11943 Fix out-of-order ZIL txtype lost on hardlinked files

11942 Panic on zil/slog replay when TX_REMOVE followed by TX_CREATE
Portions contributed by: Ryan Moeller <ryan@freqlabs.com>
Portions contributed by: Andy Fiddaman <andy@omniosce.org>
Reviewed by: Jerry Jelinek <jerry.jelinek@joyent.com>
Reviewed by: Toomas Soome <tsoome@me.com>
Approved by: Dan McDonald <danmcd@joyent.com>

32 files changed, 464 insertions, 49 deletions

diff --git a/usr/src/pkg/manifests/system-test-zfstest.mf b/usr/src/pkg/manifests/system-test-zfstest.mf
index 0f9f6932d9..4f76b99667 100644
--- a/usr/src/pkg/manifests/system-test-zfstest.mf
+++ b/usr/src/pkg/manifests/system-test-zfstest.mf
@@ -2988,6 +2988,8 @@ file path=opt/zfs-tests/tests/functional/slog/slog_012_neg mode=0555
 file path=opt/zfs-tests/tests/functional/slog/slog_013_pos mode=0555
 file path=opt/zfs-tests/tests/functional/slog/slog_014_pos mode=0555
 file path=opt/zfs-tests/tests/functional/slog/slog_015_neg mode=0555
+file path=opt/zfs-tests/tests/functional/slog/slog_replay_fs_001 mode=0555
+file path=opt/zfs-tests/tests/functional/slog/slog_replay_fs_002 mode=0555
 file path=opt/zfs-tests/tests/functional/snapshot/cleanup mode=0555
 file path=opt/zfs-tests/tests/functional/snapshot/clone_001_pos mode=0555
 file path=opt/zfs-tests/tests/functional/snapshot/deadlist_lock mode=0555
diff --git a/usr/src/test/zfs-tests/include/libtest.shlib b/usr/src/test/zfs-tests/include/libtest.shlib
index a28f015de8..363f674f03 100644
--- a/usr/src/test/zfs-tests/include/libtest.shlib
+++ b/usr/src/test/zfs-tests/include/libtest.shlib
@@ -2880,3 +2880,22 @@ function get_tunable_impl
 
 	return 1
 }
+
+#
+# Compute SHA256 digest for given file or stdin if no file given.
+# Note: file path must not contain spaces
+#
+function sha256digest
+{
+        typeset file=$1
+
+	if [ -x /usr/bin/digest ]; then
+		/usr/bin/digest -a sha256 $file
+	elif [ -x /usr/bin/sha256sum ]; then
+		/usr/bin/sha256sum -b $file | awk '{ print $1 }'
+	else
+		echo "Cannot calculate SHA256 digest"
+		return 1
+	fi
+	return 0
+}
diff --git a/usr/src/test/zfs-tests/runfiles/delphix.run b/usr/src/test/zfs-tests/runfiles/delphix.run
index 3e0b66aba8..d2e0f4df1d 100644
--- a/usr/src/test/zfs-tests/runfiles/delphix.run
+++ b/usr/src/test/zfs-tests/runfiles/delphix.run
@@ -641,7 +641,7 @@ tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos',
 tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos',
     'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg',
     'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg',
-    'slog_013_pos', 'slog_014_pos']
+    'slog_013_pos', 'slog_014_pos', 'slog_replay_fs_001', 'slog_replay_fs_002']
 
 [/opt/zfs-tests/tests/functional/snapshot]
 tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos',
diff --git a/usr/src/test/zfs-tests/runfiles/omnios.run b/usr/src/test/zfs-tests/runfiles/omnios.run
index 998f8db0ca..b198fb9674 100644
--- a/usr/src/test/zfs-tests/runfiles/omnios.run
+++ b/usr/src/test/zfs-tests/runfiles/omnios.run
@@ -640,7 +640,7 @@ tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos',
 tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos',
     'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg',
     'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg',
-    'slog_013_pos', 'slog_014_pos']
+    'slog_013_pos', 'slog_014_pos', 'slog_replay_fs_001', 'slog_replay_fs_002']
 
 [/opt/zfs-tests/tests/functional/snapshot]
 tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos',
diff --git a/usr/src/test/zfs-tests/runfiles/openindiana.run b/usr/src/test/zfs-tests/runfiles/openindiana.run
index 6392469657..8d64315e5a 100644
--- a/usr/src/test/zfs-tests/runfiles/openindiana.run
+++ b/usr/src/test/zfs-tests/runfiles/openindiana.run
@@ -640,7 +640,7 @@ tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos',
 tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos',
     'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg',
     'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg',
-    'slog_013_pos', 'slog_014_pos']
+    'slog_013_pos', 'slog_014_pos', 'slog_replay_fs_001', 'slog_replay_fs_002']
 
 [/opt/zfs-tests/tests/functional/snapshot]
 tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos',
diff --git a/usr/src/test/zfs-tests/runfiles/smartos.run b/usr/src/test/zfs-tests/runfiles/smartos.run
index 0f04168d75..ef6133bcfc 100644
--- a/usr/src/test/zfs-tests/runfiles/smartos.run
+++ b/usr/src/test/zfs-tests/runfiles/smartos.run
@@ -550,7 +550,7 @@ tests = ['scrub_mirror_001_pos', 'scrub_mirror_002_pos',
 tests = ['slog_001_pos', 'slog_002_pos', 'slog_003_pos', 'slog_004_pos',
     'slog_005_pos', 'slog_006_pos', 'slog_007_pos', 'slog_008_neg',
     'slog_009_neg', 'slog_010_neg', 'slog_011_neg', 'slog_012_neg',
-    'slog_013_pos', 'slog_014_pos']
+    'slog_013_pos', 'slog_014_pos', 'slog_replay_fs_001', 'slog_replay_fs_002']
 
 [/opt/zfs-tests/tests/functional/snapshot]
 tests = ['clone_001_pos', 'rollback_001_pos', 'rollback_002_pos',
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/setup.ksh b/usr/src/test/zfs-tests/tests/functional/slog/setup.ksh
index c5c1a82710..3521001108 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/setup.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/setup.ksh
@@ -38,13 +38,4 @@ if ! verify_slog_support ; then
 	log_unsupported "This system doesn't support separate intent logs"
 fi
 
-if [[ -d $VDEV ]]; then
-	log_must rm -rf $VDIR
-fi
-if [[ -d $VDEV2 ]]; then
-	log_must rm -rf $VDIR2
-fi
-log_must mkdir -p $VDIR $VDIR2
-log_must mkfile $MINVDEVSIZE $VDEV $SDEV $LDEV $VDEV2 $SDEV2 $LDEV2
-
 log_pass
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog.kshlib b/usr/src/test/zfs-tests/tests/functional/slog/slog.kshlib
index 493ceda60d..f96baf4967 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog.kshlib
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog.kshlib
@@ -31,10 +31,20 @@
 . $STF_SUITE/include/libtest.shlib
 . $STF_SUITE/tests/functional/slog/slog.cfg
 
+function setup
+{
+	log_must rm -rf $VDIR $VDIR2
+	log_must mkdir -p $VDIR $VDIR2
+	log_must truncate -s $MINVDEVSIZE $VDEV $SDEV $LDEV $VDEV2 $SDEV2 $LDEV2
+
+	return 0
+}
+
 function cleanup
 {
 	poolexists $TESTPOOL && destroy_pool $TESTPOOL
 	poolexists $TESTPOOL2 && destroy_pool $TESTPOOL2
+	rm -rf $TESTDIR $VDIR $VDIR2
 }
 
 #
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_001_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_001_pos.ksh
index 3d3daf5f9c..a4c35ed9e9 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_001_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_001_pos.ksh
@@ -45,6 +45,7 @@ verify_runnable "global"
 
 log_assert "Creating a pool with a log device succeeds."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_002_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_002_pos.ksh
index b056f19cdb..91904aa612 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_002_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_002_pos.ksh
@@ -46,6 +46,7 @@ verify_runnable "global"
 
 log_assert "Adding a log device to normal pool works."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_003_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_003_pos.ksh
index c647b8f54b..0b4d6ede3e 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_003_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_003_pos.ksh
@@ -46,6 +46,7 @@ verify_runnable "global"
 
 log_assert "Adding an extra log device works."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_004_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_004_pos.ksh
index 4b0b3439a2..10f28dcc00 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_004_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_004_pos.ksh
@@ -46,6 +46,7 @@ verify_runnable "global"
 
 log_assert "Attaching a log device passes."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_005_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_005_pos.ksh
index cbbb948691..4836f6f279 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_005_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_005_pos.ksh
@@ -46,6 +46,7 @@ verify_runnable "global"
 
 log_assert "Detaching a log device passes."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_006_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_006_pos.ksh
index 53e8c67ca0..24143196fd 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_006_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_006_pos.ksh
@@ -46,6 +46,7 @@ verify_runnable "global"
 
 log_assert "Replacing a log device passes."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_007_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_007_pos.ksh
index 4926fb7b31..27ac38606c 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_007_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_007_pos.ksh
@@ -48,6 +48,7 @@ verify_runnable "global"
 
 log_assert "Exporting and importing pool with log devices passes."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_008_neg.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_008_neg.ksh
index 587e0e3212..54587a0c61 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_008_neg.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_008_neg.ksh
@@ -44,6 +44,7 @@ verify_runnable "global"
 
 log_assert "A raidz/raidz2 log is not supported."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_009_neg.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_009_neg.ksh
index e7091f17b7..222f71a999 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_009_neg.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_009_neg.ksh
@@ -45,6 +45,7 @@ verify_runnable "global"
 
 log_assert "A raidz/raidz2 log can not be added to existed pool."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_010_neg.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_010_neg.ksh
index 8fe248ffbc..edd9abea09 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_010_neg.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_010_neg.ksh
@@ -46,6 +46,7 @@ verify_runnable "global"
 
 log_assert "Slog device can not be replaced with spare device."
 log_onexit cleanup
+log_must setup
 
 log_must zpool create $TESTPOOL $VDEV spare $SDEV log $LDEV
 sdev=$(random_get $SDEV)
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_011_neg.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_011_neg.ksh
index 2dad200b31..3bebc82017 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_011_neg.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_011_neg.ksh
@@ -46,6 +46,7 @@ verify_runnable "global"
 
 log_assert "Offline and online a log device passes."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_012_neg.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_012_neg.ksh
index 758563f23b..07e4e271fa 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_012_neg.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_012_neg.ksh
@@ -45,6 +45,7 @@ verify_runnable "global"
 
 log_assert "Pool can survive when one of mirror log device get corrupted."
 log_onexit cleanup
+log_must setup
 
 for type in "" "mirror" "raidz" "raidz2"
 do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_013_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_013_pos.ksh
index 3607da7928..65a1ac5d68 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_013_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_013_pos.ksh
@@ -54,6 +54,7 @@ log_assert "Verify slog device can be disk, file, lofi device or any device " \
 	"that presents a block interface."
 verify_disk_count "$DISKS" 2
 log_onexit cleanup_testenv
+log_must setup
 
 dsk1=${DISKS%% *}
 log_must zpool create $TESTPOOL ${DISKS#$dsk1}
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_014_pos.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_014_pos.ksh
index b476d497b8..efd876d554 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_014_pos.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_014_pos.ksh
@@ -45,6 +45,7 @@
 verify_runnable "global"
 
 log_assert "log device can survive when one of the pool device get corrupted."
+log_must setup
 
 for type in "mirror" "raidz" "raidz2"; do
 	for spare in "" "spare"; do
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_015_neg.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_015_neg.ksh
index 952c135a9a..0ca9e0b620 100644
--- a/usr/src/test/zfs-tests/tests/functional/slog/slog_015_neg.ksh
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_015_neg.ksh
@@ -47,6 +47,7 @@ function cleanup
 
 ORIG_TIMEOUT=$(mdb -ke "zfs_commit_timeout_pct/J" | tail -1 | awk '{print $NF}')
 log_onexit cleanup
+log_must setup
 
 for PCT in 0 1 2 4 8 16 32 64 128 256 512 1024; do
 	log_must mdb -kwe "zfs_commit_timeout_pct/Z $PCT"
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_replay_fs_001.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_replay_fs_001.ksh
new file mode 100755
index 0000000000..364113925d
--- /dev/null
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_replay_fs_001.ksh
@@ -0,0 +1,215 @@
+#!/bin/ksh -p
+#
+# 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 2007 Sun Microsystems, Inc.  All rights reserved.
+# Use is subject to license terms.
+#
+
+. $STF_SUITE/tests/functional/slog/slog.kshlib
+
+#
+# DESCRIPTION:
+#	Verify slogs are replayed correctly.  This test is a direct
+#	adaptation of the ziltest.sh script for the ZFS Test Suite.
+#
+#	The general idea is to build up an intent log from a bunch of
+#	diverse user commands without actually committing them to the
+#	file system.  Then copy the file system, replay the intent
+#	log and compare the file system and the copy.
+#
+#	To enable this automated testing of the intent log some minimal
+#	support is required of the file system.  In particular, a
+#	"freeze" command is required to flush the in-flight transactions;
+#	to stop the actual committing of transactions; and to ensure no
+#	deltas are discarded. All deltas past a freeze point are kept
+#	for replay and comparison later. Here is the flow:
+#
+# STRATEGY:
+#	1. Create an empty file system (TESTFS)
+#	2. Freeze TESTFS
+#	3. Run various user commands that create files, directories and ACLs
+#	4. Copy TESTFS to temporary location (TESTDIR/copy)
+#	5. Unmount filesystem
+#	   <at this stage TESTFS is empty again and unfrozen, and the
+#	   intent log contains a complete set of deltas to replay it>
+#	6. Remount TESTFS <which replays the intent log>
+#	7. Compare TESTFS against the TESTDIR/copy
+#
+
+verify_runnable "global"
+
+# As long as we are not running slog_015_neg, the test pool could be hanging
+# around.
+poolexists $TESTPOOL && zpool destroy -f $TESTPOOL
+
+log_assert "Replay of intent log succeeds."
+log_onexit cleanup
+log_must setup
+
+#
+# 1. Create an empty file system (TESTFS)
+#
+log_must zpool create $TESTPOOL $VDEV log mirror $LDEV
+log_must zfs set compression=on $TESTPOOL
+log_must zfs create $TESTPOOL/$TESTFS
+
+#
+# This dd command works around an issue where ZIL records aren't created
+# after freezing the pool unless a ZIL header already exists. Create a file
+# synchronously to force ZFS to write one out.
+#
+log_must dd if=/dev/zero of=/$TESTPOOL/$TESTFS/sync \
+    oflag=dsync,sync bs=1 count=1
+
+#
+# 2. Freeze TESTFS
+#
+log_must zpool freeze $TESTPOOL
+
+#
+# 3. Run various user commands that create files, directories and ACLs
+#
+
+# TX_CREATE
+log_must touch /$TESTPOOL/$TESTFS/a
+
+# TX_RENAME
+log_must mv /$TESTPOOL/$TESTFS/a /$TESTPOOL/$TESTFS/b
+
+# TX_SYMLINK
+log_must touch /$TESTPOOL/$TESTFS/c
+log_must ln -s /$TESTPOOL/$TESTFS/c /$TESTPOOL/$TESTFS/d
+
+# TX_LINK
+log_must touch /$TESTPOOL/$TESTFS/e
+log_must ln /$TESTPOOL/$TESTFS/e /$TESTPOOL/$TESTFS/f
+
+# TX_MKDIR
+log_must mkdir /$TESTPOOL/$TESTFS/dir_to_delete
+
+# TX_RMDIR
+log_must rmdir /$TESTPOOL/$TESTFS/dir_to_delete
+
+# Create a simple validation payload
+log_must mkdir -p $TESTDIR
+log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/payload bs=1k count=8
+typeset checksum=$(sha256digest /$TESTPOOL/$TESTFS/payload)
+
+# TX_WRITE (small file with ordering)
+log_must mkfile 1k /$TESTPOOL/$TESTFS/small_file
+log_must mkfile 512b /$TESTPOOL/$TESTFS/small_file
+
+# TX_CREATE, TX_MKDIR, TX_REMOVE, TX_RMDIR
+log_must cp -R /usr/dict /$TESTPOOL/$TESTFS
+log_must rm -rf /$TESTPOOL/$TESTFS/dict
+
+# TX_SETATTR
+log_must touch /$TESTPOOL/$TESTFS/setattr
+log_must chmod 567 /$TESTPOOL/$TESTFS/setattr
+log_must chgrp root /$TESTPOOL/$TESTFS/setattr
+log_must touch -cm -t 201311271200 /$TESTPOOL/$TESTFS/setattr
+
+# TX_TRUNCATE (to zero)
+log_must mkfile 4k /$TESTPOOL/$TESTFS/truncated_file
+log_must truncate -s 0 /$TESTPOOL/$TESTFS/truncated_file
+
+# TX_WRITE (large file)
+log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/large \
+    bs=128k count=64 oflag=sync
+
+# Write zeros, which compress to holes, in the middle of a file
+log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/holes.1 bs=128k count=8
+log_must dd if=/dev/zero of=/$TESTPOOL/$TESTFS/holes.1 bs=128k count=2
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/holes.2 bs=128k count=8
+log_must dd if=/dev/zero of=/$TESTPOOL/$TESTFS/holes.2 bs=128k count=2 seek=2
+
+log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/holes.3 bs=128k count=8
+log_must dd if=/dev/zero of=/$TESTPOOL/$TESTFS/holes.3 bs=128k count=2 \
+   seek=2 conv=notrunc
+
+# TX_MKXATTR
+# log_must mkdir /$TESTPOOL/$TESTFS/xattr.dir
+# log_must attr -qs fileattr -V HelloWorld /$TESTPOOL/$TESTFS/xattr.dir
+# log_must attr -qs tmpattr -V HelloWorld /$TESTPOOL/$TESTFS/xattr.dir
+# log_must attr -qr tmpattr /$TESTPOOL/$TESTFS/xattr.dir
+
+# log_must touch /$TESTPOOL/$TESTFS/xattr.file
+# log_must attr -qs fileattr -V HelloWorld /$TESTPOOL/$TESTFS/xattr.file
+# log_must attr -qs tmpattr -V HelloWorld /$TESTPOOL/$TESTFS/xattr.file
+# log_must attr -qr tmpattr /$TESTPOOL/$TESTFS/xattr.file
+
+# TX_WRITE, TX_LINK, TX_REMOVE
+# Make sure TX_REMOVE won't affect TX_WRITE if file is not destroyed
+log_must dd if=/dev/urandom of=/$TESTPOOL/$TESTFS/link_and_unlink bs=128k \
+   count=8
+log_must ln /$TESTPOOL/$TESTFS/link_and_unlink \
+   /$TESTPOOL/$TESTFS/link_and_unlink.link
+log_must rm /$TESTPOOL/$TESTFS/link_and_unlink.link
+
+#
+# 4. Copy TESTFS to temporary location (TESTDIR/copy)
+#
+log_must mkdir -p $TESTDIR/copy
+log_must cp -a /$TESTPOOL/$TESTFS/* $TESTDIR/copy/
+
+#
+# 5. Unmount filesystem and export the pool
+#
+# At this stage TESTFS is empty again and frozen, the intent log contains
+# a complete set of deltas to replay.
+#
+log_must zfs unmount /$TESTPOOL/$TESTFS
+
+log_note "Verify transactions to replay:"
+log_must zdb -iv $TESTPOOL/$TESTFS
+
+log_must zpool export $TESTPOOL
+
+#
+# 6. Remount TESTFS <which replays the intent log>
+#
+# Import the pool to unfreeze it and claim log blocks.  It has to be
+# `zpool import -f` because we can't write a frozen pool's labels!
+#
+log_must zpool import -f -d $VDIR $TESTPOOL
+
+#
+# 7. Compare TESTFS against the TESTDIR/copy
+#
+log_note "Verify current block usage:"
+log_must zdb -bcv $TESTPOOL
+
+# log_note "Verify copy of xattrs:"
+# log_must attr -l /$TESTPOOL/$TESTFS/xattr.dir
+# log_must attr -l /$TESTPOOL/$TESTFS/xattr.file
+
+log_note "Verify working set diff:"
+log_must diff -r /$TESTPOOL/$TESTFS $TESTDIR/copy
+
+log_note "Verify file checksum:"
+typeset checksum1=$(sha256digest /$TESTPOOL/$TESTFS/payload)
+[[ "$checksum1" == "$checksum" ]] || \
+    log_fail "checksum mismatch ($checksum1 != $checksum)"
+
+log_pass "Replay of intent log succeeds."
diff --git a/usr/src/test/zfs-tests/tests/functional/slog/slog_replay_fs_002.ksh b/usr/src/test/zfs-tests/tests/functional/slog/slog_replay_fs_002.ksh
new file mode 100755
index 0000000000..b334cac45b
--- /dev/null
+++ b/usr/src/test/zfs-tests/tests/functional/slog/slog_replay_fs_002.ksh
@@ -0,0 +1,137 @@
+#!/bin/ksh -p
+#
+# 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 2007 Sun Microsystems, Inc.  All rights reserved.
+# Use is subject to license terms.
+#
+
+. $STF_SUITE/tests/functional/slog/slog.kshlib
+
+#
+# DESCRIPTION:
+#	Verify slog replay correctly when TX_REMOVEs are followed by
+#	TX_CREATEs.
+#
+# STRATEGY:
+#	1. Create a file system (TESTFS) with a lot of files
+#	2. Freeze TESTFS
+#	3. Remove all files then create a lot of files
+#	4. Copy TESTFS to temporary location (TESTDIR/copy)
+#	5. Unmount filesystem
+#	   <at this stage TESTFS is empty again and unfrozen, and the
+#	   intent log contains a complete set of deltas to replay it>
+#	6. Remount TESTFS <which replays the intent log>
+#	7. Compare TESTFS against the TESTDIR/copy
+#
+
+verify_runnable "global"
+
+function cleanup_fs
+{
+	cleanup
+}
+
+log_assert "Replay of intent log succeeds."
+log_onexit cleanup_fs
+log_must setup
+
+#
+# 1. Create a file system (TESTFS) with a lot of files
+#
+log_must zpool create $TESTPOOL $VDEV log mirror $LDEV
+log_must zfs set compression=on $TESTPOOL
+log_must zfs create $TESTPOOL/$TESTFS
+
+# Prep for the test of TX_REMOVE followed by TX_CREATE
+dnsize=(legacy auto 1k 2k 4k 8k 16k)
+NFILES=200
+log_must mkdir /$TESTPOOL/$TESTFS/dir0
+log_must eval 'for i in $(seq $NFILES); do zfs set dnodesize=${dnsize[$RANDOM % ${#dnsize[@]}]} $TESTPOOL/$TESTFS; touch /$TESTPOOL/$TESTFS/dir0/file.$i; done'
+
+#
+# Reimport to reset dnode allocation pointer.
+# This is to make sure we will have TX_REMOVE and TX_CREATE on same id
+#
+log_must zpool export $TESTPOOL
+log_must zpool import -f -d $VDIR $TESTPOOL
+
+#
+# This dd command works around an issue where ZIL records aren't created
+# after freezing the pool unless a ZIL header already exists. Create a file
+# synchronously to force ZFS to write one out.
+#
+log_must dd if=/dev/zero of=/$TESTPOOL/$TESTFS/sync \
+    oflag=dsync,sync bs=1 count=1
+
+#
+# 2. Freeze TESTFS
+#
+log_must zpool freeze $TESTPOOL
+
+#
+# 3. Remove all files then create a lot of files
+#
+# TX_REMOVE followed by TX_CREATE
+log_must eval 'rm -f /$TESTPOOL/$TESTFS/dir0/*'
+log_must eval 'for i in $(seq $NFILES); do zfs set dnodesize=${dnsize[$RANDOM % ${#dnsize[@]}]} $TESTPOOL/$TESTFS; touch /$TESTPOOL/$TESTFS/dir0/file.$i; done'
+
+#
+# 4. Copy TESTFS to temporary location (TESTDIR/copy)
+#
+log_must mkdir -p $TESTDIR/copy
+log_must cp -a /$TESTPOOL/$TESTFS/* $TESTDIR/copy/
+
+#
+# 5. Unmount filesystem and export the pool
+#
+# At this stage TESTFS is empty again and frozen, the intent log contains
+# a complete set of deltas to replay.
+#
+log_must zfs unmount /$TESTPOOL/$TESTFS
+
+log_note "Verify transactions to replay:"
+log_must zdb -iv $TESTPOOL/$TESTFS
+
+log_must zpool export $TESTPOOL
+
+#
+# 6. Remount TESTFS <which replays the intent log>
+#
+# Import the pool to unfreeze it and claim log blocks.  It has to be
+# `zpool import -f` because we can't write a frozen pool's labels!
+#
+log_must zpool import -f -d $VDIR $TESTPOOL
+
+#
+# 7. Compare TESTFS against the TESTDIR/copy
+#
+log_note "Verify current block usage:"
+log_must zdb -bcv $TESTPOOL
+
+log_note "Verify number of files"
+log_must test "$(ls /$TESTPOOL/$TESTFS/dir0 | wc -l)" -eq $NFILES
+
+log_note "Verify working set diff:"
+log_must diff -r /$TESTPOOL/$TESTFS $TESTDIR/copy
+
+log_pass "Replay of intent log succeeds."
diff --git a/usr/src/uts/common/fs/zfs/dnode.c b/usr/src/uts/common/fs/zfs/dnode.c
index 90f425a800..2747d2f394 100644
--- a/usr/src/uts/common/fs/zfs/dnode.c
+++ b/usr/src/uts/common/fs/zfs/dnode.c
@@ -56,7 +56,6 @@ dnode_stats_t dnode_stats = {
 	{ "dnode_hold_free_lock_retry",		KSTAT_DATA_UINT64 },
 	{ "dnode_hold_free_overflow",		KSTAT_DATA_UINT64 },
 	{ "dnode_hold_free_refcount",		KSTAT_DATA_UINT64 },
-	{ "dnode_hold_free_txg",		KSTAT_DATA_UINT64 },
 	{ "dnode_free_interior_lock_retry",	KSTAT_DATA_UINT64 },
 	{ "dnode_allocate",			KSTAT_DATA_UINT64 },
 	{ "dnode_reallocate",			KSTAT_DATA_UINT64 },
@@ -1260,6 +1259,10 @@ dnode_buf_evict_async(void *dbu)
  * as an extra dnode slot by an large dnode, in which case it returns
  * ENOENT.
  *
+ * If the DNODE_DRY_RUN flag is set, we don't actually hold the dnode, just
+ * return whether the hold would succeed or not. tag and dnp should set to
+ * NULL in this case.
+ *
  * errors:
  * EINVAL - invalid object number or flags.
  * ENOSPC - hole too small to fulfill "slots" request (DNODE_MUST_BE_FREE)
@@ -1287,6 +1290,7 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots,
 
 	ASSERT(!(flag & DNODE_MUST_BE_ALLOCATED) || (slots == 0));
 	ASSERT(!(flag & DNODE_MUST_BE_FREE) || (slots > 0));
+	IMPLY(flag & DNODE_DRY_RUN, (tag == NULL) && (dnp == NULL));
 
 	/*
 	 * If you are holding the spa config lock as writer, you shouldn't
@@ -1316,8 +1320,11 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots,
 		if ((flag & DNODE_MUST_BE_FREE) && type != DMU_OT_NONE)
 			return (SET_ERROR(EEXIST));
 		DNODE_VERIFY(dn);
-		(void) zfs_refcount_add(&dn->dn_holds, tag);
-		*dnp = dn;
+		/* Don't actually hold if dry run, just return 0 */
+		if (!(flag & DNODE_DRY_RUN)) {
+			(void) zfs_refcount_add(&dn->dn_holds, tag);
+			*dnp = dn;
+		}
 		return (0);
 	}
 
@@ -1462,6 +1469,14 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots,
 			return (SET_ERROR(ENOENT));
 		}
 
+		/* Don't actually hold if dry run, just return 0 */
+		if (flag & DNODE_DRY_RUN) {
+			mutex_exit(&dn->dn_mtx);
+			dnode_slots_rele(dnc, idx, slots);
+			dbuf_rele(db, FTAG);
+			return (0);
+		}
+
 		DNODE_STAT_BUMP(dnode_hold_alloc_hits);
 	} else if (flag & DNODE_MUST_BE_FREE) {
 
@@ -1521,6 +1536,14 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots,
 			return (SET_ERROR(EEXIST));
 		}
 
+		/* Don't actually hold if dry run, just return 0 */
+		if (flag & DNODE_DRY_RUN) {
+			mutex_exit(&dn->dn_mtx);
+			dnode_slots_rele(dnc, idx, slots);
+			dbuf_rele(db, FTAG);
+			return (0);
+		}
+
 		dnode_set_slots(dnc, idx + 1, slots - 1, DN_SLOT_INTERIOR);
 		DNODE_STAT_BUMP(dnode_hold_free_hits);
 	} else {
@@ -1528,15 +1551,7 @@ dnode_hold_impl(objset_t *os, uint64_t object, int flag, int slots,
 		return (SET_ERROR(EINVAL));
 	}
 
-	if (dn->dn_free_txg) {
-		DNODE_STAT_BUMP(dnode_hold_free_txg);
-		type = dn->dn_type;
-		mutex_exit(&dn->dn_mtx);
-		dnode_slots_rele(dnc, idx, slots);
-		dbuf_rele(db, FTAG);
-		return (SET_ERROR((flag & DNODE_MUST_BE_ALLOCATED) ?
-		    ENOENT : EEXIST));
-	}
+	ASSERT0(dn->dn_free_txg);
 
 	if (zfs_refcount_add(&dn->dn_holds, tag) == 1)
 		dbuf_add_ref(db, dnh);
@@ -1627,6 +1642,16 @@ dnode_rele_and_unlock(dnode_t *dn, void *tag, boolean_t evicting)
 	}
 }
 
+/*
+ * Test whether we can create a dnode at the specified location.
+ */
+int
+dnode_try_claim(objset_t *os, uint64_t object, int slots)
+{
+	return (dnode_hold_impl(os, object, DNODE_MUST_BE_FREE | DNODE_DRY_RUN,
+	    slots, NULL, NULL));
+}
+
 void
 dnode_setdirty(dnode_t *dn, dmu_tx_t *tx)
 {
diff --git a/usr/src/uts/common/fs/zfs/sys/dnode.h b/usr/src/uts/common/fs/zfs/sys/dnode.h
index 6c8ec5e229..054e467bb7 100644
--- a/usr/src/uts/common/fs/zfs/sys/dnode.h
+++ b/usr/src/uts/common/fs/zfs/sys/dnode.h
@@ -46,6 +46,7 @@ extern "C" {
  */
 #define	DNODE_MUST_BE_ALLOCATED	1
 #define	DNODE_MUST_BE_FREE	2
+#define	DNODE_DRY_RUN		4
 
 /*
  * dnode_next_offset() flags.
@@ -393,6 +394,7 @@ int dnode_hold_impl(struct objset *dd, uint64_t object, int flag, int dn_slots,
 boolean_t dnode_add_ref(dnode_t *dn, void *ref);
 void dnode_rele(dnode_t *dn, void *ref);
 void dnode_rele_and_unlock(dnode_t *dn, void *tag, boolean_t evicting);
+int dnode_try_claim(objset_t *os, uint64_t object, int slots);
 void dnode_setdirty(dnode_t *dn, dmu_tx_t *tx);
 void dnode_sync(dnode_t *dn, dmu_tx_t *tx);
 void dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs,
@@ -511,11 +513,6 @@ typedef struct dnode_stats {
 	 */
 	kstat_named_t dnode_hold_free_overflow;
 	/*
-	 * Number of times a dnode_hold(...) was attempted on a dnode
-	 * which had already been unlinked in an earlier txg.
-	 */
-	kstat_named_t dnode_hold_free_txg;
-	/*
 	 * Number of times dnode_free_interior_slots() needed to retry
 	 * acquiring a slot zrl lock due to contention.
 	 */
diff --git a/usr/src/uts/common/fs/zfs/sys/zfs_znode.h b/usr/src/uts/common/fs/zfs/sys/zfs_znode.h
index c70eeec4ba..8b61ddb351 100644
--- a/usr/src/uts/common/fs/zfs/sys/zfs_znode.h
+++ b/usr/src/uts/common/fs/zfs/sys/zfs_znode.h
@@ -332,7 +332,7 @@ extern void zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 extern int zfs_log_create_txtype(zil_create_t, vsecattr_t *vsecp,
     vattr_t *vap);
 extern void zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
-    znode_t *dzp, char *name, uint64_t foid);
+    znode_t *dzp, char *name, uint64_t foid, boolean_t unlinked);
 #define	ZFS_NO_OBJECT	0	/* no object id */
 extern void zfs_log_link(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
     znode_t *dzp, znode_t *zp, char *name);
diff --git a/usr/src/uts/common/fs/zfs/zfs_log.c b/usr/src/uts/common/fs/zfs/zfs_log.c
index 7d3e1cc42a..ed39f5327d 100644
--- a/usr/src/uts/common/fs/zfs/zfs_log.c
+++ b/usr/src/uts/common/fs/zfs/zfs_log.c
@@ -354,12 +354,14 @@ zfs_log_create(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 	zil_itx_assign(zilog, itx, tx);
 }
 
+void zil_remove_async(zilog_t *zilog, uint64_t oid);
+
 /*
  * Handles both TX_REMOVE and TX_RMDIR transactions.
  */
 void
 zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
-    znode_t *dzp, char *name, uint64_t foid)
+    znode_t *dzp, char *name, uint64_t foid, boolean_t unlinked)
 {
 	itx_t *itx;
 	lr_remove_t *lr;
@@ -375,6 +377,17 @@ zfs_log_remove(zilog_t *zilog, dmu_tx_t *tx, uint64_t txtype,
 
 	itx->itx_oid = foid;
 
+	/*
+	 * Object ids can be re-instantiated in the next txg so
+	 * remove any async transactions to avoid future leaks.
+	 * This can happen if a fsync occurs on the re-instantiated
+	 * object for a WR_INDIRECT or WR_NEED_COPY write, which gets
+	 * the new file data and flushes a write record for the old object.
+	 */
+	if (unlinked) {
+		ASSERT((txtype & ~TX_CI) == TX_REMOVE);
+		zil_remove_async(zilog, foid);
+	}
 	zil_itx_assign(zilog, itx, tx);
 }
 
diff --git a/usr/src/uts/common/fs/zfs/zfs_replay.c b/usr/src/uts/common/fs/zfs/zfs_replay.c
index 969a56dc9b..2c643d6210 100644
--- a/usr/src/uts/common/fs/zfs/zfs_replay.c
+++ b/usr/src/uts/common/fs/zfs/zfs_replay.c
@@ -335,8 +335,8 @@ zfs_replay_create_acl(void *arg1, void *arg2, boolean_t byteswap)
 	xva.xva_vattr.va_nblocks = lr->lr_gen;
 	xva.xva_vattr.va_fsid = dnodesize;
 
-	error = dmu_object_info(zfsvfs->z_os, lr->lr_foid, NULL);
-	if (error != ENOENT)
+	error = dnode_try_claim(zfsvfs->z_os, objid, dnodesize >> DNODE_SHIFT);
+	if (error)
 		goto bail;
 
 	if (lr->lr_common.lrc_txtype & TX_CI)
@@ -469,8 +469,8 @@ zfs_replay_create(void *arg1, void *arg2, boolean_t byteswap)
 	xva.xva_vattr.va_nblocks = lr->lr_gen;
 	xva.xva_vattr.va_fsid = dnodesize;
 
-	error = dmu_object_info(zfsvfs->z_os, objid, NULL);
-	if (error != ENOENT)
+	error = dnode_try_claim(zfsvfs->z_os, objid, dnodesize >> DNODE_SHIFT);
+	if (error)
 		goto out;
 
 	if (lr->lr_common.lrc_txtype & TX_CI)
diff --git a/usr/src/uts/common/fs/zfs/zfs_vnops.c b/usr/src/uts/common/fs/zfs/zfs_vnops.c
index a5a8f79317..8c84e93240 100644
--- a/usr/src/uts/common/fs/zfs/zfs_vnops.c
+++ b/usr/src/uts/common/fs/zfs/zfs_vnops.c
@@ -1918,7 +1918,7 @@ top:
 	txtype = TX_REMOVE;
 	if (flags & FIGNORECASE)
 		txtype |= TX_CI;
-	zfs_log_remove(zilog, tx, txtype, dzp, name, obj);
+	zfs_log_remove(zilog, tx, txtype, dzp, name, obj, unlinked);
 
 	dmu_tx_commit(tx);
 out:
@@ -2234,7 +2234,8 @@ top:
 		uint64_t txtype = TX_RMDIR;
 		if (flags & FIGNORECASE)
 			txtype |= TX_CI;
-		zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT);
+		zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT,
+		    B_FALSE);
 	}
 
 	dmu_tx_commit(tx);
diff --git a/usr/src/uts/common/fs/zfs/zil.c b/usr/src/uts/common/fs/zfs/zil.c
index e56104f979..bd058fbd4f 100644
--- a/usr/src/uts/common/fs/zfs/zil.c
+++ b/usr/src/uts/common/fs/zfs/zil.c
@@ -1749,7 +1749,7 @@ zil_aitx_compare(const void *x1, const void *x2)
 /*
  * Remove all async itx with the given oid.
  */
-static void
+void
 zil_remove_async(zilog_t *zilog, uint64_t oid)
 {
 	uint64_t otxg, txg;
@@ -1802,16 +1802,6 @@ zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx)
 	itxs_t *itxs, *clean = NULL;
 
 	/*
-	 * Object ids can be re-instantiated in the next txg so
-	 * remove any async transactions to avoid future leaks.
-	 * This can happen if a fsync occurs on the re-instantiated
-	 * object for a WR_INDIRECT or WR_NEED_COPY write, which gets
-	 * the new file data and flushes a write record for the old object.
-	 */
-	if ((itx->itx_lr.lrc_txtype & ~TX_CI) == TX_REMOVE)
-		zil_remove_async(zilog, itx->itx_oid);
-
-	/*
 	 * Ensure the data of a renamed file is committed before the rename.
 	 */
 	if ((itx->itx_lr.lrc_txtype & ~TX_CI) == TX_RENAME)