changelist should be able to iter on mounts

Modified changelist_gather()ing for the mountpoint property.
Now instead of iterating on all dataset descendants, we read
/proc/self/mounts and iterate on the mounted descendant datasets only.

Switched changelist implementation from a uu_list_* to uu_avl_* in
order to  reduce changlist code-path's worst case time complexity.

Reviewed by: Don Brady <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Alek Pinchuk <[email protected]>
Closes #7967 

6 files changed, 366 insertions, 68 deletions

diff --git a/include/libzfs.h b/include/libzfs.h
index 08142786b..a8e3c9c40 100644
--- a/include/libzfs.h
+++ b/include/libzfs.h
@@ -26,8 +26,8 @@
  * Copyright (c) 2013 Steven Hartland. All rights reserved.
  * Copyright (c) 2016, Intel Corporation.
  * Copyright 2016 Nexenta Systems, Inc.
- * Copyright (c) 2017 Datto Inc.
  * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
+ * Copyright (c) 2018 Datto Inc.
  */
 
 #ifndef	_LIBZFS_H
@@ -621,6 +621,7 @@ extern int zfs_iter_snapshots(zfs_handle_t *, boolean_t, zfs_iter_f, void *);
 extern int zfs_iter_snapshots_sorted(zfs_handle_t *, zfs_iter_f, void *);
 extern int zfs_iter_snapspec(zfs_handle_t *, const char *, zfs_iter_f, void *);
 extern int zfs_iter_bookmarks(zfs_handle_t *, zfs_iter_f, void *);
+extern int zfs_iter_mounted(zfs_handle_t *, zfs_iter_f, void *);
 
 typedef struct get_all_cb {
 	zfs_handle_t	**cb_handles;
diff --git a/lib/libzfs/libzfs_changelist.c b/lib/libzfs/libzfs_changelist.c
index 3b95d4d36..ae9f1d179 100644
--- a/lib/libzfs/libzfs_changelist.c
+++ b/lib/libzfs/libzfs_changelist.c
@@ -26,6 +26,7 @@
  * Portions Copyright 2007 Ramprakash Jelari
  * Copyright (c) 2014, 2015 by Delphix. All rights reserved.
  * Copyright 2016 Igor Kozhukhov <[email protected]>
+ * Copyright (c) 2018 Datto Inc.
  */
 
 #include <libintl.h>
@@ -68,22 +69,21 @@ typedef struct prop_changenode {
 	int			cn_mounted;
 	int			cn_zoned;
 	boolean_t		cn_needpost;	/* is postfix() needed? */
-	uu_list_node_t		cn_listnode;
+	uu_avl_node_t		cn_treenode;
 } prop_changenode_t;
 
 struct prop_changelist {
 	zfs_prop_t		cl_prop;
 	zfs_prop_t		cl_realprop;
 	zfs_prop_t		cl_shareprop;  /* used with sharenfs/sharesmb */
-	uu_list_pool_t		*cl_pool;
-	uu_list_t		*cl_list;
+	uu_avl_pool_t		*cl_pool;
+	uu_avl_t		*cl_tree;
 	boolean_t		cl_waslegacy;
 	boolean_t		cl_allchildren;
 	boolean_t		cl_alldependents;
 	int			cl_mflags;	/* Mount flags */
 	int			cl_gflags;	/* Gather request flags */
 	boolean_t		cl_haszonedchild;
-	boolean_t		cl_sorted;
 };
 
 /*
@@ -96,14 +96,17 @@ int
 changelist_prefix(prop_changelist_t *clp)
 {
 	prop_changenode_t *cn;
+	uu_avl_walk_t *walk;
 	int ret = 0;
 
 	if (clp->cl_prop != ZFS_PROP_MOUNTPOINT &&
 	    clp->cl_prop != ZFS_PROP_SHARESMB)
 		return (0);
 
-	for (cn = uu_list_first(clp->cl_list); cn != NULL;
-	    cn = uu_list_next(clp->cl_list, cn)) {
+	if ((walk = uu_avl_walk_start(clp->cl_tree, UU_WALK_ROBUST)) == NULL)
+		return (-1);
+
+	while ((cn = uu_avl_walk_next(walk)) != NULL) {
 
 		/* if a previous loop failed, set the remaining to false */
 		if (ret == -1) {
@@ -140,6 +143,8 @@ changelist_prefix(prop_changelist_t *clp)
 		}
 	}
 
+	uu_avl_walk_end(walk);
+
 	if (ret == -1)
 		(void) changelist_postfix(clp);
 
@@ -159,6 +164,7 @@ int
 changelist_postfix(prop_changelist_t *clp)
 {
 	prop_changenode_t *cn;
+	uu_avl_walk_t *walk;
 	char shareopts[ZFS_MAXPROPLEN];
 	int errors = 0;
 	libzfs_handle_t *hdl;
@@ -170,7 +176,7 @@ changelist_postfix(prop_changelist_t *clp)
 	 * location), or have explicit mountpoints set (in which case they won't
 	 * be in the changelist).
 	 */
-	if ((cn = uu_list_last(clp->cl_list)) == NULL)
+	if ((cn = uu_avl_last(clp->cl_tree)) == NULL)
 		return (0);
 
 	if (clp->cl_prop == ZFS_PROP_MOUNTPOINT)
@@ -193,8 +199,11 @@ changelist_postfix(prop_changelist_t *clp)
 	 * datasets before mounting the children.  We walk all datasets even if
 	 * there are errors.
 	 */
-	for (cn = uu_list_last(clp->cl_list); cn != NULL;
-	    cn = uu_list_prev(clp->cl_list, cn)) {
+	if ((walk = uu_avl_walk_start(clp->cl_tree,
+	    UU_WALK_REVERSE | UU_WALK_ROBUST)) == NULL)
+		return (-1);
+
+	while ((cn = uu_avl_walk_next(walk)) != NULL) {
 
 		boolean_t sharenfs;
 		boolean_t sharesmb;
@@ -261,6 +270,8 @@ changelist_postfix(prop_changelist_t *clp)
 			errors += zfs_unshare_smb(cn->cn_handle, NULL);
 	}
 
+	uu_avl_walk_end(walk);
+
 	return (errors ? -1 : 0);
 }
 
@@ -294,10 +305,13 @@ void
 changelist_rename(prop_changelist_t *clp, const char *src, const char *dst)
 {
 	prop_changenode_t *cn;
+	uu_avl_walk_t *walk;
 	char newname[ZFS_MAX_DATASET_NAME_LEN];
 
-	for (cn = uu_list_first(clp->cl_list); cn != NULL;
-	    cn = uu_list_next(clp->cl_list, cn)) {
+	if ((walk = uu_avl_walk_start(clp->cl_tree, UU_WALK_ROBUST)) == NULL)
+		return;
+
+	while ((cn = uu_avl_walk_next(walk)) != NULL) {
 		/*
 		 * Do not rename a clone that's not in the source hierarchy.
 		 */
@@ -316,6 +330,8 @@ changelist_rename(prop_changelist_t *clp, const char *src, const char *dst)
 		(void) strlcpy(cn->cn_handle->zfs_name, newname,
 		    sizeof (cn->cn_handle->zfs_name));
 	}
+
+	uu_avl_walk_end(walk);
 }
 
 /*
@@ -326,18 +342,23 @@ int
 changelist_unshare(prop_changelist_t *clp, zfs_share_proto_t *proto)
 {
 	prop_changenode_t *cn;
+	uu_avl_walk_t *walk;
 	int ret = 0;
 
 	if (clp->cl_prop != ZFS_PROP_SHARENFS &&
 	    clp->cl_prop != ZFS_PROP_SHARESMB)
 		return (0);
 
-	for (cn = uu_list_first(clp->cl_list); cn != NULL;
-	    cn = uu_list_next(clp->cl_list, cn)) {
+	if ((walk = uu_avl_walk_start(clp->cl_tree, UU_WALK_ROBUST)) == NULL)
+		return (-1);
+
+	while ((cn = uu_avl_walk_next(walk)) != NULL) {
 		if (zfs_unshare_proto(cn->cn_handle, NULL, proto) != 0)
 			ret = -1;
 	}
 
+	uu_avl_walk_end(walk);
+
 	return (ret);
 }
 
@@ -359,17 +380,22 @@ void
 changelist_remove(prop_changelist_t *clp, const char *name)
 {
 	prop_changenode_t *cn;
+	uu_avl_walk_t *walk;
 
-	for (cn = uu_list_first(clp->cl_list); cn != NULL;
-	    cn = uu_list_next(clp->cl_list, cn)) {
+	if ((walk = uu_avl_walk_start(clp->cl_tree, UU_WALK_ROBUST)) == NULL)
+		return;
 
+	while ((cn = uu_avl_walk_next(walk)) != NULL) {
 		if (strcmp(cn->cn_handle->zfs_name, name) == 0) {
-			uu_list_remove(clp->cl_list, cn);
+			uu_avl_remove(clp->cl_tree, cn);
 			zfs_close(cn->cn_handle);
 			free(cn);
+			uu_avl_walk_end(walk);
 			return;
 		}
 	}
+
+	uu_avl_walk_end(walk);
 }
 
 /*
@@ -379,23 +405,70 @@ void
 changelist_free(prop_changelist_t *clp)
 {
 	prop_changenode_t *cn;
-	void *cookie;
 
-	if (clp->cl_list) {
-		cookie = NULL;
-		while ((cn = uu_list_teardown(clp->cl_list, &cookie)) != NULL) {
+	if (clp->cl_tree) {
+		uu_avl_walk_t *walk;
+
+		if ((walk = uu_avl_walk_start(clp->cl_tree,
+		    UU_WALK_ROBUST)) == NULL)
+			return;
+
+		while ((cn = uu_avl_walk_next(walk)) != NULL) {
+			uu_avl_remove(clp->cl_tree, cn);
 			zfs_close(cn->cn_handle);
 			free(cn);
 		}
 
-		uu_list_destroy(clp->cl_list);
+		uu_avl_walk_end(walk);
+		uu_avl_destroy(clp->cl_tree);
 	}
 	if (clp->cl_pool)
-		uu_list_pool_destroy(clp->cl_pool);
+		uu_avl_pool_destroy(clp->cl_pool);
 
 	free(clp);
 }
 
+/*
+ * Add one dataset to changelist
+ */
+static int
+changelist_add_mounted(zfs_handle_t *zhp, void *data)
+{
+	prop_changelist_t *clp = data;
+	prop_changenode_t *cn;
+	uu_avl_index_t idx;
+
+	ASSERT3U(clp->cl_prop, ==, ZFS_PROP_MOUNTPOINT);
+
+	if ((cn = zfs_alloc(zfs_get_handle(zhp),
+	    sizeof (prop_changenode_t))) == NULL) {
+		zfs_close(zhp);
+		return (ENOMEM);
+	}
+
+	cn->cn_handle = zhp;
+	cn->cn_mounted = zfs_is_mounted(zhp, NULL);
+	ASSERT3U(cn->cn_mounted, ==, B_TRUE);
+	cn->cn_shared = zfs_is_shared(zhp);
+	cn->cn_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);
+	cn->cn_needpost = B_TRUE;
+
+	/* Indicate if any child is exported to a local zone. */
+	if (getzoneid() == GLOBAL_ZONEID && cn->cn_zoned)
+		clp->cl_haszonedchild = B_TRUE;
+
+	uu_avl_node_init(cn, &cn->cn_treenode, clp->cl_pool);
+
+	if (uu_avl_find(clp->cl_tree, cn, NULL, &idx) == NULL) {
+		uu_avl_insert(clp->cl_tree, cn, idx);
+	} else {
+		free(cn);
+		zfs_close(zhp);
+	}
+
+	return (0);
+}
+
 static int
 change_one(zfs_handle_t *zhp, void *data)
 {
@@ -460,26 +533,15 @@ change_one(zfs_handle_t *zhp, void *data)
 		if (getzoneid() == GLOBAL_ZONEID && cn->cn_zoned)
 			clp->cl_haszonedchild = B_TRUE;
 
-		uu_list_node_init(cn, &cn->cn_listnode, clp->cl_pool);
+		uu_avl_node_init(cn, &cn->cn_treenode, clp->cl_pool);
 
-		if (clp->cl_sorted) {
-			uu_list_index_t idx;
+		uu_avl_index_t idx;
 
-			(void) uu_list_find(clp->cl_list, cn, NULL,
-			    &idx);
-			uu_list_insert(clp->cl_list, cn, idx);
+		if (uu_avl_find(clp->cl_tree, cn, NULL, &idx) == NULL) {
+			uu_avl_insert(clp->cl_tree, cn, idx);
 		} else {
-			/*
-			 * Add this child to beginning of the list. Children
-			 * below this one in the hierarchy will get added above
-			 * this one in the list. This produces a list in
-			 * reverse dataset name order.
-			 * This is necessary when the original mountpoint
-			 * is legacy or none.
-			 */
-			ASSERT(!clp->cl_alldependents);
-			verify(uu_list_insert_before(clp->cl_list,
-			    uu_list_first(clp->cl_list), cn) == 0);
+			free(cn);
+			zfs_close(zhp);
 		}
 
 		if (!clp->cl_alldependents)
@@ -542,7 +604,6 @@ changelist_gather(zfs_handle_t *zhp, zfs_prop_t prop, int gather_flags,
 	prop_changenode_t *cn;
 	zfs_handle_t *temp;
 	char property[ZFS_MAXPROPLEN];
-	uu_compare_fn_t *compare = NULL;
 	boolean_t legacy = B_FALSE;
 
 	if ((clp = zfs_alloc(zhp->zfs_hdl, sizeof (prop_changelist_t))) == NULL)
@@ -562,19 +623,14 @@ changelist_gather(zfs_handle_t *zhp, zfs_prop_t prop, int gather_flags,
 		    NULL, NULL, 0, B_FALSE) == 0 &&
 		    (strcmp(property, "legacy") == 0 ||
 		    strcmp(property, "none") == 0)) {
-
 			legacy = B_TRUE;
 		}
-		if (!legacy) {
-			compare = compare_mountpoints;
-			clp->cl_sorted = B_TRUE;
-		}
 	}
 
-	clp->cl_pool = uu_list_pool_create("changelist_pool",
+	clp->cl_pool = uu_avl_pool_create("changelist_pool",
 	    sizeof (prop_changenode_t),
-	    offsetof(prop_changenode_t, cn_listnode),
-	    compare, 0);
+	    offsetof(prop_changenode_t, cn_treenode),
+	    compare_mountpoints, 0);
 	if (clp->cl_pool == NULL) {
 		assert(uu_error() == UU_ERROR_NO_MEMORY);
 		(void) zfs_error(zhp->zfs_hdl, EZFS_NOMEM, "internal error");
@@ -582,12 +638,11 @@ changelist_gather(zfs_handle_t *zhp, zfs_prop_t prop, int gather_flags,
 		return (NULL);
 	}
 
-	clp->cl_list = uu_list_create(clp->cl_pool, NULL,
-	    clp->cl_sorted ? UU_LIST_SORTED : 0);
+	clp->cl_tree = uu_avl_create(clp->cl_pool, NULL, UU_DEFAULT);
 	clp->cl_gflags = gather_flags;
 	clp->cl_mflags = mnt_flags;
 
-	if (clp->cl_list == NULL) {
+	if (clp->cl_tree == NULL) {
 		assert(uu_error() == UU_ERROR_NO_MEMORY);
 		(void) zfs_error(zhp->zfs_hdl, EZFS_NOMEM, "internal error");
 		changelist_free(clp);
@@ -631,7 +686,17 @@ changelist_gather(zfs_handle_t *zhp, zfs_prop_t prop, int gather_flags,
 	else if (clp->cl_prop == ZFS_PROP_SHARESMB)
 		clp->cl_shareprop = ZFS_PROP_SHARENFS;
 
-	if (clp->cl_alldependents) {
+	if (clp->cl_prop == ZFS_PROP_MOUNTPOINT &&
+	    (clp->cl_gflags & CL_GATHER_MOUNT_ALWAYS) == 0) {
+		/*
+		 * Instead of iterating through all of the dataset children we
+		 * gather mounted dataset children from MNTTAB
+		 */
+		if (zfs_iter_mounted(zhp, changelist_add_mounted, clp) != 0) {
+			changelist_free(clp);
+			return (NULL);
+		}
+	} else if (clp->cl_alldependents) {
 		if (zfs_iter_dependents(zhp, B_TRUE, change_one, clp) != 0) {
 			changelist_free(clp);
 			return (NULL);
@@ -669,20 +734,13 @@ changelist_gather(zfs_handle_t *zhp, zfs_prop_t prop, int gather_flags,
 	cn->cn_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);
 	cn->cn_needpost = B_TRUE;
 
-	uu_list_node_init(cn, &cn->cn_listnode, clp->cl_pool);
-	if (clp->cl_sorted) {
-		uu_list_index_t idx;
-		(void) uu_list_find(clp->cl_list, cn, NULL, &idx);
-		uu_list_insert(clp->cl_list, cn, idx);
+	uu_avl_node_init(cn, &cn->cn_treenode, clp->cl_pool);
+	uu_avl_index_t idx;
+	if (uu_avl_find(clp->cl_tree, cn, NULL, &idx) == NULL) {
+		uu_avl_insert(clp->cl_tree, cn, idx);
 	} else {
-		/*
-		 * Add the target dataset to the end of the list.
-		 * The list is not really unsorted. The list will be
-		 * in reverse dataset name order. This is necessary
-		 * when the original mountpoint is legacy or none.
-		 */
-		verify(uu_list_insert_after(clp->cl_list,
-		    uu_list_last(clp->cl_list), cn) == 0);
+		free(cn);
+		zfs_close(temp);
 	}
 
 	/*
diff --git a/lib/libzfs/libzfs_iter.c b/lib/libzfs/libzfs_iter.c
index 6f9b3f465..73dc2c793 100644
--- a/lib/libzfs/libzfs_iter.c
+++ b/lib/libzfs/libzfs_iter.c
@@ -23,6 +23,7 @@
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2013, 2015 by Delphix. All rights reserved.
  * Copyright 2014 Nexenta Systems, Inc.  All rights reserved.
+ * Copyright (c) 2018 Datto Inc.
  */
 
 #include <stdio.h>
@@ -32,6 +33,7 @@
 #include <stddef.h>
 #include <libintl.h>
 #include <libzfs.h>
+#include <sys/mntent.h>
 
 #include "libzfs_impl.h"
 
@@ -522,3 +524,50 @@ zfs_iter_dependents(zfs_handle_t *zhp, boolean_t allowrecursion,
 	ida.first = B_TRUE;
 	return (iter_dependents_cb(zfs_handle_dup(zhp), &ida));
 }
+
+/*
+ * Iterate over mounted children of the specified dataset
+ */
+int
+zfs_iter_mounted(zfs_handle_t *zhp, zfs_iter_f func, void *data)
+{
+	char mnt_prop[ZFS_MAXPROPLEN];
+	struct mnttab entry;
+	zfs_handle_t *mtab_zhp;
+	size_t namelen = strlen(zhp->zfs_name);
+	FILE *mnttab;
+	int err = 0;
+
+	if ((mnttab = fopen(MNTTAB, "r")) == NULL)
+		return (ENOENT);
+
+	while (err == 0 && getmntent(mnttab, &entry) == 0) {
+		/* Ignore non-ZFS entries */
+		if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
+			continue;
+
+		/* Ignore datasets not within the provided dataset */
+		if (strncmp(entry.mnt_special, zhp->zfs_name, namelen) != 0 ||
+		    (entry.mnt_special[namelen] != '/' &&
+		    entry.mnt_special[namelen] != '@'))
+			continue;
+
+		if ((mtab_zhp = zfs_open(zhp->zfs_hdl, entry.mnt_special,
+		    ZFS_TYPE_FILESYSTEM)) == NULL)
+			continue;
+
+		/* Ignore legacy mounts as they are user managed */
+		verify(zfs_prop_get(mtab_zhp, ZFS_PROP_MOUNTPOINT, mnt_prop,
+		    sizeof (mnt_prop), NULL, NULL, 0, B_FALSE) == 0);
+		if (strcmp(mnt_prop, "legacy") == 0) {
+			zfs_close(mtab_zhp);
+			continue;
+		}
+
+		err = func(mtab_zhp, data);
+	}
+
+	fclose(mnttab);
+
+	return (err);
+}
diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run
index 95e70f043..32bc9c328 100644
--- a/tests/runfiles/linux.run
+++ b/tests/runfiles/linux.run
@@ -277,7 +277,7 @@ tags = ['functional', 'cli_root', 'zfs_unload-key']
 tests = ['zfs_unmount_001_pos', 'zfs_unmount_002_pos', 'zfs_unmount_003_pos',
     'zfs_unmount_004_pos', 'zfs_unmount_005_pos', 'zfs_unmount_006_pos',
     'zfs_unmount_007_neg', 'zfs_unmount_008_neg', 'zfs_unmount_009_pos',
-    'zfs_unmount_all_001_pos']
+    'zfs_unmount_all_001_pos', 'zfs_unmount_nested']
 tags = ['functional', 'cli_root', 'zfs_unmount']
 
 [tests/functional/cli_root/zfs_unshare]
diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_unmount/Makefile.am b/tests/zfs-tests/tests/functional/cli_root/zfs_unmount/Makefile.am
index 9ea809f26..34cbb17ae 100644
--- a/tests/zfs-tests/tests/functional/cli_root/zfs_unmount/Makefile.am
+++ b/tests/zfs-tests/tests/functional/cli_root/zfs_unmount/Makefile.am
@@ -11,7 +11,8 @@ dist_pkgdata_SCRIPTS = \
 	zfs_unmount_007_neg.ksh \
 	zfs_unmount_008_neg.ksh \
 	zfs_unmount_009_pos.ksh \
-	zfs_unmount_all_001_pos.ksh
+	zfs_unmount_all_001_pos.ksh \
+	zfs_unmount_nested.ksh
 
 dist_pkgdata_DATA = \
 	zfs_unmount.cfg \
diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_unmount/zfs_unmount_nested.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_unmount/zfs_unmount_nested.ksh
new file mode 100755
index 000000000..73eae6a25
--- /dev/null
+++ b/tests/zfs-tests/tests/functional/cli_root/zfs_unmount/zfs_unmount_nested.ksh
@@ -0,0 +1,189 @@
+#!/bin/ksh -p
+#
+# CDDL HEADER START
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy is of the CDDL is also available via the Internet
+# at http://www.illumos.org/license/CDDL.
+#
+# CDDL HEADER END
+#
+
+#
+# Copyright (c) 2018 Datto Inc.
+#
+
+. $STF_SUITE/include/libtest.shlib
+
+#
+# DESCRIPTION:
+#	zfs unmount should work on nested datasets
+#
+# STRATEGY:
+#	1. Create a set of nested datasets
+#	2. Unmount a nested dataset and make sure it is unmounted
+#	3. Ensure the dataset deeper than the one above is also unmounted
+#	4. Ensure the datasets shallower than the unmounted one is still mounted
+#	5. Repeat from step 2 with other mountpoint values and shallower nesting
+#
+
+verify_runnable "both"
+
+function nesting_cleanup
+{
+	log_must zfs destroy -fR $TESTPOOL/a
+	log_must zfs destroy -fR $TESTPOOL/b
+	log_must zfs destroy -fR $TESTPOOL/c
+	log_must zfs destroy -fR $TESTPOOL/d
+}
+
+log_onexit nesting_cleanup
+
+set -A test_depths 30 16 3
+
+dsA32=$(printf 'a/%.0s' {1..32})"a"
+log_must zfs create -p $TESTPOOL/$dsA32
+
+dsB32=$(printf 'b/%.0s' {1..32})"b"
+log_must zfs create -o mountpoint=none -p $TESTPOOL/$dsB32
+log_mustnot mount -t zfs $TESTPOOL/$dsB32 /mnt
+
+dsC32=$(printf 'c/%.0s' {1..32})"c"
+log_must zfs create -o mountpoint=legacy -p $TESTPOOL/$dsC32
+log_must mount -t zfs $TESTPOOL/$dsC32 /mnt
+
+dsD32=$(printf 'd/%.0s' {1..32})"d"
+log_must zfs create -o mountpoint=/$TESTPOOL/mnt -p $TESTPOOL/$dsD32
+
+
+for d in ${test_depths[@]}; do
+	# default mountpoint
+	ds_pre=$(printf 'a/%.0s' {1..$(($d-2))})"a"
+	ds=$(printf 'a/%.0s' {1..$(($d-1))})"a"
+	ds_post=$(printf 'a/%.0s' {1..$(($d))})"a"
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$TESTPOOL/$ds_pre (pre) not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds; then
+		log_fail "$TESTPOOL/$ds not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds_post; then
+		log_fail "$TESTPOOL/$ds_post (post) not initially mounted"
+	fi
+
+	log_must zfs snapshot $TESTPOOL/$ds@snap
+	# force snapshot mount in .zfs
+	log_must ls /$TESTPOOL/$ds/.zfs/snapshot/snap
+	log_must zfs unmount $TESTPOOL/$ds
+
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$ds_pre is not mounted"
+	fi
+	if ismounted $TESTPOOL/$ds; then
+		log_fail "$ds is mounted"
+	fi
+	if ismounted $TESTPOOL/$ds_post; then
+		log_fail "$ds_post (post) is mounted"
+	fi
+
+
+	# mountpoint=none
+	ds_pre=$(printf 'b/%.0s' {1..$(($d-2))})"b"
+	ds=$(printf 'b/%.0s' {1..$(($d-1))})"b"
+	ds_post=$(printf 'b/%.0s' {1..$(($d))})"b"
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$TESTPOOL/$ds_pre (pre) not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds; then
+		log_fail "$TESTPOOL/$ds not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds_post; then
+		log_fail "$TESTPOOL/$ds_post (post) not initially mounted"
+	fi
+
+	log_must zfs snapshot $TESTPOOL/$ds@snap
+	# force snapshot mount in .zfs
+	log_must ls /$TESTPOOL/$ds/.zfs/snapshot/snap
+	log_must zfs unmount $TESTPOOL/$ds
+
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$TESTPOOL/$ds_pre (pre) not mounted"
+	fi
+	if ismounted $TESTPOOL/$ds; then
+		log_fail "$TESTPOOL/$ds is mounted"
+	fi
+	if ismounted $TESTPOOL/$ds_post; then
+		log_fail "$TESTPOOL/$ds_post (post) is mounted"
+	fi
+
+
+	# mountpoint=legacy
+	ds_pre=$(printf 'c/%.0s' {1..$(($d-2))})"c"
+	ds=$(printf 'c/%.0s' {1..$(($d-1))})"c"
+	ds_post=$(printf 'c/%.0s' {1..$(($d))})"c"
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$TESTPOOL/$ds_pre (pre) not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds; then
+		log_fail "$TESTPOOL/$ds not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds_post; then
+		log_fail "$TESTPOOL/$ds_post (post) not initially mounted"
+	fi
+
+	log_must zfs snapshot $TESTPOOL/$ds@snap
+	# force snapshot mount in .zfs
+	log_must ls /$TESTPOOL/$ds/.zfs/snapshot/snap
+	log_must zfs unmount $TESTPOOL/$ds
+
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$TESTPOOL/$ds_pre (pre) not mounted"
+	fi
+	if ismounted $TESTPOOL/$ds; then
+		log_fail "$TESTPOOL/$ds is mounted"
+	fi
+	if ismounted $TESTPOOL/$ds_post; then
+		log_fail "$TESTPOOL/$ds_post (post) is mounted"
+	fi
+
+
+	# mountpoint=testpool/mnt
+	ds_pre=$(printf 'd/%.0s' {1..$(($d-2))})"d"
+	ds=$(printf 'd/%.0s' {1..$(($d-1))})"d"
+	ds_post=$(printf 'd/%.0s' {1..$(($d))})"d"
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$TESTPOOL/$ds_pre (pre) not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds; then
+		log_fail "$TESTPOOL/$ds not initially mounted"
+	fi
+	if ! ismounted $TESTPOOL/$ds_post; then
+		log_fail "$TESTPOOL/$ds_post (post) not initially mounted"
+	fi
+
+	log_must zfs snapshot $TESTPOOL/$ds@snap
+	# force snapshot mount in .zfs
+	log_must ls /$TESTPOOL/$ds/.zfs/snapshot/snap
+	log_must zfs unmount $TESTPOOL/$ds
+
+	if ! ismounted $TESTPOOL/$ds_pre; then
+		log_fail "$ds_pre is not mounted"
+	fi
+	if ismounted $TESTPOOL/$ds; then
+		log_fail "$ds is mounted"
+	fi
+	if ismounted $TESTPOOL/$ds_post; then
+		log_fail "$ds_post (post) is mounted"
+	fi
+done
+
+log_must zpool export $TESTPOOL
+log_must rmdir /testpool/mnt # remove the mountpoint we created
+log_must zpool import $TESTPOOL
+
+log_pass "Verified nested dataset are unmounted."