]>
Commit | Line | Data |
---|---|---|
34dc7c2f BB |
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 | /* | |
9babb374 | 22 | * Copyright 2009 Sun Microsystems, Inc. All rights reserved. |
34dc7c2f BB |
23 | * Use is subject to license terms. |
24 | */ | |
25 | ||
5ffb9d1d GW |
26 | /* |
27 | * Copyright (c) 2012 by Delphix. All rights reserved. | |
28 | */ | |
29 | ||
34dc7c2f BB |
30 | #include <sys/spa.h> |
31 | #include <sys/spa_impl.h> | |
32 | #include <sys/vdev.h> | |
33 | #include <sys/vdev_impl.h> | |
34 | #include <sys/zio.h> | |
428870ff | 35 | #include <sys/zio_checksum.h> |
34dc7c2f BB |
36 | |
37 | #include <sys/fm/fs/zfs.h> | |
38 | #include <sys/fm/protocol.h> | |
39 | #include <sys/fm/util.h> | |
40 | #include <sys/sysevent.h> | |
41 | ||
42 | /* | |
43 | * This general routine is responsible for generating all the different ZFS | |
44 | * ereports. The payload is dependent on the class, and which arguments are | |
45 | * supplied to the function: | |
46 | * | |
47 | * EREPORT POOL VDEV IO | |
48 | * block X X X | |
49 | * data X X | |
50 | * device X X | |
51 | * pool X | |
52 | * | |
53 | * If we are in a loading state, all errors are chained together by the same | |
b128c09f | 54 | * SPA-wide ENA (Error Numeric Association). |
34dc7c2f BB |
55 | * |
56 | * For isolated I/O requests, we get the ENA from the zio_t. The propagation | |
57 | * gets very complicated due to RAID-Z, gang blocks, and vdev caching. We want | |
58 | * to chain together all ereports associated with a logical piece of data. For | |
59 | * read I/Os, there are basically three 'types' of I/O, which form a roughly | |
60 | * layered diagram: | |
61 | * | |
62 | * +---------------+ | |
63 | * | Aggregate I/O | No associated logical data or device | |
64 | * +---------------+ | |
65 | * | | |
66 | * V | |
67 | * +---------------+ Reads associated with a piece of logical data. | |
68 | * | Read I/O | This includes reads on behalf of RAID-Z, | |
69 | * +---------------+ mirrors, gang blocks, retries, etc. | |
70 | * | | |
71 | * V | |
72 | * +---------------+ Reads associated with a particular device, but | |
73 | * | Physical I/O | no logical data. Issued as part of vdev caching | |
74 | * +---------------+ and I/O aggregation. | |
75 | * | |
76 | * Note that 'physical I/O' here is not the same terminology as used in the rest | |
77 | * of ZIO. Typically, 'physical I/O' simply means that there is no attached | |
78 | * blockpointer. But I/O with no associated block pointer can still be related | |
79 | * to a logical piece of data (i.e. RAID-Z requests). | |
80 | * | |
81 | * Purely physical I/O always have unique ENAs. They are not related to a | |
82 | * particular piece of logical data, and therefore cannot be chained together. | |
83 | * We still generate an ereport, but the DE doesn't correlate it with any | |
84 | * logical piece of data. When such an I/O fails, the delegated I/O requests | |
85 | * will issue a retry, which will trigger the 'real' ereport with the correct | |
86 | * ENA. | |
87 | * | |
88 | * We keep track of the ENA for a ZIO chain through the 'io_logical' member. | |
89 | * When a new logical I/O is issued, we set this to point to itself. Child I/Os | |
90 | * then inherit this pointer, so that when it is first set subsequent failures | |
b128c09f BB |
91 | * will use the same ENA. For vdev cache fill and queue aggregation I/O, |
92 | * this pointer is set to NULL, and no ereport will be generated (since it | |
93 | * doesn't actually correspond to any particular device or piece of data, | |
94 | * and the caller will always retry without caching or queueing anyway). | |
428870ff BB |
95 | * |
96 | * For checksum errors, we want to include more information about the actual | |
97 | * error which occurs. Accordingly, we build an ereport when the error is | |
98 | * noticed, but instead of sending it in immediately, we hang it off of the | |
99 | * io_cksum_report field of the logical IO. When the logical IO completes | |
100 | * (successfully or not), zfs_ereport_finish_checksum() is called with the | |
101 | * good and bad versions of the buffer (if available), and we annotate the | |
102 | * ereport with information about the differences. | |
34dc7c2f | 103 | */ |
428870ff | 104 | #ifdef _KERNEL |
12fa0466 | 105 | void |
26685276 BB |
106 | zfs_zevent_post_cb(nvlist_t *nvl, nvlist_t *detector) |
107 | { | |
108 | if (nvl) | |
109 | fm_nvlist_destroy(nvl, FM_NVA_FREE); | |
110 | ||
111 | if (detector) | |
112 | fm_nvlist_destroy(detector, FM_NVA_FREE); | |
113 | } | |
114 | ||
6078881a TH |
115 | /* |
116 | * We want to rate limit ZIO delay and checksum events so as to not | |
117 | * flood ZED when a disk is acting up. | |
118 | * | |
119 | * Returns 1 if we're ratelimiting, 0 if not. | |
120 | */ | |
121 | static int | |
122 | zfs_is_ratelimiting_event(const char *subclass, vdev_t *vd) | |
123 | { | |
124 | int rc = 0; | |
125 | /* | |
126 | * __ratelimit() returns 1 if we're *not* ratelimiting and 0 if we | |
127 | * are. Invert it to get our return value. | |
128 | */ | |
129 | if (strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) { | |
130 | rc = !zfs_ratelimit(&vd->vdev_delay_rl); | |
131 | } else if (strcmp(subclass, FM_EREPORT_ZFS_CHECKSUM) == 0) { | |
132 | rc = !zfs_ratelimit(&vd->vdev_checksum_rl); | |
133 | } | |
134 | ||
135 | if (rc) { | |
136 | /* We're rate limiting */ | |
137 | fm_erpt_dropped_increment(); | |
138 | } | |
139 | ||
140 | return (rc); | |
141 | } | |
0426c168 | 142 | |
428870ff BB |
143 | static void |
144 | zfs_ereport_start(nvlist_t **ereport_out, nvlist_t **detector_out, | |
b5256303 TC |
145 | const char *subclass, spa_t *spa, vdev_t *vd, zbookmark_phys_t *zb, |
146 | zio_t *zio, uint64_t stateoroffset, uint64_t size) | |
34dc7c2f | 147 | { |
34dc7c2f | 148 | nvlist_t *ereport, *detector; |
428870ff | 149 | |
34dc7c2f BB |
150 | uint64_t ena; |
151 | char class[64]; | |
152 | ||
153 | /* | |
428870ff BB |
154 | * If we are doing a spa_tryimport() or in recovery mode, |
155 | * ignore errors. | |
34dc7c2f | 156 | */ |
428870ff BB |
157 | if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT || |
158 | spa_load_state(spa) == SPA_LOAD_RECOVER) | |
34dc7c2f BB |
159 | return; |
160 | ||
161 | /* | |
162 | * If we are in the middle of opening a pool, and the previous attempt | |
163 | * failed, don't bother logging any new ereports - we're just going to | |
164 | * get the same diagnosis anyway. | |
165 | */ | |
428870ff | 166 | if (spa_load_state(spa) != SPA_LOAD_NONE && |
34dc7c2f BB |
167 | spa->spa_last_open_failed) |
168 | return; | |
169 | ||
b128c09f BB |
170 | if (zio != NULL) { |
171 | /* | |
172 | * If this is not a read or write zio, ignore the error. This | |
173 | * can occur if the DKIOCFLUSHWRITECACHE ioctl fails. | |
174 | */ | |
175 | if (zio->io_type != ZIO_TYPE_READ && | |
176 | zio->io_type != ZIO_TYPE_WRITE) | |
177 | return; | |
34dc7c2f | 178 | |
9babb374 BB |
179 | if (vd != NULL) { |
180 | /* | |
181 | * If the vdev has already been marked as failing due | |
182 | * to a failed probe, then ignore any subsequent I/O | |
183 | * errors, as the DE will automatically fault the vdev | |
184 | * on the first such failure. This also catches cases | |
185 | * where vdev_remove_wanted is set and the device has | |
186 | * not yet been asynchronously placed into the REMOVED | |
187 | * state. | |
188 | */ | |
428870ff | 189 | if (zio->io_vd == vd && !vdev_accessible(vd, zio)) |
9babb374 BB |
190 | return; |
191 | ||
192 | /* | |
193 | * Ignore checksum errors for reads from DTL regions of | |
194 | * leaf vdevs. | |
195 | */ | |
196 | if (zio->io_type == ZIO_TYPE_READ && | |
197 | zio->io_error == ECKSUM && | |
198 | vd->vdev_ops->vdev_op_leaf && | |
199 | vdev_dtl_contains(vd, DTL_MISSING, zio->io_txg, 1)) | |
200 | return; | |
201 | } | |
b128c09f | 202 | } |
34dc7c2f | 203 | |
428870ff BB |
204 | /* |
205 | * For probe failure, we want to avoid posting ereports if we've | |
206 | * already removed the device in the meantime. | |
207 | */ | |
208 | if (vd != NULL && | |
209 | strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) == 0 && | |
210 | (vd->vdev_remove_wanted || vd->vdev_state == VDEV_STATE_REMOVED)) | |
211 | return; | |
212 | ||
a32df59e TC |
213 | if ((strcmp(subclass, FM_EREPORT_ZFS_DELAY) == 0) && |
214 | (zio != NULL) && (!zio->io_timestamp)) { | |
215 | /* Ignore bogus delay events */ | |
216 | return; | |
217 | } | |
218 | ||
34dc7c2f BB |
219 | if ((ereport = fm_nvlist_create(NULL)) == NULL) |
220 | return; | |
221 | ||
222 | if ((detector = fm_nvlist_create(NULL)) == NULL) { | |
223 | fm_nvlist_destroy(ereport, FM_NVA_FREE); | |
224 | return; | |
225 | } | |
226 | ||
227 | /* | |
228 | * Serialize ereport generation | |
229 | */ | |
230 | mutex_enter(&spa->spa_errlist_lock); | |
231 | ||
232 | /* | |
233 | * Determine the ENA to use for this event. If we are in a loading | |
234 | * state, use a SPA-wide ENA. Otherwise, if we are in an I/O state, use | |
235 | * a root zio-wide ENA. Otherwise, simply use a unique ENA. | |
236 | */ | |
428870ff | 237 | if (spa_load_state(spa) != SPA_LOAD_NONE) { |
34dc7c2f BB |
238 | if (spa->spa_ena == 0) |
239 | spa->spa_ena = fm_ena_generate(0, FM_ENA_FMT1); | |
240 | ena = spa->spa_ena; | |
241 | } else if (zio != NULL && zio->io_logical != NULL) { | |
242 | if (zio->io_logical->io_ena == 0) | |
243 | zio->io_logical->io_ena = | |
244 | fm_ena_generate(0, FM_ENA_FMT1); | |
245 | ena = zio->io_logical->io_ena; | |
246 | } else { | |
247 | ena = fm_ena_generate(0, FM_ENA_FMT1); | |
248 | } | |
249 | ||
250 | /* | |
251 | * Construct the full class, detector, and other standard FMA fields. | |
252 | */ | |
253 | (void) snprintf(class, sizeof (class), "%s.%s", | |
254 | ZFS_ERROR_CLASS, subclass); | |
255 | ||
256 | fm_fmri_zfs_set(detector, FM_ZFS_SCHEME_VERSION, spa_guid(spa), | |
257 | vd != NULL ? vd->vdev_guid : 0); | |
258 | ||
259 | fm_ereport_set(ereport, FM_EREPORT_VERSION, class, ena, detector, NULL); | |
260 | ||
261 | /* | |
262 | * Construct the per-ereport payload, depending on which parameters are | |
263 | * passed in. | |
264 | */ | |
265 | ||
266 | /* | |
267 | * Generic payload members common to all ereports. | |
34dc7c2f | 268 | */ |
bcdb96a3 C |
269 | fm_payload_set(ereport, |
270 | FM_EREPORT_PAYLOAD_ZFS_POOL, DATA_TYPE_STRING, spa_name(spa), | |
271 | FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, DATA_TYPE_UINT64, spa_guid(spa), | |
177c91d0 DB |
272 | FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, DATA_TYPE_UINT64, |
273 | (uint64_t)spa_state(spa), | |
34dc7c2f | 274 | FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, DATA_TYPE_INT32, |
177c91d0 | 275 | (int32_t)spa_load_state(spa), NULL); |
b128c09f | 276 | |
a36cc8d2 | 277 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE, |
278 | DATA_TYPE_STRING, | |
279 | spa_get_failmode(spa) == ZIO_FAILURE_MODE_WAIT ? | |
280 | FM_EREPORT_FAILMODE_WAIT : | |
281 | spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE ? | |
282 | FM_EREPORT_FAILMODE_CONTINUE : FM_EREPORT_FAILMODE_PANIC, | |
283 | NULL); | |
34dc7c2f BB |
284 | |
285 | if (vd != NULL) { | |
286 | vdev_t *pvd = vd->vdev_parent; | |
cc92e9d0 | 287 | vdev_queue_t *vq = &vd->vdev_queue; |
904ea276 BB |
288 | vdev_stat_t *vs = &vd->vdev_stat; |
289 | vdev_t *spare_vd; | |
290 | uint64_t *spare_guids; | |
291 | char **spare_paths; | |
292 | int i, spare_count; | |
34dc7c2f BB |
293 | |
294 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, | |
295 | DATA_TYPE_UINT64, vd->vdev_guid, | |
296 | FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE, | |
297 | DATA_TYPE_STRING, vd->vdev_ops->vdev_op_type, NULL); | |
9babb374 | 298 | if (vd->vdev_path != NULL) |
34dc7c2f BB |
299 | fm_payload_set(ereport, |
300 | FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH, | |
301 | DATA_TYPE_STRING, vd->vdev_path, NULL); | |
9babb374 | 302 | if (vd->vdev_devid != NULL) |
34dc7c2f BB |
303 | fm_payload_set(ereport, |
304 | FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID, | |
305 | DATA_TYPE_STRING, vd->vdev_devid, NULL); | |
9babb374 BB |
306 | if (vd->vdev_fru != NULL) |
307 | fm_payload_set(ereport, | |
308 | FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU, | |
309 | DATA_TYPE_STRING, vd->vdev_fru, NULL); | |
6568379e TH |
310 | if (vd->vdev_enc_sysfs_path != NULL) |
311 | fm_payload_set(ereport, | |
312 | FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH, | |
313 | DATA_TYPE_STRING, vd->vdev_enc_sysfs_path, NULL); | |
32a9872b GW |
314 | if (vd->vdev_ashift) |
315 | fm_payload_set(ereport, | |
316 | FM_EREPORT_PAYLOAD_ZFS_VDEV_ASHIFT, | |
317 | DATA_TYPE_UINT64, vd->vdev_ashift, NULL); | |
34dc7c2f | 318 | |
cc92e9d0 GW |
319 | if (vq != NULL) { |
320 | fm_payload_set(ereport, | |
321 | FM_EREPORT_PAYLOAD_ZFS_VDEV_COMP_TS, | |
322 | DATA_TYPE_UINT64, vq->vq_io_complete_ts, NULL); | |
323 | fm_payload_set(ereport, | |
324 | FM_EREPORT_PAYLOAD_ZFS_VDEV_DELTA_TS, | |
325 | DATA_TYPE_UINT64, vq->vq_io_delta_ts, NULL); | |
326 | } | |
327 | ||
904ea276 BB |
328 | if (vs != NULL) { |
329 | fm_payload_set(ereport, | |
330 | FM_EREPORT_PAYLOAD_ZFS_VDEV_READ_ERRORS, | |
331 | DATA_TYPE_UINT64, vs->vs_read_errors, | |
332 | FM_EREPORT_PAYLOAD_ZFS_VDEV_WRITE_ERRORS, | |
333 | DATA_TYPE_UINT64, vs->vs_write_errors, | |
334 | FM_EREPORT_PAYLOAD_ZFS_VDEV_CKSUM_ERRORS, | |
335 | DATA_TYPE_UINT64, vs->vs_checksum_errors, NULL); | |
336 | } | |
337 | ||
34dc7c2f BB |
338 | if (pvd != NULL) { |
339 | fm_payload_set(ereport, | |
340 | FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID, | |
341 | DATA_TYPE_UINT64, pvd->vdev_guid, | |
342 | FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE, | |
343 | DATA_TYPE_STRING, pvd->vdev_ops->vdev_op_type, | |
344 | NULL); | |
345 | if (pvd->vdev_path) | |
346 | fm_payload_set(ereport, | |
347 | FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH, | |
348 | DATA_TYPE_STRING, pvd->vdev_path, NULL); | |
349 | if (pvd->vdev_devid) | |
350 | fm_payload_set(ereport, | |
351 | FM_EREPORT_PAYLOAD_ZFS_PARENT_DEVID, | |
352 | DATA_TYPE_STRING, pvd->vdev_devid, NULL); | |
353 | } | |
904ea276 BB |
354 | |
355 | spare_count = spa->spa_spares.sav_count; | |
356 | spare_paths = kmem_zalloc(sizeof (char *) * spare_count, | |
79c76d5b | 357 | KM_SLEEP); |
904ea276 | 358 | spare_guids = kmem_zalloc(sizeof (uint64_t) * spare_count, |
79c76d5b | 359 | KM_SLEEP); |
904ea276 BB |
360 | |
361 | for (i = 0; i < spare_count; i++) { | |
362 | spare_vd = spa->spa_spares.sav_vdevs[i]; | |
363 | if (spare_vd) { | |
364 | spare_paths[i] = spare_vd->vdev_path; | |
365 | spare_guids[i] = spare_vd->vdev_guid; | |
366 | } | |
367 | } | |
368 | ||
369 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_PATHS, | |
370 | DATA_TYPE_STRING_ARRAY, spare_count, spare_paths, | |
371 | FM_EREPORT_PAYLOAD_ZFS_VDEV_SPARE_GUIDS, | |
372 | DATA_TYPE_UINT64_ARRAY, spare_count, spare_guids, NULL); | |
373 | ||
374 | kmem_free(spare_guids, sizeof (uint64_t) * spare_count); | |
375 | kmem_free(spare_paths, sizeof (char *) * spare_count); | |
34dc7c2f BB |
376 | } |
377 | ||
378 | if (zio != NULL) { | |
379 | /* | |
380 | * Payload common to all I/Os. | |
381 | */ | |
382 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_ERR, | |
383 | DATA_TYPE_INT32, zio->io_error, NULL); | |
312c07ed BB |
384 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_FLAGS, |
385 | DATA_TYPE_INT32, zio->io_flags, NULL); | |
9dcb9719 BB |
386 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_STAGE, |
387 | DATA_TYPE_UINT32, zio->io_stage, NULL); | |
388 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_PIPELINE, | |
389 | DATA_TYPE_UINT32, zio->io_pipeline, NULL); | |
a69052be BB |
390 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELAY, |
391 | DATA_TYPE_UINT64, zio->io_delay, NULL); | |
cc92e9d0 GW |
392 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_TIMESTAMP, |
393 | DATA_TYPE_UINT64, zio->io_timestamp, NULL); | |
cc92e9d0 GW |
394 | fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELTA, |
395 | DATA_TYPE_UINT64, zio->io_delta, NULL); | |
34dc7c2f BB |
396 | |
397 | /* | |
398 | * If the 'size' parameter is non-zero, it indicates this is a | |
399 | * RAID-Z or other I/O where the physical offset and length are | |
400 | * provided for us, instead of within the zio_t. | |
401 | */ | |
402 | if (vd != NULL) { | |
403 | if (size) | |
404 | fm_payload_set(ereport, | |
405 | FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET, | |
406 | DATA_TYPE_UINT64, stateoroffset, | |
407 | FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE, | |
408 | DATA_TYPE_UINT64, size, NULL); | |
409 | else | |
410 | fm_payload_set(ereport, | |
411 | FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET, | |
412 | DATA_TYPE_UINT64, zio->io_offset, | |
413 | FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE, | |
414 | DATA_TYPE_UINT64, zio->io_size, NULL); | |
415 | } | |
34dc7c2f BB |
416 | } else if (vd != NULL) { |
417 | /* | |
418 | * If we have a vdev but no zio, this is a device fault, and the | |
419 | * 'stateoroffset' parameter indicates the previous state of the | |
420 | * vdev. | |
421 | */ | |
422 | fm_payload_set(ereport, | |
423 | FM_EREPORT_PAYLOAD_ZFS_PREV_STATE, | |
424 | DATA_TYPE_UINT64, stateoroffset, NULL); | |
425 | } | |
428870ff | 426 | |
b5256303 TC |
427 | /* |
428 | * Payload for I/Os with corresponding logical information. | |
429 | */ | |
430 | if (zb != NULL && (zio == NULL || zio->io_logical != NULL)) | |
431 | fm_payload_set(ereport, | |
432 | FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJSET, | |
433 | DATA_TYPE_UINT64, zb->zb_objset, | |
434 | FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJECT, | |
435 | DATA_TYPE_UINT64, zb->zb_object, | |
436 | FM_EREPORT_PAYLOAD_ZFS_ZIO_LEVEL, | |
437 | DATA_TYPE_INT64, zb->zb_level, | |
438 | FM_EREPORT_PAYLOAD_ZFS_ZIO_BLKID, | |
439 | DATA_TYPE_UINT64, zb->zb_blkid, NULL); | |
440 | ||
34dc7c2f BB |
441 | mutex_exit(&spa->spa_errlist_lock); |
442 | ||
428870ff BB |
443 | *ereport_out = ereport; |
444 | *detector_out = detector; | |
445 | } | |
446 | ||
447 | /* if it's <= 128 bytes, save the corruption directly */ | |
448 | #define ZFM_MAX_INLINE (128 / sizeof (uint64_t)) | |
449 | ||
450 | #define MAX_RANGES 16 | |
451 | ||
452 | typedef struct zfs_ecksum_info { | |
453 | /* histograms of set and cleared bits by bit number in a 64-bit word */ | |
1b18c6d7 AG |
454 | uint32_t zei_histogram_set[sizeof (uint64_t) * NBBY]; |
455 | uint32_t zei_histogram_cleared[sizeof (uint64_t) * NBBY]; | |
428870ff BB |
456 | |
457 | /* inline arrays of bits set and cleared. */ | |
458 | uint64_t zei_bits_set[ZFM_MAX_INLINE]; | |
459 | uint64_t zei_bits_cleared[ZFM_MAX_INLINE]; | |
460 | ||
461 | /* | |
462 | * for each range, the number of bits set and cleared. The Hamming | |
463 | * distance between the good and bad buffers is the sum of them all. | |
464 | */ | |
465 | uint32_t zei_range_sets[MAX_RANGES]; | |
466 | uint32_t zei_range_clears[MAX_RANGES]; | |
467 | ||
468 | struct zei_ranges { | |
469 | uint32_t zr_start; | |
470 | uint32_t zr_end; | |
471 | } zei_ranges[MAX_RANGES]; | |
472 | ||
473 | size_t zei_range_count; | |
474 | uint32_t zei_mingap; | |
475 | uint32_t zei_allowed_mingap; | |
476 | ||
477 | } zfs_ecksum_info_t; | |
478 | ||
479 | static void | |
1b18c6d7 | 480 | update_histogram(uint64_t value_arg, uint32_t *hist, uint32_t *count) |
428870ff BB |
481 | { |
482 | size_t i; | |
483 | size_t bits = 0; | |
484 | uint64_t value = BE_64(value_arg); | |
485 | ||
486 | /* We store the bits in big-endian (largest-first) order */ | |
487 | for (i = 0; i < 64; i++) { | |
488 | if (value & (1ull << i)) { | |
cf232b53 | 489 | hist[63 - i]++; |
428870ff BB |
490 | ++bits; |
491 | } | |
492 | } | |
493 | /* update the count of bits changed */ | |
494 | *count += bits; | |
495 | } | |
496 | ||
497 | /* | |
498 | * We've now filled up the range array, and need to increase "mingap" and | |
499 | * shrink the range list accordingly. zei_mingap is always the smallest | |
500 | * distance between array entries, so we set the new_allowed_gap to be | |
501 | * one greater than that. We then go through the list, joining together | |
502 | * any ranges which are closer than the new_allowed_gap. | |
503 | * | |
504 | * By construction, there will be at least one. We also update zei_mingap | |
505 | * to the new smallest gap, to prepare for our next invocation. | |
506 | */ | |
507 | static void | |
26685276 | 508 | zei_shrink_ranges(zfs_ecksum_info_t *eip) |
428870ff BB |
509 | { |
510 | uint32_t mingap = UINT32_MAX; | |
511 | uint32_t new_allowed_gap = eip->zei_mingap + 1; | |
512 | ||
513 | size_t idx, output; | |
514 | size_t max = eip->zei_range_count; | |
515 | ||
516 | struct zei_ranges *r = eip->zei_ranges; | |
517 | ||
518 | ASSERT3U(eip->zei_range_count, >, 0); | |
519 | ASSERT3U(eip->zei_range_count, <=, MAX_RANGES); | |
520 | ||
521 | output = idx = 0; | |
522 | while (idx < max - 1) { | |
523 | uint32_t start = r[idx].zr_start; | |
524 | uint32_t end = r[idx].zr_end; | |
525 | ||
526 | while (idx < max - 1) { | |
26685276 | 527 | idx++; |
428870ff | 528 | |
1c27024e DB |
529 | uint32_t nstart = r[idx].zr_start; |
530 | uint32_t nend = r[idx].zr_end; | |
531 | ||
532 | uint32_t gap = nstart - end; | |
428870ff BB |
533 | if (gap < new_allowed_gap) { |
534 | end = nend; | |
535 | continue; | |
536 | } | |
537 | if (gap < mingap) | |
538 | mingap = gap; | |
539 | break; | |
540 | } | |
541 | r[output].zr_start = start; | |
542 | r[output].zr_end = end; | |
543 | output++; | |
544 | } | |
545 | ASSERT3U(output, <, eip->zei_range_count); | |
546 | eip->zei_range_count = output; | |
547 | eip->zei_mingap = mingap; | |
548 | eip->zei_allowed_mingap = new_allowed_gap; | |
549 | } | |
550 | ||
551 | static void | |
26685276 | 552 | zei_add_range(zfs_ecksum_info_t *eip, int start, int end) |
428870ff BB |
553 | { |
554 | struct zei_ranges *r = eip->zei_ranges; | |
555 | size_t count = eip->zei_range_count; | |
556 | ||
557 | if (count >= MAX_RANGES) { | |
26685276 | 558 | zei_shrink_ranges(eip); |
428870ff BB |
559 | count = eip->zei_range_count; |
560 | } | |
561 | if (count == 0) { | |
562 | eip->zei_mingap = UINT32_MAX; | |
563 | eip->zei_allowed_mingap = 1; | |
564 | } else { | |
565 | int gap = start - r[count - 1].zr_end; | |
566 | ||
567 | if (gap < eip->zei_allowed_mingap) { | |
568 | r[count - 1].zr_end = end; | |
569 | return; | |
570 | } | |
571 | if (gap < eip->zei_mingap) | |
572 | eip->zei_mingap = gap; | |
573 | } | |
574 | r[count].zr_start = start; | |
575 | r[count].zr_end = end; | |
576 | eip->zei_range_count++; | |
577 | } | |
578 | ||
579 | static size_t | |
26685276 | 580 | zei_range_total_size(zfs_ecksum_info_t *eip) |
428870ff BB |
581 | { |
582 | struct zei_ranges *r = eip->zei_ranges; | |
583 | size_t count = eip->zei_range_count; | |
584 | size_t result = 0; | |
585 | size_t idx; | |
586 | ||
587 | for (idx = 0; idx < count; idx++) | |
588 | result += (r[idx].zr_end - r[idx].zr_start); | |
589 | ||
590 | return (result); | |
591 | } | |
592 | ||
593 | static zfs_ecksum_info_t * | |
594 | annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info, | |
84c07ada | 595 | const abd_t *goodabd, const abd_t *badabd, size_t size, |
428870ff BB |
596 | boolean_t drop_if_identical) |
597 | { | |
84c07ada GN |
598 | const uint64_t *good; |
599 | const uint64_t *bad; | |
428870ff BB |
600 | |
601 | uint64_t allset = 0; | |
602 | uint64_t allcleared = 0; | |
603 | ||
604 | size_t nui64s = size / sizeof (uint64_t); | |
605 | ||
606 | size_t inline_size; | |
607 | int no_inline = 0; | |
608 | size_t idx; | |
609 | size_t range; | |
610 | ||
611 | size_t offset = 0; | |
612 | ssize_t start = -1; | |
613 | ||
79c76d5b | 614 | zfs_ecksum_info_t *eip = kmem_zalloc(sizeof (*eip), KM_SLEEP); |
428870ff BB |
615 | |
616 | /* don't do any annotation for injected checksum errors */ | |
617 | if (info != NULL && info->zbc_injected) | |
618 | return (eip); | |
619 | ||
620 | if (info != NULL && info->zbc_has_cksum) { | |
621 | fm_payload_set(ereport, | |
622 | FM_EREPORT_PAYLOAD_ZFS_CKSUM_EXPECTED, | |
623 | DATA_TYPE_UINT64_ARRAY, | |
624 | sizeof (info->zbc_expected) / sizeof (uint64_t), | |
625 | (uint64_t *)&info->zbc_expected, | |
626 | FM_EREPORT_PAYLOAD_ZFS_CKSUM_ACTUAL, | |
627 | DATA_TYPE_UINT64_ARRAY, | |
628 | sizeof (info->zbc_actual) / sizeof (uint64_t), | |
629 | (uint64_t *)&info->zbc_actual, | |
630 | FM_EREPORT_PAYLOAD_ZFS_CKSUM_ALGO, | |
631 | DATA_TYPE_STRING, | |
632 | info->zbc_checksum_name, | |
633 | NULL); | |
634 | ||
635 | if (info->zbc_byteswapped) { | |
636 | fm_payload_set(ereport, | |
637 | FM_EREPORT_PAYLOAD_ZFS_CKSUM_BYTESWAP, | |
638 | DATA_TYPE_BOOLEAN, 1, | |
639 | NULL); | |
640 | } | |
641 | } | |
642 | ||
84c07ada | 643 | if (badabd == NULL || goodabd == NULL) |
428870ff BB |
644 | return (eip); |
645 | ||
1b18c6d7 | 646 | ASSERT3U(nui64s, <=, UINT32_MAX); |
428870ff BB |
647 | ASSERT3U(size, ==, nui64s * sizeof (uint64_t)); |
648 | ASSERT3U(size, <=, SPA_MAXBLOCKSIZE); | |
649 | ASSERT3U(size, <=, UINT32_MAX); | |
650 | ||
84c07ada GN |
651 | good = (const uint64_t *) abd_borrow_buf_copy((abd_t *)goodabd, size); |
652 | bad = (const uint64_t *) abd_borrow_buf_copy((abd_t *)badabd, size); | |
653 | ||
428870ff BB |
654 | /* build up the range list by comparing the two buffers. */ |
655 | for (idx = 0; idx < nui64s; idx++) { | |
656 | if (good[idx] == bad[idx]) { | |
657 | if (start == -1) | |
658 | continue; | |
659 | ||
26685276 | 660 | zei_add_range(eip, start, idx); |
428870ff BB |
661 | start = -1; |
662 | } else { | |
663 | if (start != -1) | |
664 | continue; | |
665 | ||
666 | start = idx; | |
667 | } | |
668 | } | |
669 | if (start != -1) | |
26685276 | 670 | zei_add_range(eip, start, idx); |
428870ff BB |
671 | |
672 | /* See if it will fit in our inline buffers */ | |
26685276 | 673 | inline_size = zei_range_total_size(eip); |
428870ff BB |
674 | if (inline_size > ZFM_MAX_INLINE) |
675 | no_inline = 1; | |
676 | ||
677 | /* | |
678 | * If there is no change and we want to drop if the buffers are | |
679 | * identical, do so. | |
680 | */ | |
681 | if (inline_size == 0 && drop_if_identical) { | |
682 | kmem_free(eip, sizeof (*eip)); | |
84c07ada GN |
683 | abd_return_buf((abd_t *)goodabd, (void *)good, size); |
684 | abd_return_buf((abd_t *)badabd, (void *)bad, size); | |
428870ff BB |
685 | return (NULL); |
686 | } | |
687 | ||
688 | /* | |
689 | * Now walk through the ranges, filling in the details of the | |
690 | * differences. Also convert our uint64_t-array offsets to byte | |
691 | * offsets. | |
692 | */ | |
693 | for (range = 0; range < eip->zei_range_count; range++) { | |
694 | size_t start = eip->zei_ranges[range].zr_start; | |
695 | size_t end = eip->zei_ranges[range].zr_end; | |
696 | ||
697 | for (idx = start; idx < end; idx++) { | |
698 | uint64_t set, cleared; | |
699 | ||
700 | // bits set in bad, but not in good | |
701 | set = ((~good[idx]) & bad[idx]); | |
702 | // bits set in good, but not in bad | |
703 | cleared = (good[idx] & (~bad[idx])); | |
704 | ||
705 | allset |= set; | |
706 | allcleared |= cleared; | |
707 | ||
708 | if (!no_inline) { | |
709 | ASSERT3U(offset, <, inline_size); | |
710 | eip->zei_bits_set[offset] = set; | |
711 | eip->zei_bits_cleared[offset] = cleared; | |
712 | offset++; | |
713 | } | |
714 | ||
715 | update_histogram(set, eip->zei_histogram_set, | |
716 | &eip->zei_range_sets[range]); | |
717 | update_histogram(cleared, eip->zei_histogram_cleared, | |
718 | &eip->zei_range_clears[range]); | |
719 | } | |
720 | ||
721 | /* convert to byte offsets */ | |
722 | eip->zei_ranges[range].zr_start *= sizeof (uint64_t); | |
723 | eip->zei_ranges[range].zr_end *= sizeof (uint64_t); | |
724 | } | |
84c07ada GN |
725 | |
726 | abd_return_buf((abd_t *)goodabd, (void *)good, size); | |
727 | abd_return_buf((abd_t *)badabd, (void *)bad, size); | |
728 | ||
428870ff BB |
729 | eip->zei_allowed_mingap *= sizeof (uint64_t); |
730 | inline_size *= sizeof (uint64_t); | |
731 | ||
732 | /* fill in ereport */ | |
733 | fm_payload_set(ereport, | |
734 | FM_EREPORT_PAYLOAD_ZFS_BAD_OFFSET_RANGES, | |
735 | DATA_TYPE_UINT32_ARRAY, 2 * eip->zei_range_count, | |
736 | (uint32_t *)eip->zei_ranges, | |
737 | FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_MIN_GAP, | |
738 | DATA_TYPE_UINT32, eip->zei_allowed_mingap, | |
739 | FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_SETS, | |
740 | DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_sets, | |
741 | FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_CLEARS, | |
742 | DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_clears, | |
743 | NULL); | |
744 | ||
745 | if (!no_inline) { | |
746 | fm_payload_set(ereport, | |
747 | FM_EREPORT_PAYLOAD_ZFS_BAD_SET_BITS, | |
748 | DATA_TYPE_UINT8_ARRAY, | |
749 | inline_size, (uint8_t *)eip->zei_bits_set, | |
750 | FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_BITS, | |
751 | DATA_TYPE_UINT8_ARRAY, | |
752 | inline_size, (uint8_t *)eip->zei_bits_cleared, | |
753 | NULL); | |
754 | } else { | |
755 | fm_payload_set(ereport, | |
756 | FM_EREPORT_PAYLOAD_ZFS_BAD_SET_HISTOGRAM, | |
1b18c6d7 | 757 | DATA_TYPE_UINT32_ARRAY, |
428870ff BB |
758 | NBBY * sizeof (uint64_t), eip->zei_histogram_set, |
759 | FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_HISTOGRAM, | |
1b18c6d7 | 760 | DATA_TYPE_UINT32_ARRAY, |
428870ff BB |
761 | NBBY * sizeof (uint64_t), eip->zei_histogram_cleared, |
762 | NULL); | |
763 | } | |
764 | return (eip); | |
765 | } | |
766 | #endif | |
767 | ||
768 | void | |
b5256303 TC |
769 | zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, |
770 | zbookmark_phys_t *zb, zio_t *zio, uint64_t stateoroffset, uint64_t size) | |
428870ff BB |
771 | { |
772 | #ifdef _KERNEL | |
773 | nvlist_t *ereport = NULL; | |
774 | nvlist_t *detector = NULL; | |
775 | ||
17b43f96 GDN |
776 | if (zfs_is_ratelimiting_event(subclass, vd)) |
777 | return; | |
778 | ||
b5256303 TC |
779 | zfs_ereport_start(&ereport, &detector, subclass, spa, vd, |
780 | zb, zio, stateoroffset, size); | |
428870ff BB |
781 | |
782 | if (ereport == NULL) | |
783 | return; | |
784 | ||
26685276 BB |
785 | /* Cleanup is handled by the callback function */ |
786 | zfs_zevent_post(ereport, detector, zfs_zevent_post_cb); | |
34dc7c2f BB |
787 | #endif |
788 | } | |
789 | ||
428870ff | 790 | void |
b5256303 | 791 | zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, zbookmark_phys_t *zb, |
428870ff BB |
792 | struct zio *zio, uint64_t offset, uint64_t length, void *arg, |
793 | zio_bad_cksum_t *info) | |
794 | { | |
6078881a TH |
795 | zio_cksum_report_t *report; |
796 | ||
797 | ||
798 | #ifdef _KERNEL | |
799 | if (zfs_is_ratelimiting_event(FM_EREPORT_ZFS_CHECKSUM, vd)) | |
800 | return; | |
801 | #endif | |
802 | ||
803 | report = kmem_zalloc(sizeof (*report), KM_SLEEP); | |
428870ff BB |
804 | |
805 | if (zio->io_vsd != NULL) | |
806 | zio->io_vsd_ops->vsd_cksum_report(zio, report, arg); | |
807 | else | |
808 | zio_vsd_default_cksum_report(zio, report, arg); | |
809 | ||
810 | /* copy the checksum failure information if it was provided */ | |
811 | if (info != NULL) { | |
79c76d5b | 812 | report->zcr_ckinfo = kmem_zalloc(sizeof (*info), KM_SLEEP); |
428870ff BB |
813 | bcopy(info, report->zcr_ckinfo, sizeof (*info)); |
814 | } | |
815 | ||
816 | report->zcr_align = 1ULL << vd->vdev_top->vdev_ashift; | |
817 | report->zcr_length = length; | |
818 | ||
819 | #ifdef _KERNEL | |
820 | zfs_ereport_start(&report->zcr_ereport, &report->zcr_detector, | |
b5256303 | 821 | FM_EREPORT_ZFS_CHECKSUM, spa, vd, zb, zio, offset, length); |
428870ff BB |
822 | |
823 | if (report->zcr_ereport == NULL) { | |
0426c168 | 824 | zfs_ereport_free_checksum(report); |
428870ff BB |
825 | return; |
826 | } | |
827 | #endif | |
828 | ||
829 | mutex_enter(&spa->spa_errlist_lock); | |
830 | report->zcr_next = zio->io_logical->io_cksum_report; | |
831 | zio->io_logical->io_cksum_report = report; | |
832 | mutex_exit(&spa->spa_errlist_lock); | |
833 | } | |
834 | ||
835 | void | |
84c07ada GN |
836 | zfs_ereport_finish_checksum(zio_cksum_report_t *report, const abd_t *good_data, |
837 | const abd_t *bad_data, boolean_t drop_if_identical) | |
428870ff BB |
838 | { |
839 | #ifdef _KERNEL | |
0426c168 IH |
840 | zfs_ecksum_info_t *info; |
841 | ||
428870ff BB |
842 | info = annotate_ecksum(report->zcr_ereport, report->zcr_ckinfo, |
843 | good_data, bad_data, report->zcr_length, drop_if_identical); | |
428870ff | 844 | if (info != NULL) |
26685276 BB |
845 | zfs_zevent_post(report->zcr_ereport, |
846 | report->zcr_detector, zfs_zevent_post_cb); | |
0426c168 IH |
847 | else |
848 | zfs_zevent_post_cb(report->zcr_ereport, report->zcr_detector); | |
428870ff | 849 | |
428870ff | 850 | report->zcr_ereport = report->zcr_detector = NULL; |
428870ff BB |
851 | if (info != NULL) |
852 | kmem_free(info, sizeof (*info)); | |
853 | #endif | |
854 | } | |
855 | ||
856 | void | |
857 | zfs_ereport_free_checksum(zio_cksum_report_t *rpt) | |
858 | { | |
859 | #ifdef _KERNEL | |
860 | if (rpt->zcr_ereport != NULL) { | |
861 | fm_nvlist_destroy(rpt->zcr_ereport, | |
862 | FM_NVA_FREE); | |
863 | fm_nvlist_destroy(rpt->zcr_detector, | |
864 | FM_NVA_FREE); | |
865 | } | |
866 | #endif | |
867 | rpt->zcr_free(rpt->zcr_cbdata, rpt->zcr_cbinfo); | |
868 | ||
869 | if (rpt->zcr_ckinfo != NULL) | |
870 | kmem_free(rpt->zcr_ckinfo, sizeof (*rpt->zcr_ckinfo)); | |
871 | ||
872 | kmem_free(rpt, sizeof (*rpt)); | |
873 | } | |
874 | ||
428870ff BB |
875 | |
876 | void | |
b5256303 | 877 | zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd, zbookmark_phys_t *zb, |
428870ff | 878 | struct zio *zio, uint64_t offset, uint64_t length, |
84c07ada | 879 | const abd_t *good_data, const abd_t *bad_data, zio_bad_cksum_t *zbc) |
428870ff BB |
880 | { |
881 | #ifdef _KERNEL | |
882 | nvlist_t *ereport = NULL; | |
883 | nvlist_t *detector = NULL; | |
884 | zfs_ecksum_info_t *info; | |
885 | ||
b5256303 TC |
886 | zfs_ereport_start(&ereport, &detector, FM_EREPORT_ZFS_CHECKSUM, |
887 | spa, vd, zb, zio, offset, length); | |
428870ff BB |
888 | |
889 | if (ereport == NULL) | |
890 | return; | |
891 | ||
892 | info = annotate_ecksum(ereport, zbc, good_data, bad_data, length, | |
893 | B_FALSE); | |
894 | ||
26685276 BB |
895 | if (info != NULL) { |
896 | zfs_zevent_post(ereport, detector, zfs_zevent_post_cb); | |
428870ff | 897 | kmem_free(info, sizeof (*info)); |
26685276 | 898 | } |
428870ff BB |
899 | #endif |
900 | } | |
901 | ||
12fa0466 DE |
902 | /* |
903 | * The 'sysevent.fs.zfs.*' events are signals posted to notify user space of | |
904 | * change in the pool. All sysevents are listed in sys/sysevent/eventdefs.h | |
905 | * and are designed to be consumed by the ZFS Event Daemon (ZED). For | |
906 | * additional details refer to the zed(8) man page. | |
907 | */ | |
908 | nvlist_t * | |
909 | zfs_event_create(spa_t *spa, vdev_t *vd, const char *type, const char *name, | |
d02ca379 | 910 | nvlist_t *aux) |
34dc7c2f | 911 | { |
12fa0466 | 912 | nvlist_t *resource = NULL; |
34dc7c2f | 913 | #ifdef _KERNEL |
34dc7c2f BB |
914 | char class[64]; |
915 | ||
428870ff | 916 | if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT) |
12fa0466 | 917 | return (NULL); |
428870ff | 918 | |
34dc7c2f | 919 | if ((resource = fm_nvlist_create(NULL)) == NULL) |
12fa0466 | 920 | return (NULL); |
34dc7c2f | 921 | |
fb390aaf | 922 | (void) snprintf(class, sizeof (class), "%s.%s.%s", type, |
34dc7c2f | 923 | ZFS_ERROR_CLASS, name); |
904ea276 BB |
924 | VERIFY0(nvlist_add_uint8(resource, FM_VERSION, FM_RSRC_VERSION)); |
925 | VERIFY0(nvlist_add_string(resource, FM_CLASS, class)); | |
bcdb96a3 C |
926 | VERIFY0(nvlist_add_string(resource, |
927 | FM_EREPORT_PAYLOAD_ZFS_POOL, spa_name(spa))); | |
904ea276 BB |
928 | VERIFY0(nvlist_add_uint64(resource, |
929 | FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, spa_guid(spa))); | |
bcdb96a3 C |
930 | VERIFY0(nvlist_add_uint64(resource, |
931 | FM_EREPORT_PAYLOAD_ZFS_POOL_STATE, spa_state(spa))); | |
904ea276 BB |
932 | VERIFY0(nvlist_add_int32(resource, |
933 | FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, spa_load_state(spa))); | |
934 | ||
26685276 | 935 | if (vd) { |
904ea276 BB |
936 | VERIFY0(nvlist_add_uint64(resource, |
937 | FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vd->vdev_guid)); | |
938 | VERIFY0(nvlist_add_uint64(resource, | |
939 | FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE, vd->vdev_state)); | |
fb390aaf HR |
940 | if (vd->vdev_path != NULL) |
941 | VERIFY0(nvlist_add_string(resource, | |
942 | FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH, vd->vdev_path)); | |
943 | if (vd->vdev_devid != NULL) | |
944 | VERIFY0(nvlist_add_string(resource, | |
945 | FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID, vd->vdev_devid)); | |
946 | if (vd->vdev_fru != NULL) | |
947 | VERIFY0(nvlist_add_string(resource, | |
948 | FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU, vd->vdev_fru)); | |
6568379e TH |
949 | if (vd->vdev_enc_sysfs_path != NULL) |
950 | VERIFY0(nvlist_add_string(resource, | |
951 | FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH, | |
952 | vd->vdev_enc_sysfs_path)); | |
12fa0466 | 953 | } |
d02ca379 | 954 | |
12fa0466 DE |
955 | /* also copy any optional payload data */ |
956 | if (aux) { | |
957 | nvpair_t *elem = NULL; | |
958 | ||
959 | while ((elem = nvlist_next_nvpair(aux, elem)) != NULL) | |
960 | (void) nvlist_add_nvpair(resource, elem); | |
26685276 | 961 | } |
34dc7c2f | 962 | |
12fa0466 DE |
963 | #endif |
964 | return (resource); | |
965 | } | |
966 | ||
967 | static void | |
968 | zfs_post_common(spa_t *spa, vdev_t *vd, const char *type, const char *name, | |
969 | nvlist_t *aux) | |
970 | { | |
971 | #ifdef _KERNEL | |
972 | nvlist_t *resource; | |
973 | ||
974 | resource = zfs_event_create(spa, vd, type, name, aux); | |
975 | if (resource) | |
976 | zfs_zevent_post(resource, NULL, zfs_zevent_post_cb); | |
34dc7c2f BB |
977 | #endif |
978 | } | |
979 | ||
34dc7c2f BB |
980 | /* |
981 | * The 'resource.fs.zfs.removed' event is an internal signal that the given vdev | |
982 | * has been removed from the system. This will cause the DE to ignore any | |
983 | * recent I/O errors, inferring that they are due to the asynchronous device | |
984 | * removal. | |
985 | */ | |
986 | void | |
987 | zfs_post_remove(spa_t *spa, vdev_t *vd) | |
988 | { | |
d02ca379 | 989 | zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_REMOVED, NULL); |
34dc7c2f BB |
990 | } |
991 | ||
992 | /* | |
993 | * The 'resource.fs.zfs.autoreplace' event is an internal signal that the pool | |
994 | * has the 'autoreplace' property set, and therefore any broken vdevs will be | |
995 | * handled by higher level logic, and no vdev fault should be generated. | |
996 | */ | |
997 | void | |
998 | zfs_post_autoreplace(spa_t *spa, vdev_t *vd) | |
999 | { | |
d02ca379 | 1000 | zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_AUTOREPLACE, NULL); |
34dc7c2f | 1001 | } |
428870ff BB |
1002 | |
1003 | /* | |
1004 | * The 'resource.fs.zfs.statechange' event is an internal signal that the | |
1005 | * given vdev has transitioned its state to DEGRADED or HEALTHY. This will | |
1006 | * cause the retire agent to repair any outstanding fault management cases | |
1007 | * open because the device was not found (fault.fs.zfs.device). | |
1008 | */ | |
1009 | void | |
d02ca379 | 1010 | zfs_post_state_change(spa_t *spa, vdev_t *vd, uint64_t laststate) |
428870ff | 1011 | { |
d02ca379 DB |
1012 | #ifdef _KERNEL |
1013 | nvlist_t *aux; | |
1014 | ||
1015 | /* | |
1016 | * Add optional supplemental keys to payload | |
1017 | */ | |
1018 | aux = fm_nvlist_create(NULL); | |
1019 | if (vd && aux) { | |
1020 | if (vd->vdev_physpath) { | |
1021 | (void) nvlist_add_string(aux, | |
1022 | FM_EREPORT_PAYLOAD_ZFS_VDEV_PHYSPATH, | |
1023 | vd->vdev_physpath); | |
1024 | } | |
1bbd8770 TH |
1025 | if (vd->vdev_enc_sysfs_path) { |
1026 | (void) nvlist_add_string(aux, | |
1027 | FM_EREPORT_PAYLOAD_ZFS_VDEV_ENC_SYSFS_PATH, | |
1028 | vd->vdev_enc_sysfs_path); | |
1029 | } | |
1030 | ||
d02ca379 DB |
1031 | (void) nvlist_add_uint64(aux, |
1032 | FM_EREPORT_PAYLOAD_ZFS_VDEV_LASTSTATE, laststate); | |
1033 | } | |
1034 | ||
1035 | zfs_post_common(spa, vd, FM_RSRC_CLASS, FM_RESOURCE_STATECHANGE, | |
1036 | aux); | |
1037 | ||
1038 | if (aux) | |
1039 | fm_nvlist_destroy(aux, FM_NVA_FREE); | |
1040 | #endif | |
fb390aaf HR |
1041 | } |
1042 | ||
26685276 BB |
1043 | #if defined(_KERNEL) && defined(HAVE_SPL) |
1044 | EXPORT_SYMBOL(zfs_ereport_post); | |
1045 | EXPORT_SYMBOL(zfs_ereport_post_checksum); | |
1046 | EXPORT_SYMBOL(zfs_post_remove); | |
1047 | EXPORT_SYMBOL(zfs_post_autoreplace); | |
1048 | EXPORT_SYMBOL(zfs_post_state_change); | |
1049 | #endif /* _KERNEL */ |