[mirror_zfs.git] / module / zfs / bpobj.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.
 * Copyright (c) 2013 by Delphix. All rights reserved.
 */

#include <sys/bpobj.h>
#include <sys/zfs_context.h>
#include <sys/refcount.h>
#include <sys/dsl_pool.h>
#include <sys/zfeature.h>
#include <sys/zap.h>

/*
 * Return an empty bpobj, preferably the empty dummy one (dp_empty_bpobj).
 */
uint64_t
bpobj_alloc_empty(objset_t *os, int blocksize, dmu_tx_t *tx)
{
	spa_t *spa = dmu_objset_spa(os);
	dsl_pool_t *dp = dmu_objset_pool(os);

	if (spa_feature_is_enabled(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
		if (!spa_feature_is_active(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
			ASSERT0(dp->dp_empty_bpobj);
			dp->dp_empty_bpobj =
			    bpobj_alloc(os, SPA_MAXBLOCKSIZE, tx);
			VERIFY(zap_add(os,
			    DMU_POOL_DIRECTORY_OBJECT,
			    DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
			    &dp->dp_empty_bpobj, tx) == 0);
		}
		spa_feature_incr(spa, SPA_FEATURE_EMPTY_BPOBJ, tx);
		ASSERT(dp->dp_empty_bpobj != 0);
		return (dp->dp_empty_bpobj);
	} else {
		return (bpobj_alloc(os, blocksize, tx));
	}
}

void
bpobj_decr_empty(objset_t *os, dmu_tx_t *tx)
{
	dsl_pool_t *dp = dmu_objset_pool(os);

	spa_feature_decr(dmu_objset_spa(os), SPA_FEATURE_EMPTY_BPOBJ, tx);
	if (!spa_feature_is_active(dmu_objset_spa(os),
	    SPA_FEATURE_EMPTY_BPOBJ)) {
		VERIFY3U(0, ==, zap_remove(dp->dp_meta_objset,
		    DMU_POOL_DIRECTORY_OBJECT,
		    DMU_POOL_EMPTY_BPOBJ, tx));
		VERIFY3U(0, ==, dmu_object_free(os, dp->dp_empty_bpobj, tx));
		dp->dp_empty_bpobj = 0;
	}
}

uint64_t
bpobj_alloc(objset_t *os, int blocksize, dmu_tx_t *tx)
{
	int size;

	if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_BPOBJ_ACCOUNT)
		size = BPOBJ_SIZE_V0;
	else if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
		size = BPOBJ_SIZE_V1;
	else
		size = sizeof (bpobj_phys_t);

	return (dmu_object_alloc(os, DMU_OT_BPOBJ, blocksize,
	    DMU_OT_BPOBJ_HDR, size, tx));
}

void
bpobj_free(objset_t *os, uint64_t obj, dmu_tx_t *tx)
{
	int64_t i;
	bpobj_t bpo;
	dmu_object_info_t doi;
	int epb;
	dmu_buf_t *dbuf = NULL;

	ASSERT(obj != dmu_objset_pool(os)->dp_empty_bpobj);
	VERIFY3U(0, ==, bpobj_open(&bpo, os, obj));

	mutex_enter(&bpo.bpo_lock);

	if (!bpo.bpo_havesubobj || bpo.bpo_phys->bpo_subobjs == 0)
		goto out;

	VERIFY3U(0, ==, dmu_object_info(os, bpo.bpo_phys->bpo_subobjs, &doi));
	epb = doi.doi_data_block_size / sizeof (uint64_t);

	for (i = bpo.bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
		uint64_t *objarray;
		uint64_t offset, blkoff;

		offset = i * sizeof (uint64_t);
		blkoff = P2PHASE(i, epb);

		if (dbuf == NULL || dbuf->db_offset > offset) {
			if (dbuf)
				dmu_buf_rele(dbuf, FTAG);
			VERIFY3U(0, ==, dmu_buf_hold(os,
			    bpo.bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0));
		}

		ASSERT3U(offset, >=, dbuf->db_offset);
		ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);

		objarray = dbuf->db_data;
		bpobj_free(os, objarray[blkoff], tx);
	}
	if (dbuf) {
		dmu_buf_rele(dbuf, FTAG);
		dbuf = NULL;
	}
	VERIFY3U(0, ==, dmu_object_free(os, bpo.bpo_phys->bpo_subobjs, tx));

out:
	mutex_exit(&bpo.bpo_lock);
	bpobj_close(&bpo);

	VERIFY3U(0, ==, dmu_object_free(os, obj, tx));
}

int
bpobj_open(bpobj_t *bpo, objset_t *os, uint64_t object)
{
	dmu_object_info_t doi;
	int err;

	err = dmu_object_info(os, object, &doi);
	if (err)
		return (err);

	bzero(bpo, sizeof (*bpo));
	mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL);

	ASSERT(bpo->bpo_dbuf == NULL);
	ASSERT(bpo->bpo_phys == NULL);
	ASSERT(object != 0);
	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ);
	ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR);

	err = dmu_bonus_hold(os, object, bpo, &bpo->bpo_dbuf);
	if (err)
		return (err);

	bpo->bpo_os = os;
	bpo->bpo_object = object;
	bpo->bpo_epb = doi.doi_data_block_size >> SPA_BLKPTRSHIFT;
	bpo->bpo_havecomp = (doi.doi_bonus_size > BPOBJ_SIZE_V0);
	bpo->bpo_havesubobj = (doi.doi_bonus_size > BPOBJ_SIZE_V1);
	bpo->bpo_phys = bpo->bpo_dbuf->db_data;
	return (0);
}

void
bpobj_close(bpobj_t *bpo)
{
	/* Lame workaround for closing a bpobj that was never opened. */
	if (bpo->bpo_object == 0)
		return;

	dmu_buf_rele(bpo->bpo_dbuf, bpo);
	if (bpo->bpo_cached_dbuf != NULL)
		dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
	bpo->bpo_dbuf = NULL;
	bpo->bpo_phys = NULL;
	bpo->bpo_cached_dbuf = NULL;
	bpo->bpo_object = 0;

	mutex_destroy(&bpo->bpo_lock);
}

static boolean_t
bpobj_hasentries(bpobj_t *bpo)
{
	return (bpo->bpo_phys->bpo_num_blkptrs != 0 ||
	    (bpo->bpo_havesubobj && bpo->bpo_phys->bpo_num_subobjs != 0));
}

static int
bpobj_iterate_impl(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx,
    boolean_t free)
{
	dmu_object_info_t doi;
	int epb;
	int64_t i;
	int err = 0;
	dmu_buf_t *dbuf = NULL;

	mutex_enter(&bpo->bpo_lock);

	if (free)
		dmu_buf_will_dirty(bpo->bpo_dbuf, tx);

	for (i = bpo->bpo_phys->bpo_num_blkptrs - 1; i >= 0; i--) {
		blkptr_t *bparray;
		blkptr_t *bp;
		uint64_t offset, blkoff;

		offset = i * sizeof (blkptr_t);
		blkoff = P2PHASE(i, bpo->bpo_epb);

		if (dbuf == NULL || dbuf->db_offset > offset) {
			if (dbuf)
				dmu_buf_rele(dbuf, FTAG);
			err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object, offset,
			    FTAG, &dbuf, 0);
			if (err)
				break;
		}

		ASSERT3U(offset, >=, dbuf->db_offset);
		ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);

		bparray = dbuf->db_data;
		bp = &bparray[blkoff];
		err = func(arg, bp, tx);
		if (err)
			break;
		if (free) {
			bpo->bpo_phys->bpo_bytes -=
			    bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
			ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
			if (bpo->bpo_havecomp) {
				bpo->bpo_phys->bpo_comp -= BP_GET_PSIZE(bp);
				bpo->bpo_phys->bpo_uncomp -= BP_GET_UCSIZE(bp);
			}
			bpo->bpo_phys->bpo_num_blkptrs--;
			ASSERT3S(bpo->bpo_phys->bpo_num_blkptrs, >=, 0);
		}
	}
	if (dbuf) {
		dmu_buf_rele(dbuf, FTAG);
		dbuf = NULL;
	}
	if (free) {
		i++;
		VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os, bpo->bpo_object,
		    i * sizeof (blkptr_t), -1ULL, tx));
	}
	if (err || !bpo->bpo_havesubobj || bpo->bpo_phys->bpo_subobjs == 0)
		goto out;

	ASSERT(bpo->bpo_havecomp);
	err = dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi);
	if (err) {
		mutex_exit(&bpo->bpo_lock);
		return (err);
	}
	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
	epb = doi.doi_data_block_size / sizeof (uint64_t);

	for (i = bpo->bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
		uint64_t *objarray;
		uint64_t offset, blkoff;
		bpobj_t sublist;
		uint64_t used_before, comp_before, uncomp_before;
		uint64_t used_after, comp_after, uncomp_after;

		offset = i * sizeof (uint64_t);
		blkoff = P2PHASE(i, epb);

		if (dbuf == NULL || dbuf->db_offset > offset) {
			if (dbuf)
				dmu_buf_rele(dbuf, FTAG);
			err = dmu_buf_hold(bpo->bpo_os,
			    bpo->bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0);
			if (err)
				break;
		}

		ASSERT3U(offset, >=, dbuf->db_offset);
		ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);

		objarray = dbuf->db_data;
		err = bpobj_open(&sublist, bpo->bpo_os, objarray[blkoff]);
		if (err)
			break;
		if (free) {
			err = bpobj_space(&sublist,
			    &used_before, &comp_before, &uncomp_before);
			if (err != 0) {
				bpobj_close(&sublist);
				break;
			}
		}
		err = bpobj_iterate_impl(&sublist, func, arg, tx, free);
		if (free) {
			VERIFY3U(0, ==, bpobj_space(&sublist,
			    &used_after, &comp_after, &uncomp_after));
			bpo->bpo_phys->bpo_bytes -= used_before - used_after;
			ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
			bpo->bpo_phys->bpo_comp -= comp_before - comp_after;
			bpo->bpo_phys->bpo_uncomp -=
			    uncomp_before - uncomp_after;
		}

		bpobj_close(&sublist);
		if (err)
			break;
		if (free) {
			err = dmu_object_free(bpo->bpo_os,
			    objarray[blkoff], tx);
			if (err)
				break;
			bpo->bpo_phys->bpo_num_subobjs--;
			ASSERT3S(bpo->bpo_phys->bpo_num_subobjs, >=, 0);
		}
	}
	if (dbuf) {
		dmu_buf_rele(dbuf, FTAG);
		dbuf = NULL;
	}
	if (free) {
		VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os,
		    bpo->bpo_phys->bpo_subobjs,
		    (i + 1) * sizeof (uint64_t), -1ULL, tx));
	}

out:
	/* If there are no entries, there should be no bytes. */
	if (!bpobj_hasentries(bpo)) {
		ASSERT0(bpo->bpo_phys->bpo_bytes);
		ASSERT0(bpo->bpo_phys->bpo_comp);
		ASSERT0(bpo->bpo_phys->bpo_uncomp);
	}

	mutex_exit(&bpo->bpo_lock);
	return (err);
}

/*
 * Iterate and remove the entries.  If func returns nonzero, iteration
 * will stop and that entry will not be removed.
 */
int
bpobj_iterate(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
{
	return (bpobj_iterate_impl(bpo, func, arg, tx, B_TRUE));
}

/*
 * Iterate the entries.  If func returns nonzero, iteration will stop.
 */
int
bpobj_iterate_nofree(bpobj_t *bpo, bpobj_itor_t func, void *arg, dmu_tx_t *tx)
{
	return (bpobj_iterate_impl(bpo, func, arg, tx, B_FALSE));
}

void
bpobj_enqueue_subobj(bpobj_t *bpo, uint64_t subobj, dmu_tx_t *tx)
{
	bpobj_t subbpo;
	uint64_t used, comp, uncomp, subsubobjs;
	ASSERTV(dmu_object_info_t doi);

	ASSERT(bpo->bpo_havesubobj);
	ASSERT(bpo->bpo_havecomp);
	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);

	if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj) {
		bpobj_decr_empty(bpo->bpo_os, tx);
		return;
	}

	VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
	VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));

	if (!bpobj_hasentries(&subbpo)) {
		/* No point in having an empty subobj. */
		bpobj_close(&subbpo);
		bpobj_free(bpo->bpo_os, subobj, tx);
		return;
	}

	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	if (bpo->bpo_phys->bpo_subobjs == 0) {
		bpo->bpo_phys->bpo_subobjs = dmu_object_alloc(bpo->bpo_os,
		    DMU_OT_BPOBJ_SUBOBJ, SPA_MAXBLOCKSIZE, DMU_OT_NONE, 0, tx);
	}

	ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi));
	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);

	mutex_enter(&bpo->bpo_lock);
	dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
	    bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
	    sizeof (subobj), &subobj, tx);
	bpo->bpo_phys->bpo_num_subobjs++;

	/*
	 * If subobj has only one block of subobjs, then move subobj's
	 * subobjs to bpo's subobj list directly.  This reduces
	 * recursion in bpobj_iterate due to nested subobjs.
	 */
	subsubobjs = subbpo.bpo_phys->bpo_subobjs;
	if (subsubobjs != 0) {
		dmu_object_info_t doi;

		VERIFY3U(0, ==, dmu_object_info(bpo->bpo_os, subsubobjs, &doi));
		if (doi.doi_max_offset == doi.doi_data_block_size) {
			dmu_buf_t *subdb;
			uint64_t numsubsub = subbpo.bpo_phys->bpo_num_subobjs;

			VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, subsubobjs,
			    0, FTAG, &subdb, 0));
			/*
			 * Make sure that we are not asking dmu_write()
			 * to write more data than we have in our buffer.
			 */
			VERIFY3U(subdb->db_size, >=,
			    numsubsub * sizeof (subobj));
			dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
			    bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
			    numsubsub * sizeof (subobj), subdb->db_data, tx);
			dmu_buf_rele(subdb, FTAG);
			bpo->bpo_phys->bpo_num_subobjs += numsubsub;

			dmu_buf_will_dirty(subbpo.bpo_dbuf, tx);
			subbpo.bpo_phys->bpo_subobjs = 0;
			VERIFY3U(0, ==, dmu_object_free(bpo->bpo_os,
			    subsubobjs, tx));
		}
	}
	bpo->bpo_phys->bpo_bytes += used;
	bpo->bpo_phys->bpo_comp += comp;
	bpo->bpo_phys->bpo_uncomp += uncomp;
	mutex_exit(&bpo->bpo_lock);

	bpobj_close(&subbpo);
}

void
bpobj_enqueue(bpobj_t *bpo, const blkptr_t *bp, dmu_tx_t *tx)
{
	blkptr_t stored_bp = *bp;
	uint64_t offset;
	int blkoff;
	blkptr_t *bparray;

	ASSERT(!BP_IS_HOLE(bp));
	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);

	if (BP_IS_EMBEDDED(bp)) {
		/*
		 * The bpobj will compress better without the payload.
		 *
		 * Note that we store EMBEDDED bp's because they have an
		 * uncompressed size, which must be accounted for.  An
		 * alternative would be to add their size to bpo_uncomp
		 * without storing the bp, but that would create additional
		 * complications: bpo_uncomp would be inconsistent with the
		 * set of BP's stored, and bpobj_iterate() wouldn't visit
		 * all the space accounted for in the bpobj.
		 */
		bzero(&stored_bp, sizeof (stored_bp));
		stored_bp.blk_prop = bp->blk_prop;
		stored_bp.blk_birth = bp->blk_birth;
	} else if (!BP_GET_DEDUP(bp)) {
		/* The bpobj will compress better without the checksum */
		bzero(&stored_bp.blk_cksum, sizeof (stored_bp.blk_cksum));
	}

	/* We never need the fill count. */
	stored_bp.blk_fill = 0;

	mutex_enter(&bpo->bpo_lock);

	offset = bpo->bpo_phys->bpo_num_blkptrs * sizeof (stored_bp);
	blkoff = P2PHASE(bpo->bpo_phys->bpo_num_blkptrs, bpo->bpo_epb);

	if (bpo->bpo_cached_dbuf == NULL ||
	    offset < bpo->bpo_cached_dbuf->db_offset ||
	    offset >= bpo->bpo_cached_dbuf->db_offset +
	    bpo->bpo_cached_dbuf->db_size) {
		if (bpo->bpo_cached_dbuf)
			dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
		VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
		    offset, bpo, &bpo->bpo_cached_dbuf, 0));
	}

	dmu_buf_will_dirty(bpo->bpo_cached_dbuf, tx);
	bparray = bpo->bpo_cached_dbuf->db_data;
	bparray[blkoff] = stored_bp;

	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	bpo->bpo_phys->bpo_num_blkptrs++;
	bpo->bpo_phys->bpo_bytes +=
	    bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
	if (bpo->bpo_havecomp) {
		bpo->bpo_phys->bpo_comp += BP_GET_PSIZE(bp);
		bpo->bpo_phys->bpo_uncomp += BP_GET_UCSIZE(bp);
	}
	mutex_exit(&bpo->bpo_lock);
}

struct space_range_arg {
	spa_t *spa;
	uint64_t mintxg;
	uint64_t maxtxg;
	uint64_t used;
	uint64_t comp;
	uint64_t uncomp;
};

/* ARGSUSED */
static int
space_range_cb(void *arg, const blkptr_t *bp, dmu_tx_t *tx)
{
	struct space_range_arg *sra = arg;

	if (bp->blk_birth > sra->mintxg && bp->blk_birth <= sra->maxtxg) {
		if (dsl_pool_sync_context(spa_get_dsl(sra->spa)))
			sra->used += bp_get_dsize_sync(sra->spa, bp);
		else
			sra->used += bp_get_dsize(sra->spa, bp);
		sra->comp += BP_GET_PSIZE(bp);
		sra->uncomp += BP_GET_UCSIZE(bp);
	}
	return (0);
}

int
bpobj_space(bpobj_t *bpo, uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
{
	mutex_enter(&bpo->bpo_lock);

	*usedp = bpo->bpo_phys->bpo_bytes;
	if (bpo->bpo_havecomp) {
		*compp = bpo->bpo_phys->bpo_comp;
		*uncompp = bpo->bpo_phys->bpo_uncomp;
		mutex_exit(&bpo->bpo_lock);
		return (0);
	} else {
		mutex_exit(&bpo->bpo_lock);
		return (bpobj_space_range(bpo, 0, UINT64_MAX,
		    usedp, compp, uncompp));
	}
}

/*
 * Return the amount of space in the bpobj which is:
 * mintxg < blk_birth <= maxtxg
 */
int
bpobj_space_range(bpobj_t *bpo, uint64_t mintxg, uint64_t maxtxg,
    uint64_t *usedp, uint64_t *compp, uint64_t *uncompp)
{
	struct space_range_arg sra = { 0 };
	int err;

	/*
	 * As an optimization, if they want the whole txg range, just
	 * get bpo_bytes rather than iterating over the bps.
	 */
	if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX && bpo->bpo_havecomp)
		return (bpobj_space(bpo, usedp, compp, uncompp));

	sra.spa = dmu_objset_spa(bpo->bpo_os);
	sra.mintxg = mintxg;
	sra.maxtxg = maxtxg;

	err = bpobj_iterate_nofree(bpo, space_range_cb, &sra, NULL);
	*usedp = sra.used;
	*compp = sra.comp;
	*uncompp = sra.uncomp;
	return (err);
}
Commit	Line	Data
428870ff 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	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
d1fada1e	23	* Copyright (c) 2013 by Delphix. All rights reserved.
428870ff BB	24	*/
	25
	26	#include <sys/bpobj.h>
	27	#include <sys/zfs_context.h>
	28	#include <sys/refcount.h>
330d06f9	29	#include <sys/dsl_pool.h>
753c3839 MA	30	#include <sys/zfeature.h>
	31	#include <sys/zap.h>
	32
	33	/*
	34	* Return an empty bpobj, preferably the empty dummy one (dp_empty_bpobj).
	35	*/
	36	uint64_t
	37	bpobj_alloc_empty(objset_t os, int blocksize, dmu_tx_t tx)
	38	{
753c3839 MA	39	spa_t *spa = dmu_objset_spa(os);
	40	dsl_pool_t *dp = dmu_objset_pool(os);
	41
fa86b5db MA	42	if (spa_feature_is_enabled(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
fa86b5db MA	43	if (!spa_feature_is_active(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
c99c9001	44	ASSERT0(dp->dp_empty_bpobj);
753c3839 MA	45	dp->dp_empty_bpobj =
	46	bpobj_alloc(os, SPA_MAXBLOCKSIZE, tx);
	47	VERIFY(zap_add(os,
	48	DMU_POOL_DIRECTORY_OBJECT,
	49	DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
	50	&dp->dp_empty_bpobj, tx) == 0);
	51	}
fa86b5db	52	spa_feature_incr(spa, SPA_FEATURE_EMPTY_BPOBJ, tx);
753c3839 MA	53	ASSERT(dp->dp_empty_bpobj != 0);
	54	return (dp->dp_empty_bpobj);
	55	} else {
	56	return (bpobj_alloc(os, blocksize, tx));
	57	}
	58	}
	59
	60	void
	61	bpobj_decr_empty(objset_t os, dmu_tx_t tx)
	62	{
753c3839 MA	63	dsl_pool_t *dp = dmu_objset_pool(os);
753c3839 MA	64
fa86b5db MA	65	spa_feature_decr(dmu_objset_spa(os), SPA_FEATURE_EMPTY_BPOBJ, tx);
	66	if (!spa_feature_is_active(dmu_objset_spa(os),
	67	SPA_FEATURE_EMPTY_BPOBJ)) {
753c3839 MA	68	VERIFY3U(0, ==, zap_remove(dp->dp_meta_objset,
	69	DMU_POOL_DIRECTORY_OBJECT,
	70	DMU_POOL_EMPTY_BPOBJ, tx));
	71	VERIFY3U(0, ==, dmu_object_free(os, dp->dp_empty_bpobj, tx));
	72	dp->dp_empty_bpobj = 0;
	73	}
	74	}
428870ff BB	75
	76	uint64_t
	77	bpobj_alloc(objset_t os, int blocksize, dmu_tx_t tx)
	78	{
	79	int size;
	80
	81	if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_BPOBJ_ACCOUNT)
	82	size = BPOBJ_SIZE_V0;
	83	else if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
	84	size = BPOBJ_SIZE_V1;
	85	else
	86	size = sizeof (bpobj_phys_t);
	87
	88	return (dmu_object_alloc(os, DMU_OT_BPOBJ, blocksize,
	89	DMU_OT_BPOBJ_HDR, size, tx));
	90	}
	91
	92	void
	93	bpobj_free(objset_t os, uint64_t obj, dmu_tx_t tx)
	94	{
	95	int64_t i;
	96	bpobj_t bpo;
	97	dmu_object_info_t doi;
	98	int epb;
	99	dmu_buf_t *dbuf = NULL;
	100
753c3839	101	ASSERT(obj != dmu_objset_pool(os)->dp_empty_bpobj);
428870ff BB	102	VERIFY3U(0, ==, bpobj_open(&bpo, os, obj));
	103
	104	mutex_enter(&bpo.bpo_lock);
	105
	106	if (!bpo.bpo_havesubobj \|\| bpo.bpo_phys->bpo_subobjs == 0)
	107	goto out;
	108
	109	VERIFY3U(0, ==, dmu_object_info(os, bpo.bpo_phys->bpo_subobjs, &doi));
	110	epb = doi.doi_data_block_size / sizeof (uint64_t);
	111
	112	for (i = bpo.bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
	113	uint64_t *objarray;
	114	uint64_t offset, blkoff;
	115
	116	offset = i * sizeof (uint64_t);
	117	blkoff = P2PHASE(i, epb);
	118
	119	if (dbuf == NULL \|\| dbuf->db_offset > offset) {
	120	if (dbuf)
	121	dmu_buf_rele(dbuf, FTAG);
	122	VERIFY3U(0, ==, dmu_buf_hold(os,
	123	bpo.bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0));
	124	}
	125
	126	ASSERT3U(offset, >=, dbuf->db_offset);
	127	ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
	128
	129	objarray = dbuf->db_data;
	130	bpobj_free(os, objarray[blkoff], tx);
	131	}
	132	if (dbuf) {
	133	dmu_buf_rele(dbuf, FTAG);
	134	dbuf = NULL;
	135	}
	136	VERIFY3U(0, ==, dmu_object_free(os, bpo.bpo_phys->bpo_subobjs, tx));
	137
	138	out:
	139	mutex_exit(&bpo.bpo_lock);
	140	bpobj_close(&bpo);
	141
	142	VERIFY3U(0, ==, dmu_object_free(os, obj, tx));
	143	}
	144
	145	int
	146	bpobj_open(bpobj_t bpo, objset_t os, uint64_t object)
	147	{
	148	dmu_object_info_t doi;
	149	int err;
	150
	151	err = dmu_object_info(os, object, &doi);
	152	if (err)
	153	return (err);
	154
	155	bzero(bpo, sizeof (*bpo));
	156	mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL);
	157
	158	ASSERT(bpo->bpo_dbuf == NULL);
	159	ASSERT(bpo->bpo_phys == NULL);
	160	ASSERT(object != 0);
	161	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ);
	162	ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR);
	163
572e2857 BB	164	err = dmu_bonus_hold(os, object, bpo, &bpo->bpo_dbuf);
	165	if (err)
	166	return (err);
	167
428870ff BB	168	bpo->bpo_os = os;
	169	bpo->bpo_object = object;
	170	bpo->bpo_epb = doi.doi_data_block_size >> SPA_BLKPTRSHIFT;
	171	bpo->bpo_havecomp = (doi.doi_bonus_size > BPOBJ_SIZE_V0);
	172	bpo->bpo_havesubobj = (doi.doi_bonus_size > BPOBJ_SIZE_V1);
428870ff BB	173	bpo->bpo_phys = bpo->bpo_dbuf->db_data;
	174	return (0);
	175	}
	176
	177	void
	178	bpobj_close(bpobj_t *bpo)
	179	{
	180	/* Lame workaround for closing a bpobj that was never opened. */
	181	if (bpo->bpo_object == 0)
	182	return;
	183
	184	dmu_buf_rele(bpo->bpo_dbuf, bpo);
	185	if (bpo->bpo_cached_dbuf != NULL)
	186	dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
	187	bpo->bpo_dbuf = NULL;
	188	bpo->bpo_phys = NULL;
	189	bpo->bpo_cached_dbuf = NULL;
572e2857	190	bpo->bpo_object = 0;
428870ff BB	191
	192	mutex_destroy(&bpo->bpo_lock);
	193	}
	194
9b67f605 MA	195	static boolean_t
	196	bpobj_hasentries(bpobj_t *bpo)
	197	{
	198	return (bpo->bpo_phys->bpo_num_blkptrs != 0 \|\|
	199	(bpo->bpo_havesubobj && bpo->bpo_phys->bpo_num_subobjs != 0));
	200	}
	201
428870ff BB	202	static int
	203	bpobj_iterate_impl(bpobj_t bpo, bpobj_itor_t func, void arg, dmu_tx_t *tx,
	204	boolean_t free)
	205	{
	206	dmu_object_info_t doi;
	207	int epb;
	208	int64_t i;
	209	int err = 0;
	210	dmu_buf_t *dbuf = NULL;
	211
	212	mutex_enter(&bpo->bpo_lock);
	213
	214	if (free)
	215	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	216
	217	for (i = bpo->bpo_phys->bpo_num_blkptrs - 1; i >= 0; i--) {
	218	blkptr_t *bparray;
	219	blkptr_t *bp;
	220	uint64_t offset, blkoff;
	221
	222	offset = i * sizeof (blkptr_t);
	223	blkoff = P2PHASE(i, bpo->bpo_epb);
	224
	225	if (dbuf == NULL \|\| dbuf->db_offset > offset) {
	226	if (dbuf)
	227	dmu_buf_rele(dbuf, FTAG);
	228	err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object, offset,
	229	FTAG, &dbuf, 0);
	230	if (err)
	231	break;
	232	}
	233
	234	ASSERT3U(offset, >=, dbuf->db_offset);
	235	ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
	236
	237	bparray = dbuf->db_data;
	238	bp = &bparray[blkoff];
	239	err = func(arg, bp, tx);
	240	if (err)
	241	break;
	242	if (free) {
	243	bpo->bpo_phys->bpo_bytes -=
	244	bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
	245	ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
	246	if (bpo->bpo_havecomp) {
	247	bpo->bpo_phys->bpo_comp -= BP_GET_PSIZE(bp);
	248	bpo->bpo_phys->bpo_uncomp -= BP_GET_UCSIZE(bp);
	249	}
	250	bpo->bpo_phys->bpo_num_blkptrs--;
	251	ASSERT3S(bpo->bpo_phys->bpo_num_blkptrs, >=, 0);
	252	}
	253	}
	254	if (dbuf) {
	255	dmu_buf_rele(dbuf, FTAG);
	256	dbuf = NULL;
	257	}
	258	if (free) {
	259	i++;
	260	VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os, bpo->bpo_object,
	261	i * sizeof (blkptr_t), -1ULL, tx));
	262	}
	263	if (err \|\| !bpo->bpo_havesubobj \|\| bpo->bpo_phys->bpo_subobjs == 0)
	264	goto out;
	265
266	ASSERT(bpo->bpo_havecomp);
267	err = dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi);
572e2857 BB	268	if (err) {
572e2857 BB	269	mutex_exit(&bpo->bpo_lock);
428870ff	270	return (err);
572e2857	271	}
fa86b5db	272	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
428870ff BB	273	epb = doi.doi_data_block_size / sizeof (uint64_t);
	274
	275	for (i = bpo->bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
	276	uint64_t *objarray;
	277	uint64_t offset, blkoff;
	278	bpobj_t sublist;
	279	uint64_t used_before, comp_before, uncomp_before;
	280	uint64_t used_after, comp_after, uncomp_after;
	281
	282	offset = i * sizeof (uint64_t);
	283	blkoff = P2PHASE(i, epb);
	284
	285	if (dbuf == NULL \|\| dbuf->db_offset > offset) {
	286	if (dbuf)
	287	dmu_buf_rele(dbuf, FTAG);
	288	err = dmu_buf_hold(bpo->bpo_os,
	289	bpo->bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0);
	290	if (err)
	291	break;
	292	}
	293
	294	ASSERT3U(offset, >=, dbuf->db_offset);
	295	ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
	296
	297	objarray = dbuf->db_data;
	298	err = bpobj_open(&sublist, bpo->bpo_os, objarray[blkoff]);
	299	if (err)
	300	break;
	301	if (free) {
	302	err = bpobj_space(&sublist,
	303	&used_before, &comp_before, &uncomp_before);
50f9ea01 WA	304	if (err != 0) {
50f9ea01 WA	305	bpobj_close(&sublist);
428870ff	306	break;
50f9ea01	307	}
428870ff BB	308	}
	309	err = bpobj_iterate_impl(&sublist, func, arg, tx, free);
	310	if (free) {
	311	VERIFY3U(0, ==, bpobj_space(&sublist,
	312	&used_after, &comp_after, &uncomp_after));
	313	bpo->bpo_phys->bpo_bytes -= used_before - used_after;
	314	ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
572e2857	315	bpo->bpo_phys->bpo_comp -= comp_before - comp_after;
428870ff BB	316	bpo->bpo_phys->bpo_uncomp -=
	317	uncomp_before - uncomp_after;
	318	}
	319
	320	bpobj_close(&sublist);
	321	if (err)
	322	break;
	323	if (free) {
	324	err = dmu_object_free(bpo->bpo_os,
	325	objarray[blkoff], tx);
	326	if (err)
	327	break;
	328	bpo->bpo_phys->bpo_num_subobjs--;
	329	ASSERT3S(bpo->bpo_phys->bpo_num_subobjs, >=, 0);
	330	}
	331	}
	332	if (dbuf) {
	333	dmu_buf_rele(dbuf, FTAG);
	334	dbuf = NULL;
	335	}
	336	if (free) {
	337	VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os,
	338	bpo->bpo_phys->bpo_subobjs,
	339	(i + 1) * sizeof (uint64_t), -1ULL, tx));
	340	}
	341
	342	out:
	343	/* If there are no entries, there should be no bytes. */
9b67f605 MA	344	if (!bpobj_hasentries(bpo)) {
	345	ASSERT0(bpo->bpo_phys->bpo_bytes);
	346	ASSERT0(bpo->bpo_phys->bpo_comp);
	347	ASSERT0(bpo->bpo_phys->bpo_uncomp);
	348	}
428870ff BB	349
	350	mutex_exit(&bpo->bpo_lock);
	351	return (err);
	352	}
	353
	354	/*
	355	* Iterate and remove the entries. If func returns nonzero, iteration
	356	* will stop and that entry will not be removed.
	357	*/
	358	int
	359	bpobj_iterate(bpobj_t bpo, bpobj_itor_t func, void arg, dmu_tx_t *tx)
	360	{
	361	return (bpobj_iterate_impl(bpo, func, arg, tx, B_TRUE));
	362	}
	363
	364	/*
	365	* Iterate the entries. If func returns nonzero, iteration will stop.
	366	*/
	367	int
	368	bpobj_iterate_nofree(bpobj_t bpo, bpobj_itor_t func, void arg, dmu_tx_t *tx)
	369	{
	370	return (bpobj_iterate_impl(bpo, func, arg, tx, B_FALSE));
	371	}
	372
	373	void
	374	bpobj_enqueue_subobj(bpobj_t bpo, uint64_t subobj, dmu_tx_t tx)
	375	{
	376	bpobj_t subbpo;
572e2857	377	uint64_t used, comp, uncomp, subsubobjs;
13fe0198	378	ASSERTV(dmu_object_info_t doi);
428870ff BB	379
	380	ASSERT(bpo->bpo_havesubobj);
	381	ASSERT(bpo->bpo_havecomp);
753c3839 MA	382	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
	383
	384	if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj) {
	385	bpobj_decr_empty(bpo->bpo_os, tx);
	386	return;
	387	}
428870ff BB	388
	389	VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
	390	VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
428870ff	391
9b67f605	392	if (!bpobj_hasentries(&subbpo)) {
428870ff	393	/* No point in having an empty subobj. */
572e2857	394	bpobj_close(&subbpo);
428870ff BB	395	bpobj_free(bpo->bpo_os, subobj, tx);
	396	return;
	397	}
	398
	399	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	400	if (bpo->bpo_phys->bpo_subobjs == 0) {
	401	bpo->bpo_phys->bpo_subobjs = dmu_object_alloc(bpo->bpo_os,
	402	DMU_OT_BPOBJ_SUBOBJ, SPA_MAXBLOCKSIZE, DMU_OT_NONE, 0, tx);
	403	}
	404
13fe0198 MA	405	ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi));
	406	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
	407
428870ff BB	408	mutex_enter(&bpo->bpo_lock);
	409	dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
	410	bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
	411	sizeof (subobj), &subobj, tx);
	412	bpo->bpo_phys->bpo_num_subobjs++;
572e2857 BB	413
	414	/*
	415	* If subobj has only one block of subobjs, then move subobj's
	416	* subobjs to bpo's subobj list directly. This reduces
	417	* recursion in bpobj_iterate due to nested subobjs.
	418	*/
	419	subsubobjs = subbpo.bpo_phys->bpo_subobjs;
	420	if (subsubobjs != 0) {
	421	dmu_object_info_t doi;
	422
	423	VERIFY3U(0, ==, dmu_object_info(bpo->bpo_os, subsubobjs, &doi));
	424	if (doi.doi_max_offset == doi.doi_data_block_size) {
	425	dmu_buf_t *subdb;
	426	uint64_t numsubsub = subbpo.bpo_phys->bpo_num_subobjs;
	427
	428	VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, subsubobjs,
	429	0, FTAG, &subdb, 0));
d1fada1e MA	430	/*
	431	* Make sure that we are not asking dmu_write()
	432	* to write more data than we have in our buffer.
	433	*/
	434	VERIFY3U(subdb->db_size, >=,
	435	numsubsub * sizeof (subobj));
572e2857 BB	436	dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
	437	bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
	438	numsubsub * sizeof (subobj), subdb->db_data, tx);
	439	dmu_buf_rele(subdb, FTAG);
	440	bpo->bpo_phys->bpo_num_subobjs += numsubsub;
	441
	442	dmu_buf_will_dirty(subbpo.bpo_dbuf, tx);
	443	subbpo.bpo_phys->bpo_subobjs = 0;
	444	VERIFY3U(0, ==, dmu_object_free(bpo->bpo_os,
	445	subsubobjs, tx));
	446	}
	447	}
428870ff BB	448	bpo->bpo_phys->bpo_bytes += used;
	449	bpo->bpo_phys->bpo_comp += comp;
	450	bpo->bpo_phys->bpo_uncomp += uncomp;
	451	mutex_exit(&bpo->bpo_lock);
572e2857 BB	452
572e2857 BB	453	bpobj_close(&subbpo);
428870ff BB	454	}
	455
	456	void
	457	bpobj_enqueue(bpobj_t bpo, const blkptr_t bp, dmu_tx_t *tx)
	458	{
	459	blkptr_t stored_bp = *bp;
	460	uint64_t offset;
	461	int blkoff;
	462	blkptr_t *bparray;
	463
	464	ASSERT(!BP_IS_HOLE(bp));
753c3839	465	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
428870ff	466
9b67f605 MA	467	if (BP_IS_EMBEDDED(bp)) {
	468	/*
	469	* The bpobj will compress better without the payload.
	470	*
	471	* Note that we store EMBEDDED bp's because they have an
	472	* uncompressed size, which must be accounted for. An
	473	* alternative would be to add their size to bpo_uncomp
	474	* without storing the bp, but that would create additional
	475	* complications: bpo_uncomp would be inconsistent with the
	476	* set of BP's stored, and bpobj_iterate() wouldn't visit
	477	* all the space accounted for in the bpobj.
	478	*/
	479	bzero(&stored_bp, sizeof (stored_bp));
	480	stored_bp.blk_prop = bp->blk_prop;
	481	stored_bp.blk_birth = bp->blk_birth;
	482	} else if (!BP_GET_DEDUP(bp)) {
	483	/* The bpobj will compress better without the checksum */
	484	bzero(&stored_bp.blk_cksum, sizeof (stored_bp.blk_cksum));
	485	}
	486
428870ff BB	487	/* We never need the fill count. */
	488	stored_bp.blk_fill = 0;
	489
428870ff BB	490	mutex_enter(&bpo->bpo_lock);
	491
	492	offset = bpo->bpo_phys->bpo_num_blkptrs * sizeof (stored_bp);
	493	blkoff = P2PHASE(bpo->bpo_phys->bpo_num_blkptrs, bpo->bpo_epb);
	494
	495	if (bpo->bpo_cached_dbuf == NULL \|\|
	496	offset < bpo->bpo_cached_dbuf->db_offset \|\|
	497	offset >= bpo->bpo_cached_dbuf->db_offset +
	498	bpo->bpo_cached_dbuf->db_size) {
	499	if (bpo->bpo_cached_dbuf)
	500	dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
	501	VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
	502	offset, bpo, &bpo->bpo_cached_dbuf, 0));
	503	}
	504
	505	dmu_buf_will_dirty(bpo->bpo_cached_dbuf, tx);
	506	bparray = bpo->bpo_cached_dbuf->db_data;
	507	bparray[blkoff] = stored_bp;
	508
	509	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	510	bpo->bpo_phys->bpo_num_blkptrs++;
	511	bpo->bpo_phys->bpo_bytes +=
	512	bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
	513	if (bpo->bpo_havecomp) {
	514	bpo->bpo_phys->bpo_comp += BP_GET_PSIZE(bp);
	515	bpo->bpo_phys->bpo_uncomp += BP_GET_UCSIZE(bp);
	516	}
	517	mutex_exit(&bpo->bpo_lock);
	518	}
	519
	520	struct space_range_arg {
	521	spa_t *spa;
	522	uint64_t mintxg;
	523	uint64_t maxtxg;
	524	uint64_t used;
	525	uint64_t comp;
	526	uint64_t uncomp;
	527	};
	528
	529	/* ARGSUSED */
	530	static int
	531	space_range_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	532	{
	533	struct space_range_arg *sra = arg;
	534
	535	if (bp->blk_birth > sra->mintxg && bp->blk_birth <= sra->maxtxg) {
330d06f9 MA	536	if (dsl_pool_sync_context(spa_get_dsl(sra->spa)))
	537	sra->used += bp_get_dsize_sync(sra->spa, bp);
	538	else
	539	sra->used += bp_get_dsize(sra->spa, bp);
428870ff BB	540	sra->comp += BP_GET_PSIZE(bp);
	541	sra->uncomp += BP_GET_UCSIZE(bp);
	542	}
	543	return (0);
	544	}
	545
	546	int
	547	bpobj_space(bpobj_t bpo, uint64_t usedp, uint64_t compp, uint64_t uncompp)
	548	{
	549	mutex_enter(&bpo->bpo_lock);
	550
	551	*usedp = bpo->bpo_phys->bpo_bytes;
	552	if (bpo->bpo_havecomp) {
	553	*compp = bpo->bpo_phys->bpo_comp;
	554	*uncompp = bpo->bpo_phys->bpo_uncomp;
	555	mutex_exit(&bpo->bpo_lock);
	556	return (0);
	557	} else {
	558	mutex_exit(&bpo->bpo_lock);
	559	return (bpobj_space_range(bpo, 0, UINT64_MAX,
	560	usedp, compp, uncompp));
	561	}
	562	}
	563
	564	/*
	565	* Return the amount of space in the bpobj which is:
	566	* mintxg < blk_birth <= maxtxg
	567	*/
	568	int
	569	bpobj_space_range(bpobj_t *bpo, uint64_t mintxg, uint64_t maxtxg,
	570	uint64_t usedp, uint64_t compp, uint64_t *uncompp)
	571	{
	572	struct space_range_arg sra = { 0 };
	573	int err;
	574
	575	/*
	576	* As an optimization, if they want the whole txg range, just
	577	* get bpo_bytes rather than iterating over the bps.
	578	*/
	579	if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX && bpo->bpo_havecomp)
	580	return (bpobj_space(bpo, usedp, compp, uncompp));
	581
	582	sra.spa = dmu_objset_spa(bpo->bpo_os);
	583	sra.mintxg = mintxg;
	584	sra.maxtxg = maxtxg;
	585
	586	err = bpobj_iterate_nofree(bpo, space_range_cb, &sra, NULL);
	587	*usedp = sra.used;
	588	*compp = sra.comp;
	589	*uncompp = sra.uncomp;
	590	return (err);
	591	}