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]
22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
26 #include <sys/dsl_pool.h>
27 #include <sys/dsl_dataset.h>
28 #include <sys/dsl_dir.h>
29 #include <sys/dsl_synctask.h>
30 #include <sys/dmu_tx.h>
31 #include <sys/dmu_objset.h>
35 #include <sys/zfs_context.h>
36 #include <sys/fs/zfs.h>
37 #include <sys/zfs_znode.h>
38 #include <sys/spa_impl.h>
40 int zfs_no_write_throttle = 0;
41 int zfs_write_limit_shift = 3; /* 1/8th of physical memory */
42 int zfs_txg_synctime = 5; /* target secs to sync a txg */
44 uint64_t zfs_write_limit_min = 32 << 20; /* min write limit is 32MB */
45 uint64_t zfs_write_limit_max = 0; /* max data payload per txg */
46 uint64_t zfs_write_limit_inflated = 0;
47 uint64_t zfs_write_limit_override = 0;
49 kmutex_t zfs_write_limit_lock;
51 static pgcnt_t old_physmem = 0;
54 dsl_pool_open_special_dir(dsl_pool_t *dp, const char *name, dsl_dir_t **ddp)
59 err = zap_lookup(dp->dp_meta_objset,
60 dp->dp_root_dir->dd_phys->dd_child_dir_zapobj,
61 name, sizeof (obj), 1, &obj);
65 return (dsl_dir_open_obj(dp, obj, name, dp, ddp));
69 dsl_pool_open_impl(spa_t *spa, uint64_t txg)
72 blkptr_t *bp = spa_get_rootblkptr(spa);
74 dp = kmem_zalloc(sizeof (dsl_pool_t), KM_SLEEP);
76 dp->dp_meta_rootbp = *bp;
77 rw_init(&dp->dp_config_rwlock, NULL, RW_DEFAULT, NULL);
78 dp->dp_write_limit = zfs_write_limit_min;
81 txg_list_create(&dp->dp_dirty_datasets,
82 offsetof(dsl_dataset_t, ds_dirty_link));
83 txg_list_create(&dp->dp_dirty_dirs,
84 offsetof(dsl_dir_t, dd_dirty_link));
85 txg_list_create(&dp->dp_sync_tasks,
86 offsetof(dsl_sync_task_group_t, dstg_node));
87 list_create(&dp->dp_synced_datasets, sizeof (dsl_dataset_t),
88 offsetof(dsl_dataset_t, ds_synced_link));
90 mutex_init(&dp->dp_lock, NULL, MUTEX_DEFAULT, NULL);
91 mutex_init(&dp->dp_scrub_cancel_lock, NULL, MUTEX_DEFAULT, NULL);
93 dp->dp_vnrele_taskq = taskq_create("zfs_vn_rele_taskq", 1, minclsyspri,
100 dsl_pool_open(spa_t *spa, uint64_t txg, dsl_pool_t **dpp)
103 dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
108 rw_enter(&dp->dp_config_rwlock, RW_WRITER);
109 err = dmu_objset_open_impl(spa, NULL, &dp->dp_meta_rootbp, &osi);
112 dp->dp_meta_objset = &osi->os;
114 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
115 DMU_POOL_ROOT_DATASET, sizeof (uint64_t), 1,
116 &dp->dp_root_dir_obj);
120 err = dsl_dir_open_obj(dp, dp->dp_root_dir_obj,
121 NULL, dp, &dp->dp_root_dir);
125 err = dsl_pool_open_special_dir(dp, MOS_DIR_NAME, &dp->dp_mos_dir);
129 if (spa_version(spa) >= SPA_VERSION_ORIGIN) {
130 err = dsl_pool_open_special_dir(dp, ORIGIN_DIR_NAME, &dd);
133 err = dsl_dataset_hold_obj(dp, dd->dd_phys->dd_head_dataset_obj,
136 err = dsl_dataset_hold_obj(dp,
137 ds->ds_phys->ds_prev_snap_obj, dp,
138 &dp->dp_origin_snap);
139 dsl_dataset_rele(ds, FTAG);
141 dsl_dir_close(dd, dp);
146 /* get scrub status */
147 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
148 DMU_POOL_SCRUB_FUNC, sizeof (uint32_t), 1,
151 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
152 DMU_POOL_SCRUB_QUEUE, sizeof (uint64_t), 1,
153 &dp->dp_scrub_queue_obj);
156 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
157 DMU_POOL_SCRUB_MIN_TXG, sizeof (uint64_t), 1,
158 &dp->dp_scrub_min_txg);
161 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
162 DMU_POOL_SCRUB_MAX_TXG, sizeof (uint64_t), 1,
163 &dp->dp_scrub_max_txg);
166 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
167 DMU_POOL_SCRUB_BOOKMARK, sizeof (uint64_t), 4,
168 &dp->dp_scrub_bookmark);
171 err = zap_lookup(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
172 DMU_POOL_SCRUB_ERRORS, sizeof (uint64_t), 1,
173 &spa->spa_scrub_errors);
176 if (spa_version(spa) < SPA_VERSION_DSL_SCRUB) {
178 * A new-type scrub was in progress on an old
179 * pool. Restart from the beginning, since the
180 * old software may have changed the pool in the
183 dsl_pool_scrub_restart(dp);
187 * It's OK if there is no scrub in progress (and if
188 * there was an I/O error, ignore it).
194 rw_exit(&dp->dp_config_rwlock);
204 dsl_pool_close(dsl_pool_t *dp)
206 /* drop our references from dsl_pool_open() */
209 * Since we held the origin_snap from "syncing" context (which
210 * includes pool-opening context), it actually only got a "ref"
211 * and not a hold, so just drop that here.
213 if (dp->dp_origin_snap)
214 dsl_dataset_drop_ref(dp->dp_origin_snap, dp);
216 dsl_dir_close(dp->dp_mos_dir, dp);
218 dsl_dir_close(dp->dp_root_dir, dp);
220 /* undo the dmu_objset_open_impl(mos) from dsl_pool_open() */
221 if (dp->dp_meta_objset)
222 dmu_objset_evict(NULL, dp->dp_meta_objset->os);
224 txg_list_destroy(&dp->dp_dirty_datasets);
225 txg_list_destroy(&dp->dp_dirty_dirs);
226 list_destroy(&dp->dp_synced_datasets);
228 arc_flush(dp->dp_spa);
230 rw_destroy(&dp->dp_config_rwlock);
231 mutex_destroy(&dp->dp_lock);
232 mutex_destroy(&dp->dp_scrub_cancel_lock);
233 taskq_destroy(dp->dp_vnrele_taskq);
235 kmem_free(dp->dp_blkstats, sizeof (zfs_all_blkstats_t));
236 kmem_free(dp, sizeof (dsl_pool_t));
240 dsl_pool_create(spa_t *spa, nvlist_t *zplprops, uint64_t txg)
243 dsl_pool_t *dp = dsl_pool_open_impl(spa, txg);
244 dmu_tx_t *tx = dmu_tx_create_assigned(dp, txg);
249 /* create and open the MOS (meta-objset) */
250 dp->dp_meta_objset = &dmu_objset_create_impl(spa,
251 NULL, &dp->dp_meta_rootbp, DMU_OST_META, tx)->os;
253 /* create the pool directory */
254 err = zap_create_claim(dp->dp_meta_objset, DMU_POOL_DIRECTORY_OBJECT,
255 DMU_OT_OBJECT_DIRECTORY, DMU_OT_NONE, 0, tx);
256 ASSERT3U(err, ==, 0);
258 /* create and open the root dir */
259 dp->dp_root_dir_obj = dsl_dir_create_sync(dp, NULL, NULL, tx);
260 VERIFY(0 == dsl_dir_open_obj(dp, dp->dp_root_dir_obj,
261 NULL, dp, &dp->dp_root_dir));
263 /* create and open the meta-objset dir */
264 (void) dsl_dir_create_sync(dp, dp->dp_root_dir, MOS_DIR_NAME, tx);
265 VERIFY(0 == dsl_pool_open_special_dir(dp,
266 MOS_DIR_NAME, &dp->dp_mos_dir));
268 if (spa_version(spa) >= SPA_VERSION_DSL_SCRUB)
269 dsl_pool_create_origin(dp, tx);
271 /* create the root dataset */
272 dsobj = dsl_dataset_create_sync_dd(dp->dp_root_dir, NULL, 0, tx);
274 /* create the root objset */
275 VERIFY(0 == dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
276 osip = dmu_objset_create_impl(dp->dp_spa, ds,
277 dsl_dataset_get_blkptr(ds), DMU_OST_ZFS, tx);
279 zfs_create_fs(&osip->os, kcred, zplprops, tx);
281 dsl_dataset_rele(ds, FTAG);
289 dsl_pool_sync(dsl_pool_t *dp, uint64_t txg)
295 dsl_sync_task_group_t *dstg;
296 objset_impl_t *mosi = dp->dp_meta_objset->os;
297 hrtime_t start, write_time;
298 uint64_t data_written;
301 tx = dmu_tx_create_assigned(dp, txg);
303 dp->dp_read_overhead = 0;
306 zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
307 while (ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) {
309 * We must not sync any non-MOS datasets twice, because
310 * we may have taken a snapshot of them. However, we
311 * may sync newly-created datasets on pass 2.
313 ASSERT(!list_link_active(&ds->ds_synced_link));
314 list_insert_tail(&dp->dp_synced_datasets, ds);
315 dsl_dataset_sync(ds, zio, tx);
317 DTRACE_PROBE(pool_sync__1setup);
320 write_time = gethrtime() - start;
322 DTRACE_PROBE(pool_sync__2rootzio);
324 for (ds = list_head(&dp->dp_synced_datasets); ds;
325 ds = list_next(&dp->dp_synced_datasets, ds))
326 dmu_objset_do_userquota_callbacks(ds->ds_user_ptr, tx);
329 * Sync the datasets again to push out the changes due to
330 * userquota updates. This must be done before we process the
331 * sync tasks, because that could cause a snapshot of a dataset
332 * whose ds_bp will be rewritten when we do this 2nd sync.
334 zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
335 while (ds = txg_list_remove(&dp->dp_dirty_datasets, txg)) {
336 ASSERT(list_link_active(&ds->ds_synced_link));
337 dmu_buf_rele(ds->ds_dbuf, ds);
338 dsl_dataset_sync(ds, zio, tx);
342 while (dstg = txg_list_remove(&dp->dp_sync_tasks, txg)) {
344 * No more sync tasks should have been added while we
347 ASSERT(spa_sync_pass(dp->dp_spa) == 1);
348 dsl_sync_task_group_sync(dstg, tx);
350 DTRACE_PROBE(pool_sync__3task);
353 while (dd = txg_list_remove(&dp->dp_dirty_dirs, txg))
354 dsl_dir_sync(dd, tx);
355 write_time += gethrtime() - start;
357 if (spa_sync_pass(dp->dp_spa) == 1)
358 dsl_pool_scrub_sync(dp, tx);
361 if (list_head(&mosi->os_dirty_dnodes[txg & TXG_MASK]) != NULL ||
362 list_head(&mosi->os_free_dnodes[txg & TXG_MASK]) != NULL) {
363 zio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED);
364 dmu_objset_sync(mosi, zio, tx);
367 dprintf_bp(&dp->dp_meta_rootbp, "meta objset rootbp is %s", "");
368 spa_set_rootblkptr(dp->dp_spa, &dp->dp_meta_rootbp);
370 write_time += gethrtime() - start;
371 DTRACE_PROBE2(pool_sync__4io, hrtime_t, write_time,
372 hrtime_t, dp->dp_read_overhead);
373 write_time -= dp->dp_read_overhead;
377 data_written = dp->dp_space_towrite[txg & TXG_MASK];
378 dp->dp_space_towrite[txg & TXG_MASK] = 0;
379 ASSERT(dp->dp_tempreserved[txg & TXG_MASK] == 0);
382 * If the write limit max has not been explicitly set, set it
383 * to a fraction of available physical memory (default 1/8th).
384 * Note that we must inflate the limit because the spa
385 * inflates write sizes to account for data replication.
386 * Check this each sync phase to catch changing memory size.
388 if (physmem != old_physmem && zfs_write_limit_shift) {
389 mutex_enter(&zfs_write_limit_lock);
390 old_physmem = physmem;
391 zfs_write_limit_max = ptob(physmem) >> zfs_write_limit_shift;
392 zfs_write_limit_inflated = MAX(zfs_write_limit_min,
393 spa_get_asize(dp->dp_spa, zfs_write_limit_max));
394 mutex_exit(&zfs_write_limit_lock);
398 * Attempt to keep the sync time consistent by adjusting the
399 * amount of write traffic allowed into each transaction group.
400 * Weight the throughput calculation towards the current value:
401 * thru = 3/4 old_thru + 1/4 new_thru
403 ASSERT(zfs_write_limit_min > 0);
404 if (data_written > zfs_write_limit_min / 8 && write_time > 0) {
405 uint64_t throughput = (data_written * NANOSEC) / write_time;
406 if (dp->dp_throughput)
407 dp->dp_throughput = throughput / 4 +
408 3 * dp->dp_throughput / 4;
410 dp->dp_throughput = throughput;
411 dp->dp_write_limit = MIN(zfs_write_limit_inflated,
412 MAX(zfs_write_limit_min,
413 dp->dp_throughput * zfs_txg_synctime));
418 dsl_pool_zil_clean(dsl_pool_t *dp)
422 while (ds = list_head(&dp->dp_synced_datasets)) {
423 list_remove(&dp->dp_synced_datasets, ds);
424 ASSERT(ds->ds_user_ptr != NULL);
425 zil_clean(((objset_impl_t *)ds->ds_user_ptr)->os_zil);
426 dmu_buf_rele(ds->ds_dbuf, ds);
431 * TRUE if the current thread is the tx_sync_thread or if we
432 * are being called from SPA context during pool initialization.
435 dsl_pool_sync_context(dsl_pool_t *dp)
437 return (curthread == dp->dp_tx.tx_sync_thread ||
438 spa_get_dsl(dp->dp_spa) == NULL);
442 dsl_pool_adjustedsize(dsl_pool_t *dp, boolean_t netfree)
444 uint64_t space, resv;
447 * Reserve about 1.6% (1/64), or at least 32MB, for allocation
449 * XXX The intent log is not accounted for, so it must fit
452 * If we're trying to assess whether it's OK to do a free,
453 * cut the reservation in half to allow forward progress
454 * (e.g. make it possible to rm(1) files from a full pool).
456 space = spa_get_dspace(dp->dp_spa);
457 resv = MAX(space >> 6, SPA_MINDEVSIZE >> 1);
461 return (space - resv);
465 dsl_pool_tempreserve_space(dsl_pool_t *dp, uint64_t space, dmu_tx_t *tx)
467 uint64_t reserved = 0;
468 uint64_t write_limit = (zfs_write_limit_override ?
469 zfs_write_limit_override : dp->dp_write_limit);
471 if (zfs_no_write_throttle) {
472 atomic_add_64(&dp->dp_tempreserved[tx->tx_txg & TXG_MASK],
478 * Check to see if we have exceeded the maximum allowed IO for
479 * this transaction group. We can do this without locks since
480 * a little slop here is ok. Note that we do the reserved check
481 * with only half the requested reserve: this is because the
482 * reserve requests are worst-case, and we really don't want to
483 * throttle based off of worst-case estimates.
485 if (write_limit > 0) {
486 reserved = dp->dp_space_towrite[tx->tx_txg & TXG_MASK]
487 + dp->dp_tempreserved[tx->tx_txg & TXG_MASK] / 2;
489 if (reserved && reserved > write_limit)
493 atomic_add_64(&dp->dp_tempreserved[tx->tx_txg & TXG_MASK], space);
496 * If this transaction group is over 7/8ths capacity, delay
497 * the caller 1 clock tick. This will slow down the "fill"
498 * rate until the sync process can catch up with us.
500 if (reserved && reserved > (write_limit - (write_limit >> 3)))
501 txg_delay(dp, tx->tx_txg, 1);
507 dsl_pool_tempreserve_clear(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx)
509 ASSERT(dp->dp_tempreserved[tx->tx_txg & TXG_MASK] >= space);
510 atomic_add_64(&dp->dp_tempreserved[tx->tx_txg & TXG_MASK], -space);
514 dsl_pool_memory_pressure(dsl_pool_t *dp)
516 uint64_t space_inuse = 0;
519 if (dp->dp_write_limit == zfs_write_limit_min)
522 for (i = 0; i < TXG_SIZE; i++) {
523 space_inuse += dp->dp_space_towrite[i];
524 space_inuse += dp->dp_tempreserved[i];
526 dp->dp_write_limit = MAX(zfs_write_limit_min,
527 MIN(dp->dp_write_limit, space_inuse / 4));
531 dsl_pool_willuse_space(dsl_pool_t *dp, int64_t space, dmu_tx_t *tx)
534 mutex_enter(&dp->dp_lock);
535 dp->dp_space_towrite[tx->tx_txg & TXG_MASK] += space;
536 mutex_exit(&dp->dp_lock);
542 upgrade_clones_cb(spa_t *spa, uint64_t dsobj, const char *dsname, void *arg)
545 dsl_dataset_t *ds, *prev = NULL;
547 dsl_pool_t *dp = spa_get_dsl(spa);
549 err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds);
553 while (ds->ds_phys->ds_prev_snap_obj != 0) {
554 err = dsl_dataset_hold_obj(dp, ds->ds_phys->ds_prev_snap_obj,
557 dsl_dataset_rele(ds, FTAG);
561 if (prev->ds_phys->ds_next_snap_obj != ds->ds_object)
563 dsl_dataset_rele(ds, FTAG);
569 prev = dp->dp_origin_snap;
572 * The $ORIGIN can't have any data, or the accounting
575 ASSERT(prev->ds_phys->ds_bp.blk_birth == 0);
577 /* The origin doesn't get attached to itself */
578 if (ds->ds_object == prev->ds_object) {
579 dsl_dataset_rele(ds, FTAG);
583 dmu_buf_will_dirty(ds->ds_dbuf, tx);
584 ds->ds_phys->ds_prev_snap_obj = prev->ds_object;
585 ds->ds_phys->ds_prev_snap_txg = prev->ds_phys->ds_creation_txg;
587 dmu_buf_will_dirty(ds->ds_dir->dd_dbuf, tx);
588 ds->ds_dir->dd_phys->dd_origin_obj = prev->ds_object;
590 dmu_buf_will_dirty(prev->ds_dbuf, tx);
591 prev->ds_phys->ds_num_children++;
593 if (ds->ds_phys->ds_next_snap_obj == 0) {
594 ASSERT(ds->ds_prev == NULL);
595 VERIFY(0 == dsl_dataset_hold_obj(dp,
596 ds->ds_phys->ds_prev_snap_obj, ds, &ds->ds_prev));
600 ASSERT(ds->ds_dir->dd_phys->dd_origin_obj == prev->ds_object);
601 ASSERT(ds->ds_phys->ds_prev_snap_obj == prev->ds_object);
603 if (prev->ds_phys->ds_next_clones_obj == 0) {
604 prev->ds_phys->ds_next_clones_obj =
605 zap_create(dp->dp_meta_objset,
606 DMU_OT_NEXT_CLONES, DMU_OT_NONE, 0, tx);
608 VERIFY(0 == zap_add_int(dp->dp_meta_objset,
609 prev->ds_phys->ds_next_clones_obj, ds->ds_object, tx));
611 dsl_dataset_rele(ds, FTAG);
612 if (prev != dp->dp_origin_snap)
613 dsl_dataset_rele(prev, FTAG);
618 dsl_pool_upgrade_clones(dsl_pool_t *dp, dmu_tx_t *tx)
620 ASSERT(dmu_tx_is_syncing(tx));
621 ASSERT(dp->dp_origin_snap != NULL);
623 (void) dmu_objset_find_spa(dp->dp_spa, NULL, upgrade_clones_cb,
624 tx, DS_FIND_CHILDREN);
628 dsl_pool_create_origin(dsl_pool_t *dp, dmu_tx_t *tx)
633 ASSERT(dmu_tx_is_syncing(tx));
634 ASSERT(dp->dp_origin_snap == NULL);
636 /* create the origin dir, ds, & snap-ds */
637 rw_enter(&dp->dp_config_rwlock, RW_WRITER);
638 dsobj = dsl_dataset_create_sync(dp->dp_root_dir, ORIGIN_DIR_NAME,
640 VERIFY(0 == dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds));
641 dsl_dataset_snapshot_sync(ds, ORIGIN_DIR_NAME, kcred, tx);
642 VERIFY(0 == dsl_dataset_hold_obj(dp, ds->ds_phys->ds_prev_snap_obj,
643 dp, &dp->dp_origin_snap));
644 dsl_dataset_rele(ds, FTAG);
645 rw_exit(&dp->dp_config_rwlock);
649 dsl_pool_vnrele_taskq(dsl_pool_t *dp)
651 return (dp->dp_vnrele_taskq);