xref: /illumos-gate/usr/src/cmd/fm/fmd/common/fmd_asru.c (revision 705e9f42)
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 /*
23  * Copyright 2010 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  */
26 
27 #include <sys/fm/protocol.h>
28 #include <uuid/uuid.h>
29 
30 #include <dirent.h>
31 #include <limits.h>
32 #include <unistd.h>
33 #include <alloca.h>
34 #include <stddef.h>
35 #include <fm/libtopo.h>
36 
37 #include <fmd_alloc.h>
38 #include <fmd_string.h>
39 #include <fmd_error.h>
40 #include <fmd_subr.h>
41 #include <fmd_protocol.h>
42 #include <fmd_event.h>
43 #include <fmd_conf.h>
44 #include <fmd_fmri.h>
45 #include <fmd_dispq.h>
46 #include <fmd_case.h>
47 #include <fmd_module.h>
48 #include <fmd_asru.h>
49 
50 #include <fmd.h>
51 
52 static const char *const _fmd_asru_events[] = {
53 	FMD_RSRC_CLASS "asru.ok",		/* UNUSABLE=0 FAULTED=0 */
54 	FMD_RSRC_CLASS "asru.degraded",		/* UNUSABLE=0 FAULTED=1 */
55 	FMD_RSRC_CLASS "asru.unknown",		/* UNUSABLE=1 FAULTED=0 */
56 	FMD_RSRC_CLASS "asru.faulted"		/* UNUSABLE=1 FAULTED=1 */
57 };
58 
59 static const char *const _fmd_asru_snames[] = {
60 	"uf", "uF", "Uf", "UF"			/* same order as above */
61 };
62 
63 volatile uint32_t fmd_asru_fake_not_present = 0;
64 
65 static uint_t
fmd_asru_strhash(fmd_asru_hash_t * ahp,const char * val)66 fmd_asru_strhash(fmd_asru_hash_t *ahp, const char *val)
67 {
68 	return (topo_fmri_strhash(ahp->ah_topo->ft_hdl, val) % ahp->ah_hashlen);
69 }
70 
71 static boolean_t
fmd_asru_strcmp(fmd_asru_hash_t * ahp,const char * a,const char * b)72 fmd_asru_strcmp(fmd_asru_hash_t *ahp, const char *a, const char *b)
73 {
74 	return (topo_fmri_strcmp(ahp->ah_topo->ft_hdl, a, b));
75 }
76 
77 static fmd_asru_t *
fmd_asru_create(fmd_asru_hash_t * ahp,const char * uuid,const char * name,nvlist_t * fmri)78 fmd_asru_create(fmd_asru_hash_t *ahp, const char *uuid,
79     const char *name, nvlist_t *fmri)
80 {
81 	fmd_asru_t *ap = fmd_zalloc(sizeof (fmd_asru_t), FMD_SLEEP);
82 	char *s;
83 
84 	(void) pthread_mutex_init(&ap->asru_lock, NULL);
85 	(void) pthread_cond_init(&ap->asru_cv, NULL);
86 
87 	ap->asru_name = fmd_strdup(name, FMD_SLEEP);
88 	if (fmri)
89 		(void) nvlist_xdup(fmri, &ap->asru_fmri, &fmd.d_nva);
90 	ap->asru_root = fmd_strdup(ahp->ah_dirpath, FMD_SLEEP);
91 	ap->asru_uuid = fmd_strdup(uuid, FMD_SLEEP);
92 	ap->asru_uuidlen = ap->asru_uuid ? strlen(ap->asru_uuid) : 0;
93 	ap->asru_refs = 1;
94 
95 	if (fmri && nvlist_lookup_string(fmri, FM_FMRI_SCHEME, &s) == 0 &&
96 	    strcmp(s, FM_FMRI_SCHEME_FMD) == 0)
97 		ap->asru_flags |= FMD_ASRU_INTERNAL;
98 
99 	return (ap);
100 }
101 
102 static void
fmd_asru_destroy(fmd_asru_t * ap)103 fmd_asru_destroy(fmd_asru_t *ap)
104 {
105 	ASSERT(MUTEX_HELD(&ap->asru_lock));
106 	ASSERT(ap->asru_refs == 0);
107 
108 	nvlist_free(ap->asru_event);
109 	fmd_strfree(ap->asru_name);
110 	nvlist_free(ap->asru_fmri);
111 	fmd_strfree(ap->asru_root);
112 	fmd_free(ap->asru_uuid, ap->asru_uuidlen + 1);
113 	fmd_free(ap, sizeof (fmd_asru_t));
114 }
115 
116 static void
fmd_asru_hash_insert(fmd_asru_hash_t * ahp,fmd_asru_t * ap)117 fmd_asru_hash_insert(fmd_asru_hash_t *ahp, fmd_asru_t *ap)
118 {
119 	uint_t h = fmd_asru_strhash(ahp, ap->asru_name);
120 
121 	ASSERT(RW_WRITE_HELD(&ahp->ah_lock));
122 	ap->asru_next = ahp->ah_hash[h];
123 	ahp->ah_hash[h] = ap;
124 	ahp->ah_count++;
125 }
126 
127 static fmd_asru_t *
fmd_asru_hold(fmd_asru_t * ap)128 fmd_asru_hold(fmd_asru_t *ap)
129 {
130 	(void) pthread_mutex_lock(&ap->asru_lock);
131 	ap->asru_refs++;
132 	ASSERT(ap->asru_refs != 0);
133 	(void) pthread_mutex_unlock(&ap->asru_lock);
134 	return (ap);
135 }
136 
137 /*
138  * Lookup an asru in the hash by name and place a hold on it.  If the asru is
139  * not found, no entry is created and NULL is returned.  This internal function
140  * is for callers who have the ah_lock held and is used by lookup_name below.
141  */
142 fmd_asru_t *
fmd_asru_hash_lookup(fmd_asru_hash_t * ahp,const char * name)143 fmd_asru_hash_lookup(fmd_asru_hash_t *ahp, const char *name)
144 {
145 	fmd_asru_t *ap;
146 	uint_t h;
147 
148 	ASSERT(RW_LOCK_HELD(&ahp->ah_lock));
149 	h = fmd_asru_strhash(ahp, name);
150 
151 	for (ap = ahp->ah_hash[h]; ap != NULL; ap = ap->asru_next) {
152 		if (fmd_asru_strcmp(ahp, ap->asru_name, name))
153 			break;
154 	}
155 
156 	if (ap != NULL)
157 		(void) fmd_asru_hold(ap);
158 	else
159 		(void) fmd_set_errno(EFMD_ASRU_NOENT);
160 
161 	return (ap);
162 }
163 
164 #define	HC_ONLY_FALSE	0
165 #define	HC_ONLY_TRUE	1
166 
167 static int
fmd_asru_replacement_state(nvlist_t * event,int hc_only)168 fmd_asru_replacement_state(nvlist_t *event, int hc_only)
169 {
170 	int ps = -1;
171 	nvlist_t *asru, *fru, *rsrc;
172 	char *s;
173 
174 	/*
175 	 * Check if there is evidence that this object is no longer present.
176 	 * In general fmd_fmri_present() should be supported on resources and/or
177 	 * frus, as those are the things that are physically present or not
178 	 * present - an asru can be spread over a number of frus some of which
179 	 * are present and some not, so fmd_fmri_present() is not generally
180 	 * meaningful. However retain a check for asru first for compatibility.
181 	 * If we have checked all three and we still get -1 then nothing knows
182 	 * whether it's present or not, so err on the safe side and treat it
183 	 * as still present.
184 	 *
185 	 * Note that if hc_only is set, then we only check status using fmris
186 	 * that are in hc-scheme.
187 	 */
188 	if (fmd_asru_fake_not_present)
189 		return (fmd_asru_fake_not_present);
190 	if (nvlist_lookup_nvlist(event, FM_FAULT_ASRU, &asru) == 0 &&
191 	    (hc_only == HC_ONLY_FALSE || (nvlist_lookup_string(asru,
192 	    FM_FMRI_SCHEME, &s) == 0 && strcmp(s, FM_FMRI_SCHEME_HC) == 0)))
193 		ps = fmd_fmri_replaced(asru);
194 	if (ps == -1 || ps == FMD_OBJ_STATE_UNKNOWN) {
195 		if (nvlist_lookup_nvlist(event, FM_FAULT_RESOURCE,
196 		    &rsrc) == 0 && (hc_only == HC_ONLY_FALSE ||
197 		    (nvlist_lookup_string(rsrc, FM_FMRI_SCHEME, &s) == 0 &&
198 		    strcmp(s, FM_FMRI_SCHEME_HC) == 0))) {
199 			if (ps == -1) {
200 				ps = fmd_fmri_replaced(rsrc);
201 			} else {
202 				/* see if we can improve on UNKNOWN */
203 				int ps2 = fmd_fmri_replaced(rsrc);
204 				if (ps2 == FMD_OBJ_STATE_STILL_PRESENT ||
205 				    ps2 == FMD_OBJ_STATE_REPLACED)
206 					ps = ps2;
207 			}
208 		}
209 	}
210 	if (ps == -1 || ps == FMD_OBJ_STATE_UNKNOWN) {
211 		if (nvlist_lookup_nvlist(event, FM_FAULT_FRU, &fru) == 0 &&
212 		    (hc_only == HC_ONLY_FALSE || (nvlist_lookup_string(fru,
213 		    FM_FMRI_SCHEME, &s) == 0 &&
214 		    strcmp(s, FM_FMRI_SCHEME_HC) == 0))) {
215 			if (ps == -1) {
216 				ps = fmd_fmri_replaced(fru);
217 			} else {
218 				/* see if we can improve on UNKNOWN */
219 				int ps2 = fmd_fmri_replaced(fru);
220 				if (ps2 == FMD_OBJ_STATE_STILL_PRESENT ||
221 				    ps2 == FMD_OBJ_STATE_REPLACED)
222 					ps = ps2;
223 			}
224 		}
225 	}
226 	if (ps == -1)
227 		ps = FMD_OBJ_STATE_UNKNOWN;
228 	return (ps);
229 }
230 
231 static void
fmd_asru_asru_hash_insert(fmd_asru_hash_t * ahp,fmd_asru_link_t * alp,char * name)232 fmd_asru_asru_hash_insert(fmd_asru_hash_t *ahp, fmd_asru_link_t *alp,
233     char *name)
234 {
235 	uint_t h = fmd_asru_strhash(ahp, name);
236 
237 	ASSERT(RW_WRITE_HELD(&ahp->ah_lock));
238 	alp->al_asru_next = ahp->ah_asru_hash[h];
239 	ahp->ah_asru_hash[h] = alp;
240 	ahp->ah_al_count++;
241 }
242 
243 static void
fmd_asru_case_hash_insert(fmd_asru_hash_t * ahp,fmd_asru_link_t * alp,char * name)244 fmd_asru_case_hash_insert(fmd_asru_hash_t *ahp, fmd_asru_link_t *alp,
245     char *name)
246 {
247 	uint_t h = fmd_asru_strhash(ahp, name);
248 
249 	ASSERT(RW_WRITE_HELD(&ahp->ah_lock));
250 	alp->al_case_next = ahp->ah_case_hash[h];
251 	ahp->ah_case_hash[h] = alp;
252 }
253 
254 static void
fmd_asru_fru_hash_insert(fmd_asru_hash_t * ahp,fmd_asru_link_t * alp,char * name)255 fmd_asru_fru_hash_insert(fmd_asru_hash_t *ahp, fmd_asru_link_t *alp, char *name)
256 {
257 	uint_t h = fmd_asru_strhash(ahp, name);
258 
259 	ASSERT(RW_WRITE_HELD(&ahp->ah_lock));
260 	alp->al_fru_next = ahp->ah_fru_hash[h];
261 	ahp->ah_fru_hash[h] = alp;
262 }
263 
264 static void
fmd_asru_label_hash_insert(fmd_asru_hash_t * ahp,fmd_asru_link_t * alp,char * name)265 fmd_asru_label_hash_insert(fmd_asru_hash_t *ahp, fmd_asru_link_t *alp,
266     char *name)
267 {
268 	uint_t h = fmd_asru_strhash(ahp, name);
269 
270 	ASSERT(RW_WRITE_HELD(&ahp->ah_lock));
271 	alp->al_label_next = ahp->ah_label_hash[h];
272 	ahp->ah_label_hash[h] = alp;
273 }
274 
275 static void
fmd_asru_rsrc_hash_insert(fmd_asru_hash_t * ahp,fmd_asru_link_t * alp,char * name)276 fmd_asru_rsrc_hash_insert(fmd_asru_hash_t *ahp, fmd_asru_link_t *alp,
277     char *name)
278 {
279 	uint_t h = fmd_asru_strhash(ahp, name);
280 
281 	ASSERT(RW_WRITE_HELD(&ahp->ah_lock));
282 	alp->al_rsrc_next = ahp->ah_rsrc_hash[h];
283 	ahp->ah_rsrc_hash[h] = alp;
284 }
285 
286 static void
fmd_asru_al_destroy(fmd_asru_link_t * alp)287 fmd_asru_al_destroy(fmd_asru_link_t *alp)
288 {
289 	ASSERT(alp->al_refs == 0);
290 	ASSERT(MUTEX_HELD(&alp->al_asru->asru_lock));
291 
292 	if (alp->al_log != NULL)
293 		fmd_log_rele(alp->al_log);
294 
295 	fmd_free(alp->al_uuid, alp->al_uuidlen + 1);
296 	nvlist_free(alp->al_event);
297 	fmd_strfree(alp->al_rsrc_name);
298 	fmd_strfree(alp->al_case_uuid);
299 	fmd_strfree(alp->al_fru_name);
300 	fmd_strfree(alp->al_asru_name);
301 	fmd_strfree(alp->al_label);
302 	nvlist_free(alp->al_asru_fmri);
303 	fmd_free(alp, sizeof (fmd_asru_link_t));
304 }
305 
306 static fmd_asru_link_t *
fmd_asru_al_hold(fmd_asru_link_t * alp)307 fmd_asru_al_hold(fmd_asru_link_t *alp)
308 {
309 	fmd_asru_t *ap = alp->al_asru;
310 
311 	(void) pthread_mutex_lock(&ap->asru_lock);
312 	ap->asru_refs++;
313 	alp->al_refs++;
314 	ASSERT(alp->al_refs != 0);
315 	(void) pthread_mutex_unlock(&ap->asru_lock);
316 	return (alp);
317 }
318 
319 static void fmd_asru_destroy(fmd_asru_t *ap);
320 
321 /*ARGSUSED*/
322 static void
fmd_asru_al_hash_release(fmd_asru_hash_t * ahp,fmd_asru_link_t * alp)323 fmd_asru_al_hash_release(fmd_asru_hash_t *ahp, fmd_asru_link_t *alp)
324 {
325 	fmd_asru_t *ap = alp->al_asru;
326 
327 	(void) pthread_mutex_lock(&ap->asru_lock);
328 	ASSERT(alp->al_refs != 0);
329 	if (--alp->al_refs == 0)
330 		fmd_asru_al_destroy(alp);
331 	ASSERT(ap->asru_refs != 0);
332 	if (--ap->asru_refs == 0)
333 		fmd_asru_destroy(ap);
334 	else
335 		(void) pthread_mutex_unlock(&ap->asru_lock);
336 }
337 
338 static int
fmd_asru_get_namestr(nvlist_t * nvl,char ** name,ssize_t * namelen)339 fmd_asru_get_namestr(nvlist_t *nvl, char **name, ssize_t *namelen)
340 {
341 	if ((*namelen = fmd_fmri_nvl2str(nvl, NULL, 0)) == -1)
342 		return (EFMD_ASRU_FMRI);
343 	*name = fmd_alloc(*namelen + 1, FMD_SLEEP);
344 	if (fmd_fmri_nvl2str(nvl, *name, *namelen + 1) == -1) {
345 		if (*name != NULL)
346 			fmd_free(*name, *namelen + 1);
347 		return (EFMD_ASRU_FMRI);
348 	}
349 	return (0);
350 }
351 
352 static fmd_asru_link_t *
fmd_asru_al_create(fmd_asru_hash_t * ahp,nvlist_t * nvl,fmd_case_t * cp,const char * al_uuid)353 fmd_asru_al_create(fmd_asru_hash_t *ahp, nvlist_t *nvl, fmd_case_t *cp,
354     const char *al_uuid)
355 {
356 	nvlist_t *asru = NULL, *fru, *rsrc;
357 	int got_rsrc = 0, got_asru = 0, got_fru = 0;
358 	ssize_t fru_namelen, rsrc_namelen, asru_namelen;
359 	char *asru_name, *rsrc_name, *fru_name, *name, *label;
360 	fmd_asru_link_t *alp;
361 	fmd_asru_t *ap;
362 	boolean_t msg;
363 	fmd_case_impl_t *cip = (fmd_case_impl_t *)cp;
364 
365 	if (nvlist_lookup_nvlist(nvl, FM_FAULT_ASRU, &asru) == 0 &&
366 	    fmd_asru_get_namestr(asru, &asru_name, &asru_namelen) == 0)
367 		got_asru = 1;
368 	if (nvlist_lookup_nvlist(nvl, FM_FAULT_FRU, &fru) == 0 &&
369 	    fmd_asru_get_namestr(fru, &fru_name, &fru_namelen) == 0)
370 		got_fru = 1;
371 	if (nvlist_lookup_nvlist(nvl, FM_FAULT_RESOURCE, &rsrc) == 0 &&
372 	    fmd_asru_get_namestr(rsrc, &rsrc_name, &rsrc_namelen) == 0)
373 		got_rsrc = 1;
374 	if (nvlist_lookup_string(nvl, FM_FAULT_LOCATION, &label) != 0)
375 		label = "";
376 
377 	/*
378 	 * Grab the rwlock as a writer; Then create and insert the asru with
379 	 * ahp->ah_lock held and hash it in. We'll then drop the rwlock and
380 	 * proceed to initializing the asru.
381 	 */
382 	(void) pthread_rwlock_wrlock(&ahp->ah_lock);
383 
384 	/*
385 	 * Create and initialise the per-fault "link" structure.
386 	 */
387 	alp = fmd_zalloc(sizeof (fmd_asru_link_t), FMD_SLEEP);
388 	if (got_asru)
389 		(void) nvlist_xdup(asru, &alp->al_asru_fmri, &fmd.d_nva);
390 	alp->al_uuid = fmd_strdup(al_uuid, FMD_SLEEP);
391 	alp->al_uuidlen = strlen(alp->al_uuid);
392 	alp->al_refs = 1;
393 
394 	/*
395 	 * If this is the first fault for this asru, then create the per-asru
396 	 * structure and link into the hash.
397 	 */
398 	name = got_asru ? asru_name : "";
399 	if ((ap = fmd_asru_hash_lookup(ahp, name)) == NULL) {
400 		ap = fmd_asru_create(ahp, al_uuid, name, got_asru ? asru :
401 		    NULL);
402 		fmd_asru_hash_insert(ahp, ap);
403 	} else
404 		nvlist_free(ap->asru_event);
405 	(void) nvlist_xdup(nvl, &ap->asru_event, &fmd.d_nva);
406 
407 	/*
408 	 * Put the link structure on the list associated with the per-asru
409 	 * structure. Then put the link structure on the various hashes.
410 	 */
411 	fmd_list_append(&ap->asru_list, (fmd_list_t *)alp);
412 	alp->al_asru = ap;
413 	alp->al_asru_name = got_asru ? asru_name : fmd_strdup("", FMD_SLEEP);
414 	fmd_asru_asru_hash_insert(ahp, alp, alp->al_asru_name);
415 	alp->al_fru_name = got_fru ? fru_name : fmd_strdup("", FMD_SLEEP);
416 	fmd_asru_fru_hash_insert(ahp, alp, alp->al_fru_name);
417 	alp->al_rsrc_name = got_rsrc ? rsrc_name : fmd_strdup("", FMD_SLEEP);
418 	fmd_asru_rsrc_hash_insert(ahp, alp, alp->al_rsrc_name);
419 	alp->al_label = fmd_strdup(label, FMD_SLEEP);
420 	fmd_asru_label_hash_insert(ahp, alp, label);
421 	alp->al_case_uuid = fmd_strdup(cip->ci_uuid, FMD_SLEEP);
422 	fmd_asru_case_hash_insert(ahp, alp, cip->ci_uuid);
423 	(void) pthread_mutex_lock(&ap->asru_lock);
424 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
425 
426 	ap->asru_case = alp->al_case = cp;
427 	if (nvlist_lookup_boolean_value(nvl, FM_SUSPECT_MESSAGE, &msg) == 0 &&
428 	    msg == B_FALSE)
429 		ap->asru_flags |= FMD_ASRU_INVISIBLE;
430 	(void) nvlist_xdup(nvl, &alp->al_event, &fmd.d_nva);
431 	ap->asru_flags |= FMD_ASRU_VALID;
432 	(void) pthread_cond_broadcast(&ap->asru_cv);
433 	(void) pthread_mutex_unlock(&ap->asru_lock);
434 	return (alp);
435 }
436 
437 static void
fmd_asru_hash_recreate(fmd_log_t * lp,fmd_event_t * ep,fmd_asru_hash_t * ahp)438 fmd_asru_hash_recreate(fmd_log_t *lp, fmd_event_t *ep, fmd_asru_hash_t *ahp)
439 {
440 	nvlist_t *nvl = FMD_EVENT_NVL(ep);
441 	boolean_t faulty = FMD_B_FALSE, unusable = FMD_B_FALSE;
442 	int ps;
443 	boolean_t repaired = FMD_B_FALSE, replaced = FMD_B_FALSE;
444 	boolean_t acquitted = FMD_B_FALSE, resolved = FMD_B_FALSE;
445 	nvlist_t *flt, *flt_copy, *asru;
446 	char *case_uuid = NULL, *case_code = NULL;
447 	fmd_asru_t *ap;
448 	fmd_asru_link_t *alp;
449 	fmd_case_t *cp;
450 	int64_t *diag_time;
451 	nvlist_t *de_fmri, *de_fmri_dup;
452 	uint_t nelem;
453 	topo_hdl_t *thp;
454 	char *class;
455 	nvlist_t *rsrc;
456 	int err;
457 	boolean_t injected;
458 
459 	/*
460 	 * Extract the most recent values of 'faulty' from the event log.
461 	 */
462 	if (nvlist_lookup_boolean_value(nvl, FM_RSRC_ASRU_FAULTY,
463 	    &faulty) != 0) {
464 		fmd_error(EFMD_ASRU_EVENT, "failed to reload asru %s: "
465 		    "invalid event log record\n", lp->log_name);
466 		ahp->ah_error = EFMD_ASRU_EVENT;
467 		return;
468 	}
469 	if (nvlist_lookup_nvlist(nvl, FM_RSRC_ASRU_EVENT, &flt) != 0) {
470 		fmd_error(EFMD_ASRU_EVENT, "failed to reload asru %s: "
471 		    "invalid event log record\n", lp->log_name);
472 		ahp->ah_error = EFMD_ASRU_EVENT;
473 		return;
474 	}
475 	(void) nvlist_lookup_string(nvl, FM_RSRC_ASRU_UUID, &case_uuid);
476 	(void) nvlist_lookup_string(nvl, FM_RSRC_ASRU_CODE, &case_code);
477 	(void) nvlist_lookup_boolean_value(nvl, FM_RSRC_ASRU_UNUSABLE,
478 	    &unusable);
479 	(void) nvlist_lookup_boolean_value(nvl, FM_RSRC_ASRU_REPAIRED,
480 	    &repaired);
481 	(void) nvlist_lookup_boolean_value(nvl, FM_RSRC_ASRU_REPLACED,
482 	    &replaced);
483 	(void) nvlist_lookup_boolean_value(nvl, FM_RSRC_ASRU_ACQUITTED,
484 	    &acquitted);
485 	(void) nvlist_lookup_boolean_value(nvl, FM_RSRC_ASRU_RESOLVED,
486 	    &resolved);
487 
488 	/*
489 	 * Attempt to recreate the case in CLOSED, REPAIRED or RESOLVED state
490 	 * (depending on whether the faulty/resolved bits are set).
491 	 * If the case is already present, fmd_case_recreate() will return it.
492 	 * If not, we'll create a new orphaned case. Either way,  we use the
493 	 * ASRU event to insert a suspect into the partially-restored case.
494 	 */
495 	fmd_module_lock(fmd.d_rmod);
496 	cp = fmd_case_recreate(fmd.d_rmod, NULL, faulty ? FMD_CASE_CLOSED :
497 	    resolved ? FMD_CASE_RESOLVED : FMD_CASE_REPAIRED, case_uuid,
498 	    case_code);
499 	fmd_case_hold(cp);
500 	fmd_module_unlock(fmd.d_rmod);
501 	if (nvlist_lookup_boolean_value(nvl, FM_SUSPECT_INJECTED,
502 	    &injected) == 0 && injected)
503 		fmd_case_set_injected(cp);
504 	if (nvlist_lookup_int64_array(nvl, FM_SUSPECT_DIAG_TIME, &diag_time,
505 	    &nelem) == 0 && nelem >= 2)
506 		fmd_case_settime(cp, diag_time[0], diag_time[1]);
507 	else
508 		fmd_case_settime(cp, lp->log_stat.st_ctime, 0);
509 	if (nvlist_lookup_nvlist(nvl, FM_SUSPECT_DE, &de_fmri) == 0) {
510 		(void) nvlist_xdup(de_fmri, &de_fmri_dup, &fmd.d_nva);
511 		fmd_case_set_de_fmri(cp, de_fmri_dup);
512 	}
513 	(void) nvlist_xdup(flt, &flt_copy, &fmd.d_nva);
514 
515 	/*
516 	 * For faults with a resource, re-evaluate the asru from the resource.
517 	 */
518 	thp = fmd_fmri_topo_hold(TOPO_VERSION);
519 	if (nvlist_lookup_string(flt_copy, FM_CLASS, &class) == 0 &&
520 	    strncmp(class, "fault", 5) == 0 &&
521 	    nvlist_lookup_nvlist(flt_copy, FM_FAULT_RESOURCE, &rsrc) == 0 &&
522 	    rsrc != NULL &&
523 	    (fmd_fmri_replaced(rsrc) != FMD_OBJ_STATE_REPLACED) &&
524 	    topo_fmri_asru(thp, rsrc, &asru, &err) == 0) {
525 		(void) nvlist_remove(flt_copy, FM_FAULT_ASRU, DATA_TYPE_NVLIST);
526 		(void) nvlist_add_nvlist(flt_copy, FM_FAULT_ASRU, asru);
527 		nvlist_free(asru);
528 	}
529 	fmd_fmri_topo_rele(thp);
530 
531 	(void) nvlist_xdup(flt_copy, &flt, &fmd.d_nva);
532 
533 	fmd_case_recreate_suspect(cp, flt_copy);
534 
535 	/*
536 	 * Now create the resource cache entries.
537 	 */
538 	alp = fmd_asru_al_create(ahp, flt, cp, fmd_strbasename(lp->log_name));
539 	ap = alp->al_asru;
540 
541 	/*
542 	 * Check to see if the resource is still present in the system.
543 	 */
544 	ps = fmd_asru_replacement_state(flt, HC_ONLY_FALSE);
545 	if (ps == FMD_OBJ_STATE_REPLACED) {
546 		replaced = FMD_B_TRUE;
547 	} else if (ps == FMD_OBJ_STATE_STILL_PRESENT ||
548 	    ps == FMD_OBJ_STATE_UNKNOWN) {
549 		ap->asru_flags |= FMD_ASRU_PRESENT;
550 		if (nvlist_lookup_nvlist(alp->al_event, FM_FAULT_ASRU,
551 		    &asru) == 0) {
552 			int us;
553 
554 			switch (fmd_fmri_service_state(asru)) {
555 			case FMD_SERVICE_STATE_UNUSABLE:
556 				unusable = FMD_B_TRUE;
557 				break;
558 			case FMD_SERVICE_STATE_OK:
559 			case FMD_SERVICE_STATE_ISOLATE_PENDING:
560 			case FMD_SERVICE_STATE_DEGRADED:
561 				unusable = FMD_B_FALSE;
562 				break;
563 			case FMD_SERVICE_STATE_UNKNOWN:
564 			case -1:
565 				/* not supported by scheme */
566 				us = fmd_fmri_unusable(asru);
567 				if (us > 0)
568 					unusable = FMD_B_TRUE;
569 				else if (us == 0)
570 					unusable = FMD_B_FALSE;
571 				break;
572 			}
573 		}
574 	}
575 
576 	nvlist_free(flt);
577 
578 	ap->asru_flags |= FMD_ASRU_RECREATED;
579 	if (faulty) {
580 		alp->al_flags |= FMD_ASRU_FAULTY;
581 		ap->asru_flags |= FMD_ASRU_FAULTY;
582 	}
583 	if (unusable) {
584 		alp->al_flags |= FMD_ASRU_UNUSABLE;
585 		ap->asru_flags |= FMD_ASRU_UNUSABLE;
586 	}
587 	if (replaced)
588 		alp->al_reason = FMD_ASRU_REPLACED;
589 	else if (repaired)
590 		alp->al_reason = FMD_ASRU_REPAIRED;
591 	else if (acquitted)
592 		alp->al_reason = FMD_ASRU_ACQUITTED;
593 	else
594 		alp->al_reason = FMD_ASRU_REMOVED;
595 
596 	TRACE((FMD_DBG_ASRU, "asru %s recreated as %p (%s)", alp->al_uuid,
597 	    (void *)ap, _fmd_asru_snames[ap->asru_flags & FMD_ASRU_STATE]));
598 }
599 
600 static void
fmd_asru_hash_discard(fmd_asru_hash_t * ahp,const char * uuid,int err)601 fmd_asru_hash_discard(fmd_asru_hash_t *ahp, const char *uuid, int err)
602 {
603 	char src[PATH_MAX], dst[PATH_MAX];
604 
605 	(void) snprintf(src, PATH_MAX, "%s/%s", ahp->ah_dirpath, uuid);
606 	(void) snprintf(dst, PATH_MAX, "%s/%s-", ahp->ah_dirpath, uuid);
607 
608 	if (err != 0)
609 		err = rename(src, dst);
610 	else
611 		err = unlink(src);
612 
613 	if (err != 0 && errno != ENOENT)
614 		fmd_error(EFMD_ASRU_EVENT, "failed to rename log %s", src);
615 }
616 
617 /*
618  * Open a saved log file and restore it into the ASRU hash.  If we can't even
619  * open the log, rename the log file to <uuid>- to indicate it is corrupt.  If
620  * fmd_log_replay() fails, we either delete the file (if it has reached the
621  * upper limit on cache age) or rename it for debugging if it was corrupted.
622  */
623 static void
fmd_asru_hash_logopen(fmd_asru_hash_t * ahp,const char * uuid)624 fmd_asru_hash_logopen(fmd_asru_hash_t *ahp, const char *uuid)
625 {
626 	fmd_log_t *lp = fmd_log_tryopen(ahp->ah_dirpath, uuid, FMD_LOG_ASRU);
627 	uint_t n;
628 
629 	if (lp == NULL) {
630 		fmd_asru_hash_discard(ahp, uuid, errno);
631 		return;
632 	}
633 
634 	ahp->ah_error = 0;
635 	n = ahp->ah_al_count;
636 
637 	fmd_log_replay(lp, (fmd_log_f *)fmd_asru_hash_recreate, ahp);
638 	fmd_log_rele(lp);
639 
640 	if (ahp->ah_al_count == n)
641 		fmd_asru_hash_discard(ahp, uuid, ahp->ah_error);
642 }
643 
644 void
fmd_asru_hash_refresh(fmd_asru_hash_t * ahp)645 fmd_asru_hash_refresh(fmd_asru_hash_t *ahp)
646 {
647 	struct dirent *dp;
648 	DIR *dirp;
649 	int zero;
650 
651 	if ((dirp = opendir(ahp->ah_dirpath)) == NULL) {
652 		fmd_error(EFMD_ASRU_NODIR,
653 		    "failed to open asru cache directory %s", ahp->ah_dirpath);
654 		return;
655 	}
656 
657 	(void) fmd_conf_getprop(fmd.d_conf, "rsrc.zero", &zero);
658 
659 	(void) pthread_rwlock_wrlock(&ahp->ah_lock);
660 
661 	while ((dp = readdir(dirp)) != NULL) {
662 		if (dp->d_name[0] == '.')
663 			continue; /* skip "." and ".." */
664 
665 		if (zero)
666 			fmd_asru_hash_discard(ahp, dp->d_name, 0);
667 		else if (!fmd_strmatch(dp->d_name, "*-"))
668 			fmd_asru_hash_logopen(ahp, dp->d_name);
669 	}
670 
671 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
672 	(void) closedir(dirp);
673 }
674 
675 /*
676  * If the resource is present and faulty but not unusable, replay the fault
677  * event that caused it be marked faulty.  This will cause the agent
678  * subscribing to this fault class to again disable the resource.
679  */
680 /*ARGSUSED*/
681 static void
fmd_asru_hash_replay_asru(fmd_asru_t * ap,void * data)682 fmd_asru_hash_replay_asru(fmd_asru_t *ap, void *data)
683 {
684 	fmd_event_t *e;
685 	nvlist_t *nvl;
686 	char *class;
687 
688 	if (ap->asru_event != NULL && (ap->asru_flags & (FMD_ASRU_STATE |
689 	    FMD_ASRU_PRESENT)) == (FMD_ASRU_FAULTY | FMD_ASRU_PRESENT)) {
690 
691 		fmd_dprintf(FMD_DBG_ASRU,
692 		    "replaying fault event for %s", ap->asru_name);
693 
694 		(void) nvlist_xdup(ap->asru_event, &nvl, &fmd.d_nva);
695 		(void) nvlist_lookup_string(nvl, FM_CLASS, &class);
696 
697 		(void) nvlist_add_string(nvl, FMD_EVN_UUID,
698 		    ((fmd_case_impl_t *)ap->asru_case)->ci_uuid);
699 
700 		e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, class);
701 		fmd_dispq_dispatch(fmd.d_disp, e, class);
702 	}
703 }
704 
705 void
fmd_asru_hash_replay(fmd_asru_hash_t * ahp)706 fmd_asru_hash_replay(fmd_asru_hash_t *ahp)
707 {
708 	fmd_asru_hash_apply(ahp, fmd_asru_hash_replay_asru, NULL);
709 }
710 
711 /*
712  * Check if the resource is still present. If not, and if the rsrc.age time
713  * has expired, then do an implicit repair on the resource.
714  */
715 /*ARGSUSED*/
716 static void
fmd_asru_repair_if_aged(fmd_asru_link_t * alp,void * arg)717 fmd_asru_repair_if_aged(fmd_asru_link_t *alp, void *arg)
718 {
719 	struct timeval tv;
720 	fmd_log_t *lp;
721 	hrtime_t hrt;
722 	int ps;
723 	int err;
724 	fmd_asru_rep_arg_t fara;
725 
726 	if (!(alp->al_flags & FMD_ASRU_FAULTY))
727 		return;
728 
729 	/*
730 	 * Checking for aged resources only happens on the diagnosing side
731 	 * not on a proxy.
732 	 */
733 	if (alp->al_flags & FMD_ASRU_PROXY)
734 		return;
735 
736 	ps = fmd_asru_replacement_state(alp->al_event, HC_ONLY_FALSE);
737 	if (ps == FMD_OBJ_STATE_REPLACED) {
738 		fara.fara_reason = FMD_ASRU_REPLACED;
739 		fara.fara_bywhat = FARA_ALL;
740 		fara.fara_rval = &err;
741 		fmd_asru_repaired(alp, &fara);
742 	} else if (ps == FMD_OBJ_STATE_NOT_PRESENT) {
743 		fmd_time_gettimeofday(&tv);
744 		lp = fmd_log_open(alp->al_asru->asru_root, alp->al_uuid,
745 		    FMD_LOG_ASRU);
746 		if (lp == NULL)
747 			return;
748 		hrt = (hrtime_t)(tv.tv_sec - lp->log_stat.st_mtime);
749 		fmd_log_rele(lp);
750 		if (hrt * NANOSEC >= fmd.d_asrus->ah_lifetime) {
751 			fara.fara_reason = FMD_ASRU_REMOVED;
752 			fara.fara_bywhat = FARA_ALL;
753 			fara.fara_rval = &err;
754 			fmd_asru_repaired(alp, &fara);
755 		}
756 	}
757 }
758 
759 /*ARGSUSED*/
760 void
fmd_asru_check_if_aged(fmd_asru_link_t * alp,void * arg)761 fmd_asru_check_if_aged(fmd_asru_link_t *alp, void *arg)
762 {
763 	struct timeval tv;
764 	fmd_log_t *lp;
765 	hrtime_t hrt;
766 
767 	/*
768 	 * Case must be in resolved state for this to be called. So modified
769 	 * time on resource cache entry should be the time the resolve occurred.
770 	 * Return 0 if not yet hit rsrc.aged.
771 	 */
772 	fmd_time_gettimeofday(&tv);
773 	lp = fmd_log_open(alp->al_asru->asru_root, alp->al_uuid, FMD_LOG_ASRU);
774 	if (lp == NULL)
775 		return;
776 	hrt = (hrtime_t)(tv.tv_sec - lp->log_stat.st_mtime);
777 	fmd_log_rele(lp);
778 	if (hrt * NANOSEC < fmd.d_asrus->ah_lifetime)
779 		*(int *)arg = 0;
780 }
781 
782 /*ARGSUSED*/
783 void
fmd_asru_most_recent(fmd_asru_link_t * alp,void * arg)784 fmd_asru_most_recent(fmd_asru_link_t *alp, void *arg)
785 {
786 	fmd_log_t *lp;
787 	uint64_t hrt;
788 
789 	/*
790 	 * Find most recent modified time of a set of resource cache entries.
791 	 */
792 	lp = fmd_log_open(alp->al_asru->asru_root, alp->al_uuid, FMD_LOG_ASRU);
793 	if (lp == NULL)
794 		return;
795 	hrt = lp->log_stat.st_mtime;
796 	fmd_log_rele(lp);
797 	if (*(uint64_t *)arg < hrt)
798 		*(uint64_t *)arg = hrt;
799 }
800 
801 void
fmd_asru_clear_aged_rsrcs()802 fmd_asru_clear_aged_rsrcs()
803 {
804 	int check_if_aged = 1;
805 	fmd_asru_al_hash_apply(fmd.d_asrus, fmd_asru_repair_if_aged, NULL);
806 	fmd_case_hash_apply(fmd.d_cases, fmd_case_discard_resolved,
807 	    &check_if_aged);
808 }
809 
810 fmd_asru_hash_t *
fmd_asru_hash_create(const char * root,const char * dir)811 fmd_asru_hash_create(const char *root, const char *dir)
812 {
813 	fmd_asru_hash_t *ahp;
814 	char path[PATH_MAX];
815 
816 	ahp = fmd_alloc(sizeof (fmd_asru_hash_t), FMD_SLEEP);
817 	(void) pthread_rwlock_init(&ahp->ah_lock, NULL);
818 	ahp->ah_hashlen = fmd.d_str_buckets;
819 	ahp->ah_hash = fmd_zalloc(sizeof (void *) * ahp->ah_hashlen, FMD_SLEEP);
820 	ahp->ah_asru_hash = fmd_zalloc(sizeof (void *) * ahp->ah_hashlen,
821 	    FMD_SLEEP);
822 	ahp->ah_case_hash = fmd_zalloc(sizeof (void *) * ahp->ah_hashlen,
823 	    FMD_SLEEP);
824 	ahp->ah_fru_hash = fmd_zalloc(sizeof (void *) * ahp->ah_hashlen,
825 	    FMD_SLEEP);
826 	ahp->ah_label_hash = fmd_zalloc(sizeof (void *) * ahp->ah_hashlen,
827 	    FMD_SLEEP);
828 	ahp->ah_rsrc_hash = fmd_zalloc(sizeof (void *) * ahp->ah_hashlen,
829 	    FMD_SLEEP);
830 	(void) snprintf(path, sizeof (path), "%s/%s", root, dir);
831 	ahp->ah_dirpath = fmd_strdup(path, FMD_SLEEP);
832 	(void) fmd_conf_getprop(fmd.d_conf, "rsrc.age", &ahp->ah_lifetime);
833 	(void) fmd_conf_getprop(fmd.d_conf, "fakenotpresent",
834 	    (uint32_t *)&fmd_asru_fake_not_present);
835 	ahp->ah_al_count = 0;
836 	ahp->ah_count = 0;
837 	ahp->ah_error = 0;
838 	ahp->ah_topo = fmd_topo_hold();
839 
840 	return (ahp);
841 }
842 
843 void
fmd_asru_hash_destroy(fmd_asru_hash_t * ahp)844 fmd_asru_hash_destroy(fmd_asru_hash_t *ahp)
845 {
846 	fmd_asru_link_t *alp, *np;
847 	uint_t i;
848 
849 	for (i = 0; i < ahp->ah_hashlen; i++) {
850 		for (alp = ahp->ah_case_hash[i]; alp != NULL; alp = np) {
851 			np = alp->al_case_next;
852 			alp->al_case_next = NULL;
853 			fmd_case_rele(alp->al_case);
854 			alp->al_case = NULL;
855 			fmd_asru_al_hash_release(ahp, alp);
856 		}
857 	}
858 
859 	fmd_strfree(ahp->ah_dirpath);
860 	fmd_free(ahp->ah_hash, sizeof (void *) * ahp->ah_hashlen);
861 	fmd_free(ahp->ah_asru_hash, sizeof (void *) * ahp->ah_hashlen);
862 	fmd_free(ahp->ah_case_hash, sizeof (void *) * ahp->ah_hashlen);
863 	fmd_free(ahp->ah_fru_hash, sizeof (void *) * ahp->ah_hashlen);
864 	fmd_free(ahp->ah_label_hash, sizeof (void *) * ahp->ah_hashlen);
865 	fmd_free(ahp->ah_rsrc_hash, sizeof (void *) * ahp->ah_hashlen);
866 	fmd_topo_rele(ahp->ah_topo);
867 	fmd_free(ahp, sizeof (fmd_asru_hash_t));
868 }
869 
870 /*
871  * Take a snapshot of the ASRU database by placing an additional hold on each
872  * member in an auxiliary array, and then call 'func' for each ASRU.
873  */
874 void
fmd_asru_hash_apply(fmd_asru_hash_t * ahp,void (* func)(fmd_asru_t *,void *),void * arg)875 fmd_asru_hash_apply(fmd_asru_hash_t *ahp,
876     void (*func)(fmd_asru_t *, void *), void *arg)
877 {
878 	fmd_asru_t *ap, **aps, **app;
879 	uint_t apc, i;
880 
881 	(void) pthread_rwlock_rdlock(&ahp->ah_lock);
882 
883 	aps = app = fmd_alloc(ahp->ah_count * sizeof (fmd_asru_t *), FMD_SLEEP);
884 	apc = ahp->ah_count;
885 
886 	for (i = 0; i < ahp->ah_hashlen; i++) {
887 		for (ap = ahp->ah_hash[i]; ap != NULL; ap = ap->asru_next)
888 			*app++ = fmd_asru_hold(ap);
889 	}
890 
891 	ASSERT(app == aps + apc);
892 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
893 
894 	for (i = 0; i < apc; i++) {
895 		if (aps[i]->asru_fmri != NULL)
896 			func(aps[i], arg);
897 		fmd_asru_hash_release(ahp, aps[i]);
898 	}
899 
900 	fmd_free(aps, apc * sizeof (fmd_asru_t *));
901 }
902 
903 void
fmd_asru_al_hash_apply(fmd_asru_hash_t * ahp,void (* func)(fmd_asru_link_t *,void *),void * arg)904 fmd_asru_al_hash_apply(fmd_asru_hash_t *ahp,
905     void (*func)(fmd_asru_link_t *, void *), void *arg)
906 {
907 	fmd_asru_link_t *alp, **alps, **alpp;
908 	uint_t alpc, i;
909 
910 	(void) pthread_rwlock_rdlock(&ahp->ah_lock);
911 
912 	alps = alpp = fmd_alloc(ahp->ah_al_count * sizeof (fmd_asru_link_t *),
913 	    FMD_SLEEP);
914 	alpc = ahp->ah_al_count;
915 
916 	for (i = 0; i < ahp->ah_hashlen; i++) {
917 		for (alp = ahp->ah_case_hash[i]; alp != NULL;
918 		    alp = alp->al_case_next)
919 			*alpp++ = fmd_asru_al_hold(alp);
920 	}
921 
922 	ASSERT(alpp == alps + alpc);
923 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
924 
925 	for (i = 0; i < alpc; i++) {
926 		func(alps[i], arg);
927 		fmd_asru_al_hash_release(ahp, alps[i]);
928 	}
929 
930 	fmd_free(alps, alpc * sizeof (fmd_asru_link_t *));
931 }
932 
933 static void
fmd_asru_do_hash_apply(fmd_asru_hash_t * ahp,const char * name,void (* func)(fmd_asru_link_t *,void *),void * arg,fmd_asru_link_t ** hash,size_t match_offset,size_t next_offset)934 fmd_asru_do_hash_apply(fmd_asru_hash_t *ahp, const char *name,
935     void (*func)(fmd_asru_link_t *, void *), void *arg,
936     fmd_asru_link_t **hash, size_t match_offset, size_t next_offset)
937 {
938 	fmd_asru_link_t *alp, **alps, **alpp;
939 	uint_t alpc = 0, i;
940 	uint_t h;
941 
942 	(void) pthread_rwlock_rdlock(&ahp->ah_lock);
943 
944 	h = fmd_asru_strhash(ahp, name);
945 
946 	for (alp = hash[h]; alp != NULL; alp =
947 	    /* LINTED pointer alignment */
948 	    FMD_ASRU_AL_HASH_NEXT(alp, next_offset))
949 		if (fmd_asru_strcmp(ahp,
950 		    /* LINTED pointer alignment */
951 		    FMD_ASRU_AL_HASH_NAME(alp, match_offset), name))
952 			alpc++;
953 
954 	alps = alpp = fmd_alloc(alpc * sizeof (fmd_asru_link_t *), FMD_SLEEP);
955 
956 	for (alp = hash[h]; alp != NULL; alp =
957 	    /* LINTED pointer alignment */
958 	    FMD_ASRU_AL_HASH_NEXT(alp, next_offset))
959 		if (fmd_asru_strcmp(ahp,
960 		    /* LINTED pointer alignment */
961 		    FMD_ASRU_AL_HASH_NAME(alp, match_offset), name))
962 			*alpp++ = fmd_asru_al_hold(alp);
963 
964 	ASSERT(alpp == alps + alpc);
965 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
966 
967 	for (i = 0; i < alpc; i++) {
968 		func(alps[i], arg);
969 		fmd_asru_al_hash_release(ahp, alps[i]);
970 	}
971 
972 	fmd_free(alps, alpc * sizeof (fmd_asru_link_t *));
973 }
974 
975 void
fmd_asru_hash_apply_by_asru(fmd_asru_hash_t * ahp,const char * name,void (* func)(fmd_asru_link_t *,void *),void * arg)976 fmd_asru_hash_apply_by_asru(fmd_asru_hash_t *ahp, const char *name,
977     void (*func)(fmd_asru_link_t *, void *), void *arg)
978 {
979 	fmd_asru_do_hash_apply(ahp, name, func, arg, ahp->ah_asru_hash,
980 	    offsetof(fmd_asru_link_t, al_asru_name),
981 	    offsetof(fmd_asru_link_t, al_asru_next));
982 }
983 
984 void
fmd_asru_hash_apply_by_case(fmd_asru_hash_t * ahp,fmd_case_t * cp,void (* func)(fmd_asru_link_t *,void *),void * arg)985 fmd_asru_hash_apply_by_case(fmd_asru_hash_t *ahp, fmd_case_t *cp,
986 	void (*func)(fmd_asru_link_t *, void *), void *arg)
987 {
988 	fmd_asru_do_hash_apply(ahp, ((fmd_case_impl_t *)cp)->ci_uuid, func, arg,
989 	    ahp->ah_case_hash, offsetof(fmd_asru_link_t, al_case_uuid),
990 	    offsetof(fmd_asru_link_t, al_case_next));
991 }
992 
993 void
fmd_asru_hash_apply_by_fru(fmd_asru_hash_t * ahp,const char * name,void (* func)(fmd_asru_link_t *,void *),void * arg)994 fmd_asru_hash_apply_by_fru(fmd_asru_hash_t *ahp, const char *name,
995     void (*func)(fmd_asru_link_t *, void *), void *arg)
996 {
997 	fmd_asru_do_hash_apply(ahp, name, func, arg, ahp->ah_fru_hash,
998 	    offsetof(fmd_asru_link_t, al_fru_name),
999 	    offsetof(fmd_asru_link_t, al_fru_next));
1000 }
1001 
1002 void
fmd_asru_hash_apply_by_rsrc(fmd_asru_hash_t * ahp,const char * name,void (* func)(fmd_asru_link_t *,void *),void * arg)1003 fmd_asru_hash_apply_by_rsrc(fmd_asru_hash_t *ahp, const char *name,
1004     void (*func)(fmd_asru_link_t *, void *), void *arg)
1005 {
1006 	fmd_asru_do_hash_apply(ahp, name, func, arg, ahp->ah_rsrc_hash,
1007 	    offsetof(fmd_asru_link_t, al_rsrc_name),
1008 	    offsetof(fmd_asru_link_t, al_rsrc_next));
1009 }
1010 
1011 void
fmd_asru_hash_apply_by_label(fmd_asru_hash_t * ahp,const char * name,void (* func)(fmd_asru_link_t *,void *),void * arg)1012 fmd_asru_hash_apply_by_label(fmd_asru_hash_t *ahp, const char *name,
1013     void (*func)(fmd_asru_link_t *, void *), void *arg)
1014 {
1015 	fmd_asru_do_hash_apply(ahp, name, func, arg, ahp->ah_label_hash,
1016 	    offsetof(fmd_asru_link_t, al_label),
1017 	    offsetof(fmd_asru_link_t, al_label_next));
1018 }
1019 
1020 /*
1021  * Lookup an asru in the hash by name and place a hold on it.  If the asru is
1022  * not found, no entry is created and NULL is returned.
1023  */
1024 fmd_asru_t *
fmd_asru_hash_lookup_name(fmd_asru_hash_t * ahp,const char * name)1025 fmd_asru_hash_lookup_name(fmd_asru_hash_t *ahp, const char *name)
1026 {
1027 	fmd_asru_t *ap;
1028 
1029 	(void) pthread_rwlock_rdlock(&ahp->ah_lock);
1030 	ap = fmd_asru_hash_lookup(ahp, name);
1031 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
1032 
1033 	return (ap);
1034 }
1035 
1036 /*
1037  * Create a resource cache entry using the fault event "nvl" for one of the
1038  * suspects from the case "cp".
1039  *
1040  * The fault event can have the following components :  FM_FAULT_ASRU,
1041  * FM_FAULT_FRU, FM_FAULT_RESOURCE. These should be set by the Diagnosis Engine
1042  * when calling fmd_nvl_create_fault(). In the general case, these are all
1043  * optional and an entry will always be added into the cache even if one or all
1044  * of these fields is missing.
1045  *
1046  * However, for hardware faults the recommended practice is that the fault
1047  * event should always have the FM_FAULT_RESOURCE field present and that this
1048  * should be represented in hc-scheme.
1049  *
1050  * Currently the DE should also add the FM_FAULT_ASRU and FM_FAULT_FRU fields
1051  * where known, though at some future stage fmd might be able to fill these
1052  * in automatically from the topology.
1053  */
1054 fmd_asru_link_t *
fmd_asru_hash_create_entry(fmd_asru_hash_t * ahp,fmd_case_t * cp,nvlist_t * nvl)1055 fmd_asru_hash_create_entry(fmd_asru_hash_t *ahp, fmd_case_t *cp, nvlist_t *nvl)
1056 {
1057 	char *parsed_uuid;
1058 	uuid_t uuid;
1059 	int uuidlen;
1060 	fmd_asru_link_t *alp;
1061 
1062 	/*
1063 	 * Generate a UUID for the ASRU.  libuuid cleverly gives us no
1064 	 * interface for specifying or learning the buffer size.  Sigh.
1065 	 * The spec says 36 bytes but we use a tunable just to be safe.
1066 	 */
1067 	(void) fmd_conf_getprop(fmd.d_conf, "uuidlen", &uuidlen);
1068 	parsed_uuid = fmd_zalloc(uuidlen + 1, FMD_SLEEP);
1069 	uuid_generate(uuid);
1070 	uuid_unparse(uuid, parsed_uuid);
1071 
1072 	/*
1073 	 * Now create the resource cache entries.
1074 	 */
1075 	fmd_case_hold_locked(cp);
1076 	alp = fmd_asru_al_create(ahp, nvl, cp, parsed_uuid);
1077 	TRACE((FMD_DBG_ASRU, "asru %s created as %p",
1078 	    alp->al_uuid, (void *)alp->al_asru));
1079 
1080 	fmd_free(parsed_uuid, uuidlen + 1);
1081 	return (alp);
1082 
1083 }
1084 
1085 /*
1086  * Release the reference count on an asru obtained using fmd_asru_hash_lookup.
1087  * We take 'ahp' for symmetry and in case we need to use it in future work.
1088  */
1089 /*ARGSUSED*/
1090 void
fmd_asru_hash_release(fmd_asru_hash_t * ahp,fmd_asru_t * ap)1091 fmd_asru_hash_release(fmd_asru_hash_t *ahp, fmd_asru_t *ap)
1092 {
1093 	(void) pthread_mutex_lock(&ap->asru_lock);
1094 
1095 	ASSERT(ap->asru_refs != 0);
1096 	if (--ap->asru_refs == 0)
1097 		fmd_asru_destroy(ap);
1098 	else
1099 		(void) pthread_mutex_unlock(&ap->asru_lock);
1100 }
1101 
1102 static void
fmd_asru_do_delete_entry(fmd_asru_hash_t * ahp,fmd_case_t * cp,fmd_asru_link_t ** hash,size_t next_offset,char * name)1103 fmd_asru_do_delete_entry(fmd_asru_hash_t *ahp, fmd_case_t *cp,
1104     fmd_asru_link_t **hash, size_t next_offset, char *name)
1105 {
1106 	uint_t h;
1107 	fmd_asru_link_t *alp, **pp, *alpnext, **alpnextp;
1108 
1109 	(void) pthread_rwlock_wrlock(&ahp->ah_lock);
1110 	h = fmd_asru_strhash(ahp, name);
1111 	pp = &hash[h];
1112 	for (alp = *pp; alp != NULL; alp = alpnext) {
1113 		/* LINTED pointer alignment */
1114 		alpnextp = FMD_ASRU_AL_HASH_NEXTP(alp, next_offset);
1115 		alpnext = *alpnextp;
1116 		if (alp->al_case == cp) {
1117 			*pp = *alpnextp;
1118 			*alpnextp = NULL;
1119 		} else
1120 			pp = alpnextp;
1121 	}
1122 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
1123 }
1124 
1125 static void
fmd_asru_do_hash_delete(fmd_asru_hash_t * ahp,fmd_case_susp_t * cis,fmd_case_t * cp,fmd_asru_link_t ** hash,size_t next_offset,char * nvname)1126 fmd_asru_do_hash_delete(fmd_asru_hash_t *ahp, fmd_case_susp_t *cis,
1127     fmd_case_t *cp, fmd_asru_link_t **hash, size_t next_offset, char *nvname)
1128 {
1129 	nvlist_t *nvl;
1130 	char *name = NULL;
1131 	ssize_t namelen;
1132 
1133 	if (nvlist_lookup_nvlist(cis->cis_nvl, nvname, &nvl) == 0 &&
1134 	    (namelen = fmd_fmri_nvl2str(nvl, NULL, 0)) != -1 &&
1135 	    (name = fmd_alloc(namelen + 1, FMD_SLEEP)) != NULL) {
1136 		if (fmd_fmri_nvl2str(nvl, name, namelen + 1) != -1)
1137 			fmd_asru_do_delete_entry(ahp, cp, hash, next_offset,
1138 			    name);
1139 		fmd_free(name, namelen + 1);
1140 	} else
1141 		fmd_asru_do_delete_entry(ahp, cp, hash, next_offset, "");
1142 }
1143 
1144 void
fmd_asru_hash_delete_case(fmd_asru_hash_t * ahp,fmd_case_t * cp)1145 fmd_asru_hash_delete_case(fmd_asru_hash_t *ahp, fmd_case_t *cp)
1146 {
1147 	fmd_case_impl_t *cip = (fmd_case_impl_t *)cp;
1148 	fmd_case_susp_t *cis;
1149 	fmd_asru_link_t *alp, **plp, *alpnext;
1150 	fmd_asru_t *ap;
1151 	char path[PATH_MAX];
1152 	char *label;
1153 	uint_t h;
1154 
1155 	/*
1156 	 * first delete hash entries for each suspect
1157 	 */
1158 	for (cis = cip->ci_suspects; cis != NULL; cis = cis->cis_next) {
1159 		fmd_asru_do_hash_delete(ahp, cis, cp, ahp->ah_fru_hash,
1160 		    offsetof(fmd_asru_link_t, al_fru_next), FM_FAULT_FRU);
1161 		fmd_asru_do_hash_delete(ahp, cis, cp, ahp->ah_rsrc_hash,
1162 		    offsetof(fmd_asru_link_t, al_rsrc_next), FM_FAULT_RESOURCE);
1163 		if (nvlist_lookup_string(cis->cis_nvl, FM_FAULT_LOCATION,
1164 		    &label) != 0)
1165 			label = "";
1166 		fmd_asru_do_delete_entry(ahp, cp, ahp->ah_label_hash,
1167 		    offsetof(fmd_asru_link_t, al_label_next), label);
1168 		fmd_asru_do_hash_delete(ahp, cis, cp, ahp->ah_asru_hash,
1169 		    offsetof(fmd_asru_link_t, al_asru_next), FM_FAULT_ASRU);
1170 	}
1171 
1172 	/*
1173 	 * then delete associated case hash entries
1174 	 */
1175 	(void) pthread_rwlock_wrlock(&ahp->ah_lock);
1176 	h = fmd_asru_strhash(ahp, cip->ci_uuid);
1177 	plp = &ahp->ah_case_hash[h];
1178 	for (alp = *plp; alp != NULL; alp = alpnext) {
1179 		alpnext = alp->al_case_next;
1180 		if (alp->al_case == cp) {
1181 			*plp = alp->al_case_next;
1182 			alp->al_case_next = NULL;
1183 			ASSERT(ahp->ah_al_count != 0);
1184 			ahp->ah_al_count--;
1185 
1186 			/*
1187 			 * decrement case ref.
1188 			 */
1189 			fmd_case_rele_locked(cp);
1190 			alp->al_case = NULL;
1191 
1192 			/*
1193 			 * If we found a matching ASRU, unlink its log file and
1194 			 * then release the hash entry. Note that it may still
1195 			 * be referenced if another thread is manipulating it;
1196 			 * this is ok because once we unlink, the log file will
1197 			 * not be restored, and the log data will be freed when
1198 			 * all of the referencing threads release their
1199 			 * respective references.
1200 			 */
1201 			(void) snprintf(path, sizeof (path), "%s/%s",
1202 			    ahp->ah_dirpath, alp->al_uuid);
1203 			if (cip->ci_xprt == NULL && unlink(path) != 0)
1204 				fmd_error(EFMD_ASRU_UNLINK,
1205 				    "failed to unlink asru %s", path);
1206 
1207 			/*
1208 			 * Now unlink from the global per-resource cache
1209 			 * and if this is the last link then remove that from
1210 			 * it's own hash too.
1211 			 */
1212 			ap = alp->al_asru;
1213 			(void) pthread_mutex_lock(&ap->asru_lock);
1214 			fmd_list_delete(&ap->asru_list, alp);
1215 			if (ap->asru_list.l_next == NULL) {
1216 				uint_t h;
1217 				fmd_asru_t *ap2, **pp;
1218 				fmd_asru_t *apnext, **apnextp;
1219 
1220 				ASSERT(ahp->ah_count != 0);
1221 				ahp->ah_count--;
1222 				h = fmd_asru_strhash(ahp, ap->asru_name);
1223 				pp = &ahp->ah_hash[h];
1224 				for (ap2 = *pp; ap2 != NULL; ap2 = apnext) {
1225 					apnextp = &ap2->asru_next;
1226 					apnext = *apnextp;
1227 					if (ap2 == ap) {
1228 						*pp = *apnextp;
1229 						*apnextp = NULL;
1230 					} else
1231 						pp = apnextp;
1232 				}
1233 			}
1234 			(void) pthread_mutex_unlock(&ap->asru_lock);
1235 			fmd_asru_al_hash_release(ahp, alp);
1236 		} else
1237 			plp = &alp->al_case_next;
1238 	}
1239 	(void) pthread_rwlock_unlock(&ahp->ah_lock);
1240 }
1241 
1242 typedef struct {
1243 	nvlist_t *farc_parent_fmri;
1244 	uint8_t farc_reason;
1245 } fmd_asru_farc_t;
1246 
1247 static void
fmd_asru_repair_containee(fmd_asru_link_t * alp,void * arg)1248 fmd_asru_repair_containee(fmd_asru_link_t *alp, void *arg)
1249 {
1250 	fmd_asru_farc_t *farcp = (fmd_asru_farc_t *)arg;
1251 
1252 	if ((alp->al_asru->asru_flags & FMD_ASRU_INVISIBLE) &&
1253 	    alp->al_asru_fmri &&
1254 	    fmd_fmri_contains(farcp->farc_parent_fmri, alp->al_asru_fmri) > 0) {
1255 		if (fmd_asru_clrflags(alp, FMD_ASRU_FAULTY,
1256 		    farcp->farc_reason)) {
1257 			if (alp->al_flags & FMD_ASRU_PROXY)
1258 				fmd_case_xprt_updated(alp->al_case);
1259 			else
1260 				fmd_case_update(alp->al_case);
1261 		}
1262 	}
1263 }
1264 
1265 static void
fmd_asru_do_repair_containees(fmd_asru_link_t * alp,uint8_t reason)1266 fmd_asru_do_repair_containees(fmd_asru_link_t *alp, uint8_t reason)
1267 {
1268 	int flags;
1269 
1270 	/*
1271 	 * Check if all entries associated with this asru are acquitted and
1272 	 * if so acquit containees. Don't try to repair containees on proxy
1273 	 * side unless we have local asru.
1274 	 */
1275 	if (alp->al_asru_fmri != NULL && (!(alp->al_flags & FMD_ASRU_PROXY) ||
1276 	    (alp->al_flags & FMD_ASRU_PROXY_WITH_ASRU))) {
1277 		(void) pthread_mutex_lock(&alp->al_asru->asru_lock);
1278 		flags = alp->al_asru->asru_flags;
1279 		(void) pthread_mutex_unlock(&alp->al_asru->asru_lock);
1280 		if (!(flags & (FMD_ASRU_FAULTY | FMD_ASRU_INVISIBLE))) {
1281 			fmd_asru_farc_t farc;
1282 
1283 			farc.farc_parent_fmri = alp->al_asru_fmri;
1284 			farc.farc_reason = reason;
1285 			fmd_asru_al_hash_apply(fmd.d_asrus,
1286 			    fmd_asru_repair_containee, &farc);
1287 		}
1288 	}
1289 }
1290 
1291 void
fmd_asru_repaired(fmd_asru_link_t * alp,void * arg)1292 fmd_asru_repaired(fmd_asru_link_t *alp, void *arg)
1293 {
1294 	int cleared;
1295 	fmd_asru_rep_arg_t *farap = (fmd_asru_rep_arg_t *)arg;
1296 
1297 	/*
1298 	 * don't allow remote repair over readonly transport
1299 	 */
1300 	if (alp->al_flags & FMD_ASRU_PROXY_RDONLY)
1301 		return;
1302 
1303 	/*
1304 	 * don't allow repair etc by asru on proxy unless asru is local
1305 	 */
1306 	if (farap->fara_bywhat == FARA_BY_ASRU &&
1307 	    (alp->al_flags & FMD_ASRU_PROXY) &&
1308 	    !(alp->al_flags & FMD_ASRU_PROXY_WITH_ASRU))
1309 		return;
1310 	/*
1311 	 * For acquit, need to check both name and uuid if specified
1312 	 */
1313 	if (farap->fara_reason == FMD_ASRU_ACQUITTED &&
1314 	    farap->fara_rval != NULL && strcmp(farap->fara_uuid, "") != 0 &&
1315 	    strcmp(farap->fara_uuid, alp->al_case_uuid) != 0)
1316 		return;
1317 
1318 	/*
1319 	 * For replaced, verify it has been replaced if we have serial number.
1320 	 * If not set *farap->fara_rval to FARA_ERR_RSRCNOTR.
1321 	 */
1322 	if (farap->fara_reason == FMD_ASRU_REPLACED &&
1323 	    !(alp->al_flags & FMD_ASRU_PROXY_EXTERNAL) &&
1324 	    fmd_asru_replacement_state(alp->al_event,
1325 	    (alp->al_flags & FMD_ASRU_PROXY) ? HC_ONLY_TRUE : HC_ONLY_FALSE) ==
1326 	    FMD_OBJ_STATE_STILL_PRESENT) {
1327 		if (farap->fara_rval)
1328 			*farap->fara_rval = FARA_ERR_RSRCNOTR;
1329 		return;
1330 	}
1331 
1332 	cleared = fmd_asru_clrflags(alp, FMD_ASRU_FAULTY, farap->fara_reason);
1333 	fmd_asru_do_repair_containees(alp, farap->fara_reason);
1334 
1335 	/*
1336 	 * if called from fmd_adm_*() and we really did clear the bit then
1337 	 * we need to do a case update to see if the associated case can be
1338 	 * repaired. No need to do this if called from fmd_case_*() (ie
1339 	 * when arg is NULL) as the case will be explicitly repaired anyway.
1340 	 */
1341 	if (farap->fara_rval) {
1342 		/*
1343 		 * *farap->fara_rval defaults to FARA_ERR_RSRCNOTF (not found).
1344 		 * If we find a valid cache entry which we repair then we
1345 		 * set it to FARA_OK. However we don't want to do this if
1346 		 * we have already set it to FARA_ERR_RSRCNOTR (not replaced)
1347 		 * in a previous iteration (see above). So only set it to
1348 		 * FARA_OK if the current value is still FARA_ERR_RSRCNOTF.
1349 		 */
1350 		if (*farap->fara_rval == FARA_ERR_RSRCNOTF)
1351 			*farap->fara_rval = FARA_OK;
1352 		if (cleared) {
1353 			if (alp->al_flags & FMD_ASRU_PROXY)
1354 				fmd_case_xprt_updated(alp->al_case);
1355 			else
1356 				fmd_case_update(alp->al_case);
1357 		}
1358 	}
1359 }
1360 
1361 /*
1362  * Discard the case associated with this alp if it is in resolved state.
1363  * Called on "fmadm flush".
1364  */
1365 /*ARGSUSED*/
1366 void
fmd_asru_flush(fmd_asru_link_t * alp,void * arg)1367 fmd_asru_flush(fmd_asru_link_t *alp, void *arg)
1368 {
1369 	int check_if_aged = 0;
1370 	int *rval = (int *)arg;
1371 
1372 	if (alp->al_case)
1373 		fmd_case_discard_resolved(alp->al_case, &check_if_aged);
1374 	*rval = 0;
1375 }
1376 
1377 /*
1378  * This is only called for proxied faults. Set various flags so we can
1379  * find the nature of the transport from the resource cache code.
1380  */
1381 /*ARGSUSED*/
1382 void
fmd_asru_set_on_proxy(fmd_asru_link_t * alp,void * arg)1383 fmd_asru_set_on_proxy(fmd_asru_link_t *alp, void *arg)
1384 {
1385 	fmd_asru_set_on_proxy_t *entryp = (fmd_asru_set_on_proxy_t *)arg;
1386 
1387 	if (*entryp->fasp_countp >= entryp->fasp_maxcount)
1388 		return;
1389 
1390 	/*
1391 	 * Note that this is a proxy fault and save whetehr transport is
1392 	 * RDONLY or EXTERNAL.
1393 	 */
1394 	alp->al_flags |= FMD_ASRU_PROXY;
1395 	alp->al_asru->asru_flags |= FMD_ASRU_PROXY;
1396 
1397 	if (entryp->fasp_proxy_external) {
1398 		alp->al_flags |= FMD_ASRU_PROXY_EXTERNAL;
1399 		alp->al_asru->asru_flags |= FMD_ASRU_PROXY_EXTERNAL;
1400 	}
1401 
1402 	if (entryp->fasp_proxy_rdonly)
1403 		alp->al_flags |= FMD_ASRU_PROXY_RDONLY;
1404 
1405 	/*
1406 	 * Save whether asru is accessible in local domain
1407 	 */
1408 	if (entryp->fasp_proxy_asru[*entryp->fasp_countp]) {
1409 		alp->al_flags |= FMD_ASRU_PROXY_WITH_ASRU;
1410 		alp->al_asru->asru_flags |= FMD_ASRU_PROXY_WITH_ASRU;
1411 	}
1412 	(*entryp->fasp_countp)++;
1413 }
1414 
1415 /*ARGSUSED*/
1416 void
fmd_asru_update_containees(fmd_asru_link_t * alp,void * arg)1417 fmd_asru_update_containees(fmd_asru_link_t *alp, void *arg)
1418 {
1419 	fmd_asru_do_repair_containees(alp, alp->al_reason);
1420 }
1421 
1422 /*
1423  * This function is used for fault proxying. It updates the resource status in
1424  * the resource cache based on information that has come from the other side of
1425  * the transport. This can be called on either the proxy side or the
1426  * diagnosing side.
1427  */
1428 void
fmd_asru_update_status(fmd_asru_link_t * alp,void * arg)1429 fmd_asru_update_status(fmd_asru_link_t *alp, void *arg)
1430 {
1431 	fmd_asru_update_status_t *entryp = (fmd_asru_update_status_t *)arg;
1432 	uint8_t status;
1433 
1434 	if (*entryp->faus_countp >= entryp->faus_maxcount)
1435 		return;
1436 
1437 	status = entryp->faus_ba[*entryp->faus_countp];
1438 
1439 	/*
1440 	 * For proxy, if there is no asru on the proxy side, but there is on
1441 	 * the diag side, then take the diag side asru status.
1442 	 * For diag, if there is an asru on the proxy side, then take the proxy
1443 	 * side asru status.
1444 	 */
1445 	if (entryp->faus_is_proxy ?
1446 	    (entryp->faus_diag_asru[*entryp->faus_countp] &&
1447 	    !entryp->faus_proxy_asru[*entryp->faus_countp]) :
1448 	    entryp->faus_proxy_asru[*entryp->faus_countp]) {
1449 		if (status & FM_SUSPECT_DEGRADED)
1450 			alp->al_flags |= FMD_ASRU_DEGRADED;
1451 		else
1452 			alp->al_flags &= ~FMD_ASRU_DEGRADED;
1453 		if (status & FM_SUSPECT_UNUSABLE)
1454 			(void) fmd_asru_setflags(alp, FMD_ASRU_UNUSABLE);
1455 		else
1456 			(void) fmd_asru_clrflags(alp, FMD_ASRU_UNUSABLE, 0);
1457 	}
1458 
1459 	/*
1460 	 * Update the faulty status too.
1461 	 */
1462 	if (!(status & FM_SUSPECT_FAULTY))
1463 		(void) fmd_asru_clrflags(alp, FMD_ASRU_FAULTY,
1464 		    (status & FM_SUSPECT_REPAIRED) ? FMD_ASRU_REPAIRED :
1465 		    (status & FM_SUSPECT_REPLACED) ? FMD_ASRU_REPLACED :
1466 		    (status & FM_SUSPECT_ACQUITTED) ? FMD_ASRU_ACQUITTED :
1467 		    FMD_ASRU_REMOVED);
1468 	else if (entryp->faus_is_proxy)
1469 		(void) fmd_asru_setflags(alp, FMD_ASRU_FAULTY);
1470 
1471 	/*
1472 	 * for proxy only, update the present status too.
1473 	 */
1474 	if (entryp->faus_is_proxy) {
1475 		if (!(status & FM_SUSPECT_NOT_PRESENT)) {
1476 			alp->al_flags |= FMD_ASRU_PRESENT;
1477 			alp->al_asru->asru_flags |= FMD_ASRU_PRESENT;
1478 		} else {
1479 			alp->al_flags &= ~FMD_ASRU_PRESENT;
1480 			alp->al_asru->asru_flags &= ~FMD_ASRU_PRESENT;
1481 		}
1482 	}
1483 	(*entryp->faus_countp)++;
1484 }
1485 
1486 /*
1487  * This function is called on the diagnosing side when fault proxying is
1488  * in use and the proxy has sent a uuclose. It updates the status of the
1489  * resource cache entries.
1490  */
1491 void
fmd_asru_close_status(fmd_asru_link_t * alp,void * arg)1492 fmd_asru_close_status(fmd_asru_link_t *alp, void *arg)
1493 {
1494 	fmd_asru_close_status_t *entryp = (fmd_asru_close_status_t *)arg;
1495 
1496 	if (*entryp->facs_countp >= entryp->facs_maxcount)
1497 		return;
1498 	alp->al_flags &= ~FMD_ASRU_DEGRADED;
1499 	(void) fmd_asru_setflags(alp, FMD_ASRU_UNUSABLE);
1500 	(*entryp->facs_countp)++;
1501 }
1502 
1503 static void
fmd_asru_logevent(fmd_asru_link_t * alp)1504 fmd_asru_logevent(fmd_asru_link_t *alp)
1505 {
1506 	fmd_asru_t *ap = alp->al_asru;
1507 	boolean_t faulty = (alp->al_flags & FMD_ASRU_FAULTY) != 0;
1508 	boolean_t unusable = (alp->al_flags & FMD_ASRU_UNUSABLE) != 0;
1509 	boolean_t message = (ap->asru_flags & FMD_ASRU_INVISIBLE) == 0;
1510 	boolean_t repaired = (alp->al_reason == FMD_ASRU_REPAIRED);
1511 	boolean_t replaced = (alp->al_reason == FMD_ASRU_REPLACED);
1512 	boolean_t acquitted = (alp->al_reason == FMD_ASRU_ACQUITTED);
1513 
1514 	fmd_case_impl_t *cip;
1515 	fmd_event_t *e;
1516 	fmd_log_t *lp;
1517 	nvlist_t *nvl;
1518 	char *class;
1519 
1520 	ASSERT(MUTEX_HELD(&ap->asru_lock));
1521 	cip = (fmd_case_impl_t *)alp->al_case;
1522 	ASSERT(cip != NULL);
1523 
1524 	/*
1525 	 * Don't log to disk on proxy side
1526 	 */
1527 	if (cip->ci_xprt != NULL)
1528 		return;
1529 
1530 	if ((lp = alp->al_log) == NULL)
1531 		lp = fmd_log_open(ap->asru_root, alp->al_uuid, FMD_LOG_ASRU);
1532 
1533 	if (lp == NULL)
1534 		return; /* can't log events if we can't open the log */
1535 
1536 	nvl = fmd_protocol_rsrc_asru(_fmd_asru_events[faulty | (unusable << 1)],
1537 	    alp->al_asru_fmri, cip->ci_uuid, cip->ci_code, faulty, unusable,
1538 	    message, alp->al_event, &cip->ci_tv, repaired, replaced, acquitted,
1539 	    cip->ci_state == FMD_CASE_RESOLVED, cip->ci_diag_de == NULL ?
1540 	    cip->ci_mod->mod_fmri : cip->ci_diag_de, cip->ci_injected == 1);
1541 
1542 	(void) nvlist_lookup_string(nvl, FM_CLASS, &class);
1543 	e = fmd_event_create(FMD_EVT_PROTOCOL, FMD_HRT_NOW, nvl, class);
1544 
1545 	fmd_event_hold(e);
1546 	fmd_log_append(lp, e, NULL);
1547 	fmd_event_rele(e);
1548 
1549 	/*
1550 	 * For now, we close the log file after every update to conserve file
1551 	 * descriptors and daemon overhead.  If this becomes a performance
1552 	 * issue this code can change to keep a fixed-size LRU cache of logs.
1553 	 */
1554 	fmd_log_rele(lp);
1555 	alp->al_log = NULL;
1556 }
1557 
1558 int
fmd_asru_setflags(fmd_asru_link_t * alp,uint_t sflag)1559 fmd_asru_setflags(fmd_asru_link_t *alp, uint_t sflag)
1560 {
1561 	fmd_asru_t *ap = alp->al_asru;
1562 	uint_t nstate, ostate;
1563 
1564 	ASSERT(!(sflag & ~FMD_ASRU_STATE));
1565 	ASSERT(sflag != FMD_ASRU_STATE);
1566 
1567 	(void) pthread_mutex_lock(&ap->asru_lock);
1568 
1569 	ostate = alp->al_flags & FMD_ASRU_STATE;
1570 	alp->al_flags |= sflag;
1571 	nstate = alp->al_flags & FMD_ASRU_STATE;
1572 
1573 	if (nstate == ostate) {
1574 		(void) pthread_mutex_unlock(&ap->asru_lock);
1575 		return (0);
1576 	}
1577 
1578 	ap->asru_flags |= sflag;
1579 	TRACE((FMD_DBG_ASRU, "asru %s %s->%s", alp->al_uuid,
1580 	    _fmd_asru_snames[ostate], _fmd_asru_snames[nstate]));
1581 
1582 	fmd_asru_logevent(alp);
1583 
1584 	(void) pthread_cond_broadcast(&ap->asru_cv);
1585 	(void) pthread_mutex_unlock(&ap->asru_lock);
1586 	return (1);
1587 }
1588 
1589 int
fmd_asru_clrflags(fmd_asru_link_t * alp,uint_t sflag,uint8_t reason)1590 fmd_asru_clrflags(fmd_asru_link_t *alp, uint_t sflag, uint8_t reason)
1591 {
1592 	fmd_asru_t *ap = alp->al_asru;
1593 	fmd_asru_link_t *nalp;
1594 	uint_t nstate, ostate, flags = 0;
1595 
1596 	ASSERT(!(sflag & ~FMD_ASRU_STATE));
1597 	ASSERT(sflag != FMD_ASRU_STATE);
1598 
1599 	(void) pthread_mutex_lock(&ap->asru_lock);
1600 
1601 	ostate = alp->al_flags & FMD_ASRU_STATE;
1602 	alp->al_flags &= ~sflag;
1603 	nstate = alp->al_flags & FMD_ASRU_STATE;
1604 
1605 	if (nstate == ostate) {
1606 		if (reason > alp->al_reason &&
1607 		    ((fmd_case_impl_t *)alp->al_case)->ci_state <
1608 		    FMD_CASE_REPAIRED) {
1609 			alp->al_reason = reason;
1610 			fmd_asru_logevent(alp);
1611 			(void) pthread_cond_broadcast(&ap->asru_cv);
1612 		}
1613 		(void) pthread_mutex_unlock(&ap->asru_lock);
1614 		return (0);
1615 	}
1616 	if (reason > alp->al_reason)
1617 		alp->al_reason = reason;
1618 
1619 	if (sflag == FMD_ASRU_UNUSABLE)
1620 		ap->asru_flags &= ~sflag;
1621 	else if (sflag == FMD_ASRU_FAULTY) {
1622 		/*
1623 		 * only clear the faulty bit if all links are clear
1624 		 */
1625 		for (nalp = fmd_list_next(&ap->asru_list); nalp != NULL;
1626 		    nalp = fmd_list_next(nalp))
1627 			flags |= nalp->al_flags;
1628 		if (!(flags & FMD_ASRU_FAULTY))
1629 			ap->asru_flags &= ~sflag;
1630 	}
1631 
1632 	TRACE((FMD_DBG_ASRU, "asru %s %s->%s", alp->al_uuid,
1633 	    _fmd_asru_snames[ostate], _fmd_asru_snames[nstate]));
1634 
1635 	fmd_asru_logevent(alp);
1636 
1637 	(void) pthread_cond_broadcast(&ap->asru_cv);
1638 	(void) pthread_mutex_unlock(&ap->asru_lock);
1639 
1640 	return (1);
1641 }
1642 
1643 /*ARGSUSED*/
1644 void
fmd_asru_log_resolved(fmd_asru_link_t * alp,void * unused)1645 fmd_asru_log_resolved(fmd_asru_link_t *alp, void *unused)
1646 {
1647 	fmd_asru_t *ap = alp->al_asru;
1648 
1649 	(void) pthread_mutex_lock(&ap->asru_lock);
1650 	fmd_asru_logevent(alp);
1651 	(void) pthread_cond_broadcast(&ap->asru_cv);
1652 	(void) pthread_mutex_unlock(&ap->asru_lock);
1653 }
1654 
1655 /*
1656  * Report the current known state of the link entry (ie this particular fault
1657  * affecting this particular ASRU).
1658  */
1659 int
fmd_asru_al_getstate(fmd_asru_link_t * alp)1660 fmd_asru_al_getstate(fmd_asru_link_t *alp)
1661 {
1662 	int us, st = (alp->al_flags & (FMD_ASRU_FAULTY | FMD_ASRU_UNUSABLE));
1663 	nvlist_t *asru;
1664 	int ps = FMD_OBJ_STATE_UNKNOWN;
1665 
1666 	/*
1667 	 * For fault proxying with an EXTERNAL transport, believe the presence
1668 	 * state as sent by the diagnosing side. Otherwise find the presence
1669 	 * state here. Note that if fault proxying with an INTERNAL transport
1670 	 * we can only trust the presence state where we are using hc-scheme
1671 	 * fmris which should be consistant across domains in the same system -
1672 	 * other schemes can refer to different devices in different domains.
1673 	 */
1674 	if (!(alp->al_flags & FMD_ASRU_PROXY_EXTERNAL)) {
1675 		ps = fmd_asru_replacement_state(alp->al_event, (alp->al_flags &
1676 		    FMD_ASRU_PROXY)? HC_ONLY_TRUE : HC_ONLY_FALSE);
1677 		if (ps == FMD_OBJ_STATE_NOT_PRESENT)
1678 			return (st | FMD_ASRU_UNUSABLE);
1679 		if (ps == FMD_OBJ_STATE_REPLACED) {
1680 			if (alp->al_reason < FMD_ASRU_REPLACED)
1681 				alp->al_reason = FMD_ASRU_REPLACED;
1682 			return (st | FMD_ASRU_UNUSABLE);
1683 		}
1684 	}
1685 	if (ps == FMD_OBJ_STATE_UNKNOWN && (alp->al_flags & FMD_ASRU_PROXY))
1686 		st |= (alp->al_flags & (FMD_ASRU_DEGRADED | FMD_ASRU_PRESENT));
1687 	else
1688 		st |= (alp->al_flags & (FMD_ASRU_DEGRADED)) | FMD_ASRU_PRESENT;
1689 
1690 	/*
1691 	 * For fault proxying, unless we have a local ASRU, then believe the
1692 	 * service state sent by the diagnosing side. Otherwise find the service
1693 	 * state here. Try fmd_fmri_service_state() first, but if that's not
1694 	 * supported by the scheme then fall back to fmd_fmri_unusable().
1695 	 */
1696 	if ((!(alp->al_flags & FMD_ASRU_PROXY) ||
1697 	    (alp->al_flags & FMD_ASRU_PROXY_WITH_ASRU)) &&
1698 	    nvlist_lookup_nvlist(alp->al_event, FM_FAULT_ASRU, &asru) == 0) {
1699 		us = fmd_fmri_service_state(asru);
1700 		if (us == -1 || us == FMD_SERVICE_STATE_UNKNOWN) {
1701 			/* not supported by scheme - try fmd_fmri_unusable */
1702 			us = fmd_fmri_unusable(asru);
1703 			if (us > 0)
1704 				st |= FMD_ASRU_UNUSABLE;
1705 			else if (us == 0)
1706 				st &= ~FMD_ASRU_UNUSABLE;
1707 		} else {
1708 			if (us == FMD_SERVICE_STATE_UNUSABLE) {
1709 				st &= ~FMD_ASRU_DEGRADED;
1710 				st |= FMD_ASRU_UNUSABLE;
1711 			} else if (us == FMD_SERVICE_STATE_OK) {
1712 				st &= ~(FMD_ASRU_DEGRADED | FMD_ASRU_UNUSABLE);
1713 			} else if (us == FMD_SERVICE_STATE_ISOLATE_PENDING) {
1714 				st &= ~(FMD_ASRU_DEGRADED | FMD_ASRU_UNUSABLE);
1715 			} else if (us == FMD_SERVICE_STATE_DEGRADED) {
1716 				st &= ~FMD_ASRU_UNUSABLE;
1717 				st |= FMD_ASRU_DEGRADED;
1718 			}
1719 		}
1720 	}
1721 	return (st);
1722 }
1723 
1724 /*
1725  * Report the current known state of the ASRU by refreshing its unusable status
1726  * based upon the routines provided by the scheme module.  If the unusable bit
1727  * is different, we do *not* generate a state change here because that change
1728  * may be unrelated to fmd activities and therefore we have no case or event.
1729  * The absence of the transition is harmless as this function is only provided
1730  * for RPC observability and fmd's clients are only concerned with ASRU_FAULTY.
1731  */
1732 int
fmd_asru_getstate(fmd_asru_t * ap)1733 fmd_asru_getstate(fmd_asru_t *ap)
1734 {
1735 	int us, st, p = -1;
1736 	char *s;
1737 
1738 	/* do not report non-fmd non-present resources */
1739 	if (!(ap->asru_flags & FMD_ASRU_INTERNAL)) {
1740 		/*
1741 		 * As with fmd_asru_al_getstate(), we can only trust the
1742 		 * local presence state on a proxy if the transport is
1743 		 * internal and the scheme is hc. Otherwise we believe the
1744 		 * state as sent by the diagnosing side.
1745 		 */
1746 		if (!(ap->asru_flags & FMD_ASRU_PROXY) ||
1747 		    (!(ap->asru_flags & FMD_ASRU_PROXY_EXTERNAL) &&
1748 		    (nvlist_lookup_string(ap->asru_fmri, FM_FMRI_SCHEME,
1749 		    &s) == 0 && strcmp(s, FM_FMRI_SCHEME_HC) == 0))) {
1750 			if (fmd_asru_fake_not_present >=
1751 			    FMD_OBJ_STATE_REPLACED)
1752 				return (0);
1753 			p = fmd_fmri_present(ap->asru_fmri);
1754 		}
1755 		if (p == 0 || (p < 0 && !(ap->asru_flags & FMD_ASRU_PROXY) ||
1756 		    !(ap->asru_flags & FMD_ASRU_PRESENT)))
1757 			return (0);
1758 	}
1759 
1760 	/*
1761 	 * As with fmd_asru_al_getstate(), we can only trust the local unusable
1762 	 * state on a proxy if there is a local ASRU.
1763 	 */
1764 	st = ap->asru_flags & (FMD_ASRU_FAULTY | FMD_ASRU_UNUSABLE);
1765 	if (!(ap->asru_flags & FMD_ASRU_PROXY) ||
1766 	    (ap->asru_flags & FMD_ASRU_PROXY_WITH_ASRU)) {
1767 		us = fmd_fmri_unusable(ap->asru_fmri);
1768 		if (us > 0)
1769 			st |= FMD_ASRU_UNUSABLE;
1770 		else if (us == 0)
1771 			st &= ~FMD_ASRU_UNUSABLE;
1772 	}
1773 	return (st);
1774 }
1775