1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
|
/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License Version 1.0 (CDDL-1.0).
* You can obtain a copy of the license from the top-level file
* "OPENSOLARIS.LICENSE" or at <http://opensource.org/licenses/CDDL-1.0>.
* You may not use this file except in compliance with the license.
*
* CDDL HEADER END
*/
/*
* Copyright (c) 2016, Intel Corporation.
* Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
* Copyright (c) 2021 Hewlett Packard Enterprise Development LP
*/
#include <libnvpair.h>
#include <libzfs.h>
#include <stddef.h>
#include <stdlib.h>
#include <string.h>
#include <sys/list.h>
#include <sys/time.h>
#include <sys/sysevent/eventdefs.h>
#include <sys/sysevent/dev.h>
#include <sys/fm/protocol.h>
#include <sys/fm/fs/zfs.h>
#include <pthread.h>
#include <unistd.h>
#include "zfs_agents.h"
#include "fmd_api.h"
#include "../zed_log.h"
/*
* agent dispatch code
*/
static pthread_mutex_t agent_lock = PTHREAD_MUTEX_INITIALIZER;
static pthread_cond_t agent_cond = PTHREAD_COND_INITIALIZER;
static list_t agent_events; /* list of pending events */
static int agent_exiting;
typedef struct agent_event {
char ae_class[64];
char ae_subclass[32];
nvlist_t *ae_nvl;
list_node_t ae_node;
} agent_event_t;
pthread_t g_agents_tid;
libzfs_handle_t *g_zfs_hdl;
/* guid search data */
typedef enum device_type {
DEVICE_TYPE_L2ARC, /* l2arc device */
DEVICE_TYPE_SPARE, /* spare device */
DEVICE_TYPE_PRIMARY /* any primary pool storage device */
} device_type_t;
typedef struct guid_search {
uint64_t gs_pool_guid;
uint64_t gs_vdev_guid;
const char *gs_devid;
device_type_t gs_vdev_type;
uint64_t gs_vdev_expandtime; /* vdev expansion time */
} guid_search_t;
/*
* Walks the vdev tree recursively looking for a matching devid.
* Returns B_TRUE as soon as a matching device is found, B_FALSE otherwise.
*/
static boolean_t
zfs_agent_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *arg)
{
guid_search_t *gsp = arg;
const char *path = NULL;
uint_t c, children;
nvlist_t **child;
uint64_t vdev_guid;
/*
* First iterate over any children.
*/
if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_CHILDREN,
&child, &children) == 0) {
for (c = 0; c < children; c++) {
if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
gsp->gs_vdev_type = DEVICE_TYPE_PRIMARY;
return (B_TRUE);
}
}
}
/*
* Iterate over any spares and cache devices
*/
if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_SPARES,
&child, &children) == 0) {
for (c = 0; c < children; c++) {
if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
gsp->gs_vdev_type = DEVICE_TYPE_SPARE;
return (B_TRUE);
}
}
}
if (nvlist_lookup_nvlist_array(nvl, ZPOOL_CONFIG_L2CACHE,
&child, &children) == 0) {
for (c = 0; c < children; c++) {
if (zfs_agent_iter_vdev(zhp, child[c], gsp)) {
gsp->gs_vdev_type = DEVICE_TYPE_L2ARC;
return (B_TRUE);
}
}
}
/*
* On a devid match, grab the vdev guid and expansion time, if any.
*/
if (gsp->gs_devid != NULL &&
(nvlist_lookup_string(nvl, ZPOOL_CONFIG_DEVID, &path) == 0) &&
(strcmp(gsp->gs_devid, path) == 0)) {
(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID,
&gsp->gs_vdev_guid);
(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_EXPANSION_TIME,
&gsp->gs_vdev_expandtime);
return (B_TRUE);
}
/*
* Otherwise, on a vdev guid match, grab the devid and expansion
* time. The devid might be missing on removal since its not part
* of blkid cache and L2ARC VDEV does not contain pool guid in its
* blkid, so this is a special case for L2ARC VDEV.
*/
else if (gsp->gs_vdev_guid != 0 && gsp->gs_devid == NULL &&
nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID, &vdev_guid) == 0 &&
gsp->gs_vdev_guid == vdev_guid) {
(void) nvlist_lookup_string(nvl, ZPOOL_CONFIG_DEVID,
&gsp->gs_devid);
(void) nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_EXPANSION_TIME,
&gsp->gs_vdev_expandtime);
return (B_TRUE);
}
return (B_FALSE);
}
static int
zfs_agent_iter_pool(zpool_handle_t *zhp, void *arg)
{
guid_search_t *gsp = arg;
nvlist_t *config, *nvl;
/*
* For each vdev in this pool, look for a match by devid
*/
if ((config = zpool_get_config(zhp, NULL)) != NULL) {
if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
&nvl) == 0) {
(void) zfs_agent_iter_vdev(zhp, nvl, gsp);
}
}
/*
* if a match was found then grab the pool guid
*/
if (gsp->gs_vdev_guid && gsp->gs_devid) {
(void) nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
&gsp->gs_pool_guid);
}
zpool_close(zhp);
return (gsp->gs_devid != NULL && gsp->gs_vdev_guid != 0);
}
void
zfs_agent_post_event(const char *class, const char *subclass, nvlist_t *nvl)
{
agent_event_t *event;
if (subclass == NULL)
subclass = "";
event = malloc(sizeof (agent_event_t));
if (event == NULL || nvlist_dup(nvl, &event->ae_nvl, 0) != 0) {
if (event)
free(event);
return;
}
if (strcmp(class, "sysevent.fs.zfs.vdev_check") == 0) {
class = EC_ZFS;
subclass = ESC_ZFS_VDEV_CHECK;
}
/*
* On Linux, we don't get the expected FM_RESOURCE_REMOVED ereport
* from the vdev_disk layer after a hot unplug. Fortunately we do
* get an EC_DEV_REMOVE from our disk monitor and it is a suitable
* proxy so we remap it here for the benefit of the diagnosis engine.
* Starting in OpenZFS 2.0, we do get FM_RESOURCE_REMOVED from the spa
* layer. Processing multiple FM_RESOURCE_REMOVED events is not harmful.
*/
if ((strcmp(class, EC_DEV_REMOVE) == 0) &&
(strcmp(subclass, ESC_DISK) == 0) &&
(nvlist_exists(nvl, ZFS_EV_VDEV_GUID) ||
nvlist_exists(nvl, DEV_IDENTIFIER))) {
nvlist_t *payload = event->ae_nvl;
struct timeval tv;
int64_t tod[2];
uint64_t pool_guid = 0, vdev_guid = 0;
guid_search_t search = { 0 };
device_type_t devtype = DEVICE_TYPE_PRIMARY;
const char *devid = NULL;
class = "resource.fs.zfs.removed";
subclass = "";
(void) nvlist_add_string(payload, FM_CLASS, class);
(void) nvlist_lookup_string(nvl, DEV_IDENTIFIER, &devid);
(void) nvlist_lookup_uint64(nvl, ZFS_EV_POOL_GUID, &pool_guid);
(void) nvlist_lookup_uint64(nvl, ZFS_EV_VDEV_GUID, &vdev_guid);
(void) gettimeofday(&tv, NULL);
tod[0] = tv.tv_sec;
tod[1] = tv.tv_usec;
(void) nvlist_add_int64_array(payload, FM_EREPORT_TIME, tod, 2);
/*
* If devid is missing but vdev_guid is available, find devid
* and pool_guid from vdev_guid.
* For multipath, spare and l2arc devices ZFS_EV_VDEV_GUID or
* ZFS_EV_POOL_GUID may be missing so find them.
*/
if (devid == NULL || pool_guid == 0 || vdev_guid == 0) {
if (devid == NULL)
search.gs_vdev_guid = vdev_guid;
else
search.gs_devid = devid;
zpool_iter(g_zfs_hdl, zfs_agent_iter_pool, &search);
if (devid == NULL)
devid = search.gs_devid;
if (pool_guid == 0)
pool_guid = search.gs_pool_guid;
if (vdev_guid == 0)
vdev_guid = search.gs_vdev_guid;
devtype = search.gs_vdev_type;
}
/*
* We want to avoid reporting "remove" events coming from
* libudev for VDEVs which were expanded recently (10s) and
* avoid activating spares in response to partitions being
* deleted and created in rapid succession.
*/
if (search.gs_vdev_expandtime != 0 &&
search.gs_vdev_expandtime + 10 > tv.tv_sec) {
zed_log_msg(LOG_INFO, "agent post event: ignoring '%s' "
"for recently expanded device '%s'", EC_DEV_REMOVE,
devid);
fnvlist_free(payload);
free(event);
goto out;
}
(void) nvlist_add_uint64(payload,
FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, pool_guid);
(void) nvlist_add_uint64(payload,
FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vdev_guid);
switch (devtype) {
case DEVICE_TYPE_L2ARC:
(void) nvlist_add_string(payload,
FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
VDEV_TYPE_L2CACHE);
break;
case DEVICE_TYPE_SPARE:
(void) nvlist_add_string(payload,
FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE, VDEV_TYPE_SPARE);
break;
case DEVICE_TYPE_PRIMARY:
(void) nvlist_add_string(payload,
FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE, VDEV_TYPE_DISK);
break;
}
zed_log_msg(LOG_INFO, "agent post event: mapping '%s' to '%s'",
EC_DEV_REMOVE, class);
}
(void) strlcpy(event->ae_class, class, sizeof (event->ae_class));
(void) strlcpy(event->ae_subclass, subclass,
sizeof (event->ae_subclass));
(void) pthread_mutex_lock(&agent_lock);
list_insert_tail(&agent_events, event);
(void) pthread_mutex_unlock(&agent_lock);
out:
(void) pthread_cond_signal(&agent_cond);
}
static void
zfs_agent_dispatch(const char *class, const char *subclass, nvlist_t *nvl)
{
/*
* The diagnosis engine subscribes to the following events.
* On illumos these subscriptions reside in:
* /usr/lib/fm/fmd/plugins/zfs-diagnosis.conf
*/
if (strstr(class, "ereport.fs.zfs.") != NULL ||
strstr(class, "resource.fs.zfs.") != NULL ||
strcmp(class, "sysevent.fs.zfs.vdev_remove") == 0 ||
strcmp(class, "sysevent.fs.zfs.vdev_remove_dev") == 0 ||
strcmp(class, "sysevent.fs.zfs.pool_destroy") == 0) {
fmd_module_recv(fmd_module_hdl("zfs-diagnosis"), nvl, class);
}
/*
* The retire agent subscribes to the following events.
* On illumos these subscriptions reside in:
* /usr/lib/fm/fmd/plugins/zfs-retire.conf
*
* NOTE: faults events come directly from our diagnosis engine
* and will not pass through the zfs kernel module.
*/
if (strcmp(class, FM_LIST_SUSPECT_CLASS) == 0 ||
strcmp(class, "resource.fs.zfs.removed") == 0 ||
strcmp(class, "resource.fs.zfs.statechange") == 0 ||
strcmp(class, "sysevent.fs.zfs.vdev_remove") == 0) {
fmd_module_recv(fmd_module_hdl("zfs-retire"), nvl, class);
}
/*
* The SLM module only consumes disk events and vdev check events
*
* NOTE: disk events come directly from disk monitor and will
* not pass through the zfs kernel module.
*/
if (strstr(class, "EC_dev_") != NULL ||
strcmp(class, EC_ZFS) == 0) {
(void) zfs_slm_event(class, subclass, nvl);
}
}
/*
* Events are consumed and dispatched from this thread
* An agent can also post an event so event list lock
* is not held when calling an agent.
* One event is consumed at a time.
*/
static void *
zfs_agent_consumer_thread(void *arg)
{
(void) arg;
for (;;) {
agent_event_t *event;
(void) pthread_mutex_lock(&agent_lock);
/* wait for an event to show up */
while (!agent_exiting && list_is_empty(&agent_events))
(void) pthread_cond_wait(&agent_cond, &agent_lock);
if (agent_exiting) {
(void) pthread_mutex_unlock(&agent_lock);
zed_log_msg(LOG_INFO, "zfs_agent_consumer_thread: "
"exiting");
return (NULL);
}
if ((event = list_remove_head(&agent_events)) != NULL) {
(void) pthread_mutex_unlock(&agent_lock);
/* dispatch to all event subscribers */
zfs_agent_dispatch(event->ae_class, event->ae_subclass,
event->ae_nvl);
nvlist_free(event->ae_nvl);
free(event);
continue;
}
(void) pthread_mutex_unlock(&agent_lock);
}
return (NULL);
}
void
zfs_agent_init(libzfs_handle_t *zfs_hdl)
{
fmd_hdl_t *hdl;
g_zfs_hdl = zfs_hdl;
if (zfs_slm_init() != 0)
zed_log_die("Failed to initialize zfs slm");
zed_log_msg(LOG_INFO, "Add Agent: init");
hdl = fmd_module_hdl("zfs-diagnosis");
_zfs_diagnosis_init(hdl);
if (!fmd_module_initialized(hdl))
zed_log_die("Failed to initialize zfs diagnosis");
hdl = fmd_module_hdl("zfs-retire");
_zfs_retire_init(hdl);
if (!fmd_module_initialized(hdl))
zed_log_die("Failed to initialize zfs retire");
list_create(&agent_events, sizeof (agent_event_t),
offsetof(struct agent_event, ae_node));
if (pthread_create(&g_agents_tid, NULL, zfs_agent_consumer_thread,
NULL) != 0) {
list_destroy(&agent_events);
zed_log_die("Failed to initialize agents");
}
pthread_setname_np(g_agents_tid, "agents");
}
void
zfs_agent_fini(void)
{
fmd_hdl_t *hdl;
agent_event_t *event;
agent_exiting = 1;
(void) pthread_cond_signal(&agent_cond);
/* wait for zfs_enum_pools thread to complete */
(void) pthread_join(g_agents_tid, NULL);
/* drain any pending events */
while ((event = list_remove_head(&agent_events)) != NULL) {
nvlist_free(event->ae_nvl);
free(event);
}
list_destroy(&agent_events);
if ((hdl = fmd_module_hdl("zfs-retire")) != NULL) {
_zfs_retire_fini(hdl);
fmd_hdl_unregister(hdl);
}
if ((hdl = fmd_module_hdl("zfs-diagnosis")) != NULL) {
_zfs_diagnosis_fini(hdl);
fmd_hdl_unregister(hdl);
}
zed_log_msg(LOG_INFO, "Add Agent: fini");
zfs_slm_fini();
g_zfs_hdl = NULL;
}
|