate ire_delete(ire); 31547c478bd9Sstevel@tonic-gate *ire_p = NULL; 31557c478bd9Sstevel@tonic-gate return (EINVAL); 31567c478bd9Sstevel@tonic-gate } 31577c478bd9Sstevel@tonic-gate 31587c478bd9Sstevel@tonic-gate /* Make sure the address is properly masked. */ 31597c478bd9Sstevel@tonic-gate ire->ire_addr &= ire->ire_mask; 31607c478bd9Sstevel@tonic-gate 31617c478bd9Sstevel@tonic-gate /* 31627c478bd9Sstevel@tonic-gate * ip_newroute/ip_newroute_multi are unable to prevent the deletion 31637c478bd9Sstevel@tonic-gate * of the interface route while adding an IRE_CACHE for an on-link 31647c478bd9Sstevel@tonic-gate * destination in the IRE_IF_RESOLVER case, since the ire has to 31657c478bd9Sstevel@tonic-gate * go to ARP and return. We can't do a REFHOLD on the 31667c478bd9Sstevel@tonic-gate * associated interface ire for fear of ARP freeing the message. 31677c478bd9Sstevel@tonic-gate * Here we look up the interface ire in the forwarding table and 31687c478bd9Sstevel@tonic-gate * make sure that the interface route has not been deleted. 31697c478bd9Sstevel@tonic-gate */ 31707c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE && ire->ire_gateway_addr == 0 && 31717c478bd9Sstevel@tonic-gate ((ill_t *)ire->ire_stq->q_ptr)->ill_net_type == IRE_IF_RESOLVER) { 3172c793af95Ssangeeta 31737c478bd9Sstevel@tonic-gate ASSERT(ire->ire_max_fragp == NULL); 31747c478bd9Sstevel@tonic-gate if (CLASSD(ire->ire_addr) && !(ire->ire_flags & RTF_SETSRC)) { 31757c478bd9Sstevel@tonic-gate /* 31767c478bd9Sstevel@tonic-gate * The ihandle that we used in ip_newroute_multi 31777c478bd9Sstevel@tonic-gate * comes from the interface route corresponding 31787c478bd9Sstevel@tonic-gate * to ire_ipif. Lookup here to see if it exists 31797c478bd9Sstevel@tonic-gate * still. 31807c478bd9Sstevel@tonic-gate * If the ire has a source address assigned using 31817c478bd9Sstevel@tonic-gate * RTF_SETSRC, ire_ipif is the logical interface holding 31827c478bd9Sstevel@tonic-gate * this source address, so we can't use it to check for 31837c478bd9Sstevel@tonic-gate * the existence of the interface route. Instead we rely 31847c478bd9Sstevel@tonic-gate * on the brute force ihandle search in 31857c478bd9Sstevel@tonic-gate * ire_ihandle_lookup_onlink() below. 31867c478bd9Sstevel@tonic-gate */ 31877c478bd9Sstevel@tonic-gate pire = ipif_to_ire(ire->ire_ipif); 31887c478bd9Sstevel@tonic-gate if (pire == NULL) { 31897c478bd9Sstevel@tonic-gate ire_delete(ire); 31907c478bd9Sstevel@tonic-gate *ire_p = NULL; 31917c478bd9Sstevel@tonic-gate return (EINVAL); 31927c478bd9Sstevel@tonic-gate } else if (pire->ire_ihandle != ire->ire_ihandle) { 31937c478bd9Sstevel@tonic-gate ire_refrele(pire); 31947c478bd9Sstevel@tonic-gate ire_delete(ire); 31957c478bd9Sstevel@tonic-gate *ire_p = NULL; 31967c478bd9Sstevel@tonic-gate return (EINVAL); 31977c478bd9Sstevel@tonic-gate } 31987c478bd9Sstevel@tonic-gate } else { 31997c478bd9Sstevel@tonic-gate pire = ire_ihandle_lookup_onlink(ire); 32007c478bd9Sstevel@tonic-gate if (pire == NULL) { 32017c478bd9Sstevel@tonic-gate ire_delete(ire); 32027c478bd9Sstevel@tonic-gate *ire_p = NULL; 32037c478bd9Sstevel@tonic-gate return (EINVAL); 32047c478bd9Sstevel@tonic-gate } 32057c478bd9Sstevel@tonic-gate } 32067c478bd9Sstevel@tonic-gate /* Prevent pire from getting deleted */ 32077c478bd9Sstevel@tonic-gate IRB_REFHOLD(pire->ire_bucket); 32087c478bd9Sstevel@tonic-gate /* Has it been removed already ? */ 32097c478bd9Sstevel@tonic-gate if (pire->ire_marks & IRE_MARK_CONDEMNED) { 32107c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 32117c478bd9Sstevel@tonic-gate ire_refrele(pire); 32127c478bd9Sstevel@tonic-gate ire_delete(ire); 32137c478bd9Sstevel@tonic-gate *ire_p = NULL; 32147c478bd9Sstevel@tonic-gate return (EINVAL); 32157c478bd9Sstevel@tonic-gate } 32167c478bd9Sstevel@tonic-gate } else { 32177c478bd9Sstevel@tonic-gate ASSERT(ire->ire_max_fragp != NULL); 32187c478bd9Sstevel@tonic-gate } 32197c478bd9Sstevel@tonic-gate flags = (MATCH_IRE_MASK | MATCH_IRE_TYPE | MATCH_IRE_GW); 32207c478bd9Sstevel@tonic-gate 32217c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 32227c478bd9Sstevel@tonic-gate /* 32237c478bd9Sstevel@tonic-gate * We use MATCH_IRE_IPIF while adding IRE_CACHES only 32247c478bd9Sstevel@tonic-gate * for historic reasons and to maintain symmetry with 32257c478bd9Sstevel@tonic-gate * IPv6 code path. Historically this was used by 32267c478bd9Sstevel@tonic-gate * multicast code to create multiple IRE_CACHES on 32277c478bd9Sstevel@tonic-gate * a single ill with different ipifs. This was used 32287c478bd9Sstevel@tonic-gate * so that multicast packets leaving the node had the 32297c478bd9Sstevel@tonic-gate * right source address. This is no longer needed as 32307c478bd9Sstevel@tonic-gate * ip_wput initializes the address correctly. 32317c478bd9Sstevel@tonic-gate */ 32327c478bd9Sstevel@tonic-gate flags |= MATCH_IRE_IPIF; 32337c478bd9Sstevel@tonic-gate /* 32347c478bd9Sstevel@tonic-gate * If we are creating hidden ires, make sure we search on 32357c478bd9Sstevel@tonic-gate * this ill (MATCH_IRE_ILL) and a hidden ire, 32367c478bd9Sstevel@tonic-gate * while we are searching for duplicates below. Otherwise we 32377c478bd9Sstevel@tonic-gate * could potentially find an IRE on some other interface 32387c478bd9Sstevel@tonic-gate * and it may not be a IRE marked with IRE_MARK_HIDDEN. We 32397c478bd9Sstevel@tonic-gate * shouldn't do this as this will lead to an infinite loop 32407c478bd9Sstevel@tonic-gate * (if we get to ip_wput again) eventually we need an hidden 32417c478bd9Sstevel@tonic-gate * ire for this packet to go out. MATCH_IRE_ILL is explicitly 32427c478bd9Sstevel@tonic-gate * done below. 32437c478bd9Sstevel@tonic-gate */ 32447c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE && 32457c478bd9Sstevel@tonic-gate (ire->ire_marks & IRE_MARK_HIDDEN)) 32467c478bd9Sstevel@tonic-gate flags |= (MATCH_IRE_MARK_HIDDEN); 32477c478bd9Sstevel@tonic-gate } 3248c793af95Ssangeeta if ((ire->ire_type & IRE_CACHETABLE) == 0) { 3249c793af95Ssangeeta irb_ptr = ire_get_bucket(ire); 3250c793af95Ssangeeta need_refrele = B_TRUE; 3251c793af95Ssangeeta if (irb_ptr == NULL) { 3252c793af95Ssangeeta /* 3253c793af95Ssangeeta * This assumes that the ire has not added 3254c793af95Ssangeeta * a reference to the ipif. 3255c793af95Ssangeeta */ 3256c793af95Ssangeeta ire->ire_ipif = NULL; 3257c793af95Ssangeeta ire_delete(ire); 3258c793af95Ssangeeta if (pire != NULL) { 3259c793af95Ssangeeta IRB_REFRELE(pire->ire_bucket); 3260c793af95Ssangeeta ire_refrele(pire); 3261c793af95Ssangeeta } 3262c793af95Ssangeeta *ire_p = NULL; 3263c793af95Ssangeeta return (EINVAL); 3264c793af95Ssangeeta } 3265c793af95Ssangeeta } else { 3266*f4b3ec61Sdh155122 irb_ptr = &(ipst->ips_ip_cache_table[IRE_ADDR_HASH( 3267*f4b3ec61Sdh155122 ire->ire_addr, ipst->ips_ip_cache_table_size)]); 3268c793af95Ssangeeta } 32697c478bd9Sstevel@tonic-gate 32707c478bd9Sstevel@tonic-gate /* 32717c478bd9Sstevel@tonic-gate * Start the atomic add of the ire. Grab the ill locks, 32727c478bd9Sstevel@tonic-gate * ill_g_usesrc_lock and the bucket lock. Check for condemned 32737c478bd9Sstevel@tonic-gate * 32747c478bd9Sstevel@tonic-gate * If ipif or ill is changing ire_atomic_start() may queue the 32757c478bd9Sstevel@tonic-gate * request and return EINPROGRESS. 3276*f4b3ec61Sdh155122 * To avoid lock order problems, get the ndp4->ndp_g_lock. 32777c478bd9Sstevel@tonic-gate */ 3278*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ndp4->ndp_g_lock); 32797c478bd9Sstevel@tonic-gate error = ire_atomic_start(irb_ptr, ire, q, mp, func); 32807c478bd9Sstevel@tonic-gate if (error != 0) { 3281*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 32827c478bd9Sstevel@tonic-gate /* 32837c478bd9Sstevel@tonic-gate * We don't know whether it is a valid ipif or not. 32847c478bd9Sstevel@tonic-gate * So, set it to NULL. This assumes that the ire has not added 32857c478bd9Sstevel@tonic-gate * a reference to the ipif. 32867c478bd9Sstevel@tonic-gate */ 32877c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 32887c478bd9Sstevel@tonic-gate ire_delete(ire); 32897c478bd9Sstevel@tonic-gate if (pire != NULL) { 32907c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 32917c478bd9Sstevel@tonic-gate ire_refrele(pire); 32927c478bd9Sstevel@tonic-gate } 32937c478bd9Sstevel@tonic-gate *ire_p = NULL; 3294c793af95Ssangeeta if (need_refrele) 3295c793af95Ssangeeta IRB_REFRELE(irb_ptr); 32967c478bd9Sstevel@tonic-gate return (error); 32977c478bd9Sstevel@tonic-gate } 32987c478bd9Sstevel@tonic-gate /* 32997c478bd9Sstevel@tonic-gate * To avoid creating ires having stale values for the ire_max_frag 33007c478bd9Sstevel@tonic-gate * we get the latest value atomically here. For more details 33017c478bd9Sstevel@tonic-gate * see the block comment in ip_sioctl_mtu and in DL_NOTE_SDU_CHANGE 33027c478bd9Sstevel@tonic-gate * in ip_rput_dlpi_writer 33037c478bd9Sstevel@tonic-gate */ 33047c478bd9Sstevel@tonic-gate if (ire->ire_max_fragp == NULL) { 33057c478bd9Sstevel@tonic-gate if (CLASSD(ire->ire_addr)) 33067c478bd9Sstevel@tonic-gate ire->ire_max_frag = ire->ire_ipif->ipif_mtu; 33077c478bd9Sstevel@tonic-gate else 33087c478bd9Sstevel@tonic-gate ire->ire_max_frag = pire->ire_max_frag; 33097c478bd9Sstevel@tonic-gate } else { 33107c478bd9Sstevel@tonic-gate uint_t max_frag; 33117c478bd9Sstevel@tonic-gate 33127c478bd9Sstevel@tonic-gate max_frag = *ire->ire_max_fragp; 33137c478bd9Sstevel@tonic-gate ire->ire_max_fragp = NULL; 33147c478bd9Sstevel@tonic-gate ire->ire_max_frag = max_frag; 33157c478bd9Sstevel@tonic-gate } 33167c478bd9Sstevel@tonic-gate /* 33177c478bd9Sstevel@tonic-gate * Atomically check for duplicate and insert in the table. 33187c478bd9Sstevel@tonic-gate */ 33197c478bd9Sstevel@tonic-gate for (ire1 = irb_ptr->irb_ire; ire1 != NULL; ire1 = ire1->ire_next) { 33207c478bd9Sstevel@tonic-gate if (ire1->ire_marks & IRE_MARK_CONDEMNED) 33217c478bd9Sstevel@tonic-gate continue; 33227c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 33237c478bd9Sstevel@tonic-gate /* 33247c478bd9Sstevel@tonic-gate * We do MATCH_IRE_ILL implicitly here for IREs 33257c478bd9Sstevel@tonic-gate * with a non-null ire_ipif, including IRE_CACHEs. 33267c478bd9Sstevel@tonic-gate * As ire_ipif and ire_stq could point to two 33277c478bd9Sstevel@tonic-gate * different ills, we can't pass just ire_ipif to 33287c478bd9Sstevel@tonic-gate * ire_match_args and get a match on both ills. 33297c478bd9Sstevel@tonic-gate * This is just needed for duplicate checks here and 33307c478bd9Sstevel@tonic-gate * so we don't add an extra argument to 33317c478bd9Sstevel@tonic-gate * ire_match_args for this. Do it locally. 33327c478bd9Sstevel@tonic-gate * 33337c478bd9Sstevel@tonic-gate * NOTE : Currently there is no part of the code 33347c478bd9Sstevel@tonic-gate * that asks for both MATH_IRE_IPIF and MATCH_IRE_ILL 33357c478bd9Sstevel@tonic-gate * match for IRE_CACHEs. Thus we don't want to 33367c478bd9Sstevel@tonic-gate * extend the arguments to ire_match_args. 33377c478bd9Sstevel@tonic-gate */ 33387c478bd9Sstevel@tonic-gate if (ire1->ire_stq != ire->ire_stq) 33397c478bd9Sstevel@tonic-gate continue; 33407c478bd9Sstevel@tonic-gate /* 33417c478bd9Sstevel@tonic-gate * Multiroute IRE_CACHEs for a given destination can 33427c478bd9Sstevel@tonic-gate * have the same ire_ipif, typically if their source 33437c478bd9Sstevel@tonic-gate * address is forced using RTF_SETSRC, and the same 33447c478bd9Sstevel@tonic-gate * send-to queue. We differentiate them using the parent 33457c478bd9Sstevel@tonic-gate * handle. 33467c478bd9Sstevel@tonic-gate */ 33477c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_CACHE && 33487c478bd9Sstevel@tonic-gate (ire1->ire_flags & RTF_MULTIRT) && 33497c478bd9Sstevel@tonic-gate (ire->ire_flags & RTF_MULTIRT) && 33507c478bd9Sstevel@tonic-gate (ire1->ire_phandle != ire->ire_phandle)) 33517c478bd9Sstevel@tonic-gate continue; 33527c478bd9Sstevel@tonic-gate } 33537c478bd9Sstevel@tonic-gate if (ire1->ire_zoneid != ire->ire_zoneid) 33547c478bd9Sstevel@tonic-gate continue; 33557c478bd9Sstevel@tonic-gate if (ire_match_args(ire1, ire->ire_addr, ire->ire_mask, 33567c478bd9Sstevel@tonic-gate ire->ire_gateway_addr, ire->ire_type, ire->ire_ipif, 335745916cd2Sjpk ire->ire_zoneid, 0, NULL, flags)) { 33587c478bd9Sstevel@tonic-gate /* 33597c478bd9Sstevel@tonic-gate * Return the old ire after doing a REFHOLD. 33607c478bd9Sstevel@tonic-gate * As most of the callers continue to use the IRE 33617c478bd9Sstevel@tonic-gate * after adding, we return a held ire. This will 33627c478bd9Sstevel@tonic-gate * avoid a lookup in the caller again. If the callers 33637c478bd9Sstevel@tonic-gate * don't want to use it, they need to do a REFRELE. 33647c478bd9Sstevel@tonic-gate */ 33657c478bd9Sstevel@tonic-gate ip1dbg(("found dup ire existing %p new %p", 33667c478bd9Sstevel@tonic-gate (void *)ire1, (void *)ire)); 33677c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire1); 33687c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 3369*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 33707c478bd9Sstevel@tonic-gate ire_delete(ire); 33717c478bd9Sstevel@tonic-gate if (pire != NULL) { 33727c478bd9Sstevel@tonic-gate /* 33737c478bd9Sstevel@tonic-gate * Assert that it is not removed from the 33747c478bd9Sstevel@tonic-gate * list yet. 33757c478bd9Sstevel@tonic-gate */ 33767c478bd9Sstevel@tonic-gate ASSERT(pire->ire_ptpn != NULL); 33777c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 33787c478bd9Sstevel@tonic-gate ire_refrele(pire); 33797c478bd9Sstevel@tonic-gate } 33807c478bd9Sstevel@tonic-gate *ire_p = ire1; 3381c793af95Ssangeeta if (need_refrele) 3382c793af95Ssangeeta IRB_REFRELE(irb_ptr); 33837c478bd9Sstevel@tonic-gate return (0); 33847c478bd9Sstevel@tonic-gate } 33857c478bd9Sstevel@tonic-gate } 3386c793af95Ssangeeta if (ire->ire_type & IRE_CACHE) { 3387c793af95Ssangeeta ASSERT(ire->ire_stq != NULL); 3388c793af95Ssangeeta nce = ndp_lookup_v4(ire_to_ill(ire), 3389c793af95Ssangeeta ((ire->ire_gateway_addr != INADDR_ANY) ? 3390c793af95Ssangeeta &ire->ire_gateway_addr : &ire->ire_addr), 3391c793af95Ssangeeta B_TRUE); 3392c793af95Ssangeeta if (nce != NULL) 3393c793af95Ssangeeta mutex_enter(&nce->nce_lock); 3394c793af95Ssangeeta /* 3395c793af95Ssangeeta * if the nce is NCE_F_CONDEMNED, or if it is not ND_REACHABLE 3396c793af95Ssangeeta * and the caller has prohibited the addition of incomplete 3397c793af95Ssangeeta * ire's, we fail the add. Note that nce_state could be 3398c793af95Ssangeeta * something other than ND_REACHABLE if nce_reinit has just 3399c793af95Ssangeeta * kicked in and reset the nce. 3400c793af95Ssangeeta */ 3401c793af95Ssangeeta if ((nce == NULL) || 3402c793af95Ssangeeta (nce->nce_flags & NCE_F_CONDEMNED) || 3403c793af95Ssangeeta (!allow_unresolved && 3404107875b0Ssangeeta (nce->nce_state != ND_REACHABLE))) { 3405c793af95Ssangeeta if (nce != NULL) 3406c793af95Ssangeeta mutex_exit(&nce->nce_lock); 3407c793af95Ssangeeta ire_atomic_end(irb_ptr, ire); 3408*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 3409c793af95Ssangeeta if (nce != NULL) 3410c793af95Ssangeeta NCE_REFRELE(nce); 3411c793af95Ssangeeta DTRACE_PROBE1(ire__no__nce, ire_t *, ire); 3412c793af95Ssangeeta ire_delete(ire); 3413c793af95Ssangeeta if (pire != NULL) { 3414c793af95Ssangeeta IRB_REFRELE(pire->ire_bucket); 3415c793af95Ssangeeta ire_refrele(pire); 3416c793af95Ssangeeta } 3417c793af95Ssangeeta *ire_p = NULL; 3418c793af95Ssangeeta if (need_refrele) 3419c793af95Ssangeeta IRB_REFRELE(irb_ptr); 3420c793af95Ssangeeta return (EINVAL); 3421c793af95Ssangeeta } else { 3422c793af95Ssangeeta ire->ire_nce = nce; 3423c793af95Ssangeeta mutex_exit(&nce->nce_lock); 3424c793af95Ssangeeta /* 3425c793af95Ssangeeta * We are associating this nce to the ire, so 3426c793af95Ssangeeta * change the nce ref taken in ndp_lookup_v4() from 3427c793af95Ssangeeta * NCE_REFHOLD to NCE_REFHOLD_NOTR 3428c793af95Ssangeeta */ 3429c793af95Ssangeeta NCE_REFHOLD_TO_REFHOLD_NOTR(ire->ire_nce); 3430c793af95Ssangeeta } 3431c793af95Ssangeeta } 34327c478bd9Sstevel@tonic-gate /* 34337c478bd9Sstevel@tonic-gate * Make it easy for ip_wput_ire() to hit multiple broadcast ires by 34347c478bd9Sstevel@tonic-gate * grouping identical addresses together on the hash chain. We also 34357c478bd9Sstevel@tonic-gate * don't want to send multiple copies out if there are two ills part 34367c478bd9Sstevel@tonic-gate * of the same group. Thus we group the ires with same addr and same 34377c478bd9Sstevel@tonic-gate * ill group together so that ip_wput_ire can easily skip all the 34387c478bd9Sstevel@tonic-gate * ires with same addr and same group after sending the first copy. 34397c478bd9Sstevel@tonic-gate * We do this only for IRE_BROADCASTs as ip_wput_ire is currently 34407c478bd9Sstevel@tonic-gate * interested in such groupings only for broadcasts. 34417c478bd9Sstevel@tonic-gate * 34427c478bd9Sstevel@tonic-gate * NOTE : If the interfaces are brought up first and then grouped, 34437c478bd9Sstevel@tonic-gate * illgrp_insert will handle it. We come here when the interfaces 34447c478bd9Sstevel@tonic-gate * are already in group and we are bringing them UP. 34457c478bd9Sstevel@tonic-gate * 34467c478bd9Sstevel@tonic-gate * Find the first entry that matches ire_addr. *irep will be null 34477c478bd9Sstevel@tonic-gate * if no match. 34487c478bd9Sstevel@tonic-gate */ 34497c478bd9Sstevel@tonic-gate irep = (ire_t **)irb_ptr; 34507c478bd9Sstevel@tonic-gate while ((ire1 = *irep) != NULL && ire->ire_addr != ire1->ire_addr) 34517c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 34527c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_BROADCAST && *irep != NULL) { 34537c478bd9Sstevel@tonic-gate /* 34547c478bd9Sstevel@tonic-gate * We found some ire (i.e *irep) with a matching addr. We 34557c478bd9Sstevel@tonic-gate * want to group ires with same addr and same ill group 34567c478bd9Sstevel@tonic-gate * together. 34577c478bd9Sstevel@tonic-gate * 34587c478bd9Sstevel@tonic-gate * First get to the entry that matches our address and 34597c478bd9Sstevel@tonic-gate * ill group i.e stop as soon as we find the first ire 34607c478bd9Sstevel@tonic-gate * matching the ill group and address. If there is only 34617c478bd9Sstevel@tonic-gate * an address match, we should walk and look for some 34627c478bd9Sstevel@tonic-gate * group match. These are some of the possible scenarios : 34637c478bd9Sstevel@tonic-gate * 34647c478bd9Sstevel@tonic-gate * 1) There are no groups at all i.e all ire's ill_group 34657c478bd9Sstevel@tonic-gate * are NULL. In that case we will essentially group 34667c478bd9Sstevel@tonic-gate * all the ires with the same addr together. Same as 34677c478bd9Sstevel@tonic-gate * the "else" block of this "if". 34687c478bd9Sstevel@tonic-gate * 34697c478bd9Sstevel@tonic-gate * 2) There are some groups and this ire's ill_group is 34707c478bd9Sstevel@tonic-gate * NULL. In this case, we will first find the group 34717c478bd9Sstevel@tonic-gate * that matches the address and a NULL group. Then 34727c478bd9Sstevel@tonic-gate * we will insert the ire at the end of that group. 34737c478bd9Sstevel@tonic-gate * 34747c478bd9Sstevel@tonic-gate * 3) There are some groups and this ires's ill_group is 34757c478bd9Sstevel@tonic-gate * non-NULL. In this case we will first find the group 34767c478bd9Sstevel@tonic-gate * that matches the address and the ill_group. Then 34777c478bd9Sstevel@tonic-gate * we will insert the ire at the end of that group. 34787c478bd9Sstevel@tonic-gate */ 34797c478bd9Sstevel@tonic-gate /* LINTED : constant in conditional context */ 34807c478bd9Sstevel@tonic-gate while (1) { 34817c478bd9Sstevel@tonic-gate ire1 = *irep; 34827c478bd9Sstevel@tonic-gate if ((ire1->ire_next == NULL) || 34837c478bd9Sstevel@tonic-gate (ire1->ire_next->ire_addr != ire->ire_addr) || 34847c478bd9Sstevel@tonic-gate (ire1->ire_type != IRE_BROADCAST) || 34857c478bd9Sstevel@tonic-gate (ire1->ire_ipif->ipif_ill->ill_group == 34867c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ill->ill_group)) 34877c478bd9Sstevel@tonic-gate break; 34887c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 34897c478bd9Sstevel@tonic-gate } 34907c478bd9Sstevel@tonic-gate ASSERT(*irep != NULL); 34917c478bd9Sstevel@tonic-gate irep = &((*irep)->ire_next); 34927c478bd9Sstevel@tonic-gate 34937c478bd9Sstevel@tonic-gate /* 34947c478bd9Sstevel@tonic-gate * Either we have hit the end of the list or the address 34957c478bd9Sstevel@tonic-gate * did not match or the group *matched*. If we found 34967c478bd9Sstevel@tonic-gate * a match on the group, skip to the end of the group. 34977c478bd9Sstevel@tonic-gate */ 34987c478bd9Sstevel@tonic-gate while (*irep != NULL) { 34997c478bd9Sstevel@tonic-gate ire1 = *irep; 35007c478bd9Sstevel@tonic-gate if ((ire1->ire_addr != ire->ire_addr) || 35017c478bd9Sstevel@tonic-gate (ire1->ire_type != IRE_BROADCAST) || 35027c478bd9Sstevel@tonic-gate (ire1->ire_ipif->ipif_ill->ill_group != 35037c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ill->ill_group)) 35047c478bd9Sstevel@tonic-gate break; 35057c478bd9Sstevel@tonic-gate if (ire1->ire_ipif->ipif_ill->ill_group == NULL && 35067c478bd9Sstevel@tonic-gate ire1->ire_ipif == ire->ire_ipif) { 35077c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 35087c478bd9Sstevel@tonic-gate break; 35097c478bd9Sstevel@tonic-gate } 35107c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 35117c478bd9Sstevel@tonic-gate } 35127c478bd9Sstevel@tonic-gate } else if (*irep != NULL) { 35137c478bd9Sstevel@tonic-gate /* 35147c478bd9Sstevel@tonic-gate * Find the last ire which matches ire_addr. 35157c478bd9Sstevel@tonic-gate * Needed to do tail insertion among entries with the same 35167c478bd9Sstevel@tonic-gate * ire_addr. 35177c478bd9Sstevel@tonic-gate */ 35187c478bd9Sstevel@tonic-gate while (ire->ire_addr == ire1->ire_addr) { 35197c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 35207c478bd9Sstevel@tonic-gate ire1 = *irep; 35217c478bd9Sstevel@tonic-gate if (ire1 == NULL) 35227c478bd9Sstevel@tonic-gate break; 35237c478bd9Sstevel@tonic-gate } 35247c478bd9Sstevel@tonic-gate } 35257c478bd9Sstevel@tonic-gate 35267c478bd9Sstevel@tonic-gate /* Insert at *irep */ 35277c478bd9Sstevel@tonic-gate ire1 = *irep; 35287c478bd9Sstevel@tonic-gate if (ire1 != NULL) 35297c478bd9Sstevel@tonic-gate ire1->ire_ptpn = &ire->ire_next; 35307c478bd9Sstevel@tonic-gate ire->ire_next = ire1; 35317c478bd9Sstevel@tonic-gate /* Link the new one in. */ 35327c478bd9Sstevel@tonic-gate ire->ire_ptpn = irep; 35337c478bd9Sstevel@tonic-gate 35347c478bd9Sstevel@tonic-gate /* 35357c478bd9Sstevel@tonic-gate * ire_walk routines de-reference ire_next without holding 35367c478bd9Sstevel@tonic-gate * a lock. Before we point to the new ire, we want to make 35377c478bd9Sstevel@tonic-gate * sure the store that sets the ire_next of the new ire 35387c478bd9Sstevel@tonic-gate * reaches global visibility, so that ire_walk routines 35397c478bd9Sstevel@tonic-gate * don't see a truncated list of ires i.e if the ire_next 35407c478bd9Sstevel@tonic-gate * of the new ire gets set after we do "*irep = ire" due 35417c478bd9Sstevel@tonic-gate * to re-ordering, the ire_walk thread will see a NULL 35427c478bd9Sstevel@tonic-gate * once it accesses the ire_next of the new ire. 35437c478bd9Sstevel@tonic-gate * membar_producer() makes sure that the following store 35447c478bd9Sstevel@tonic-gate * happens *after* all of the above stores. 35457c478bd9Sstevel@tonic-gate */ 35467c478bd9Sstevel@tonic-gate membar_producer(); 35477c478bd9Sstevel@tonic-gate *irep = ire; 35487c478bd9Sstevel@tonic-gate ire->ire_bucket = irb_ptr; 35497c478bd9Sstevel@tonic-gate /* 35507c478bd9Sstevel@tonic-gate * We return a bumped up IRE above. Keep it symmetrical 35517c478bd9Sstevel@tonic-gate * so that the callers will always have to release. This 35527c478bd9Sstevel@tonic-gate * helps the callers of this function because they continue 35537c478bd9Sstevel@tonic-gate * to use the IRE after adding and hence they don't have to 35547c478bd9Sstevel@tonic-gate * lookup again after we return the IRE. 35557c478bd9Sstevel@tonic-gate * 35567c478bd9Sstevel@tonic-gate * NOTE : We don't have to use atomics as this is appearing 35577c478bd9Sstevel@tonic-gate * in the list for the first time and no one else can bump 35587c478bd9Sstevel@tonic-gate * up the reference count on this yet. 35597c478bd9Sstevel@tonic-gate */ 35607c478bd9Sstevel@tonic-gate IRE_REFHOLD_LOCKED(ire); 3561*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_inserted); 3562c793af95Ssangeeta 35637c478bd9Sstevel@tonic-gate irb_ptr->irb_ire_cnt++; 3564c793af95Ssangeeta if (irb_ptr->irb_marks & IRB_MARK_FTABLE) 3565c793af95Ssangeeta irb_ptr->irb_nire++; 3566c793af95Ssangeeta 35677c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_TEMPORARY) 35687c478bd9Sstevel@tonic-gate irb_ptr->irb_tmp_ire_cnt++; 35697c478bd9Sstevel@tonic-gate 35707c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 35717c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ire_cnt++; 35727c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) { 35737c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 35747c478bd9Sstevel@tonic-gate stq_ill->ill_ire_cnt++; 35757c478bd9Sstevel@tonic-gate } 35767c478bd9Sstevel@tonic-gate } else { 35777c478bd9Sstevel@tonic-gate ASSERT(ire->ire_stq == NULL); 35787c478bd9Sstevel@tonic-gate } 35797c478bd9Sstevel@tonic-gate 35807c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 3581*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ndp4->ndp_g_lock); 35827c478bd9Sstevel@tonic-gate 35837c478bd9Sstevel@tonic-gate if (pire != NULL) { 35847c478bd9Sstevel@tonic-gate /* Assert that it is not removed from the list yet */ 35857c478bd9Sstevel@tonic-gate ASSERT(pire->ire_ptpn != NULL); 35867c478bd9Sstevel@tonic-gate IRB_REFRELE(pire->ire_bucket); 35877c478bd9Sstevel@tonic-gate ire_refrele(pire); 35887c478bd9Sstevel@tonic-gate } 35897c478bd9Sstevel@tonic-gate 35907c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) { 35917c478bd9Sstevel@tonic-gate /* 3592c793af95Ssangeeta * For ire's with host mask see if there is an entry 35937c478bd9Sstevel@tonic-gate * in the cache. If there is one flush the whole cache as 35947c478bd9Sstevel@tonic-gate * there might be multiple entries due to RTF_MULTIRT (CGTP). 35957c478bd9Sstevel@tonic-gate * If no entry is found than there is no need to flush the 35967c478bd9Sstevel@tonic-gate * cache. 35977c478bd9Sstevel@tonic-gate */ 35987c478bd9Sstevel@tonic-gate if (ire->ire_mask == IP_HOST_MASK) { 35997c478bd9Sstevel@tonic-gate ire_t *lire; 36007c478bd9Sstevel@tonic-gate lire = ire_ctable_lookup(ire->ire_addr, NULL, IRE_CACHE, 3601*f4b3ec61Sdh155122 NULL, ALL_ZONES, NULL, MATCH_IRE_TYPE, ipst); 36027c478bd9Sstevel@tonic-gate if (lire != NULL) { 36037c478bd9Sstevel@tonic-gate ire_refrele(lire); 36047c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire, IRE_FLUSH_ADD); 36057c478bd9Sstevel@tonic-gate } 36067c478bd9Sstevel@tonic-gate } else { 36077c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire, IRE_FLUSH_ADD); 36087c478bd9Sstevel@tonic-gate } 36097c478bd9Sstevel@tonic-gate } 36107c478bd9Sstevel@tonic-gate /* 36117c478bd9Sstevel@tonic-gate * We had to delay the fast path probe until the ire is inserted 36127c478bd9Sstevel@tonic-gate * in the list. Otherwise the fast path ack won't find the ire in 36137c478bd9Sstevel@tonic-gate * the table. 36147c478bd9Sstevel@tonic-gate */ 3615516bda92Ssowmini if (ire->ire_type == IRE_CACHE || 3616516bda92Ssowmini (ire->ire_type == IRE_BROADCAST && ire->ire_stq != NULL)) { 3617516bda92Ssowmini ASSERT(ire->ire_nce != NULL); 3618516bda92Ssowmini nce_fastpath(ire->ire_nce); 3619516bda92Ssowmini } 36207c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) 36217c478bd9Sstevel@tonic-gate ASSERT(!MUTEX_HELD(&ire->ire_ipif->ipif_ill->ill_lock)); 36227c478bd9Sstevel@tonic-gate *ire_p = ire; 3623c793af95Ssangeeta if (need_refrele) { 3624c793af95Ssangeeta IRB_REFRELE(irb_ptr); 3625c793af95Ssangeeta } 36267c478bd9Sstevel@tonic-gate return (0); 36277c478bd9Sstevel@tonic-gate } 36287c478bd9Sstevel@tonic-gate 36297c478bd9Sstevel@tonic-gate /* 36307c478bd9Sstevel@tonic-gate * IRB_REFRELE is the only caller of the function. ire_unlink calls to 36317c478bd9Sstevel@tonic-gate * do the final cleanup for this ire. 36327c478bd9Sstevel@tonic-gate */ 36337c478bd9Sstevel@tonic-gate void 36347c478bd9Sstevel@tonic-gate ire_cleanup(ire_t *ire) 36357c478bd9Sstevel@tonic-gate { 36367c478bd9Sstevel@tonic-gate ire_t *ire_next; 3637*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 36387c478bd9Sstevel@tonic-gate 36397c478bd9Sstevel@tonic-gate ASSERT(ire != NULL); 36407c478bd9Sstevel@tonic-gate 36417c478bd9Sstevel@tonic-gate while (ire != NULL) { 36427c478bd9Sstevel@tonic-gate ire_next = ire->ire_next; 36437c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV4_VERSION) { 36447c478bd9Sstevel@tonic-gate ire_delete_v4(ire); 3645*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, 3646*f4b3ec61Sdh155122 ire_stats_deleted); 36477c478bd9Sstevel@tonic-gate } else { 36487c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV6_VERSION); 36497c478bd9Sstevel@tonic-gate ire_delete_v6(ire); 3650*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v6, 3651*f4b3ec61Sdh155122 ire_stats_deleted); 36527c478bd9Sstevel@tonic-gate } 36537c478bd9Sstevel@tonic-gate /* 36547c478bd9Sstevel@tonic-gate * Now it's really out of the list. Before doing the 36557c478bd9Sstevel@tonic-gate * REFRELE, set ire_next to NULL as ire_inactive asserts 36567c478bd9Sstevel@tonic-gate * so. 36577c478bd9Sstevel@tonic-gate */ 36587c478bd9Sstevel@tonic-gate ire->ire_next = NULL; 36597c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 36607c478bd9Sstevel@tonic-gate ire = ire_next; 36617c478bd9Sstevel@tonic-gate } 36627c478bd9Sstevel@tonic-gate } 36637c478bd9Sstevel@tonic-gate 36647c478bd9Sstevel@tonic-gate /* 36657c478bd9Sstevel@tonic-gate * IRB_REFRELE is the only caller of the function. It calls to unlink 36667c478bd9Sstevel@tonic-gate * all the CONDEMNED ires from this bucket. 36677c478bd9Sstevel@tonic-gate */ 36687c478bd9Sstevel@tonic-gate ire_t * 36697c478bd9Sstevel@tonic-gate ire_unlink(irb_t *irb) 36707c478bd9Sstevel@tonic-gate { 36717c478bd9Sstevel@tonic-gate ire_t *ire; 36727c478bd9Sstevel@tonic-gate ire_t *ire1; 36737c478bd9Sstevel@tonic-gate ire_t **ptpn; 36747c478bd9Sstevel@tonic-gate ire_t *ire_list = NULL; 36757c478bd9Sstevel@tonic-gate 36767c478bd9Sstevel@tonic-gate ASSERT(RW_WRITE_HELD(&irb->irb_lock)); 3677c793af95Ssangeeta ASSERT(((irb->irb_marks & IRB_MARK_FTABLE) && irb->irb_refcnt == 1) || 3678c793af95Ssangeeta (irb->irb_refcnt == 0)); 3679c793af95Ssangeeta ASSERT(irb->irb_marks & IRB_MARK_CONDEMNED); 36807c478bd9Sstevel@tonic-gate ASSERT(irb->irb_ire != NULL); 36817c478bd9Sstevel@tonic-gate 36827c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; ire = ire1) { 3683*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 3684*f4b3ec61Sdh155122 36857c478bd9Sstevel@tonic-gate ire1 = ire->ire_next; 36867c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) { 36877c478bd9Sstevel@tonic-gate ptpn = ire->ire_ptpn; 36887c478bd9Sstevel@tonic-gate ire1 = ire->ire_next; 36897c478bd9Sstevel@tonic-gate if (ire1) 36907c478bd9Sstevel@tonic-gate ire1->ire_ptpn = ptpn; 36917c478bd9Sstevel@tonic-gate *ptpn = ire1; 36927c478bd9Sstevel@tonic-gate ire->ire_ptpn = NULL; 36937c478bd9Sstevel@tonic-gate ire->ire_next = NULL; 36947c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT) { 36957c478bd9Sstevel@tonic-gate /* 36967c478bd9Sstevel@tonic-gate * IRE is out of the list. We need to adjust 36977c478bd9Sstevel@tonic-gate * the accounting before the caller drops 36987c478bd9Sstevel@tonic-gate * the lock. 36997c478bd9Sstevel@tonic-gate */ 37007c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 3701*f4b3ec61Sdh155122 ASSERT(ipst-> 3702*f4b3ec61Sdh155122 ips_ipv6_ire_default_count != 3703*f4b3ec61Sdh155122 0); 3704*f4b3ec61Sdh155122 ipst->ips_ipv6_ire_default_count--; 37057c478bd9Sstevel@tonic-gate } 37067c478bd9Sstevel@tonic-gate } 37077c478bd9Sstevel@tonic-gate /* 37087c478bd9Sstevel@tonic-gate * We need to call ire_delete_v4 or ire_delete_v6 37097c478bd9Sstevel@tonic-gate * to clean up the cache or the redirects pointing at 37107c478bd9Sstevel@tonic-gate * the default gateway. We need to drop the lock 37117c478bd9Sstevel@tonic-gate * as ire_flush_cache/ire_delete_host_redircts require 37127c478bd9Sstevel@tonic-gate * so. But we can't drop the lock, as ire_unlink needs 37137c478bd9Sstevel@tonic-gate * to atomically remove the ires from the list. 37147c478bd9Sstevel@tonic-gate * So, create a temporary list of CONDEMNED ires 37157c478bd9Sstevel@tonic-gate * for doing ire_delete_v4/ire_delete_v6 operations 37167c478bd9Sstevel@tonic-gate * later on. 37177c478bd9Sstevel@tonic-gate */ 37187c478bd9Sstevel@tonic-gate ire->ire_next = ire_list; 37197c478bd9Sstevel@tonic-gate ire_list = ire; 37207c478bd9Sstevel@tonic-gate } 37217c478bd9Sstevel@tonic-gate } 3722c793af95Ssangeeta irb->irb_marks &= ~IRB_MARK_CONDEMNED; 37237c478bd9Sstevel@tonic-gate return (ire_list); 37247c478bd9Sstevel@tonic-gate } 37257c478bd9Sstevel@tonic-gate 37267c478bd9Sstevel@tonic-gate /* 37277c478bd9Sstevel@tonic-gate * Delete all the cache entries with this 'addr'. When IP gets a gratuitous 3728c793af95Ssangeeta * ARP message on any of its interface queue, it scans the nce table and 3729c793af95Ssangeeta * deletes and calls ndp_delete() for the appropriate nce. This action 3730c793af95Ssangeeta * also deletes all the neighbor/ire cache entries for that address. 3731c793af95Ssangeeta * This function is called from ip_arp_news in ip.c and also for 3732c793af95Ssangeeta * ARP ioctl processing in ip_if.c. ip_ire_clookup_and_delete returns 3733c793af95Ssangeeta * true if it finds a nce entry which is used by ip_arp_news to determine if 3734c793af95Ssangeeta * it needs to do an ire_walk_v4. The return value is also used for the 3735c793af95Ssangeeta * same purpose by ARP IOCTL processing * in ip_if.c when deleting 3736c793af95Ssangeeta * ARP entries. For SIOC*IFARP ioctls in addition to the address, 3737c793af95Ssangeeta * ip_if->ipif_ill also needs to be matched. 37387c478bd9Sstevel@tonic-gate */ 37397c478bd9Sstevel@tonic-gate boolean_t 3740*f4b3ec61Sdh155122 ip_ire_clookup_and_delete(ipaddr_t addr, ipif_t *ipif, ip_stack_t *ipst) 37417c478bd9Sstevel@tonic-gate { 37427c478bd9Sstevel@tonic-gate ill_t *ill; 3743c793af95Ssangeeta nce_t *nce; 37447c478bd9Sstevel@tonic-gate 3745c793af95Ssangeeta ill = (ipif ? ipif->ipif_ill : NULL); 37467c478bd9Sstevel@tonic-gate 3747c793af95Ssangeeta if (ill != NULL) { 3748c793af95Ssangeeta /* 3749c793af95Ssangeeta * clean up the nce (and any relevant ire's) that matches 3750c793af95Ssangeeta * on addr and ill. 3751c793af95Ssangeeta */ 3752c793af95Ssangeeta nce = ndp_lookup_v4(ill, &addr, B_FALSE); 3753c793af95Ssangeeta if (nce != NULL) { 3754c793af95Ssangeeta ndp_delete(nce); 3755c793af95Ssangeeta return (B_TRUE); 37567c478bd9Sstevel@tonic-gate } 3757c793af95Ssangeeta } else { 3758c793af95Ssangeeta /* 3759c793af95Ssangeeta * ill is wildcard. clean up all nce's and 3760c793af95Ssangeeta * ire's that match on addr 3761c793af95Ssangeeta */ 3762c793af95Ssangeeta nce_clookup_t cl; 37637c478bd9Sstevel@tonic-gate 3764c793af95Ssangeeta cl.ncecl_addr = addr; 3765c793af95Ssangeeta cl.ncecl_found = B_FALSE; 37667c478bd9Sstevel@tonic-gate 3767*f4b3ec61Sdh155122 ndp_walk_common(ipst->ips_ndp4, NULL, 3768c793af95Ssangeeta (pfi_t)ip_nce_clookup_and_delete, (uchar_t *)&cl, B_TRUE); 3769c793af95Ssangeeta 3770c793af95Ssangeeta /* 3771c793af95Ssangeeta * ncecl_found would be set by ip_nce_clookup_and_delete if 3772c793af95Ssangeeta * we found a matching nce. 3773c793af95Ssangeeta */ 3774c793af95Ssangeeta return (cl.ncecl_found); 3775c793af95Ssangeeta } 3776c793af95Ssangeeta return (B_FALSE); 3777c793af95Ssangeeta 3778c793af95Ssangeeta } 3779c793af95Ssangeeta 3780c793af95Ssangeeta /* Delete the supplied nce if its nce_addr matches the supplied address */ 3781c793af95Ssangeeta static void 3782c793af95Ssangeeta ip_nce_clookup_and_delete(nce_t *nce, void *arg) 3783c793af95Ssangeeta { 3784c793af95Ssangeeta nce_clookup_t *cl = (nce_clookup_t *)arg; 3785c793af95Ssangeeta ipaddr_t nce_addr; 3786c793af95Ssangeeta 3787c793af95Ssangeeta IN6_V4MAPPED_TO_IPADDR(&nce->nce_addr, nce_addr); 3788c793af95Ssangeeta if (nce_addr == cl->ncecl_addr) { 3789c793af95Ssangeeta cl->ncecl_found = B_TRUE; 3790c793af95Ssangeeta /* clean up the nce (and any relevant ire's) */ 3791c793af95Ssangeeta ndp_delete(nce); 3792c793af95Ssangeeta } 3793c793af95Ssangeeta } 3794c793af95Ssangeeta 3795c793af95Ssangeeta /* 3796c793af95Ssangeeta * Clean up the radix node for this ire. Must be called by IRB_REFRELE 3797c793af95Ssangeeta * when there are no ire's left in the bucket. Returns TRUE if the bucket 3798c793af95Ssangeeta * is deleted and freed. 3799c793af95Ssangeeta */ 3800c793af95Ssangeeta boolean_t 3801c793af95Ssangeeta irb_inactive(irb_t *irb) 3802c793af95Ssangeeta { 3803c793af95Ssangeeta struct rt_entry *rt; 3804c793af95Ssangeeta struct radix_node *rn; 3805*f4b3ec61Sdh155122 ip_stack_t *ipst = irb->irb_ipst; 3806*f4b3ec61Sdh155122 3807*f4b3ec61Sdh155122 ASSERT(irb->irb_ipst != NULL); 3808c793af95Ssangeeta 3809c793af95Ssangeeta rt = IRB2RT(irb); 3810c793af95Ssangeeta rn = (struct radix_node *)rt; 3811c793af95Ssangeeta 3812c793af95Ssangeeta /* first remove it from the radix tree. */ 3813*f4b3ec61Sdh155122 RADIX_NODE_HEAD_WLOCK(ipst->ips_ip_ftable); 3814c793af95Ssangeeta rw_enter(&irb->irb_lock, RW_WRITER); 3815c793af95Ssangeeta if (irb->irb_refcnt == 1 && irb->irb_nire == 0) { 3816*f4b3ec61Sdh155122 rn = ipst->ips_ip_ftable->rnh_deladdr(rn->rn_key, rn->rn_mask, 3817*f4b3ec61Sdh155122 ipst->ips_ip_ftable); 3818c793af95Ssangeeta DTRACE_PROBE1(irb__free, rt_t *, rt); 3819c793af95Ssangeeta ASSERT((void *)rn == (void *)rt); 3820c793af95Ssangeeta Free(rt, rt_entry_cache); 3821c793af95Ssangeeta /* irb_lock is freed */ 3822*f4b3ec61Sdh155122 RADIX_NODE_HEAD_UNLOCK(ipst->ips_ip_ftable); 3823c793af95Ssangeeta return (B_TRUE); 3824c793af95Ssangeeta } 3825c793af95Ssangeeta rw_exit(&irb->irb_lock); 3826*f4b3ec61Sdh155122 RADIX_NODE_HEAD_UNLOCK(ipst->ips_ip_ftable); 3827c793af95Ssangeeta return (B_FALSE); 38287c478bd9Sstevel@tonic-gate } 38297c478bd9Sstevel@tonic-gate 38307c478bd9Sstevel@tonic-gate /* 38317c478bd9Sstevel@tonic-gate * Delete the specified IRE. 38327c478bd9Sstevel@tonic-gate */ 38337c478bd9Sstevel@tonic-gate void 38347c478bd9Sstevel@tonic-gate ire_delete(ire_t *ire) 38357c478bd9Sstevel@tonic-gate { 38367c478bd9Sstevel@tonic-gate ire_t *ire1; 38377c478bd9Sstevel@tonic-gate ire_t **ptpn; 38387c478bd9Sstevel@tonic-gate irb_t *irb; 3839*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 38407c478bd9Sstevel@tonic-gate 3841c793af95Ssangeeta if ((irb = ire->ire_bucket) == NULL) { 38427c478bd9Sstevel@tonic-gate /* 38437c478bd9Sstevel@tonic-gate * It was never inserted in the list. Should call REFRELE 38447c478bd9Sstevel@tonic-gate * to free this IRE. 38457c478bd9Sstevel@tonic-gate */ 38467c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 38477c478bd9Sstevel@tonic-gate return; 38487c478bd9Sstevel@tonic-gate } 38497c478bd9Sstevel@tonic-gate 38507c478bd9Sstevel@tonic-gate rw_enter(&irb->irb_lock, RW_WRITER); 38517c478bd9Sstevel@tonic-gate 3852c793af95Ssangeeta if (irb->irb_rr_origin == ire) { 3853c793af95Ssangeeta irb->irb_rr_origin = NULL; 3854c793af95Ssangeeta } 3855c793af95Ssangeeta 38567c478bd9Sstevel@tonic-gate /* 38577c478bd9Sstevel@tonic-gate * In case of V4 we might still be waiting for fastpath ack. 38587c478bd9Sstevel@tonic-gate */ 3859516bda92Ssowmini if (ire->ire_ipversion == IPV4_VERSION && 3860516bda92Ssowmini (ire->ire_type == IRE_CACHE || 3861516bda92Ssowmini (ire->ire_type == IRE_BROADCAST && ire->ire_stq != NULL))) { 3862516bda92Ssowmini ASSERT(ire->ire_nce != NULL); 3863516bda92Ssowmini nce_fastpath_list_delete(ire->ire_nce); 38647c478bd9Sstevel@tonic-gate } 38657c478bd9Sstevel@tonic-gate 38667c478bd9Sstevel@tonic-gate if (ire->ire_ptpn == NULL) { 38677c478bd9Sstevel@tonic-gate /* 38687c478bd9Sstevel@tonic-gate * Some other thread has removed us from the list. 38697c478bd9Sstevel@tonic-gate * It should have done the REFRELE for us. 38707c478bd9Sstevel@tonic-gate */ 38717c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 38727c478bd9Sstevel@tonic-gate return; 38737c478bd9Sstevel@tonic-gate } 38747c478bd9Sstevel@tonic-gate 38757c478bd9Sstevel@tonic-gate if (irb->irb_refcnt != 0) { 38767c478bd9Sstevel@tonic-gate /* 38777c478bd9Sstevel@tonic-gate * The last thread to leave this bucket will 38787c478bd9Sstevel@tonic-gate * delete this ire. 38797c478bd9Sstevel@tonic-gate */ 38807c478bd9Sstevel@tonic-gate if (!(ire->ire_marks & IRE_MARK_CONDEMNED)) { 38817c478bd9Sstevel@tonic-gate irb->irb_ire_cnt--; 38827c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_TEMPORARY) 38837c478bd9Sstevel@tonic-gate irb->irb_tmp_ire_cnt--; 38847c478bd9Sstevel@tonic-gate ire->ire_marks |= IRE_MARK_CONDEMNED; 38857c478bd9Sstevel@tonic-gate } 3886c793af95Ssangeeta irb->irb_marks |= IRB_MARK_CONDEMNED; 38877c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 38887c478bd9Sstevel@tonic-gate return; 38897c478bd9Sstevel@tonic-gate } 38907c478bd9Sstevel@tonic-gate 38917c478bd9Sstevel@tonic-gate /* 38927c478bd9Sstevel@tonic-gate * Normally to delete an ire, we walk the bucket. While we 38937c478bd9Sstevel@tonic-gate * walk the bucket, we normally bump up irb_refcnt and hence 38947c478bd9Sstevel@tonic-gate * we return from above where we mark CONDEMNED and the ire 38957c478bd9Sstevel@tonic-gate * gets deleted from ire_unlink. This case is where somebody 38967c478bd9Sstevel@tonic-gate * knows the ire e.g by doing a lookup, and wants to delete the 38977c478bd9Sstevel@tonic-gate * IRE. irb_refcnt would be 0 in this case if nobody is walking 38987c478bd9Sstevel@tonic-gate * the bucket. 38997c478bd9Sstevel@tonic-gate */ 39007c478bd9Sstevel@tonic-gate ptpn = ire->ire_ptpn; 39017c478bd9Sstevel@tonic-gate ire1 = ire->ire_next; 39027c478bd9Sstevel@tonic-gate if (ire1 != NULL) 39037c478bd9Sstevel@tonic-gate ire1->ire_ptpn = ptpn; 39047c478bd9Sstevel@tonic-gate ASSERT(ptpn != NULL); 39057c478bd9Sstevel@tonic-gate *ptpn = ire1; 39067c478bd9Sstevel@tonic-gate ire->ire_ptpn = NULL; 39077c478bd9Sstevel@tonic-gate ire->ire_next = NULL; 39087c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 3909*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v6, ire_stats_deleted); 39107c478bd9Sstevel@tonic-gate } else { 3911*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_deleted); 39127c478bd9Sstevel@tonic-gate } 39137c478bd9Sstevel@tonic-gate /* 39147c478bd9Sstevel@tonic-gate * ip_wput/ip_wput_v6 checks this flag to see whether 39157c478bd9Sstevel@tonic-gate * it should still use the cached ire or not. 39167c478bd9Sstevel@tonic-gate */ 39177c478bd9Sstevel@tonic-gate ire->ire_marks |= IRE_MARK_CONDEMNED; 39187c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT) { 39197c478bd9Sstevel@tonic-gate /* 39207c478bd9Sstevel@tonic-gate * IRE is out of the list. We need to adjust the 39217c478bd9Sstevel@tonic-gate * accounting before we drop the lock. 39227c478bd9Sstevel@tonic-gate */ 39237c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 3924*f4b3ec61Sdh155122 ASSERT(ipst->ips_ipv6_ire_default_count != 0); 3925*f4b3ec61Sdh155122 ipst->ips_ipv6_ire_default_count--; 39267c478bd9Sstevel@tonic-gate } 39277c478bd9Sstevel@tonic-gate } 39287c478bd9Sstevel@tonic-gate irb->irb_ire_cnt--; 3929c793af95Ssangeeta 39307c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_TEMPORARY) 39317c478bd9Sstevel@tonic-gate irb->irb_tmp_ire_cnt--; 39327c478bd9Sstevel@tonic-gate rw_exit(&irb->irb_lock); 39337c478bd9Sstevel@tonic-gate 39347c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 39357c478bd9Sstevel@tonic-gate ire_delete_v6(ire); 39367c478bd9Sstevel@tonic-gate } else { 39377c478bd9Sstevel@tonic-gate ire_delete_v4(ire); 39387c478bd9Sstevel@tonic-gate } 39397c478bd9Sstevel@tonic-gate /* 39407c478bd9Sstevel@tonic-gate * We removed it from the list. Decrement the 39417c478bd9Sstevel@tonic-gate * reference count. 39427c478bd9Sstevel@tonic-gate */ 39437c478bd9Sstevel@tonic-gate IRE_REFRELE_NOTR(ire); 39447c478bd9Sstevel@tonic-gate } 39457c478bd9Sstevel@tonic-gate 39467c478bd9Sstevel@tonic-gate /* 39477c478bd9Sstevel@tonic-gate * Delete the specified IRE. 39487c478bd9Sstevel@tonic-gate * All calls should use ire_delete(). 39497c478bd9Sstevel@tonic-gate * Sometimes called as writer though not required by this function. 39507c478bd9Sstevel@tonic-gate * 39517c478bd9Sstevel@tonic-gate * NOTE : This function is called only if the ire was added 39527c478bd9Sstevel@tonic-gate * in the list. 39537c478bd9Sstevel@tonic-gate */ 39547c478bd9Sstevel@tonic-gate static void 39557c478bd9Sstevel@tonic-gate ire_delete_v4(ire_t *ire) 39567c478bd9Sstevel@tonic-gate { 3957*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 3958*f4b3ec61Sdh155122 39597c478bd9Sstevel@tonic-gate ASSERT(ire->ire_refcnt >= 1); 39607c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 39617c478bd9Sstevel@tonic-gate 39627c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 39637c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire, IRE_FLUSH_DELETE); 39647c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT) { 39657c478bd9Sstevel@tonic-gate /* 39667c478bd9Sstevel@tonic-gate * when a default gateway is going away 39677c478bd9Sstevel@tonic-gate * delete all the host redirects pointing at that 39687c478bd9Sstevel@tonic-gate * gateway. 39697c478bd9Sstevel@tonic-gate */ 3970*f4b3ec61Sdh155122 ire_delete_host_redirects(ire->ire_gateway_addr, ipst); 39717c478bd9Sstevel@tonic-gate } 39727c478bd9Sstevel@tonic-gate } 39737c478bd9Sstevel@tonic-gate 39747c478bd9Sstevel@tonic-gate /* 39757c478bd9Sstevel@tonic-gate * IRE_REFRELE/ire_refrele are the only caller of the function. It calls 39767c478bd9Sstevel@tonic-gate * to free the ire when the reference count goes to zero. 39777c478bd9Sstevel@tonic-gate */ 39787c478bd9Sstevel@tonic-gate void 39797c478bd9Sstevel@tonic-gate ire_inactive(ire_t *ire) 39807c478bd9Sstevel@tonic-gate { 39817c478bd9Sstevel@tonic-gate nce_t *nce; 39827c478bd9Sstevel@tonic-gate ill_t *ill = NULL; 39837c478bd9Sstevel@tonic-gate ill_t *stq_ill = NULL; 39847c478bd9Sstevel@tonic-gate ill_t *in_ill = NULL; 39857c478bd9Sstevel@tonic-gate ipif_t *ipif; 39867c478bd9Sstevel@tonic-gate boolean_t need_wakeup = B_FALSE; 3987c793af95Ssangeeta irb_t *irb; 3988*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 39897c478bd9Sstevel@tonic-gate 39907c478bd9Sstevel@tonic-gate ASSERT(ire->ire_refcnt == 0); 39917c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ptpn == NULL); 39927c478bd9Sstevel@tonic-gate ASSERT(ire->ire_next == NULL); 39937c478bd9Sstevel@tonic-gate 3994c793af95Ssangeeta if (ire->ire_gw_secattr != NULL) { 3995c793af95Ssangeeta ire_gw_secattr_free(ire->ire_gw_secattr); 3996c793af95Ssangeeta ire->ire_gw_secattr = NULL; 3997c793af95Ssangeeta } 3998c793af95Ssangeeta 3999c793af95Ssangeeta if (ire->ire_mp != NULL) { 4000c793af95Ssangeeta ASSERT(ire->ire_bucket == NULL); 4001c793af95Ssangeeta mutex_destroy(&ire->ire_lock); 4002*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_freed); 4003c793af95Ssangeeta if (ire->ire_nce != NULL) 4004c793af95Ssangeeta NCE_REFRELE_NOTR(ire->ire_nce); 4005c793af95Ssangeeta freeb(ire->ire_mp); 4006c793af95Ssangeeta return; 4007c793af95Ssangeeta } 4008c793af95Ssangeeta 40097c478bd9Sstevel@tonic-gate if ((nce = ire->ire_nce) != NULL) { 40107c478bd9Sstevel@tonic-gate NCE_REFRELE_NOTR(nce); 40117c478bd9Sstevel@tonic-gate ire->ire_nce = NULL; 40127c478bd9Sstevel@tonic-gate } 4013c793af95Ssangeeta 40147c478bd9Sstevel@tonic-gate if (ire->ire_ipif == NULL) 40157c478bd9Sstevel@tonic-gate goto end; 40167c478bd9Sstevel@tonic-gate 40177c478bd9Sstevel@tonic-gate ipif = ire->ire_ipif; 40187c478bd9Sstevel@tonic-gate ill = ipif->ipif_ill; 40197c478bd9Sstevel@tonic-gate 40207c478bd9Sstevel@tonic-gate if (ire->ire_bucket == NULL) { 40217c478bd9Sstevel@tonic-gate /* The ire was never inserted in the table. */ 40227c478bd9Sstevel@tonic-gate goto end; 40237c478bd9Sstevel@tonic-gate } 40247c478bd9Sstevel@tonic-gate 40257c478bd9Sstevel@tonic-gate /* 40267c478bd9Sstevel@tonic-gate * ipif_ire_cnt on this ipif goes down by 1. If the ire_stq is 40277c478bd9Sstevel@tonic-gate * non-null ill_ire_count also goes down by 1. If the in_ill is 40287c478bd9Sstevel@tonic-gate * non-null either ill_mrtun_refcnt or ill_srcif_refcnt goes down by 1. 40297c478bd9Sstevel@tonic-gate * 40307c478bd9Sstevel@tonic-gate * The ipif that is associated with an ire is ire->ire_ipif and 40317c478bd9Sstevel@tonic-gate * hence when the ire->ire_ipif->ipif_ire_cnt drops to zero we call 40327c478bd9Sstevel@tonic-gate * ipif_ill_refrele_tail. Usually stq_ill is null or the same as 40337c478bd9Sstevel@tonic-gate * ire->ire_ipif->ipif_ill. So nothing more needs to be done. Only 40347c478bd9Sstevel@tonic-gate * in the case of IRE_CACHES when IPMP is used, stq_ill can be 40357c478bd9Sstevel@tonic-gate * different. If this is different from ire->ire_ipif->ipif_ill and 40367c478bd9Sstevel@tonic-gate * if the ill_ire_cnt on the stq_ill also has dropped to zero, we call 40377c478bd9Sstevel@tonic-gate * ipif_ill_refrele_tail on the stq_ill. If mobile ip is in use 40387c478bd9Sstevel@tonic-gate * in_ill could be non-null. If it is a reverse tunnel related ire 40397c478bd9Sstevel@tonic-gate * ill_mrtun_refcnt is non-zero. If it is forward tunnel related ire 40407c478bd9Sstevel@tonic-gate * ill_srcif_refcnt is non-null. 40417c478bd9Sstevel@tonic-gate */ 40427c478bd9Sstevel@tonic-gate 40437c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) 40447c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 40457c478bd9Sstevel@tonic-gate if (ire->ire_in_ill != NULL) 40467c478bd9Sstevel@tonic-gate in_ill = ire->ire_in_ill; 40477c478bd9Sstevel@tonic-gate 40487c478bd9Sstevel@tonic-gate if ((stq_ill == NULL || stq_ill == ill) && (in_ill == NULL)) { 40497c478bd9Sstevel@tonic-gate /* Optimize the most common case */ 40507c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 40517c478bd9Sstevel@tonic-gate ASSERT(ipif->ipif_ire_cnt != 0); 40527c478bd9Sstevel@tonic-gate ipif->ipif_ire_cnt--; 40537c478bd9Sstevel@tonic-gate if (ipif->ipif_ire_cnt == 0) 40547c478bd9Sstevel@tonic-gate need_wakeup = B_TRUE; 40557c478bd9Sstevel@tonic-gate if (stq_ill != NULL) { 40567c478bd9Sstevel@tonic-gate ASSERT(stq_ill->ill_ire_cnt != 0); 40577c478bd9Sstevel@tonic-gate stq_ill->ill_ire_cnt--; 40587c478bd9Sstevel@tonic-gate if (stq_ill->ill_ire_cnt == 0) 40597c478bd9Sstevel@tonic-gate need_wakeup = B_TRUE; 40607c478bd9Sstevel@tonic-gate } 40617c478bd9Sstevel@tonic-gate if (need_wakeup) { 40627c478bd9Sstevel@tonic-gate /* Drops the ill lock */ 40637c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(ill); 40647c478bd9Sstevel@tonic-gate } else { 40657c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 40667c478bd9Sstevel@tonic-gate } 40677c478bd9Sstevel@tonic-gate } else { 40687c478bd9Sstevel@tonic-gate /* 40697c478bd9Sstevel@tonic-gate * We can't grab all the ill locks at the same time. 40707c478bd9Sstevel@tonic-gate * It can lead to recursive lock enter in the call to 40717c478bd9Sstevel@tonic-gate * ipif_ill_refrele_tail and later. Instead do it 1 at 40727c478bd9Sstevel@tonic-gate * a time. 40737c478bd9Sstevel@tonic-gate */ 40747c478bd9Sstevel@tonic-gate mutex_enter(&ill->ill_lock); 40757c478bd9Sstevel@tonic-gate ASSERT(ipif->ipif_ire_cnt != 0); 40767c478bd9Sstevel@tonic-gate ipif->ipif_ire_cnt--; 40777c478bd9Sstevel@tonic-gate if (ipif->ipif_ire_cnt == 0) { 40787c478bd9Sstevel@tonic-gate /* Drops the lock */ 40797c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(ill); 40807c478bd9Sstevel@tonic-gate } else { 40817c478bd9Sstevel@tonic-gate mutex_exit(&ill->ill_lock); 40827c478bd9Sstevel@tonic-gate } 40837c478bd9Sstevel@tonic-gate if (stq_ill != NULL) { 40847c478bd9Sstevel@tonic-gate mutex_enter(&stq_ill->ill_lock); 40857c478bd9Sstevel@tonic-gate ASSERT(stq_ill->ill_ire_cnt != 0); 40867c478bd9Sstevel@tonic-gate stq_ill->ill_ire_cnt--; 40877c478bd9Sstevel@tonic-gate if (stq_ill->ill_ire_cnt == 0) { 40887c478bd9Sstevel@tonic-gate /* Drops the ill lock */ 40897c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(stq_ill); 40907c478bd9Sstevel@tonic-gate } else { 40917c478bd9Sstevel@tonic-gate mutex_exit(&stq_ill->ill_lock); 40927c478bd9Sstevel@tonic-gate } 40937c478bd9Sstevel@tonic-gate } 40947c478bd9Sstevel@tonic-gate if (in_ill != NULL) { 40957c478bd9Sstevel@tonic-gate mutex_enter(&in_ill->ill_lock); 40967c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_MIPRTUN) { 40977c478bd9Sstevel@tonic-gate /* 40987c478bd9Sstevel@tonic-gate * Mobile IP reverse tunnel ire. 40997c478bd9Sstevel@tonic-gate * Decrement table count and the 41007c478bd9Sstevel@tonic-gate * ill reference count. This signifies 41017c478bd9Sstevel@tonic-gate * mipagent is deleting reverse tunnel 41027c478bd9Sstevel@tonic-gate * route for a particular mobile node. 41037c478bd9Sstevel@tonic-gate */ 4104*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_mrtun_lock); 4105*f4b3ec61Sdh155122 ipst->ips_ire_mrtun_count--; 4106*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_mrtun_lock); 41077c478bd9Sstevel@tonic-gate ASSERT(in_ill->ill_mrtun_refcnt != 0); 41087c478bd9Sstevel@tonic-gate in_ill->ill_mrtun_refcnt--; 41097c478bd9Sstevel@tonic-gate if (in_ill->ill_mrtun_refcnt == 0) { 41107c478bd9Sstevel@tonic-gate /* Drops the ill lock */ 41117c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(in_ill); 41127c478bd9Sstevel@tonic-gate } else { 41137c478bd9Sstevel@tonic-gate mutex_exit(&in_ill->ill_lock); 41147c478bd9Sstevel@tonic-gate } 41157c478bd9Sstevel@tonic-gate } else { 4116*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_srcif_table_lock); 4117*f4b3ec61Sdh155122 ipst->ips_ire_srcif_table_count--; 4118*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_srcif_table_lock); 41197c478bd9Sstevel@tonic-gate ASSERT(in_ill->ill_srcif_refcnt != 0); 41207c478bd9Sstevel@tonic-gate in_ill->ill_srcif_refcnt--; 41217c478bd9Sstevel@tonic-gate if (in_ill->ill_srcif_refcnt == 0) { 41227c478bd9Sstevel@tonic-gate /* Drops the ill lock */ 41237c478bd9Sstevel@tonic-gate ipif_ill_refrele_tail(in_ill); 41247c478bd9Sstevel@tonic-gate } else { 41257c478bd9Sstevel@tonic-gate mutex_exit(&in_ill->ill_lock); 41267c478bd9Sstevel@tonic-gate } 41277c478bd9Sstevel@tonic-gate } 41287c478bd9Sstevel@tonic-gate } 41297c478bd9Sstevel@tonic-gate } 41307c478bd9Sstevel@tonic-gate end: 41317c478bd9Sstevel@tonic-gate /* This should be true for both V4 and V6 */ 41327c478bd9Sstevel@tonic-gate 4133c793af95Ssangeeta if ((ire->ire_type & IRE_FORWARDTABLE) && 4134c793af95Ssangeeta (ire->ire_ipversion == IPV4_VERSION) && 4135c793af95Ssangeeta ((irb = ire->ire_bucket) != NULL)) { 4136c793af95Ssangeeta rw_enter(&irb->irb_lock, RW_WRITER); 4137c793af95Ssangeeta irb->irb_nire--; 4138c793af95Ssangeeta /* 4139c793af95Ssangeeta * Instead of examining the conditions for freeing 4140c793af95Ssangeeta * the radix node here, we do it by calling 4141c793af95Ssangeeta * IRB_REFRELE which is a single point in the code 4142c793af95Ssangeeta * that embeds that logic. Bump up the refcnt to 4143c793af95Ssangeeta * be able to call IRB_REFRELE 4144c793af95Ssangeeta */ 4145c793af95Ssangeeta IRB_REFHOLD_LOCKED(irb); 4146c793af95Ssangeeta rw_exit(&irb->irb_lock); 4147c793af95Ssangeeta IRB_REFRELE(irb); 4148c793af95Ssangeeta } 41497c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 41507c478bd9Sstevel@tonic-gate 41517c478bd9Sstevel@tonic-gate if (ire->ire_in_ill != NULL) { 41527c478bd9Sstevel@tonic-gate ire->ire_in_ill = NULL; 41537c478bd9Sstevel@tonic-gate } 41547c478bd9Sstevel@tonic-gate 41557c478bd9Sstevel@tonic-gate #ifdef IRE_DEBUG 41567c478bd9Sstevel@tonic-gate ire_trace_inactive(ire); 41577c478bd9Sstevel@tonic-gate #endif 41587c478bd9Sstevel@tonic-gate mutex_destroy(&ire->ire_lock); 41597c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 4160*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v6, ire_stats_freed); 41617c478bd9Sstevel@tonic-gate } else { 4162*f4b3ec61Sdh155122 BUMP_IRE_STATS(ipst->ips_ire_stats_v4, ire_stats_freed); 41637c478bd9Sstevel@tonic-gate } 4164c793af95Ssangeeta ASSERT(ire->ire_mp == NULL); 41657c478bd9Sstevel@tonic-gate /* Has been allocated out of the cache */ 41667c478bd9Sstevel@tonic-gate kmem_cache_free(ire_cache, ire); 41677c478bd9Sstevel@tonic-gate } 41687c478bd9Sstevel@tonic-gate 41697c478bd9Sstevel@tonic-gate /* 41706bdb8e66Sdd193516 * ire_walk routine to delete all IRE_CACHE/IRE_HOST types redirect 41716bdb8e66Sdd193516 * entries that have a given gateway address. 41727c478bd9Sstevel@tonic-gate */ 41737c478bd9Sstevel@tonic-gate void 41747c478bd9Sstevel@tonic-gate ire_delete_cache_gw(ire_t *ire, char *cp) 41757c478bd9Sstevel@tonic-gate { 41767c478bd9Sstevel@tonic-gate ipaddr_t gw_addr; 41777c478bd9Sstevel@tonic-gate 41786bdb8e66Sdd193516 if (!(ire->ire_type & IRE_CACHE) && 41796bdb8e66Sdd193516 !(ire->ire_flags & RTF_DYNAMIC)) 41807c478bd9Sstevel@tonic-gate return; 41817c478bd9Sstevel@tonic-gate 41827c478bd9Sstevel@tonic-gate bcopy(cp, &gw_addr, sizeof (gw_addr)); 41837c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == gw_addr) { 41847c478bd9Sstevel@tonic-gate ip1dbg(("ire_delete_cache_gw: deleted 0x%x type %d to 0x%x\n", 41857c478bd9Sstevel@tonic-gate (int)ntohl(ire->ire_addr), ire->ire_type, 41867c478bd9Sstevel@tonic-gate (int)ntohl(ire->ire_gateway_addr))); 41877c478bd9Sstevel@tonic-gate ire_delete(ire); 41887c478bd9Sstevel@tonic-gate } 41897c478bd9Sstevel@tonic-gate } 41907c478bd9Sstevel@tonic-gate 41917c478bd9Sstevel@tonic-gate /* 41927c478bd9Sstevel@tonic-gate * Remove all IRE_CACHE entries that match the ire specified. 41937c478bd9Sstevel@tonic-gate * 41947c478bd9Sstevel@tonic-gate * The flag argument indicates if the flush request is due to addition 41957c478bd9Sstevel@tonic-gate * of new route (IRE_FLUSH_ADD) or deletion of old route (IRE_FLUSH_DELETE). 41967c478bd9Sstevel@tonic-gate * 41977c478bd9Sstevel@tonic-gate * This routine takes only the IREs from the forwarding table and flushes 41987c478bd9Sstevel@tonic-gate * the corresponding entries from the cache table. 41997c478bd9Sstevel@tonic-gate * 42007c478bd9Sstevel@tonic-gate * When flushing due to the deletion of an old route, it 42017c478bd9Sstevel@tonic-gate * just checks the cache handles (ire_phandle and ire_ihandle) and 42027c478bd9Sstevel@tonic-gate * deletes the ones that match. 42037c478bd9Sstevel@tonic-gate * 42047c478bd9Sstevel@tonic-gate * When flushing due to the creation of a new route, it checks 42057c478bd9Sstevel@tonic-gate * if a cache entry's address matches the one in the IRE and 42067c478bd9Sstevel@tonic-gate * that the cache entry's parent has a less specific mask than the 42077c478bd9Sstevel@tonic-gate * one in IRE. The destination of such a cache entry could be the 42087c478bd9Sstevel@tonic-gate * gateway for other cache entries, so we need to flush those as 42097c478bd9Sstevel@tonic-gate * well by looking for gateway addresses matching the IRE's address. 42107c478bd9Sstevel@tonic-gate */ 42117c478bd9Sstevel@tonic-gate void 42127c478bd9Sstevel@tonic-gate ire_flush_cache_v4(ire_t *ire, int flag) 42137c478bd9Sstevel@tonic-gate { 42147c478bd9Sstevel@tonic-gate int i; 42157c478bd9Sstevel@tonic-gate ire_t *cire; 42167c478bd9Sstevel@tonic-gate irb_t *irb; 4217*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 42187c478bd9Sstevel@tonic-gate 42197c478bd9Sstevel@tonic-gate if (ire->ire_type & IRE_CACHE) 42207c478bd9Sstevel@tonic-gate return; 42217c478bd9Sstevel@tonic-gate 42227c478bd9Sstevel@tonic-gate /* 42237c478bd9Sstevel@tonic-gate * If a default is just created, there is no point 42247c478bd9Sstevel@tonic-gate * in going through the cache, as there will not be any 42257c478bd9Sstevel@tonic-gate * cached ires. 42267c478bd9Sstevel@tonic-gate */ 42277c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_DEFAULT && flag == IRE_FLUSH_ADD) 42287c478bd9Sstevel@tonic-gate return; 42297c478bd9Sstevel@tonic-gate if (flag == IRE_FLUSH_ADD) { 42307c478bd9Sstevel@tonic-gate /* 42317c478bd9Sstevel@tonic-gate * This selective flush is due to the addition of 42327c478bd9Sstevel@tonic-gate * new IRE. 42337c478bd9Sstevel@tonic-gate */ 4234*f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 4235*f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[i]; 42367c478bd9Sstevel@tonic-gate if ((cire = irb->irb_ire) == NULL) 42377c478bd9Sstevel@tonic-gate continue; 42387c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 42397c478bd9Sstevel@tonic-gate for (cire = irb->irb_ire; cire != NULL; 42407c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 42417c478bd9Sstevel@tonic-gate if (cire->ire_type != IRE_CACHE) 42427c478bd9Sstevel@tonic-gate continue; 42437c478bd9Sstevel@tonic-gate /* 42447c478bd9Sstevel@tonic-gate * If 'cire' belongs to the same subnet 42457c478bd9Sstevel@tonic-gate * as the new ire being added, and 'cire' 42467c478bd9Sstevel@tonic-gate * is derived from a prefix that is less 42477c478bd9Sstevel@tonic-gate * specific than the new ire being added, 42487c478bd9Sstevel@tonic-gate * we need to flush 'cire'; for instance, 42497c478bd9Sstevel@tonic-gate * when a new interface comes up. 42507c478bd9Sstevel@tonic-gate */ 42517c478bd9Sstevel@tonic-gate if (((cire->ire_addr & ire->ire_mask) == 42527c478bd9Sstevel@tonic-gate (ire->ire_addr & ire->ire_mask)) && 42537c478bd9Sstevel@tonic-gate (ip_mask_to_plen(cire->ire_cmask) <= 42547c478bd9Sstevel@tonic-gate ire->ire_masklen)) { 42557c478bd9Sstevel@tonic-gate ire_delete(cire); 42567c478bd9Sstevel@tonic-gate continue; 42577c478bd9Sstevel@tonic-gate } 42587c478bd9Sstevel@tonic-gate /* 42597c478bd9Sstevel@tonic-gate * This is the case when the ire_gateway_addr 42607c478bd9Sstevel@tonic-gate * of 'cire' belongs to the same subnet as 42617c478bd9Sstevel@tonic-gate * the new ire being added. 42627c478bd9Sstevel@tonic-gate * Flushing such ires is sometimes required to 42637c478bd9Sstevel@tonic-gate * avoid misrouting: say we have a machine with 42647c478bd9Sstevel@tonic-gate * two interfaces (I1 and I2), a default router 42657c478bd9Sstevel@tonic-gate * R on the I1 subnet, and a host route to an 42667c478bd9Sstevel@tonic-gate * off-link destination D with a gateway G on 42677c478bd9Sstevel@tonic-gate * the I2 subnet. 42687c478bd9Sstevel@tonic-gate * Under normal operation, we will have an 42697c478bd9Sstevel@tonic-gate * on-link cache entry for G and an off-link 42707c478bd9Sstevel@tonic-gate * cache entry for D with G as ire_gateway_addr, 42717c478bd9Sstevel@tonic-gate * traffic to D will reach its destination 42727c478bd9Sstevel@tonic-gate * through gateway G. 42737c478bd9Sstevel@tonic-gate * If the administrator does 'ifconfig I2 down', 42747c478bd9Sstevel@tonic-gate * the cache entries for D and G will be 42757c478bd9Sstevel@tonic-gate * flushed. However, G will now be resolved as 42767c478bd9Sstevel@tonic-gate * an off-link destination using R (the default 42777c478bd9Sstevel@tonic-gate * router) as gateway. Then D will also be 42787c478bd9Sstevel@tonic-gate * resolved as an off-link destination using G 42797c478bd9Sstevel@tonic-gate * as gateway - this behavior is due to 42807c478bd9Sstevel@tonic-gate * compatibility reasons, see comment in 42817c478bd9Sstevel@tonic-gate * ire_ihandle_lookup_offlink(). Traffic to D 42827c478bd9Sstevel@tonic-gate * will go to the router R and probably won't 42837c478bd9Sstevel@tonic-gate * reach the destination. 42847c478bd9Sstevel@tonic-gate * The administrator then does 'ifconfig I2 up'. 42857c478bd9Sstevel@tonic-gate * Since G is on the I2 subnet, this routine 42867c478bd9Sstevel@tonic-gate * will flush its cache entry. It must also 42877c478bd9Sstevel@tonic-gate * flush the cache entry for D, otherwise 42887c478bd9Sstevel@tonic-gate * traffic will stay misrouted until the IRE 42897c478bd9Sstevel@tonic-gate * times out. 42907c478bd9Sstevel@tonic-gate */ 42917c478bd9Sstevel@tonic-gate if ((cire->ire_gateway_addr & ire->ire_mask) == 42927c478bd9Sstevel@tonic-gate (ire->ire_addr & ire->ire_mask)) { 42937c478bd9Sstevel@tonic-gate ire_delete(cire); 42947c478bd9Sstevel@tonic-gate continue; 42957c478bd9Sstevel@tonic-gate } 42967c478bd9Sstevel@tonic-gate } 42977c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 42987c478bd9Sstevel@tonic-gate } 42997c478bd9Sstevel@tonic-gate } else { 43007c478bd9Sstevel@tonic-gate /* 43017c478bd9Sstevel@tonic-gate * delete the cache entries based on 43027c478bd9Sstevel@tonic-gate * handle in the IRE as this IRE is 43037c478bd9Sstevel@tonic-gate * being deleted/changed. 43047c478bd9Sstevel@tonic-gate */ 4305*f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 4306*f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[i]; 43077c478bd9Sstevel@tonic-gate if ((cire = irb->irb_ire) == NULL) 43087c478bd9Sstevel@tonic-gate continue; 43097c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 43107c478bd9Sstevel@tonic-gate for (cire = irb->irb_ire; cire != NULL; 43117c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 43127c478bd9Sstevel@tonic-gate if (cire->ire_type != IRE_CACHE) 43137c478bd9Sstevel@tonic-gate continue; 43147c478bd9Sstevel@tonic-gate if ((cire->ire_phandle == 0 || 43157c478bd9Sstevel@tonic-gate cire->ire_phandle != ire->ire_phandle) && 43167c478bd9Sstevel@tonic-gate (cire->ire_ihandle == 0 || 43177c478bd9Sstevel@tonic-gate cire->ire_ihandle != ire->ire_ihandle)) 43187c478bd9Sstevel@tonic-gate continue; 43197c478bd9Sstevel@tonic-gate ire_delete(cire); 43207c478bd9Sstevel@tonic-gate } 43217c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 43227c478bd9Sstevel@tonic-gate } 43237c478bd9Sstevel@tonic-gate } 43247c478bd9Sstevel@tonic-gate } 43257c478bd9Sstevel@tonic-gate 43267c478bd9Sstevel@tonic-gate /* 43277c478bd9Sstevel@tonic-gate * Matches the arguments passed with the values in the ire. 43287c478bd9Sstevel@tonic-gate * 43297c478bd9Sstevel@tonic-gate * Note: for match types that match using "ipif" passed in, ipif 43307c478bd9Sstevel@tonic-gate * must be checked for non-NULL before calling this routine. 43317c478bd9Sstevel@tonic-gate */ 4332c793af95Ssangeeta boolean_t 43337c478bd9Sstevel@tonic-gate ire_match_args(ire_t *ire, ipaddr_t addr, ipaddr_t mask, ipaddr_t gateway, 433445916cd2Sjpk int type, const ipif_t *ipif, zoneid_t zoneid, uint32_t ihandle, 433545916cd2Sjpk const ts_label_t *tsl, int match_flags) 43367c478bd9Sstevel@tonic-gate { 43377c478bd9Sstevel@tonic-gate ill_t *ire_ill = NULL, *dst_ill; 43387c478bd9Sstevel@tonic-gate ill_t *ipif_ill = NULL; 43397c478bd9Sstevel@tonic-gate ill_group_t *ire_ill_group = NULL; 43407c478bd9Sstevel@tonic-gate ill_group_t *ipif_ill_group = NULL; 43417c478bd9Sstevel@tonic-gate 43427c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 43437c478bd9Sstevel@tonic-gate ASSERT((ire->ire_addr & ~ire->ire_mask) == 0); 43447c478bd9Sstevel@tonic-gate ASSERT((!(match_flags & (MATCH_IRE_ILL|MATCH_IRE_ILL_GROUP))) || 43457c478bd9Sstevel@tonic-gate (ipif != NULL && !ipif->ipif_isv6)); 43467c478bd9Sstevel@tonic-gate ASSERT(!(match_flags & MATCH_IRE_WQ)); 43477c478bd9Sstevel@tonic-gate 43487c478bd9Sstevel@tonic-gate /* 43497c478bd9Sstevel@tonic-gate * HIDDEN cache entries have to be looked up specifically with 43507c478bd9Sstevel@tonic-gate * MATCH_IRE_MARK_HIDDEN. MATCH_IRE_MARK_HIDDEN is usually set 43517c478bd9Sstevel@tonic-gate * when the interface is FAILED or INACTIVE. In that case, 43527c478bd9Sstevel@tonic-gate * any IRE_CACHES that exists should be marked with 43537c478bd9Sstevel@tonic-gate * IRE_MARK_HIDDEN. So, we don't really need to match below 43547c478bd9Sstevel@tonic-gate * for IRE_MARK_HIDDEN. But we do so for consistency. 43557c478bd9Sstevel@tonic-gate */ 43567c478bd9Sstevel@tonic-gate if (!(match_flags & MATCH_IRE_MARK_HIDDEN) && 43577c478bd9Sstevel@tonic-gate (ire->ire_marks & IRE_MARK_HIDDEN)) 43587c478bd9Sstevel@tonic-gate return (B_FALSE); 43597c478bd9Sstevel@tonic-gate 436043d18f1cSpriyanka /* 436143d18f1cSpriyanka * MATCH_IRE_MARK_PRIVATE_ADDR is set when IP_NEXTHOP option 436243d18f1cSpriyanka * is used. In that case the routing table is bypassed and the 436343d18f1cSpriyanka * packets are sent directly to the specified nexthop. The 436443d18f1cSpriyanka * IRE_CACHE entry representing this route should be marked 436543d18f1cSpriyanka * with IRE_MARK_PRIVATE_ADDR. 436643d18f1cSpriyanka */ 436743d18f1cSpriyanka 436843d18f1cSpriyanka if (!(match_flags & MATCH_IRE_MARK_PRIVATE_ADDR) && 436943d18f1cSpriyanka (ire->ire_marks & IRE_MARK_PRIVATE_ADDR)) 437043d18f1cSpriyanka return (B_FALSE); 437143d18f1cSpriyanka 437245916cd2Sjpk if (zoneid != ALL_ZONES && zoneid != ire->ire_zoneid && 437345916cd2Sjpk ire->ire_zoneid != ALL_ZONES) { 43747c478bd9Sstevel@tonic-gate /* 43757c478bd9Sstevel@tonic-gate * If MATCH_IRE_ZONEONLY has been set and the supplied zoneid is 43767c478bd9Sstevel@tonic-gate * valid and does not match that of ire_zoneid, a failure to 43777c478bd9Sstevel@tonic-gate * match is reported at this point. Otherwise, since some IREs 43787c478bd9Sstevel@tonic-gate * that are available in the global zone can be used in local 43797c478bd9Sstevel@tonic-gate * zones, additional checks need to be performed: 43807c478bd9Sstevel@tonic-gate * 43817c478bd9Sstevel@tonic-gate * IRE_BROADCAST, IRE_CACHE and IRE_LOOPBACK 43827c478bd9Sstevel@tonic-gate * entries should never be matched in this situation. 43837c478bd9Sstevel@tonic-gate * 43847c478bd9Sstevel@tonic-gate * IRE entries that have an interface associated with them 43857c478bd9Sstevel@tonic-gate * should in general not match unless they are an IRE_LOCAL 43867c478bd9Sstevel@tonic-gate * or in the case when MATCH_IRE_DEFAULT has been set in 43877c478bd9Sstevel@tonic-gate * the caller. In the case of the former, checking of the 43887c478bd9Sstevel@tonic-gate * other fields supplied should take place. 43897c478bd9Sstevel@tonic-gate * 43907c478bd9Sstevel@tonic-gate * In the case where MATCH_IRE_DEFAULT has been set, 43917c478bd9Sstevel@tonic-gate * all of the ipif's associated with the IRE's ill are 43927c478bd9Sstevel@tonic-gate * checked to see if there is a matching zoneid. If any 43937c478bd9Sstevel@tonic-gate * one ipif has a matching zoneid, this IRE is a 43947c478bd9Sstevel@tonic-gate * potential candidate so checking of the other fields 43957c478bd9Sstevel@tonic-gate * takes place. 43967c478bd9Sstevel@tonic-gate * 43977c478bd9Sstevel@tonic-gate * In the case where the IRE_INTERFACE has a usable source 43987c478bd9Sstevel@tonic-gate * address (indicated by ill_usesrc_ifindex) in the 43997c478bd9Sstevel@tonic-gate * correct zone then it's permitted to return this IRE 44007c478bd9Sstevel@tonic-gate */ 44017c478bd9Sstevel@tonic-gate if (match_flags & MATCH_IRE_ZONEONLY) 44027c478bd9Sstevel@tonic-gate return (B_FALSE); 44037c478bd9Sstevel@tonic-gate if (ire->ire_type & (IRE_BROADCAST | IRE_CACHE | IRE_LOOPBACK)) 44047c478bd9Sstevel@tonic-gate return (B_FALSE); 44057c478bd9Sstevel@tonic-gate /* 44067c478bd9Sstevel@tonic-gate * Note, IRE_INTERFACE can have the stq as NULL. For 44077c478bd9Sstevel@tonic-gate * example, if the default multicast route is tied to 44087c478bd9Sstevel@tonic-gate * the loopback address. 44097c478bd9Sstevel@tonic-gate */ 44107c478bd9Sstevel@tonic-gate if ((ire->ire_type & IRE_INTERFACE) && 44117c478bd9Sstevel@tonic-gate (ire->ire_stq != NULL)) { 44127c478bd9Sstevel@tonic-gate dst_ill = (ill_t *)ire->ire_stq->q_ptr; 44137c478bd9Sstevel@tonic-gate /* 44147c478bd9Sstevel@tonic-gate * If there is a usable source address in the 44157c478bd9Sstevel@tonic-gate * zone, then it's ok to return an 44167c478bd9Sstevel@tonic-gate * IRE_INTERFACE 44177c478bd9Sstevel@tonic-gate */ 44187c478bd9Sstevel@tonic-gate if (ipif_usesrc_avail(dst_ill, zoneid)) { 44197c478bd9Sstevel@tonic-gate ip3dbg(("ire_match_args: dst_ill %p match %d\n", 44207c478bd9Sstevel@tonic-gate (void *)dst_ill, 44217c478bd9Sstevel@tonic-gate (ire->ire_addr == (addr & mask)))); 44227c478bd9Sstevel@tonic-gate } else { 44237c478bd9Sstevel@tonic-gate ip3dbg(("ire_match_args: src_ipif NULL" 44247c478bd9Sstevel@tonic-gate " dst_ill %p\n", (void *)dst_ill)); 44257c478bd9Sstevel@tonic-gate return (B_FALSE); 44267c478bd9Sstevel@tonic-gate } 44277c478bd9Sstevel@tonic-gate } 44287c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL && ire->ire_type != IRE_LOCAL && 44297c478bd9Sstevel@tonic-gate !(ire->ire_type & IRE_INTERFACE)) { 44307c478bd9Sstevel@tonic-gate ipif_t *tipif; 44317c478bd9Sstevel@tonic-gate 44327c478bd9Sstevel@tonic-gate if ((match_flags & MATCH_IRE_DEFAULT) == 0) { 44337c478bd9Sstevel@tonic-gate return (B_FALSE); 44347c478bd9Sstevel@tonic-gate } 44357c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_ipif->ipif_ill->ill_lock); 44367c478bd9Sstevel@tonic-gate for (tipif = ire->ire_ipif->ipif_ill->ill_ipif; 44377c478bd9Sstevel@tonic-gate tipif != NULL; tipif = tipif->ipif_next) { 44387c478bd9Sstevel@tonic-gate if (IPIF_CAN_LOOKUP(tipif) && 44397c478bd9Sstevel@tonic-gate (tipif->ipif_flags & IPIF_UP) && 444045916cd2Sjpk (tipif->ipif_zoneid == zoneid || 444145916cd2Sjpk tipif->ipif_zoneid == ALL_ZONES)) 44427c478bd9Sstevel@tonic-gate break; 44437c478bd9Sstevel@tonic-gate } 44447c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_ipif->ipif_ill->ill_lock); 44457c478bd9Sstevel@tonic-gate if (tipif == NULL) { 44467c478bd9Sstevel@tonic-gate return (B_FALSE); 44477c478bd9Sstevel@tonic-gate } 44487c478bd9Sstevel@tonic-gate } 44497c478bd9Sstevel@tonic-gate } 44507c478bd9Sstevel@tonic-gate 44517c478bd9Sstevel@tonic-gate /* 44527c478bd9Sstevel@tonic-gate * For IRE_CACHES, MATCH_IRE_ILL/ILL_GROUP really means that 44537c478bd9Sstevel@tonic-gate * somebody wants to send out on a particular interface which 44547c478bd9Sstevel@tonic-gate * is given by ire_stq and hence use ire_stq to derive the ill 44557c478bd9Sstevel@tonic-gate * value. ire_ipif for IRE_CACHES is just the means of getting 44567c478bd9Sstevel@tonic-gate * a source address i.e ire_src_addr = ire->ire_ipif->ipif_src_addr. 44577c478bd9Sstevel@tonic-gate * ire_to_ill does the right thing for this. 44587c478bd9Sstevel@tonic-gate */ 44597c478bd9Sstevel@tonic-gate if (match_flags & (MATCH_IRE_ILL|MATCH_IRE_ILL_GROUP)) { 44607c478bd9Sstevel@tonic-gate ire_ill = ire_to_ill(ire); 44617c478bd9Sstevel@tonic-gate if (ire_ill != NULL) 44627c478bd9Sstevel@tonic-gate ire_ill_group = ire_ill->ill_group; 44637c478bd9Sstevel@tonic-gate ipif_ill = ipif->ipif_ill; 44647c478bd9Sstevel@tonic-gate ipif_ill_group = ipif_ill->ill_group; 44657c478bd9Sstevel@tonic-gate } 44667c478bd9Sstevel@tonic-gate 44677c478bd9Sstevel@tonic-gate if ((ire->ire_addr == (addr & mask)) && 44687c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_GW)) || 44697c478bd9Sstevel@tonic-gate (ire->ire_gateway_addr == gateway)) && 44707c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_TYPE)) || 44717c478bd9Sstevel@tonic-gate (ire->ire_type & type)) && 44727c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_SRC)) || 44737c478bd9Sstevel@tonic-gate (ire->ire_src_addr == ipif->ipif_src_addr)) && 44747c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_IPIF)) || 44757c478bd9Sstevel@tonic-gate (ire->ire_ipif == ipif)) && 44767c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_MARK_HIDDEN)) || 44777c478bd9Sstevel@tonic-gate (ire->ire_type != IRE_CACHE || 44787c478bd9Sstevel@tonic-gate ire->ire_marks & IRE_MARK_HIDDEN)) && 447943d18f1cSpriyanka ((!(match_flags & MATCH_IRE_MARK_PRIVATE_ADDR)) || 448043d18f1cSpriyanka (ire->ire_type != IRE_CACHE || 448143d18f1cSpriyanka ire->ire_marks & IRE_MARK_PRIVATE_ADDR)) && 44827c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL)) || 44837c478bd9Sstevel@tonic-gate (ire_ill == ipif_ill)) && 44847c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_IHANDLE)) || 44857c478bd9Sstevel@tonic-gate (ire->ire_ihandle == ihandle)) && 4486c793af95Ssangeeta ((!(match_flags & MATCH_IRE_MASK)) || 4487c793af95Ssangeeta (ire->ire_mask == mask)) && 44887c478bd9Sstevel@tonic-gate ((!(match_flags & MATCH_IRE_ILL_GROUP)) || 44897c478bd9Sstevel@tonic-gate (ire_ill == ipif_ill) || 44907c478bd9Sstevel@tonic-gate (ire_ill_group != NULL && 449145916cd2Sjpk ire_ill_group == ipif_ill_group)) && 449245916cd2Sjpk ((!(match_flags & MATCH_IRE_SECATTR)) || 449345916cd2Sjpk (!is_system_labeled()) || 449445916cd2Sjpk (tsol_ire_match_gwattr(ire, tsl) == 0))) { 44957c478bd9Sstevel@tonic-gate /* We found the matched IRE */ 44967c478bd9Sstevel@tonic-gate return (B_TRUE); 44977c478bd9Sstevel@tonic-gate } 44987c478bd9Sstevel@tonic-gate return (B_FALSE); 44997c478bd9Sstevel@tonic-gate } 45007c478bd9Sstevel@tonic-gate 45017c478bd9Sstevel@tonic-gate 45027c478bd9Sstevel@tonic-gate /* 45037c478bd9Sstevel@tonic-gate * Lookup for a route in all the tables 45047c478bd9Sstevel@tonic-gate */ 45057c478bd9Sstevel@tonic-gate ire_t * 45067c478bd9Sstevel@tonic-gate ire_route_lookup(ipaddr_t addr, ipaddr_t mask, ipaddr_t gateway, 450745916cd2Sjpk int type, const ipif_t *ipif, ire_t **pire, zoneid_t zoneid, 4508*f4b3ec61Sdh155122 const ts_label_t *tsl, int flags, ip_stack_t *ipst) 45097c478bd9Sstevel@tonic-gate { 45107c478bd9Sstevel@tonic-gate ire_t *ire = NULL; 45117c478bd9Sstevel@tonic-gate 45127c478bd9Sstevel@tonic-gate /* 45137c478bd9Sstevel@tonic-gate * ire_match_args() will dereference ipif MATCH_IRE_SRC or 45147c478bd9Sstevel@tonic-gate * MATCH_IRE_ILL is set. 45157c478bd9Sstevel@tonic-gate */ 45167c478bd9Sstevel@tonic-gate if ((flags & (MATCH_IRE_SRC | MATCH_IRE_ILL | MATCH_IRE_ILL_GROUP)) && 45177c478bd9Sstevel@tonic-gate (ipif == NULL)) 45187c478bd9Sstevel@tonic-gate return (NULL); 45197c478bd9Sstevel@tonic-gate 45207c478bd9Sstevel@tonic-gate /* 45217c478bd9Sstevel@tonic-gate * might be asking for a cache lookup, 45227c478bd9Sstevel@tonic-gate * This is not best way to lookup cache, 45237c478bd9Sstevel@tonic-gate * user should call ire_cache_lookup directly. 45247c478bd9Sstevel@tonic-gate * 45257c478bd9Sstevel@tonic-gate * If MATCH_IRE_TYPE was set, first lookup in the cache table and then 45267c478bd9Sstevel@tonic-gate * in the forwarding table, if the applicable type flags were set. 45277c478bd9Sstevel@tonic-gate */ 45287c478bd9Sstevel@tonic-gate if ((flags & MATCH_IRE_TYPE) == 0 || (type & IRE_CACHETABLE) != 0) { 45297c478bd9Sstevel@tonic-gate ire = ire_ctable_lookup(addr, gateway, type, ipif, zoneid, 4530*f4b3ec61Sdh155122 tsl, flags, ipst); 45317c478bd9Sstevel@tonic-gate if (ire != NULL) 45327c478bd9Sstevel@tonic-gate return (ire); 45337c478bd9Sstevel@tonic-gate } 45347c478bd9Sstevel@tonic-gate if ((flags & MATCH_IRE_TYPE) == 0 || (type & IRE_FORWARDTABLE) != 0) { 45357c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(addr, mask, gateway, type, ipif, pire, 4536*f4b3ec61Sdh155122 zoneid, 0, tsl, flags, ipst); 45377c478bd9Sstevel@tonic-gate } 45387c478bd9Sstevel@tonic-gate return (ire); 45397c478bd9Sstevel@tonic-gate } 45407c478bd9Sstevel@tonic-gate 45417c478bd9Sstevel@tonic-gate 45427c478bd9Sstevel@tonic-gate /* 454345916cd2Sjpk * Delete the IRE cache for the gateway and all IRE caches whose 454445916cd2Sjpk * ire_gateway_addr points to this gateway, and allow them to 454545916cd2Sjpk * be created on demand by ip_newroute. 454645916cd2Sjpk */ 454745916cd2Sjpk void 4548*f4b3ec61Sdh155122 ire_clookup_delete_cache_gw(ipaddr_t addr, zoneid_t zoneid, ip_stack_t *ipst) 454945916cd2Sjpk { 455045916cd2Sjpk irb_t *irb; 455145916cd2Sjpk ire_t *ire; 455245916cd2Sjpk 4553*f4b3ec61Sdh155122 irb = &ipst->ips_ip_cache_table[IRE_ADDR_HASH(addr, 4554*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size)]; 455545916cd2Sjpk IRB_REFHOLD(irb); 455645916cd2Sjpk for (ire = irb->irb_ire; ire != NULL; ire = ire->ire_next) { 455745916cd2Sjpk if (ire->ire_marks & IRE_MARK_CONDEMNED) 455845916cd2Sjpk continue; 455945916cd2Sjpk 456045916cd2Sjpk ASSERT(ire->ire_mask == IP_HOST_MASK); 456145916cd2Sjpk ASSERT(ire->ire_type != IRE_MIPRTUN && ire->ire_in_ill == NULL); 456245916cd2Sjpk if (ire_match_args(ire, addr, ire->ire_mask, 0, IRE_CACHE, 456345916cd2Sjpk NULL, zoneid, 0, NULL, MATCH_IRE_TYPE)) { 456445916cd2Sjpk ire_delete(ire); 456545916cd2Sjpk } 456645916cd2Sjpk } 456745916cd2Sjpk IRB_REFRELE(irb); 456845916cd2Sjpk 4569*f4b3ec61Sdh155122 ire_walk_v4(ire_delete_cache_gw, &addr, zoneid, ipst); 457045916cd2Sjpk } 457145916cd2Sjpk 457245916cd2Sjpk /* 45737c478bd9Sstevel@tonic-gate * Looks up cache table for a route. 45747c478bd9Sstevel@tonic-gate * specific lookup can be indicated by 45757c478bd9Sstevel@tonic-gate * passing the MATCH_* flags and the 45767c478bd9Sstevel@tonic-gate * necessary parameters. 45777c478bd9Sstevel@tonic-gate */ 45787c478bd9Sstevel@tonic-gate ire_t * 457945916cd2Sjpk ire_ctable_lookup(ipaddr_t addr, ipaddr_t gateway, int type, const ipif_t *ipif, 4580*f4b3ec61Sdh155122 zoneid_t zoneid, const ts_label_t *tsl, int flags, ip_stack_t *ipst) 45817c478bd9Sstevel@tonic-gate { 45827c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 45837c478bd9Sstevel@tonic-gate ire_t *ire; 45847c478bd9Sstevel@tonic-gate 45857c478bd9Sstevel@tonic-gate /* 45867c478bd9Sstevel@tonic-gate * ire_match_args() will dereference ipif MATCH_IRE_SRC or 45877c478bd9Sstevel@tonic-gate * MATCH_IRE_ILL is set. 45887c478bd9Sstevel@tonic-gate */ 45897c478bd9Sstevel@tonic-gate if ((flags & (MATCH_IRE_SRC | MATCH_IRE_ILL | MATCH_IRE_ILL_GROUP)) && 45907c478bd9Sstevel@tonic-gate (ipif == NULL)) 45917c478bd9Sstevel@tonic-gate return (NULL); 45927c478bd9Sstevel@tonic-gate 4593*f4b3ec61Sdh155122 irb_ptr = &ipst->ips_ip_cache_table[IRE_ADDR_HASH(addr, 4594*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size)]; 45957c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_READER); 45967c478bd9Sstevel@tonic-gate for (ire = irb_ptr->irb_ire; ire != NULL; ire = ire->ire_next) { 45977c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) 45987c478bd9Sstevel@tonic-gate continue; 45997c478bd9Sstevel@tonic-gate ASSERT(ire->ire_mask == IP_HOST_MASK); 46007c478bd9Sstevel@tonic-gate ASSERT(ire->ire_type != IRE_MIPRTUN && ire->ire_in_ill == NULL); 46017c478bd9Sstevel@tonic-gate if (ire_match_args(ire, addr, ire->ire_mask, gateway, type, 460245916cd2Sjpk ipif, zoneid, 0, tsl, flags)) { 46037c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 46047c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 46057c478bd9Sstevel@tonic-gate return (ire); 46067c478bd9Sstevel@tonic-gate } 46077c478bd9Sstevel@tonic-gate } 46087c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 46097c478bd9Sstevel@tonic-gate return (NULL); 46107c478bd9Sstevel@tonic-gate } 46117c478bd9Sstevel@tonic-gate 46127c478bd9Sstevel@tonic-gate /* 46135597b60aSnordmark * Check whether the IRE_LOCAL and the IRE potentially used to transmit 46145597b60aSnordmark * (could be an IRE_CACHE, IRE_BROADCAST, or IRE_INTERFACE) are part of 46155597b60aSnordmark * the same ill group. 46165597b60aSnordmark */ 46175597b60aSnordmark boolean_t 46185597b60aSnordmark ire_local_same_ill_group(ire_t *ire_local, ire_t *xmit_ire) 46195597b60aSnordmark { 46205597b60aSnordmark ill_t *recv_ill, *xmit_ill; 46215597b60aSnordmark ill_group_t *recv_group, *xmit_group; 46225597b60aSnordmark 46232bc4236aSnordmark ASSERT(ire_local->ire_type & (IRE_LOCAL|IRE_LOOPBACK)); 4624aa2eaee6Snordmark ASSERT(xmit_ire->ire_type & (IRE_CACHETABLE|IRE_INTERFACE)); 46255597b60aSnordmark 46262bc4236aSnordmark recv_ill = ire_to_ill(ire_local); 46272bc4236aSnordmark xmit_ill = ire_to_ill(xmit_ire); 46282bc4236aSnordmark 46292bc4236aSnordmark ASSERT(recv_ill != NULL); 46302bc4236aSnordmark ASSERT(xmit_ill != NULL); 46315597b60aSnordmark 46325597b60aSnordmark if (recv_ill == xmit_ill) 46335597b60aSnordmark return (B_TRUE); 46345597b60aSnordmark 46355597b60aSnordmark recv_group = recv_ill->ill_group; 46365597b60aSnordmark xmit_group = xmit_ill->ill_group; 46375597b60aSnordmark 46385597b60aSnordmark if (recv_group != NULL && recv_group == xmit_group) 46395597b60aSnordmark return (B_TRUE); 46405597b60aSnordmark 46415597b60aSnordmark return (B_FALSE); 46425597b60aSnordmark } 46435597b60aSnordmark 46445597b60aSnordmark /* 46455597b60aSnordmark * Check if the IRE_LOCAL uses the same ill (group) as another route would use. 4646aa2eaee6Snordmark * If there is no alternate route, or the alternate is a REJECT or BLACKHOLE, 4647aa2eaee6Snordmark * then we don't allow this IRE_LOCAL to be used. 46485597b60aSnordmark */ 46495597b60aSnordmark boolean_t 46505597b60aSnordmark ire_local_ok_across_zones(ire_t *ire_local, zoneid_t zoneid, void *addr, 4651*f4b3ec61Sdh155122 const ts_label_t *tsl, ip_stack_t *ipst) 46525597b60aSnordmark { 46535597b60aSnordmark ire_t *alt_ire; 46545597b60aSnordmark boolean_t rval; 46555597b60aSnordmark 46565597b60aSnordmark if (ire_local->ire_ipversion == IPV4_VERSION) { 46575597b60aSnordmark alt_ire = ire_ftable_lookup(*((ipaddr_t *)addr), 0, 0, 0, NULL, 46585597b60aSnordmark NULL, zoneid, 0, tsl, 46595597b60aSnordmark MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT | 4660*f4b3ec61Sdh155122 MATCH_IRE_RJ_BHOLE, ipst); 46615597b60aSnordmark } else { 46625597b60aSnordmark alt_ire = ire_ftable_lookup_v6((in6_addr_t *)addr, NULL, NULL, 46635597b60aSnordmark 0, NULL, NULL, zoneid, 0, tsl, 46645597b60aSnordmark MATCH_IRE_RECURSIVE | MATCH_IRE_DEFAULT | 4665*f4b3ec61Sdh155122 MATCH_IRE_RJ_BHOLE, ipst); 46665597b60aSnordmark } 46675597b60aSnordmark 46685597b60aSnordmark if (alt_ire == NULL) 46695597b60aSnordmark return (B_FALSE); 46705597b60aSnordmark 4671aa2eaee6Snordmark if (alt_ire->ire_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 4672aa2eaee6Snordmark ire_refrele(alt_ire); 4673aa2eaee6Snordmark return (B_FALSE); 4674aa2eaee6Snordmark } 46755597b60aSnordmark rval = ire_local_same_ill_group(ire_local, alt_ire); 46765597b60aSnordmark 46775597b60aSnordmark ire_refrele(alt_ire); 46785597b60aSnordmark return (rval); 46795597b60aSnordmark } 46805597b60aSnordmark 46815597b60aSnordmark /* 46827c478bd9Sstevel@tonic-gate * Lookup cache. Don't return IRE_MARK_HIDDEN entries. Callers 46837c478bd9Sstevel@tonic-gate * should use ire_ctable_lookup with MATCH_IRE_MARK_HIDDEN to get 46847c478bd9Sstevel@tonic-gate * to the hidden ones. 46855597b60aSnordmark * 46865597b60aSnordmark * In general the zoneid has to match (where ALL_ZONES match all of them). 46875597b60aSnordmark * But for IRE_LOCAL we also need to handle the case where L2 should 46885597b60aSnordmark * conceptually loop back the packet. This is necessary since neither 46895597b60aSnordmark * Ethernet drivers nor Ethernet hardware loops back packets sent to their 46905597b60aSnordmark * own MAC address. This loopback is needed when the normal 46915597b60aSnordmark * routes (ignoring IREs with different zoneids) would send out the packet on 46925597b60aSnordmark * the same ill (or ill group) as the ill with which this IRE_LOCAL is 46935597b60aSnordmark * associated. 46945597b60aSnordmark * 46955597b60aSnordmark * Earlier versions of this code always matched an IRE_LOCAL independently of 46965597b60aSnordmark * the zoneid. We preserve that earlier behavior when 46975597b60aSnordmark * ip_restrict_interzone_loopback is turned off. 46987c478bd9Sstevel@tonic-gate */ 46997c478bd9Sstevel@tonic-gate ire_t * 4700*f4b3ec61Sdh155122 ire_cache_lookup(ipaddr_t addr, zoneid_t zoneid, const ts_label_t *tsl, 4701*f4b3ec61Sdh155122 ip_stack_t *ipst) 47027c478bd9Sstevel@tonic-gate { 47037c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 47047c478bd9Sstevel@tonic-gate ire_t *ire; 47057c478bd9Sstevel@tonic-gate 4706*f4b3ec61Sdh155122 irb_ptr = &ipst->ips_ip_cache_table[IRE_ADDR_HASH(addr, 4707*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size)]; 47087c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_READER); 47097c478bd9Sstevel@tonic-gate for (ire = irb_ptr->irb_ire; ire != NULL; ire = ire->ire_next) { 471043d18f1cSpriyanka if (ire->ire_marks & (IRE_MARK_CONDEMNED | 471143d18f1cSpriyanka IRE_MARK_HIDDEN | IRE_MARK_PRIVATE_ADDR)) { 47127c478bd9Sstevel@tonic-gate continue; 471343d18f1cSpriyanka } 47147c478bd9Sstevel@tonic-gate if (ire->ire_addr == addr) { 471545916cd2Sjpk /* 471645916cd2Sjpk * Finally, check if the security policy has any 471745916cd2Sjpk * restriction on using this route for the specified 471845916cd2Sjpk * message. 471945916cd2Sjpk */ 472045916cd2Sjpk if (tsl != NULL && 472145916cd2Sjpk ire->ire_gw_secattr != NULL && 472245916cd2Sjpk tsol_ire_match_gwattr(ire, tsl) != 0) { 472345916cd2Sjpk continue; 472445916cd2Sjpk } 472545916cd2Sjpk 47267c478bd9Sstevel@tonic-gate if (zoneid == ALL_ZONES || ire->ire_zoneid == zoneid || 47275597b60aSnordmark ire->ire_zoneid == ALL_ZONES) { 47285597b60aSnordmark IRE_REFHOLD(ire); 47295597b60aSnordmark rw_exit(&irb_ptr->irb_lock); 47305597b60aSnordmark return (ire); 47315597b60aSnordmark } 47325597b60aSnordmark 47335597b60aSnordmark if (ire->ire_type == IRE_LOCAL) { 4734*f4b3ec61Sdh155122 if (ipst->ips_ip_restrict_interzone_loopback && 47355597b60aSnordmark !ire_local_ok_across_zones(ire, zoneid, 4736*f4b3ec61Sdh155122 &addr, tsl, ipst)) 47375597b60aSnordmark continue; 47385597b60aSnordmark 47397c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 47407c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 47417c478bd9Sstevel@tonic-gate return (ire); 47427c478bd9Sstevel@tonic-gate } 47437c478bd9Sstevel@tonic-gate } 47447c478bd9Sstevel@tonic-gate } 47457c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 47467c478bd9Sstevel@tonic-gate return (NULL); 47477c478bd9Sstevel@tonic-gate } 47487c478bd9Sstevel@tonic-gate 47497c478bd9Sstevel@tonic-gate /* 47507c478bd9Sstevel@tonic-gate * Locate the interface ire that is tied to the cache ire 'cire' via 47517c478bd9Sstevel@tonic-gate * cire->ire_ihandle. 47527c478bd9Sstevel@tonic-gate * 47537c478bd9Sstevel@tonic-gate * We are trying to create the cache ire for an offlink destn based 47547c478bd9Sstevel@tonic-gate * on the cache ire of the gateway in 'cire'. 'pire' is the prefix ire 47557c478bd9Sstevel@tonic-gate * as found by ip_newroute(). We are called from ip_newroute() in 47567c478bd9Sstevel@tonic-gate * the IRE_CACHE case. 47577c478bd9Sstevel@tonic-gate */ 47587c478bd9Sstevel@tonic-gate ire_t * 47597c478bd9Sstevel@tonic-gate ire_ihandle_lookup_offlink(ire_t *cire, ire_t *pire) 47607c478bd9Sstevel@tonic-gate { 47617c478bd9Sstevel@tonic-gate ire_t *ire; 47627c478bd9Sstevel@tonic-gate int match_flags; 47637c478bd9Sstevel@tonic-gate ipaddr_t gw_addr; 47647c478bd9Sstevel@tonic-gate ipif_t *gw_ipif; 4765*f4b3ec61Sdh155122 ip_stack_t *ipst = cire->ire_ipst; 47667c478bd9Sstevel@tonic-gate 47677c478bd9Sstevel@tonic-gate ASSERT(cire != NULL && pire != NULL); 47687c478bd9Sstevel@tonic-gate 47697c478bd9Sstevel@tonic-gate /* 47707c478bd9Sstevel@tonic-gate * We don't need to specify the zoneid to ire_ftable_lookup() below 47717c478bd9Sstevel@tonic-gate * because the ihandle refers to an ipif which can be in only one zone. 47727c478bd9Sstevel@tonic-gate */ 47737c478bd9Sstevel@tonic-gate match_flags = MATCH_IRE_TYPE | MATCH_IRE_IHANDLE | MATCH_IRE_MASK; 47747c478bd9Sstevel@tonic-gate /* 47757c478bd9Sstevel@tonic-gate * ip_newroute calls ire_ftable_lookup with MATCH_IRE_ILL only 47767c478bd9Sstevel@tonic-gate * for on-link hosts. We should never be here for onlink. 47777c478bd9Sstevel@tonic-gate * Thus, use MATCH_IRE_ILL_GROUP. 47787c478bd9Sstevel@tonic-gate */ 47797c478bd9Sstevel@tonic-gate if (pire->ire_ipif != NULL) 47807c478bd9Sstevel@tonic-gate match_flags |= MATCH_IRE_ILL_GROUP; 47817c478bd9Sstevel@tonic-gate /* 47827c478bd9Sstevel@tonic-gate * We know that the mask of the interface ire equals cire->ire_cmask. 47837c478bd9Sstevel@tonic-gate * (When ip_newroute() created 'cire' for the gateway it set its 47847c478bd9Sstevel@tonic-gate * cmask from the interface ire's mask) 47857c478bd9Sstevel@tonic-gate */ 47867c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(cire->ire_addr, cire->ire_cmask, 0, 47877c478bd9Sstevel@tonic-gate IRE_INTERFACE, pire->ire_ipif, NULL, ALL_ZONES, cire->ire_ihandle, 4788*f4b3ec61Sdh155122 NULL, match_flags, ipst); 47897c478bd9Sstevel@tonic-gate if (ire != NULL) 47907c478bd9Sstevel@tonic-gate return (ire); 47917c478bd9Sstevel@tonic-gate /* 47927c478bd9Sstevel@tonic-gate * If we didn't find an interface ire above, we can't declare failure. 47937c478bd9Sstevel@tonic-gate * For backwards compatibility, we need to support prefix routes 47947c478bd9Sstevel@tonic-gate * pointing to next hop gateways that are not on-link. 47957c478bd9Sstevel@tonic-gate * 47967c478bd9Sstevel@tonic-gate * Assume we are trying to ping some offlink destn, and we have the 47977c478bd9Sstevel@tonic-gate * routing table below. 47987c478bd9Sstevel@tonic-gate * 47997c478bd9Sstevel@tonic-gate * Eg. default - gw1 <--- pire (line 1) 48007c478bd9Sstevel@tonic-gate * gw1 - gw2 (line 2) 48017c478bd9Sstevel@tonic-gate * gw2 - hme0 (line 3) 48027c478bd9Sstevel@tonic-gate * 48037c478bd9Sstevel@tonic-gate * If we already have a cache ire for gw1 in 'cire', the 48047c478bd9Sstevel@tonic-gate * ire_ftable_lookup above would have failed, since there is no 48057c478bd9Sstevel@tonic-gate * interface ire to reach gw1. We will fallthru below. 48067c478bd9Sstevel@tonic-gate * 48077c478bd9Sstevel@tonic-gate * Here we duplicate the steps that ire_ftable_lookup() did in 48087c478bd9Sstevel@tonic-gate * getting 'cire' from 'pire', in the MATCH_IRE_RECURSIVE case. 48097c478bd9Sstevel@tonic-gate * The differences are the following 48107c478bd9Sstevel@tonic-gate * i. We want the interface ire only, so we call ire_ftable_lookup() 48117c478bd9Sstevel@tonic-gate * instead of ire_route_lookup() 48127c478bd9Sstevel@tonic-gate * ii. We look for only prefix routes in the 1st call below. 48137c478bd9Sstevel@tonic-gate * ii. We want to match on the ihandle in the 2nd call below. 48147c478bd9Sstevel@tonic-gate */ 48157c478bd9Sstevel@tonic-gate match_flags = MATCH_IRE_TYPE; 48167c478bd9Sstevel@tonic-gate if (pire->ire_ipif != NULL) 48177c478bd9Sstevel@tonic-gate match_flags |= MATCH_IRE_ILL_GROUP; 48187c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(pire->ire_gateway_addr, 0, 0, IRE_OFFSUBNET, 4819*f4b3ec61Sdh155122 pire->ire_ipif, NULL, ALL_ZONES, 0, NULL, match_flags, ipst); 48207c478bd9Sstevel@tonic-gate if (ire == NULL) 48217c478bd9Sstevel@tonic-gate return (NULL); 48227c478bd9Sstevel@tonic-gate /* 48237c478bd9Sstevel@tonic-gate * At this point 'ire' corresponds to the entry shown in line 2. 48247c478bd9Sstevel@tonic-gate * gw_addr is 'gw2' in the example above. 48257c478bd9Sstevel@tonic-gate */ 48267c478bd9Sstevel@tonic-gate gw_addr = ire->ire_gateway_addr; 48277c478bd9Sstevel@tonic-gate gw_ipif = ire->ire_ipif; 48287c478bd9Sstevel@tonic-gate ire_refrele(ire); 48297c478bd9Sstevel@tonic-gate 48307c478bd9Sstevel@tonic-gate match_flags |= MATCH_IRE_IHANDLE; 48317c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(gw_addr, 0, 0, IRE_INTERFACE, 4832*f4b3ec61Sdh155122 gw_ipif, NULL, ALL_ZONES, cire->ire_ihandle, NULL, match_flags, 4833*f4b3ec61Sdh155122 ipst); 48347c478bd9Sstevel@tonic-gate return (ire); 48357c478bd9Sstevel@tonic-gate } 48367c478bd9Sstevel@tonic-gate 48377c478bd9Sstevel@tonic-gate /* 48387c478bd9Sstevel@tonic-gate * ire_mrtun_lookup() is called by ip_rput() when packet is to be 48397c478bd9Sstevel@tonic-gate * tunneled through reverse tunnel. This is only supported for 48407c478bd9Sstevel@tonic-gate * IPv4 packets 48417c478bd9Sstevel@tonic-gate */ 48427c478bd9Sstevel@tonic-gate 48437c478bd9Sstevel@tonic-gate ire_t * 48447c478bd9Sstevel@tonic-gate ire_mrtun_lookup(ipaddr_t srcaddr, ill_t *ill) 48457c478bd9Sstevel@tonic-gate { 48467c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 48477c478bd9Sstevel@tonic-gate ire_t *ire; 4848*f4b3ec61Sdh155122 ip_stack_t *ipst = ill->ill_ipst; 48497c478bd9Sstevel@tonic-gate 48507c478bd9Sstevel@tonic-gate ASSERT(ill != NULL); 48517c478bd9Sstevel@tonic-gate ASSERT(!(ill->ill_isv6)); 48527c478bd9Sstevel@tonic-gate 4853*f4b3ec61Sdh155122 if (ipst->ips_ip_mrtun_table == NULL) 48547c478bd9Sstevel@tonic-gate return (NULL); 4855*f4b3ec61Sdh155122 irb_ptr = &ipst->ips_ip_mrtun_table[IRE_ADDR_HASH(srcaddr, 4856*f4b3ec61Sdh155122 IP_MRTUN_TABLE_SIZE)]; 48577c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_READER); 48587c478bd9Sstevel@tonic-gate for (ire = irb_ptr->irb_ire; ire != NULL; ire = ire->ire_next) { 48597c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) 48607c478bd9Sstevel@tonic-gate continue; 48617c478bd9Sstevel@tonic-gate if ((ire->ire_in_src_addr == srcaddr) && 48627c478bd9Sstevel@tonic-gate ire->ire_in_ill == ill) { 48637c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 48647c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 48657c478bd9Sstevel@tonic-gate return (ire); 48667c478bd9Sstevel@tonic-gate } 48677c478bd9Sstevel@tonic-gate } 48687c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 48697c478bd9Sstevel@tonic-gate return (NULL); 48707c478bd9Sstevel@tonic-gate } 48717c478bd9Sstevel@tonic-gate 48727c478bd9Sstevel@tonic-gate /* 48737c478bd9Sstevel@tonic-gate * Return the IRE_LOOPBACK, IRE_IF_RESOLVER or IRE_IF_NORESOLVER 48747c478bd9Sstevel@tonic-gate * ire associated with the specified ipif. 48757c478bd9Sstevel@tonic-gate * 48767c478bd9Sstevel@tonic-gate * This might occasionally be called when IPIF_UP is not set since 48777c478bd9Sstevel@tonic-gate * the IP_MULTICAST_IF as well as creating interface routes 48787c478bd9Sstevel@tonic-gate * allows specifying a down ipif (ipif_lookup* match ipifs that are down). 48797c478bd9Sstevel@tonic-gate * 48807c478bd9Sstevel@tonic-gate * Note that if IPIF_NOLOCAL, IPIF_NOXMIT, or IPIF_DEPRECATED is set on 48817c478bd9Sstevel@tonic-gate * the ipif, this routine might return NULL. 48827c478bd9Sstevel@tonic-gate */ 48837c478bd9Sstevel@tonic-gate ire_t * 488445916cd2Sjpk ipif_to_ire(const ipif_t *ipif) 48857c478bd9Sstevel@tonic-gate { 48867c478bd9Sstevel@tonic-gate ire_t *ire; 4887*f4b3ec61Sdh155122 ip_stack_t *ipst = ipif->ipif_ill->ill_ipst; 48887c478bd9Sstevel@tonic-gate 48897c478bd9Sstevel@tonic-gate ASSERT(!ipif->ipif_isv6); 48907c478bd9Sstevel@tonic-gate if (ipif->ipif_ire_type == IRE_LOOPBACK) { 48917c478bd9Sstevel@tonic-gate ire = ire_ctable_lookup(ipif->ipif_lcl_addr, 0, IRE_LOOPBACK, 4892*f4b3ec61Sdh155122 ipif, ALL_ZONES, NULL, (MATCH_IRE_TYPE | MATCH_IRE_IPIF), 4893*f4b3ec61Sdh155122 ipst); 48947c478bd9Sstevel@tonic-gate } else if (ipif->ipif_flags & IPIF_POINTOPOINT) { 48957c478bd9Sstevel@tonic-gate /* In this case we need to lookup destination address. */ 48967c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(ipif->ipif_pp_dst_addr, IP_HOST_MASK, 0, 489745916cd2Sjpk IRE_INTERFACE, ipif, NULL, ALL_ZONES, 0, NULL, 4898*f4b3ec61Sdh155122 (MATCH_IRE_TYPE | MATCH_IRE_IPIF | MATCH_IRE_MASK), ipst); 48997c478bd9Sstevel@tonic-gate } else { 49007c478bd9Sstevel@tonic-gate ire = ire_ftable_lookup(ipif->ipif_subnet, 49017c478bd9Sstevel@tonic-gate ipif->ipif_net_mask, 0, IRE_INTERFACE, ipif, NULL, 490245916cd2Sjpk ALL_ZONES, 0, NULL, (MATCH_IRE_TYPE | MATCH_IRE_IPIF | 4903*f4b3ec61Sdh155122 MATCH_IRE_MASK), ipst); 49047c478bd9Sstevel@tonic-gate } 49057c478bd9Sstevel@tonic-gate return (ire); 49067c478bd9Sstevel@tonic-gate } 49077c478bd9Sstevel@tonic-gate 49087c478bd9Sstevel@tonic-gate /* 49097c478bd9Sstevel@tonic-gate * ire_walk function. 49107c478bd9Sstevel@tonic-gate * Count the number of IRE_CACHE entries in different categories. 49117c478bd9Sstevel@tonic-gate */ 49127c478bd9Sstevel@tonic-gate void 49137c478bd9Sstevel@tonic-gate ire_cache_count(ire_t *ire, char *arg) 49147c478bd9Sstevel@tonic-gate { 49157c478bd9Sstevel@tonic-gate ire_cache_count_t *icc = (ire_cache_count_t *)arg; 49167c478bd9Sstevel@tonic-gate 49177c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 49187c478bd9Sstevel@tonic-gate return; 49197c478bd9Sstevel@tonic-gate 49207c478bd9Sstevel@tonic-gate icc->icc_total++; 49217c478bd9Sstevel@tonic-gate 49227c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 49237c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 49247c478bd9Sstevel@tonic-gate if (IN6_IS_ADDR_UNSPECIFIED(&ire->ire_gateway_addr_v6)) { 49257c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 49267c478bd9Sstevel@tonic-gate icc->icc_onlink++; 49277c478bd9Sstevel@tonic-gate return; 49287c478bd9Sstevel@tonic-gate } 49297c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 49307c478bd9Sstevel@tonic-gate } else { 49317c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == 0) { 49327c478bd9Sstevel@tonic-gate icc->icc_onlink++; 49337c478bd9Sstevel@tonic-gate return; 49347c478bd9Sstevel@tonic-gate } 49357c478bd9Sstevel@tonic-gate } 49367c478bd9Sstevel@tonic-gate 49377c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipif != NULL); 49387c478bd9Sstevel@tonic-gate if (ire->ire_max_frag < ire->ire_ipif->ipif_mtu) 49397c478bd9Sstevel@tonic-gate icc->icc_pmtu++; 49407c478bd9Sstevel@tonic-gate else if (ire->ire_tire_mark != ire->ire_ob_pkt_count + 49417c478bd9Sstevel@tonic-gate ire->ire_ib_pkt_count) 49427c478bd9Sstevel@tonic-gate icc->icc_offlink++; 49437c478bd9Sstevel@tonic-gate else 49447c478bd9Sstevel@tonic-gate icc->icc_unused++; 49457c478bd9Sstevel@tonic-gate } 49467c478bd9Sstevel@tonic-gate 49477c478bd9Sstevel@tonic-gate /* 49487c478bd9Sstevel@tonic-gate * ire_walk function called by ip_trash_ire_reclaim(). 49497c478bd9Sstevel@tonic-gate * Free a fraction of the IRE_CACHE cache entries. The fractions are 49507c478bd9Sstevel@tonic-gate * different for different categories of IRE_CACHE entries. 49517c478bd9Sstevel@tonic-gate * A fraction of zero means to not free any in that category. 49527c478bd9Sstevel@tonic-gate * Use the hash bucket id plus lbolt as a random number. Thus if the fraction 49537c478bd9Sstevel@tonic-gate * is N then every Nth hash bucket chain will be freed. 49547c478bd9Sstevel@tonic-gate */ 49557c478bd9Sstevel@tonic-gate void 49567c478bd9Sstevel@tonic-gate ire_cache_reclaim(ire_t *ire, char *arg) 49577c478bd9Sstevel@tonic-gate { 49587c478bd9Sstevel@tonic-gate ire_cache_reclaim_t *icr = (ire_cache_reclaim_t *)arg; 49597c478bd9Sstevel@tonic-gate uint_t rand; 4960*f4b3ec61Sdh155122 ip_stack_t *ipst = icr->icr_ipst; 49617c478bd9Sstevel@tonic-gate 49627c478bd9Sstevel@tonic-gate if (ire->ire_type != IRE_CACHE) 49637c478bd9Sstevel@tonic-gate return; 49647c478bd9Sstevel@tonic-gate 49657c478bd9Sstevel@tonic-gate if (ire->ire_ipversion == IPV6_VERSION) { 49667c478bd9Sstevel@tonic-gate rand = (uint_t)lbolt + 4967*f4b3ec61Sdh155122 IRE_ADDR_HASH_V6(ire->ire_addr_v6, 4968*f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size); 49697c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 49707c478bd9Sstevel@tonic-gate if (IN6_IS_ADDR_UNSPECIFIED(&ire->ire_gateway_addr_v6)) { 49717c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 49727c478bd9Sstevel@tonic-gate if (icr->icr_onlink != 0 && 49737c478bd9Sstevel@tonic-gate (rand/icr->icr_onlink)*icr->icr_onlink == rand) { 49747c478bd9Sstevel@tonic-gate ire_delete(ire); 49757c478bd9Sstevel@tonic-gate return; 49767c478bd9Sstevel@tonic-gate } 49777c478bd9Sstevel@tonic-gate goto done; 49787c478bd9Sstevel@tonic-gate } 49797c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 49807c478bd9Sstevel@tonic-gate } else { 49817c478bd9Sstevel@tonic-gate rand = (uint_t)lbolt + 4982*f4b3ec61Sdh155122 IRE_ADDR_HASH(ire->ire_addr, ipst->ips_ip_cache_table_size); 49837c478bd9Sstevel@tonic-gate if (ire->ire_gateway_addr == 0) { 49847c478bd9Sstevel@tonic-gate if (icr->icr_onlink != 0 && 49857c478bd9Sstevel@tonic-gate (rand/icr->icr_onlink)*icr->icr_onlink == rand) { 49867c478bd9Sstevel@tonic-gate ire_delete(ire); 49877c478bd9Sstevel@tonic-gate return; 49887c478bd9Sstevel@tonic-gate } 49897c478bd9Sstevel@tonic-gate goto done; 49907c478bd9Sstevel@tonic-gate } 49917c478bd9Sstevel@tonic-gate } 49927c478bd9Sstevel@tonic-gate /* Not onlink IRE */ 49937c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipif != NULL); 49947c478bd9Sstevel@tonic-gate if (ire->ire_max_frag < ire->ire_ipif->ipif_mtu) { 49957c478bd9Sstevel@tonic-gate /* Use ptmu fraction */ 49967c478bd9Sstevel@tonic-gate if (icr->icr_pmtu != 0 && 49977c478bd9Sstevel@tonic-gate (rand/icr->icr_pmtu)*icr->icr_pmtu == rand) { 49987c478bd9Sstevel@tonic-gate ire_delete(ire); 49997c478bd9Sstevel@tonic-gate return; 50007c478bd9Sstevel@tonic-gate } 50017c478bd9Sstevel@tonic-gate } else if (ire->ire_tire_mark != ire->ire_ob_pkt_count + 50027c478bd9Sstevel@tonic-gate ire->ire_ib_pkt_count) { 50037c478bd9Sstevel@tonic-gate /* Use offlink fraction */ 50047c478bd9Sstevel@tonic-gate if (icr->icr_offlink != 0 && 50057c478bd9Sstevel@tonic-gate (rand/icr->icr_offlink)*icr->icr_offlink == rand) { 50067c478bd9Sstevel@tonic-gate ire_delete(ire); 50077c478bd9Sstevel@tonic-gate return; 50087c478bd9Sstevel@tonic-gate } 50097c478bd9Sstevel@tonic-gate } else { 50107c478bd9Sstevel@tonic-gate /* Use unused fraction */ 50117c478bd9Sstevel@tonic-gate if (icr->icr_unused != 0 && 50127c478bd9Sstevel@tonic-gate (rand/icr->icr_unused)*icr->icr_unused == rand) { 50137c478bd9Sstevel@tonic-gate ire_delete(ire); 50147c478bd9Sstevel@tonic-gate return; 50157c478bd9Sstevel@tonic-gate } 50167c478bd9Sstevel@tonic-gate } 50177c478bd9Sstevel@tonic-gate done: 50187c478bd9Sstevel@tonic-gate /* 50197c478bd9Sstevel@tonic-gate * Update tire_mark so that those that haven't been used since this 50207c478bd9Sstevel@tonic-gate * reclaim will be considered unused next time we reclaim. 50217c478bd9Sstevel@tonic-gate */ 50227c478bd9Sstevel@tonic-gate ire->ire_tire_mark = ire->ire_ob_pkt_count + ire->ire_ib_pkt_count; 50237c478bd9Sstevel@tonic-gate } 50247c478bd9Sstevel@tonic-gate 50257c478bd9Sstevel@tonic-gate static void 50267c478bd9Sstevel@tonic-gate power2_roundup(uint32_t *value) 50277c478bd9Sstevel@tonic-gate { 50287c478bd9Sstevel@tonic-gate int i; 50297c478bd9Sstevel@tonic-gate 50307c478bd9Sstevel@tonic-gate for (i = 1; i < 31; i++) { 50317c478bd9Sstevel@tonic-gate if (*value <= (1 << i)) 50327c478bd9Sstevel@tonic-gate break; 50337c478bd9Sstevel@tonic-gate } 50347c478bd9Sstevel@tonic-gate *value = (1 << i); 50357c478bd9Sstevel@tonic-gate } 50367c478bd9Sstevel@tonic-gate 5037*f4b3ec61Sdh155122 /* Global init for all zones */ 50387c478bd9Sstevel@tonic-gate void 5039*f4b3ec61Sdh155122 ip_ire_g_init() 50407c478bd9Sstevel@tonic-gate { 50417c478bd9Sstevel@tonic-gate /* 50427c478bd9Sstevel@tonic-gate * Create ire caches, ire_reclaim() 50437c478bd9Sstevel@tonic-gate * will give IRE_CACHE back to system when needed. 50447c478bd9Sstevel@tonic-gate * This needs to be done here before anything else, since 50457c478bd9Sstevel@tonic-gate * ire_add() expects the cache to be created. 50467c478bd9Sstevel@tonic-gate */ 50477c478bd9Sstevel@tonic-gate ire_cache = kmem_cache_create("ire_cache", 50487c478bd9Sstevel@tonic-gate sizeof (ire_t), 0, ip_ire_constructor, 50497c478bd9Sstevel@tonic-gate ip_ire_destructor, ip_trash_ire_reclaim, NULL, NULL, 0); 50507c478bd9Sstevel@tonic-gate 5051*f4b3ec61Sdh155122 rt_entry_cache = kmem_cache_create("rt_entry", 5052*f4b3ec61Sdh155122 sizeof (struct rt_entry), 0, NULL, NULL, NULL, NULL, NULL, 0); 5053*f4b3ec61Sdh155122 5054*f4b3ec61Sdh155122 /* 5055*f4b3ec61Sdh155122 * Have radix code setup kmem caches etc. 5056*f4b3ec61Sdh155122 */ 5057*f4b3ec61Sdh155122 rn_init(); 5058*f4b3ec61Sdh155122 } 5059*f4b3ec61Sdh155122 5060*f4b3ec61Sdh155122 void 5061*f4b3ec61Sdh155122 ip_ire_init(ip_stack_t *ipst) 5062*f4b3ec61Sdh155122 { 5063*f4b3ec61Sdh155122 int i; 5064*f4b3ec61Sdh155122 uint32_t mem_cnt; 5065*f4b3ec61Sdh155122 uint32_t cpu_cnt; 5066*f4b3ec61Sdh155122 uint32_t min_cnt; 5067*f4b3ec61Sdh155122 pgcnt_t mem_avail; 5068*f4b3ec61Sdh155122 5069*f4b3ec61Sdh155122 /* 5070*f4b3ec61Sdh155122 * ip_ire_max_bucket_cnt is sized below based on the memory 5071*f4b3ec61Sdh155122 * size and the cpu speed of the machine. This is upper 5072*f4b3ec61Sdh155122 * bounded by the compile time value of ip_ire_max_bucket_cnt 5073*f4b3ec61Sdh155122 * and is lower bounded by the compile time value of 5074*f4b3ec61Sdh155122 * ip_ire_min_bucket_cnt. Similar logic applies to 5075*f4b3ec61Sdh155122 * ip6_ire_max_bucket_cnt. 5076*f4b3ec61Sdh155122 * 5077*f4b3ec61Sdh155122 * We calculate this for each IP Instances in order to use 5078*f4b3ec61Sdh155122 * the kmem_avail and ip_ire_{min,max}_bucket_cnt that are 5079*f4b3ec61Sdh155122 * in effect when the zone is booted. 5080*f4b3ec61Sdh155122 */ 5081*f4b3ec61Sdh155122 mem_avail = kmem_avail(); 5082*f4b3ec61Sdh155122 mem_cnt = (mem_avail >> ip_ire_mem_ratio) / 5083*f4b3ec61Sdh155122 ip_cache_table_size / sizeof (ire_t); 5084*f4b3ec61Sdh155122 cpu_cnt = CPU->cpu_type_info.pi_clock >> ip_ire_cpu_ratio; 5085*f4b3ec61Sdh155122 5086*f4b3ec61Sdh155122 min_cnt = MIN(cpu_cnt, mem_cnt); 5087*f4b3ec61Sdh155122 if (min_cnt < ip_ire_min_bucket_cnt) 5088*f4b3ec61Sdh155122 min_cnt = ip_ire_min_bucket_cnt; 5089*f4b3ec61Sdh155122 if (ip_ire_max_bucket_cnt > min_cnt) { 5090*f4b3ec61Sdh155122 ip_ire_max_bucket_cnt = min_cnt; 5091*f4b3ec61Sdh155122 } 5092*f4b3ec61Sdh155122 5093*f4b3ec61Sdh155122 mem_cnt = (mem_avail >> ip_ire_mem_ratio) / 5094*f4b3ec61Sdh155122 ip6_cache_table_size / sizeof (ire_t); 5095*f4b3ec61Sdh155122 min_cnt = MIN(cpu_cnt, mem_cnt); 5096*f4b3ec61Sdh155122 if (min_cnt < ip6_ire_min_bucket_cnt) 5097*f4b3ec61Sdh155122 min_cnt = ip6_ire_min_bucket_cnt; 5098*f4b3ec61Sdh155122 if (ip6_ire_max_bucket_cnt > min_cnt) { 5099*f4b3ec61Sdh155122 ip6_ire_max_bucket_cnt = min_cnt; 5100*f4b3ec61Sdh155122 } 5101*f4b3ec61Sdh155122 5102*f4b3ec61Sdh155122 mutex_init(&ipst->ips_ire_ft_init_lock, NULL, MUTEX_DEFAULT, 0); 5103*f4b3ec61Sdh155122 mutex_init(&ipst->ips_ire_handle_lock, NULL, MUTEX_DEFAULT, NULL); 5104*f4b3ec61Sdh155122 mutex_init(&ipst->ips_ire_mrtun_lock, NULL, MUTEX_DEFAULT, NULL); 5105*f4b3ec61Sdh155122 mutex_init(&ipst->ips_ire_srcif_table_lock, NULL, MUTEX_DEFAULT, NULL); 5106*f4b3ec61Sdh155122 5107*f4b3ec61Sdh155122 (void) rn_inithead((void **)&ipst->ips_ip_ftable, 32); 5108*f4b3ec61Sdh155122 5109*f4b3ec61Sdh155122 5110*f4b3ec61Sdh155122 /* Calculate the IPv4 cache table size. */ 5111*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size = MAX(ip_cache_table_size, 5112*f4b3ec61Sdh155122 ((mem_avail >> ip_ire_mem_ratio) / sizeof (ire_t) / 5113*f4b3ec61Sdh155122 ip_ire_max_bucket_cnt)); 5114*f4b3ec61Sdh155122 if (ipst->ips_ip_cache_table_size > ip_max_cache_table_size) 5115*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size = ip_max_cache_table_size; 5116*f4b3ec61Sdh155122 /* 5117*f4b3ec61Sdh155122 * Make sure that the table size is always a power of 2. The 5118*f4b3ec61Sdh155122 * hash macro IRE_ADDR_HASH() depends on that. 5119*f4b3ec61Sdh155122 */ 5120*f4b3ec61Sdh155122 power2_roundup(&ipst->ips_ip_cache_table_size); 5121*f4b3ec61Sdh155122 5122*f4b3ec61Sdh155122 ipst->ips_ip_cache_table = kmem_zalloc(ipst->ips_ip_cache_table_size * 5123*f4b3ec61Sdh155122 sizeof (irb_t), KM_SLEEP); 5124*f4b3ec61Sdh155122 5125*f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 5126*f4b3ec61Sdh155122 rw_init(&ipst->ips_ip_cache_table[i].irb_lock, NULL, 5127*f4b3ec61Sdh155122 RW_DEFAULT, NULL); 5128*f4b3ec61Sdh155122 } 5129*f4b3ec61Sdh155122 5130*f4b3ec61Sdh155122 /* Calculate the IPv6 cache table size. */ 5131*f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size = MAX(ip6_cache_table_size, 5132*f4b3ec61Sdh155122 ((mem_avail >> ip_ire_mem_ratio) / sizeof (ire_t) / 5133*f4b3ec61Sdh155122 ip6_ire_max_bucket_cnt)); 5134*f4b3ec61Sdh155122 if (ipst->ips_ip6_cache_table_size > ip6_max_cache_table_size) 5135*f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size = ip6_max_cache_table_size; 5136*f4b3ec61Sdh155122 /* 5137*f4b3ec61Sdh155122 * Make sure that the table size is always a power of 2. The 5138*f4b3ec61Sdh155122 * hash macro IRE_ADDR_HASH_V6() depends on that. 5139*f4b3ec61Sdh155122 */ 5140*f4b3ec61Sdh155122 power2_roundup(&ipst->ips_ip6_cache_table_size); 5141*f4b3ec61Sdh155122 5142*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6 = kmem_zalloc( 5143*f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size * sizeof (irb_t), KM_SLEEP); 5144*f4b3ec61Sdh155122 5145*f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip6_cache_table_size; i++) { 5146*f4b3ec61Sdh155122 rw_init(&ipst->ips_ip_cache_table_v6[i].irb_lock, NULL, 5147*f4b3ec61Sdh155122 RW_DEFAULT, NULL); 5148*f4b3ec61Sdh155122 } 5149*f4b3ec61Sdh155122 51507c478bd9Sstevel@tonic-gate /* 51517c478bd9Sstevel@tonic-gate * Initialize ip_mrtun_table to NULL now, it will be 51527c478bd9Sstevel@tonic-gate * populated by ip_rt_add if reverse tunnel is created 51537c478bd9Sstevel@tonic-gate */ 5154*f4b3ec61Sdh155122 ipst->ips_ip_mrtun_table = NULL; 51557c478bd9Sstevel@tonic-gate 51567c478bd9Sstevel@tonic-gate /* 51577c478bd9Sstevel@tonic-gate * Make sure that the forwarding table size is a power of 2. 51587c478bd9Sstevel@tonic-gate * The IRE*_ADDR_HASH() macroes depend on that. 51597c478bd9Sstevel@tonic-gate */ 5160*f4b3ec61Sdh155122 ipst->ips_ip6_ftable_hash_size = ip6_ftable_hash_size; 5161*f4b3ec61Sdh155122 power2_roundup(&ipst->ips_ip6_ftable_hash_size); 5162*f4b3ec61Sdh155122 5163*f4b3ec61Sdh155122 ipst->ips_ire_handle = 1; 51647c478bd9Sstevel@tonic-gate } 51657c478bd9Sstevel@tonic-gate 51667c478bd9Sstevel@tonic-gate void 5167*f4b3ec61Sdh155122 ip_ire_g_fini(void) 5168*f4b3ec61Sdh155122 { 5169*f4b3ec61Sdh155122 kmem_cache_destroy(ire_cache); 5170*f4b3ec61Sdh155122 kmem_cache_destroy(rt_entry_cache); 5171*f4b3ec61Sdh155122 5172*f4b3ec61Sdh155122 rn_fini(); 5173*f4b3ec61Sdh155122 } 5174*f4b3ec61Sdh155122 5175*f4b3ec61Sdh155122 void 5176*f4b3ec61Sdh155122 ip_ire_fini(ip_stack_t *ipst) 51777c478bd9Sstevel@tonic-gate { 51787c478bd9Sstevel@tonic-gate int i; 51797c478bd9Sstevel@tonic-gate 5180*f4b3ec61Sdh155122 /* 5181*f4b3ec61Sdh155122 * Delete all IREs - assumes that the ill/ipifs have 5182*f4b3ec61Sdh155122 * been removed so what remains are just the ftable and IRE_CACHE. 5183*f4b3ec61Sdh155122 */ 5184*f4b3ec61Sdh155122 ire_walk_ill_mrtun(0, 0, ire_delete, NULL, NULL, ipst); 5185*f4b3ec61Sdh155122 ire_walk(ire_delete, NULL, ipst); 5186c793af95Ssangeeta 5187*f4b3ec61Sdh155122 rn_freehead(ipst->ips_ip_ftable); 5188*f4b3ec61Sdh155122 ipst->ips_ip_ftable = NULL; 51897c478bd9Sstevel@tonic-gate 5190*f4b3ec61Sdh155122 mutex_destroy(&ipst->ips_ire_ft_init_lock); 5191*f4b3ec61Sdh155122 mutex_destroy(&ipst->ips_ire_handle_lock); 5192*f4b3ec61Sdh155122 mutex_destroy(&ipst->ips_ire_mrtun_lock); 5193*f4b3ec61Sdh155122 mutex_destroy(&ipst->ips_ire_srcif_table_lock); 5194*f4b3ec61Sdh155122 5195*f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip_cache_table_size; i++) { 5196*f4b3ec61Sdh155122 ASSERT(ipst->ips_ip_cache_table[i].irb_ire == NULL); 5197*f4b3ec61Sdh155122 rw_destroy(&ipst->ips_ip_cache_table[i].irb_lock); 51987c478bd9Sstevel@tonic-gate } 5199*f4b3ec61Sdh155122 kmem_free(ipst->ips_ip_cache_table, 5200*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_size * sizeof (irb_t)); 5201*f4b3ec61Sdh155122 ipst->ips_ip_cache_table = NULL; 52027c478bd9Sstevel@tonic-gate 5203*f4b3ec61Sdh155122 for (i = 0; i < ipst->ips_ip6_cache_table_size; i++) { 5204*f4b3ec61Sdh155122 ASSERT(ipst->ips_ip_cache_table_v6[i].irb_ire == NULL); 5205*f4b3ec61Sdh155122 rw_destroy(&ipst->ips_ip_cache_table_v6[i].irb_lock); 52067c478bd9Sstevel@tonic-gate } 5207*f4b3ec61Sdh155122 kmem_free(ipst->ips_ip_cache_table_v6, 5208*f4b3ec61Sdh155122 ipst->ips_ip6_cache_table_size * sizeof (irb_t)); 5209*f4b3ec61Sdh155122 ipst->ips_ip_cache_table_v6 = NULL; 52107c478bd9Sstevel@tonic-gate 5211*f4b3ec61Sdh155122 if (ipst->ips_ip_mrtun_table != NULL) { 52127c478bd9Sstevel@tonic-gate for (i = 0; i < IP_MRTUN_TABLE_SIZE; i++) { 5213*f4b3ec61Sdh155122 ASSERT(ipst->ips_ip_mrtun_table[i].irb_ire == NULL); 5214*f4b3ec61Sdh155122 rw_destroy(&ipst->ips_ip_mrtun_table[i].irb_lock); 52157c478bd9Sstevel@tonic-gate } 5216*f4b3ec61Sdh155122 kmem_free(ipst->ips_ip_mrtun_table, 5217*f4b3ec61Sdh155122 IP_MRTUN_TABLE_SIZE * sizeof (irb_t)); 5218*f4b3ec61Sdh155122 ipst->ips_ip_mrtun_table = NULL; 52197c478bd9Sstevel@tonic-gate } 5220*f4b3ec61Sdh155122 5221*f4b3ec61Sdh155122 for (i = 0; i < IP6_MASK_TABLE_SIZE; i++) { 5222*f4b3ec61Sdh155122 irb_t *ptr; 5223*f4b3ec61Sdh155122 int j; 5224*f4b3ec61Sdh155122 5225*f4b3ec61Sdh155122 if ((ptr = ipst->ips_ip_forwarding_table_v6[i]) == NULL) 5226*f4b3ec61Sdh155122 continue; 5227*f4b3ec61Sdh155122 5228*f4b3ec61Sdh155122 for (j = 0; j < ipst->ips_ip6_ftable_hash_size; j++) { 5229*f4b3ec61Sdh155122 ASSERT(ptr[j].irb_ire == NULL); 5230*f4b3ec61Sdh155122 rw_destroy(&ptr[j].irb_lock); 5231*f4b3ec61Sdh155122 } 5232*f4b3ec61Sdh155122 mi_free(ptr); 5233*f4b3ec61Sdh155122 ipst->ips_ip_forwarding_table_v6[i] = NULL; 5234*f4b3ec61Sdh155122 } 52357c478bd9Sstevel@tonic-gate } 52367c478bd9Sstevel@tonic-gate 52377c478bd9Sstevel@tonic-gate int 52387c478bd9Sstevel@tonic-gate ire_add_mrtun(ire_t **ire_p, queue_t *q, mblk_t *mp, ipsq_func_t func) 52397c478bd9Sstevel@tonic-gate { 52407c478bd9Sstevel@tonic-gate ire_t *ire1; 52417c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 52427c478bd9Sstevel@tonic-gate ire_t **irep; 5243*f4b3ec61Sdh155122 ire_t *ire = *ire_p; 52447c478bd9Sstevel@tonic-gate int i; 52457c478bd9Sstevel@tonic-gate uint_t max_frag; 52467c478bd9Sstevel@tonic-gate ill_t *stq_ill; 52477c478bd9Sstevel@tonic-gate int error; 5248*f4b3ec61Sdh155122 ip_stack_t *ipst = ire->ire_ipst; 52497c478bd9Sstevel@tonic-gate 52507c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 52517c478bd9Sstevel@tonic-gate /* Is ip_mrtun_table empty ? */ 52527c478bd9Sstevel@tonic-gate 5253*f4b3ec61Sdh155122 if (ipst->ips_ip_mrtun_table == NULL) { 52547c478bd9Sstevel@tonic-gate /* create the mrtun table */ 5255*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_mrtun_lock); 5256*f4b3ec61Sdh155122 if (ipst->ips_ip_mrtun_table == NULL) { 5257*f4b3ec61Sdh155122 ipst->ips_ip_mrtun_table = kmem_zalloc( 5258*f4b3ec61Sdh155122 IP_MRTUN_TABLE_SIZE * sizeof (irb_t), KM_NOSLEEP); 52597c478bd9Sstevel@tonic-gate 5260*f4b3ec61Sdh155122 if (ipst->ips_ip_mrtun_table == NULL) { 52617c478bd9Sstevel@tonic-gate ip2dbg(("ire_add_mrtun: allocation failure\n")); 5262*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_mrtun_lock); 52637c478bd9Sstevel@tonic-gate ire_refrele(ire); 52647c478bd9Sstevel@tonic-gate *ire_p = NULL; 52657c478bd9Sstevel@tonic-gate return (ENOMEM); 52667c478bd9Sstevel@tonic-gate } 52677c478bd9Sstevel@tonic-gate 52687c478bd9Sstevel@tonic-gate for (i = 0; i < IP_MRTUN_TABLE_SIZE; i++) { 5269*f4b3ec61Sdh155122 rw_init(&ipst->ips_ip_mrtun_table[i].irb_lock, NULL, 52707c478bd9Sstevel@tonic-gate RW_DEFAULT, NULL); 52717c478bd9Sstevel@tonic-gate } 52727c478bd9Sstevel@tonic-gate ip2dbg(("ire_add_mrtun: mrtun table is created\n")); 52737c478bd9Sstevel@tonic-gate } 52747c478bd9Sstevel@tonic-gate /* some other thread got it and created the table */ 5275*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_mrtun_lock); 52767c478bd9Sstevel@tonic-gate } 52777c478bd9Sstevel@tonic-gate 52787c478bd9Sstevel@tonic-gate /* 52797c478bd9Sstevel@tonic-gate * Check for duplicate in the bucket and insert in the table 52807c478bd9Sstevel@tonic-gate */ 5281*f4b3ec61Sdh155122 irb_ptr = &(ipst->ips_ip_mrtun_table[IRE_ADDR_HASH(ire->ire_in_src_addr, 52827c478bd9Sstevel@tonic-gate IP_MRTUN_TABLE_SIZE)]); 52837c478bd9Sstevel@tonic-gate 52847c478bd9Sstevel@tonic-gate /* 52857c478bd9Sstevel@tonic-gate * Start the atomic add of the ire. Grab the ill locks, 52867c478bd9Sstevel@tonic-gate * ill_g_usesrc_lock and the bucket lock. 52877c478bd9Sstevel@tonic-gate * 52887c478bd9Sstevel@tonic-gate * If ipif or ill is changing ire_atomic_start() may queue the 52897c478bd9Sstevel@tonic-gate * request and return EINPROGRESS. 52907c478bd9Sstevel@tonic-gate */ 52917c478bd9Sstevel@tonic-gate error = ire_atomic_start(irb_ptr, ire, q, mp, func); 52927c478bd9Sstevel@tonic-gate if (error != 0) { 52937c478bd9Sstevel@tonic-gate /* 52947c478bd9Sstevel@tonic-gate * We don't know whether it is a valid ipif or not. 52957c478bd9Sstevel@tonic-gate * So, set it to NULL. This assumes that the ire has not added 52967c478bd9Sstevel@tonic-gate * a reference to the ipif. 52977c478bd9Sstevel@tonic-gate */ 52987c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 52997c478bd9Sstevel@tonic-gate ire_delete(ire); 53007c478bd9Sstevel@tonic-gate ip1dbg(("ire_add_mrtun: ire_atomic_start failed\n")); 53017c478bd9Sstevel@tonic-gate *ire_p = NULL; 53027c478bd9Sstevel@tonic-gate return (error); 53037c478bd9Sstevel@tonic-gate } 53047c478bd9Sstevel@tonic-gate for (ire1 = irb_ptr->irb_ire; ire1 != NULL; ire1 = ire1->ire_next) { 53057c478bd9Sstevel@tonic-gate if (ire1->ire_marks & IRE_MARK_CONDEMNED) 53067c478bd9Sstevel@tonic-gate continue; 53077c478bd9Sstevel@tonic-gate /* has anyone inserted the route in the meanwhile ? */ 53087c478bd9Sstevel@tonic-gate if (ire1->ire_in_ill == ire->ire_in_ill && 53097c478bd9Sstevel@tonic-gate ire1->ire_in_src_addr == ire->ire_in_src_addr) { 53107c478bd9Sstevel@tonic-gate ip1dbg(("ire_add_mrtun: Duplicate entry exists\n")); 53117c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire1); 53127c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 53137c478bd9Sstevel@tonic-gate ire_delete(ire); 53147c478bd9Sstevel@tonic-gate /* Return the old ire */ 53157c478bd9Sstevel@tonic-gate *ire_p = ire1; 53167c478bd9Sstevel@tonic-gate return (0); 53177c478bd9Sstevel@tonic-gate } 53187c478bd9Sstevel@tonic-gate } 53197c478bd9Sstevel@tonic-gate 53207c478bd9Sstevel@tonic-gate /* Atomically set the ire_max_frag */ 53217c478bd9Sstevel@tonic-gate max_frag = *ire->ire_max_fragp; 53227c478bd9Sstevel@tonic-gate ire->ire_max_fragp = NULL; 53237c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(max_frag, IP_MAXPACKET); 5324c793af95Ssangeeta ASSERT(ire->ire_type != IRE_CACHE); 53257c478bd9Sstevel@tonic-gate irep = (ire_t **)irb_ptr; 53267c478bd9Sstevel@tonic-gate if (*irep != NULL) { 53277c478bd9Sstevel@tonic-gate /* Find the last ire which matches ire_in_src_addr */ 53287c478bd9Sstevel@tonic-gate ire1 = *irep; 53297c478bd9Sstevel@tonic-gate while (ire1->ire_in_src_addr == ire->ire_in_src_addr) { 53307c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 53317c478bd9Sstevel@tonic-gate ire1 = *irep; 53327c478bd9Sstevel@tonic-gate if (ire1 == NULL) 53337c478bd9Sstevel@tonic-gate break; 53347c478bd9Sstevel@tonic-gate } 53357c478bd9Sstevel@tonic-gate } 53367c478bd9Sstevel@tonic-gate ire1 = *irep; 53377c478bd9Sstevel@tonic-gate if (ire1 != NULL) 53387c478bd9Sstevel@tonic-gate ire1->ire_ptpn = &ire->ire_next; 53397c478bd9Sstevel@tonic-gate ire->ire_next = ire1; 53407c478bd9Sstevel@tonic-gate /* Link the new one in. */ 53417c478bd9Sstevel@tonic-gate ire->ire_ptpn = irep; 53427c478bd9Sstevel@tonic-gate membar_producer(); 53437c478bd9Sstevel@tonic-gate *irep = ire; 53447c478bd9Sstevel@tonic-gate ire->ire_bucket = irb_ptr; 53457c478bd9Sstevel@tonic-gate IRE_REFHOLD_LOCKED(ire); 53467c478bd9Sstevel@tonic-gate 53477c478bd9Sstevel@tonic-gate ip2dbg(("ire_add_mrtun: created and linked ire %p\n", (void *)*irep)); 53487c478bd9Sstevel@tonic-gate 53497c478bd9Sstevel@tonic-gate /* 53507c478bd9Sstevel@tonic-gate * Protect ire_mrtun_count and ill_mrtun_refcnt from 53517c478bd9Sstevel@tonic-gate * another thread trying to add ire in the table 53527c478bd9Sstevel@tonic-gate */ 5353*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_mrtun_lock); 5354*f4b3ec61Sdh155122 ipst->ips_ire_mrtun_count++; 5355*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_mrtun_lock); 53567c478bd9Sstevel@tonic-gate /* 53577c478bd9Sstevel@tonic-gate * ill_mrtun_refcnt is protected by the ill_lock held via 53587c478bd9Sstevel@tonic-gate * ire_atomic_start 53597c478bd9Sstevel@tonic-gate */ 53607c478bd9Sstevel@tonic-gate ire->ire_in_ill->ill_mrtun_refcnt++; 53617c478bd9Sstevel@tonic-gate 53627c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 53637c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ire_cnt++; 53647c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) { 53657c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 53667c478bd9Sstevel@tonic-gate stq_ill->ill_ire_cnt++; 53677c478bd9Sstevel@tonic-gate } 53687c478bd9Sstevel@tonic-gate } else { 53697c478bd9Sstevel@tonic-gate ASSERT(ire->ire_stq == NULL); 53707c478bd9Sstevel@tonic-gate } 53717c478bd9Sstevel@tonic-gate 53727c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 5373516bda92Ssowmini nce_fastpath(ire->ire_nce); 53747c478bd9Sstevel@tonic-gate *ire_p = ire; 53757c478bd9Sstevel@tonic-gate return (0); 53767c478bd9Sstevel@tonic-gate } 53777c478bd9Sstevel@tonic-gate 53787c478bd9Sstevel@tonic-gate 53797c478bd9Sstevel@tonic-gate /* Walks down the mrtun table */ 53807c478bd9Sstevel@tonic-gate 53817c478bd9Sstevel@tonic-gate void 53827c478bd9Sstevel@tonic-gate ire_walk_ill_mrtun(uint_t match_flags, uint_t ire_type, pfv_t func, void *arg, 5383*f4b3ec61Sdh155122 ill_t *ill, ip_stack_t *ipst) 53847c478bd9Sstevel@tonic-gate { 53857c478bd9Sstevel@tonic-gate irb_t *irb; 53867c478bd9Sstevel@tonic-gate ire_t *ire; 53877c478bd9Sstevel@tonic-gate int i; 53887c478bd9Sstevel@tonic-gate int ret; 53897c478bd9Sstevel@tonic-gate 53907c478bd9Sstevel@tonic-gate ASSERT((!(match_flags & (MATCH_IRE_WQ | MATCH_IRE_ILL | 53917c478bd9Sstevel@tonic-gate MATCH_IRE_ILL_GROUP))) || (ill != NULL)); 53927c478bd9Sstevel@tonic-gate ASSERT(match_flags == 0 || ire_type == IRE_MIPRTUN); 53937c478bd9Sstevel@tonic-gate 5394*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_mrtun_lock); 5395*f4b3ec61Sdh155122 if (ipst->ips_ire_mrtun_count == 0) { 5396*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_mrtun_lock); 53977c478bd9Sstevel@tonic-gate return; 53987c478bd9Sstevel@tonic-gate } 5399*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_mrtun_lock); 54007c478bd9Sstevel@tonic-gate 54017c478bd9Sstevel@tonic-gate ip2dbg(("ire_walk_ill_mrtun:walking the reverse tunnel table \n")); 54027c478bd9Sstevel@tonic-gate for (i = 0; i < IP_MRTUN_TABLE_SIZE; i++) { 54037c478bd9Sstevel@tonic-gate 5404*f4b3ec61Sdh155122 irb = &(ipst->ips_ip_mrtun_table[i]); 54057c478bd9Sstevel@tonic-gate if (irb->irb_ire == NULL) 54067c478bd9Sstevel@tonic-gate continue; 54077c478bd9Sstevel@tonic-gate IRB_REFHOLD(irb); 54087c478bd9Sstevel@tonic-gate for (ire = irb->irb_ire; ire != NULL; 54097c478bd9Sstevel@tonic-gate ire = ire->ire_next) { 54107c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 54117c478bd9Sstevel@tonic-gate if (match_flags != 0) { 54127c478bd9Sstevel@tonic-gate ret = ire_walk_ill_match( 54137c478bd9Sstevel@tonic-gate match_flags, ire_type, 5414*f4b3ec61Sdh155122 ire, ill, ALL_ZONES, ipst); 54157c478bd9Sstevel@tonic-gate } 54167c478bd9Sstevel@tonic-gate if (match_flags == 0 || ret) 54177c478bd9Sstevel@tonic-gate (*func)(ire, arg); 54187c478bd9Sstevel@tonic-gate } 54197c478bd9Sstevel@tonic-gate IRB_REFRELE(irb); 54207c478bd9Sstevel@tonic-gate } 54217c478bd9Sstevel@tonic-gate } 54227c478bd9Sstevel@tonic-gate 54237c478bd9Sstevel@tonic-gate /* 54247c478bd9Sstevel@tonic-gate * Source interface based lookup routine (IPV4 only). 54257c478bd9Sstevel@tonic-gate * This routine is called only when RTA_SRCIFP bitflag is set 54267c478bd9Sstevel@tonic-gate * by routing socket while adding/deleting the route and it is 54277c478bd9Sstevel@tonic-gate * also called from ip_rput() when packets arrive from an interface 54287c478bd9Sstevel@tonic-gate * for which ill_srcif_ref_cnt is positive. This function is useful 54297c478bd9Sstevel@tonic-gate * when a packet coming from one interface must be forwarded to another 54307c478bd9Sstevel@tonic-gate * designated interface to reach the correct node. This function is also 54317c478bd9Sstevel@tonic-gate * called from ip_newroute when the link-layer address of an ire is resolved. 54327c478bd9Sstevel@tonic-gate * We need to make sure that ip_newroute searches for IRE_IF_RESOLVER type 54337c478bd9Sstevel@tonic-gate * ires--thus the ire_type parameter is needed. 54347c478bd9Sstevel@tonic-gate */ 54357c478bd9Sstevel@tonic-gate 54367c478bd9Sstevel@tonic-gate ire_t * 54377c478bd9Sstevel@tonic-gate ire_srcif_table_lookup(ipaddr_t dst_addr, int ire_type, ipif_t *ipif, 54387c478bd9Sstevel@tonic-gate ill_t *in_ill, int flags) 54397c478bd9Sstevel@tonic-gate { 54407c478bd9Sstevel@tonic-gate irb_t *irb_ptr; 54417c478bd9Sstevel@tonic-gate ire_t *ire; 54427c478bd9Sstevel@tonic-gate irb_t *ire_srcif_table; 54437c478bd9Sstevel@tonic-gate 54447c478bd9Sstevel@tonic-gate ASSERT(in_ill != NULL && !in_ill->ill_isv6); 54457c478bd9Sstevel@tonic-gate ASSERT(!(flags & (MATCH_IRE_ILL|MATCH_IRE_ILL_GROUP)) || 54467c478bd9Sstevel@tonic-gate (ipif != NULL && !ipif->ipif_isv6)); 54477c478bd9Sstevel@tonic-gate 54487c478bd9Sstevel@tonic-gate /* 54497c478bd9Sstevel@tonic-gate * No need to lock the ill since it is refheld by the caller of this 54507c478bd9Sstevel@tonic-gate * function 54517c478bd9Sstevel@tonic-gate */ 54527c478bd9Sstevel@tonic-gate if (in_ill->ill_srcif_table == NULL) { 54537c478bd9Sstevel@tonic-gate return (NULL); 54547c478bd9Sstevel@tonic-gate } 54557c478bd9Sstevel@tonic-gate 54567c478bd9Sstevel@tonic-gate if (!(flags & MATCH_IRE_TYPE)) { 54577c478bd9Sstevel@tonic-gate flags |= MATCH_IRE_TYPE; 54587c478bd9Sstevel@tonic-gate ire_type = IRE_INTERFACE; 54597c478bd9Sstevel@tonic-gate } 54607c478bd9Sstevel@tonic-gate ire_srcif_table = in_ill->ill_srcif_table; 54617c478bd9Sstevel@tonic-gate irb_ptr = &ire_srcif_table[IRE_ADDR_HASH(dst_addr, 54627c478bd9Sstevel@tonic-gate IP_SRCIF_TABLE_SIZE)]; 54637c478bd9Sstevel@tonic-gate rw_enter(&irb_ptr->irb_lock, RW_READER); 54647c478bd9Sstevel@tonic-gate for (ire = irb_ptr->irb_ire; ire != NULL; ire = ire->ire_next) { 54657c478bd9Sstevel@tonic-gate if (ire->ire_marks & IRE_MARK_CONDEMNED) 54667c478bd9Sstevel@tonic-gate continue; 54677c478bd9Sstevel@tonic-gate if (ire_match_args(ire, dst_addr, ire->ire_mask, 0, 546845916cd2Sjpk ire_type, ipif, ire->ire_zoneid, 0, NULL, flags)) { 54697c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire); 54707c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 54717c478bd9Sstevel@tonic-gate return (ire); 54727c478bd9Sstevel@tonic-gate } 54737c478bd9Sstevel@tonic-gate } 54747c478bd9Sstevel@tonic-gate /* Not Found */ 54757c478bd9Sstevel@tonic-gate rw_exit(&irb_ptr->irb_lock); 54767c478bd9Sstevel@tonic-gate return (NULL); 54777c478bd9Sstevel@tonic-gate } 54787c478bd9Sstevel@tonic-gate 54797c478bd9Sstevel@tonic-gate 54807c478bd9Sstevel@tonic-gate /* 54817c478bd9Sstevel@tonic-gate * Adds the ire into the special routing table which is hanging off of 54827c478bd9Sstevel@tonic-gate * the src_ipif->ipif_ill. It also increments the refcnt in the ill. 54837c478bd9Sstevel@tonic-gate * The forward table contains only IRE_IF_RESOLVER, IRE_IF_NORESOLVER 54847c478bd9Sstevel@tonic-gate * i,e. IRE_INTERFACE entries. Originally the dlureq_mp field is NULL 54857c478bd9Sstevel@tonic-gate * for IRE_IF_RESOLVER entry because we do not have the dst_addr's 54867c478bd9Sstevel@tonic-gate * link-layer address at the time of addition. 54877c478bd9Sstevel@tonic-gate * Upon resolving the address from ARP, dlureq_mp field is updated with 54887c478bd9Sstevel@tonic-gate * proper information in ire_update_srcif_v4. 54897c478bd9Sstevel@tonic-gate */ 54907c478bd9Sstevel@tonic-gate static int 54917c478bd9Sstevel@tonic-gate ire_add_srcif_v4(ire_t **ire_p, queue_t *q, mblk_t *mp, ipsq_func_t func) 54927c478bd9Sstevel@tonic-gate { 54937c478bd9Sstevel@tonic-gate ire_t *ire1; 54947c478bd9Sstevel@tonic-gate irb_t *ire_srcifp_table = NULL; 54957c478bd9Sstevel@tonic-gate irb_t *irb_ptr = NULL; 54967c478bd9Sstevel@tonic-gate ire_t **irep; 54977c478bd9Sstevel@tonic-gate ire_t *ire; 54987c478bd9Sstevel@tonic-gate int flags; 54997c478bd9Sstevel@tonic-gate int i; 55007c478bd9Sstevel@tonic-gate ill_t *stq_ill; 55017c478bd9Sstevel@tonic-gate uint_t max_frag; 55027c478bd9Sstevel@tonic-gate int error = 0; 5503*f4b3ec61Sdh155122 ip_stack_t *ipst; 55047c478bd9Sstevel@tonic-gate 55057c478bd9Sstevel@tonic-gate ire = *ire_p; 5506*f4b3ec61Sdh155122 ipst = ire->ire_ipst; 55077c478bd9Sstevel@tonic-gate ASSERT(ire->ire_in_ill != NULL); 55087c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 55097c478bd9Sstevel@tonic-gate ASSERT(ire->ire_type == IRE_IF_NORESOLVER || 55107c478bd9Sstevel@tonic-gate ire->ire_type == IRE_IF_RESOLVER); 55117c478bd9Sstevel@tonic-gate 55127c478bd9Sstevel@tonic-gate ire->ire_mask = IP_HOST_MASK; 5513c793af95Ssangeeta /* 5514c793af95Ssangeeta * Update ire_nce->nce_res_mp with NULL value upon creation; 5515c793af95Ssangeeta * first free the default res_mp created by ire_nce_init. 5516c793af95Ssangeeta */ 5517c793af95Ssangeeta freeb(ire->ire_nce->nce_res_mp); 55187c478bd9Sstevel@tonic-gate if (ire->ire_type == IRE_IF_RESOLVER) { 55197c478bd9Sstevel@tonic-gate /* 55207c478bd9Sstevel@tonic-gate * assign NULL now, it will be updated 55217c478bd9Sstevel@tonic-gate * with correct value upon returning from 55227c478bd9Sstevel@tonic-gate * ARP 55237c478bd9Sstevel@tonic-gate */ 5524c793af95Ssangeeta ire->ire_nce->nce_res_mp = NULL; 55257c478bd9Sstevel@tonic-gate } else { 5526c793af95Ssangeeta ire->ire_nce->nce_res_mp = ill_dlur_gen(NULL, 55277c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ill->ill_phys_addr_length, 55287c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ill->ill_sap, 55297c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ill->ill_sap_length); 55307c478bd9Sstevel@tonic-gate } 55317c478bd9Sstevel@tonic-gate /* Make sure the address is properly masked. */ 55327c478bd9Sstevel@tonic-gate ire->ire_addr &= ire->ire_mask; 55337c478bd9Sstevel@tonic-gate 55347c478bd9Sstevel@tonic-gate ASSERT(ire->ire_max_fragp != NULL); 55357c478bd9Sstevel@tonic-gate max_frag = *ire->ire_max_fragp; 55367c478bd9Sstevel@tonic-gate ire->ire_max_fragp = NULL; 55377c478bd9Sstevel@tonic-gate ire->ire_max_frag = MIN(max_frag, IP_MAXPACKET); 55387c478bd9Sstevel@tonic-gate 55397c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_in_ill->ill_lock); 55407c478bd9Sstevel@tonic-gate if (ire->ire_in_ill->ill_srcif_table == NULL) { 55417c478bd9Sstevel@tonic-gate /* create the incoming interface based table */ 5542*f4b3ec61Sdh155122 ire->ire_in_ill->ill_srcif_table = kmem_zalloc( 5543*f4b3ec61Sdh155122 IP_SRCIF_TABLE_SIZE * sizeof (irb_t), KM_NOSLEEP); 55447c478bd9Sstevel@tonic-gate if (ire->ire_in_ill->ill_srcif_table == NULL) { 55457c478bd9Sstevel@tonic-gate ip1dbg(("ire_add_srcif_v4: Allocation fail\n")); 55467c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_in_ill->ill_lock); 55477c478bd9Sstevel@tonic-gate ire_delete(ire); 55487c478bd9Sstevel@tonic-gate *ire_p = NULL; 55497c478bd9Sstevel@tonic-gate return (ENOMEM); 55507c478bd9Sstevel@tonic-gate } 55517c478bd9Sstevel@tonic-gate ire_srcifp_table = ire->ire_in_ill->ill_srcif_table; 55527c478bd9Sstevel@tonic-gate for (i = 0; i < IP_SRCIF_TABLE_SIZE; i++) { 55537c478bd9Sstevel@tonic-gate rw_init(&ire_srcifp_table[i].irb_lock, NULL, 55547c478bd9Sstevel@tonic-gate RW_DEFAULT, NULL); 55557c478bd9Sstevel@tonic-gate } 55567c478bd9Sstevel@tonic-gate ip2dbg(("ire_add_srcif_v4: table created for ill %p\n", 55577c478bd9Sstevel@tonic-gate (void *)ire->ire_in_ill)); 55587c478bd9Sstevel@tonic-gate } 55597c478bd9Sstevel@tonic-gate /* Check for duplicate and insert */ 55607c478bd9Sstevel@tonic-gate ASSERT(ire->ire_in_ill->ill_srcif_table != NULL); 55617c478bd9Sstevel@tonic-gate irb_ptr = 55627c478bd9Sstevel@tonic-gate &(ire->ire_in_ill->ill_srcif_table[IRE_ADDR_HASH(ire->ire_addr, 55637c478bd9Sstevel@tonic-gate IP_SRCIF_TABLE_SIZE)]); 55647c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_in_ill->ill_lock); 55657c478bd9Sstevel@tonic-gate flags = (MATCH_IRE_MASK | MATCH_IRE_TYPE | MATCH_IRE_GW); 55667c478bd9Sstevel@tonic-gate flags |= MATCH_IRE_IPIF; 55677c478bd9Sstevel@tonic-gate 55687c478bd9Sstevel@tonic-gate /* 55697c478bd9Sstevel@tonic-gate * Start the atomic add of the ire. Grab the ill locks, 55707c478bd9Sstevel@tonic-gate * ill_g_usesrc_lock and the bucket lock. 55717c478bd9Sstevel@tonic-gate * 55727c478bd9Sstevel@tonic-gate * If ipif or ill is changing ire_atomic_start() may queue the 55737c478bd9Sstevel@tonic-gate * request and return EINPROGRESS. 55747c478bd9Sstevel@tonic-gate */ 55757c478bd9Sstevel@tonic-gate error = ire_atomic_start(irb_ptr, ire, q, mp, func); 55767c478bd9Sstevel@tonic-gate if (error != 0) { 55777c478bd9Sstevel@tonic-gate /* 55787c478bd9Sstevel@tonic-gate * We don't know whether it is a valid ipif or not. 55797c478bd9Sstevel@tonic-gate * So, set it to NULL. This assumes that the ire has not added 55807c478bd9Sstevel@tonic-gate * a reference to the ipif. 55817c478bd9Sstevel@tonic-gate */ 55827c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 55837c478bd9Sstevel@tonic-gate ire_delete(ire); 55847c478bd9Sstevel@tonic-gate ip1dbg(("ire_add_srcif_v4: ire_atomic_start failed\n")); 55857c478bd9Sstevel@tonic-gate *ire_p = NULL; 55867c478bd9Sstevel@tonic-gate return (error); 55877c478bd9Sstevel@tonic-gate } 55887c478bd9Sstevel@tonic-gate for (ire1 = irb_ptr->irb_ire; ire1 != NULL; ire1 = ire1->ire_next) { 55897c478bd9Sstevel@tonic-gate if (ire1->ire_marks & IRE_MARK_CONDEMNED) 55907c478bd9Sstevel@tonic-gate continue; 55917c478bd9Sstevel@tonic-gate if (ire1->ire_zoneid != ire->ire_zoneid) 55927c478bd9Sstevel@tonic-gate continue; 55937c478bd9Sstevel@tonic-gate /* Has anyone inserted route in the meanwhile ? */ 55947c478bd9Sstevel@tonic-gate if (ire_match_args(ire1, ire->ire_addr, ire->ire_mask, 0, 559545916cd2Sjpk ire->ire_type, ire->ire_ipif, ire->ire_zoneid, 0, NULL, 559645916cd2Sjpk flags)) { 55977c478bd9Sstevel@tonic-gate ip1dbg(("ire_add_srcif_v4 : Duplicate entry exists\n")); 55987c478bd9Sstevel@tonic-gate IRE_REFHOLD(ire1); 55997c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 56007c478bd9Sstevel@tonic-gate ire_delete(ire); 56017c478bd9Sstevel@tonic-gate /* Return old ire as in ire_add_v4 */ 56027c478bd9Sstevel@tonic-gate *ire_p = ire1; 56037c478bd9Sstevel@tonic-gate return (0); 56047c478bd9Sstevel@tonic-gate } 56057c478bd9Sstevel@tonic-gate } 56067c478bd9Sstevel@tonic-gate irep = (ire_t **)irb_ptr; 56077c478bd9Sstevel@tonic-gate if (*irep != NULL) { 56087c478bd9Sstevel@tonic-gate /* Find the last ire which matches ire_addr */ 56097c478bd9Sstevel@tonic-gate ire1 = *irep; 56107c478bd9Sstevel@tonic-gate while (ire1->ire_addr == ire->ire_addr) { 56117c478bd9Sstevel@tonic-gate irep = &ire1->ire_next; 56127c478bd9Sstevel@tonic-gate ire1 = *irep; 56137c478bd9Sstevel@tonic-gate if (ire1 == NULL) 56147c478bd9Sstevel@tonic-gate break; 56157c478bd9Sstevel@tonic-gate } 56167c478bd9Sstevel@tonic-gate } 56177c478bd9Sstevel@tonic-gate ire1 = *irep; 56187c478bd9Sstevel@tonic-gate if (ire1 != NULL) 56197c478bd9Sstevel@tonic-gate ire1->ire_ptpn = &ire->ire_next; 56207c478bd9Sstevel@tonic-gate ire->ire_next = ire1; 56217c478bd9Sstevel@tonic-gate /* Link the new one in. */ 56227c478bd9Sstevel@tonic-gate ire->ire_ptpn = irep; 56237c478bd9Sstevel@tonic-gate membar_producer(); 56247c478bd9Sstevel@tonic-gate *irep = ire; 56257c478bd9Sstevel@tonic-gate ire->ire_bucket = irb_ptr; 56267c478bd9Sstevel@tonic-gate IRE_REFHOLD_LOCKED(ire); 56277c478bd9Sstevel@tonic-gate 56287c478bd9Sstevel@tonic-gate /* 56297c478bd9Sstevel@tonic-gate * Protect ire_in_ill->ill_srcif_refcnt and table reference count. 56307c478bd9Sstevel@tonic-gate * Note, ire_atomic_start already grabs the ire_in_ill->ill_lock 56317c478bd9Sstevel@tonic-gate * so ill_srcif_refcnt is already protected. 56327c478bd9Sstevel@tonic-gate */ 56337c478bd9Sstevel@tonic-gate ire->ire_in_ill->ill_srcif_refcnt++; 5634*f4b3ec61Sdh155122 mutex_enter(&ipst->ips_ire_srcif_table_lock); 5635*f4b3ec61Sdh155122 ipst->ips_ire_srcif_table_count++; 5636*f4b3ec61Sdh155122 mutex_exit(&ipst->ips_ire_srcif_table_lock); 56377c478bd9Sstevel@tonic-gate irb_ptr->irb_ire_cnt++; 56387c478bd9Sstevel@tonic-gate if (ire->ire_ipif != NULL) { 56397c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_ire_cnt++; 56407c478bd9Sstevel@tonic-gate if (ire->ire_stq != NULL) { 56417c478bd9Sstevel@tonic-gate stq_ill = (ill_t *)ire->ire_stq->q_ptr; 56427c478bd9Sstevel@tonic-gate stq_ill->ill_ire_cnt++; 56437c478bd9Sstevel@tonic-gate } 56447c478bd9Sstevel@tonic-gate } else { 56457c478bd9Sstevel@tonic-gate ASSERT(ire->ire_stq == NULL); 56467c478bd9Sstevel@tonic-gate } 56477c478bd9Sstevel@tonic-gate 56487c478bd9Sstevel@tonic-gate ire_atomic_end(irb_ptr, ire); 56497c478bd9Sstevel@tonic-gate *ire_p = ire; 56507c478bd9Sstevel@tonic-gate return (0); 56517c478bd9Sstevel@tonic-gate } 56527c478bd9Sstevel@tonic-gate 56537c478bd9Sstevel@tonic-gate 56547c478bd9Sstevel@tonic-gate /* 56557c478bd9Sstevel@tonic-gate * This function is called by ire_add_then_send when ARP request comes 56567c478bd9Sstevel@tonic-gate * back to ip_wput->ire_add_then_send for resolved ire in the interface 56577c478bd9Sstevel@tonic-gate * based routing table. At this point, it only needs to update the resolver 56587c478bd9Sstevel@tonic-gate * information for the ire. The passed ire is returned to the caller as it 56597c478bd9Sstevel@tonic-gate * is the ire which is created as mblk. 56607c478bd9Sstevel@tonic-gate */ 56617c478bd9Sstevel@tonic-gate static ire_t * 56627c478bd9Sstevel@tonic-gate ire_update_srcif_v4(ire_t *ire) 56637c478bd9Sstevel@tonic-gate { 56647c478bd9Sstevel@tonic-gate ire_t *ire1; 56657c478bd9Sstevel@tonic-gate irb_t *irb; 56667c478bd9Sstevel@tonic-gate int error; 56677c478bd9Sstevel@tonic-gate 56687c478bd9Sstevel@tonic-gate ASSERT(ire->ire_type != IRE_MIPRTUN && 56697c478bd9Sstevel@tonic-gate ire->ire_ipif->ipif_net_type == IRE_IF_RESOLVER); 56707c478bd9Sstevel@tonic-gate ASSERT(ire->ire_ipversion == IPV4_VERSION); 56717c478bd9Sstevel@tonic-gate 56727c478bd9Sstevel@tonic-gate /* 56737c478bd9Sstevel@tonic-gate * This ire is from ARP. Update 5674c793af95Ssangeeta * ire_nce->nce_res_mp info 56757c478bd9Sstevel@tonic-gate */ 56767c478bd9Sstevel@tonic-gate ire1 = ire_srcif_table_lookup(ire->ire_addr, 56777c478bd9Sstevel@tonic-gate IRE_IF_RESOLVER, ire->ire_ipif, 56787c478bd9Sstevel@tonic-gate ire->ire_in_ill, 56797c478bd9Sstevel@tonic-gate MATCH_IRE_ILL | MATCH_IRE_TYPE); 56807c478bd9Sstevel@tonic-gate if (ire1 == NULL) { 56817c478bd9Sstevel@tonic-gate /* Mobile node registration expired ? */ 56827c478bd9Sstevel@tonic-gate ire_delete(ire); 56837c478bd9Sstevel@tonic-gate return (NULL); 56847c478bd9Sstevel@tonic-gate } 56857c478bd9Sstevel@tonic-gate irb = ire1->ire_bucket; 56867c478bd9Sstevel@tonic-gate ASSERT(irb != NULL); 56877c478bd9Sstevel@tonic-gate /* 56887c478bd9Sstevel@tonic-gate * Start the atomic add of the ire. Grab the ill locks, 56897c478bd9Sstevel@tonic-gate * ill_g_usesrc_lock and the bucket lock. 56907c478bd9Sstevel@tonic-gate */ 56917c478bd9Sstevel@tonic-gate error = ire_atomic_start(irb, ire1, NULL, NULL, NULL); 56927c478bd9Sstevel@tonic-gate if (error != 0) { 56937c478bd9Sstevel@tonic-gate /* 56947c478bd9Sstevel@tonic-gate * We don't know whether it is a valid ipif or not. 56957c478bd9Sstevel@tonic-gate * So, set it to NULL. This assumes that the ire has not added 56967c478bd9Sstevel@tonic-gate * a reference to the ipif. 56977c478bd9Sstevel@tonic-gate */ 56987c478bd9Sstevel@tonic-gate ire->ire_ipif = NULL; 56997c478bd9Sstevel@tonic-gate ire_delete(ire); 57007c478bd9Sstevel@tonic-gate ip1dbg(("ire_update_srcif_v4: ire_atomic_start failed\n")); 57017c478bd9Sstevel@tonic-gate return (NULL); 57027c478bd9Sstevel@tonic-gate } 57037c478bd9Sstevel@tonic-gate ASSERT(ire->ire_max_fragp == NULL); 57047c478bd9Sstevel@tonic-gate ire->ire_max_frag = ire1->ire_max_frag; 57057c478bd9Sstevel@tonic-gate /* 57067c478bd9Sstevel@tonic-gate * Update resolver information and 57077c478bd9Sstevel@tonic-gate * send-to queue. 57087c478bd9Sstevel@tonic-gate */ 5709c793af95Ssangeeta ASSERT(ire->ire_nce->nce_res_mp != NULL); 5710c793af95Ssangeeta ire1->ire_nce->nce_res_mp = copyb(ire->ire_nce->nce_res_mp); 5711c793af95Ssangeeta if (ire1->ire_nce->nce_res_mp == NULL) { 57127c478bd9Sstevel@tonic-gate ip0dbg(("ire_update_srcif: copyb failed\n")); 57137c478bd9Sstevel@tonic-gate ire_refrele(ire1); 57147c478bd9Sstevel@tonic-gate ire_refrele(ire); 57157c478bd9Sstevel@tonic-gate ire_atomic_end(irb, ire1); 57167c478bd9Sstevel@tonic-gate return (NULL); 57177c478bd9Sstevel@tonic-gate } 57187c478bd9Sstevel@tonic-gate ire1->ire_stq = ire->ire_stq; 57197c478bd9Sstevel@tonic-gate 5720c793af95Ssangeeta ASSERT(ire->ire_nce->nce_fp_mp == NULL); 57217c478bd9Sstevel@tonic-gate 57227c478bd9Sstevel@tonic-gate ire_atomic_end(irb, ire1); 57237c478bd9Sstevel@tonic-gate ire_refrele(ire1); 57247c478bd9Sstevel@tonic-gate /* Return the passed ire */ 57257c478bd9Sstevel@tonic-gate return (ire); /* Update done */ 57267c478bd9Sstevel@tonic-gate } 57277c478bd9Sstevel@tonic-gate 57287c478bd9Sstevel@tonic-gate 57297c478bd9Sstevel@tonic-gate /* 57307c478bd9Sstevel@tonic-gate * Check if another multirt route resolution is needed. 57317c478bd9Sstevel@tonic-gate * B_TRUE is returned is there remain a resolvable route, 57327c478bd9Sstevel@tonic-gate * or if no route for that dst is resolved yet. 57337c478bd9Sstevel@tonic-gate * B_FALSE is returned if all routes for that dst are resolved 57347c478bd9Sstevel@tonic-gate * or if the remaining unresolved routes are actually not 57357c478bd9Sstevel@tonic-gate * resolvable. 57367c478bd9Sstevel@tonic-gate * This only works in the global zone. 57377c478bd9Sstevel@tonic-gate */ 57387c478bd9Sstevel@tonic-gate boolean_t 5739*f4b3ec61Sdh155122 ire_multirt_need_resolve(ipaddr_t dst, const ts_label_t *tsl, ip_stack_t *ipst) 57407c478bd9Sstevel@tonic-gate { 57417c478bd9Sstevel@tonic-gate ire_t *first_fire; 57427c478bd9Sstevel@tonic-gate ire_t *first_cire; 57437c478bd9Sstevel@tonic-gate ire_t *fire; 57447c478bd9Sstevel@tonic-gate ire_t *cire; 57457c478bd9Sstevel@tonic-gate irb_t *firb; 57467c478bd9Sstevel@tonic-gate irb_t *cirb; 57477c478bd9Sstevel@tonic-gate int unres_cnt = 0; 57487c478bd9Sstevel@tonic-gate boolean_t resolvable = B_FALSE; 57497c478bd9Sstevel@tonic-gate 57507c478bd9Sstevel@tonic-gate /* Retrieve the first IRE_HOST that matches the destination */ 57517c478bd9Sstevel@tonic-gate first_fire = ire_ftable_lookup(dst, IP_HOST_MASK, 0, IRE_HOST, NULL, 575245916cd2Sjpk NULL, ALL_ZONES, 0, tsl, 5753*f4b3ec61Sdh155122 MATCH_IRE_MASK | MATCH_IRE_TYPE | MATCH_IRE_SECATTR, ipst); 57547c478bd9Sstevel@tonic-gate 57557c478bd9Sstevel@tonic-gate /* No route at all */ 57567c478bd9Sstevel@tonic-gate if (first_fire == NULL) { 57577c478bd9Sstevel@tonic-gate return (B_TRUE); 57587c478bd9Sstevel@tonic-gate } 57597c478bd9Sstevel@tonic-gate 57607c478bd9Sstevel@tonic-gate firb = first_fire->ire_bucket; 57617c478bd9Sstevel@tonic-gate ASSERT(firb != NULL); 57627c478bd9Sstevel@tonic-gate 57637c478bd9Sstevel@tonic-gate /* Retrieve the first IRE_CACHE ire for that destination. */ 5764*f4b3ec61Sdh155122 first_cire = ire_cache_lookup(dst, GLOBAL_ZONEID, tsl, ipst); 57657c478bd9Sstevel@tonic-gate 57667c478bd9Sstevel@tonic-gate /* No resolved route. */ 57677c478bd9Sstevel@tonic-gate if (first_cire == NULL) { 57687c478bd9Sstevel@tonic-gate ire_refrele(first_fire); 57697c478bd9Sstevel@tonic-gate return (B_TRUE); 57707c478bd9Sstevel@tonic-gate } 57717c478bd9Sstevel@tonic-gate 57727c478bd9Sstevel@tonic-gate /* 57737c478bd9Sstevel@tonic-gate * At least one route is resolved. Here we look through the forward 57747c478bd9Sstevel@tonic-gate * and cache tables, to compare the number of declared routes 57757c478bd9Sstevel@tonic-gate * with the number of resolved routes. The search for a resolvable 57767c478bd9Sstevel@tonic-gate * route is performed only if at least one route remains 57777c478bd9Sstevel@tonic-gate * unresolved. 57787c478bd9Sstevel@tonic-gate */ 57797c478bd9Sstevel@tonic-gate cirb = first_cire->ire_bucket; 57807c478bd9Sstevel@tonic-gate ASSERT(cirb != NULL); 57817c478bd9Sstevel@tonic-gate 57827c478bd9Sstevel@tonic-gate /* Count the number of routes to that dest that are declared. */ 57837c478bd9Sstevel@tonic-gate IRB_REFHOLD(firb); 57847c478bd9Sstevel@tonic-gate for (fire = first_fire; fire != NULL; fire = fire->ire_next) { 57857c478bd9Sstevel@tonic-gate if (!(fire->ire_flags & RTF_MULTIRT)) 57867c478bd9Sstevel@tonic-gate continue; 57877c478bd9Sstevel@tonic-gate if (fire->ire_addr != dst) 57887c478bd9Sstevel@tonic-gate continue; 57897c478bd9Sstevel@tonic-gate unres_cnt++; 57907c478bd9Sstevel@tonic-gate } 57917c478bd9Sstevel@tonic-gate IRB_REFRELE(firb); 57927c478bd9Sstevel@tonic-gate 57937c478bd9Sstevel@tonic-gate /* Then subtract the number of routes to that dst that are resolved */ 57947c478bd9Sstevel@tonic-gate IRB_REFHOLD(cirb); 57957c478bd9Sstevel@tonic-gate for (cire = first_cire; cire != NULL; cire = cire->ire_next) { 57967c478bd9Sstevel@tonic-gate if (!(cire->ire_flags & RTF_MULTIRT)) 57977c478bd9Sstevel@tonic-gate continue; 57987c478bd9Sstevel@tonic-gate if (cire->ire_addr != dst) 57997c478bd9Sstevel@tonic-gate continue; 58007c478bd9Sstevel@tonic-gate if (cire->ire_marks & (IRE_MARK_CONDEMNED | IRE_MARK_HIDDEN)) 58017c478bd9Sstevel@tonic-gate continue; 58027c478bd9Sstevel@tonic-gate unres_cnt--; 58037c478bd9Sstevel@tonic-gate } 58047c478bd9Sstevel@tonic-gate IRB_REFRELE(cirb); 58057c478bd9Sstevel@tonic-gate 58067c478bd9Sstevel@tonic-gate /* At least one route is unresolved; search for a resolvable route. */ 58077c478bd9Sstevel@tonic-gate if (unres_cnt > 0) 58087c478bd9Sstevel@tonic-gate resolvable = ire_multirt_lookup(&first_cire, &first_fire, 5809*f4b3ec61Sdh155122 MULTIRT_USESTAMP | MULTIRT_CACHEGW, tsl, ipst); 58107c478bd9Sstevel@tonic-gate 58117c478bd9Sstevel@tonic-gate if (first_fire != NULL) 58127c478bd9Sstevel@tonic-gate ire_refrele(first_fire); 58137c478bd9Sstevel@tonic-gate 58147c478bd9Sstevel@tonic-gate if (first_cire != NULL) 58157c478bd9Sstevel@tonic-gate ire_refrele(first_cire); 58167c478bd9Sstevel@tonic-gate 58177c478bd9Sstevel@tonic-gate return (resolvable); 58187c478bd9Sstevel@tonic-gate } 58197c478bd9Sstevel@tonic-gate 58207c478bd9Sstevel@tonic-gate 58217c478bd9Sstevel@tonic-gate /* 58227c478bd9Sstevel@tonic-gate * Explore a forward_table bucket, starting from fire_arg. 58237c478bd9Sstevel@tonic-gate * fire_arg MUST be an IRE_HOST entry. 58247c478bd9Sstevel@tonic-gate * 58257c478bd9Sstevel@tonic-gate * Return B_TRUE and update *ire_arg and *fire_arg 58267c478bd9Sstevel@tonic-gate * if at least one resolvable route is found. *ire_arg 58277c478bd9Sstevel@tonic-gate * is the IRE entry for *fire_arg's gateway. 58287c478bd9Sstevel@tonic-gate * 58297c478bd9Sstevel@tonic-gate * Return B_FALSE otherwise (all routes are resolved or 58307c478bd9Sstevel@tonic-gate * the remaining unresolved routes are all unresolvable). 58317c478bd9Sstevel@tonic-gate * 58327c478bd9Sstevel@tonic-gate * The IRE selection relies on a priority mechanism 58337c478bd9Sstevel@tonic-gate * driven by the flags passed in by the caller. 58347c478bd9Sstevel@tonic-gate * The caller, such as ip_newroute_ipif(), can get the most 58357c478bd9Sstevel@tonic-gate * relevant ire at each stage of a multiple route resolution. 58367c478bd9Sstevel@tonic-gate * 58377c478bd9Sstevel@tonic-gate * The rules are: 58387c478bd9Sstevel@tonic-gate * 58397c478bd9Sstevel@tonic-gate * - if MULTIRT_CACHEGW is specified in flags, IRE_CACHETABLE 58407c478bd9Sstevel@tonic-gate * ires are preferred for the gateway. This gives the highest 58417c478bd9Sstevel@tonic-gate * priority to routes that can be resolved without using 58427c478bd9Sstevel@tonic-gate * a resolver. 58437c478bd9Sstevel@tonic-gate * 58447c478bd9Sstevel@tonic-gate * - if MULTIRT_CACHEGW is not specified, or if MULTIRT_CACHEGW 58457c478bd9Sstevel@tonic-gate * is specified but no IRE_CACHETABLE ire entry for the gateway 58467c478bd9Sstevel@tonic-gate * is found, the following rules apply. 58477c478bd9Sstevel@tonic-gate * 58487c478bd9Sstevel@tonic-gate * - if MULTIRT_USESTAMP is specified in flags, IRE_INTERFACE 58497c478bd9Sstevel@tonic-gate * ires for the gateway, that have not been tried since 58507c478bd9Sstevel@tonic-gate * a configurable amount of time, are preferred. 58517c478bd9Sstevel@tonic-gate * This applies when a resolver must be invoked for 58527c478bd9Sstevel@tonic-gate * a missing route, but we don't want to use the resolver 58537c478bd9Sstevel@tonic-gate * upon each packet emission. If no such resolver is found, 58547c478bd9Sstevel@tonic-gate * B_FALSE is returned. 58557c478bd9Sstevel@tonic-gate * The MULTIRT_USESTAMP flag can be combined with 58567c478bd9Sstevel@tonic-gate * MULTIRT_CACHEGW. 58577c478bd9Sstevel@tonic-gate * 58587c478bd9Sstevel@tonic-gate * - if MULTIRT_USESTAMP is not specified in flags, the first 58597c478bd9Sstevel@tonic-gate * unresolved but resolvable route is selected. 58607c478bd9Sstevel@tonic-gate * 58617c478bd9Sstevel@tonic-gate * - Otherwise, there is no resolvalble route, and 58627c478bd9Sstevel@tonic-gate * B_FALSE is returned. 58637c478bd9Sstevel@tonic-gate * 58647c478bd9Sstevel@tonic-gate * At last, MULTIRT_SETSTAMP can be specified in flags to 58657c478bd9Sstevel@tonic-gate * request the timestamp of unresolvable routes to 58667c478bd9Sstevel@tonic-gate * be refreshed. This prevents the useless exploration 58677c478bd9Sstevel@tonic-gate * of those routes for a while, when MULTIRT_USESTAMP is used. 58687c478bd9Sstevel@tonic-gate * 58697c478bd9Sstevel@tonic-gate * This only works in the global zone. 58707c478bd9Sstevel@tonic-gate */ 58717c478bd9Sstevel@tonic-gate boolean_t 587245916cd2Sjpk ire_multirt_lookup(ire_t **ire_arg, ire_t **fire_arg, uint32_t flags, 5873*f4b3ec61Sdh155122 const ts_label_t *tsl, ip_stack_t *ipst) 58747c478bd9Sstevel@tonic-gate { 58757c478bd9Sstevel@tonic-gate clock_t delta; 58767c478bd9Sstevel@tonic-gate ire_t *best_fire = NULL; 58777c478bd9Sstevel@tonic-gate ire_t *best_cire = NULL; 58787c478bd9Sstevel@tonic-gate ire_t *first_fire; 58797c478bd9Sstevel@tonic-gate ire_t *first_cire; 58807c478bd9Sstevel@tonic-gate ire_t *fire; 58817c478bd9Sstevel@tonic-gate ire_t *cire; 58827c478bd9Sstevel@tonic-gate irb_t *firb = NULL; 58837c478bd9Sstevel@tonic-gate irb_t *cirb = NULL; 58847c478bd9Sstevel@tonic-gate ire_t *gw_ire; 58857c478bd9Sstevel@tonic-gate boolean_t already_resolved; 58867c478bd9Sstevel@tonic-gate boolean_t res; 58877c478bd9Sstevel@tonic-gate ipaddr_t dst; 58887c478bd9Sstevel@tonic-gate ipaddr_t gw; 58897c478bd9Sstevel@tonic-gate 58907c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: *ire_arg %p, *fire_arg %p, flags %04x\n", 58917c478bd9Sstevel@tonic-gate (void *)*ire_arg, (void *)*fire_arg, flags)); 58927c478bd9Sstevel@tonic-gate 58937c478bd9Sstevel@tonic-gate ASSERT(ire_arg != NULL); 58947c478bd9Sstevel@tonic-gate ASSERT(fire_arg != NULL); 58957c478bd9Sstevel@tonic-gate 58967c478bd9Sstevel@tonic-gate /* Not an IRE_HOST ire; give up. */ 58977c478bd9Sstevel@tonic-gate if ((*fire_arg == NULL) || ((*fire_arg)->ire_type != IRE_HOST)) { 58987c478bd9Sstevel@tonic-gate return (B_FALSE); 58997c478bd9Sstevel@tonic-gate } 59007c478bd9Sstevel@tonic-gate 59017c478bd9Sstevel@tonic-gate /* This is the first IRE_HOST ire for that destination. */ 59027c478bd9Sstevel@tonic-gate first_fire = *fire_arg; 59037c478bd9Sstevel@tonic-gate firb = first_fire->ire_bucket; 59047c478bd9Sstevel@tonic-gate ASSERT(firb != NULL); 59057c478bd9Sstevel@tonic-gate 59067c478bd9Sstevel@tonic-gate dst = first_fire->ire_addr; 59077c478bd9Sstevel@tonic-gate 59087c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: dst %08x\n", ntohl(dst))); 59097c478bd9Sstevel@tonic-gate 59107c478bd9Sstevel@tonic-gate /* 59117c478bd9Sstevel@tonic-gate * Retrieve the first IRE_CACHE ire for that destination; 59127c478bd9Sstevel@tonic-gate * if we don't find one, no route for that dest is 59137c478bd9Sstevel@tonic-gate * resolved yet. 59147c478bd9Sstevel@tonic-gate */ 5915*f4b3ec61Sdh155122 first_cire = ire_cache_lookup(dst, GLOBAL_ZONEID, tsl, ipst); 59167c478bd9Sstevel@tonic-gate if (first_cire != NULL) { 59177c478bd9Sstevel@tonic-gate cirb = first_cire->ire_bucket; 59187c478bd9Sstevel@tonic-gate } 59197c478bd9Sstevel@tonic-gate 59207c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: first_cire %p\n", (void *)first_cire)); 59217c478bd9Sstevel@tonic-gate 59227c478bd9Sstevel@tonic-gate /* 59237c478bd9Sstevel@tonic-gate * Search for a resolvable route, giving the top priority 59247c478bd9Sstevel@tonic-gate * to routes that can be resolved without any call to the resolver. 59257c478bd9Sstevel@tonic-gate */ 59267c478bd9Sstevel@tonic-gate IRB_REFHOLD(firb); 59277c478bd9Sstevel@tonic-gate 59287c478bd9Sstevel@tonic-gate if (!CLASSD(dst)) { 59297c478bd9Sstevel@tonic-gate /* 59307c478bd9Sstevel@tonic-gate * For all multiroute IRE_HOST ires for that destination, 59317c478bd9Sstevel@tonic-gate * check if the route via the IRE_HOST's gateway is 59327c478bd9Sstevel@tonic-gate * resolved yet. 59337c478bd9Sstevel@tonic-gate */ 59347c478bd9Sstevel@tonic-gate for (fire = first_fire; fire != NULL; fire = fire->ire_next) { 59357c478bd9Sstevel@tonic-gate 59367c478bd9Sstevel@tonic-gate if (!(fire->ire_flags & RTF_MULTIRT)) 59377c478bd9Sstevel@tonic-gate continue; 59387c478bd9Sstevel@tonic-gate if (fire->ire_addr != dst) 59397c478bd9Sstevel@tonic-gate continue; 59407c478bd9Sstevel@tonic-gate 594145916cd2Sjpk if (fire->ire_gw_secattr != NULL && 594245916cd2Sjpk tsol_ire_match_gwattr(fire, tsl) != 0) { 594345916cd2Sjpk continue; 594445916cd2Sjpk } 594545916cd2Sjpk 59467c478bd9Sstevel@tonic-gate gw = fire->ire_gateway_addr; 59477c478bd9Sstevel@tonic-gate 59487c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: fire %p, " 59497c478bd9Sstevel@tonic-gate "ire_addr %08x, ire_gateway_addr %08x\n", 59507c478bd9Sstevel@tonic-gate (void *)fire, ntohl(fire->ire_addr), ntohl(gw))); 59517c478bd9Sstevel@tonic-gate 59527c478bd9Sstevel@tonic-gate already_resolved = B_FALSE; 59537c478bd9Sstevel@tonic-gate 59547c478bd9Sstevel@tonic-gate if (first_cire != NULL) { 59557c478bd9Sstevel@tonic-gate ASSERT(cirb != NULL); 59567c478bd9Sstevel@tonic-gate 59577c478bd9Sstevel@tonic-gate IRB_REFHOLD(cirb); 59587c478bd9Sstevel@tonic-gate /* 59597c478bd9Sstevel@tonic-gate * For all IRE_CACHE ires for that 59607c478bd9Sstevel@tonic-gate * destination. 59617c478bd9Sstevel@tonic-gate */ 59627c478bd9Sstevel@tonic-gate for (cire = first_cire; 59637c478bd9Sstevel@tonic-gate cire != NULL; 59647c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 59657c478bd9Sstevel@tonic-gate 59667c478bd9Sstevel@tonic-gate if (!(cire->ire_flags & RTF_MULTIRT)) 59677c478bd9Sstevel@tonic-gate continue; 59687c478bd9Sstevel@tonic-gate if (cire->ire_addr != dst) 59697c478bd9Sstevel@tonic-gate continue; 59707c478bd9Sstevel@tonic-gate if (cire->ire_marks & 59717c478bd9Sstevel@tonic-gate (IRE_MARK_CONDEMNED | 59727c478bd9Sstevel@tonic-gate IRE_MARK_HIDDEN)) 59737c478bd9Sstevel@tonic-gate continue; 597445916cd2Sjpk 597545916cd2Sjpk if (cire->ire_gw_secattr != NULL && 597645916cd2Sjpk tsol_ire_match_gwattr(cire, 597745916cd2Sjpk tsl) != 0) { 597845916cd2Sjpk continue; 597945916cd2Sjpk } 598045916cd2Sjpk 59817c478bd9Sstevel@tonic-gate /* 59827c478bd9Sstevel@tonic-gate * Check if the IRE_CACHE's gateway 59837c478bd9Sstevel@tonic-gate * matches the IRE_HOST's gateway. 59847c478bd9Sstevel@tonic-gate */ 59857c478bd9Sstevel@tonic-gate if (cire->ire_gateway_addr == gw) { 59867c478bd9Sstevel@tonic-gate already_resolved = B_TRUE; 59877c478bd9Sstevel@tonic-gate break; 59887c478bd9Sstevel@tonic-gate } 59897c478bd9Sstevel@tonic-gate } 59907c478bd9Sstevel@tonic-gate IRB_REFRELE(cirb); 59917c478bd9Sstevel@tonic-gate } 59927c478bd9Sstevel@tonic-gate 59937c478bd9Sstevel@tonic-gate /* 59947c478bd9Sstevel@tonic-gate * This route is already resolved; 59957c478bd9Sstevel@tonic-gate * proceed with next one. 59967c478bd9Sstevel@tonic-gate */ 59977c478bd9Sstevel@tonic-gate if (already_resolved) { 59987c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: found cire %p, " 59997c478bd9Sstevel@tonic-gate "already resolved\n", (void *)cire)); 60007c478bd9Sstevel@tonic-gate continue; 60017c478bd9Sstevel@tonic-gate } 60027c478bd9Sstevel@tonic-gate 60037c478bd9Sstevel@tonic-gate /* 60047c478bd9Sstevel@tonic-gate * The route is unresolved; is it actually 60057c478bd9Sstevel@tonic-gate * resolvable, i.e. is there a cache or a resolver 60067c478bd9Sstevel@tonic-gate * for the gateway? 60077c478bd9Sstevel@tonic-gate */ 60087c478bd9Sstevel@tonic-gate gw_ire = ire_route_lookup(gw, 0, 0, 0, NULL, NULL, 600945916cd2Sjpk ALL_ZONES, tsl, 6010*f4b3ec61Sdh155122 MATCH_IRE_RECURSIVE | MATCH_IRE_SECATTR, ipst); 60117c478bd9Sstevel@tonic-gate 60127c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: looked up gw_ire %p\n", 60137c478bd9Sstevel@tonic-gate (void *)gw_ire)); 60147c478bd9Sstevel@tonic-gate 60157c478bd9Sstevel@tonic-gate /* 60167c478bd9Sstevel@tonic-gate * If gw_ire is typed IRE_CACHETABLE, 60177c478bd9Sstevel@tonic-gate * this route can be resolved without any call to the 60187c478bd9Sstevel@tonic-gate * resolver. If the MULTIRT_CACHEGW flag is set, 60197c478bd9Sstevel@tonic-gate * give the top priority to this ire and exit the 60207c478bd9Sstevel@tonic-gate * loop. 60217c478bd9Sstevel@tonic-gate * This is typically the case when an ARP reply 60227c478bd9Sstevel@tonic-gate * is processed through ip_wput_nondata(). 60237c478bd9Sstevel@tonic-gate */ 60247c478bd9Sstevel@tonic-gate if ((flags & MULTIRT_CACHEGW) && 60257c478bd9Sstevel@tonic-gate (gw_ire != NULL) && 60267c478bd9Sstevel@tonic-gate (gw_ire->ire_type & IRE_CACHETABLE)) { 6027c793af95Ssangeeta ASSERT(gw_ire->ire_nce == NULL || 6028c793af95Ssangeeta gw_ire->ire_nce->nce_state == ND_REACHABLE); 60297c478bd9Sstevel@tonic-gate /* 60307c478bd9Sstevel@tonic-gate * Release the resolver associated to the 60317c478bd9Sstevel@tonic-gate * previous candidate best ire, if any. 60327c478bd9Sstevel@tonic-gate */ 60337c478bd9Sstevel@tonic-gate if (best_cire != NULL) { 60347c478bd9Sstevel@tonic-gate ire_refrele(best_cire); 60357c478bd9Sstevel@tonic-gate ASSERT(best_fire != NULL); 60367c478bd9Sstevel@tonic-gate } 60377c478bd9Sstevel@tonic-gate 60387c478bd9Sstevel@tonic-gate best_fire = fire; 60397c478bd9Sstevel@tonic-gate best_cire = gw_ire; 60407c478bd9Sstevel@tonic-gate 60417c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: found top prio " 60427c478bd9Sstevel@tonic-gate "best_fire %p, best_cire %p\n", 60437c478bd9Sstevel@tonic-gate (void *)best_fire, (void *)best_cire)); 60447c478bd9Sstevel@tonic-gate break; 60457c478bd9Sstevel@tonic-gate } 60467c478bd9Sstevel@tonic-gate 60477c478bd9Sstevel@tonic-gate /* 60487c478bd9Sstevel@tonic-gate * Compute the time elapsed since our preceding 60497c478bd9Sstevel@tonic-gate * attempt to resolve that route. 60507c478bd9Sstevel@tonic-gate * If the MULTIRT_USESTAMP flag is set, we take that 60517c478bd9Sstevel@tonic-gate * route into account only if this time interval 60527c478bd9Sstevel@tonic-gate * exceeds ip_multirt_resolution_interval; 60537c478bd9Sstevel@tonic-gate * this prevents us from attempting to resolve a 60547c478bd9Sstevel@tonic-gate * broken route upon each sending of a packet. 60557c478bd9Sstevel@tonic-gate */ 60567c478bd9Sstevel@tonic-gate delta = lbolt - fire->ire_last_used_time; 60577c478bd9Sstevel@tonic-gate delta = TICK_TO_MSEC(delta); 60587c478bd9Sstevel@tonic-gate 6059*f4b3ec61Sdh155122 res = (boolean_t)((delta > 6060*f4b3ec61Sdh155122 ipst->ips_ip_multirt_resolution_interval) || 60617c478bd9Sstevel@tonic-gate (!(flags & MULTIRT_USESTAMP))); 60627c478bd9Sstevel@tonic-gate 60637c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: fire %p, delta %lu, " 60647c478bd9Sstevel@tonic-gate "res %d\n", 60657c478bd9Sstevel@tonic-gate (void *)fire, delta, res)); 60667c478bd9Sstevel@tonic-gate 60677c478bd9Sstevel@tonic-gate if (res) { 60687c478bd9Sstevel@tonic-gate /* 60697c478bd9Sstevel@tonic-gate * We are here if MULTIRT_USESTAMP flag is set 60707c478bd9Sstevel@tonic-gate * and the resolver for fire's gateway 60717c478bd9Sstevel@tonic-gate * has not been tried since 60727c478bd9Sstevel@tonic-gate * ip_multirt_resolution_interval, or if 60737c478bd9Sstevel@tonic-gate * MULTIRT_USESTAMP is not set but gw_ire did 60747c478bd9Sstevel@tonic-gate * not fill the conditions for MULTIRT_CACHEGW, 60757c478bd9Sstevel@tonic-gate * or if neither MULTIRT_USESTAMP nor 60767c478bd9Sstevel@tonic-gate * MULTIRT_CACHEGW are set. 60777c478bd9Sstevel@tonic-gate */ 60787c478bd9Sstevel@tonic-gate if (gw_ire != NULL) { 60797c478bd9Sstevel@tonic-gate if (best_fire == NULL) { 60807c478bd9Sstevel@tonic-gate ASSERT(best_cire == NULL); 60817c478bd9Sstevel@tonic-gate 60827c478bd9Sstevel@tonic-gate best_fire = fire; 60837c478bd9Sstevel@tonic-gate best_cire = gw_ire; 60847c478bd9Sstevel@tonic-gate 60857c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup:" 60867c478bd9Sstevel@tonic-gate "found candidate " 60877c478bd9Sstevel@tonic-gate "best_fire %p, " 60887c478bd9Sstevel@tonic-gate "best_cire %p\n", 60897c478bd9Sstevel@tonic-gate (void *)best_fire, 60907c478bd9Sstevel@tonic-gate (void *)best_cire)); 60917c478bd9Sstevel@tonic-gate 60927c478bd9Sstevel@tonic-gate /* 60937c478bd9Sstevel@tonic-gate * If MULTIRT_CACHEGW is not 60947c478bd9Sstevel@tonic-gate * set, we ignore the top 60957c478bd9Sstevel@tonic-gate * priority ires that can 60967c478bd9Sstevel@tonic-gate * be resolved without any 60977c478bd9Sstevel@tonic-gate * call to the resolver; 60987c478bd9Sstevel@tonic-gate * In that case, there is 60997c478bd9Sstevel@tonic-gate * actually no need 61007c478bd9Sstevel@tonic-gate * to continue the loop. 61017c478bd9Sstevel@tonic-gate */ 61027c478bd9Sstevel@tonic-gate if (!(flags & 61037c478bd9Sstevel@tonic-gate MULTIRT_CACHEGW)) { 61047c478bd9Sstevel@tonic-gate break; 61057c478bd9Sstevel@tonic-gate } 61067c478bd9Sstevel@tonic-gate continue; 61077c478bd9Sstevel@tonic-gate } 61087c478bd9Sstevel@tonic-gate } else { 61097c478bd9Sstevel@tonic-gate /* 61107c478bd9Sstevel@tonic-gate * No resolver for the gateway: the 61117c478bd9Sstevel@tonic-gate * route is not resolvable. 61127c478bd9Sstevel@tonic-gate * If the MULTIRT_SETSTAMP flag is 61137c478bd9Sstevel@tonic-gate * set, we stamp the IRE_HOST ire, 61147c478bd9Sstevel@tonic-gate * so we will not select it again 61157c478bd9Sstevel@tonic-gate * during this resolution interval. 61167c478bd9Sstevel@tonic-gate */ 61177c478bd9Sstevel@tonic-gate if (flags & MULTIRT_SETSTAMP) 61187c478bd9Sstevel@tonic-gate fire->ire_last_used_time = 61197c478bd9Sstevel@tonic-gate lbolt; 61207c478bd9Sstevel@tonic-gate } 61217c478bd9Sstevel@tonic-gate } 61227c478bd9Sstevel@tonic-gate 61237c478bd9Sstevel@tonic-gate if (gw_ire != NULL) 61247c478bd9Sstevel@tonic-gate ire_refrele(gw_ire); 61257c478bd9Sstevel@tonic-gate } 61267c478bd9Sstevel@tonic-gate } else { /* CLASSD(dst) */ 61277c478bd9Sstevel@tonic-gate 61287c478bd9Sstevel@tonic-gate for (fire = first_fire; 61297c478bd9Sstevel@tonic-gate fire != NULL; 61307c478bd9Sstevel@tonic-gate fire = fire->ire_next) { 61317c478bd9Sstevel@tonic-gate 61327c478bd9Sstevel@tonic-gate if (!(fire->ire_flags & RTF_MULTIRT)) 61337c478bd9Sstevel@tonic-gate continue; 61347c478bd9Sstevel@tonic-gate if (fire->ire_addr != dst) 61357c478bd9Sstevel@tonic-gate continue; 61367c478bd9Sstevel@tonic-gate 613745916cd2Sjpk if (fire->ire_gw_secattr != NULL && 613845916cd2Sjpk tsol_ire_match_gwattr(fire, tsl) != 0) { 613945916cd2Sjpk continue; 614045916cd2Sjpk } 614145916cd2Sjpk 61427c478bd9Sstevel@tonic-gate already_resolved = B_FALSE; 61437c478bd9Sstevel@tonic-gate 61447c478bd9Sstevel@tonic-gate gw = fire->ire_gateway_addr; 61457c478bd9Sstevel@tonic-gate 61467c478bd9Sstevel@tonic-gate gw_ire = ire_ftable_lookup(gw, 0, 0, IRE_INTERFACE, 614745916cd2Sjpk NULL, NULL, ALL_ZONES, 0, tsl, 614845916cd2Sjpk MATCH_IRE_RECURSIVE | MATCH_IRE_TYPE | 6149*f4b3ec61Sdh155122 MATCH_IRE_SECATTR, ipst); 61507c478bd9Sstevel@tonic-gate 61517c478bd9Sstevel@tonic-gate /* No resolver for the gateway; we skip this ire. */ 61527c478bd9Sstevel@tonic-gate if (gw_ire == NULL) { 61537c478bd9Sstevel@tonic-gate continue; 61547c478bd9Sstevel@tonic-gate } 6155c793af95Ssangeeta ASSERT(gw_ire->ire_nce == NULL || 6156c793af95Ssangeeta gw_ire->ire_nce->nce_state == ND_REACHABLE); 61577c478bd9Sstevel@tonic-gate 61587c478bd9Sstevel@tonic-gate if (first_cire != NULL) { 61597c478bd9Sstevel@tonic-gate 61607c478bd9Sstevel@tonic-gate IRB_REFHOLD(cirb); 61617c478bd9Sstevel@tonic-gate /* 61627c478bd9Sstevel@tonic-gate * For all IRE_CACHE ires for that 61637c478bd9Sstevel@tonic-gate * destination. 61647c478bd9Sstevel@tonic-gate */ 61657c478bd9Sstevel@tonic-gate for (cire = first_cire; 61667c478bd9Sstevel@tonic-gate cire != NULL; 61677c478bd9Sstevel@tonic-gate cire = cire->ire_next) { 61687c478bd9Sstevel@tonic-gate 61697c478bd9Sstevel@tonic-gate if (!(cire->ire_flags & RTF_MULTIRT)) 61707c478bd9Sstevel@tonic-gate continue; 61717c478bd9Sstevel@tonic-gate if (cire->ire_addr != dst) 61727c478bd9Sstevel@tonic-gate continue; 61737c478bd9Sstevel@tonic-gate if (cire->ire_marks & 61747c478bd9Sstevel@tonic-gate (IRE_MARK_CONDEMNED | 61757c478bd9Sstevel@tonic-gate IRE_MARK_HIDDEN)) 61767c478bd9Sstevel@tonic-gate continue; 61777c478bd9Sstevel@tonic-gate 617845916cd2Sjpk if (cire->ire_gw_secattr != NULL && 617945916cd2Sjpk tsol_ire_match_gwattr(cire, 618045916cd2Sjpk tsl) != 0) { 618145916cd2Sjpk continue; 618245916cd2Sjpk } 618345916cd2Sjpk 61847c478bd9Sstevel@tonic-gate /* 61857c478bd9Sstevel@tonic-gate * Cache entries are linked to the 61867c478bd9Sstevel@tonic-gate * parent routes using the parent handle 61877c478bd9Sstevel@tonic-gate * (ire_phandle). If no cache entry has 61887c478bd9Sstevel@tonic-gate * the same handle as fire, fire is 61897c478bd9Sstevel@tonic-gate * still unresolved. 61907c478bd9Sstevel@tonic-gate */ 61917c478bd9Sstevel@tonic-gate ASSERT(cire->ire_phandle != 0); 61927c478bd9Sstevel@tonic-gate if (cire->ire_phandle == 61937c478bd9Sstevel@tonic-gate fire->ire_phandle) { 61947c478bd9Sstevel@tonic-gate already_resolved = B_TRUE; 61957c478bd9Sstevel@tonic-gate break; 61967c478bd9Sstevel@tonic-gate } 61977c478bd9Sstevel@tonic-gate } 61987c478bd9Sstevel@tonic-gate IRB_REFRELE(cirb); 61997c478bd9Sstevel@tonic-gate } 62007c478bd9Sstevel@tonic-gate 62017c478bd9Sstevel@tonic-gate /* 62027c478bd9Sstevel@tonic-gate * This route is already resolved; proceed with 62037c478bd9Sstevel@tonic-gate * next one. 62047c478bd9Sstevel@tonic-gate */ 62057c478bd9Sstevel@tonic-gate if (already_resolved) { 62067c478bd9Sstevel@tonic-gate ire_refrele(gw_ire); 62077c478bd9Sstevel@tonic-gate continue; 62087c478bd9Sstevel@tonic-gate } 62097c478bd9Sstevel@tonic-gate 62107c478bd9Sstevel@tonic-gate /* 62117c478bd9Sstevel@tonic-gate * Compute the time elapsed since our preceding 62127c478bd9Sstevel@tonic-gate * attempt to resolve that route. 62137c478bd9Sstevel@tonic-gate * If the MULTIRT_USESTAMP flag is set, we take 62147c478bd9Sstevel@tonic-gate * that route into account only if this time 62157c478bd9Sstevel@tonic-gate * interval exceeds ip_multirt_resolution_interval; 62167c478bd9Sstevel@tonic-gate * this prevents us from attempting to resolve a 62177c478bd9Sstevel@tonic-gate * broken route upon each sending of a packet. 62187c478bd9Sstevel@tonic-gate */ 62197c478bd9Sstevel@tonic-gate delta = lbolt - fire->ire_last_used_time; 62207c478bd9Sstevel@tonic-gate delta = TICK_TO_MSEC(delta); 62217c478bd9Sstevel@tonic-gate 6222*f4b3ec61Sdh155122 res = (boolean_t)((delta > 6223*f4b3ec61Sdh155122 ipst->ips_ip_multirt_resolution_interval) || 62247c478bd9Sstevel@tonic-gate (!(flags & MULTIRT_USESTAMP))); 62257c478bd9Sstevel@tonic-gate 62267c478bd9Sstevel@tonic-gate ip3dbg(("ire_multirt_lookup: fire %p, delta %lx, " 62277c478bd9Sstevel@tonic-gate "flags %04x, res %d\n", 62287c478bd9Sstevel@tonic-gate (void *)fire, delta, flags, res)); 62297c478bd9Sstevel@tonic-gate 62307c478bd9Sstevel@tonic-gate if (res) { 62317c478bd9Sstevel@tonic-gate if (best_cire != NULL) { 62327c478bd9Sstevel@tonic-gate /* 62337c478bd9Sstevel@tonic-gate * Release the resolver associated 62347c478bd9Sstevel@tonic-gate * to the preceding candidate best 62357c478bd9Sstevel@tonic-gate * ire, if any. 62367c478bd9Sstevel@tonic-gate */ 62377c478bd9Sstevel@tonic-gate ire_refrele(best_cire); 62387c478bd9Sstevel@tonic-gate ASSERT(best_fire != NULL); 62397c478bd9Sstevel@tonic-gate } 62407c478bd9Sstevel@tonic-gate best_fire = fire; 62417c478bd9Sstevel@tonic-gate best_cire = gw_ire; 62427c478bd9Sstevel@tonic-gate continue; 62437c478bd9Sstevel@tonic-gate } 62447c478bd9Sstevel@tonic-gate 62457c478bd9Sstevel@tonic-gate ire_refrele(gw_ire); 62467c478bd9Sstevel@tonic-gate } 62477c478bd9Sstevel@tonic-gate } 62487c478bd9Sstevel@tonic-gate 62497c478bd9Sstevel@tonic-gate if (best_fire != NULL) { 62507c478bd9Sstevel@tonic-gate IRE_REFHOLD(best_fire); 62517c478bd9Sstevel@tonic-gate } 62527c478bd9Sstevel@tonic-gate IRB_REFRELE(firb); 62537c478bd9Sstevel@tonic-gate 62547c478bd9Sstevel@tonic-gate /* Release the first IRE_CACHE we initially looked up, if any. */ 62557c478bd9Sstevel@tonic-gate if (first_cire != NULL) 62567c478bd9Sstevel@tonic-gate ire_refrele(first_cire); 62577c478bd9Sstevel@tonic-gate 62587c478bd9Sstevel@tonic-gate /* Found a resolvable route. */ 62597c478bd9Sstevel@tonic-gate if (best_fire != NULL) { 62607c478bd9Sstevel@tonic-gate ASSERT(best_cire != NULL); 62617c478bd9Sstevel@tonic-gate 62627c478bd9Sstevel@tonic-gate if (*fire_arg != NULL) 62637c478bd9Sstevel@tonic-gate ire_refrele(*fire_arg); 62647c478bd9Sstevel@tonic-gate if (*ire_arg != NULL) 62657c478bd9Sstevel@tonic-gate ire_refrele(*ire_arg); 62667c478bd9Sstevel@tonic-gate 62677c478bd9Sstevel@tonic-gate /* 62687c478bd9Sstevel@tonic-gate * Update the passed-in arguments with the 62697c478bd9Sstevel@tonic-gate * resolvable multirt route we found. 62707c478bd9Sstevel@tonic-gate */ 62717c478bd9Sstevel@tonic-gate *fire_arg = best_fire; 62727c478bd9Sstevel@tonic-gate *ire_arg = best_cire; 62737c478bd9Sstevel@tonic-gate 62747c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: returning B_TRUE, " 62757c478bd9Sstevel@tonic-gate "*fire_arg %p, *ire_arg %p\n", 62767c478bd9Sstevel@tonic-gate (void *)best_fire, (void *)best_cire)); 62777c478bd9Sstevel@tonic-gate 62787c478bd9Sstevel@tonic-gate return (B_TRUE); 62797c478bd9Sstevel@tonic-gate } 62807c478bd9Sstevel@tonic-gate 62817c478bd9Sstevel@tonic-gate ASSERT(best_cire == NULL); 62827c478bd9Sstevel@tonic-gate 62837c478bd9Sstevel@tonic-gate ip2dbg(("ire_multirt_lookup: returning B_FALSE, *fire_arg %p, " 62847c478bd9Sstevel@tonic-gate "*ire_arg %p\n", 62857c478bd9Sstevel@tonic-gate (void *)*fire_arg, (void *)*ire_arg)); 62867c478bd9Sstevel@tonic-gate 62877c478bd9Sstevel@tonic-gate /* No resolvable route. */ 62887c478bd9Sstevel@tonic-gate return (B_FALSE); 62897c478bd9Sstevel@tonic-gate } 62907c478bd9Sstevel@tonic-gate 62917c478bd9Sstevel@tonic-gate /* 62927c478bd9Sstevel@tonic-gate * IRE iterator for inbound and loopback broadcast processing. 62937c478bd9Sstevel@tonic-gate * Given an IRE_BROADCAST ire, walk the ires with the same destination 62947c478bd9Sstevel@tonic-gate * address, but skip over the passed-in ire. Returns the next ire without 62957c478bd9Sstevel@tonic-gate * a hold - assumes that the caller holds a reference on the IRE bucket. 62967c478bd9Sstevel@tonic-gate */ 62977c478bd9Sstevel@tonic-gate ire_t * 62987c478bd9Sstevel@tonic-gate ire_get_next_bcast_ire(ire_t *curr, ire_t *ire) 62997c478bd9Sstevel@tonic-gate { 63007c478bd9Sstevel@tonic-gate ill_t *ill; 63017c478bd9Sstevel@tonic-gate 63027c478bd9Sstevel@tonic-gate if (curr == NULL) { 63037c478bd9Sstevel@tonic-gate for (curr = ire->ire_bucket->irb_ire; curr != NULL; 63047c478bd9Sstevel@tonic-gate curr = curr->ire_next) { 63057c478bd9Sstevel@tonic-gate if (curr->ire_addr == ire->ire_addr) 63067c478bd9Sstevel@tonic-gate break; 63077c478bd9Sstevel@tonic-gate } 63087c478bd9Sstevel@tonic-gate } else { 63097c478bd9Sstevel@tonic-gate curr = curr->ire_next; 63107c478bd9Sstevel@tonic-gate } 63117c478bd9Sstevel@tonic-gate ill = ire_to_ill(ire); 63127c478bd9Sstevel@tonic-gate for (; curr != NULL; curr = curr->ire_next) { 63137c478bd9Sstevel@tonic-gate if (curr->ire_addr != ire->ire_addr) { 63147c478bd9Sstevel@tonic-gate /* 63157c478bd9Sstevel@tonic-gate * All the IREs to a given destination are contiguous; 63167c478bd9Sstevel@tonic-gate * break out once the address doesn't match. 63177c478bd9Sstevel@tonic-gate */ 63187c478bd9Sstevel@tonic-gate break; 63197c478bd9Sstevel@tonic-gate } 63207c478bd9Sstevel@tonic-gate if (curr == ire) { 63217c478bd9Sstevel@tonic-gate /* skip over the passed-in ire */ 63227c478bd9Sstevel@tonic-gate continue; 63237c478bd9Sstevel@tonic-gate } 63247c478bd9Sstevel@tonic-gate if ((curr->ire_stq != NULL && ire->ire_stq == NULL) || 63257c478bd9Sstevel@tonic-gate (curr->ire_stq == NULL && ire->ire_stq != NULL)) { 63267c478bd9Sstevel@tonic-gate /* 63277c478bd9Sstevel@tonic-gate * If the passed-in ire is loopback, skip over 63287c478bd9Sstevel@tonic-gate * non-loopback ires and vice versa. 63297c478bd9Sstevel@tonic-gate */ 63307c478bd9Sstevel@tonic-gate continue; 63317c478bd9Sstevel@tonic-gate } 63327c478bd9Sstevel@tonic-gate if (ire_to_ill(curr) != ill) { 63337c478bd9Sstevel@tonic-gate /* skip over IREs going through a different interface */ 63347c478bd9Sstevel@tonic-gate continue; 63357c478bd9Sstevel@tonic-gate } 63367c478bd9Sstevel@tonic-gate if (curr->ire_marks & IRE_MARK_CONDEMNED) { 63377c478bd9Sstevel@tonic-gate /* skip over deleted IREs */ 63387c478bd9Sstevel@tonic-gate continue; 63397c478bd9Sstevel@tonic-gate } 63407c478bd9Sstevel@tonic-gate return (curr); 63417c478bd9Sstevel@tonic-gate } 63427c478bd9Sstevel@tonic-gate return (NULL); 63437c478bd9Sstevel@tonic-gate } 63447c478bd9Sstevel@tonic-gate 63457c478bd9Sstevel@tonic-gate #ifdef IRE_DEBUG 63467c478bd9Sstevel@tonic-gate th_trace_t * 63477c478bd9Sstevel@tonic-gate th_trace_ire_lookup(ire_t *ire) 63487c478bd9Sstevel@tonic-gate { 63497c478bd9Sstevel@tonic-gate int bucket_id; 63507c478bd9Sstevel@tonic-gate th_trace_t *th_trace; 63517c478bd9Sstevel@tonic-gate 63527c478bd9Sstevel@tonic-gate ASSERT(MUTEX_HELD(&ire->ire_lock)); 63537c478bd9Sstevel@tonic-gate 63547c478bd9Sstevel@tonic-gate bucket_id = IP_TR_HASH(curthread); 63557c478bd9Sstevel@tonic-gate ASSERT(bucket_id < IP_TR_HASH_MAX); 63567c478bd9Sstevel@tonic-gate 63577c478bd9Sstevel@tonic-gate for (th_trace = ire->ire_trace[bucket_id]; th_trace != NULL; 63587c478bd9Sstevel@tonic-gate th_trace = th_trace->th_next) { 63597c478bd9Sstevel@tonic-gate if (th_trace->th_id == curthread) 63607c478bd9Sstevel@tonic-gate return (th_trace); 63617c478bd9Sstevel@tonic-gate } 63627c478bd9Sstevel@tonic-gate return (NULL); 63637c478bd9Sstevel@tonic-gate } 63647c478bd9Sstevel@tonic-gate 63657c478bd9Sstevel@tonic-gate void 63667c478bd9Sstevel@tonic-gate ire_trace_ref(ire_t *ire) 63677c478bd9Sstevel@tonic-gate { 63687c478bd9Sstevel@tonic-gate int bucket_id; 63697c478bd9Sstevel@tonic-gate th_trace_t *th_trace; 63707c478bd9Sstevel@tonic-gate 63717c478bd9Sstevel@tonic-gate /* 63727c478bd9Sstevel@tonic-gate * Attempt to locate the trace buffer for the curthread. 63737c478bd9Sstevel@tonic-gate * If it does not exist, then allocate a new trace buffer 63747c478bd9Sstevel@tonic-gate * and link it in list of trace bufs for this ipif, at the head 63757c478bd9Sstevel@tonic-gate */ 63767c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 63777c478bd9Sstevel@tonic-gate if (ire->ire_trace_disable == B_TRUE) { 63787c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 63797c478bd9Sstevel@tonic-gate return; 63807c478bd9Sstevel@tonic-gate } 63817c478bd9Sstevel@tonic-gate th_trace = th_trace_ire_lookup(ire); 63827c478bd9Sstevel@tonic-gate if (th_trace == NULL) { 63837c478bd9Sstevel@tonic-gate bucket_id = IP_TR_HASH(curthread); 63847c478bd9Sstevel@tonic-gate th_trace = (th_trace_t *)kmem_zalloc(sizeof (th_trace_t), 63857c478bd9Sstevel@tonic-gate KM_NOSLEEP); 63867c478bd9Sstevel@tonic-gate if (th_trace == NULL) { 63877c478bd9Sstevel@tonic-gate ire->ire_trace_disable = B_TRUE; 63887c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 63897c478bd9Sstevel@tonic-gate ire_trace_inactive(ire); 63907c478bd9Sstevel@tonic-gate return; 63917c478bd9Sstevel@tonic-gate } 63927c478bd9Sstevel@tonic-gate 63937c478bd9Sstevel@tonic-gate th_trace->th_id = curthread; 63947c478bd9Sstevel@tonic-gate th_trace->th_next = ire->ire_trace[bucket_id]; 63957c478bd9Sstevel@tonic-gate th_trace->th_prev = &ire->ire_trace[bucket_id]; 63967c478bd9Sstevel@tonic-gate if (th_trace->th_next != NULL) 63977c478bd9Sstevel@tonic-gate th_trace->th_next->th_prev = &th_trace->th_next; 63987c478bd9Sstevel@tonic-gate ire->ire_trace[bucket_id] = th_trace; 63997c478bd9Sstevel@tonic-gate } 64007c478bd9Sstevel@tonic-gate ASSERT(th_trace->th_refcnt < TR_BUF_MAX - 1); 64017c478bd9Sstevel@tonic-gate th_trace->th_refcnt++; 64027c478bd9Sstevel@tonic-gate th_trace_rrecord(th_trace); 64037c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 64047c478bd9Sstevel@tonic-gate } 64057c478bd9Sstevel@tonic-gate 64067c478bd9Sstevel@tonic-gate void 64077c478bd9Sstevel@tonic-gate ire_trace_free(th_trace_t *th_trace) 64087c478bd9Sstevel@tonic-gate { 64097c478bd9Sstevel@tonic-gate /* unlink th_trace and free it */ 64107c478bd9Sstevel@tonic-gate *th_trace->th_prev = th_trace->th_next; 64117c478bd9Sstevel@tonic-gate if (th_trace->th_next != NULL) 64127c478bd9Sstevel@tonic-gate th_trace->th_next->th_prev = th_trace->th_prev; 64137c478bd9Sstevel@tonic-gate th_trace->th_next = NULL; 64147c478bd9Sstevel@tonic-gate th_trace->th_prev = NULL; 64157c478bd9Sstevel@tonic-gate kmem_free(th_trace, sizeof (th_trace_t)); 64167c478bd9Sstevel@tonic-gate } 64177c478bd9Sstevel@tonic-gate 64187c478bd9Sstevel@tonic-gate void 64197c478bd9Sstevel@tonic-gate ire_untrace_ref(ire_t *ire) 64207c478bd9Sstevel@tonic-gate { 64217c478bd9Sstevel@tonic-gate th_trace_t *th_trace; 64227c478bd9Sstevel@tonic-gate 64237c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 64247c478bd9Sstevel@tonic-gate 64257c478bd9Sstevel@tonic-gate if (ire->ire_trace_disable == B_TRUE) { 64267c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 64277c478bd9Sstevel@tonic-gate return; 64287c478bd9Sstevel@tonic-gate } 64297c478bd9Sstevel@tonic-gate 64307c478bd9Sstevel@tonic-gate th_trace = th_trace_ire_lookup(ire); 64317c478bd9Sstevel@tonic-gate ASSERT(th_trace != NULL && th_trace->th_refcnt > 0); 64327c478bd9Sstevel@tonic-gate th_trace_rrecord(th_trace); 64337c478bd9Sstevel@tonic-gate th_trace->th_refcnt--; 64347c478bd9Sstevel@tonic-gate 64357c478bd9Sstevel@tonic-gate if (th_trace->th_refcnt == 0) 64367c478bd9Sstevel@tonic-gate ire_trace_free(th_trace); 64377c478bd9Sstevel@tonic-gate 64387c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 64397c478bd9Sstevel@tonic-gate } 64407c478bd9Sstevel@tonic-gate 64417c478bd9Sstevel@tonic-gate static void 64427c478bd9Sstevel@tonic-gate ire_trace_inactive(ire_t *ire) 64437c478bd9Sstevel@tonic-gate { 64447c478bd9Sstevel@tonic-gate th_trace_t *th_trace; 64457c478bd9Sstevel@tonic-gate int i; 64467c478bd9Sstevel@tonic-gate 64477c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 64487c478bd9Sstevel@tonic-gate for (i = 0; i < IP_TR_HASH_MAX; i++) { 64497c478bd9Sstevel@tonic-gate while (ire->ire_trace[i] != NULL) { 64507c478bd9Sstevel@tonic-gate th_trace = ire->ire_trace[i]; 64517c478bd9Sstevel@tonic-gate 64527c478bd9Sstevel@tonic-gate /* unlink th_trace and free it */ 64537c478bd9Sstevel@tonic-gate ire->ire_trace[i] = th_trace->th_next; 64547c478bd9Sstevel@tonic-gate if (th_trace->th_next != NULL) 64557c478bd9Sstevel@tonic-gate th_trace->th_next->th_prev = 64567c478bd9Sstevel@tonic-gate &ire->ire_trace[i]; 64577c478bd9Sstevel@tonic-gate 64587c478bd9Sstevel@tonic-gate th_trace->th_next = NULL; 64597c478bd9Sstevel@tonic-gate th_trace->th_prev = NULL; 64607c478bd9Sstevel@tonic-gate kmem_free(th_trace, sizeof (th_trace_t)); 64617c478bd9Sstevel@tonic-gate } 64627c478bd9Sstevel@tonic-gate } 64637c478bd9Sstevel@tonic-gate 64647c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 64657c478bd9Sstevel@tonic-gate } 64667c478bd9Sstevel@tonic-gate 64677c478bd9Sstevel@tonic-gate /* ARGSUSED */ 64687c478bd9Sstevel@tonic-gate void 64697c478bd9Sstevel@tonic-gate ire_thread_exit(ire_t *ire, caddr_t arg) 64707c478bd9Sstevel@tonic-gate { 64717c478bd9Sstevel@tonic-gate th_trace_t *th_trace; 64727c478bd9Sstevel@tonic-gate 64737c478bd9Sstevel@tonic-gate mutex_enter(&ire->ire_lock); 64747c478bd9Sstevel@tonic-gate th_trace = th_trace_ire_lookup(ire); 64757c478bd9Sstevel@tonic-gate if (th_trace == NULL) { 64767c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 64777c478bd9Sstevel@tonic-gate return; 64787c478bd9Sstevel@tonic-gate } 64797c478bd9Sstevel@tonic-gate ASSERT(th_trace->th_refcnt == 0); 64807c478bd9Sstevel@tonic-gate 64817c478bd9Sstevel@tonic-gate ire_trace_free(th_trace); 64827c478bd9Sstevel@tonic-gate mutex_exit(&ire->ire_lock); 64837c478bd9Sstevel@tonic-gate } 64847c478bd9Sstevel@tonic-gate 64857c478bd9Sstevel@tonic-gate #endif 6486c793af95Ssangeeta 6487c793af95Ssangeeta /* 6488c793af95Ssangeeta * Generate a message chain with an arp request to resolve the in_ire. 6489c793af95Ssangeeta * It is assumed that in_ire itself is currently in the ire cache table, 6490c793af95Ssangeeta * so we create a fake_ire filled with enough information about ire_addr etc. 6491c793af95Ssangeeta * to retrieve in_ire when the DL_UNITDATA response from the resolver 6492c793af95Ssangeeta * comes back. The fake_ire itself is created by calling esballoc with 6493c793af95Ssangeeta * the fr_rtnp (free routine) set to ire_freemblk. This routine will be 6494c793af95Ssangeeta * invoked when the mblk containing fake_ire is freed. 6495c793af95Ssangeeta */ 6496c793af95Ssangeeta void 6497c793af95Ssangeeta ire_arpresolve(ire_t *in_ire, ill_t *dst_ill) 6498c793af95Ssangeeta { 6499c793af95Ssangeeta areq_t *areq; 6500c793af95Ssangeeta ipaddr_t *addrp; 6501c793af95Ssangeeta mblk_t *ire_mp, *dlureq_mp; 6502c793af95Ssangeeta ire_t *ire, *buf; 6503c793af95Ssangeeta size_t bufsize; 6504c793af95Ssangeeta frtn_t *frtnp; 6505c793af95Ssangeeta ill_t *ill; 6506*f4b3ec61Sdh155122 ip_stack_t *ipst = dst_ill->ill_ipst; 6507c793af95Ssangeeta 6508c793af95Ssangeeta /* 6509c793af95Ssangeeta * Construct message chain for the resolver 6510c793af95Ssangeeta * of the form: 6511c793af95Ssangeeta * ARP_REQ_MBLK-->IRE_MBLK 6512c793af95Ssangeeta * 6513c793af95Ssangeeta * NOTE : If the response does not 6514c793af95Ssangeeta * come back, ARP frees the packet. For this reason, 6515c793af95Ssangeeta * we can't REFHOLD the bucket of save_ire to prevent 6516c793af95Ssangeeta * deletions. We may not be able to REFRELE the bucket 6517c793af95Ssangeeta * if the response never comes back. Thus, before 6518c793af95Ssangeeta * adding the ire, ire_add_v4 will make sure that the 6519c793af95Ssangeeta * interface route does not get deleted. This is the 6520c793af95Ssangeeta * only case unlike ip_newroute_v6, ip_newroute_ipif_v6 6521c793af95Ssangeeta * where we can always prevent deletions because of 6522c793af95Ssangeeta * the synchronous nature of adding IRES i.e 6523c793af95Ssangeeta * ire_add_then_send is called after creating the IRE. 6524c793af95Ssangeeta */ 6525c793af95Ssangeeta 6526c793af95Ssangeeta /* 6527c793af95Ssangeeta * We use esballoc to allocate the second part(the ire_t size mblk) 6528c793af95Ssangeeta * of the message chain depicted above. THis mblk will be freed 6529c793af95Ssangeeta * by arp when there is a timeout, and otherwise passed to IP 6530c793af95Ssangeeta * and IP will * free it after processing the ARP response. 6531c793af95Ssangeeta */ 6532c793af95Ssangeeta 6533c793af95Ssangeeta bufsize = sizeof (ire_t) + sizeof (frtn_t); 6534c793af95Ssangeeta buf = kmem_alloc(bufsize, KM_NOSLEEP); 6535c793af95Ssangeeta if (buf == NULL) { 6536c793af95Ssangeeta ip1dbg(("ire_arpresolver:alloc buffer failed\n ")); 6537c793af95Ssangeeta return; 6538c793af95Ssangeeta } 6539c793af95Ssangeeta frtnp = (frtn_t *)(buf + 1); 6540c793af95Ssangeeta frtnp->free_arg = (caddr_t)buf; 6541c793af95Ssangeeta frtnp->free_func = ire_freemblk; 6542c793af95Ssangeeta 6543c793af95Ssangeeta ire_mp = esballoc((unsigned char *)buf, bufsize, BPRI_MED, frtnp); 6544c793af95Ssangeeta 6545c793af95Ssangeeta if (ire_mp == NULL) { 6546c793af95Ssangeeta ip1dbg(("ire_arpresolve: esballoc failed\n")); 6547c793af95Ssangeeta kmem_free(buf, bufsize); 6548c793af95Ssangeeta return; 6549c793af95Ssangeeta } 6550c793af95Ssangeeta ASSERT(in_ire->ire_nce != NULL); 6551c793af95Ssangeeta dlureq_mp = copyb(dst_ill->ill_resolver_mp); 6552c793af95Ssangeeta if (dlureq_mp == NULL) { 6553c793af95Ssangeeta kmem_free(buf, bufsize); 6554c793af95Ssangeeta return; 6555c793af95Ssangeeta } 6556c793af95Ssangeeta 6557c793af95Ssangeeta ire_mp->b_datap->db_type = IRE_ARPRESOLVE_TYPE; 6558c793af95Ssangeeta ire = (ire_t *)buf; 6559c793af95Ssangeeta /* 6560c793af95Ssangeeta * keep enough info in the fake ire so that we can pull up 6561c793af95Ssangeeta * the incomplete ire (in_ire) after result comes back from 6562c793af95Ssangeeta * arp and make it complete. 6563c793af95Ssangeeta */ 6564c793af95Ssangeeta *ire = ire_null; 6565c793af95Ssangeeta ire->ire_u = in_ire->ire_u; 6566c793af95Ssangeeta ire->ire_ipif_seqid = in_ire->ire_ipif_seqid; 6567c793af95Ssangeeta ire->ire_ipif = in_ire->ire_ipif; 6568c793af95Ssangeeta ire->ire_stq = in_ire->ire_stq; 6569c793af95Ssangeeta ill = ire_to_ill(ire); 6570c793af95Ssangeeta ire->ire_stq_ifindex = ill->ill_phyint->phyint_ifindex; 6571c793af95Ssangeeta ire->ire_zoneid = in_ire->ire_zoneid; 6572*f4b3ec61Sdh155122 ire->ire_ipst = ipst; 6573*f4b3ec61Sdh155122 6574c793af95Ssangeeta /* 6575c793af95Ssangeeta * ire_freemblk will be called when ire_mp is freed, both for 6576c793af95Ssangeeta * successful and failed arp resolution. IRE_MARK_UNCACHED will be set 6577c793af95Ssangeeta * when the arp resolution failed. 6578c793af95Ssangeeta */ 6579c793af95Ssangeeta ire->ire_marks |= IRE_MARK_UNCACHED; 6580c793af95Ssangeeta ire->ire_mp = ire_mp; 6581c793af95Ssangeeta ire_mp->b_wptr = (uchar_t *)&ire[1]; 6582c793af95Ssangeeta ire_mp->b_cont = NULL; 6583c793af95Ssangeeta ASSERT(dlureq_mp != NULL); 6584c793af95Ssangeeta linkb(dlureq_mp, ire_mp); 6585c793af95Ssangeeta 6586c793af95Ssangeeta /* 6587c793af95Ssangeeta * Fill in the source and dest addrs for the resolver. 6588c793af95Ssangeeta * NOTE: this depends on memory layouts imposed by 6589c793af95Ssangeeta * ill_init(). 6590c793af95Ssangeeta */ 6591c793af95Ssangeeta areq = (areq_t *)dlureq_mp->b_rptr; 6592c793af95Ssangeeta addrp = (ipaddr_t *)((char *)areq + areq->areq_sender_addr_offset); 6593c793af95Ssangeeta *addrp = ire->ire_src_addr; 6594c793af95Ssangeeta 6595c793af95Ssangeeta addrp = (ipaddr_t *)((char *)areq + areq->areq_target_addr_offset); 6596c793af95Ssangeeta if (ire->ire_gateway_addr != INADDR_ANY) { 6597c793af95Ssangeeta *addrp = ire->ire_gateway_addr; 6598c793af95Ssangeeta } else { 6599c793af95Ssangeeta *addrp = ire->ire_addr; 6600c793af95Ssangeeta } 6601c793af95Ssangeeta 6602c793af95Ssangeeta /* Up to the resolver. */ 6603c793af95Ssangeeta if (canputnext(dst_ill->ill_rq)) { 6604c793af95Ssangeeta putnext(dst_ill->ill_rq, dlureq_mp); 6605c793af95Ssangeeta } else { 6606c793af95Ssangeeta /* Prepare for cleanup */ 6607c793af95Ssangeeta freemsg(dlureq_mp); 6608c793af95Ssangeeta } 6609c793af95Ssangeeta } 6610c793af95Ssangeeta 6611c793af95Ssangeeta /* 6612c793af95Ssangeeta * Esballoc free function for AR_ENTRY_QUERY request to clean up any 6613c793af95Ssangeeta * unresolved ire_t and/or nce_t structures when ARP resolution fails. 6614c793af95Ssangeeta * 6615c793af95Ssangeeta * This function can be called by ARP via free routine for ire_mp or 6616c793af95Ssangeeta * by IPv4(both host and forwarding path) via ire_delete 6617c793af95Ssangeeta * in case ARP resolution fails. 6618c793af95Ssangeeta * NOTE: Since IP is MT, ARP can call into IP but not vice versa 6619c793af95Ssangeeta * (for IP to talk to ARP, it still has to send AR* messages). 6620c793af95Ssangeeta * 6621c793af95Ssangeeta * Note that the ARP/IP merge should replace the functioanlity by providing 6622c793af95Ssangeeta * direct function calls to clean up unresolved entries in ire/nce lists. 6623c793af95Ssangeeta */ 6624c793af95Ssangeeta void 6625c793af95Ssangeeta ire_freemblk(ire_t *ire_mp) 6626c793af95Ssangeeta { 6627c793af95Ssangeeta nce_t *nce = NULL; 6628c793af95Ssangeeta ill_t *ill; 6629*f4b3ec61Sdh155122 ip_stack_t *ipst; 6630c793af95Ssangeeta 6631c793af95Ssangeeta ASSERT(ire_mp != NULL); 6632c793af95Ssangeeta 6633c793af95Ssangeeta if ((ire_mp->ire_addr == NULL) && (ire_mp->ire_gateway_addr == NULL)) { 6634c793af95Ssangeeta ip1dbg(("ire_freemblk(0x%p) ire_addr is NULL\n", 6635c793af95Ssangeeta (void *)ire_mp)); 6636c793af95Ssangeeta goto cleanup; 6637c793af95Ssangeeta } 6638c793af95Ssangeeta if ((ire_mp->ire_marks & IRE_MARK_UNCACHED) == 0) { 6639c793af95Ssangeeta goto cleanup; /* everything succeeded. just free and return */ 6640c793af95Ssangeeta } 6641c793af95Ssangeeta 6642c793af95Ssangeeta /* 6643c793af95Ssangeeta * the arp information corresponding to this ire_mp was not 6644c793af95Ssangeeta * transferred to a ire_cache entry. Need 6645c793af95Ssangeeta * to clean up incomplete ire's and nce, if necessary. 6646c793af95Ssangeeta */ 6647c793af95Ssangeeta ASSERT(ire_mp->ire_stq != NULL); 6648c793af95Ssangeeta ASSERT(ire_mp->ire_stq_ifindex != 0); 6649*f4b3ec61Sdh155122 ASSERT(ire_mp->ire_ipst != NULL); 6650*f4b3ec61Sdh155122 6651*f4b3ec61Sdh155122 ipst = ire_mp->ire_ipst; 6652*f4b3ec61Sdh155122 6653c793af95Ssangeeta /* 6654c793af95Ssangeeta * Get any nce's corresponding to this ire_mp. We first have to 6655c793af95Ssangeeta * make sure that the ill is still around. 6656c793af95Ssangeeta */ 6657*f4b3ec61Sdh155122 ill = ill_lookup_on_ifindex(ire_mp->ire_stq_ifindex, 6658*f4b3ec61Sdh155122 B_FALSE, NULL, NULL, NULL, NULL, ipst); 6659c793af95Ssangeeta if (ill == NULL || (ire_mp->ire_stq != ill->ill_wq) || 6660c793af95Ssangeeta (ill->ill_state_flags & ILL_CONDEMNED)) { 6661c793af95Ssangeeta /* 6662c793af95Ssangeeta * ill went away. no nce to clean up. 6663c793af95Ssangeeta * Note that the ill_state_flags could be set to 6664c793af95Ssangeeta * ILL_CONDEMNED after this point, but if we know 6665c793af95Ssangeeta * that it is CONDEMNED now, we just bail out quickly. 6666c793af95Ssangeeta */ 6667c793af95Ssangeeta if (ill != NULL) 6668c793af95Ssangeeta ill_refrele(ill); 6669c793af95Ssangeeta goto cleanup; 6670c793af95Ssangeeta } 6671c793af95Ssangeeta nce = ndp_lookup_v4(ill, 6672c793af95Ssangeeta ((ire_mp->ire_gateway_addr != INADDR_ANY) ? 6673c793af95Ssangeeta &ire_mp->ire_gateway_addr : &ire_mp->ire_addr), 6674c793af95Ssangeeta B_FALSE); 6675c793af95Ssangeeta ill_refrele(ill); 6676c793af95Ssangeeta 6677c793af95Ssangeeta if ((nce != NULL) && (nce->nce_state != ND_REACHABLE)) { 6678c793af95Ssangeeta /* 6679c793af95Ssangeeta * some incomplete nce was found. 6680c793af95Ssangeeta */ 6681c793af95Ssangeeta DTRACE_PROBE2(ire__freemblk__arp__resolv__fail, 6682c793af95Ssangeeta nce_t *, nce, ire_t *, ire_mp); 6683c793af95Ssangeeta /* 6684c793af95Ssangeeta * Send the icmp_unreachable messages for the queued mblks in 6685c793af95Ssangeeta * ire->ire_nce->nce_qd_mp, since ARP resolution failed 6686c793af95Ssangeeta * for this ire 6687c793af95Ssangeeta */ 6688c793af95Ssangeeta arp_resolv_failed(nce); 6689c793af95Ssangeeta /* 6690c793af95Ssangeeta * Delete the nce and clean up all ire's pointing at this nce 6691c793af95Ssangeeta * in the cachetable 6692c793af95Ssangeeta */ 6693c793af95Ssangeeta ndp_delete(nce); 6694c793af95Ssangeeta } 6695c793af95Ssangeeta if (nce != NULL) 6696c793af95Ssangeeta NCE_REFRELE(nce); /* release the ref taken by ndp_lookup_v4 */ 6697c793af95Ssangeeta 6698c793af95Ssangeeta cleanup: 6699c793af95Ssangeeta /* 6700c793af95Ssangeeta * Get rid of the ire buffer 6701c793af95Ssangeeta * We call kmem_free here(instead of ire_delete()), since 6702c793af95Ssangeeta * this is the freeb's callback. 6703c793af95Ssangeeta */ 6704c793af95Ssangeeta kmem_free(ire_mp, sizeof (ire_t) + sizeof (frtn_t)); 6705c793af95Ssangeeta } 6706c793af95Ssangeeta 6707c793af95Ssangeeta 6708c793af95Ssangeeta /* 6709c793af95Ssangeeta * create the neighbor cache entry nce_t for IRE_CACHE and 6710c793af95Ssangeeta * non-loopback IRE_BROADCAST ire's. Note that IRE_BROADCAST 6711c793af95Ssangeeta * (non-loopback) entries have the nce_res_mp set to the 6712c793af95Ssangeeta * template passed in (generated from ill_bcast_mp); IRE_CACHE ire's 6713c793af95Ssangeeta * contain the information for the nexthop (ire_gateway_addr) in the 6714c793af95Ssangeeta * case of indirect routes, and for the dst itself (ire_addr) in the 6715c793af95Ssangeeta * case of direct routes, with the nce_res_mp containing a template 6716c793af95Ssangeeta * DL_UNITDATA request. 6717c793af95Ssangeeta * 6718c793af95Ssangeeta * This function always consumes res_mp and fp_mp. 6719c793af95Ssangeeta * 6720c793af95Ssangeeta * The actual association of the ire_nce to the nce created here is 6721c793af95Ssangeeta * typically done in ire_add_v4 for IRE_CACHE entries. Exceptions 6722c793af95Ssangeeta * to this rule are SO_DONTROUTE ire's (IRE_MARK_NO_ADD), for which 6723c793af95Ssangeeta * the ire_nce assignment is done in ire_add_then_send, and mobile-ip 6724c793af95Ssangeeta * where the assignment is done in ire_add_mrtun(). 6725c793af95Ssangeeta */ 6726c793af95Ssangeeta int 6727c793af95Ssangeeta ire_nce_init(ire_t *ire, mblk_t *fp_mp, mblk_t *res_mp) 6728c793af95Ssangeeta { 6729c793af95Ssangeeta in_addr_t addr4, mask4; 6730c793af95Ssangeeta int err; 6731c793af95Ssangeeta nce_t *arpce = NULL; 6732c793af95Ssangeeta ill_t *ire_ill; 6733c793af95Ssangeeta uint16_t nce_state, nce_flags; 6734*f4b3ec61Sdh155122 ip_stack_t *ipst; 6735c793af95Ssangeeta 6736c793af95Ssangeeta if (ire->ire_stq == NULL) { 6737c793af95Ssangeeta if (res_mp) 6738c793af95Ssangeeta freemsg(res_mp); 6739c793af95Ssangeeta if (fp_mp) 6740c793af95Ssangeeta freemsg(fp_mp); 6741c793af95Ssangeeta return (0); /* no need to create nce for local/loopback */ 6742c793af95Ssangeeta } 6743c793af95Ssangeeta 6744c793af95Ssangeeta mask4 = IP_HOST_MASK; 6745c793af95Ssangeeta switch (ire->ire_type) { 6746c793af95Ssangeeta case IRE_CACHE: 6747c793af95Ssangeeta if (ire->ire_gateway_addr != INADDR_ANY) 6748c793af95Ssangeeta addr4 = ire->ire_gateway_addr; /* 'G' route */ 6749c793af95Ssangeeta else 6750c793af95Ssangeeta addr4 = ire->ire_addr; /* direct route */ 6751c793af95Ssangeeta break; 6752c793af95Ssangeeta case IRE_BROADCAST: 6753c793af95Ssangeeta addr4 = ire->ire_addr; 6754c793af95Ssangeeta break; 6755c793af95Ssangeeta default: 6756c793af95Ssangeeta if (res_mp) 6757c793af95Ssangeeta freemsg(res_mp); 6758c793af95Ssangeeta if (fp_mp) 6759c793af95Ssangeeta freemsg(fp_mp); 6760c793af95Ssangeeta return (0); 6761c793af95Ssangeeta } 6762c793af95Ssangeeta 6763c793af95Ssangeeta /* 6764c793af95Ssangeeta * ire_ipif is picked based on RTF_SETSRC, usesrc etc. 6765c793af95Ssangeeta * rules in ire_forward_src_ipif. We want the dlureq_mp 6766c793af95Ssangeeta * for the outgoing interface, which we get from the ire_stq. 6767c793af95Ssangeeta */ 6768c793af95Ssangeeta ire_ill = ire_to_ill(ire); 6769*f4b3ec61Sdh155122 ipst = ire_ill->ill_ipst; 6770c793af95Ssangeeta 6771c793af95Ssangeeta /* 6772c793af95Ssangeeta * if we are creating an nce for the first time, and this is 6773c793af95Ssangeeta * a NORESOLVER interface, atomically create the nce in the 6774c793af95Ssangeeta * REACHABLE state; else create it in the ND_INITIAL state. 6775c793af95Ssangeeta */ 6776c793af95Ssangeeta if (ire_ill->ill_net_type == IRE_IF_NORESOLVER) { 6777c793af95Ssangeeta nce_state = ND_REACHABLE; 6778c793af95Ssangeeta nce_flags = NCE_F_PERMANENT; 6779c793af95Ssangeeta } else { 6780c793af95Ssangeeta if (fp_mp != NULL) 6781c793af95Ssangeeta nce_state = ND_REACHABLE; 6782c793af95Ssangeeta else 6783c793af95Ssangeeta nce_state = ND_INITIAL; 6784c793af95Ssangeeta nce_flags = 0; 6785c793af95Ssangeeta } 6786c793af95Ssangeeta 6787c793af95Ssangeeta err = ndp_lookup_then_add(ire_ill, NULL, 6788c793af95Ssangeeta &addr4, &mask4, NULL, 0, nce_flags, nce_state, &arpce, 6789c793af95Ssangeeta fp_mp, res_mp); 6790c793af95Ssangeeta 6791c793af95Ssangeeta ip1dbg(("ire 0x%p addr 0x%lx mask 0x%lx type 0x%x; " 6792c793af95Ssangeeta "found nce 0x%p err %d\n", (void *)ire, (ulong_t)addr4, 6793c793af95Ssangeeta (ulong_t)mask4, ire->ire_type, (void *)arpce, err)); 6794c793af95Ssangeeta 6795c793af95Ssangeeta switch (err) { 6796c793af95Ssangeeta case 0: 6797c793af95Ssangeeta break; 6798c793af95Ssangeeta case EEXIST: 6799c793af95Ssangeeta /* 6800c793af95Ssangeeta * return a pointer to an existing nce_t; 6801c793af95Ssangeeta * note that the ire-nce mapping is many-one, i.e., 6802c793af95Ssangeeta * multiple ire's could point to the same nce_t; 6803c793af95Ssangeeta */ 6804c793af95Ssangeeta if (fp_mp != NULL) { 6805c793af95Ssangeeta freemsg(fp_mp); 6806c793af95Ssangeeta } 6807c793af95Ssangeeta if (res_mp != NULL) { 6808c793af95Ssangeeta freemsg(res_mp); 6809c793af95Ssangeeta } 6810c793af95Ssangeeta break; 6811c793af95Ssangeeta default: 6812c793af95Ssangeeta DTRACE_PROBE2(nce__init__fail, ill_t *, ire_ill, int, err); 6813c793af95Ssangeeta if (res_mp) 6814c793af95Ssangeeta freemsg(res_mp); 6815c793af95Ssangeeta if (fp_mp) 6816c793af95Ssangeeta freemsg(fp_mp); 6817c793af95Ssangeeta return (EINVAL); 6818c793af95Ssangeeta } 6819c793af95Ssangeeta #if DEBUG 6820c793af95Ssangeeta /* 6821c793af95Ssangeeta * if an nce_fp_mp was passed in, we should be picking up an 6822c793af95Ssangeeta * existing nce_t in the ND_REACHABLE state. 6823c793af95Ssangeeta */ 6824c793af95Ssangeeta mutex_enter(&arpce->nce_lock); 6825c793af95Ssangeeta ASSERT(arpce->nce_fp_mp == NULL || arpce->nce_state == ND_REACHABLE); 6826c793af95Ssangeeta mutex_exit(&arpce->nce_lock); 6827c793af95Ssangeeta #endif 6828c793af95Ssangeeta if (ire->ire_type == IRE_BROADCAST) { 6829c793af95Ssangeeta /* 6830c793af95Ssangeeta * Two bcast ires are created for each interface; 6831c793af95Ssangeeta * 1. loopback copy (which does not have an 6832c793af95Ssangeeta * ire_stq, and therefore has no ire_nce), and, 6833c793af95Ssangeeta * 2. the non-loopback copy, which has the nce_res_mp 6834c793af95Ssangeeta * initialized to a copy of the ill_bcast_mp, and 6835c793af95Ssangeeta * is marked as ND_REACHABLE at this point. 6836c793af95Ssangeeta * This nce does not undergo any further state changes, 6837c793af95Ssangeeta * and exists as long as the interface is plumbed. 6838c793af95Ssangeeta * Note: we do the ire_nce assignment here for IRE_BROADCAST 6839c793af95Ssangeeta * because some functions like ill_mark_bcast() inline the 6840c793af95Ssangeeta * ire_add functionality; 6841c793af95Ssangeeta */ 6842c793af95Ssangeeta mutex_enter(&arpce->nce_lock); 6843c793af95Ssangeeta arpce->nce_state = ND_REACHABLE; 6844516bda92Ssowmini arpce->nce_flags |= (NCE_F_PERMANENT | NCE_F_BCAST); 6845c793af95Ssangeeta arpce->nce_last = TICK_TO_MSEC(lbolt64); 6846c793af95Ssangeeta ire->ire_nce = arpce; 6847c793af95Ssangeeta mutex_exit(&arpce->nce_lock); 6848c793af95Ssangeeta /* 6849c793af95Ssangeeta * We are associating this nce to the ire, 6850c793af95Ssangeeta * so change the nce ref taken in 6851c793af95Ssangeeta * ndp_lookup_then_add_v4() from 6852c793af95Ssangeeta * NCE_REFHOLD to NCE_REFHOLD_NOTR 6853c793af95Ssangeeta */ 6854c793af95Ssangeeta NCE_REFHOLD_TO_REFHOLD_NOTR(ire->ire_nce); 6855c793af95Ssangeeta } else { 6856*f4b3ec61Sdh155122 if (NCE_EXPIRED(arpce, ipst)) 6857c793af95Ssangeeta arpce = nce_reinit(arpce); 6858c793af95Ssangeeta if (arpce != NULL) { 6859c793af95Ssangeeta /* 6860c793af95Ssangeeta * We are not using this nce_t just yet so release 6861c793af95Ssangeeta * the ref taken in ndp_lookup_then_add_v4() 6862c793af95Ssangeeta */ 6863c793af95Ssangeeta NCE_REFRELE(arpce); 6864c793af95Ssangeeta } else { 6865c793af95Ssangeeta ip0dbg(("can't reinit arpce for ill 0x%p;\n", 6866c793af95Ssangeeta (void *)ire_ill)); 6867c793af95Ssangeeta } 6868c793af95Ssangeeta } 6869c793af95Ssangeeta return (0); 6870c793af95Ssangeeta } 6871