1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21 /* 22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #ifndef _INET_IP_NDP_H 27 #define _INET_IP_NDP_H 28 29 #include <sys/mutex.h> 30 #include <sys/stream.h> 31 #include <netinet/in.h> 32 #include <netinet/icmp6.h> 33 #include <inet/ip.h> 34 #include <inet/ip2mac.h> 35 36 /* 37 * Internal definitions for the kernel implementation of the IPv6 38 * Neighbor Discovery Protocol (NDP). 39 */ 40 41 #ifdef __cplusplus 42 extern "C" { 43 #endif 44 45 #ifdef _KERNEL 46 #define NCE_TABLE_SIZE 256 47 /* 48 * callbacks set up with ip2mac interface, waiting for result 49 * of neighbor resolution. 50 */ 51 typedef struct nce_cb_s { 52 list_node_t nce_cb_node; 53 void *nce_cb_id; 54 uint32_t nce_cb_flags; 55 ip2mac_callback_t *nce_cb_func; 56 void *nce_cb_arg; 57 } nce_cb_t; 58 59 #define NCE_CB_DISPATCHED 0x00000001 60 61 /* 62 * NDP Cache Entry 63 */ 64 typedef struct nce_s { 65 struct nce_s *nce_next; /* Hash chain next pointer */ 66 struct nce_s **nce_ptpn; /* Pointer to previous next */ 67 struct ill_s *nce_ill; /* Associated ill */ 68 uint16_t nce_flags; /* See below */ 69 uint16_t nce_state; /* See reachability states in if.h */ 70 int16_t nce_pcnt; /* Probe counter */ 71 uint16_t nce_rcnt; /* Retransmit counter */ 72 in6_addr_t nce_addr; /* address of the nighbor */ 73 in6_addr_t nce_mask; /* If not all ones, mask allows an */ 74 /* entry to respond to requests for a group of addresses, for */ 75 /* instantance multicast addresses */ 76 in6_addr_t nce_extract_mask; /* For mappings */ 77 uint32_t nce_ll_extract_start; /* For mappings */ 78 #define nce_first_mp_to_free nce_fp_mp 79 mblk_t *nce_fp_mp; /* link layer fast path mp */ 80 mblk_t *nce_res_mp; /* DL_UNITDATA_REQ */ 81 mblk_t *nce_qd_mp; /* Head outgoing queued packets */ 82 #define nce_last_mp_to_free nce_qd_mp 83 mblk_t *nce_timer_mp; /* NDP timer mblk */ 84 mblk_t *nce_mp; /* mblk we are in, last to be freed */ 85 uint64_t nce_last; /* Time last reachable in msec */ 86 uint32_t nce_refcnt; /* nce active usage count */ 87 kmutex_t nce_lock; /* See comments on top for what */ 88 /* this field protects */ 89 int nce_unsolicit_count; /* Unsolicited Adv count */ 90 struct nce_s *nce_fastpath; /* for fastpath list */ 91 timeout_id_t nce_timeout_id; 92 uchar_t nce_ipversion; /* IPv4(ARP)/IPv6(NDP) version */ 93 uint_t nce_defense_count; /* number of NDP conflicts */ 94 uint_t nce_defense_time; /* last time defended (secs) */ 95 uint64_t nce_init_time; /* time when it was set to ND_INITIAL */ 96 boolean_t nce_trace_disable; /* True when alloc fails */ 97 list_t nce_cb; 98 uint_t nce_cb_walker_cnt; 99 } nce_t; 100 101 /* 102 * The ndp_g_t structure contains protocol specific information needed 103 * to synchronize and manage neighbor cache entries for IPv4 and IPv6. 104 * There are 2 such structures, ips_ndp4 and ips_ndp6. 105 * ips_ndp6 contains the data structures needed for IPv6 Neighbor Discovery. 106 * ips_ndp4 has IPv4 link layer info in its nce_t structures 107 * Note that the nce_t is not currently used as the arp cache itself; 108 * it is used for the following purposes: 109 * - queue packets in nce_qd_mp while waiting for arp resolution to complete 110 * - nce_{res, fp}_mp are used to track DL_UNITDATA request/responses. 111 * - track state of ARP resolution in the nce_state; 112 * 113 * Locking notes: 114 * ndp_g_lock protects neighbor cache tables access and 115 * insertion/removal of cache entries into/from these tables. 116 * nce_lock protects nce_pcnt, nce_rcnt, nce_qd_mp nce_state, 117 * nce_res_mp, nce_refcnt and nce_last. 118 * nce_refcnt is incremented for every ire pointing to this nce and 119 * every time ndp_lookup() finds an nce. 120 * Should there be a need to obtain nce_lock and ndp_g_lock, ndp_g_lock is 121 * acquired first. 122 * To avoid becoming exclusive when deleting NCEs, ndp_walk() routine holds 123 * the ndp_g_lock (i.e global lock) and marks NCEs to be deleted with 124 * NCE_F_CONDEMNED. When all active users of such NCEs are gone the walk 125 * routine passes a list for deletion to nce_ire_delete_list(). 126 * 127 * When the link-layer address of some onlink host changes, ARP will send 128 * an AR_CN_ANNOUNCE message to ip so that stale neighbor-cache 129 * information will not get used. This message is processed in ip_arp_news() 130 * by walking the nce list, and updating as appropriate. The ndp_g_hw_change 131 * flag is set by ip_arp_news() to notify nce_t users that ip_arp_news() is 132 * in progress. 133 */ 134 typedef struct ndp_g_s { 135 kmutex_t ndp_g_lock; /* Lock protecting cache hash table */ 136 nce_t *nce_mask_entries; /* mask not all ones */ 137 nce_t *nce_hash_tbl[NCE_TABLE_SIZE]; 138 int ndp_g_walker; /* # of active thread walking hash list */ 139 boolean_t ndp_g_walker_cleanup; /* true implies defer deletion. */ 140 int ndp_g_hw_change; /* non-zero if nce flush in progress */ 141 } ndp_g_t; 142 143 #define NDP_HW_CHANGE_INCR(ndp) { \ 144 mutex_enter(&(ndp)->ndp_g_lock); \ 145 (ndp)->ndp_g_hw_change++; \ 146 mutex_exit(&(ndp)->ndp_g_lock); \ 147 } 148 149 #define NDP_HW_CHANGE_DECR(ndp) { \ 150 mutex_enter(&(ndp)->ndp_g_lock); \ 151 (ndp)->ndp_g_hw_change--; \ 152 mutex_exit(&(ndp)->ndp_g_lock); \ 153 } 154 155 /* nce_flags */ 156 #define NCE_F_PERMANENT 0x1 157 #define NCE_F_MAPPING 0x2 158 #define NCE_F_ISROUTER 0x4 159 /* unused 0x8 */ 160 #define NCE_F_NONUD 0x10 161 #define NCE_F_ANYCAST 0x20 162 #define NCE_F_CONDEMNED 0x40 163 #define NCE_F_UNSOL_ADV 0x80 164 #define NCE_F_BCAST 0x100 165 166 #define NCE_EXTERNAL_FLAGS_MASK \ 167 (NCE_F_PERMANENT | NCE_F_MAPPING | NCE_F_ISROUTER | NCE_F_NONUD | \ 168 NCE_F_ANYCAST | NCE_F_UNSOL_ADV) 169 170 /* State REACHABLE, STALE, DELAY or PROBE */ 171 #define NCE_ISREACHABLE(nce) \ 172 (((((nce)->nce_state) >= ND_REACHABLE) && \ 173 ((nce)->nce_state) <= ND_PROBE)) 174 175 /* NDP flags set in SOL/ADV requests */ 176 #define NDP_UNICAST 0x1 177 #define NDP_ISROUTER 0x2 178 #define NDP_SOLICITED 0x4 179 #define NDP_ORIDE 0x8 180 #define NDP_PROBE 0x10 181 182 /* Number of packets queued in NDP for a neighbor */ 183 #define ND_MAX_Q 4 184 185 186 #ifdef DEBUG 187 #define NCE_TRACE_REF(nce) nce_trace_ref(nce) 188 #define NCE_UNTRACE_REF(nce) nce_untrace_ref(nce) 189 #else 190 #define NCE_TRACE_REF(nce) 191 #define NCE_UNTRACE_REF(nce) 192 #endif 193 194 #define NCE_REFHOLD(nce) { \ 195 mutex_enter(&(nce)->nce_lock); \ 196 (nce)->nce_refcnt++; \ 197 ASSERT((nce)->nce_refcnt != 0); \ 198 NCE_TRACE_REF(nce); \ 199 mutex_exit(&(nce)->nce_lock); \ 200 } 201 202 #define NCE_REFHOLD_NOTR(nce) { \ 203 mutex_enter(&(nce)->nce_lock); \ 204 (nce)->nce_refcnt++; \ 205 ASSERT((nce)->nce_refcnt != 0); \ 206 mutex_exit(&(nce)->nce_lock); \ 207 } 208 209 #define NCE_REFHOLD_LOCKED(nce) { \ 210 ASSERT(MUTEX_HELD(&(nce)->nce_lock)); \ 211 (nce)->nce_refcnt++; \ 212 NCE_TRACE_REF(nce); \ 213 } 214 215 /* nce_inactive destroys the mutex thus no mutex_exit is needed */ 216 #define NCE_REFRELE(nce) { \ 217 mutex_enter(&(nce)->nce_lock); \ 218 NCE_UNTRACE_REF(nce); \ 219 ASSERT((nce)->nce_refcnt != 0); \ 220 if (--(nce)->nce_refcnt == 0) \ 221 ndp_inactive(nce); \ 222 else { \ 223 mutex_exit(&(nce)->nce_lock);\ 224 } \ 225 } 226 227 #define NCE_REFRELE_NOTR(nce) { \ 228 mutex_enter(&(nce)->nce_lock); \ 229 ASSERT((nce)->nce_refcnt != 0); \ 230 if (--(nce)->nce_refcnt == 0) \ 231 ndp_inactive(nce); \ 232 else { \ 233 mutex_exit(&(nce)->nce_lock);\ 234 } \ 235 } 236 237 #define NDP_RESTART_TIMER(nce, ms) { \ 238 ASSERT(!MUTEX_HELD(&(nce)->nce_lock)); \ 239 if ((nce)->nce_timeout_id != 0) { \ 240 /* Ok to untimeout bad id. we don't hold a lock. */ \ 241 (void) untimeout((nce)->nce_timeout_id); \ 242 } \ 243 mutex_enter(&(nce)->nce_lock); \ 244 /* Don't start the timer if the nce has been deleted */ \ 245 if (!((nce)->nce_flags & NCE_F_CONDEMNED)) \ 246 nce->nce_timeout_id = timeout(ndp_timer, nce, \ 247 MSEC_TO_TICK(ms) == 0 ? 1 : MSEC_TO_TICK(ms)); \ 248 mutex_exit(&(nce)->nce_lock); \ 249 } 250 251 /* Structure for ndp_cache_count() */ 252 typedef struct { 253 int ncc_total; /* Total number of NCEs */ 254 int ncc_host; /* NCE entries without R bit set */ 255 } ncc_cache_count_t; 256 257 /* 258 * Structure of ndp_cache_reclaim(). Each field is a fraction i.e. 1 means 259 * reclaim all, N means reclaim 1/Nth of all entries, 0 means reclaim none. 260 */ 261 typedef struct { 262 int ncr_host; /* Fraction for host entries */ 263 } nce_cache_reclaim_t; 264 265 /* 266 * Structure for nce_delete_hw_changed; specifies an IPv4 address to link-layer 267 * address mapping. Any route that has a cached copy of a mapping for that 268 * IPv4 address that doesn't match the given mapping must be purged. 269 */ 270 typedef struct { 271 ipaddr_t hwm_addr; /* IPv4 address */ 272 uint_t hwm_hwlen; /* Length of hardware address (may be 0) */ 273 uchar_t *hwm_hwaddr; /* Pointer to new hardware address, if any */ 274 } nce_hw_map_t; 275 276 /* When SAP is greater than zero address appears before SAP */ 277 #define NCE_LL_ADDR_OFFSET(ill) (((ill)->ill_sap_length) < 0 ? \ 278 (sizeof (dl_unitdata_req_t)) : \ 279 ((sizeof (dl_unitdata_req_t)) + (ABS((ill)->ill_sap_length)))) 280 281 #define NCE_LL_SAP_OFFSET(ill) (((ill)->ill_sap_length) < 0 ? \ 282 ((sizeof (dl_unitdata_req_t)) + ((ill)->ill_phys_addr_length)) : \ 283 (sizeof (dl_unitdata_req_t))) 284 285 #ifdef _BIG_ENDIAN 286 #define NCE_LL_SAP_COPY(ill, mp) \ 287 { \ 288 size_t abs_sap_len = ABS((ill)->ill_sap_length); \ 289 if (abs_sap_len > 0) { \ 290 ASSERT(abs_sap_len <= sizeof (uint32_t)); \ 291 ASSERT((mp)->b_rptr + NCE_LL_SAP_OFFSET(ill) + \ 292 abs_sap_len <= ((mp)->b_wptr)); \ 293 bcopy((uint8_t *)&(ill)->ill_sap + sizeof (ill->ill_sap) - \ 294 abs_sap_len, \ 295 ((mp)->b_rptr + NCE_LL_SAP_OFFSET(ill)), \ 296 abs_sap_len); \ 297 } \ 298 } 299 #else 300 #define NCE_LL_SAP_COPY(ill, mp) \ 301 { \ 302 size_t abs_sap_len = ABS((ill)->ill_sap_length); \ 303 if (abs_sap_len > 0) { \ 304 uint32_t abs_sap_len = ABS((ill)->ill_sap_length); \ 305 ASSERT(abs_sap_len <= sizeof (uint32_t)); \ 306 ASSERT((mp)->b_rptr + NCE_LL_SAP_OFFSET(ill) + \ 307 abs_sap_len <= ((mp)->b_wptr)); \ 308 bcopy(&((ill)->ill_sap), \ 309 ((mp)->b_rptr + NCE_LL_SAP_OFFSET(ill)), \ 310 abs_sap_len); \ 311 } \ 312 } 313 #endif 314 315 /* 316 * Exclusive-or the 6 bytes that are likely to contain the MAC 317 * address. Assumes table_size does not exceed 256. 318 * Assumes EUI-64 format for good hashing. 319 */ 320 #define NCE_ADDR_HASH_V6(addr, table_size) \ 321 (((addr).s6_addr8[8] ^ (addr).s6_addr8[9] ^ \ 322 (addr).s6_addr8[10] ^ (addr).s6_addr8[13] ^ \ 323 (addr).s6_addr8[14] ^ (addr).s6_addr8[15]) % (table_size)) 324 325 /* NDP Cache Entry Hash Table */ 326 #define NCE_TABLE_SIZE 256 327 328 extern void ndp_cache_count(nce_t *, char *); 329 extern void ndp_cache_reclaim(nce_t *, char *); 330 extern void ndp_delete(nce_t *); 331 extern void ndp_delete_per_ill(nce_t *, uchar_t *); 332 extern void ndp_fastpath_flush(nce_t *, char *); 333 extern boolean_t ndp_fastpath_update(nce_t *, void *); 334 extern nd_opt_hdr_t *ndp_get_option(nd_opt_hdr_t *, int, int); 335 extern void ndp_inactive(nce_t *); 336 extern void ndp_input(ill_t *, mblk_t *, mblk_t *); 337 extern boolean_t ndp_lookup_ipaddr(in_addr_t, netstack_t *); 338 extern nce_t *ndp_lookup_v6(ill_t *, boolean_t, const in6_addr_t *, 339 boolean_t); 340 extern nce_t *ndp_lookup_v4(ill_t *, const in_addr_t *, boolean_t); 341 extern int ndp_mcastreq(ill_t *, const in6_addr_t *, uint32_t, uint32_t, 342 mblk_t *); 343 extern int ndp_noresolver(ill_t *, const in6_addr_t *); 344 extern void ndp_process(nce_t *, uchar_t *, uint32_t, boolean_t); 345 extern int ndp_query(ill_t *, lif_nd_req_t *); 346 extern int ndp_resolver(ill_t *, const in6_addr_t *, mblk_t *, zoneid_t); 347 extern int ndp_sioc_update(ill_t *, lif_nd_req_t *); 348 extern boolean_t ndp_verify_optlen(nd_opt_hdr_t *, int); 349 extern void ndp_timer(void *); 350 extern void ndp_walk(ill_t *, pfi_t, void *, ip_stack_t *); 351 extern void ndp_walk_common(ndp_g_t *, ill_t *, pfi_t, 352 void *, boolean_t); 353 extern boolean_t ndp_restart_dad(nce_t *); 354 extern void ndp_do_recovery(ipif_t *); 355 extern void nce_resolv_failed(nce_t *); 356 extern void arp_resolv_failed(nce_t *); 357 extern void nce_fastpath_list_add(nce_t *); 358 extern void nce_fastpath_list_delete(nce_t *); 359 extern void nce_fastpath_list_dispatch(ill_t *, 360 boolean_t (*)(nce_t *, void *), void *); 361 extern void nce_queue_mp_common(nce_t *, mblk_t *, boolean_t); 362 extern void nce_delete_hw_changed(nce_t *, void *); 363 extern void nce_fastpath(nce_t *); 364 extern int ndp_add_v6(ill_t *, uchar_t *, const in6_addr_t *, 365 const in6_addr_t *, const in6_addr_t *, uint32_t, uint16_t, uint16_t, 366 nce_t **); 367 extern int ndp_lookup_then_add_v6(ill_t *, boolean_t, uchar_t *, 368 const in6_addr_t *, const in6_addr_t *, const in6_addr_t *, uint32_t, 369 uint16_t, uint16_t, nce_t **); 370 extern int ndp_lookup_then_add_v4(ill_t *, 371 const in_addr_t *, uint16_t, nce_t **, nce_t *); 372 extern void ip_ndp_resolve(nce_t *); 373 374 #ifdef DEBUG 375 extern void nce_trace_ref(nce_t *); 376 extern void nce_untrace_ref(nce_t *); 377 #endif 378 379 #endif /* _KERNEL */ 380 381 #ifdef __cplusplus 382 } 383 #endif 384 385 #endif /* _INET_IP_NDP_H */ 386