[mirror_zfs.git] / module / zfs / dsl_synctask.c

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 */

#include <sys/dmu.h>
#include <sys/dmu_tx.h>
#include <sys/dsl_pool.h>
#include <sys/dsl_dir.h>
#include <sys/dsl_synctask.h>
#include <sys/metaslab.h>

#define	DST_AVG_BLKSHIFT 14

/* ARGSUSED */
static int
dsl_null_checkfunc(void *arg1, void *arg2, dmu_tx_t *tx)
{
	return (0);
}

dsl_sync_task_group_t *
dsl_sync_task_group_create(dsl_pool_t *dp)
{
	dsl_sync_task_group_t *dstg;

	dstg = kmem_zalloc(sizeof (dsl_sync_task_group_t), KM_SLEEP);
	list_create(&dstg->dstg_tasks, sizeof (dsl_sync_task_t),
	    offsetof(dsl_sync_task_t, dst_node));
	dstg->dstg_pool = dp;

	return (dstg);
}

void
dsl_sync_task_create(dsl_sync_task_group_t *dstg,
    dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
    void *arg1, void *arg2, int blocks_modified)
{
	dsl_sync_task_t *dst;

	if (checkfunc == NULL)
		checkfunc = dsl_null_checkfunc;
	dst = kmem_zalloc(sizeof (dsl_sync_task_t), KM_SLEEP);
	dst->dst_checkfunc = checkfunc;
	dst->dst_syncfunc = syncfunc;
	dst->dst_arg1 = arg1;
	dst->dst_arg2 = arg2;
	list_insert_tail(&dstg->dstg_tasks, dst);

	dstg->dstg_space += blocks_modified << DST_AVG_BLKSHIFT;
}

int
dsl_sync_task_group_wait(dsl_sync_task_group_t *dstg)
{
	dmu_tx_t *tx;
	uint64_t txg;
	dsl_sync_task_t *dst;

top:
	tx = dmu_tx_create_dd(dstg->dstg_pool->dp_mos_dir);
	VERIFY(0 == dmu_tx_assign(tx, TXG_WAIT));

	txg = dmu_tx_get_txg(tx);

	/* Do a preliminary error check. */
	dstg->dstg_err = 0;
	rw_enter(&dstg->dstg_pool->dp_config_rwlock, RW_READER);
	for (dst = list_head(&dstg->dstg_tasks); dst;
	    dst = list_next(&dstg->dstg_tasks, dst)) {
#ifdef ZFS_DEBUG
		/*
		 * Only check half the time, otherwise, the sync-context
		 * check will almost never fail.
		 */
		if (spa_get_random(2) == 0)
			continue;
#endif
		dst->dst_err =
		    dst->dst_checkfunc(dst->dst_arg1, dst->dst_arg2, tx);
		if (dst->dst_err)
			dstg->dstg_err = dst->dst_err;
	}
	rw_exit(&dstg->dstg_pool->dp_config_rwlock);

	if (dstg->dstg_err) {
		dmu_tx_commit(tx);
		return (dstg->dstg_err);
	}

	/*
	 * We don't generally have many sync tasks, so pay the price of
	 * add_tail to get the tasks executed in the right order.
	 */
	VERIFY(0 == txg_list_add_tail(&dstg->dstg_pool->dp_sync_tasks,
	    dstg, txg));

	dmu_tx_commit(tx);

	txg_wait_synced(dstg->dstg_pool, txg);

	if (dstg->dstg_err == EAGAIN) {
		txg_wait_synced(dstg->dstg_pool, txg + TXG_DEFER_SIZE);
		goto top;
	}

	return (dstg->dstg_err);
}

void
dsl_sync_task_group_nowait(dsl_sync_task_group_t *dstg, dmu_tx_t *tx)
{
	uint64_t txg;

	dstg->dstg_nowaiter = B_TRUE;
	txg = dmu_tx_get_txg(tx);
	/*
	 * We don't generally have many sync tasks, so pay the price of
	 * add_tail to get the tasks executed in the right order.
	 */
	VERIFY(0 == txg_list_add_tail(&dstg->dstg_pool->dp_sync_tasks,
	    dstg, txg));
}

void
dsl_sync_task_group_destroy(dsl_sync_task_group_t *dstg)
{
	dsl_sync_task_t *dst;

	while ((dst = list_head(&dstg->dstg_tasks))) {
		list_remove(&dstg->dstg_tasks, dst);
		kmem_free(dst, sizeof (dsl_sync_task_t));
	}
	kmem_free(dstg, sizeof (dsl_sync_task_group_t));
}

void
dsl_sync_task_group_sync(dsl_sync_task_group_t *dstg, dmu_tx_t *tx)
{
	dsl_sync_task_t *dst;
	dsl_pool_t *dp = dstg->dstg_pool;
	uint64_t quota, used;

	ASSERT3U(dstg->dstg_err, ==, 0);

	/*
	 * Check for sufficient space.  We just check against what's
	 * on-disk; we don't want any in-flight accounting to get in our
	 * way, because open context may have already used up various
	 * in-core limits (arc_tempreserve, dsl_pool_tempreserve).
	 */
	quota = dsl_pool_adjustedsize(dp, B_FALSE) -
	    metaslab_class_get_deferred(spa_normal_class(dp->dp_spa));
	used = dp->dp_root_dir->dd_phys->dd_used_bytes;
	/* MOS space is triple-dittoed, so we multiply by 3. */
	if (dstg->dstg_space > 0 && used + dstg->dstg_space * 3 > quota) {
		dstg->dstg_err = ENOSPC;
		return;
	}

	/*
	 * Check for errors by calling checkfuncs.
	 */
	rw_enter(&dp->dp_config_rwlock, RW_WRITER);
	for (dst = list_head(&dstg->dstg_tasks); dst;
	    dst = list_next(&dstg->dstg_tasks, dst)) {
		dst->dst_err =
		    dst->dst_checkfunc(dst->dst_arg1, dst->dst_arg2, tx);
		if (dst->dst_err)
			dstg->dstg_err = dst->dst_err;
	}

	if (dstg->dstg_err == 0) {
		/*
		 * Execute sync tasks.
		 */
		for (dst = list_head(&dstg->dstg_tasks); dst;
		    dst = list_next(&dstg->dstg_tasks, dst)) {
			dst->dst_syncfunc(dst->dst_arg1, dst->dst_arg2, tx);
		}
	}
	rw_exit(&dp->dp_config_rwlock);

	if (dstg->dstg_nowaiter)
		dsl_sync_task_group_destroy(dstg);
}

int
dsl_sync_task_do(dsl_pool_t *dp,
    dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
    void *arg1, void *arg2, int blocks_modified)
{
	dsl_sync_task_group_t *dstg;
	int err;

	ASSERT(spa_writeable(dp->dp_spa));

	dstg = dsl_sync_task_group_create(dp);
	dsl_sync_task_create(dstg, checkfunc, syncfunc,
	    arg1, arg2, blocks_modified);
	err = dsl_sync_task_group_wait(dstg);
	dsl_sync_task_group_destroy(dstg);
	return (err);
}

void
dsl_sync_task_do_nowait(dsl_pool_t *dp,
    dsl_checkfunc_t *checkfunc, dsl_syncfunc_t *syncfunc,
    void *arg1, void *arg2, int blocks_modified, dmu_tx_t *tx)
{
	dsl_sync_task_group_t *dstg;

	if (!spa_writeable(dp->dp_spa))
		return;

	dstg = dsl_sync_task_group_create(dp);
	dsl_sync_task_create(dstg, checkfunc, syncfunc,
	    arg1, arg2, blocks_modified);
	dsl_sync_task_group_nowait(dstg, tx);
}

#if defined(_KERNEL) && defined(HAVE_SPL)
EXPORT_SYMBOL(dsl_sync_task_do);
EXPORT_SYMBOL(dsl_sync_task_do_nowait);
#endif
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	/*
428870ff	22	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
34dc7c2f BB	23	*/
34dc7c2f BB	24
34dc7c2f BB	25	#include <sys/dmu.h>
	26	#include <sys/dmu_tx.h>
	27	#include <sys/dsl_pool.h>
	28	#include <sys/dsl_dir.h>
	29	#include <sys/dsl_synctask.h>
428870ff	30	#include <sys/metaslab.h>
34dc7c2f BB	31
	32	#define DST_AVG_BLKSHIFT 14
	33
	34	/* ARGSUSED */
	35	static int
	36	dsl_null_checkfunc(void arg1, void arg2, dmu_tx_t *tx)
	37	{
	38	return (0);
	39	}
	40
	41	dsl_sync_task_group_t *
	42	dsl_sync_task_group_create(dsl_pool_t *dp)
	43	{
	44	dsl_sync_task_group_t *dstg;
	45
	46	dstg = kmem_zalloc(sizeof (dsl_sync_task_group_t), KM_SLEEP);
	47	list_create(&dstg->dstg_tasks, sizeof (dsl_sync_task_t),
	48	offsetof(dsl_sync_task_t, dst_node));
	49	dstg->dstg_pool = dp;
34dc7c2f BB	50
	51	return (dstg);
	52	}
	53
	54	void
	55	dsl_sync_task_create(dsl_sync_task_group_t *dstg,
	56	dsl_checkfunc_t checkfunc, dsl_syncfunc_t syncfunc,
	57	void arg1, void arg2, int blocks_modified)
	58	{
	59	dsl_sync_task_t *dst;
	60
	61	if (checkfunc == NULL)
	62	checkfunc = dsl_null_checkfunc;
	63	dst = kmem_zalloc(sizeof (dsl_sync_task_t), KM_SLEEP);
	64	dst->dst_checkfunc = checkfunc;
	65	dst->dst_syncfunc = syncfunc;
	66	dst->dst_arg1 = arg1;
	67	dst->dst_arg2 = arg2;
	68	list_insert_tail(&dstg->dstg_tasks, dst);
	69
	70	dstg->dstg_space += blocks_modified << DST_AVG_BLKSHIFT;
	71	}
	72
	73	int
	74	dsl_sync_task_group_wait(dsl_sync_task_group_t *dstg)
	75	{
	76	dmu_tx_t *tx;
	77	uint64_t txg;
	78	dsl_sync_task_t *dst;
	79
	80	top:
	81	tx = dmu_tx_create_dd(dstg->dstg_pool->dp_mos_dir);
	82	VERIFY(0 == dmu_tx_assign(tx, TXG_WAIT));
	83
	84	txg = dmu_tx_get_txg(tx);
	85
	86	/* Do a preliminary error check. */
	87	dstg->dstg_err = 0;
	88	rw_enter(&dstg->dstg_pool->dp_config_rwlock, RW_READER);
	89	for (dst = list_head(&dstg->dstg_tasks); dst;
	90	dst = list_next(&dstg->dstg_tasks, dst)) {
	91	#ifdef ZFS_DEBUG
	92	/*
	93	* Only check half the time, otherwise, the sync-context
	94	* check will almost never fail.
	95	*/
	96	if (spa_get_random(2) == 0)
	97	continue;
	98	#endif
	99	dst->dst_err =
	100	dst->dst_checkfunc(dst->dst_arg1, dst->dst_arg2, tx);
	101	if (dst->dst_err)
	102	dstg->dstg_err = dst->dst_err;
	103	}
	104	rw_exit(&dstg->dstg_pool->dp_config_rwlock);
	105
	106	if (dstg->dstg_err) {
	107	dmu_tx_commit(tx);
	108	return (dstg->dstg_err);
	109	}
	110
428870ff BB	111	/*
	112	* We don't generally have many sync tasks, so pay the price of
	113	* add_tail to get the tasks executed in the right order.
	114	*/
	115	VERIFY(0 == txg_list_add_tail(&dstg->dstg_pool->dp_sync_tasks,
	116	dstg, txg));
34dc7c2f BB	117
	118	dmu_tx_commit(tx);
	119
	120	txg_wait_synced(dstg->dstg_pool, txg);
	121
428870ff BB	122	if (dstg->dstg_err == EAGAIN) {
428870ff BB	123	txg_wait_synced(dstg->dstg_pool, txg + TXG_DEFER_SIZE);
34dc7c2f	124	goto top;
428870ff	125	}
34dc7c2f BB	126
	127	return (dstg->dstg_err);
	128	}
	129
	130	void
	131	dsl_sync_task_group_nowait(dsl_sync_task_group_t dstg, dmu_tx_t tx)
	132	{
	133	uint64_t txg;
	134
	135	dstg->dstg_nowaiter = B_TRUE;
	136	txg = dmu_tx_get_txg(tx);
428870ff BB	137	/*
	138	* We don't generally have many sync tasks, so pay the price of
	139	* add_tail to get the tasks executed in the right order.
	140	*/
	141	VERIFY(0 == txg_list_add_tail(&dstg->dstg_pool->dp_sync_tasks,
	142	dstg, txg));
34dc7c2f BB	143	}
	144
	145	void
	146	dsl_sync_task_group_destroy(dsl_sync_task_group_t *dstg)
	147	{
	148	dsl_sync_task_t *dst;
	149
c65aa5b2	150	while ((dst = list_head(&dstg->dstg_tasks))) {
34dc7c2f BB	151	list_remove(&dstg->dstg_tasks, dst);
	152	kmem_free(dst, sizeof (dsl_sync_task_t));
	153	}
	154	kmem_free(dstg, sizeof (dsl_sync_task_group_t));
	155	}
	156
	157	void
	158	dsl_sync_task_group_sync(dsl_sync_task_group_t dstg, dmu_tx_t tx)
	159	{
	160	dsl_sync_task_t *dst;
428870ff BB	161	dsl_pool_t *dp = dstg->dstg_pool;
428870ff BB	162	uint64_t quota, used;
34dc7c2f BB	163
	164	ASSERT3U(dstg->dstg_err, ==, 0);
	165
	166	/*
428870ff BB	167	* Check for sufficient space. We just check against what's
	168	* on-disk; we don't want any in-flight accounting to get in our
	169	* way, because open context may have already used up various
	170	* in-core limits (arc_tempreserve, dsl_pool_tempreserve).
34dc7c2f	171	*/
428870ff BB	172	quota = dsl_pool_adjustedsize(dp, B_FALSE) -
	173	metaslab_class_get_deferred(spa_normal_class(dp->dp_spa));
	174	used = dp->dp_root_dir->dd_phys->dd_used_bytes;
	175	/* MOS space is triple-dittoed, so we multiply by 3. */
	176	if (dstg->dstg_space > 0 && used + dstg->dstg_space * 3 > quota) {
	177	dstg->dstg_err = ENOSPC;
34dc7c2f	178	return;
428870ff	179	}
34dc7c2f BB	180
	181	/*
	182	* Check for errors by calling checkfuncs.
	183	*/
428870ff	184	rw_enter(&dp->dp_config_rwlock, RW_WRITER);
34dc7c2f BB	185	for (dst = list_head(&dstg->dstg_tasks); dst;
	186	dst = list_next(&dstg->dstg_tasks, dst)) {
	187	dst->dst_err =
	188	dst->dst_checkfunc(dst->dst_arg1, dst->dst_arg2, tx);
	189	if (dst->dst_err)
	190	dstg->dstg_err = dst->dst_err;
	191	}
	192
	193	if (dstg->dstg_err == 0) {
	194	/*
	195	* Execute sync tasks.
	196	*/
	197	for (dst = list_head(&dstg->dstg_tasks); dst;
	198	dst = list_next(&dstg->dstg_tasks, dst)) {
428870ff	199	dst->dst_syncfunc(dst->dst_arg1, dst->dst_arg2, tx);
34dc7c2f BB	200	}
34dc7c2f BB	201	}
428870ff	202	rw_exit(&dp->dp_config_rwlock);
34dc7c2f BB	203
	204	if (dstg->dstg_nowaiter)
	205	dsl_sync_task_group_destroy(dstg);
	206	}
	207
	208	int
	209	dsl_sync_task_do(dsl_pool_t *dp,
	210	dsl_checkfunc_t checkfunc, dsl_syncfunc_t syncfunc,
	211	void arg1, void arg2, int blocks_modified)
	212	{
	213	dsl_sync_task_group_t *dstg;
	214	int err;
	215
572e2857 BB	216	ASSERT(spa_writeable(dp->dp_spa));
572e2857 BB	217
34dc7c2f BB	218	dstg = dsl_sync_task_group_create(dp);
	219	dsl_sync_task_create(dstg, checkfunc, syncfunc,
	220	arg1, arg2, blocks_modified);
	221	err = dsl_sync_task_group_wait(dstg);
	222	dsl_sync_task_group_destroy(dstg);
	223	return (err);
	224	}
	225
	226	void
	227	dsl_sync_task_do_nowait(dsl_pool_t *dp,
	228	dsl_checkfunc_t checkfunc, dsl_syncfunc_t syncfunc,
	229	void arg1, void arg2, int blocks_modified, dmu_tx_t *tx)
	230	{
	231	dsl_sync_task_group_t *dstg;
	232
572e2857 BB	233	if (!spa_writeable(dp->dp_spa))
	234	return;
	235
34dc7c2f BB	236	dstg = dsl_sync_task_group_create(dp);
	237	dsl_sync_task_create(dstg, checkfunc, syncfunc,
	238	arg1, arg2, blocks_modified);
	239	dsl_sync_task_group_nowait(dstg, tx);
	240	}
c28b2279 BB	241
	242	#if defined(_KERNEL) && defined(HAVE_SPL)
	243	EXPORT_SYMBOL(dsl_sync_task_do);
	244	EXPORT_SYMBOL(dsl_sync_task_do_nowait);
	245	#endif