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 2006 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #ifndef _MPD_TABLES_H 27 #define _MPD_TABLES_H 28 29 #pragma ident "%Z%%M% %I% %E% SMI" 30 31 #ifdef __cplusplus 32 extern "C" { 33 #endif 34 35 /* 36 * Terminology: 37 * 38 * phyint: A NIC eg. hme0. This is represented as 'struct phyint' 39 * 40 * phyint instance: A protocol instance of a phyint. Eg. the IPv4 instance of 41 * hme0 or the IPv6 instance of hme0. (struct phyint_instance) 42 * 43 * logint: A logical interface eg. hme0:1 (struct logint) 44 * 45 * phyint_group: A group of phyints i.e. physical interfaces that are 46 * (i) connected to the same level 2 topology e.g. the same ethernet 47 * switch AND 48 * (ii) share the same phyint group name. 49 * Load spreading and failover occur across members of the same phyint group. 50 * phyint group members must be homogenous. i.e. if a phyint belonging to a 51 * phyint group has a IPv6 protocol instance, then all members of the phyint 52 * group, must have IPv6 protocol instances. (struct phyint_group) 53 */ 54 55 /* 56 * Parameter passed to try_failover(), indicating the type of failover 57 * that is requested. 58 */ 59 #define FAILOVER_NORMAL 1 /* Failover to another phyint */ 60 /* that is preferably a standby */ 61 #define FAILOVER_TO_NONSTANDBY 2 /* Failover to non-standby phyint */ 62 #define FAILOVER_TO_ANY 3 /* Failover to any available phyint */ 63 64 #define MAXDEFERREDRTT 1 /* Maximum number of deferred rtts */ 65 66 /* 67 * Status of the phyint, expressed by the return code of failure_state() 68 */ 69 #define PHYINT_OK 0 /* No failure detected */ 70 #define PHYINT_FAILURE 1 /* NIC failure detected */ 71 #define GROUP_FAILURE 2 /* All NICs have failed */ 72 73 /* 74 * Return values of phyint_inst_update_from_k() 75 */ 76 #define PI_OK 1 /* Phyint matches in the kernel */ 77 #define PI_DELETED 2 /* Phyint has vanished in the kernel */ 78 #define PI_IFINDEX_CHANGED 3 /* Phyint's ifindex has changed */ 79 #define PI_IOCTL_ERROR 4 /* Some ioctl error */ 80 #define PI_GROUP_CHANGED 5 /* The phyint has changed group. */ 81 82 /* 83 * Though IFF_POINTOPOINT is a logint property, for the purpose of 84 * failover, we treat it as a phyint property. Note that we cannot failover 85 * individual logints. 86 */ 87 #define PHYINT_FLAGS(flags) \ 88 (((flags) & (IFF_STANDBY | IFF_INACTIVE | IFF_FAILED | IFF_OFFLINE | \ 89 IFF_POINTOPOINT | IFF_RUNNING)) | (handle_link_notifications ? \ 90 0 : IFF_RUNNING)) 91 92 /* A Phyint can have up to 2 instances, the IPv4 and the IPv6 instance */ 93 #define PHYINT_INSTANCE(pi, af) \ 94 ((af) == AF_INET ? (pi)->pi_v4 : (pi)->pi_v6) 95 96 /* 97 * A phyint instance is probe *enabled* if it has been configured with a probe 98 * address (i.e., an IFF_NOFAILOVER address). It is probe *capable* if it is 99 * also able to send probes (i.e., has one or more targets available). 100 */ 101 #define PROBE_ENABLED(pii) \ 102 (((pii) != NULL) && ((pii)->pii_probe_logint != NULL)) 103 104 #define PROBE_CAPABLE(pii) \ 105 (PROBE_ENABLED(pii) && ((pii)->pii_ntargets != 0)) 106 107 /* Subtract b from a modulo n. i.e. (a - b) mod n */ 108 #define MOD_SUB(a, b, n) \ 109 ((((a) + (n)) - (b)) % (n)) 110 111 /* Increment modulo n */ 112 #define MOD_INCR(a, n) \ 113 (((a) + 1) % (n)) 114 115 /* Decrement modulo n */ 116 #define MOD_DCR(a, n) \ 117 MOD_SUB(a, 1, n) 118 119 /* 120 * 'index' represents an index into the circular probe stats array of 121 * size PROBE_STATS_COUNT. 0 <= index < PROBE_STATS_COUNT. This is used 122 * to access members of the pii_probes[] array defined in the phyint_instance 123 * structure. 124 */ 125 #define PROBE_INDEX_PREV(index) \ 126 MOD_DCR(index, PROBE_STATS_COUNT) 127 128 #define PROBE_INDEX_NEXT(index) \ 129 MOD_INCR(index, PROBE_STATS_COUNT) 130 131 132 /* 133 * If we receive more than LINK_UP_PERMIN "link up" notifications in a minute, 134 * then don't actually perform the repair operation until we've dropped back 135 * below the threshold (or we have a probe address and our probes indicate 136 * that the link is functioning again). This is to prevent link flapping in 137 * the case where we don't have a probe address. 138 */ 139 #define LINK_UP_PERMIN 2 140 141 #define LINK_DOWN(pi) ((pi)->pi_link_state == 0) 142 #define LINK_UP(pi) (!LINK_DOWN(pi)) 143 #define FLAGS_TO_LINK_STATE(pi) (((pi)->pi_flags & IFF_RUNNING) != 0) 144 #define UPDATE_LINK_STATE(pi) ((pi)->pi_link_state = \ 145 FLAGS_TO_LINK_STATE(pi) ? 1 : 0) 146 #define INIT_LINK_STATE(pi) ((pi)->pi_link_state = 1) 147 148 /* 149 * Phyint group states; see below for the phyint group definition. 150 */ 151 enum pg_state { 152 PG_RUNNING = 1, /* at least one interface in group is working */ 153 PG_FAILED = 2 /* group has failed completely */ 154 }; 155 156 /* 157 * Convenience macro to check if the whole group has failed. 158 */ 159 #define GROUP_FAILED(pg) ((pg)->pg_groupfailed) 160 161 /* 162 * A doubly linked list of all phyint groups in the system. 163 * A phyint group is identified by its group name. 164 */ 165 struct phyint_group { 166 char pg_name[LIFNAMSIZ + 1]; /* Phyint group name */ 167 struct phyint *pg_phyint; /* List of phyints in this group */ 168 struct phyint_group *pg_next; /* Next phyint group */ 169 struct phyint_group *pg_prev; /* Prev phyint group */ 170 uint64_t pg_sig; /* Current signature of this group */ 171 int pg_probeint; /* Interval between probes */ 172 int pg_fdt; /* Time needed to detect failure */ 173 uint_t 174 pg_groupfailed : 1; /* The whole group has failed */ 175 }; 176 177 /* 178 * Phyint states; see below for the phyint definition. 179 */ 180 enum pi_state { 181 PI_NOTARGETS = 1, /* Phyint has no targets */ 182 PI_RUNNING = 2, /* Phyint is functioning */ 183 PI_FAILED = 3, /* Phyint is failed */ 184 PI_OFFLINE = 4 /* Phyint is offline */ 185 }; 186 187 /* 188 * Representation of a NIC or a phyint. There is a list of all known phyints. 189 * There is also a list of phyints belonging to a phyint group, one list 190 * per phyint group. 191 */ 192 struct phyint { 193 char pi_name[LIFNAMSIZ + 1]; /* Phyint name eg. le0 */ 194 struct phyint_instance *pi_v4; /* The IPv4 instance */ 195 struct phyint_instance *pi_v6; /* The IPv6 instance */ 196 struct phyint_group *pi_group; /* Pointer to the group */ 197 struct phyint *pi_next; /* List of all phyints */ 198 struct phyint *pi_prev; /* List of all phyints */ 199 struct phyint *pi_pgnext; /* List of phyints in this group */ 200 struct phyint *pi_pgprev; /* List of phyints in this group */ 201 uint_t pi_ifindex; /* interface index */ 202 enum pi_state pi_state; /* State of the phyint */ 203 uint64_t pi_flags; /* Phyint flags from kernel */ 204 uint16_t pi_icmpid; /* icmp id in icmp echo request */ 205 /* 206 * The pi_whenup array is a circular buffer of the most recent 207 * times (in milliseconds since some arbitrary point of time in 208 * the past) that the interface was brought up; pi_whendx identifies 209 * the oldest element of the array. 210 */ 211 uint_t pi_whenup[LINK_UP_PERMIN]; 212 unsigned int pi_whendx; 213 214 uint_t 215 pi_empty : 1, /* failover done, empty */ 216 pi_full : 1, /* failback done, full */ 217 /* More details in probe.c */ 218 pi_taddrmsg_printed : 1, /* testaddr msg printed */ 219 pi_cfgmsg_printed : 1, /* bad config msg printed */ 220 pi_lfmsg_printed : 1, /* link-flapping msg printed */ 221 pi_link_state : 1; /* interface link state */ 222 }; 223 224 /* 225 * A doubly linked list of all phyint_instances each of which contains a 226 * doubly linked list of logical interfaces and targets. For eg. if both 227 * IPv4 and IPv6 are used over hme0, we have 2 phyint instances, 1 for each 228 * protocol. 229 */ 230 struct phyint_instance { 231 struct phyint_instance *pii_next; /* List of all phyint insts */ 232 struct phyint_instance *pii_prev; /* List of all phyint insts */ 233 234 struct phyint *pii_phyint; /* Back pointer to the phyint */ 235 struct target *pii_targets; /* List of targets on this link */ 236 struct logint *pii_probe_logint; /* IFF_NOFAILOVER addr for probing */ 237 struct logint *pii_logint; /* Doubly linked list of logical ifs */ 238 239 int pii_probe_sock; /* Socket for ICMP Probe packets */ 240 int pii_af; /* Address family */ 241 uint16_t pii_rack; /* highest acknowledged seq number */ 242 uint16_t pii_snxt; /* sequence number of next probe */ 243 uint_t pii_snxt_time; /* actual next probe time that */ 244 /* includes some randomness */ 245 246 uint_t pii_snxt_basetime; /* strictly periodic base probe time */ 247 /* for all periodic probes */ 248 uint_t pii_fd_snxt_basetime; /* strictly periodic base probe time */ 249 /* for failure detection probes */ 250 251 hrtime_t pii_fd_hrtime; /* hrtime_t before which we should */ 252 /* not send probes out this pii */ 253 254 uint64_t pii_flags; /* Phyint flags from kernel */ 255 256 struct probe_stats { 257 struct target *pr_target; /* Probe Target */ 258 uint_t pr_time_sent; /* Time probe was sent */ 259 uint_t pr_status; /* probe status as below */ 260 #define PR_UNUSED 0 /* Probe slot unused */ 261 #define PR_UNACKED 1 /* Probe is unacknowledged */ 262 #define PR_ACKED 2 /* Probe has been acknowledged */ 263 #define PR_LOST 3 /* Probe is declared lost */ 264 union { 265 uint_t tl; /* time probe is declared lost */ 266 uint_t ta; /* time probe is acked */ 267 } prt; 268 #define pr_time_lost prt.tl 269 #define pr_time_acked prt.ta 270 } pii_probes[PROBE_STATS_COUNT]; 271 272 uint_t 273 pii_in_use : 1, /* To detect removed phyints */ 274 pii_basetime_inited : 1, /* probe time initialized */ 275 pii_targets_are_routers : 1; /* routers or hosts ? */ 276 277 uint_t pii_probe_next; /* next index to use in pii_probes[] */ 278 struct target *pii_target_next; /* next target for probing */ 279 struct target *pii_rtt_target_next; 280 /* next target for rtt probes */ 281 282 int pii_ntargets; /* Number of active targets */ 283 struct stats { /* Cumulative statistics */ 284 uint64_t lost; /* Number of probes lost */ 285 uint64_t acked; /* Number of probes acked */ 286 uint64_t sent; /* Number of probes sent */ 287 uint64_t unknown; /* Number of ambiguous */ 288 /* probe acks */ 289 } pii_cum_stats; 290 }; 291 292 #define pii_name pii_phyint->pi_name 293 #define pii_ifindex pii_phyint->pi_ifindex 294 #define pii_state pii_phyint->pi_state 295 #define pii_icmpid pii_phyint->pi_icmpid 296 297 #define PR_STATUS_VALID(status) ((status) <= PR_LOST) 298 299 300 /* 301 * A doubly linked list of prefixes or logicals, hanging off the 302 * phyint instance. 303 */ 304 struct logint { 305 struct logint *li_next; /* Next logint of this phyint inst. */ 306 struct logint *li_prev; /* Prev logint of this phyint inst. */ 307 struct phyint_instance *li_phyint_inst; 308 /* Back pointer to phyint inst. */ 309 310 char li_name[LIFNAMSIZ + 1]; /* name Eg. hme0:1 */ 311 struct in6_addr li_addr; /* IP address */ 312 struct in6_addr li_dstaddr; /* Dst IP address for pointopoint */ 313 struct in6_addr li_subnet; /* prefix / subnet */ 314 uint_t li_subnet_len; /* prefix / subnet length */ 315 uint64_t li_flags; /* IFF_* flags */ 316 uint_t li_oifindex; /* original ifindex (SIOCGLIFOINDEX) */ 317 uint_t 318 li_in_use : 1, /* flag to detect deleted logints */ 319 li_dupaddr : 1, /* This test address is not unique */ 320 li_dupaddrmsg_printed : 1; 321 /* Error has been logged to console */ 322 }; 323 324 325 /* 326 * Doubly-linked list of probe targets on a phyint instance. Probe targets are 327 * usually onlink routers. If no onlink routers can be found, onlink hosts 328 * are used. 329 */ 330 struct target { 331 struct target *tg_next; /* Next target for this phyint inst. */ 332 struct target *tg_prev; /* Prev target for this phyint inst. */ 333 struct phyint_instance *tg_phyint_inst; 334 /* Back pointer to phyint instance */ 335 336 struct in6_addr tg_address; /* Target IP address */ 337 int tg_status; /* Status of the target below */ 338 #define TG_ACTIVE 1 /* active probe target */ 339 #define TG_UNUSED 2 /* target not in use now */ 340 #define TG_SLOW 3 /* rtt is high - Not in use now */ 341 #define TG_DEAD 4 /* Target is not responding */ 342 343 hrtime_t tg_latime; /* Target's last active time */ 344 int tg_rtt_sa; /* Scaled round trip time(RTT) avg. */ 345 int tg_rtt_sd; /* Scaled RTT deviation */ 346 int tg_crtt; /* Conservative RTT = A + 4D */ 347 uint32_t 348 tg_in_use : 1; /* In use flag */ 349 int tg_deferred[MAXDEFERREDRTT + 1]; 350 /* Deferred rtt data points */ 351 int tg_num_deferred; 352 /* Number of deferred rtt data points */ 353 }; 354 355 #define TG_STATUS_VALID(status) \ 356 (((status) >= TG_ACTIVE) && ((status) <= TG_DEAD)) 357 358 /* 359 * Statistics about consecutive probe failures are passed around between 360 * functions in this structure. 361 */ 362 struct probe_fail_count 363 { 364 uint_t pf_tff; /* Earliest time of failure in a series */ 365 int pf_nfail; /* Number of consecutive probe failures */ 366 int pf_nfail_tg; /* Number of consecutive probe fails for */ 367 /* some given target 'tg' */ 368 }; 369 370 /* 371 * Statistics about consecutive probe successes is passed around between 372 * functions in this structure. 373 */ 374 struct probe_success_count 375 { 376 uint_t ps_tls; /* Most recent time of probe success */ 377 boolean_t ps_tls_valid; /* is ps_tls valid */ 378 int ps_nsucc; /* Number of consecutive probe successes */ 379 /* starting from the most recent */ 380 int ps_nsucc_tg; /* Number of consecutive probe successes */ 381 /* for some given target 'tg' */ 382 }; 383 384 /* 385 * Statistics about missed probes that were never sent. 386 * Happens due to scheduling delay. 387 */ 388 389 struct probes_missed 390 { 391 uint_t pm_nprobes; /* Cumulative number of missed probes */ 392 uint_t pm_ntimes; /* Total number of occassions */ 393 }; 394 395 /* 396 * Globals 397 */ 398 extern struct phyint *phyints; /* List of all phyints */ 399 extern struct phyint_group *phyint_groups; /* List of all phyint groups */ 400 extern struct phyint_group *phyint_anongroup; /* Pointer to the anon group */ 401 extern struct phyint_instance *phyint_instances; 402 /* List of all phyint instances */ 403 extern struct probes_missed probes_missed; 404 /* statistics about missed probes */ 405 406 /* 407 * Function prototypes 408 */ 409 extern int phyint_init(void); 410 extern struct phyint *phyint_lookup(const char *name); 411 extern struct phyint_instance *phyint_inst_lookup(int af, char *name); 412 extern struct phyint_instance *phyint_inst_init_from_k(int af, char *name); 413 extern struct phyint_instance *phyint_inst_other(struct phyint_instance *pii); 414 extern int phyint_inst_update_from_k(struct phyint_instance *pii); 415 extern void phyint_inst_delete(struct phyint_instance *pii); 416 extern uint_t phyint_inst_timer(struct phyint_instance *pii); 417 extern boolean_t phyint_inst_sockinit(struct phyint_instance *pii); 418 419 extern void phyint_newtype(struct phyint *pi); 420 extern void phyint_chstate(struct phyint *pi, enum pi_state state); 421 extern void phyint_group_chstate(struct phyint_group *pg, enum pg_state state); 422 extern void phyint_check_for_repair(struct phyint *pi); 423 424 extern void logint_init_from_k(struct phyint_instance *pii, char *li_name); 425 extern void logint_delete(struct logint *li); 426 427 extern struct target *target_lookup(struct phyint_instance *pii, 428 struct in6_addr addr); 429 extern void target_create(struct phyint_instance *pii, 430 struct in6_addr addr, boolean_t is_router); 431 extern void target_delete(struct target *tg); 432 extern struct target *target_next(struct target *tg); 433 extern void target_add(struct phyint_instance *pii, struct in6_addr addr, 434 boolean_t is_router); 435 436 extern void in_data(struct phyint_instance *pii); 437 extern void in6_data(struct phyint_instance *pii); 438 439 extern int try_failover(struct phyint *pi, int failover_type); 440 extern int try_failback(struct phyint *pi, boolean_t check_only); 441 extern int do_failback(struct phyint *pi, boolean_t check_only); 442 extern boolean_t change_lif_flags(struct phyint *pi, uint64_t flags, 443 boolean_t setfl); 444 445 extern void logperror_pii(struct phyint_instance *pii, char *str); 446 extern void logperror_li(struct logint *li, char *str); 447 extern char *pr_addr(int af, struct in6_addr addr, char *abuf, int len); 448 extern void phyint_inst_print_all(void); 449 450 extern int logint_upcount(struct phyint *pi); 451 extern void restore_phyint(struct phyint *pi); 452 extern void reset_crtt_all(struct phyint *pi); 453 extern int failure_state(struct phyint_instance *pii); 454 extern void process_link_state_changes(void); 455 extern void clear_pii_probe_stats(struct phyint_instance *pii); 456 extern void start_timer(struct phyint_instance *pii); 457 458 extern boolean_t own_address(int af, struct in6_addr addr); 459 460 extern void close_probe_socket(struct phyint_instance *pii, boolean_t flag); 461 462 extern unsigned int getifinfo(const char *, ipmp_ifinfo_t **); 463 extern unsigned int getgroupinfo(const char *, ipmp_groupinfo_t **); 464 extern unsigned int getgrouplist(ipmp_grouplist_t **); 465 extern unsigned int getsnap(ipmp_snap_t **); 466 467 #ifdef __cplusplus 468 } 469 #endif 470 471 #endif /* _MPD_TABLES_H */ 472