1 /* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License, Version 1.0 only 6 * (the "License"). You may not use this file except in compliance 7 * with the License. 8 * 9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 10 * or http://www.opensolaris.org/os/licensing. 11 * See the License for the specific language governing permissions 12 * and limitations under the License. 13 * 14 * When distributing Covered Code, include this CDDL HEADER in each 15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 16 * If applicable, add the following below this CDDL HEADER, with the 17 * fields enclosed by brackets "[]" replaced with your own identifying 18 * information: Portions Copyright [yyyy] [name of copyright owner] 19 * 20 * CDDL HEADER END 21 */ 22 /* 23 * Copyright 2004 Sun Microsystems, Inc. All rights reserved. 24 * Use is subject to license terms. 25 */ 26 27 #ifndef _MPD_TABLES_H 28 #define _MPD_TABLES_H 29 30 #pragma ident "%Z%%M% %I% %E% SMI" 31 32 #ifdef __cplusplus 33 extern "C" { 34 #endif 35 36 /* 37 * Terminology: 38 * 39 * phyint: A NIC eg. hme0. This is represented as 'struct phyint' 40 * 41 * phyint instance: A protocol instance of a phyint. Eg. the IPv4 instance of 42 * hme0 or the IPv6 instance of hme0. (struct phyint_instance) 43 * 44 * logint: A logical interface eg. hme0:1 (struct logint) 45 * 46 * phyint_group: A group of phyints i.e. physical interfaces that are 47 * (i) connected to the same level 2 topology e.g. the same ethernet 48 * switch AND 49 * (ii) share the same phyint group name. 50 * Load spreading and failover occur across members of the same phyint group. 51 * phyint group members must be homogenous. i.e. if a phyint belonging to a 52 * phyint group has a IPv6 protocol instance, then all members of the phyint 53 * group, must have IPv6 protocol instances. (struct phyint_group) 54 */ 55 56 /* 57 * Parameter passed to try_failover(), indicating the type of failover 58 * that is requested. 59 */ 60 #define FAILOVER_NORMAL 1 /* Failover to another phyint */ 61 /* that is preferably a standby */ 62 #define FAILOVER_TO_NONSTANDBY 2 /* Failover to non-standby phyint */ 63 #define FAILOVER_TO_ANY 3 /* Failover to any available phyint */ 64 65 #define MAXDEFERREDRTT 1 /* Maximum number of deferred rtts */ 66 67 /* 68 * Status of the phyint, expressed by the return code of failure_state() 69 */ 70 #define PHYINT_OK 0 /* No failure detected */ 71 #define PHYINT_FAILURE 1 /* NIC failure detected */ 72 #define GROUP_FAILURE 2 /* All NICs have failed */ 73 74 /* 75 * Return values of phyint_inst_update_from_k() 76 */ 77 #define PI_OK 1 /* Phyint matches in the kernel */ 78 #define PI_DELETED 2 /* Phyint has vanished in the kernel */ 79 #define PI_IFINDEX_CHANGED 3 /* Phyint's ifindex has changed */ 80 #define PI_IOCTL_ERROR 4 /* Some ioctl error */ 81 #define PI_GROUP_CHANGED 5 /* The phyint has changed group. */ 82 83 /* 84 * Though IFF_POINTOPOINT is a logint property, for the purpose of 85 * failover, we treat it as a phyint property. Note that we cannot failover 86 * individual logints. 87 */ 88 #define PHYINT_FLAGS(flags) \ 89 (((flags) & (IFF_STANDBY | IFF_INACTIVE | IFF_FAILED | IFF_OFFLINE | \ 90 IFF_POINTOPOINT | IFF_RUNNING)) | (handle_link_notifications ? \ 91 0 : IFF_RUNNING)) 92 93 /* A Phyint can have up to 2 instances, the IPv4 and the IPv6 instance */ 94 #define PHYINT_INSTANCE(pi, af) \ 95 ((af) == AF_INET ? (pi)->pi_v4 : (pi)->pi_v6) 96 97 /* 98 * A phyint instance is probe *enabled* if it has been configured with a probe 99 * address (i.e., an IFF_NOFAILOVER address). It is probe *capable* if it is 100 * also able to send probes (i.e., has one or more targets available). 101 */ 102 #define PROBE_ENABLED(pii) \ 103 (((pii) != NULL) && ((pii)->pii_probe_logint != NULL)) 104 105 #define PROBE_CAPABLE(pii) \ 106 (PROBE_ENABLED(pii) && ((pii)->pii_ntargets != 0)) 107 108 /* Subtract b from a modulo n. i.e. (a - b) mod n */ 109 #define MOD_SUB(a, b, n) \ 110 ((((a) + (n)) - (b)) % (n)) 111 112 /* Increment modulo n */ 113 #define MOD_INCR(a, n) \ 114 (((a) + 1) % (n)) 115 116 /* Decrement modulo n */ 117 #define MOD_DCR(a, n) \ 118 MOD_SUB(a, 1, n) 119 120 /* 121 * 'index' represents an index into the circular probe stats array of 122 * size PROBE_STATS_COUNT. 0 <= index < PROBE_STATS_COUNT. This is used 123 * to access members of the pii_probes[] array defined in the phyint_instance 124 * structure. 125 */ 126 #define PROBE_INDEX_PREV(index) \ 127 MOD_DCR(index, PROBE_STATS_COUNT) 128 129 #define PROBE_INDEX_NEXT(index) \ 130 MOD_INCR(index, PROBE_STATS_COUNT) 131 132 133 /* 134 * If we receive more than LINK_UP_PERMIN "link up" notifications in a minute, 135 * then don't actually perform the repair operation until we've dropped back 136 * below the threshold (or we have a probe address and our probes indicate 137 * that the link is functioning again). This is to prevent link flapping in 138 * the case where we don't have a probe address. 139 */ 140 #define LINK_UP_PERMIN 2 141 142 #define LINK_DOWN(pi) ((pi)->pi_link_state == 0) 143 #define LINK_UP(pi) (!LINK_DOWN(pi)) 144 #define FLAGS_TO_LINK_STATE(pi) (((pi)->pi_flags & IFF_RUNNING) != 0) 145 #define UPDATE_LINK_STATE(pi) ((pi)->pi_link_state = \ 146 FLAGS_TO_LINK_STATE(pi) ? 1 : 0) 147 #define INIT_LINK_STATE(pi) ((pi)->pi_link_state = 1) 148 149 #define SINGLETON_GROUP(pi) ((pi) != NULL && \ 150 (pi)->pi_group->pg_name[0] != '\0' && \ 151 (pi)->pi_group->pg_phyint == (pi) && (pi)->pi_pgnext == NULL) 152 153 /* 154 * Phyint group states; see below for the phyint group definition. 155 */ 156 enum pg_state { 157 PG_RUNNING = 1, /* at least one interface in group is working */ 158 PG_FAILED = 2 /* group has failed completely */ 159 }; 160 161 /* 162 * Convenience macro to check if the whole group has failed. 163 */ 164 #define GROUP_FAILED(pg) ((pg)->pg_groupfailed) 165 166 /* 167 * A doubly linked list of all phyint groups in the system. 168 * A phyint group is identified by its group name. 169 */ 170 struct phyint_group { 171 char pg_name[LIFNAMSIZ + 1]; /* Phyint group name */ 172 struct phyint *pg_phyint; /* List of phyints in this group */ 173 struct phyint_group *pg_next; /* Next phyint group */ 174 struct phyint_group *pg_prev; /* Prev phyint group */ 175 uint64_t pg_sig; /* Current signature of this group */ 176 int pg_probeint; /* Interval between probes */ 177 int pg_fdt; /* Time needed to detect failure */ 178 uint_t 179 pg_groupfailed : 1; /* The whole group has failed */ 180 }; 181 182 /* 183 * Phyint states; see below for the phyint definition. 184 */ 185 enum pi_state { 186 PI_NOTARGETS = 1, /* Phyint has no targets */ 187 PI_RUNNING = 2, /* Phyint is functioning */ 188 PI_FAILED = 3, /* Phyint is failed */ 189 PI_OFFLINE = 4 /* Phyint is offline */ 190 }; 191 192 /* 193 * Representation of a NIC or a phyint. There is a list of all known phyints. 194 * There is also a list of phyints belonging to a phyint group, one list 195 * per phyint group. 196 */ 197 struct phyint { 198 char pi_name[LIFNAMSIZ + 1]; /* Phyint name eg. le0 */ 199 struct phyint_instance *pi_v4; /* The IPv4 instance */ 200 struct phyint_instance *pi_v6; /* The IPv6 instance */ 201 struct phyint_group *pi_group; /* Pointer to the group */ 202 struct phyint *pi_next; /* List of all phyints */ 203 struct phyint *pi_prev; /* List of all phyints */ 204 struct phyint *pi_pgnext; /* List of phyints in this group */ 205 struct phyint *pi_pgprev; /* List of phyints in this group */ 206 uint_t pi_ifindex; /* interface index */ 207 enum pi_state pi_state; /* State of the phyint */ 208 uint64_t pi_flags; /* Phyint flags from kernel */ 209 uint16_t pi_icmpid; /* icmp id in icmp echo request */ 210 /* 211 * The pi_whenup array is a circular buffer of the most recent 212 * times (in milliseconds since some arbitrary point of time in 213 * the past) that the interface was brought up; pi_whendx identifies 214 * the oldest element of the array. 215 */ 216 uint_t pi_whenup[LINK_UP_PERMIN]; 217 unsigned int pi_whendx; 218 219 uint_t 220 pi_empty : 1, /* failover done, empty */ 221 pi_full : 1, /* failback done, full */ 222 /* More details in probe.c */ 223 pi_taddrmsg_printed : 1, /* testaddr msg printed */ 224 pi_cfgmsg_printed : 1, /* bad config msg printed */ 225 pi_lfmsg_printed : 1, /* link-flapping msg printed */ 226 pi_link_state : 1; /* interface link state */ 227 }; 228 229 /* 230 * A doubly linked list of all phyint_instances each of which contains a 231 * doubly linked list of logical interfaces and targets. For eg. if both 232 * IPv4 and IPv6 are used over hme0, we have 2 phyint instances, 1 for each 233 * protocol. 234 */ 235 struct phyint_instance { 236 struct phyint_instance *pii_next; /* List of all phyint insts */ 237 struct phyint_instance *pii_prev; /* List of all phyint insts */ 238 239 struct phyint *pii_phyint; /* Back pointer to the phyint */ 240 struct target *pii_targets; /* List of targets on this link */ 241 struct logint *pii_probe_logint; /* IFF_NOFAILOVER addr for probing */ 242 struct logint *pii_logint; /* Doubly linked list of logical ifs */ 243 244 int pii_probe_sock; /* Socket for ICMP Probe packets */ 245 int pii_af; /* Address family */ 246 uint16_t pii_rack; /* highest acknowledged seq number */ 247 uint16_t pii_snxt; /* sequence number of next probe */ 248 uint_t pii_snxt_time; /* actual next probe time that */ 249 /* includes some randomness */ 250 251 uint_t pii_snxt_basetime; /* strictly periodic base probe time */ 252 /* for all periodic probes */ 253 uint_t pii_fd_snxt_basetime; /* strictly periodic base probe time */ 254 /* for failure detection probes */ 255 256 hrtime_t pii_fd_hrtime; /* hrtime_t before which we should */ 257 /* not send probes out this pii */ 258 259 uint64_t pii_flags; /* Phyint flags from kernel */ 260 261 struct probe_stats { 262 struct target *pr_target; /* Probe Target */ 263 uint_t pr_time_sent; /* Time probe was sent */ 264 uint_t pr_status; /* probe status as below */ 265 #define PR_UNUSED 0 /* Probe slot unused */ 266 #define PR_UNACKED 1 /* Probe is unacknowledged */ 267 #define PR_ACKED 2 /* Probe has been acknowledged */ 268 #define PR_LOST 3 /* Probe is declared lost */ 269 union { 270 uint_t tl; /* time probe is declared lost */ 271 uint_t ta; /* time probe is acked */ 272 } prt; 273 #define pr_time_lost prt.tl 274 #define pr_time_acked prt.ta 275 } pii_probes[PROBE_STATS_COUNT]; 276 277 uint_t 278 pii_in_use : 1, /* To detect removed phyints */ 279 pii_basetime_inited : 1, /* probe time initialized */ 280 pii_targets_are_routers : 1; /* routers or hosts ? */ 281 282 uint_t pii_probe_next; /* next index to use in pii_probes[] */ 283 struct target *pii_target_next; /* next target for probing */ 284 struct target *pii_rtt_target_next; 285 /* next target for rtt probes */ 286 287 int pii_ntargets; /* Number of active targets */ 288 struct stats { /* Cumulative statistics */ 289 uint64_t lost; /* Number of probes lost */ 290 uint64_t acked; /* Number of probes acked */ 291 uint64_t sent; /* Number of probes sent */ 292 uint64_t unknown; /* Number of ambiguous */ 293 /* probe acks */ 294 } pii_cum_stats; 295 }; 296 297 #define pii_name pii_phyint->pi_name 298 #define pii_ifindex pii_phyint->pi_ifindex 299 #define pii_state pii_phyint->pi_state 300 #define pii_icmpid pii_phyint->pi_icmpid 301 302 #define PR_STATUS_VALID(status) ((status) <= PR_LOST) 303 304 305 /* 306 * A doubly linked list of prefixes or logicals, hanging off the 307 * phyint instance. 308 */ 309 struct logint { 310 struct logint *li_next; /* Next logint of this phyint inst. */ 311 struct logint *li_prev; /* Prev logint of this phyint inst. */ 312 struct phyint_instance *li_phyint_inst; 313 /* Back pointer to phyint inst. */ 314 315 char li_name[LIFNAMSIZ + 1]; /* name Eg. hme0:1 */ 316 struct in6_addr li_addr; /* IP address */ 317 struct in6_addr li_dstaddr; /* Dst IP address for pointopoint */ 318 struct in6_addr li_subnet; /* prefix / subnet */ 319 uint_t li_subnet_len; /* prefix / subnet length */ 320 uint64_t li_flags; /* IFF_* flags */ 321 uint_t li_oifindex; /* original ifindex (SIOCGLIFOINDEX) */ 322 uint_t 323 li_in_use : 1, /* flag to detect deleted logints */ 324 li_dupaddr : 1, /* This test address is not unique */ 325 li_dupaddrmsg_printed : 1; 326 /* Error has been logged to console */ 327 }; 328 329 330 /* 331 * Doubly-linked list of probe targets on a phyint instance. Probe targets are 332 * usually onlink routers. If no onlink routers can be found, onlink hosts 333 * are used. 334 */ 335 struct target { 336 struct target *tg_next; /* Next target for this phyint inst. */ 337 struct target *tg_prev; /* Prev target for this phyint inst. */ 338 struct phyint_instance *tg_phyint_inst; 339 /* Back pointer to phyint instance */ 340 341 struct in6_addr tg_address; /* Target IP address */ 342 int tg_status; /* Status of the target below */ 343 #define TG_ACTIVE 1 /* active probe target */ 344 #define TG_UNUSED 2 /* target not in use now */ 345 #define TG_SLOW 3 /* rtt is high - Not in use now */ 346 #define TG_DEAD 4 /* Target is not responding */ 347 348 hrtime_t tg_latime; /* Target's last active time */ 349 int tg_rtt_sa; /* Scaled round trip time(RTT) avg. */ 350 int tg_rtt_sd; /* Scaled RTT deviation */ 351 int tg_crtt; /* Conservative RTT = A + 4D */ 352 uint32_t 353 tg_in_use : 1; /* In use flag */ 354 int tg_deferred[MAXDEFERREDRTT + 1]; 355 /* Deferred rtt data points */ 356 int tg_num_deferred; 357 /* Number of deferred rtt data points */ 358 }; 359 360 #define TG_STATUS_VALID(status) \ 361 (((status) >= TG_ACTIVE) && ((status) <= TG_DEAD)) 362 363 /* 364 * Statistics about consecutive probe failures are passed around between 365 * functions in this structure. 366 */ 367 struct probe_fail_count 368 { 369 uint_t pf_tff; /* Earliest time of failure in a series */ 370 int pf_nfail; /* Number of consecutive probe failures */ 371 int pf_nfail_tg; /* Number of consecutive probe fails for */ 372 /* some given target 'tg' */ 373 }; 374 375 /* 376 * Statistics about consecutive probe successes is passed around between 377 * functions in this structure. 378 */ 379 struct probe_success_count 380 { 381 uint_t ps_tls; /* Most recent time of probe success */ 382 boolean_t ps_tls_valid; /* is ps_tls valid */ 383 int ps_nsucc; /* Number of consecutive probe successes */ 384 /* starting from the most recent */ 385 int ps_nsucc_tg; /* Number of consecutive probe successes */ 386 /* for some given target 'tg' */ 387 }; 388 389 /* 390 * Statistics about missed probes that were never sent. 391 * Happens due to scheduling delay. 392 */ 393 394 struct probes_missed 395 { 396 uint_t pm_nprobes; /* Cumulative number of missed probes */ 397 uint_t pm_ntimes; /* Total number of occassions */ 398 }; 399 400 /* 401 * Globals 402 */ 403 extern struct phyint *phyints; /* List of all phyints */ 404 extern struct phyint_group *phyint_groups; /* List of all phyint groups */ 405 extern struct phyint_group *phyint_anongroup; /* Pointer to the anon group */ 406 extern struct phyint_instance *phyint_instances; 407 /* List of all phyint instances */ 408 extern struct probes_missed probes_missed; 409 /* statistics about missed probes */ 410 411 /* 412 * Function prototypes 413 */ 414 extern int phyint_init(void); 415 extern struct phyint *phyint_lookup(const char *name); 416 extern struct phyint_instance *phyint_inst_lookup(int af, char *name); 417 extern struct phyint_instance *phyint_inst_init_from_k(int af, char *name); 418 extern struct phyint_instance *phyint_inst_other(struct phyint_instance *pii); 419 extern int phyint_inst_update_from_k(struct phyint_instance *pii); 420 extern void phyint_inst_delete(struct phyint_instance *pii); 421 extern uint_t phyint_inst_timer(struct phyint_instance *pii); 422 extern boolean_t phyint_inst_sockinit(struct phyint_instance *pii); 423 424 extern void phyint_newtype(struct phyint *pi); 425 extern void phyint_chstate(struct phyint *pi, enum pi_state state); 426 extern void phyint_group_chstate(struct phyint_group *pg, enum pg_state state); 427 extern void phyint_check_for_repair(struct phyint *pi); 428 429 extern void logint_init_from_k(struct phyint_instance *pii, char *li_name); 430 extern void logint_delete(struct logint *li); 431 432 extern struct target *target_lookup(struct phyint_instance *pii, 433 struct in6_addr addr); 434 extern void target_create(struct phyint_instance *pii, 435 struct in6_addr addr, boolean_t is_router); 436 extern void target_delete(struct target *tg); 437 extern struct target *target_next(struct target *tg); 438 extern void target_add(struct phyint_instance *pii, struct in6_addr addr, 439 boolean_t is_router); 440 441 extern void in_data(struct phyint_instance *pii); 442 extern void in6_data(struct phyint_instance *pii); 443 444 extern int try_failover(struct phyint *pi, int failover_type); 445 extern int try_failback(struct phyint *pi, boolean_t check_only); 446 extern int do_failback(struct phyint *pi, boolean_t check_only); 447 extern boolean_t change_lif_flags(struct phyint *pi, uint64_t flags, 448 boolean_t setfl); 449 450 extern void logperror_pii(struct phyint_instance *pii, char *str); 451 extern void logperror_li(struct logint *li, char *str); 452 extern char *pr_addr(int af, struct in6_addr addr, char *abuf, int len); 453 extern void phyint_inst_print_all(void); 454 455 extern int logint_upcount(struct phyint *pi); 456 extern void restore_phyint(struct phyint *pi); 457 extern void reset_crtt_all(struct phyint *pi); 458 extern int failure_state(struct phyint_instance *pii); 459 extern void process_link_state_changes(void); 460 extern void clear_pii_probe_stats(struct phyint_instance *pii); 461 extern void start_timer(struct phyint_instance *pii); 462 463 extern boolean_t own_address(int af, struct in6_addr addr); 464 465 extern void close_probe_socket(struct phyint_instance *pii, boolean_t flag); 466 467 extern unsigned int getifinfo(const char *, ipmp_ifinfo_t **); 468 extern unsigned int getgroupinfo(const char *, ipmp_groupinfo_t **); 469 extern unsigned int getgrouplist(ipmp_grouplist_t **); 470 extern unsigned int getsnap(ipmp_snap_t **); 471 472 #ifdef __cplusplus 473 } 474 #endif 475 476 #endif /* _MPD_TABLES_H */ 477