]>
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 | ||
22 | /* | |
428870ff | 23 | * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved. |
eca7b760 | 24 | * Copyright (c) 2011, 2015 by Delphix. All rights reserved. |
12fa0466 DE |
25 | * Copyright (c) 2014 Integros [integros.com] |
26 | * Copyright 2017 Joyent, Inc. | |
34dc7c2f BB |
27 | */ |
28 | ||
34dc7c2f BB |
29 | #include <sys/spa.h> |
30 | #include <sys/spa_impl.h> | |
31 | #include <sys/zap.h> | |
32 | #include <sys/dsl_synctask.h> | |
33 | #include <sys/dmu_tx.h> | |
34 | #include <sys/dmu_objset.h> | |
6f1ffb06 MA |
35 | #include <sys/dsl_dataset.h> |
36 | #include <sys/dsl_dir.h> | |
34dc7c2f BB |
37 | #include <sys/cmn_err.h> |
38 | #include <sys/sunddi.h> | |
6f1ffb06 | 39 | #include <sys/cred.h> |
428870ff | 40 | #include "zfs_comutil.h" |
34dc7c2f BB |
41 | #ifdef _KERNEL |
42 | #include <sys/zone.h> | |
43 | #endif | |
44 | ||
45 | /* | |
46 | * Routines to manage the on-disk history log. | |
47 | * | |
48 | * The history log is stored as a dmu object containing | |
49 | * <packed record length, record nvlist> tuples. | |
50 | * | |
d5884c34 | 51 | * Where "record nvlist" is an nvlist containing uint64_ts and strings, and |
34dc7c2f BB |
52 | * "packed record length" is the packed length of the "record nvlist" stored |
53 | * as a little endian uint64_t. | |
54 | * | |
55 | * The log is implemented as a ring buffer, though the original creation | |
56 | * of the pool ('zpool create') is never overwritten. | |
57 | * | |
58 | * The history log is tracked as object 'spa_t::spa_history'. The bonus buffer | |
59 | * of 'spa_history' stores the offsets for logging/retrieving history as | |
60 | * 'spa_history_phys_t'. 'sh_pool_create_len' is the ending offset in bytes of | |
61 | * where the 'zpool create' record is stored. This allows us to never | |
62 | * overwrite the original creation of the pool. 'sh_phys_max_off' is the | |
63 | * physical ending offset in bytes of the log. This tells you the length of | |
64 | * the buffer. 'sh_eof' is the logical EOF (in bytes). Whenever a record | |
65 | * is added, 'sh_eof' is incremented by the the size of the record. | |
66 | * 'sh_eof' is never decremented. 'sh_bof' is the logical BOF (in bytes). | |
67 | * This is where the consumer should start reading from after reading in | |
68 | * the 'zpool create' portion of the log. | |
69 | * | |
70 | * 'sh_records_lost' keeps track of how many records have been overwritten | |
71 | * and permanently lost. | |
72 | */ | |
73 | ||
74 | /* convert a logical offset to physical */ | |
75 | static uint64_t | |
76 | spa_history_log_to_phys(uint64_t log_off, spa_history_phys_t *shpp) | |
77 | { | |
78 | uint64_t phys_len; | |
79 | ||
80 | phys_len = shpp->sh_phys_max_off - shpp->sh_pool_create_len; | |
81 | return ((log_off - shpp->sh_pool_create_len) % phys_len | |
82 | + shpp->sh_pool_create_len); | |
83 | } | |
84 | ||
85 | void | |
86 | spa_history_create_obj(spa_t *spa, dmu_tx_t *tx) | |
87 | { | |
88 | dmu_buf_t *dbp; | |
89 | spa_history_phys_t *shpp; | |
90 | objset_t *mos = spa->spa_meta_objset; | |
91 | ||
d5e024cb | 92 | ASSERT0(spa->spa_history); |
34dc7c2f | 93 | spa->spa_history = dmu_object_alloc(mos, DMU_OT_SPA_HISTORY, |
f1512ee6 | 94 | SPA_OLD_MAXBLOCKSIZE, DMU_OT_SPA_HISTORY_OFFSETS, |
34dc7c2f BB |
95 | sizeof (spa_history_phys_t), tx); |
96 | ||
d5e024cb | 97 | VERIFY0(zap_add(mos, DMU_POOL_DIRECTORY_OBJECT, |
34dc7c2f | 98 | DMU_POOL_HISTORY, sizeof (uint64_t), 1, |
d5e024cb | 99 | &spa->spa_history, tx)); |
34dc7c2f | 100 | |
d5e024cb BB |
101 | VERIFY0(dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)); |
102 | ASSERT3U(dbp->db_size, >=, sizeof (spa_history_phys_t)); | |
34dc7c2f BB |
103 | |
104 | shpp = dbp->db_data; | |
105 | dmu_buf_will_dirty(dbp, tx); | |
106 | ||
107 | /* | |
108 | * Figure out maximum size of history log. We set it at | |
330d06f9 | 109 | * 0.1% of pool size, with a max of 1G and min of 128KB. |
34dc7c2f | 110 | */ |
428870ff | 111 | shpp->sh_phys_max_off = |
330d06f9 MA |
112 | metaslab_class_get_dspace(spa_normal_class(spa)) / 1000; |
113 | shpp->sh_phys_max_off = MIN(shpp->sh_phys_max_off, 1<<30); | |
34dc7c2f BB |
114 | shpp->sh_phys_max_off = MAX(shpp->sh_phys_max_off, 128<<10); |
115 | ||
116 | dmu_buf_rele(dbp, FTAG); | |
117 | } | |
118 | ||
119 | /* | |
120 | * Change 'sh_bof' to the beginning of the next record. | |
121 | */ | |
122 | static int | |
123 | spa_history_advance_bof(spa_t *spa, spa_history_phys_t *shpp) | |
124 | { | |
125 | objset_t *mos = spa->spa_meta_objset; | |
126 | uint64_t firstread, reclen, phys_bof; | |
127 | char buf[sizeof (reclen)]; | |
128 | int err; | |
129 | ||
130 | phys_bof = spa_history_log_to_phys(shpp->sh_bof, shpp); | |
131 | firstread = MIN(sizeof (reclen), shpp->sh_phys_max_off - phys_bof); | |
132 | ||
133 | if ((err = dmu_read(mos, spa->spa_history, phys_bof, firstread, | |
9babb374 | 134 | buf, DMU_READ_PREFETCH)) != 0) |
34dc7c2f BB |
135 | return (err); |
136 | if (firstread != sizeof (reclen)) { | |
137 | if ((err = dmu_read(mos, spa->spa_history, | |
138 | shpp->sh_pool_create_len, sizeof (reclen) - firstread, | |
9babb374 | 139 | buf + firstread, DMU_READ_PREFETCH)) != 0) |
34dc7c2f BB |
140 | return (err); |
141 | } | |
142 | ||
143 | reclen = LE_64(*((uint64_t *)buf)); | |
144 | shpp->sh_bof += reclen + sizeof (reclen); | |
145 | shpp->sh_records_lost++; | |
146 | return (0); | |
147 | } | |
148 | ||
149 | static int | |
150 | spa_history_write(spa_t *spa, void *buf, uint64_t len, spa_history_phys_t *shpp, | |
151 | dmu_tx_t *tx) | |
152 | { | |
153 | uint64_t firstwrite, phys_eof; | |
154 | objset_t *mos = spa->spa_meta_objset; | |
155 | int err; | |
156 | ||
157 | ASSERT(MUTEX_HELD(&spa->spa_history_lock)); | |
158 | ||
159 | /* see if we need to reset logical BOF */ | |
160 | while (shpp->sh_phys_max_off - shpp->sh_pool_create_len - | |
161 | (shpp->sh_eof - shpp->sh_bof) <= len) { | |
162 | if ((err = spa_history_advance_bof(spa, shpp)) != 0) { | |
163 | return (err); | |
164 | } | |
165 | } | |
166 | ||
167 | phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp); | |
168 | firstwrite = MIN(len, shpp->sh_phys_max_off - phys_eof); | |
169 | shpp->sh_eof += len; | |
170 | dmu_write(mos, spa->spa_history, phys_eof, firstwrite, buf, tx); | |
171 | ||
172 | len -= firstwrite; | |
173 | if (len > 0) { | |
174 | /* write out the rest at the beginning of physical file */ | |
175 | dmu_write(mos, spa->spa_history, shpp->sh_pool_create_len, | |
176 | len, (char *)buf + firstwrite, tx); | |
177 | } | |
178 | ||
179 | return (0); | |
180 | } | |
181 | ||
182 | static char * | |
0bc8fd78 | 183 | spa_history_zone(void) |
34dc7c2f BB |
184 | { |
185 | #ifdef _KERNEL | |
c28b2279 | 186 | return ("linux"); |
34dc7c2f | 187 | #else |
6f1ffb06 | 188 | return (NULL); |
34dc7c2f BB |
189 | #endif |
190 | } | |
191 | ||
12fa0466 DE |
192 | /* |
193 | * Post a history sysevent. | |
194 | * | |
195 | * The nvlist_t* passed into this function will be transformed into a new | |
196 | * nvlist where: | |
197 | * | |
198 | * 1. Nested nvlists will be flattened to a single level | |
199 | * 2. Keys will have their names normalized (to remove any problematic | |
200 | * characters, such as whitespace) | |
201 | * | |
202 | * The nvlist_t passed into this function will duplicated and should be freed | |
203 | * by caller. | |
204 | * | |
205 | */ | |
206 | static void | |
207 | spa_history_log_notify(spa_t *spa, nvlist_t *nvl) | |
208 | { | |
209 | nvlist_t *hist_nvl = fnvlist_alloc(); | |
210 | uint64_t uint64; | |
211 | char *string; | |
212 | ||
213 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_CMD, &string) == 0) | |
214 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_CMD, string); | |
215 | ||
216 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME, &string) == 0) | |
217 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_INT_NAME, string); | |
218 | ||
219 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_ZONE, &string) == 0) | |
220 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_ZONE, string); | |
221 | ||
222 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_HOST, &string) == 0) | |
223 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_HOST, string); | |
224 | ||
225 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME, &string) == 0) | |
226 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_DSNAME, string); | |
227 | ||
228 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR, &string) == 0) | |
229 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_INT_STR, string); | |
230 | ||
231 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL, &string) == 0) | |
232 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_IOCTL, string); | |
233 | ||
234 | if (nvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME, &string) == 0) | |
235 | fnvlist_add_string(hist_nvl, ZFS_EV_HIST_INT_NAME, string); | |
236 | ||
237 | if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID, &uint64) == 0) | |
238 | fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_DSID, uint64); | |
239 | ||
240 | if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG, &uint64) == 0) | |
241 | fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_TXG, uint64); | |
242 | ||
243 | if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_TIME, &uint64) == 0) | |
244 | fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_TIME, uint64); | |
245 | ||
246 | if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_WHO, &uint64) == 0) | |
247 | fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_WHO, uint64); | |
248 | ||
249 | if (nvlist_lookup_uint64(nvl, ZPOOL_HIST_INT_EVENT, &uint64) == 0) | |
250 | fnvlist_add_uint64(hist_nvl, ZFS_EV_HIST_INT_EVENT, uint64); | |
251 | ||
252 | spa_event_notify(spa, NULL, hist_nvl, ESC_ZFS_HISTORY_EVENT); | |
253 | ||
254 | nvlist_free(hist_nvl); | |
255 | } | |
256 | ||
34dc7c2f BB |
257 | /* |
258 | * Write out a history event. | |
259 | */ | |
428870ff | 260 | /*ARGSUSED*/ |
34dc7c2f | 261 | static void |
13fe0198 | 262 | spa_history_log_sync(void *arg, dmu_tx_t *tx) |
34dc7c2f | 263 | { |
13fe0198 MA |
264 | nvlist_t *nvl = arg; |
265 | spa_t *spa = dmu_tx_pool(tx)->dp_spa; | |
34dc7c2f BB |
266 | objset_t *mos = spa->spa_meta_objset; |
267 | dmu_buf_t *dbp; | |
268 | spa_history_phys_t *shpp; | |
269 | size_t reclen; | |
270 | uint64_t le_len; | |
34dc7c2f BB |
271 | char *record_packed = NULL; |
272 | int ret; | |
273 | ||
274 | /* | |
275 | * If we have an older pool that doesn't have a command | |
276 | * history object, create it now. | |
277 | */ | |
278 | mutex_enter(&spa->spa_history_lock); | |
279 | if (!spa->spa_history) | |
280 | spa_history_create_obj(spa, tx); | |
281 | mutex_exit(&spa->spa_history_lock); | |
282 | ||
283 | /* | |
284 | * Get the offset of where we need to write via the bonus buffer. | |
285 | * Update the offset when the write completes. | |
286 | */ | |
13fe0198 | 287 | VERIFY0(dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)); |
34dc7c2f BB |
288 | shpp = dbp->db_data; |
289 | ||
290 | dmu_buf_will_dirty(dbp, tx); | |
291 | ||
292 | #ifdef ZFS_DEBUG | |
293 | { | |
294 | dmu_object_info_t doi; | |
295 | dmu_object_info_from_db(dbp, &doi); | |
296 | ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS); | |
297 | } | |
298 | #endif | |
299 | ||
6f1ffb06 | 300 | fnvlist_add_uint64(nvl, ZPOOL_HIST_TIME, gethrestime_sec()); |
f0e324f2 BB |
301 | fnvlist_add_string(nvl, ZPOOL_HIST_HOST, utsname()->nodename); |
302 | ||
6f1ffb06 MA |
303 | if (nvlist_exists(nvl, ZPOOL_HIST_CMD)) { |
304 | zfs_dbgmsg("command: %s", | |
305 | fnvlist_lookup_string(nvl, ZPOOL_HIST_CMD)); | |
306 | } else if (nvlist_exists(nvl, ZPOOL_HIST_INT_NAME)) { | |
307 | if (nvlist_exists(nvl, ZPOOL_HIST_DSNAME)) { | |
308 | zfs_dbgmsg("txg %lld %s %s (id %llu) %s", | |
309 | fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG), | |
310 | fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME), | |
311 | fnvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME), | |
312 | fnvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID), | |
313 | fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR)); | |
314 | } else { | |
315 | zfs_dbgmsg("txg %lld %s %s", | |
316 | fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG), | |
317 | fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME), | |
318 | fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR)); | |
319 | } | |
12fa0466 DE |
320 | /* |
321 | * The history sysevent is posted only for internal history | |
322 | * messages to show what has happened, not how it happened. For | |
323 | * example, the following command: | |
324 | * | |
325 | * # zfs destroy -r tank/foo | |
326 | * | |
327 | * will result in one sysevent posted per dataset that is | |
328 | * destroyed as a result of the command - which could be more | |
329 | * than one event in total. By contrast, if the sysevent was | |
330 | * posted as a result of the ZPOOL_HIST_CMD key being present | |
331 | * it would result in only one sysevent being posted with the | |
332 | * full command line arguments, requiring the consumer to know | |
333 | * how to parse and understand zfs(1M) command invocations. | |
334 | */ | |
335 | spa_history_log_notify(spa, nvl); | |
6f1ffb06 MA |
336 | } else if (nvlist_exists(nvl, ZPOOL_HIST_IOCTL)) { |
337 | zfs_dbgmsg("ioctl %s", | |
338 | fnvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL)); | |
34dc7c2f BB |
339 | } |
340 | ||
6f1ffb06 | 341 | VERIFY3U(nvlist_pack(nvl, &record_packed, &reclen, NV_ENCODE_NATIVE, |
79c76d5b | 342 | KM_SLEEP), ==, 0); |
34dc7c2f BB |
343 | |
344 | mutex_enter(&spa->spa_history_lock); | |
34dc7c2f BB |
345 | |
346 | /* write out the packed length as little endian */ | |
347 | le_len = LE_64((uint64_t)reclen); | |
348 | ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx); | |
349 | if (!ret) | |
350 | ret = spa_history_write(spa, record_packed, reclen, shpp, tx); | |
351 | ||
6f1ffb06 MA |
352 | /* The first command is the create, which we keep forever */ |
353 | if (ret == 0 && shpp->sh_pool_create_len == 0 && | |
354 | nvlist_exists(nvl, ZPOOL_HIST_CMD)) { | |
355 | shpp->sh_pool_create_len = shpp->sh_bof = shpp->sh_eof; | |
34dc7c2f BB |
356 | } |
357 | ||
358 | mutex_exit(&spa->spa_history_lock); | |
6f1ffb06 | 359 | fnvlist_pack_free(record_packed, reclen); |
34dc7c2f | 360 | dmu_buf_rele(dbp, FTAG); |
6f1ffb06 | 361 | fnvlist_free(nvl); |
34dc7c2f BB |
362 | } |
363 | ||
364 | /* | |
365 | * Write out a history event. | |
366 | */ | |
367 | int | |
6f1ffb06 MA |
368 | spa_history_log(spa_t *spa, const char *msg) |
369 | { | |
370 | int err; | |
79c76d5b | 371 | nvlist_t *nvl = fnvlist_alloc(); |
6f1ffb06 MA |
372 | |
373 | fnvlist_add_string(nvl, ZPOOL_HIST_CMD, msg); | |
374 | err = spa_history_log_nvl(spa, nvl); | |
375 | fnvlist_free(nvl); | |
376 | return (err); | |
377 | } | |
378 | ||
379 | int | |
380 | spa_history_log_nvl(spa_t *spa, nvlist_t *nvl) | |
34dc7c2f | 381 | { |
428870ff BB |
382 | int err = 0; |
383 | dmu_tx_t *tx; | |
b5256303 | 384 | nvlist_t *nvarg, *in_nvl = NULL; |
34dc7c2f | 385 | |
6f1ffb06 | 386 | if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa)) |
2e528b49 | 387 | return (SET_ERROR(EINVAL)); |
34dc7c2f | 388 | |
b5256303 TC |
389 | err = nvlist_lookup_nvlist(nvl, ZPOOL_HIST_INPUT_NVL, &in_nvl); |
390 | if (err == 0) { | |
391 | (void) nvlist_remove_all(in_nvl, ZPOOL_HIDDEN_ARGS); | |
392 | } | |
393 | ||
428870ff BB |
394 | tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); |
395 | err = dmu_tx_assign(tx, TXG_WAIT); | |
396 | if (err) { | |
397 | dmu_tx_abort(tx); | |
398 | return (err); | |
399 | } | |
400 | ||
79c76d5b | 401 | VERIFY0(nvlist_dup(nvl, &nvarg, KM_SLEEP)); |
6f1ffb06 MA |
402 | if (spa_history_zone() != NULL) { |
403 | fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE, | |
404 | spa_history_zone()); | |
405 | } | |
406 | fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED())); | |
428870ff BB |
407 | |
408 | /* Kick this off asynchronously; errors are ignored. */ | |
13fe0198 | 409 | dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync, |
3d45fdd6 | 410 | nvarg, 0, ZFS_SPACE_CHECK_NONE, tx); |
428870ff BB |
411 | dmu_tx_commit(tx); |
412 | ||
6f1ffb06 | 413 | /* spa_history_log_sync will free nvl */ |
428870ff | 414 | return (err); |
6f1ffb06 | 415 | |
34dc7c2f BB |
416 | } |
417 | ||
418 | /* | |
419 | * Read out the command history. | |
420 | */ | |
421 | int | |
422 | spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf) | |
423 | { | |
424 | objset_t *mos = spa->spa_meta_objset; | |
425 | dmu_buf_t *dbp; | |
426 | uint64_t read_len, phys_read_off, phys_eof; | |
427 | uint64_t leftover = 0; | |
428 | spa_history_phys_t *shpp; | |
429 | int err; | |
430 | ||
431 | /* | |
6f1ffb06 | 432 | * If the command history doesn't exist (older pool), |
34dc7c2f BB |
433 | * that's ok, just return ENOENT. |
434 | */ | |
435 | if (!spa->spa_history) | |
2e528b49 | 436 | return (SET_ERROR(ENOENT)); |
34dc7c2f | 437 | |
428870ff BB |
438 | /* |
439 | * The history is logged asynchronously, so when they request | |
440 | * the first chunk of history, make sure everything has been | |
441 | * synced to disk so that we get it. | |
442 | */ | |
443 | if (*offp == 0 && spa_writeable(spa)) | |
444 | txg_wait_synced(spa_get_dsl(spa), 0); | |
445 | ||
34dc7c2f BB |
446 | if ((err = dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)) != 0) |
447 | return (err); | |
448 | shpp = dbp->db_data; | |
449 | ||
450 | #ifdef ZFS_DEBUG | |
451 | { | |
452 | dmu_object_info_t doi; | |
453 | dmu_object_info_from_db(dbp, &doi); | |
454 | ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS); | |
455 | } | |
456 | #endif | |
457 | ||
458 | mutex_enter(&spa->spa_history_lock); | |
459 | phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp); | |
460 | ||
461 | if (*offp < shpp->sh_pool_create_len) { | |
462 | /* read in just the zpool create history */ | |
463 | phys_read_off = *offp; | |
464 | read_len = MIN(*len, shpp->sh_pool_create_len - | |
465 | phys_read_off); | |
466 | } else { | |
467 | /* | |
468 | * Need to reset passed in offset to BOF if the passed in | |
469 | * offset has since been overwritten. | |
470 | */ | |
471 | *offp = MAX(*offp, shpp->sh_bof); | |
472 | phys_read_off = spa_history_log_to_phys(*offp, shpp); | |
473 | ||
474 | /* | |
475 | * Read up to the minimum of what the user passed down or | |
476 | * the EOF (physical or logical). If we hit physical EOF, | |
477 | * use 'leftover' to read from the physical BOF. | |
478 | */ | |
479 | if (phys_read_off <= phys_eof) { | |
480 | read_len = MIN(*len, phys_eof - phys_read_off); | |
481 | } else { | |
482 | read_len = MIN(*len, | |
483 | shpp->sh_phys_max_off - phys_read_off); | |
484 | if (phys_read_off + *len > shpp->sh_phys_max_off) { | |
485 | leftover = MIN(*len - read_len, | |
486 | phys_eof - shpp->sh_pool_create_len); | |
487 | } | |
488 | } | |
489 | } | |
490 | ||
491 | /* offset for consumer to use next */ | |
492 | *offp += read_len + leftover; | |
493 | ||
494 | /* tell the consumer how much you actually read */ | |
495 | *len = read_len + leftover; | |
496 | ||
497 | if (read_len == 0) { | |
498 | mutex_exit(&spa->spa_history_lock); | |
499 | dmu_buf_rele(dbp, FTAG); | |
500 | return (0); | |
501 | } | |
502 | ||
9babb374 BB |
503 | err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf, |
504 | DMU_READ_PREFETCH); | |
34dc7c2f BB |
505 | if (leftover && err == 0) { |
506 | err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len, | |
9babb374 | 507 | leftover, buf + read_len, DMU_READ_PREFETCH); |
34dc7c2f BB |
508 | } |
509 | mutex_exit(&spa->spa_history_lock); | |
510 | ||
511 | dmu_buf_rele(dbp, FTAG); | |
512 | return (err); | |
513 | } | |
514 | ||
6f1ffb06 MA |
515 | /* |
516 | * The nvlist will be consumed by this call. | |
517 | */ | |
45d1cae3 | 518 | static void |
6f1ffb06 | 519 | log_internal(nvlist_t *nvl, const char *operation, spa_t *spa, |
428870ff | 520 | dmu_tx_t *tx, const char *fmt, va_list adx) |
34dc7c2f | 521 | { |
6f1ffb06 | 522 | char *msg; |
34dc7c2f | 523 | |
b128c09f BB |
524 | /* |
525 | * If this is part of creating a pool, not everything is | |
526 | * initialized yet, so don't bother logging the internal events. | |
6f1ffb06 | 527 | * Likewise if the pool is not writeable. |
b128c09f | 528 | */ |
d5e024cb | 529 | if (spa_is_initializing(spa) || !spa_writeable(spa)) { |
6f1ffb06 | 530 | fnvlist_free(nvl); |
b128c09f | 531 | return; |
6f1ffb06 | 532 | } |
b128c09f | 533 | |
841c9d43 | 534 | msg = kmem_vasprintf(fmt, adx); |
6f1ffb06 | 535 | fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg); |
841c9d43 | 536 | strfree(msg); |
6f1ffb06 MA |
537 | |
538 | fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation); | |
539 | fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg); | |
34dc7c2f BB |
540 | |
541 | if (dmu_tx_is_syncing(tx)) { | |
13fe0198 | 542 | spa_history_log_sync(nvl, tx); |
34dc7c2f | 543 | } else { |
13fe0198 | 544 | dsl_sync_task_nowait(spa_get_dsl(spa), |
3d45fdd6 | 545 | spa_history_log_sync, nvl, 0, ZFS_SPACE_CHECK_NONE, tx); |
34dc7c2f | 546 | } |
6f1ffb06 | 547 | /* spa_history_log_sync() will free nvl */ |
34dc7c2f | 548 | } |
45d1cae3 BB |
549 | |
550 | void | |
6f1ffb06 | 551 | spa_history_log_internal(spa_t *spa, const char *operation, |
428870ff | 552 | dmu_tx_t *tx, const char *fmt, ...) |
45d1cae3 BB |
553 | { |
554 | dmu_tx_t *htx = tx; | |
555 | va_list adx; | |
556 | ||
557 | /* create a tx if we didn't get one */ | |
558 | if (tx == NULL) { | |
559 | htx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir); | |
560 | if (dmu_tx_assign(htx, TXG_WAIT) != 0) { | |
561 | dmu_tx_abort(htx); | |
562 | return; | |
563 | } | |
564 | } | |
565 | ||
566 | va_start(adx, fmt); | |
79c76d5b | 567 | log_internal(fnvlist_alloc(), operation, spa, htx, fmt, adx); |
45d1cae3 BB |
568 | va_end(adx); |
569 | ||
570 | /* if we didn't get a tx from the caller, commit the one we made */ | |
571 | if (tx == NULL) | |
572 | dmu_tx_commit(htx); | |
573 | } | |
574 | ||
575 | void | |
6f1ffb06 MA |
576 | spa_history_log_internal_ds(dsl_dataset_t *ds, const char *operation, |
577 | dmu_tx_t *tx, const char *fmt, ...) | |
578 | { | |
579 | va_list adx; | |
eca7b760 | 580 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
79c76d5b | 581 | nvlist_t *nvl = fnvlist_alloc(); |
6f1ffb06 MA |
582 | |
583 | ASSERT(tx != NULL); | |
584 | ||
585 | dsl_dataset_name(ds, namebuf); | |
6f1ffb06 MA |
586 | fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf); |
587 | fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID, ds->ds_object); | |
588 | ||
589 | va_start(adx, fmt); | |
590 | log_internal(nvl, operation, dsl_dataset_get_spa(ds), tx, fmt, adx); | |
591 | va_end(adx); | |
592 | } | |
593 | ||
594 | void | |
595 | spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation, | |
596 | dmu_tx_t *tx, const char *fmt, ...) | |
597 | { | |
598 | va_list adx; | |
eca7b760 | 599 | char namebuf[ZFS_MAX_DATASET_NAME_LEN]; |
79c76d5b | 600 | nvlist_t *nvl = fnvlist_alloc(); |
6f1ffb06 MA |
601 | |
602 | ASSERT(tx != NULL); | |
603 | ||
604 | dsl_dir_name(dd, namebuf); | |
6f1ffb06 MA |
605 | fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf); |
606 | fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID, | |
d683ddbb | 607 | dsl_dir_phys(dd)->dd_head_dataset_obj); |
6f1ffb06 MA |
608 | |
609 | va_start(adx, fmt); | |
610 | log_internal(nvl, operation, dd->dd_pool->dp_spa, tx, fmt, adx); | |
611 | va_end(adx); | |
612 | } | |
613 | ||
614 | void | |
d5e024cb | 615 | spa_history_log_version(spa_t *spa, const char *operation, dmu_tx_t *tx) |
45d1cae3 | 616 | { |
f0e324f2 BB |
617 | utsname_t *u = utsname(); |
618 | ||
d5e024cb | 619 | spa_history_log_internal(spa, operation, tx, |
bff245dd | 620 | "pool version %llu; software version %llu/%llu; uts %s %s %s %s", |
13fe0198 | 621 | (u_longlong_t)spa_version(spa), SPA_VERSION, ZPL_VERSION, |
f0e324f2 | 622 | u->nodename, u->release, u->version, u->machine); |
45d1cae3 | 623 | } |
c28b2279 | 624 | |
93ce2b4c | 625 | #if defined(_KERNEL) |
c28b2279 BB |
626 | EXPORT_SYMBOL(spa_history_create_obj); |
627 | EXPORT_SYMBOL(spa_history_get); | |
628 | EXPORT_SYMBOL(spa_history_log); | |
629 | EXPORT_SYMBOL(spa_history_log_internal); | |
630 | EXPORT_SYMBOL(spa_history_log_version); | |
631 | #endif |