[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) 2011, 2016 by Delphix. All rights reserved.
 * Copyright (c) 2017 Datto Inc.
 */

#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_OLD_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 (!bpobj_hasentries(bpo))
		goto out;

	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) {
		VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os, bpo->bpo_object,
		    (i + 1) * sizeof (blkptr_t), DMU_OBJECT_END, 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), DMU_OBJECT_END, 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;

	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;
	}

	mutex_enter(&bpo->bpo_lock);
	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_OLD_MAXBLOCKSIZE,
		    DMU_OT_NONE, 0, tx);
	}

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

	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.
df7eeccc	23	* Copyright (c) 2011, 2016 by Delphix. All rights reserved.
9210e43a	24	* Copyright (c) 2017 Datto Inc.
428870ff BB	25	*/
	26
	27	#include <sys/bpobj.h>
	28	#include <sys/zfs_context.h>
	29	#include <sys/refcount.h>
330d06f9	30	#include <sys/dsl_pool.h>
753c3839 MA	31	#include <sys/zfeature.h>
	32	#include <sys/zap.h>
	33
	34	/*
	35	* Return an empty bpobj, preferably the empty dummy one (dp_empty_bpobj).
	36	*/
	37	uint64_t
	38	bpobj_alloc_empty(objset_t os, int blocksize, dmu_tx_t tx)
	39	{
753c3839 MA	40	spa_t *spa = dmu_objset_spa(os);
	41	dsl_pool_t *dp = dmu_objset_pool(os);
	42
fa86b5db MA	43	if (spa_feature_is_enabled(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
fa86b5db MA	44	if (!spa_feature_is_active(spa, SPA_FEATURE_EMPTY_BPOBJ)) {
c99c9001	45	ASSERT0(dp->dp_empty_bpobj);
753c3839	46	dp->dp_empty_bpobj =
f1512ee6	47	bpobj_alloc(os, SPA_OLD_MAXBLOCKSIZE, tx);
753c3839 MA	48	VERIFY(zap_add(os,
	49	DMU_POOL_DIRECTORY_OBJECT,
	50	DMU_POOL_EMPTY_BPOBJ, sizeof (uint64_t), 1,
	51	&dp->dp_empty_bpobj, tx) == 0);
	52	}
fa86b5db	53	spa_feature_incr(spa, SPA_FEATURE_EMPTY_BPOBJ, tx);
753c3839 MA	54	ASSERT(dp->dp_empty_bpobj != 0);
	55	return (dp->dp_empty_bpobj);
	56	} else {
	57	return (bpobj_alloc(os, blocksize, tx));
	58	}
	59	}
	60
	61	void
	62	bpobj_decr_empty(objset_t os, dmu_tx_t tx)
	63	{
753c3839 MA	64	dsl_pool_t *dp = dmu_objset_pool(os);
753c3839 MA	65
fa86b5db MA	66	spa_feature_decr(dmu_objset_spa(os), SPA_FEATURE_EMPTY_BPOBJ, tx);
	67	if (!spa_feature_is_active(dmu_objset_spa(os),
	68	SPA_FEATURE_EMPTY_BPOBJ)) {
753c3839 MA	69	VERIFY3U(0, ==, zap_remove(dp->dp_meta_objset,
	70	DMU_POOL_DIRECTORY_OBJECT,
	71	DMU_POOL_EMPTY_BPOBJ, tx));
	72	VERIFY3U(0, ==, dmu_object_free(os, dp->dp_empty_bpobj, tx));
	73	dp->dp_empty_bpobj = 0;
	74	}
	75	}
428870ff BB	76
	77	uint64_t
	78	bpobj_alloc(objset_t os, int blocksize, dmu_tx_t tx)
	79	{
	80	int size;
	81
	82	if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_BPOBJ_ACCOUNT)
	83	size = BPOBJ_SIZE_V0;
	84	else if (spa_version(dmu_objset_spa(os)) < SPA_VERSION_DEADLISTS)
	85	size = BPOBJ_SIZE_V1;
	86	else
	87	size = sizeof (bpobj_phys_t);
	88
	89	return (dmu_object_alloc(os, DMU_OT_BPOBJ, blocksize,
	90	DMU_OT_BPOBJ_HDR, size, tx));
	91	}
	92
	93	void
	94	bpobj_free(objset_t os, uint64_t obj, dmu_tx_t tx)
	95	{
	96	int64_t i;
	97	bpobj_t bpo;
	98	dmu_object_info_t doi;
	99	int epb;
	100	dmu_buf_t *dbuf = NULL;
	101
753c3839	102	ASSERT(obj != dmu_objset_pool(os)->dp_empty_bpobj);
428870ff BB	103	VERIFY3U(0, ==, bpobj_open(&bpo, os, obj));
	104
	105	mutex_enter(&bpo.bpo_lock);
	106
	107	if (!bpo.bpo_havesubobj \|\| bpo.bpo_phys->bpo_subobjs == 0)
	108	goto out;
	109
	110	VERIFY3U(0, ==, dmu_object_info(os, bpo.bpo_phys->bpo_subobjs, &doi));
	111	epb = doi.doi_data_block_size / sizeof (uint64_t);
	112
	113	for (i = bpo.bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
	114	uint64_t *objarray;
	115	uint64_t offset, blkoff;
	116
	117	offset = i * sizeof (uint64_t);
	118	blkoff = P2PHASE(i, epb);
	119
	120	if (dbuf == NULL \|\| dbuf->db_offset > offset) {
	121	if (dbuf)
	122	dmu_buf_rele(dbuf, FTAG);
	123	VERIFY3U(0, ==, dmu_buf_hold(os,
	124	bpo.bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0));
	125	}
	126
	127	ASSERT3U(offset, >=, dbuf->db_offset);
	128	ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
	129
	130	objarray = dbuf->db_data;
	131	bpobj_free(os, objarray[blkoff], tx);
	132	}
	133	if (dbuf) {
	134	dmu_buf_rele(dbuf, FTAG);
	135	dbuf = NULL;
	136	}
	137	VERIFY3U(0, ==, dmu_object_free(os, bpo.bpo_phys->bpo_subobjs, tx));
	138
	139	out:
	140	mutex_exit(&bpo.bpo_lock);
	141	bpobj_close(&bpo);
	142
	143	VERIFY3U(0, ==, dmu_object_free(os, obj, tx));
	144	}
	145
	146	int
	147	bpobj_open(bpobj_t bpo, objset_t os, uint64_t object)
	148	{
	149	dmu_object_info_t doi;
	150	int err;
	151
	152	err = dmu_object_info(os, object, &doi);
	153	if (err)
	154	return (err);
	155
	156	bzero(bpo, sizeof (*bpo));
	157	mutex_init(&bpo->bpo_lock, NULL, MUTEX_DEFAULT, NULL);
	158
	159	ASSERT(bpo->bpo_dbuf == NULL);
	160	ASSERT(bpo->bpo_phys == NULL);
	161	ASSERT(object != 0);
	162	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ);
	163	ASSERT3U(doi.doi_bonus_type, ==, DMU_OT_BPOBJ_HDR);
	164
572e2857 BB	165	err = dmu_bonus_hold(os, object, bpo, &bpo->bpo_dbuf);
	166	if (err)
	167	return (err);
	168
428870ff BB	169	bpo->bpo_os = os;
	170	bpo->bpo_object = object;
	171	bpo->bpo_epb = doi.doi_data_block_size >> SPA_BLKPTRSHIFT;
	172	bpo->bpo_havecomp = (doi.doi_bonus_size > BPOBJ_SIZE_V0);
	173	bpo->bpo_havesubobj = (doi.doi_bonus_size > BPOBJ_SIZE_V1);
428870ff BB	174	bpo->bpo_phys = bpo->bpo_dbuf->db_data;
	175	return (0);
	176	}
	177
	178	void
	179	bpobj_close(bpobj_t *bpo)
	180	{
	181	/* Lame workaround for closing a bpobj that was never opened. */
	182	if (bpo->bpo_object == 0)
	183	return;
	184
	185	dmu_buf_rele(bpo->bpo_dbuf, bpo);
	186	if (bpo->bpo_cached_dbuf != NULL)
	187	dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
	188	bpo->bpo_dbuf = NULL;
	189	bpo->bpo_phys = NULL;
	190	bpo->bpo_cached_dbuf = NULL;
572e2857	191	bpo->bpo_object = 0;
428870ff BB	192
	193	mutex_destroy(&bpo->bpo_lock);
	194	}
	195
9b67f605 MA	196	static boolean_t
	197	bpobj_hasentries(bpobj_t *bpo)
	198	{
	199	return (bpo->bpo_phys->bpo_num_blkptrs != 0 \|\|
	200	(bpo->bpo_havesubobj && bpo->bpo_phys->bpo_num_subobjs != 0));
	201	}
	202
428870ff BB	203	static int
	204	bpobj_iterate_impl(bpobj_t bpo, bpobj_itor_t func, void arg, dmu_tx_t *tx,
	205	boolean_t free)
	206	{
	207	dmu_object_info_t doi;
	208	int epb;
	209	int64_t i;
	210	int err = 0;
	211	dmu_buf_t *dbuf = NULL;
	212
	213	mutex_enter(&bpo->bpo_lock);
	214
9210e43a AP	215	if (!bpobj_hasentries(bpo))
	216	goto out;
	217
428870ff BB	218	if (free)
	219	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	220
	221	for (i = bpo->bpo_phys->bpo_num_blkptrs - 1; i >= 0; i--) {
	222	blkptr_t *bparray;
	223	blkptr_t *bp;
	224	uint64_t offset, blkoff;
	225
	226	offset = i * sizeof (blkptr_t);
	227	blkoff = P2PHASE(i, bpo->bpo_epb);
	228
	229	if (dbuf == NULL \|\| dbuf->db_offset > offset) {
	230	if (dbuf)
	231	dmu_buf_rele(dbuf, FTAG);
	232	err = dmu_buf_hold(bpo->bpo_os, bpo->bpo_object, offset,
	233	FTAG, &dbuf, 0);
	234	if (err)
	235	break;
	236	}
	237
	238	ASSERT3U(offset, >=, dbuf->db_offset);
	239	ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
	240
	241	bparray = dbuf->db_data;
	242	bp = &bparray[blkoff];
	243	err = func(arg, bp, tx);
	244	if (err)
	245	break;
	246	if (free) {
	247	bpo->bpo_phys->bpo_bytes -=
	248	bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
	249	ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
	250	if (bpo->bpo_havecomp) {
	251	bpo->bpo_phys->bpo_comp -= BP_GET_PSIZE(bp);
	252	bpo->bpo_phys->bpo_uncomp -= BP_GET_UCSIZE(bp);
	253	}
	254	bpo->bpo_phys->bpo_num_blkptrs--;
	255	ASSERT3S(bpo->bpo_phys->bpo_num_blkptrs, >=, 0);
	256	}
	257	}
	258	if (dbuf) {
	259	dmu_buf_rele(dbuf, FTAG);
	260	dbuf = NULL;
	261	}
	262	if (free) {
428870ff	263	VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os, bpo->bpo_object,
ee45fbd8	264	(i + 1) * sizeof (blkptr_t), DMU_OBJECT_END, tx));
428870ff BB	265	}
	266	if (err \|\| !bpo->bpo_havesubobj \|\| bpo->bpo_phys->bpo_subobjs == 0)
	267	goto out;
	268
	269	ASSERT(bpo->bpo_havecomp);
	270	err = dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi);
572e2857 BB	271	if (err) {
572e2857 BB	272	mutex_exit(&bpo->bpo_lock);
428870ff	273	return (err);
572e2857	274	}
fa86b5db	275	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
428870ff BB	276	epb = doi.doi_data_block_size / sizeof (uint64_t);
	277
	278	for (i = bpo->bpo_phys->bpo_num_subobjs - 1; i >= 0; i--) {
	279	uint64_t *objarray;
	280	uint64_t offset, blkoff;
	281	bpobj_t sublist;
	282	uint64_t used_before, comp_before, uncomp_before;
	283	uint64_t used_after, comp_after, uncomp_after;
	284
	285	offset = i * sizeof (uint64_t);
	286	blkoff = P2PHASE(i, epb);
	287
	288	if (dbuf == NULL \|\| dbuf->db_offset > offset) {
	289	if (dbuf)
	290	dmu_buf_rele(dbuf, FTAG);
	291	err = dmu_buf_hold(bpo->bpo_os,
	292	bpo->bpo_phys->bpo_subobjs, offset, FTAG, &dbuf, 0);
	293	if (err)
	294	break;
	295	}
	296
	297	ASSERT3U(offset, >=, dbuf->db_offset);
	298	ASSERT3U(offset, <, dbuf->db_offset + dbuf->db_size);
	299
	300	objarray = dbuf->db_data;
	301	err = bpobj_open(&sublist, bpo->bpo_os, objarray[blkoff]);
	302	if (err)
	303	break;
	304	if (free) {
	305	err = bpobj_space(&sublist,
	306	&used_before, &comp_before, &uncomp_before);
50f9ea01 WA	307	if (err != 0) {
50f9ea01 WA	308	bpobj_close(&sublist);
428870ff	309	break;
50f9ea01	310	}
428870ff BB	311	}
	312	err = bpobj_iterate_impl(&sublist, func, arg, tx, free);
	313	if (free) {
	314	VERIFY3U(0, ==, bpobj_space(&sublist,
	315	&used_after, &comp_after, &uncomp_after));
	316	bpo->bpo_phys->bpo_bytes -= used_before - used_after;
	317	ASSERT3S(bpo->bpo_phys->bpo_bytes, >=, 0);
572e2857	318	bpo->bpo_phys->bpo_comp -= comp_before - comp_after;
428870ff BB	319	bpo->bpo_phys->bpo_uncomp -=
	320	uncomp_before - uncomp_after;
	321	}
	322
	323	bpobj_close(&sublist);
	324	if (err)
	325	break;
	326	if (free) {
	327	err = dmu_object_free(bpo->bpo_os,
	328	objarray[blkoff], tx);
	329	if (err)
	330	break;
	331	bpo->bpo_phys->bpo_num_subobjs--;
	332	ASSERT3S(bpo->bpo_phys->bpo_num_subobjs, >=, 0);
	333	}
	334	}
	335	if (dbuf) {
	336	dmu_buf_rele(dbuf, FTAG);
	337	dbuf = NULL;
	338	}
	339	if (free) {
	340	VERIFY3U(0, ==, dmu_free_range(bpo->bpo_os,
	341	bpo->bpo_phys->bpo_subobjs,
ee45fbd8	342	(i + 1) * sizeof (uint64_t), DMU_OBJECT_END, tx));
428870ff BB	343	}
	344
	345	out:
	346	/* If there are no entries, there should be no bytes. */
9b67f605 MA	347	if (!bpobj_hasentries(bpo)) {
	348	ASSERT0(bpo->bpo_phys->bpo_bytes);
	349	ASSERT0(bpo->bpo_phys->bpo_comp);
	350	ASSERT0(bpo->bpo_phys->bpo_uncomp);
	351	}
428870ff BB	352
	353	mutex_exit(&bpo->bpo_lock);
	354	return (err);
	355	}
	356
	357	/*
	358	* Iterate and remove the entries. If func returns nonzero, iteration
	359	* will stop and that entry will not be removed.
	360	*/
	361	int
	362	bpobj_iterate(bpobj_t bpo, bpobj_itor_t func, void arg, dmu_tx_t *tx)
	363	{
	364	return (bpobj_iterate_impl(bpo, func, arg, tx, B_TRUE));
	365	}
	366
	367	/*
	368	* Iterate the entries. If func returns nonzero, iteration will stop.
	369	*/
	370	int
	371	bpobj_iterate_nofree(bpobj_t bpo, bpobj_itor_t func, void arg, dmu_tx_t *tx)
	372	{
	373	return (bpobj_iterate_impl(bpo, func, arg, tx, B_FALSE));
	374	}
	375
	376	void
	377	bpobj_enqueue_subobj(bpobj_t bpo, uint64_t subobj, dmu_tx_t tx)
	378	{
	379	bpobj_t subbpo;
572e2857	380	uint64_t used, comp, uncomp, subsubobjs;
428870ff BB	381
	382	ASSERT(bpo->bpo_havesubobj);
	383	ASSERT(bpo->bpo_havecomp);
753c3839 MA	384	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
	385
	386	if (subobj == dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj) {
	387	bpobj_decr_empty(bpo->bpo_os, tx);
	388	return;
	389	}
428870ff BB	390
	391	VERIFY3U(0, ==, bpobj_open(&subbpo, bpo->bpo_os, subobj));
	392	VERIFY3U(0, ==, bpobj_space(&subbpo, &used, &comp, &uncomp));
428870ff	393
9b67f605	394	if (!bpobj_hasentries(&subbpo)) {
428870ff	395	/* No point in having an empty subobj. */
572e2857	396	bpobj_close(&subbpo);
428870ff BB	397	bpobj_free(bpo->bpo_os, subobj, tx);
	398	return;
	399	}
	400
df7eeccc	401	mutex_enter(&bpo->bpo_lock);
428870ff BB	402	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	403	if (bpo->bpo_phys->bpo_subobjs == 0) {
	404	bpo->bpo_phys->bpo_subobjs = dmu_object_alloc(bpo->bpo_os,
f1512ee6 MA	405	DMU_OT_BPOBJ_SUBOBJ, SPA_OLD_MAXBLOCKSIZE,
f1512ee6 MA	406	DMU_OT_NONE, 0, tx);
428870ff BB	407	}
428870ff BB	408
1c27024e	409	ASSERTV(dmu_object_info_t doi);
13fe0198 MA	410	ASSERT0(dmu_object_info(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs, &doi));
	411	ASSERT3U(doi.doi_type, ==, DMU_OT_BPOBJ_SUBOBJ);
	412
428870ff BB	413	dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
	414	bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
	415	sizeof (subobj), &subobj, tx);
	416	bpo->bpo_phys->bpo_num_subobjs++;
572e2857 BB	417
	418	/*
	419	* If subobj has only one block of subobjs, then move subobj's
	420	* subobjs to bpo's subobj list directly. This reduces
	421	* recursion in bpobj_iterate due to nested subobjs.
	422	*/
	423	subsubobjs = subbpo.bpo_phys->bpo_subobjs;
	424	if (subsubobjs != 0) {
	425	dmu_object_info_t doi;
	426
	427	VERIFY3U(0, ==, dmu_object_info(bpo->bpo_os, subsubobjs, &doi));
	428	if (doi.doi_max_offset == doi.doi_data_block_size) {
	429	dmu_buf_t *subdb;
	430	uint64_t numsubsub = subbpo.bpo_phys->bpo_num_subobjs;
	431
	432	VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, subsubobjs,
	433	0, FTAG, &subdb, 0));
d1fada1e MA	434	/*
	435	* Make sure that we are not asking dmu_write()
	436	* to write more data than we have in our buffer.
	437	*/
	438	VERIFY3U(subdb->db_size, >=,
	439	numsubsub * sizeof (subobj));
572e2857 BB	440	dmu_write(bpo->bpo_os, bpo->bpo_phys->bpo_subobjs,
	441	bpo->bpo_phys->bpo_num_subobjs * sizeof (subobj),
	442	numsubsub * sizeof (subobj), subdb->db_data, tx);
	443	dmu_buf_rele(subdb, FTAG);
	444	bpo->bpo_phys->bpo_num_subobjs += numsubsub;
	445
	446	dmu_buf_will_dirty(subbpo.bpo_dbuf, tx);
	447	subbpo.bpo_phys->bpo_subobjs = 0;
	448	VERIFY3U(0, ==, dmu_object_free(bpo->bpo_os,
	449	subsubobjs, tx));
	450	}
	451	}
428870ff BB	452	bpo->bpo_phys->bpo_bytes += used;
	453	bpo->bpo_phys->bpo_comp += comp;
	454	bpo->bpo_phys->bpo_uncomp += uncomp;
	455	mutex_exit(&bpo->bpo_lock);
572e2857 BB	456
572e2857 BB	457	bpobj_close(&subbpo);
428870ff BB	458	}
	459
	460	void
	461	bpobj_enqueue(bpobj_t bpo, const blkptr_t bp, dmu_tx_t *tx)
	462	{
	463	blkptr_t stored_bp = *bp;
	464	uint64_t offset;
	465	int blkoff;
	466	blkptr_t *bparray;
	467
	468	ASSERT(!BP_IS_HOLE(bp));
753c3839	469	ASSERT(bpo->bpo_object != dmu_objset_pool(bpo->bpo_os)->dp_empty_bpobj);
428870ff	470
9b67f605 MA	471	if (BP_IS_EMBEDDED(bp)) {
	472	/*
	473	* The bpobj will compress better without the payload.
	474	*
	475	* Note that we store EMBEDDED bp's because they have an
	476	* uncompressed size, which must be accounted for. An
	477	* alternative would be to add their size to bpo_uncomp
	478	* without storing the bp, but that would create additional
	479	* complications: bpo_uncomp would be inconsistent with the
	480	* set of BP's stored, and bpobj_iterate() wouldn't visit
	481	* all the space accounted for in the bpobj.
	482	*/
	483	bzero(&stored_bp, sizeof (stored_bp));
	484	stored_bp.blk_prop = bp->blk_prop;
	485	stored_bp.blk_birth = bp->blk_birth;
	486	} else if (!BP_GET_DEDUP(bp)) {
	487	/* The bpobj will compress better without the checksum */
	488	bzero(&stored_bp.blk_cksum, sizeof (stored_bp.blk_cksum));
	489	}
	490
428870ff BB	491	/* We never need the fill count. */
	492	stored_bp.blk_fill = 0;
	493
428870ff BB	494	mutex_enter(&bpo->bpo_lock);
	495
	496	offset = bpo->bpo_phys->bpo_num_blkptrs * sizeof (stored_bp);
	497	blkoff = P2PHASE(bpo->bpo_phys->bpo_num_blkptrs, bpo->bpo_epb);
	498
	499	if (bpo->bpo_cached_dbuf == NULL \|\|
	500	offset < bpo->bpo_cached_dbuf->db_offset \|\|
	501	offset >= bpo->bpo_cached_dbuf->db_offset +
	502	bpo->bpo_cached_dbuf->db_size) {
	503	if (bpo->bpo_cached_dbuf)
	504	dmu_buf_rele(bpo->bpo_cached_dbuf, bpo);
	505	VERIFY3U(0, ==, dmu_buf_hold(bpo->bpo_os, bpo->bpo_object,
	506	offset, bpo, &bpo->bpo_cached_dbuf, 0));
	507	}
	508
	509	dmu_buf_will_dirty(bpo->bpo_cached_dbuf, tx);
	510	bparray = bpo->bpo_cached_dbuf->db_data;
	511	bparray[blkoff] = stored_bp;
	512
	513	dmu_buf_will_dirty(bpo->bpo_dbuf, tx);
	514	bpo->bpo_phys->bpo_num_blkptrs++;
	515	bpo->bpo_phys->bpo_bytes +=
	516	bp_get_dsize_sync(dmu_objset_spa(bpo->bpo_os), bp);
	517	if (bpo->bpo_havecomp) {
	518	bpo->bpo_phys->bpo_comp += BP_GET_PSIZE(bp);
	519	bpo->bpo_phys->bpo_uncomp += BP_GET_UCSIZE(bp);
	520	}
	521	mutex_exit(&bpo->bpo_lock);
	522	}
	523
	524	struct space_range_arg {
	525	spa_t *spa;
	526	uint64_t mintxg;
	527	uint64_t maxtxg;
	528	uint64_t used;
	529	uint64_t comp;
	530	uint64_t uncomp;
	531	};
	532
	533	/* ARGSUSED */
	534	static int
	535	space_range_cb(void arg, const blkptr_t bp, dmu_tx_t *tx)
	536	{
	537	struct space_range_arg *sra = arg;
	538
	539	if (bp->blk_birth > sra->mintxg && bp->blk_birth <= sra->maxtxg) {
330d06f9 MA	540	if (dsl_pool_sync_context(spa_get_dsl(sra->spa)))
	541	sra->used += bp_get_dsize_sync(sra->spa, bp);
	542	else
	543	sra->used += bp_get_dsize(sra->spa, bp);
428870ff BB	544	sra->comp += BP_GET_PSIZE(bp);
	545	sra->uncomp += BP_GET_UCSIZE(bp);
	546	}
	547	return (0);
	548	}
	549
	550	int
	551	bpobj_space(bpobj_t bpo, uint64_t usedp, uint64_t compp, uint64_t uncompp)
	552	{
	553	mutex_enter(&bpo->bpo_lock);
	554
	555	*usedp = bpo->bpo_phys->bpo_bytes;
	556	if (bpo->bpo_havecomp) {
	557	*compp = bpo->bpo_phys->bpo_comp;
	558	*uncompp = bpo->bpo_phys->bpo_uncomp;
	559	mutex_exit(&bpo->bpo_lock);
	560	return (0);
	561	} else {
	562	mutex_exit(&bpo->bpo_lock);
	563	return (bpobj_space_range(bpo, 0, UINT64_MAX,
	564	usedp, compp, uncompp));
	565	}
	566	}
	567
	568	/*
	569	* Return the amount of space in the bpobj which is:
	570	* mintxg < blk_birth <= maxtxg
	571	*/
	572	int
	573	bpobj_space_range(bpobj_t *bpo, uint64_t mintxg, uint64_t maxtxg,
	574	uint64_t usedp, uint64_t compp, uint64_t *uncompp)
	575	{
	576	struct space_range_arg sra = { 0 };
	577	int err;
	578
	579	/*
	580	* As an optimization, if they want the whole txg range, just
	581	* get bpo_bytes rather than iterating over the bps.
	582	*/
	583	if (mintxg < TXG_INITIAL && maxtxg == UINT64_MAX && bpo->bpo_havecomp)
	584	return (bpobj_space(bpo, usedp, compp, uncompp));
	585
	586	sra.spa = dmu_objset_spa(bpo->bpo_os);
	587	sra.mintxg = mintxg;
	588	sra.maxtxg = maxtxg;
	589
	590	err = bpobj_iterate_nofree(bpo, space_range_cb, &sra, NULL);
	591	*usedp = sra.used;
	592	*compp = sra.comp;
	593	*uncompp = sra.uncomp;
	594	return (err);
	595	}