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 https://opensource.org/licenses/CDDL-1.0.
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 (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
23 *
24 * Copyright (c) 2016, Intel Corporation.
25 * Copyright (c) 2023, Klara Inc.
26 */
27
28 /*
29 * This file implements the minimal FMD module API required to support the
30 * fault logic modules in ZED. This support includes module registration,
31 * memory allocation, module property accessors, basic case management,
32 * one-shot timers and SERD engines.
33 *
34 * In the ZED runtime, the modules are called from a single thread so no
35 * locking is required in this emulated FMD environment.
36 */
37
38 #include <sys/types.h>
39 #include <sys/fm/protocol.h>
40 #include <uuid/uuid.h>
41 #include <signal.h>
42 #include <string.h>
43 #include <time.h>
44
45 #include "fmd_api.h"
46 #include "fmd_serd.h"
47
48 #include "zfs_agents.h"
49 #include "../zed_log.h"
50
51 typedef struct fmd_modstat {
52 fmd_stat_t ms_accepted; /* total events accepted by module */
53 fmd_stat_t ms_caseopen; /* cases currently open */
54 fmd_stat_t ms_casesolved; /* total cases solved by module */
55 fmd_stat_t ms_caseclosed; /* total cases closed by module */
56 } fmd_modstat_t;
57
58 typedef struct fmd_module {
59 const char *mod_name; /* basename of module (ro) */
60 const fmd_hdl_info_t *mod_info; /* module info registered with handle */
61 void *mod_spec; /* fmd_hdl_get/setspecific data value */
62 fmd_stat_t *mod_ustat; /* module specific custom stats */
63 uint_t mod_ustat_cnt; /* count of ustat stats */
64 fmd_modstat_t mod_stats; /* fmd built-in per-module statistics */
65 fmd_serd_hash_t mod_serds; /* hash of serd engs owned by module */
66 char *mod_vers; /* a copy of module version string */
67 } fmd_module_t;
68
69 /*
70 * ZED has two FMD hardwired module instances
71 */
72 fmd_module_t zfs_retire_module;
73 fmd_module_t zfs_diagnosis_module;
74
75 /*
76 * Enable a reasonable set of defaults for libumem debugging on DEBUG builds.
77 */
78
79 #ifdef DEBUG
80 const char *
_umem_debug_init(void)81 _umem_debug_init(void)
82 {
83 return ("default,verbose"); /* $UMEM_DEBUG setting */
84 }
85
86 const char *
_umem_logging_init(void)87 _umem_logging_init(void)
88 {
89 return ("fail,contents"); /* $UMEM_LOGGING setting */
90 }
91 #endif
92
93 /*
94 * Register a module with fmd and finish module initialization.
95 * Returns an integer indicating whether it succeeded (zero) or
96 * failed (non-zero).
97 */
98 int
fmd_hdl_register(fmd_hdl_t * hdl,int version,const fmd_hdl_info_t * mip)99 fmd_hdl_register(fmd_hdl_t *hdl, int version, const fmd_hdl_info_t *mip)
100 {
101 (void) version;
102 fmd_module_t *mp = (fmd_module_t *)hdl;
103
104 mp->mod_info = mip;
105 mp->mod_name = mip->fmdi_desc + 4; /* drop 'ZFS ' prefix */
106 mp->mod_spec = NULL;
107
108 /* bare minimum module stats */
109 (void) strcpy(mp->mod_stats.ms_accepted.fmds_name, "fmd.accepted");
110 (void) strcpy(mp->mod_stats.ms_caseopen.fmds_name, "fmd.caseopen");
111 (void) strcpy(mp->mod_stats.ms_casesolved.fmds_name, "fmd.casesolved");
112 (void) strcpy(mp->mod_stats.ms_caseclosed.fmds_name, "fmd.caseclosed");
113
114 fmd_serd_hash_create(&mp->mod_serds);
115
116 fmd_hdl_debug(hdl, "register module");
117
118 return (0);
119 }
120
121 void
fmd_hdl_unregister(fmd_hdl_t * hdl)122 fmd_hdl_unregister(fmd_hdl_t *hdl)
123 {
124 fmd_module_t *mp = (fmd_module_t *)hdl;
125 fmd_modstat_t *msp = &mp->mod_stats;
126 const fmd_hdl_ops_t *ops = mp->mod_info->fmdi_ops;
127
128 /* dump generic module stats */
129 fmd_hdl_debug(hdl, "%s: %llu", msp->ms_accepted.fmds_name,
130 msp->ms_accepted.fmds_value.ui64);
131 if (ops->fmdo_close != NULL) {
132 fmd_hdl_debug(hdl, "%s: %llu", msp->ms_caseopen.fmds_name,
133 msp->ms_caseopen.fmds_value.ui64);
134 fmd_hdl_debug(hdl, "%s: %llu", msp->ms_casesolved.fmds_name,
135 msp->ms_casesolved.fmds_value.ui64);
136 fmd_hdl_debug(hdl, "%s: %llu", msp->ms_caseclosed.fmds_name,
137 msp->ms_caseclosed.fmds_value.ui64);
138 }
139
140 /* dump module specific stats */
141 if (mp->mod_ustat != NULL) {
142 int i;
143
144 for (i = 0; i < mp->mod_ustat_cnt; i++) {
145 fmd_hdl_debug(hdl, "%s: %llu",
146 mp->mod_ustat[i].fmds_name,
147 mp->mod_ustat[i].fmds_value.ui64);
148 }
149 }
150
151 fmd_serd_hash_destroy(&mp->mod_serds);
152
153 fmd_hdl_debug(hdl, "unregister module");
154 }
155
156 /*
157 * fmd_hdl_setspecific() is used to associate a data pointer with
158 * the specified handle for the duration of the module's lifetime.
159 * This pointer can be retrieved using fmd_hdl_getspecific().
160 */
161 void
fmd_hdl_setspecific(fmd_hdl_t * hdl,void * spec)162 fmd_hdl_setspecific(fmd_hdl_t *hdl, void *spec)
163 {
164 fmd_module_t *mp = (fmd_module_t *)hdl;
165
166 mp->mod_spec = spec;
167 }
168
169 /*
170 * Return the module-specific data pointer previously associated
171 * with the handle using fmd_hdl_setspecific().
172 */
173 void *
fmd_hdl_getspecific(fmd_hdl_t * hdl)174 fmd_hdl_getspecific(fmd_hdl_t *hdl)
175 {
176 fmd_module_t *mp = (fmd_module_t *)hdl;
177
178 return (mp->mod_spec);
179 }
180
181 void *
fmd_hdl_alloc(fmd_hdl_t * hdl,size_t size,int flags)182 fmd_hdl_alloc(fmd_hdl_t *hdl, size_t size, int flags)
183 {
184 (void) hdl;
185 return (umem_alloc(size, flags));
186 }
187
188 void *
fmd_hdl_zalloc(fmd_hdl_t * hdl,size_t size,int flags)189 fmd_hdl_zalloc(fmd_hdl_t *hdl, size_t size, int flags)
190 {
191 (void) hdl;
192 return (umem_zalloc(size, flags));
193 }
194
195 void
fmd_hdl_free(fmd_hdl_t * hdl,void * data,size_t size)196 fmd_hdl_free(fmd_hdl_t *hdl, void *data, size_t size)
197 {
198 (void) hdl;
199 umem_free(data, size);
200 }
201
202 /*
203 * Record a module debug message using the specified format.
204 */
205 void
fmd_hdl_debug(fmd_hdl_t * hdl,const char * format,...)206 fmd_hdl_debug(fmd_hdl_t *hdl, const char *format, ...)
207 {
208 char message[256];
209 va_list vargs;
210 fmd_module_t *mp = (fmd_module_t *)hdl;
211
212 va_start(vargs, format);
213 (void) vsnprintf(message, sizeof (message), format, vargs);
214 va_end(vargs);
215
216 /* prefix message with module name */
217 zed_log_msg(LOG_INFO, "%s: %s", mp->mod_name, message);
218 }
219
220 /* Property Retrieval */
221
222 int32_t
fmd_prop_get_int32(fmd_hdl_t * hdl,const char * name)223 fmd_prop_get_int32(fmd_hdl_t *hdl, const char *name)
224 {
225 (void) hdl;
226
227 /*
228 * These can be looked up in mp->modinfo->fmdi_props
229 * For now we just hard code for phase 2. In the
230 * future, there can be a ZED based override.
231 */
232 if (strcmp(name, "spare_on_remove") == 0)
233 return (1);
234
235 return (0);
236 }
237
238 /* FMD Statistics */
239
240 fmd_stat_t *
fmd_stat_create(fmd_hdl_t * hdl,uint_t flags,uint_t nstats,fmd_stat_t * statv)241 fmd_stat_create(fmd_hdl_t *hdl, uint_t flags, uint_t nstats, fmd_stat_t *statv)
242 {
243 fmd_module_t *mp = (fmd_module_t *)hdl;
244
245 if (flags == FMD_STAT_NOALLOC) {
246 mp->mod_ustat = statv;
247 mp->mod_ustat_cnt = nstats;
248 }
249
250 return (statv);
251 }
252
253 /* Case Management */
254
255 fmd_case_t *
fmd_case_open(fmd_hdl_t * hdl,void * data)256 fmd_case_open(fmd_hdl_t *hdl, void *data)
257 {
258 fmd_module_t *mp = (fmd_module_t *)hdl;
259 uuid_t uuid;
260
261 fmd_case_t *cp;
262
263 cp = fmd_hdl_zalloc(hdl, sizeof (fmd_case_t), FMD_SLEEP);
264 cp->ci_mod = hdl;
265 cp->ci_state = FMD_CASE_UNSOLVED;
266 cp->ci_flags = FMD_CF_DIRTY;
267 cp->ci_data = data;
268 cp->ci_bufptr = NULL;
269 cp->ci_bufsiz = 0;
270
271 uuid_generate(uuid);
272 uuid_unparse(uuid, cp->ci_uuid);
273
274 fmd_hdl_debug(hdl, "case opened (%s)", cp->ci_uuid);
275 mp->mod_stats.ms_caseopen.fmds_value.ui64++;
276
277 return (cp);
278 }
279
280 void
fmd_case_solve(fmd_hdl_t * hdl,fmd_case_t * cp)281 fmd_case_solve(fmd_hdl_t *hdl, fmd_case_t *cp)
282 {
283 fmd_module_t *mp = (fmd_module_t *)hdl;
284
285 /*
286 * For ZED, the event was already sent from fmd_case_add_suspect()
287 */
288
289 if (cp->ci_state >= FMD_CASE_SOLVED)
290 fmd_hdl_debug(hdl, "case is already solved or closed");
291
292 cp->ci_state = FMD_CASE_SOLVED;
293
294 fmd_hdl_debug(hdl, "case solved (%s)", cp->ci_uuid);
295 mp->mod_stats.ms_casesolved.fmds_value.ui64++;
296 }
297
298 void
fmd_case_close(fmd_hdl_t * hdl,fmd_case_t * cp)299 fmd_case_close(fmd_hdl_t *hdl, fmd_case_t *cp)
300 {
301 fmd_module_t *mp = (fmd_module_t *)hdl;
302 const fmd_hdl_ops_t *ops = mp->mod_info->fmdi_ops;
303
304 fmd_hdl_debug(hdl, "case closed (%s)", cp->ci_uuid);
305
306 if (ops->fmdo_close != NULL)
307 ops->fmdo_close(hdl, cp);
308
309 mp->mod_stats.ms_caseopen.fmds_value.ui64--;
310 mp->mod_stats.ms_caseclosed.fmds_value.ui64++;
311
312 if (cp->ci_bufptr != NULL && cp->ci_bufsiz > 0)
313 fmd_hdl_free(hdl, cp->ci_bufptr, cp->ci_bufsiz);
314
315 fmd_hdl_free(hdl, cp, sizeof (fmd_case_t));
316 }
317
318 void
fmd_case_uuresolved(fmd_hdl_t * hdl,const char * uuid)319 fmd_case_uuresolved(fmd_hdl_t *hdl, const char *uuid)
320 {
321 fmd_hdl_debug(hdl, "case resolved by uuid (%s)", uuid);
322 }
323
324 boolean_t
fmd_case_solved(fmd_hdl_t * hdl,fmd_case_t * cp)325 fmd_case_solved(fmd_hdl_t *hdl, fmd_case_t *cp)
326 {
327 (void) hdl;
328 return (cp->ci_state >= FMD_CASE_SOLVED);
329 }
330
331 void
fmd_case_add_ereport(fmd_hdl_t * hdl,fmd_case_t * cp,fmd_event_t * ep)332 fmd_case_add_ereport(fmd_hdl_t *hdl, fmd_case_t *cp, fmd_event_t *ep)
333 {
334 (void) hdl, (void) cp, (void) ep;
335 }
336
337 static void
zed_log_fault(nvlist_t * nvl,const char * uuid,const char * code)338 zed_log_fault(nvlist_t *nvl, const char *uuid, const char *code)
339 {
340 nvlist_t *rsrc;
341 const char *strval;
342 uint64_t guid;
343 uint8_t byte;
344
345 zed_log_msg(LOG_INFO, "\nzed_fault_event:");
346
347 if (uuid != NULL)
348 zed_log_msg(LOG_INFO, "\t%s: %s", FM_SUSPECT_UUID, uuid);
349 if (nvlist_lookup_string(nvl, FM_CLASS, &strval) == 0)
350 zed_log_msg(LOG_INFO, "\t%s: %s", FM_CLASS, strval);
351 if (code != NULL)
352 zed_log_msg(LOG_INFO, "\t%s: %s", FM_SUSPECT_DIAG_CODE, code);
353 if (nvlist_lookup_uint8(nvl, FM_FAULT_CERTAINTY, &byte) == 0)
354 zed_log_msg(LOG_INFO, "\t%s: %hhu", FM_FAULT_CERTAINTY, byte);
355 if (nvlist_lookup_nvlist(nvl, FM_FAULT_RESOURCE, &rsrc) == 0) {
356 if (nvlist_lookup_string(rsrc, FM_FMRI_SCHEME, &strval) == 0)
357 zed_log_msg(LOG_INFO, "\t%s: %s", FM_FMRI_SCHEME,
358 strval);
359 if (nvlist_lookup_uint64(rsrc, FM_FMRI_ZFS_POOL, &guid) == 0)
360 zed_log_msg(LOG_INFO, "\t%s: %llu", FM_FMRI_ZFS_POOL,
361 guid);
362 if (nvlist_lookup_uint64(rsrc, FM_FMRI_ZFS_VDEV, &guid) == 0)
363 zed_log_msg(LOG_INFO, "\t%s: %llu \n", FM_FMRI_ZFS_VDEV,
364 guid);
365 }
366 }
367
368 static const char *
fmd_fault_mkcode(nvlist_t * fault)369 fmd_fault_mkcode(nvlist_t *fault)
370 {
371 const char *class;
372 const char *code = "-";
373
374 /*
375 * Note: message codes come from: openzfs/usr/src/cmd/fm/dicts/ZFS.po
376 */
377 if (nvlist_lookup_string(fault, FM_CLASS, &class) == 0) {
378 if (strcmp(class, "fault.fs.zfs.vdev.io") == 0)
379 code = "ZFS-8000-FD";
380 else if (strcmp(class, "fault.fs.zfs.vdev.checksum") == 0)
381 code = "ZFS-8000-GH";
382 else if (strcmp(class, "fault.fs.zfs.io_failure_wait") == 0)
383 code = "ZFS-8000-HC";
384 else if (strcmp(class, "fault.fs.zfs.io_failure_continue") == 0)
385 code = "ZFS-8000-JQ";
386 else if (strcmp(class, "fault.fs.zfs.log_replay") == 0)
387 code = "ZFS-8000-K4";
388 else if (strcmp(class, "fault.fs.zfs.pool") == 0)
389 code = "ZFS-8000-CS";
390 else if (strcmp(class, "fault.fs.zfs.device") == 0)
391 code = "ZFS-8000-D3";
392
393 }
394 return (code);
395 }
396
397 void
fmd_case_add_suspect(fmd_hdl_t * hdl,fmd_case_t * cp,nvlist_t * fault)398 fmd_case_add_suspect(fmd_hdl_t *hdl, fmd_case_t *cp, nvlist_t *fault)
399 {
400 nvlist_t *nvl;
401 const char *code = fmd_fault_mkcode(fault);
402 int64_t tod[2];
403 int err = 0;
404
405 /*
406 * payload derived from fmd_protocol_list()
407 */
408
409 (void) gettimeofday(&cp->ci_tv, NULL);
410 tod[0] = cp->ci_tv.tv_sec;
411 tod[1] = cp->ci_tv.tv_usec;
412
413 nvl = fmd_nvl_alloc(hdl, FMD_SLEEP);
414
415 err |= nvlist_add_uint8(nvl, FM_VERSION, FM_SUSPECT_VERSION);
416 err |= nvlist_add_string(nvl, FM_CLASS, FM_LIST_SUSPECT_CLASS);
417 err |= nvlist_add_string(nvl, FM_SUSPECT_UUID, cp->ci_uuid);
418 err |= nvlist_add_string(nvl, FM_SUSPECT_DIAG_CODE, code);
419 err |= nvlist_add_int64_array(nvl, FM_SUSPECT_DIAG_TIME, tod, 2);
420 err |= nvlist_add_uint32(nvl, FM_SUSPECT_FAULT_SZ, 1);
421 err |= nvlist_add_nvlist_array(nvl, FM_SUSPECT_FAULT_LIST,
422 (const nvlist_t **)&fault, 1);
423
424 if (err)
425 zed_log_die("failed to populate nvlist");
426
427 zed_log_fault(fault, cp->ci_uuid, code);
428 zfs_agent_post_event(FM_LIST_SUSPECT_CLASS, NULL, nvl);
429
430 nvlist_free(nvl);
431 nvlist_free(fault);
432 }
433
434 void
fmd_case_setspecific(fmd_hdl_t * hdl,fmd_case_t * cp,void * data)435 fmd_case_setspecific(fmd_hdl_t *hdl, fmd_case_t *cp, void *data)
436 {
437 (void) hdl;
438 cp->ci_data = data;
439 }
440
441 void *
fmd_case_getspecific(fmd_hdl_t * hdl,fmd_case_t * cp)442 fmd_case_getspecific(fmd_hdl_t *hdl, fmd_case_t *cp)
443 {
444 (void) hdl;
445 return (cp->ci_data);
446 }
447
448 void
fmd_buf_create(fmd_hdl_t * hdl,fmd_case_t * cp,const char * name,size_t size)449 fmd_buf_create(fmd_hdl_t *hdl, fmd_case_t *cp, const char *name, size_t size)
450 {
451 assert(strcmp(name, "data") == 0), (void) name;
452 assert(cp->ci_bufptr == NULL);
453 assert(size < (1024 * 1024));
454
455 cp->ci_bufptr = fmd_hdl_alloc(hdl, size, FMD_SLEEP);
456 cp->ci_bufsiz = size;
457 }
458
459 void
fmd_buf_read(fmd_hdl_t * hdl,fmd_case_t * cp,const char * name,void * buf,size_t size)460 fmd_buf_read(fmd_hdl_t *hdl, fmd_case_t *cp,
461 const char *name, void *buf, size_t size)
462 {
463 (void) hdl;
464 assert(strcmp(name, "data") == 0), (void) name;
465 assert(cp->ci_bufptr != NULL);
466 assert(size <= cp->ci_bufsiz);
467
468 memcpy(buf, cp->ci_bufptr, size);
469 }
470
471 void
fmd_buf_write(fmd_hdl_t * hdl,fmd_case_t * cp,const char * name,const void * buf,size_t size)472 fmd_buf_write(fmd_hdl_t *hdl, fmd_case_t *cp,
473 const char *name, const void *buf, size_t size)
474 {
475 (void) hdl;
476 assert(strcmp(name, "data") == 0), (void) name;
477 assert(cp->ci_bufptr != NULL);
478 assert(cp->ci_bufsiz >= size);
479
480 memcpy(cp->ci_bufptr, buf, size);
481 }
482
483 /* SERD Engines */
484
485 void
fmd_serd_create(fmd_hdl_t * hdl,const char * name,uint_t n,hrtime_t t)486 fmd_serd_create(fmd_hdl_t *hdl, const char *name, uint_t n, hrtime_t t)
487 {
488 fmd_module_t *mp = (fmd_module_t *)hdl;
489
490 if (fmd_serd_eng_lookup(&mp->mod_serds, name) != NULL) {
491 zed_log_msg(LOG_ERR, "failed to create SERD engine '%s': "
492 " name already exists", name);
493 return;
494 }
495
496 (void) fmd_serd_eng_insert(&mp->mod_serds, name, n, t);
497 }
498
499 void
fmd_serd_destroy(fmd_hdl_t * hdl,const char * name)500 fmd_serd_destroy(fmd_hdl_t *hdl, const char *name)
501 {
502 fmd_module_t *mp = (fmd_module_t *)hdl;
503
504 fmd_serd_eng_delete(&mp->mod_serds, name);
505
506 fmd_hdl_debug(hdl, "serd_destroy %s", name);
507 }
508
509 int
fmd_serd_exists(fmd_hdl_t * hdl,const char * name)510 fmd_serd_exists(fmd_hdl_t *hdl, const char *name)
511 {
512 fmd_module_t *mp = (fmd_module_t *)hdl;
513
514 return (fmd_serd_eng_lookup(&mp->mod_serds, name) != NULL);
515 }
516
517 int
fmd_serd_active(fmd_hdl_t * hdl,const char * name)518 fmd_serd_active(fmd_hdl_t *hdl, const char *name)
519 {
520 fmd_module_t *mp = (fmd_module_t *)hdl;
521 fmd_serd_eng_t *sgp;
522
523 if ((sgp = fmd_serd_eng_lookup(&mp->mod_serds, name)) == NULL) {
524 zed_log_msg(LOG_ERR, "serd engine '%s' does not exist", name);
525 return (0);
526 }
527 return (fmd_serd_eng_fired(sgp) || !fmd_serd_eng_empty(sgp));
528 }
529
530 void
fmd_serd_reset(fmd_hdl_t * hdl,const char * name)531 fmd_serd_reset(fmd_hdl_t *hdl, const char *name)
532 {
533 fmd_module_t *mp = (fmd_module_t *)hdl;
534 fmd_serd_eng_t *sgp;
535
536 if ((sgp = fmd_serd_eng_lookup(&mp->mod_serds, name)) == NULL) {
537 zed_log_msg(LOG_ERR, "serd engine '%s' does not exist", name);
538 } else {
539 fmd_serd_eng_reset(sgp);
540 fmd_hdl_debug(hdl, "serd_reset %s", name);
541 }
542 }
543
544 int
fmd_serd_record(fmd_hdl_t * hdl,const char * name,fmd_event_t * ep)545 fmd_serd_record(fmd_hdl_t *hdl, const char *name, fmd_event_t *ep)
546 {
547 fmd_module_t *mp = (fmd_module_t *)hdl;
548 fmd_serd_eng_t *sgp;
549
550 if ((sgp = fmd_serd_eng_lookup(&mp->mod_serds, name)) == NULL) {
551 zed_log_msg(LOG_ERR, "failed to add record to SERD engine '%s'",
552 name);
553 return (0);
554 }
555 return (fmd_serd_eng_record(sgp, ep->ev_hrt));
556 }
557
558 void
fmd_serd_gc(fmd_hdl_t * hdl)559 fmd_serd_gc(fmd_hdl_t *hdl)
560 {
561 fmd_module_t *mp = (fmd_module_t *)hdl;
562
563 fmd_serd_hash_apply(&mp->mod_serds, fmd_serd_eng_gc, NULL);
564 }
565
566 /* FMD Timers */
567
568 static void
_timer_notify(union sigval sv)569 _timer_notify(union sigval sv)
570 {
571 fmd_timer_t *ftp = sv.sival_ptr;
572 fmd_hdl_t *hdl = ftp->ft_hdl;
573 fmd_module_t *mp = (fmd_module_t *)hdl;
574 const fmd_hdl_ops_t *ops = mp->mod_info->fmdi_ops;
575 struct itimerspec its;
576
577 fmd_hdl_debug(hdl, "%s timer fired (%p)", mp->mod_name, ftp->ft_tid);
578
579 /* disarm the timer */
580 memset(&its, 0, sizeof (struct itimerspec));
581 timer_settime(ftp->ft_tid, 0, &its, NULL);
582
583 /* Note that the fmdo_timeout can remove this timer */
584 if (ops->fmdo_timeout != NULL)
585 ops->fmdo_timeout(hdl, ftp, ftp->ft_arg);
586 }
587
588 /*
589 * Install a new timer which will fire at least delta nanoseconds after the
590 * current time. After the timeout has expired, the module's fmdo_timeout
591 * entry point is called.
592 */
593 fmd_timer_t *
fmd_timer_install(fmd_hdl_t * hdl,void * arg,fmd_event_t * ep,hrtime_t delta)594 fmd_timer_install(fmd_hdl_t *hdl, void *arg, fmd_event_t *ep, hrtime_t delta)
595 {
596 (void) ep;
597 struct sigevent sev;
598 struct itimerspec its;
599 fmd_timer_t *ftp;
600
601 ftp = fmd_hdl_alloc(hdl, sizeof (fmd_timer_t), FMD_SLEEP);
602 ftp->ft_arg = arg;
603 ftp->ft_hdl = hdl;
604
605 its.it_value.tv_sec = delta / 1000000000;
606 its.it_value.tv_nsec = delta % 1000000000;
607 its.it_interval.tv_sec = its.it_value.tv_sec;
608 its.it_interval.tv_nsec = its.it_value.tv_nsec;
609
610 sev.sigev_notify = SIGEV_THREAD;
611 sev.sigev_notify_function = _timer_notify;
612 sev.sigev_notify_attributes = NULL;
613 sev.sigev_value.sival_ptr = ftp;
614 sev.sigev_signo = 0;
615
616 timer_create(CLOCK_REALTIME, &sev, &ftp->ft_tid);
617 timer_settime(ftp->ft_tid, 0, &its, NULL);
618
619 fmd_hdl_debug(hdl, "installing timer for %d secs (%p)",
620 (int)its.it_value.tv_sec, ftp->ft_tid);
621
622 return (ftp);
623 }
624
625 void
fmd_timer_remove(fmd_hdl_t * hdl,fmd_timer_t * ftp)626 fmd_timer_remove(fmd_hdl_t *hdl, fmd_timer_t *ftp)
627 {
628 fmd_hdl_debug(hdl, "removing timer (%p)", ftp->ft_tid);
629
630 timer_delete(ftp->ft_tid);
631
632 fmd_hdl_free(hdl, ftp, sizeof (fmd_timer_t));
633 }
634
635 /* Name-Value Pair Lists */
636
637 nvlist_t *
fmd_nvl_create_fault(fmd_hdl_t * hdl,const char * class,uint8_t certainty,nvlist_t * asru,nvlist_t * fru,nvlist_t * resource)638 fmd_nvl_create_fault(fmd_hdl_t *hdl, const char *class, uint8_t certainty,
639 nvlist_t *asru, nvlist_t *fru, nvlist_t *resource)
640 {
641 (void) hdl;
642 nvlist_t *nvl;
643 int err = 0;
644
645 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
646 zed_log_die("failed to xalloc fault nvlist");
647
648 err |= nvlist_add_uint8(nvl, FM_VERSION, FM_FAULT_VERSION);
649 err |= nvlist_add_string(nvl, FM_CLASS, class);
650 err |= nvlist_add_uint8(nvl, FM_FAULT_CERTAINTY, certainty);
651
652 if (asru != NULL)
653 err |= nvlist_add_nvlist(nvl, FM_FAULT_ASRU, asru);
654 if (fru != NULL)
655 err |= nvlist_add_nvlist(nvl, FM_FAULT_FRU, fru);
656 if (resource != NULL)
657 err |= nvlist_add_nvlist(nvl, FM_FAULT_RESOURCE, resource);
658
659 if (err)
660 zed_log_die("failed to populate nvlist: %s\n", strerror(err));
661
662 return (nvl);
663 }
664
665 /*
666 * sourced from fmd_string.c
667 */
668 static int
fmd_strmatch(const char * s,const char * p)669 fmd_strmatch(const char *s, const char *p)
670 {
671 char c;
672
673 if (p == NULL)
674 return (0);
675
676 if (s == NULL)
677 s = ""; /* treat NULL string as the empty string */
678
679 do {
680 if ((c = *p++) == '\0')
681 return (*s == '\0');
682
683 if (c == '*') {
684 while (*p == '*')
685 p++; /* consecutive *'s can be collapsed */
686
687 if (*p == '\0')
688 return (1);
689
690 while (*s != '\0') {
691 if (fmd_strmatch(s++, p) != 0)
692 return (1);
693 }
694
695 return (0);
696 }
697 } while (c == *s++);
698
699 return (0);
700 }
701
702 int
fmd_nvl_class_match(fmd_hdl_t * hdl,nvlist_t * nvl,const char * pattern)703 fmd_nvl_class_match(fmd_hdl_t *hdl, nvlist_t *nvl, const char *pattern)
704 {
705 (void) hdl;
706 const char *class;
707
708 return (nvl != NULL &&
709 nvlist_lookup_string(nvl, FM_CLASS, &class) == 0 &&
710 fmd_strmatch(class, pattern));
711 }
712
713 nvlist_t *
fmd_nvl_alloc(fmd_hdl_t * hdl,int flags)714 fmd_nvl_alloc(fmd_hdl_t *hdl, int flags)
715 {
716 (void) hdl, (void) flags;
717 nvlist_t *nvl = NULL;
718
719 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
720 return (NULL);
721
722 return (nvl);
723 }
724
725
726 /*
727 * ZED Agent specific APIs
728 */
729
730 fmd_hdl_t *
fmd_module_hdl(const char * name)731 fmd_module_hdl(const char *name)
732 {
733 if (strcmp(name, "zfs-retire") == 0)
734 return ((fmd_hdl_t *)&zfs_retire_module);
735 if (strcmp(name, "zfs-diagnosis") == 0)
736 return ((fmd_hdl_t *)&zfs_diagnosis_module);
737
738 return (NULL);
739 }
740
741 boolean_t
fmd_module_initialized(fmd_hdl_t * hdl)742 fmd_module_initialized(fmd_hdl_t *hdl)
743 {
744 fmd_module_t *mp = (fmd_module_t *)hdl;
745
746 return (mp->mod_info != NULL);
747 }
748
749 /*
750 * fmd_module_recv is called for each event that is received by
751 * the fault manager that has a class that matches one of the
752 * module's subscriptions.
753 */
754 void
fmd_module_recv(fmd_hdl_t * hdl,nvlist_t * nvl,const char * class)755 fmd_module_recv(fmd_hdl_t *hdl, nvlist_t *nvl, const char *class)
756 {
757 fmd_module_t *mp = (fmd_module_t *)hdl;
758 const fmd_hdl_ops_t *ops = mp->mod_info->fmdi_ops;
759 fmd_event_t faux_event = {0};
760 int64_t *tv;
761 uint_t n;
762
763 /*
764 * Will need to normalized this if we persistently store the case data
765 */
766 if (nvlist_lookup_int64_array(nvl, FM_EREPORT_TIME, &tv, &n) == 0)
767 faux_event.ev_hrt = tv[0] * NANOSEC + tv[1];
768 else
769 faux_event.ev_hrt = 0;
770
771 ops->fmdo_recv(hdl, &faux_event, nvl, class);
772
773 mp->mod_stats.ms_accepted.fmds_value.ui64++;
774
775 /* TBD - should we initiate fm_module_gc() periodically? */
776 }
777