diff --git a/cmd/zfs/zfs_main.c b/cmd/zfs/zfs_main.c
index c361e18b9faa..50ac59fba747 100644
--- a/cmd/zfs/zfs_main.c
+++ b/cmd/zfs/zfs_main.c
@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  * Copyright (c) 2013 Steven Hartland.  All rights reserved.
  * Copyright 2013 Nexenta Systems, Inc. All rights reserved.
@@ -6636,6 +6636,9 @@ zfs_do_bookmark(int argc, char **argv)
 		case ENOTSUP:
 			err_msg = "bookmark feature not enabled";
 			break;
+		case ENOSPC:
+			err_msg = "out of space";
+			break;
 		default:
 			err_msg = "unknown error";
 			break;
@@ -6644,7 +6647,7 @@ zfs_do_bookmark(int argc, char **argv)
 		    dgettext(TEXT_DOMAIN, err_msg));
 	}
 
-	return (ret);
+	return (ret != 0);
 
 usage:
 	usage(B_FALSE);
diff --git a/cmd/zhack/zhack.c b/cmd/zhack/zhack.c
index 19493bd09410..3f86714fe31d 100644
--- a/cmd/zhack/zhack.c
+++ b/cmd/zhack/zhack.c
@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
@@ -362,7 +362,7 @@ zhack_do_feature_enable(int argc, char **argv)
 		    feature.fi_guid);
 
 	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
-	    zhack_feature_enable_sync, &feature, 5));
+	    zhack_feature_enable_sync, &feature, 5, ZFS_SPACE_CHECK_NORMAL));
 
 	spa_close(spa, FTAG);
 
@@ -473,7 +473,8 @@ zhack_do_feature_ref(int argc, char **argv)
 	}
 
 	VERIFY0(dsl_sync_task(spa_name(spa), NULL,
-	    decr ? feature_decr_sync : feature_incr_sync, &feature, 5));
+	    decr ? feature_decr_sync : feature_incr_sync, &feature,
+	    5, ZFS_SPACE_CHECK_NORMAL));
 
 	spa_close(spa, FTAG);
 }
diff --git a/include/sys/dsl_synctask.h b/include/sys/dsl_synctask.h
index ef86fb64cf0c..6139303c1564 100644
--- a/include/sys/dsl_synctask.h
+++ b/include/sys/dsl_synctask.h
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #ifndef	_SYS_DSL_SYNCTASK_H
@@ -38,11 +38,41 @@ struct dsl_pool;
 typedef int (dsl_checkfunc_t)(void *, dmu_tx_t *);
 typedef void (dsl_syncfunc_t)(void *, dmu_tx_t *);
 
+typedef enum zfs_space_check {
+	/*
+	 * Normal space check: if there is less than 3.2% free space,
+	 * the operation will fail.  Operations which are logically
+	 * creating things should use this (e.g. "zfs create", "zfs snapshot").
+	 * User writes (via the ZPL / ZVOL) also fail at this point.
+	 */
+	ZFS_SPACE_CHECK_NORMAL,
+
+	/*
+	 * Space check allows use of half the slop space.  If there
+	 * is less than 1.6% free space, the operation will fail.  Most
+	 * operations should use this (e.g. "zfs set", "zfs rename"),
+	 * because we want them to succeed even after user writes are failing,
+	 * so that they can be used as part of the space recovery process.
+	 */
+	ZFS_SPACE_CHECK_RESERVED,
+
+	/*
+	 * No space check is performed.  Only operations which we expect to
+	 * result in a net reduction in space should use this
+	 * (e.g. "zfs destroy". Setting quotas & reservations also uses
+	 * this because it needs to circumvent the quota/reservation checks).
+	 *
+	 * See also the comments above spa_slop_shift.
+	 */
+	ZFS_SPACE_CHECK_NONE,
+} zfs_space_check_t;
+
 typedef struct dsl_sync_task {
 	txg_node_t dst_node;
 	struct dsl_pool *dst_pool;
 	uint64_t dst_txg;
 	int dst_space;
+	zfs_space_check_t dst_space_check;
 	dsl_checkfunc_t *dst_checkfunc;
 	dsl_syncfunc_t *dst_syncfunc;
 	void *dst_arg;
@@ -50,11 +80,11 @@ typedef struct dsl_sync_task {
 	boolean_t dst_nowaiter;
 } dsl_sync_task_t;
 
-void dsl_sync_task_sync(dsl_sync_task_t *dst, dmu_tx_t *tx);
-int dsl_sync_task(const char *pool, dsl_checkfunc_t *checkfunc,
-    dsl_syncfunc_t *syncfunc, void *arg, int blocks_modified);
-void dsl_sync_task_nowait(struct dsl_pool *dp, dsl_syncfunc_t *syncfunc,
-    void *arg, int blocks_modified, dmu_tx_t *tx);
+void dsl_sync_task_sync(dsl_sync_task_t *, dmu_tx_t *);
+int dsl_sync_task(const char *, dsl_checkfunc_t *,
+    dsl_syncfunc_t *, void *, int, zfs_space_check_t);
+void dsl_sync_task_nowait(struct dsl_pool *, dsl_syncfunc_t *,
+    void *, int, zfs_space_check_t, dmu_tx_t *);
 
 #ifdef	__cplusplus
 }
diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c
index 6399a0a75254..6d42bf8a28b7 100644
--- a/module/zfs/dmu_objset.c
+++ b/module/zfs/dmu_objset.c
@@ -865,7 +865,8 @@ dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags,
 	doca.doca_type = type;
 
 	return (dsl_sync_task(name,
-	    dmu_objset_create_check, dmu_objset_create_sync, &doca, 5));
+	    dmu_objset_create_check, dmu_objset_create_sync, &doca,
+	    5, ZFS_SPACE_CHECK_NORMAL));
 }
 
 typedef struct dmu_objset_clone_arg {
@@ -964,7 +965,8 @@ dmu_objset_clone(const char *clone, const char *origin)
 	doca.doca_cred = CRED();
 
 	return (dsl_sync_task(clone,
-	    dmu_objset_clone_check, dmu_objset_clone_sync, &doca, 5));
+	    dmu_objset_clone_check, dmu_objset_clone_sync, &doca,
+	    5, ZFS_SPACE_CHECK_NORMAL));
 }
 
 int
diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c
index 108fa098b985..c1cac2e6762b 100644
--- a/module/zfs/dmu_send.c
+++ b/module/zfs/dmu_send.c
@@ -1191,7 +1191,7 @@ dmu_recv_begin(char *tofs, char *tosnap, struct drr_begin *drrb,
 	drba.drba_cred = CRED();
 
 	return (dsl_sync_task(tofs, dmu_recv_begin_check, dmu_recv_begin_sync,
-	    &drba, 5));
+	    &drba, 5, ZFS_SPACE_CHECK_NORMAL));
 }
 
 struct restorearg {
@@ -2108,7 +2108,7 @@ dmu_recv_existing_end(dmu_recv_cookie_t *drc)
 
 	error = dsl_sync_task(drc->drc_tofs,
 	    dmu_recv_end_check, dmu_recv_end_sync, drc,
-	    dmu_recv_end_modified_blocks);
+	    dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL);
 
 	if (error != 0)
 		dmu_recv_cleanup_ds(drc);
@@ -2122,7 +2122,7 @@ dmu_recv_new_end(dmu_recv_cookie_t *drc)
 
 	error = dsl_sync_task(drc->drc_tofs,
 	    dmu_recv_end_check, dmu_recv_end_sync, drc,
-	    dmu_recv_end_modified_blocks);
+	    dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL);
 
 	if (error != 0) {
 		dmu_recv_cleanup_ds(drc);
diff --git a/module/zfs/dsl_bookmark.c b/module/zfs/dsl_bookmark.c
index a4cb361d4e7a..447a3a2dc3a2 100644
--- a/module/zfs/dsl_bookmark.c
+++ b/module/zfs/dsl_bookmark.c
@@ -13,7 +13,7 @@
  * CDDL HEADER END
  */
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/zfs_context.h>
@@ -249,7 +249,8 @@ dsl_bookmark_create(nvlist_t *bmarks, nvlist_t *errors)
 	dbca.dbca_errors = errors;
 
 	return (dsl_sync_task(nvpair_name(pair), dsl_bookmark_create_check,
-	    dsl_bookmark_create_sync, &dbca, fnvlist_num_pairs(bmarks)));
+	    dsl_bookmark_create_sync, &dbca,
+	    fnvlist_num_pairs(bmarks), ZFS_SPACE_CHECK_NORMAL));
 }
 
 int
@@ -454,7 +455,8 @@ dsl_bookmark_destroy(nvlist_t *bmarks, nvlist_t *errors)
 	dbda.dbda_success = fnvlist_alloc();
 
 	rv = dsl_sync_task(nvpair_name(pair), dsl_bookmark_destroy_check,
-	    dsl_bookmark_destroy_sync, &dbda, fnvlist_num_pairs(bmarks));
+	    dsl_bookmark_destroy_sync, &dbda, fnvlist_num_pairs(bmarks),
+	    ZFS_SPACE_CHECK_RESERVED);
 	fnvlist_free(dbda.dbda_success);
 	return (rv);
 }
diff --git a/module/zfs/dsl_dataset.c b/module/zfs/dsl_dataset.c
index de2703868e44..b444fca64a68 100644
--- a/module/zfs/dsl_dataset.c
+++ b/module/zfs/dsl_dataset.c
@@ -1406,7 +1406,7 @@ dsl_dataset_snapshot(nvlist_t *snaps, nvlist_t *props, nvlist_t *errors)
 	if (error == 0) {
 		error = dsl_sync_task(firstname, dsl_dataset_snapshot_check,
 		    dsl_dataset_snapshot_sync, &ddsa,
-		    fnvlist_num_pairs(snaps) * 3);
+		    fnvlist_num_pairs(snaps) * 3, ZFS_SPACE_CHECK_NORMAL);
 	}
 
 	if (suspended != NULL) {
@@ -1518,7 +1518,7 @@ dsl_dataset_snapshot_tmp(const char *fsname, const char *snapname,
 	}
 
 	error = dsl_sync_task(fsname, dsl_dataset_snapshot_tmp_check,
-	    dsl_dataset_snapshot_tmp_sync, &ddsta, 3);
+	    dsl_dataset_snapshot_tmp_sync, &ddsta, 3, ZFS_SPACE_CHECK_RESERVED);
 
 	if (needsuspend)
 		zil_resume(cookie);
@@ -1874,7 +1874,8 @@ dsl_dataset_rename_snapshot(const char *fsname,
 	ddrsa.ddrsa_recursive = recursive;
 
 	error = dsl_sync_task(fsname, dsl_dataset_rename_snapshot_check,
-	    dsl_dataset_rename_snapshot_sync, &ddrsa, 1);
+	    dsl_dataset_rename_snapshot_sync, &ddrsa,
+	    1, ZFS_SPACE_CHECK_RESERVED);
 
 	if (error)
 	    return (SET_ERROR(error));
@@ -2064,7 +2065,8 @@ dsl_dataset_rollback(const char *fsname, void *owner, nvlist_t *result)
 	ddra.ddra_result = result;
 
 	return (dsl_sync_task(fsname, dsl_dataset_rollback_check,
-	    dsl_dataset_rollback_sync, &ddra, 1));
+	    dsl_dataset_rollback_sync, &ddra,
+	    1, ZFS_SPACE_CHECK_RESERVED));
 }
 
 struct promotenode {
@@ -2595,7 +2597,8 @@ dsl_dataset_promote(const char *name, char *conflsnap)
 	ddpa.cr = CRED();
 
 	return (dsl_sync_task(name, dsl_dataset_promote_check,
-	    dsl_dataset_promote_sync, &ddpa, 2 + numsnaps));
+	    dsl_dataset_promote_sync, &ddpa,
+	    2 + numsnaps, ZFS_SPACE_CHECK_RESERVED));
 }
 
 int
@@ -2949,7 +2952,7 @@ dsl_dataset_set_refquota(const char *dsname, zprop_source_t source,
 	ddsqra.ddsqra_value = refquota;
 
 	return (dsl_sync_task(dsname, dsl_dataset_set_refquota_check,
-	    dsl_dataset_set_refquota_sync, &ddsqra, 0));
+	    dsl_dataset_set_refquota_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 static int
@@ -3064,7 +3067,8 @@ dsl_dataset_set_refreservation(const char *dsname, zprop_source_t source,
 	ddsqra.ddsqra_value = refreservation;
 
 	return (dsl_sync_task(dsname, dsl_dataset_set_refreservation_check,
-	    dsl_dataset_set_refreservation_sync, &ddsqra, 0));
+	    dsl_dataset_set_refreservation_sync, &ddsqra,
+	    0, ZFS_SPACE_CHECK_NONE));
 }
 
 /*
diff --git a/module/zfs/dsl_deleg.c b/module/zfs/dsl_deleg.c
index 23bbe5b94692..952422be2381 100644
--- a/module/zfs/dsl_deleg.c
+++ b/module/zfs/dsl_deleg.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  */
 
 /*
@@ -282,7 +282,7 @@ dsl_deleg_set(const char *ddname, nvlist_t *nvp, boolean_t unset)
 
 	return (dsl_sync_task(ddname, dsl_deleg_check,
 	    unset ? dsl_deleg_unset_sync : dsl_deleg_set_sync,
-	    &dda, fnvlist_num_pairs(nvp)));
+	    &dda, fnvlist_num_pairs(nvp), ZFS_SPACE_CHECK_RESERVED));
 }
 
 /*
diff --git a/module/zfs/dsl_destroy.c b/module/zfs/dsl_destroy.c
index 6a52ed22f003..4623f5dd5b9d 100644
--- a/module/zfs/dsl_destroy.c
+++ b/module/zfs/dsl_destroy.c
@@ -520,7 +520,7 @@ dsl_destroy_snapshots_nvl(nvlist_t *snaps, boolean_t defer,
 
 	error = dsl_sync_task(nvpair_name(pair),
 	    dsl_destroy_snapshot_check, dsl_destroy_snapshot_sync,
-	    &dsda, 0);
+	    &dsda, 0, ZFS_SPACE_CHECK_NONE);
 	fnvlist_free(dsda.dsda_successful_snaps);
 
 	return (error);
@@ -918,7 +918,8 @@ dsl_destroy_head(const char *name)
 		objset_t *os;
 
 		error = dsl_sync_task(name, dsl_destroy_head_check,
-		    dsl_destroy_head_begin_sync, &ddha, 0);
+		    dsl_destroy_head_begin_sync, &ddha,
+		    0, ZFS_SPACE_CHECK_NONE);
 		if (error != 0)
 			return (error);
 
@@ -944,7 +945,7 @@ dsl_destroy_head(const char *name)
 	}
 
 	return (dsl_sync_task(name, dsl_destroy_head_check,
-	    dsl_destroy_head_sync, &ddha, 0));
+	    dsl_destroy_head_sync, &ddha, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 /*
diff --git a/module/zfs/dsl_dir.c b/module/zfs/dsl_dir.c
index 186b33203396..ba6c24486463 100644
--- a/module/zfs/dsl_dir.c
+++ b/module/zfs/dsl_dir.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 Martin Matuska. All rights reserved.
  * Copyright (c) 2014 Joyent, Inc. All rights reserved.
  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
@@ -665,7 +665,8 @@ dsl_dir_activate_fs_ss_limit(const char *ddname)
 	int error;
 
 	error = dsl_sync_task(ddname, dsl_dir_actv_fs_ss_limit_check,
-	    dsl_dir_actv_fs_ss_limit_sync, (void *)ddname, 0);
+	    dsl_dir_actv_fs_ss_limit_sync, (void *)ddname, 0,
+	    ZFS_SPACE_CHECK_RESERVED);
 
 	if (error == EALREADY)
 		error = 0;
@@ -1529,7 +1530,7 @@ dsl_dir_set_quota(const char *ddname, zprop_source_t source, uint64_t quota)
 	ddsqra.ddsqra_value = quota;
 
 	return (dsl_sync_task(ddname, dsl_dir_set_quota_check,
-	    dsl_dir_set_quota_sync, &ddsqra, 0));
+	    dsl_dir_set_quota_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 int
@@ -1650,7 +1651,7 @@ dsl_dir_set_reservation(const char *ddname, zprop_source_t source,
 	ddsqra.ddsqra_value = reservation;
 
 	return (dsl_sync_task(ddname, dsl_dir_set_reservation_check,
-	    dsl_dir_set_reservation_sync, &ddsqra, 0));
+	    dsl_dir_set_reservation_sync, &ddsqra, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 static dsl_dir_t *
@@ -1932,7 +1933,8 @@ dsl_dir_rename(const char *oldname, const char *newname)
 	ddra.ddra_cred = CRED();
 
 	return (dsl_sync_task(oldname,
-	    dsl_dir_rename_check, dsl_dir_rename_sync, &ddra, 3));
+	    dsl_dir_rename_check, dsl_dir_rename_sync, &ddra,
+	    3, ZFS_SPACE_CHECK_RESERVED));
 }
 
 int
diff --git a/module/zfs/dsl_prop.c b/module/zfs/dsl_prop.c
index 69bcd190f2e0..28b101eee547 100644
--- a/module/zfs/dsl_prop.c
+++ b/module/zfs/dsl_prop.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 Martin Matuska. All rights reserved.
  */
 
@@ -865,7 +865,7 @@ dsl_props_set(const char *dsname, zprop_source_t source, nvlist_t *props)
 		nblks = 2 * fnvlist_num_pairs(props);
 
 	return (dsl_sync_task(dsname, dsl_props_set_check, dsl_props_set_sync,
-	    &dpsa, nblks));
+	    &dpsa, nblks, ZFS_SPACE_CHECK_RESERVED));
 }
 
 typedef enum dsl_prop_getflags {
diff --git a/module/zfs/dsl_scan.c b/module/zfs/dsl_scan.c
index 5f70b18719c8..31e693149f68 100644
--- a/module/zfs/dsl_scan.c
+++ b/module/zfs/dsl_scan.c
@@ -388,7 +388,7 @@ int
 dsl_scan_cancel(dsl_pool_t *dp)
 {
 	return (dsl_sync_task(spa_name(dp->dp_spa), dsl_scan_cancel_check,
-	    dsl_scan_cancel_sync, NULL, 3));
+	    dsl_scan_cancel_sync, NULL, 3, ZFS_SPACE_CHECK_RESERVED));
 }
 
 static void dsl_scan_visitbp(blkptr_t *bp, const zbookmark_phys_t *zb,
@@ -1848,7 +1848,7 @@ dsl_scan(dsl_pool_t *dp, pool_scan_func_t func)
 	(void) spa_vdev_state_exit(spa, NULL, 0);
 
 	return (dsl_sync_task(spa_name(spa), dsl_scan_setup_check,
-	    dsl_scan_setup_sync, &func, 0));
+	    dsl_scan_setup_sync, &func, 0, ZFS_SPACE_CHECK_NONE));
 }
 
 #if defined(_KERNEL) && defined(HAVE_SPL)
diff --git a/module/zfs/dsl_synctask.c b/module/zfs/dsl_synctask.c
index 8b11cd957154..28130d25711a 100644
--- a/module/zfs/dsl_synctask.c
+++ b/module/zfs/dsl_synctask.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/dmu.h>
@@ -64,7 +64,8 @@ dsl_null_checkfunc(void *arg, dmu_tx_t *tx)
  */
 int
 dsl_sync_task(const char *pool, dsl_checkfunc_t *checkfunc,
-    dsl_syncfunc_t *syncfunc, void *arg, int blocks_modified)
+    dsl_syncfunc_t *syncfunc, void *arg,
+    int blocks_modified, zfs_space_check_t space_check)
 {
 	spa_t *spa;
 	dmu_tx_t *tx;
@@ -84,6 +85,7 @@ dsl_sync_task(const char *pool, dsl_checkfunc_t *checkfunc,
 	dst.dst_pool = dp;
 	dst.dst_txg = dmu_tx_get_txg(tx);
 	dst.dst_space = blocks_modified << DST_AVG_BLKSHIFT;
+	dst.dst_space_check = space_check;
 	dst.dst_checkfunc = checkfunc != NULL ? checkfunc : dsl_null_checkfunc;
 	dst.dst_syncfunc = syncfunc;
 	dst.dst_arg = arg;
@@ -117,13 +119,14 @@ dsl_sync_task(const char *pool, dsl_checkfunc_t *checkfunc,
 
 void
 dsl_sync_task_nowait(dsl_pool_t *dp, dsl_syncfunc_t *syncfunc, void *arg,
-    int blocks_modified, dmu_tx_t *tx)
+    int blocks_modified, zfs_space_check_t space_check, dmu_tx_t *tx)
 {
 	dsl_sync_task_t *dst = kmem_zalloc(sizeof (*dst), KM_SLEEP);
 
 	dst->dst_pool = dp;
 	dst->dst_txg = dmu_tx_get_txg(tx);
 	dst->dst_space = blocks_modified << DST_AVG_BLKSHIFT;
+	dst->dst_space_check = space_check;
 	dst->dst_checkfunc = dsl_null_checkfunc;
 	dst->dst_syncfunc = syncfunc;
 	dst->dst_arg = arg;
@@ -140,25 +143,34 @@ void
 dsl_sync_task_sync(dsl_sync_task_t *dst, dmu_tx_t *tx)
 {
 	dsl_pool_t *dp = dst->dst_pool;
-	uint64_t quota, used;
 
 	ASSERT0(dst->dst_error);
 
 	/*
-	 * 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).
+	 * Check for sufficient space.
+	 *
+	 * When the sync task was created, the caller specified the
+	 * type of space checking required.  See the comment in
+	 * zfs_space_check_t for details on the semantics of each
+	 * type of space checking.
+	 *
+	 * 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 = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes;
-	/* MOS space is triple-dittoed, so we multiply by 3. */
-	if (dst->dst_space > 0 && used + dst->dst_space * 3 > quota) {
-		dst->dst_error = SET_ERROR(ENOSPC);
-		if (dst->dst_nowaiter)
-			kmem_free(dst, sizeof (*dst));
-		return;
+	if (dst->dst_space_check != ZFS_SPACE_CHECK_NONE) {
+		uint64_t quota = dsl_pool_adjustedsize(dp,
+		    dst->dst_space_check == ZFS_SPACE_CHECK_RESERVED) -
+		    metaslab_class_get_deferred(spa_normal_class(dp->dp_spa));
+		uint64_t used = dsl_dir_phys(dp->dp_root_dir)->dd_used_bytes;
+		/* MOS space is triple-dittoed, so we multiply by 3. */
+		if (dst->dst_space > 0 && used + dst->dst_space * 3 > quota) {
+			dst->dst_error = SET_ERROR(ENOSPC);
+			if (dst->dst_nowaiter)
+				kmem_free(dst, sizeof (*dst));
+			return;
+		}
 	}
 
 	/*
diff --git a/module/zfs/dsl_userhold.c b/module/zfs/dsl_userhold.c
index 4007d4c5ea35..1b234ed480f9 100644
--- a/module/zfs/dsl_userhold.c
+++ b/module/zfs/dsl_userhold.c
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  */
 
@@ -319,7 +319,8 @@ dsl_dataset_user_hold(nvlist_t *holds, minor_t cleanup_minor, nvlist_t *errlist)
 	dduha.dduha_minor = cleanup_minor;
 
 	ret = dsl_sync_task(nvpair_name(pair), dsl_dataset_user_hold_check,
-	    dsl_dataset_user_hold_sync, &dduha, fnvlist_num_pairs(holds));
+	    dsl_dataset_user_hold_sync, &dduha,
+	    fnvlist_num_pairs(holds), ZFS_SPACE_CHECK_RESERVED);
 	fnvlist_free(dduha.dduha_chkholds);
 
 	return (ret);
@@ -609,7 +610,7 @@ dsl_dataset_user_release_impl(nvlist_t *holds, nvlist_t *errlist,
 	    KM_SLEEP));
 
 	error = dsl_sync_task(pool, dsl_dataset_user_release_check,
-	    dsl_dataset_user_release_sync, &ddura, 0);
+	    dsl_dataset_user_release_sync, &ddura, 0, ZFS_SPACE_CHECK_NONE);
 	fnvlist_free(ddura.ddura_todelete);
 	fnvlist_free(ddura.ddura_chkholds);
 
diff --git a/module/zfs/spa.c b/module/zfs/spa.c
index 376f5ae13db3..53aa05d01fb1 100644
--- a/module/zfs/spa.c
+++ b/module/zfs/spa.c
@@ -666,7 +666,8 @@ spa_prop_set(spa_t *spa, nvlist_t *nvp)
 			 * feature descriptions object.
 			 */
 			error = dsl_sync_task(spa->spa_name, NULL,
-			    spa_sync_version, &ver, 6);
+			    spa_sync_version, &ver,
+			    6, ZFS_SPACE_CHECK_RESERVED);
 			if (error)
 				return (error);
 			continue;
@@ -678,7 +679,7 @@ spa_prop_set(spa_t *spa, nvlist_t *nvp)
 
 	if (need_sync) {
 		return (dsl_sync_task(spa->spa_name, NULL, spa_sync_props,
-		    nvp, 6));
+		    nvp, 6, ZFS_SPACE_CHECK_RESERVED));
 	}
 
 	return (0);
@@ -759,7 +760,7 @@ spa_change_guid(spa_t *spa)
 	guid = spa_generate_guid(NULL);
 
 	error = dsl_sync_task(spa->spa_name, spa_change_guid_check,
-	    spa_change_guid_sync, &guid, 5);
+	    spa_change_guid_sync, &guid, 5, ZFS_SPACE_CHECK_RESERVED);
 
 	if (error == 0) {
 		spa_config_sync(spa, B_FALSE, B_TRUE);
diff --git a/module/zfs/spa_history.c b/module/zfs/spa_history.c
index 8e0033d949da..1041c8f572f2 100644
--- a/module/zfs/spa_history.c
+++ b/module/zfs/spa_history.c
@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2006, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  */
 
 #include <sys/spa.h>
@@ -323,7 +323,7 @@ spa_history_log_nvl(spa_t *spa, nvlist_t *nvl)
 
 	/* Kick this off asynchronously; errors are ignored. */
 	dsl_sync_task_nowait(spa_get_dsl(spa), spa_history_log_sync,
-	    nvarg, 0, tx);
+	    nvarg, 0, ZFS_SPACE_CHECK_NONE, tx);
 	dmu_tx_commit(tx);
 
 	/* spa_history_log_sync will free nvl */
@@ -458,7 +458,7 @@ log_internal(nvlist_t *nvl, const char *operation, spa_t *spa,
 		spa_history_log_sync(nvl, tx);
 	} else {
 		dsl_sync_task_nowait(spa_get_dsl(spa),
-		    spa_history_log_sync, nvl, 0, tx);
+		    spa_history_log_sync, nvl, 0, ZFS_SPACE_CHECK_NONE, tx);
 	}
 	/* spa_history_log_sync() will free nvl */
 }
diff --git a/module/zfs/spa_misc.c b/module/zfs/spa_misc.c
index c17043657497..8e5fa683ef42 100644
--- a/module/zfs/spa_misc.c
+++ b/module/zfs/spa_misc.c
@@ -311,6 +311,32 @@ int zfs_deadman_enabled = 1;
  */
 int spa_asize_inflation = 24;
 
+/*
+ * Normally, we don't allow the last 3.2% (1/(2^spa_slop_shift)) of space in
+ * the pool to be consumed.  This ensures that we don't run the pool
+ * completely out of space, due to unaccounted changes (e.g. to the MOS).
+ * It also limits the worst-case time to allocate space.  If we have
+ * less than this amount of free space, most ZPL operations (e.g. write,
+ * create) will return ENOSPC.
+ *
+ * Certain operations (e.g. file removal, most administrative actions) can
+ * use half the slop space.  They will only return ENOSPC if less than half
+ * the slop space is free.  Typically, once the pool has less than the slop
+ * space free, the user will use these operations to free up space in the pool.
+ * These are the operations that call dsl_pool_adjustedsize() with the netfree
+ * argument set to TRUE.
+ *
+ * A very restricted set of operations are always permitted, regardless of
+ * the amount of free space.  These are the operations that call
+ * dsl_sync_task(ZFS_SPACE_CHECK_NONE), e.g. "zfs destroy".  If these
+ * operations result in a net increase in the amount of space used,
+ * it is possible to run the pool completely out of space, causing it to
+ * be permanently read-only.
+ *
+ * See also the comments in zfs_space_check_t.
+ */
+int spa_slop_shift = 5;
+
 /*
  * ==========================================================================
  * SPA config locking
@@ -1564,6 +1590,18 @@ spa_get_asize(spa_t *spa, uint64_t lsize)
 	return (lsize * spa_asize_inflation);
 }
 
+/*
+ * Return the amount of slop space in bytes.  It is 1/32 of the pool (3.2%),
+ * or at least 32MB.
+ *
+ * See the comment above spa_slop_shift for details.
+ */
+uint64_t
+spa_get_slop_space(spa_t *spa) {
+	uint64_t space = spa_get_dspace(spa);
+	return (MAX(space >> spa_slop_shift, SPA_MINDEVSIZE >> 1));
+}
+
 uint64_t
 spa_get_dspace(spa_t *spa)
 {
diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c
index 121f99919c61..caaa3a21c71c 100644
--- a/module/zfs/zfs_ioctl.c
+++ b/module/zfs/zfs_ioctl.c
@@ -26,7 +26,7 @@
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  * Copyright (c) 2014, Joyent, Inc. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  * Copyright (c) 2014, Nexenta Systems, Inc. All rights reserved.