xref: /illumos-gate/usr/src/cmd/cmd-inet/usr.lib/in.mpathd/mpd_tables.h (revision d656abb5804319b33c85955a73ee450ef7ff9739)
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	_MPD_TABLES_H
27 #define	_MPD_TABLES_H
28 
29 #ifdef	__cplusplus
30 extern "C" {
31 #endif
32 
33 /*
34  * Terminology:
35  *
36  * phyint: A NIC eg. hme0. This is represented as 'struct phyint'
37  *
38  * phyint instance: A protocol instance of a phyint. Eg. the IPv4 instance of
39  * 	hme0 or the IPv6 instance of hme0. (struct phyint_instance)
40  *
41  * logint: A logical interface eg. hme0:1 (struct logint)
42  *
43  * phyint_group: A group of phyints i.e. physical interfaces that are
44  *	(i) connected to the same level 2 topology e.g. the same ethernet
45  *	    switch AND
46  *	(ii) share the same phyint group name.
47  * Load spreading and failover occur across members of the same phyint group.
48  * phyint group members must be homogeneous. i.e. if a phyint belonging to a
49  * phyint group has a IPv6 protocol instance, then all members of the phyint
50  * group, must have IPv6 protocol instances. (struct phyint_group)
51  */
52 
53 #define	MAXDEFERREDRTT		1	/* Maximum number of deferred rtts */
54 
55 /*
56  * Status of the phyint, expressed by the return code of failure_state()
57  */
58 #define	PHYINT_OK	0		/* No failure detected */
59 #define	PHYINT_FAILURE	1		/* NIC failure detected */
60 #define	GROUP_FAILURE	2		/* All NICs have failed */
61 
62 /*
63  * Return values of phyint_inst_update_from_k()
64  */
65 #define	PI_OK			1	/* Phyint matches in the kernel */
66 #define	PI_DELETED		2	/* Phyint has vanished in the kernel */
67 #define	PI_IFINDEX_CHANGED	3	/* Phyint's ifindex has changed */
68 #define	PI_IOCTL_ERROR		4	/* Some ioctl error */
69 #define	PI_GROUP_CHANGED	5	/* The phyint has changed group. */
70 
71 #define	PHYINT_FLAGS(flags)	\
72 	(((flags) & (IFF_STANDBY | IFF_INACTIVE | IFF_FAILED | IFF_OFFLINE | \
73 	IFF_RUNNING)) | (handle_link_notifications ? 0 : IFF_RUNNING))
74 
75 /* A Phyint can have up to 2 instances, the IPv4 and the IPv6 instance */
76 #define	PHYINT_INSTANCE(pi, af)	\
77 	((af) == AF_INET ? (pi)->pi_v4 : (pi)->pi_v6)
78 
79 /*
80  * A phyint instance is probe *enabled* if it has been configured with a
81  * unique probe address (i.e., an IFF_NOFAILOVER address).  It is probe
82  * *capable* if it is also able to send probes (i.e., has one or more
83  * targets available).
84  */
85 #define	PROBE_ENABLED(pii) \
86 	(((pii) != NULL) && ((pii)->pii_probe_sock != -1) &&	\
87 	((pii)->pii_probe_logint != NULL) &&			\
88 	(((pii)->pii_probe_logint->li_dupaddr == 0)))
89 
90 #define	PROBE_CAPABLE(pii) \
91 	(PROBE_ENABLED(pii) && ((pii)->pii_ntargets != 0))
92 
93 /* Subtract b from a modulo n. i.e. (a - b) mod n  */
94 #define	MOD_SUB(a, b, n)	\
95 	((((a) + (n)) - (b)) % (n))
96 
97 /* Increment modulo n */
98 #define	MOD_INCR(a, n)		\
99 	(((a) + 1) % (n))
100 
101 /* Decrement modulo n */
102 #define	MOD_DCR(a, n)		\
103 	MOD_SUB(a, 1, n)
104 
105 /*
106  * 'index' represents an index into the circular probe stats array of
107  * size PROBE_STATS_COUNT.  0 <= index < PROBE_STATS_COUNT. This is used
108  * to access members of the pii_probes[] array defined in the phyint_instance
109  * structure.
110  */
111 #define	PROBE_INDEX_PREV(index)	\
112 	MOD_DCR(index, PROBE_STATS_COUNT)
113 
114 #define	PROBE_INDEX_NEXT(index)	\
115 	MOD_INCR(index, PROBE_STATS_COUNT)
116 
117 
118 /*
119  * If we receive more than LINK_UP_PERMIN "link up" notifications in a minute,
120  * then don't actually perform the repair operation until we've dropped back
121  * below the threshold (or we have a probe address and our probes indicate
122  * that the link is functioning again).  This is to prevent link flapping in
123  * the case where we don't have a probe address.
124  */
125 #define	LINK_UP_PERMIN	2
126 
127 #define	LINK_DOWN(pi) ((pi)->pi_link_state == 0)
128 #define	LINK_UP(pi) (!LINK_DOWN(pi))
129 #define	FLAGS_TO_LINK_STATE(pi) (((pi)->pi_flags & IFF_RUNNING) != 0)
130 #define	UPDATE_LINK_STATE(pi) ((pi)->pi_link_state = \
131 	FLAGS_TO_LINK_STATE(pi) ? 1 : 0)
132 #define	INIT_LINK_STATE(pi) ((pi)->pi_link_state = 1)
133 
134 /*
135  * Phyint group states; see below for the phyint group definition.
136  */
137 enum pg_state {
138 	PG_OK = 1,	/* all interfaces in the group are working */
139 	PG_DEGRADED,	/* some interfaces in the group are unusable */
140 	PG_FAILED	/* all interfaces in the group are unusable */
141 };
142 
143 /*
144  * Convenience macro to check if the whole group has failed.
145  */
146 #define	GROUP_FAILED(pg)	((pg)->pg_state == PG_FAILED)
147 
148 /*
149  * A doubly linked list of all phyint groups in the system.
150  * A phyint group is identified by its group name.
151  */
152 struct phyint_group {
153 	char pg_name[LIFGRNAMSIZ];	/* Phyint group name */
154 	struct phyint *pg_phyint;	/* List of phyints in this group */
155 	struct phyint_group *pg_next;	/* Next phyint group */
156 	struct phyint_group *pg_prev;	/* Prev phyint group */
157 	uint64_t 	pg_sig;		/* Current signature of this group */
158 	int		pg_probeint;	/* Interval between probes */
159 	int		pg_fdt;		/* Time needed to detect failure */
160 	enum pg_state	pg_state;	/* Current group state */
161 	boolean_t	pg_in_use;	/* To detect removed groups */
162 	struct addrlist	*pg_addrs;	/* Data addresses in this group */
163 	boolean_t pg_failmsg_printed;	/* Group failure msg printed */
164 };
165 
166 /*
167  * Phyint states; see below for the phyint definition.
168  */
169 enum pi_state {
170 	PI_NOTARGETS	= 1,	/* Phyint has no targets */
171 	PI_RUNNING	= 2,	/* Phyint is functioning */
172 	PI_FAILED	= 3,	/* Phyint is failed */
173 	PI_OFFLINE	= 4	/* Phyint is offline */
174 };
175 
176 /*
177  * Representation of a NIC or a phyint. There is a list of all known phyints.
178  * There is also a list of phyints belonging to a phyint group, one list
179  * per phyint group.
180  */
181 struct phyint {
182 	char	pi_name[LIFNAMSIZ + 1]; /* Phyint name eg. le0 */
183 	struct phyint_instance *pi_v4;	/* The IPv4 instance */
184 	struct phyint_instance *pi_v6;	/* The IPv6 instance */
185 	struct phyint_group *pi_group;	/* Pointer to the group */
186 	struct phyint	*pi_next;	/* List of all phyints */
187 	struct phyint	*pi_prev;	/* List of all phyints */
188 	struct phyint	*pi_pgnext;	/* List of phyints in this group */
189 	struct phyint	*pi_pgprev;	/* List of phyints in this group */
190 	uint_t		pi_ifindex;	/* interface index */
191 	enum pi_state	pi_state;	/* State of the phyint */
192 	uint64_t	pi_flags;	/* Phyint flags from kernel */
193 	uint16_t	pi_icmpid;	/* icmp id in icmp echo request */
194 	uint64_t	pi_taddrthresh;	/* time (in secs) to delay logging */
195 					/* about missing test addresses */
196 	dlpi_handle_t	pi_dh;		/* DLPI handle to underlying link */
197 	uint_t		pi_notes; 	/* enabled DLPI notifications */
198 	uchar_t		pi_hwaddr[DLPI_PHYSADDR_MAX]; /* phyint's hw address */
199 	size_t		pi_hwaddrlen;	/* phyint's hw address length */
200 
201 	/*
202 	 * The pi_whenup array is a circular buffer of the most recent
203 	 * times (in milliseconds since some arbitrary point of time in
204 	 * the past) that the interface was brought up; pi_whendx identifies
205 	 * the oldest element of the array.
206 	 */
207 	uint_t		pi_whenup[LINK_UP_PERMIN];
208 	unsigned int	pi_whendx;
209 
210 	uint_t
211 		pi_taddrmsg_printed : 1,	/* testaddr msg printed */
212 		pi_duptaddrmsg_printed : 1,	/* dup testaddr msg printed */
213 		pi_cfgmsg_printed : 1,	/* bad config msg printed */
214 		pi_lfmsg_printed : 1,   /* link-flapping msg printed */
215 		pi_link_state : 1,	/* interface link state */
216 		pi_hwaddrdup : 1; 	/* disabled due to dup hw address */
217 };
218 
219 /*
220  * A doubly linked list of all phyint_instances each of which contains a
221  * doubly linked list of logical interfaces and targets. For eg. if both
222  * IPv4 and IPv6 are used over hme0, we have 2 phyint instances, 1 for each
223  * protocol.
224  */
225 struct phyint_instance {
226 	struct phyint_instance	*pii_next;	/* List of all phyint insts */
227 	struct phyint_instance	*pii_prev;	/* List of all phyint insts */
228 
229 	struct phyint	*pii_phyint;	/* Back pointer to the phyint */
230 	struct target	*pii_targets;	/* List of targets on this link */
231 	struct logint	*pii_probe_logint; /* IFF_NOFAILOVER addr for probing */
232 	struct logint	*pii_logint;	/* Doubly linked list of logical ifs */
233 
234 	int	pii_probe_sock;		/* Socket for ICMP Probe packets */
235 	int	pii_af;			/* Address family */
236 	uint16_t pii_rack;		/* highest acknowledged seq number */
237 	uint16_t pii_snxt;		/* sequence number of next probe */
238 	uint_t	pii_snxt_time;		/* actual next probe time that */
239 					/* includes some randomness */
240 
241 	uint_t	pii_snxt_basetime; 	/* strictly periodic base probe time */
242 					/* for all periodic probes */
243 	uint_t	pii_fd_snxt_basetime; 	/* strictly periodic base probe time */
244 					/* for failure detection probes */
245 
246 	hrtime_t 	pii_fd_hrtime;	/* hrtime_t before which we should */
247 					/* not send probes out this pii */
248 
249 	uint64_t	pii_flags;	/* Phyint flags from kernel */
250 
251 	struct probe_stats {
252 		uint_t		pr_id;		/* Full ID of probe */
253 		struct target	*pr_target;	/* Probe Target */
254 		uint_t		pr_time_lost; 	/* Time probe declared lost */
255 		struct timeval	pr_tv_sent;	/* Wall time probe was sent */
256 		hrtime_t pr_hrtime_start;	/* hrtime probe op started */
257 		hrtime_t pr_hrtime_sent;	/* hrtime probe was sent */
258 		hrtime_t pr_hrtime_ackrecv; 	/* hrtime probe ack received */
259 		hrtime_t pr_hrtime_ackproc;	/* hrtime probe ack processed */
260 		uint_t	pr_status;	/* probe status as below */
261 #define	PR_UNUSED	0		/* Probe slot unused */
262 #define	PR_UNACKED	1		/* Probe is unacknowledged */
263 #define	PR_ACKED	2		/* Probe has been acknowledged */
264 #define	PR_LOST		3		/* Probe is declared lost */
265 	} pii_probes[PROBE_STATS_COUNT];
266 
267 	uint_t
268 		pii_in_use : 1,			/* To detect removed phyints */
269 		pii_basetime_inited : 1,	/* probe time initialized */
270 		pii_targets_are_routers : 1;	/* routers or hosts ? */
271 
272 	uint_t	pii_probe_next;		/* next index to use in pii_probes[] */
273 	struct target *pii_target_next;	/* next target for probing */
274 	struct target *pii_rtt_target_next;
275 					/* next target for rtt probes */
276 
277 	int	pii_ntargets;		/* Number of active targets */
278 	struct stats {			/* Cumulative statistics */
279 		uint64_t	lost;		/* Number of probes lost */
280 		uint64_t	acked;		/* Number of probes acked */
281 		uint64_t	sent;		/* Number of probes sent */
282 		uint64_t	unknown;	/* Number of ambiguous */
283 						/* probe acks */
284 	} pii_cum_stats;
285 };
286 
287 #define	pii_name	pii_phyint->pi_name
288 #define	pii_ifindex	pii_phyint->pi_ifindex
289 #define	pii_state	pii_phyint->pi_state
290 #define	pii_icmpid	pii_phyint->pi_icmpid
291 
292 #define	PR_STATUS_VALID(status)		((status) <= PR_LOST)
293 
294 
295 /*
296  * A doubly linked list of prefixes or logicals, hanging off the
297  * phyint instance.
298  */
299 struct logint {
300 	struct logint	*li_next;	/* Next logint of this phyint inst. */
301 	struct logint	*li_prev;	/* Prev logint of this phyint inst. */
302 	struct phyint_instance	*li_phyint_inst;
303 					/* Back pointer to phyint inst. */
304 
305 	char		li_name[LIFNAMSIZ + 1];	/* name Eg. hme0:1 */
306 	struct in6_addr	li_addr;	/* IP address */
307 	struct in6_addr	li_dstaddr;	/* Dst IP address for pointopoint */
308 	struct in6_addr	li_subnet;	/* prefix / subnet */
309 	uint_t		li_subnet_len;	/* prefix / subnet length */
310 	uint64_t	li_flags;	/* IFF_* flags */
311 	uint_t
312 			li_in_use : 1,	/* flag to detect deleted logints */
313 			li_dupaddr : 1;	/* test address is not unique */
314 };
315 
316 
317 /*
318  * Doubly-linked list of probe targets on a phyint instance. Probe targets are
319  * usually onlink routers. If no onlink routers can be found, onlink hosts
320  * are used.
321  */
322 struct target {
323 	struct target	*tg_next;	/* Next target for this phyint inst. */
324 	struct target	*tg_prev;	/* Prev target for this phyint inst. */
325 	struct phyint_instance	*tg_phyint_inst;
326 					/* Back pointer to phyint instance */
327 
328 	struct in6_addr	tg_address;	/* Target IP address */
329 	int		tg_status;	/* Status of the target below */
330 #define	TG_ACTIVE	1		/* active probe target */
331 #define	TG_UNUSED	2		/* target not in use now */
332 #define	TG_SLOW		3		/* rtt is high - Not in use now */
333 #define	TG_DEAD		4		/* Target is not responding */
334 
335 	hrtime_t	tg_latime;	/* Target's last active time */
336 	int64_t		tg_rtt_sa;	/* Scaled RTT average (in ns) */
337 	int64_t		tg_rtt_sd;	/* Scaled RTT deviation (in ns) */
338 	int		tg_crtt;	/* Conservative RTT = A + 4D (in ms) */
339 	uint32_t
340 			tg_in_use : 1;	/* In use flag */
341 	int64_t		tg_deferred[MAXDEFERREDRTT + 1];
342 					/* Deferred rtt data points */
343 	int		tg_num_deferred;
344 					/* Number of deferred rtt data points */
345 };
346 
347 #define	TG_STATUS_VALID(status) \
348 	(((status) >= TG_ACTIVE) && ((status) <= TG_DEAD))
349 
350 /*
351  * Statistics about consecutive probe failures are passed around between
352  * functions in this structure.
353  */
354 struct probe_fail_count
355 {
356 	uint_t	pf_tff;		/* Earliest time of failure in a series */
357 	int	pf_nfail;	/* Number of consecutive probe failures */
358 	int	pf_nfail_tg;	/* Number of consecutive probe fails for */
359 				/* some given target 'tg' */
360 };
361 
362 /*
363  * Statistics about consecutive probe successes is passed around between
364  * functions in this structure.
365  */
366 struct probe_success_count
367 {
368 	uint_t ps_tls;		/* Most recent time of probe success */
369 	boolean_t ps_tls_valid;	/* is ps_tls valid */
370 	int	ps_nsucc;	/* Number of consecutive probe successes */
371 				/* starting from the most recent */
372 	int	ps_nsucc_tg;	/* Number of consecutive probe successes */
373 				/* for some given target 'tg' */
374 };
375 
376 /*
377  * Statistics about missed probes that were never sent.
378  * Happens due to scheduling delay.
379  */
380 
381 struct probes_missed
382 {
383 	uint_t	pm_nprobes;	/* Cumulative number of missed probes */
384 	uint_t	pm_ntimes;	/* Total number of occasions */
385 };
386 
387 typedef struct addrlist {
388 	struct addrlist		*al_next; 		/* next address */
389 	char			al_name[LIFNAMSIZ];	/* address lif name */
390 	uint64_t		al_flags;		/* address flags */
391 	struct sockaddr_storage	al_addr; 		/* address */
392 } addrlist_t;
393 
394 /*
395  * Globals
396  */
397 extern addrlist_t *localaddrs;
398 			/* List of all local addresses, including local zones */
399 extern struct phyint *phyints;		/* List of all phyints */
400 extern struct phyint_group *phyint_groups; /* List of all phyint groups */
401 extern struct phyint_group *phyint_anongroup; /* Pointer to the anon group */
402 extern struct phyint_instance *phyint_instances;
403 					/* List of all phyint instances */
404 extern struct probes_missed probes_missed;
405 					/* statistics about missed probes */
406 
407 /*
408  * Function prototypes
409  */
410 extern int phyint_init(void);
411 extern struct phyint *phyint_lookup(const char *name);
412 extern struct phyint_instance *phyint_inst_lookup(int af, char *name);
413 extern struct phyint_instance *phyint_inst_init_from_k(int af, char *name);
414 extern struct phyint_instance *phyint_inst_other(struct phyint_instance *pii);
415 extern int phyint_inst_update_from_k(struct phyint_instance *pii);
416 extern void phyint_inst_delete(struct phyint_instance *pii);
417 extern uint_t phyint_inst_timer(struct phyint_instance *pii);
418 extern boolean_t phyint_inst_sockinit(struct phyint_instance *pii);
419 
420 extern void phyint_changed(struct phyint *pi);
421 extern void phyint_chstate(struct phyint *pi, enum pi_state state);
422 extern void phyint_group_chstate(struct phyint_group *pg, enum pg_state state);
423 extern struct phyint_group *phyint_group_create(const char *pg_name);
424 extern struct phyint_group *phyint_group_lookup(const char *pg_name);
425 extern void phyint_group_insert(struct phyint_group *pg);
426 extern void phyint_group_delete(struct phyint_group *pg);
427 extern void phyint_group_refresh_state(struct phyint_group *pg);
428 extern void phyint_check_for_repair(struct phyint *pi);
429 extern void phyint_transition_to_running(struct phyint *pi);
430 extern void phyint_activate_another(struct phyint *pi);
431 extern int phyint_offline(struct phyint *pi, unsigned int);
432 extern int phyint_undo_offline(struct phyint *pi);
433 
434 extern void logint_init_from_k(struct phyint_instance *pii, char *li_name);
435 extern void logint_delete(struct logint *li);
436 
437 extern struct target *target_lookup(struct phyint_instance *pii,
438     struct in6_addr addr);
439 extern void target_create(struct phyint_instance *pii,
440     struct in6_addr addr, boolean_t is_router);
441 extern void target_delete(struct target *tg);
442 extern struct target *target_next(struct target *tg);
443 extern void target_add(struct phyint_instance *pii, struct in6_addr addr,
444     boolean_t is_router);
445 
446 extern void in_data(struct phyint_instance *pii);
447 extern void in6_data(struct phyint_instance *pii);
448 
449 extern void logperror_pii(struct phyint_instance *pii, const char *str);
450 extern void logperror_li(struct logint *li, const char *str);
451 extern char *pr_addr(int af, struct in6_addr addr, char *abuf, int len);
452 extern void addr2storage(int af, const struct in6_addr *addr,
453     struct sockaddr_storage *ssp);
454 extern void phyint_inst_print_all(void);
455 extern boolean_t prefix_equal(struct in6_addr, struct in6_addr, uint_t);
456 
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 extern void stop_probing(struct phyint *pi);
463 
464 extern boolean_t own_address(struct in6_addr addr);
465 extern boolean_t change_pif_flags(struct phyint *pi, uint64_t set,
466     uint64_t clear);
467 
468 extern void close_probe_socket(struct phyint_instance *pii, boolean_t flag);
469 extern int probe_state_event(struct probe_stats *, struct phyint_instance *);
470 extern void probe_chstate(struct probe_stats *, struct phyint_instance *, int);
471 
472 extern unsigned int getgraddrinfo(const char *, struct sockaddr_storage *,
473     ipmp_addrinfo_t **);
474 extern unsigned int getifinfo(const char *, ipmp_ifinfo_t **);
475 extern unsigned int getgroupinfo(const char *, ipmp_groupinfo_t **);
476 extern unsigned int getgrouplist(ipmp_grouplist_t **);
477 extern unsigned int getsnap(ipmp_snap_t **);
478 
479 extern boolean_t addrlist_add(addrlist_t **, const char *, uint64_t,
480     struct sockaddr_storage *);
481 extern void addrlist_free(addrlist_t **);
482 
483 #ifdef	__cplusplus
484 }
485 #endif
486 
487 #endif	/* _MPD_TABLES_H */
488