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.
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.
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]
23 * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
28 #include <sys/spa_impl.h>
30 #include <sys/dsl_synctask.h>
31 #include <sys/dmu_tx.h>
32 #include <sys/dmu_objset.h>
33 #include <sys/dsl_dataset.h>
34 #include <sys/dsl_dir.h>
35 #include <sys/utsname.h>
36 #include <sys/sunddi.h>
38 #include "zfs_comutil.h"
40 #include <sys/cmn_err.h>
45 * Routines to manage the on-disk history log.
47 * The history log is stored as a dmu object containing
48 * <packed record length, record nvlist> tuples.
50 * Where "record nvlist" is a nvlist containing uint64_ts and strings, and
51 * "packed record length" is the packed length of the "record nvlist" stored
52 * as a little endian uint64_t.
54 * The log is implemented as a ring buffer, though the original creation
55 * of the pool ('zpool create') is never overwritten.
57 * The history log is tracked as object 'spa_t::spa_history'. The bonus buffer
58 * of 'spa_history' stores the offsets for logging/retrieving history as
59 * 'spa_history_phys_t'. 'sh_pool_create_len' is the ending offset in bytes of
60 * where the 'zpool create' record is stored. This allows us to never
61 * overwrite the original creation of the pool. 'sh_phys_max_off' is the
62 * physical ending offset in bytes of the log. This tells you the length of
63 * the buffer. 'sh_eof' is the logical EOF (in bytes). Whenever a record
64 * is added, 'sh_eof' is incremented by the the size of the record.
65 * 'sh_eof' is never decremented. 'sh_bof' is the logical BOF (in bytes).
66 * This is where the consumer should start reading from after reading in
67 * the 'zpool create' portion of the log.
69 * 'sh_records_lost' keeps track of how many records have been overwritten
70 * and permanently lost.
73 /* convert a logical offset to physical */
75 spa_history_log_to_phys(uint64_t log_off, spa_history_phys_t *shpp)
79 phys_len = shpp->sh_phys_max_off - shpp->sh_pool_create_len;
80 return ((log_off - shpp->sh_pool_create_len) % phys_len
81 + shpp->sh_pool_create_len);
85 spa_history_create_obj(spa_t *spa, dmu_tx_t *tx)
88 spa_history_phys_t *shpp;
89 objset_t *mos = spa->spa_meta_objset;
91 ASSERT(spa->spa_history == 0);
92 spa->spa_history = dmu_object_alloc(mos, DMU_OT_SPA_HISTORY,
93 SPA_OLD_MAXBLOCKSIZE, DMU_OT_SPA_HISTORY_OFFSETS,
94 sizeof (spa_history_phys_t), tx);
96 VERIFY(zap_add(mos, DMU_POOL_DIRECTORY_OBJECT,
97 DMU_POOL_HISTORY, sizeof (uint64_t), 1,
98 &spa->spa_history, tx) == 0);
100 VERIFY(0 == dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
101 ASSERT(dbp->db_size >= sizeof (spa_history_phys_t));
104 dmu_buf_will_dirty(dbp, tx);
107 * Figure out maximum size of history log. We set it at
108 * 0.1% of pool size, with a max of 1G and min of 128KB.
110 shpp->sh_phys_max_off =
111 metaslab_class_get_dspace(spa_normal_class(spa)) / 1000;
112 shpp->sh_phys_max_off = MIN(shpp->sh_phys_max_off, 1<<30);
113 shpp->sh_phys_max_off = MAX(shpp->sh_phys_max_off, 128<<10);
115 dmu_buf_rele(dbp, FTAG);
119 * Change 'sh_bof' to the beginning of the next record.
122 spa_history_advance_bof(spa_t *spa, spa_history_phys_t *shpp)
124 objset_t *mos = spa->spa_meta_objset;
125 uint64_t firstread, reclen, phys_bof;
126 char buf[sizeof (reclen)];
129 phys_bof = spa_history_log_to_phys(shpp->sh_bof, shpp);
130 firstread = MIN(sizeof (reclen), shpp->sh_phys_max_off - phys_bof);
132 if ((err = dmu_read(mos, spa->spa_history, phys_bof, firstread,
133 buf, DMU_READ_PREFETCH)) != 0)
135 if (firstread != sizeof (reclen)) {
136 if ((err = dmu_read(mos, spa->spa_history,
137 shpp->sh_pool_create_len, sizeof (reclen) - firstread,
138 buf + firstread, DMU_READ_PREFETCH)) != 0)
142 reclen = LE_64(*((uint64_t *)buf));
143 shpp->sh_bof += reclen + sizeof (reclen);
144 shpp->sh_records_lost++;
149 spa_history_write(spa_t *spa, void *buf, uint64_t len, spa_history_phys_t *shpp,
152 uint64_t firstwrite, phys_eof;
153 objset_t *mos = spa->spa_meta_objset;
156 ASSERT(MUTEX_HELD(&spa->spa_history_lock));
158 /* see if we need to reset logical BOF */
159 while (shpp->sh_phys_max_off - shpp->sh_pool_create_len -
160 (shpp->sh_eof - shpp->sh_bof) <= len) {
161 if ((err = spa_history_advance_bof(spa, shpp)) != 0) {
166 phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
167 firstwrite = MIN(len, shpp->sh_phys_max_off - phys_eof);
169 dmu_write(mos, spa->spa_history, phys_eof, firstwrite, buf, tx);
173 /* write out the rest at the beginning of physical file */
174 dmu_write(mos, spa->spa_history, shpp->sh_pool_create_len,
175 len, (char *)buf + firstwrite, tx);
182 spa_history_zone(void)
185 /* XXX: pr_hostname can be changed by default from within a jail! */
186 if (jailed(curthread->td_ucred))
187 return (curthread->td_ucred->cr_prison->pr_hostname);
193 * Write out a history event.
197 spa_history_log_sync(void *arg, dmu_tx_t *tx)
200 spa_t *spa = dmu_tx_pool(tx)->dp_spa;
201 objset_t *mos = spa->spa_meta_objset;
203 spa_history_phys_t *shpp;
206 char *record_packed = NULL;
210 * If we have an older pool that doesn't have a command
211 * history object, create it now.
213 mutex_enter(&spa->spa_history_lock);
214 if (!spa->spa_history)
215 spa_history_create_obj(spa, tx);
216 mutex_exit(&spa->spa_history_lock);
219 * Get the offset of where we need to write via the bonus buffer.
220 * Update the offset when the write completes.
222 VERIFY0(dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp));
225 dmu_buf_will_dirty(dbp, tx);
229 dmu_object_info_t doi;
230 dmu_object_info_from_db(dbp, &doi);
231 ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
235 fnvlist_add_uint64(nvl, ZPOOL_HIST_TIME, gethrestime_sec());
237 fnvlist_add_string(nvl, ZPOOL_HIST_HOST, utsname.nodename);
239 if (nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
240 zfs_dbgmsg("command: %s",
241 fnvlist_lookup_string(nvl, ZPOOL_HIST_CMD));
242 } else if (nvlist_exists(nvl, ZPOOL_HIST_INT_NAME)) {
243 if (nvlist_exists(nvl, ZPOOL_HIST_DSNAME)) {
244 zfs_dbgmsg("txg %lld %s %s (id %llu) %s",
245 fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
246 fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
247 fnvlist_lookup_string(nvl, ZPOOL_HIST_DSNAME),
248 fnvlist_lookup_uint64(nvl, ZPOOL_HIST_DSID),
249 fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
251 zfs_dbgmsg("txg %lld %s %s",
252 fnvlist_lookup_uint64(nvl, ZPOOL_HIST_TXG),
253 fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_NAME),
254 fnvlist_lookup_string(nvl, ZPOOL_HIST_INT_STR));
256 } else if (nvlist_exists(nvl, ZPOOL_HIST_IOCTL)) {
257 zfs_dbgmsg("ioctl %s",
258 fnvlist_lookup_string(nvl, ZPOOL_HIST_IOCTL));
261 record_packed = fnvlist_pack(nvl, &reclen);
263 mutex_enter(&spa->spa_history_lock);
265 /* write out the packed length as little endian */
266 le_len = LE_64((uint64_t)reclen);
267 ret = spa_history_write(spa, &le_len, sizeof (le_len), shpp, tx);
269 ret = spa_history_write(spa, record_packed, reclen, shpp, tx);
271 /* The first command is the create, which we keep forever */
272 if (ret == 0 && shpp->sh_pool_create_len == 0 &&
273 nvlist_exists(nvl, ZPOOL_HIST_CMD)) {
274 shpp->sh_pool_create_len = shpp->sh_bof = shpp->sh_eof;
277 mutex_exit(&spa->spa_history_lock);
278 fnvlist_pack_free(record_packed, reclen);
279 dmu_buf_rele(dbp, FTAG);
284 * Write out a history event.
287 spa_history_log(spa_t *spa, const char *msg)
290 nvlist_t *nvl = fnvlist_alloc();
292 fnvlist_add_string(nvl, ZPOOL_HIST_CMD, msg);
293 err = spa_history_log_nvl(spa, nvl);
299 spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
305 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY)
308 if (spa_version(spa) < SPA_VERSION_ZPOOL_HISTORY || !spa_writeable(spa))
309 return (SET_ERROR(EINVAL));
311 tx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
312 err = dmu_tx_assign(tx, TXG_WAIT);
318 nvarg = fnvlist_dup(nvl);
319 if (spa_history_zone() != NULL) {
320 fnvlist_add_string(nvarg, ZPOOL_HIST_ZONE,
323 fnvlist_add_uint64(nvarg, ZPOOL_HIST_WHO, crgetruid(CRED()));
325 /* Kick this off asynchronously; errors are ignored. */
326 dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync,
327 nvarg, 0, ZFS_SPACE_CHECK_NONE, tx);
330 /* spa_history_log_sync will free nvl */
336 * Read out the command history.
339 spa_history_get(spa_t *spa, uint64_t *offp, uint64_t *len, char *buf)
341 objset_t *mos = spa->spa_meta_objset;
343 uint64_t read_len, phys_read_off, phys_eof;
344 uint64_t leftover = 0;
345 spa_history_phys_t *shpp;
349 * If the command history doesn't exist (older pool),
350 * that's ok, just return ENOENT.
352 if (!spa->spa_history)
353 return (SET_ERROR(ENOENT));
356 * The history is logged asynchronously, so when they request
357 * the first chunk of history, make sure everything has been
358 * synced to disk so that we get it.
360 if (*offp == 0 && spa_writeable(spa))
361 txg_wait_synced(spa_get_dsl(spa), 0);
363 if ((err = dmu_bonus_hold(mos, spa->spa_history, FTAG, &dbp)) != 0)
369 dmu_object_info_t doi;
370 dmu_object_info_from_db(dbp, &doi);
371 ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_SPA_HISTORY_OFFSETS);
375 mutex_enter(&spa->spa_history_lock);
376 phys_eof = spa_history_log_to_phys(shpp->sh_eof, shpp);
378 if (*offp < shpp->sh_pool_create_len) {
379 /* read in just the zpool create history */
380 phys_read_off = *offp;
381 read_len = MIN(*len, shpp->sh_pool_create_len -
385 * Need to reset passed in offset to BOF if the passed in
386 * offset has since been overwritten.
388 *offp = MAX(*offp, shpp->sh_bof);
389 phys_read_off = spa_history_log_to_phys(*offp, shpp);
392 * Read up to the minimum of what the user passed down or
393 * the EOF (physical or logical). If we hit physical EOF,
394 * use 'leftover' to read from the physical BOF.
396 if (phys_read_off <= phys_eof) {
397 read_len = MIN(*len, phys_eof - phys_read_off);
400 shpp->sh_phys_max_off - phys_read_off);
401 if (phys_read_off + *len > shpp->sh_phys_max_off) {
402 leftover = MIN(*len - read_len,
403 phys_eof - shpp->sh_pool_create_len);
408 /* offset for consumer to use next */
409 *offp += read_len + leftover;
411 /* tell the consumer how much you actually read */
412 *len = read_len + leftover;
415 mutex_exit(&spa->spa_history_lock);
416 dmu_buf_rele(dbp, FTAG);
420 err = dmu_read(mos, spa->spa_history, phys_read_off, read_len, buf,
422 if (leftover && err == 0) {
423 err = dmu_read(mos, spa->spa_history, shpp->sh_pool_create_len,
424 leftover, buf + read_len, DMU_READ_PREFETCH);
426 mutex_exit(&spa->spa_history_lock);
428 dmu_buf_rele(dbp, FTAG);
433 * The nvlist will be consumed by this call.
436 log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
437 dmu_tx_t *tx, const char *fmt, va_list adx)
443 * If this is part of creating a pool, not everything is
444 * initialized yet, so don't bother logging the internal events.
445 * Likewise if the pool is not writeable.
447 if (tx->tx_txg == TXG_INITIAL || !spa_writeable(spa)) {
454 msg = kmem_alloc(vsnprintf(NULL, 0, fmt, adx) + 1, KM_SLEEP);
455 (void) vsprintf(msg, fmt, adx2);
456 fnvlist_add_string(nvl, ZPOOL_HIST_INT_STR, msg);
461 fnvlist_add_string(nvl, ZPOOL_HIST_INT_NAME, operation);
462 fnvlist_add_uint64(nvl, ZPOOL_HIST_TXG, tx->tx_txg);
464 if (dmu_tx_is_syncing(tx)) {
465 spa_history_log_sync(nvl, tx);
467 dsl_sync_task_nowait(spa_get_dsl(spa),
468 spa_history_log_sync, nvl, 0, ZFS_SPACE_CHECK_NONE, tx);
470 /* spa_history_log_sync() will free nvl */
474 spa_history_log_internal(spa_t *spa, const char *operation,
475 dmu_tx_t *tx, const char *fmt, ...)
480 /* create a tx if we didn't get one */
482 htx = dmu_tx_create_dd(spa_get_dsl(spa)->dp_mos_dir);
483 if (dmu_tx_assign(htx, TXG_WAIT) != 0) {
490 log_internal(fnvlist_alloc(), operation, spa, htx, fmt, adx);
493 /* if we didn't get a tx from the caller, commit the one we made */
499 spa_history_log_internal_ds(dsl_dataset_t *ds, const char *operation,
500 dmu_tx_t *tx, const char *fmt, ...)
503 char namebuf[MAXNAMELEN];
504 nvlist_t *nvl = fnvlist_alloc();
508 dsl_dataset_name(ds, namebuf);
509 fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf);
510 fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID, ds->ds_object);
513 log_internal(nvl, operation, dsl_dataset_get_spa(ds), tx, fmt, adx);
518 spa_history_log_internal_dd(dsl_dir_t *dd, const char *operation,
519 dmu_tx_t *tx, const char *fmt, ...)
522 char namebuf[MAXNAMELEN];
523 nvlist_t *nvl = fnvlist_alloc();
527 dsl_dir_name(dd, namebuf);
528 fnvlist_add_string(nvl, ZPOOL_HIST_DSNAME, namebuf);
529 fnvlist_add_uint64(nvl, ZPOOL_HIST_DSID,
530 dd->dd_phys->dd_head_dataset_obj);
533 log_internal(nvl, operation, dd->dd_pool->dp_spa, tx, fmt, adx);
538 spa_history_log_version(spa_t *spa, const char *operation)
540 spa_history_log_internal(spa, operation, NULL,
541 "pool version %llu; software version %llu/%d; uts %s %s %s %s",
542 (u_longlong_t)spa_version(spa), SPA_VERSION, ZPL_VERSION,
543 utsname.nodename, utsname.release, utsname.version,