OpenZFS 7104 - increase indirect block size

Authored by: Matthew Ahrens <[email protected]> Reviewed by: George Wilson <[email protected]> Reviewed by: Paul Dagnelie <[email protected]> Reviewed by: Dan McDonald <[email protected]> Approved by: Robert Mustacchi <[email protected]> Ported-by: George Melikov <[email protected]> OpenZFS-issue: https://www.illumos.org/issues/7104 OpenZFS-commit: openzfs/openzfs@4b5c8e9 TEST_ZILTEST_SKIP="yes" TEST_ZFSTESTS_DISKSIZE=4G
openzfs · Feb 1, 2017 · 8ba114a · 8ba114a
1 parent 456079d
commit 8ba114a
Show file tree

Hide file tree

Showing 52 changed files with 209 additions and 150 deletions.
diff --git a/include/sys/dnode.h b/include/sys/dnode.h
@@ -64,7 +64,7 @@ extern "C" {
  * 4 levels of indirect blocks would not be able to guarantee addressing an
  * entire object, so 5 levels will be used, but 5 * (20 - 7) = 65.
  */
-#define	DN_MAX_INDBLKSHIFT	14	/* 16k */
+#define	DN_MAX_INDBLKSHIFT	17	/* 128k */
 #define	DNODE_BLOCK_SHIFT	14	/* 16k */
 #define	DNODE_CORE_SIZE		64	/* 64 bytes for dnode sans blkptrs */
 #define	DN_MAX_OBJECT_SHIFT	48	/* 256 trillion (zfs_fid_t limit) */
@@ -101,6 +101,11 @@ extern "C" {
 
 #define	DNODES_PER_BLOCK_SHIFT	(DNODE_BLOCK_SHIFT - DNODE_SHIFT)
 #define	DNODES_PER_BLOCK	(1ULL << DNODES_PER_BLOCK_SHIFT)
+
+/*
+ * This is inaccurate if the indblkshift of the particular object is not the
+ * max.  But it's only used by userland to calculate the zvol reservation.
+ */
 #define	DNODES_PER_LEVEL_SHIFT	(DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
 #define	DNODES_PER_LEVEL	(1ULL << DNODES_PER_LEVEL_SHIFT)
 

diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c
@@ -858,11 +858,17 @@ dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
 
 		/*
 		 * Determine the number of levels necessary for the meta-dnode
-		 * to contain DN_MAX_OBJECT dnodes.
+		 * to contain DN_MAX_OBJECT dnodes.  Note that in order to
+		 * ensure that we do not overflow 64 bits, there has to be
+		 * a nlevels that gives us a number of blocks > DN_MAX_OBJECT
+		 * but < 2^64.  Therefore,
+		 * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT) (10) must be
+		 * less than (64 - log2(DN_MAX_OBJECT)) (16).
 		 */
-		while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift +
+		while ((uint64_t)mdn->dn_nblkptr <<
+		    (mdn->dn_datablkshift - DNODE_SHIFT +
 		    (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) <
-		    DN_MAX_OBJECT * sizeof (dnode_phys_t))
+		    DN_MAX_OBJECT)
 			levels++;
 
 		mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] =

diff --git a/module/zfs/spa_misc.c b/module/zfs/spa_misc.c
@@ -335,9 +335,14 @@ int spa_asize_inflation = 24;
  * it is possible to run the pool completely out of space, causing it to
  * be permanently read-only.
  *
+ * Note that on very small pools, the slop space will be larger than
+ * 3.2%, in an effort to have it be at least spa_min_slop (128MB),
+ * but we never allow it to be more than half the pool size.
+ *
  * See also the comments in zfs_space_check_t.
  */
 int spa_slop_shift = 5;
+uint64_t spa_min_slop = 128 * 1024 * 1024;
 
 /*
  * ==========================================================================
@@ -1607,15 +1612,16 @@ spa_get_asize(spa_t *spa, uint64_t lsize)
 
 /*
  * Return the amount of slop space in bytes.  It is 1/32 of the pool (3.2%),
- * or at least 32MB.
+ * or at least 128MB, unless that would cause it to be more than half the
+ * pool size.
  *
  * 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));
+	return (MAX(space >> spa_slop_shift, MIN(space >> 1, spa_min_slop)));
 }
 
 uint64_t

diff --git a/scripts/zconfig.sh b/scripts/zconfig.sh
@@ -360,7 +360,7 @@ test_6() {
 	# Create a pool and volume.
 	${ZFS_SH} zfs="spa_config_path=${TMP_CACHE}" || fail 1
 	${ZPOOL_CREATE_SH} -p ${POOL_NAME} -c lo-raid0 || fail 2
-	${ZFS} create -V 800M ${FULL_ZVOL_NAME} || fail 3
+	${ZFS} create -s -V 800M ${FULL_ZVOL_NAME} || fail 3
 	${ZFS} set snapdev=visible ${FULL_ZVOL_NAME} || fail 3
 	label /dev/zvol/${FULL_ZVOL_NAME} msdos || fail 4
 	partition /dev/zvol/${FULL_ZVOL_NAME} primary 1 -1 || fail 4

diff --git a/scripts/zfs-tests.sh b/scripts/zfs-tests.sh
@@ -39,7 +39,7 @@ QUIET=
 CLEANUP=1
 CLEANUPALL=0
 LOOPBACK=1
-FILESIZE="2G"
+FILESIZE="4G"
 RUNFILE=${RUNFILE:-"linux.run"}
 FILEDIR=${FILEDIR:-/var/tmp}
 DISKS=${DISKS:-""}
@@ -165,7 +165,7 @@ OPTIONS:
 	-k          Disable cleanup after test failure
 	-f          Use files only, disables block device tests
 	-d DIR      Use DIR for files and loopback devices
-	-s SIZE     Use vdevs of SIZE (default: 2G)
+	-s SIZE     Use vdevs of SIZE (default: 4G)
 	-r RUNFILE  Run tests in RUNFILE (default: linux.run)
 
 EXAMPLES:

diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run
@@ -529,7 +529,8 @@ tests = ['quota_001_pos', 'quota_002_pos', 'quota_003_pos',
 tests = ['raidz_001_neg', 'raidz_002_pos']
 
 [tests/functional/redundancy]
-tests = ['redundancy_001_pos', 'redundancy_002_pos', 'redundancy_003_pos']
+tests = ['redundancy_001_pos', 'redundancy_002_pos', 'redundancy_003_pos',
+    'redundancy_004_neg']
 
 [tests/functional/refquota]
 tests = ['refquota_001_pos', 'refquota_002_pos', 'refquota_003_pos',

diff --git a/tests/zfs-tests/include/default.cfg.in b/tests/zfs-tests/include/default.cfg.in
@@ -160,6 +160,12 @@ export BIGVOLSIZE=1eb
 # Default to limit disks to be checked
 export MAX_FINDDISKSNUM=6
 
+# Default minimum size for file based vdevs in the test suite
+export MINVDEVSIZE=$((256 * 1024 * 1024))
+
+# Minimum vdev size possible as defined in the OS
+export SPA_MINDEVSIZE=$((64 * 1024 * 1024))
+
 # For iscsi target support
 export ISCSITGTFILE=/tmp/iscsitgt_file
 export ISCSITGT_FMRI=svc:/system/iscsitgt:default

diff --git a/tests/zfs-tests/include/libtest.shlib b/tests/zfs-tests/include/libtest.shlib
@@ -1542,7 +1542,7 @@ function zfs_zones_setup #zone_name zone_root zone_ip
 	#
 	if verify_slog_support ; then
 		typeset sdevs="/var/tmp/sdev1 /var/tmp/sdev2"
-		log_must $MKFILE 100M $sdevs
+		log_must $MKFILE $MINVDEVSIZE $sdevs
 		log_must $ZPOOL add $pool_name log mirror $sdevs
 	fi
 
@@ -2512,7 +2512,7 @@ function verify_slog_support
 	typeset sdev=$dir/b
 
 	$MKDIR -p $dir
-	$MKFILE 64M $vdev $sdev
+	$MKFILE $MINVDEVSIZE $vdev $sdev
 
 	typeset -i ret=0
 	if ! $ZPOOL create -n $pool $vdev log $sdev > /dev/null 2>&1; then

diff --git a/tests/zfs-tests/tests/functional/bootfs/bootfs_001_pos.ksh b/tests/zfs-tests/tests/functional/bootfs/bootfs_001_pos.ksh
@@ -27,6 +27,10 @@
 # Copyright 2015 Nexenta Systems, Inc.
 #
 
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 
 #
@@ -62,7 +66,7 @@ log_onexit cleanup
 
 typeset VDEV=$TESTDIR/bootfs_001_pos_a.$$.dat
 
-log_must $MKFILE 400m $VDEV
+log_must $MKFILE $MINVDEVSIZE $VDEV
 create_pool "$TESTPOOL" "$VDEV"
 log_must $ZFS create $TESTPOOL/$TESTFS
 

diff --git a/tests/zfs-tests/tests/functional/bootfs/bootfs_002_neg.ksh b/tests/zfs-tests/tests/functional/bootfs/bootfs_002_neg.ksh
@@ -27,6 +27,10 @@
 # Copyright 2015 Nexenta Systems, Inc.
 #
 
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 
 #

diff --git a/tests/zfs-tests/tests/functional/bootfs/bootfs_003_pos.ksh b/tests/zfs-tests/tests/functional/bootfs/bootfs_003_pos.ksh
@@ -25,6 +25,10 @@
 # Use is subject to license terms.
 #
 
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 
 #
@@ -59,7 +63,7 @@ fi
 log_onexit cleanup
 
 log_assert "Valid pool names are accepted by zpool set bootfs"
-$MKFILE 64m $TESTDIR/bootfs_003.$$.dat
+$MKFILE $MINVDEVSIZE $TESTDIR/bootfs_003.$$.dat
 
 typeset -i i=0;
 

diff --git a/tests/zfs-tests/tests/functional/bootfs/bootfs_004_neg.ksh b/tests/zfs-tests/tests/functional/bootfs/bootfs_004_neg.ksh
@@ -25,6 +25,10 @@
 # Use is subject to license terms.
 #
 
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 
 #
@@ -74,7 +78,7 @@ pools[${#pools[@]}]="$bigname"
 
 
 
-$MKFILE 64m $TESTDIR/bootfs_004.$$.dat
+$MKFILE $MINVDEVSIZE $TESTDIR/bootfs_004.$$.dat
 
 typeset -i i=0;
 

diff --git a/tests/zfs-tests/tests/functional/bootfs/bootfs_005_neg.ksh b/tests/zfs-tests/tests/functional/bootfs/bootfs_005_neg.ksh
@@ -25,6 +25,10 @@
 # Use is subject to license terms.
 #
 
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 . $STF_SUITE/tests/functional/cli_root/zpool_upgrade/zpool_upgrade.kshlib
 

diff --git a/tests/zfs-tests/tests/functional/bootfs/bootfs_006_pos.ksh b/tests/zfs-tests/tests/functional/bootfs/bootfs_006_pos.ksh
@@ -25,6 +25,10 @@
 # Use is subject to license terms.
 #
 
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 
 #
@@ -92,7 +96,7 @@ log_assert "Pools of correct vdev types accept boot property"
 
 
 log_onexit cleanup
-log_must $MKFILE 64m $VDEV1 $VDEV2 $VDEV3 $VDEV4
+log_must $MKFILE $MINVDEVSIZE $VDEV1 $VDEV2 $VDEV3 $VDEV4
 
 
 ## the following configurations are supported bootable pools

diff --git a/tests/zfs-tests/tests/functional/bootfs/bootfs_008_neg.ksh b/tests/zfs-tests/tests/functional/bootfs/bootfs_008_neg.ksh
@@ -25,6 +25,10 @@
 # Use is subject to license terms.
 #
 
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 
 #
@@ -60,7 +64,7 @@ typeset COMP_FS=$TESTPOOL/COMP_FS
 log_onexit cleanup
 log_assert $assert_msg
 
-log_must $MKFILE 300m $VDEV
+log_must $MKFILE $MINVDEVSIZE $VDEV
 log_must $ZPOOL create $TESTPOOL $VDEV
 log_must $ZFS create $COMP_FS
 

diff --git a/tests/zfs-tests/tests/functional/cache/cache.cfg b/tests/zfs-tests/tests/functional/cache/cache.cfg
@@ -25,7 +25,7 @@
 #
 
 #
-# Copyright (c) 2013 by Delphix. All rights reserved.
+# Copyright (c) 2013, 2015 by Delphix. All rights reserved.
 #
 
 . $STF_SUITE/include/libtest.shlib
@@ -60,7 +60,7 @@ function set_disks
 set_disks
 set_device_dir
 
-export SIZE=64M
+export SIZE=$MINVDEVSIZE
 
 export VDIR=$TESTDIR/disk.cache
 export VDIR2=$TESTDIR/disk2.cache

diff --git a/tests/zfs-tests/tests/functional/cachefile/cachefile_004_pos.ksh b/tests/zfs-tests/tests/functional/cachefile/cachefile_004_pos.ksh
@@ -26,7 +26,7 @@
 #
 
 #
-# Copyright (c) 2013 by Delphix. All rights reserved.
+# Copyright (c) 2013, 2015 by Delphix. All rights reserved.
 #
 
 . $STF_SUITE/include/libtest.shlib
@@ -84,7 +84,7 @@ log_must $ZPOOL create $TESTPOOL $DISKS
 mntpnt=$(get_prop mountpoint $TESTPOOL)
 typeset -i i=0
 while ((i < 2)); do
-	log_must $MKFILE 64M $mntpnt/vdev$i
+	log_must $MKFILE $MINVDEVSIZE $mntpnt/vdev$i
 	eval vdev$i=$mntpnt/vdev$i
 	((i += 1))
 done

diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_destroy/zfs_destroy_015_pos.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_destroy/zfs_destroy_015_pos.ksh
@@ -11,7 +11,7 @@
 #
 
 #
-# Copyright (c) 2012 by Delphix. All rights reserved.
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
 #
 
 # DESCRIPTION
@@ -139,7 +139,7 @@ log_must snapexists $TESTPOOL/$TESTFS1@snap3
 
 log_note "zfs destroy for snapshots from different pools"
 VIRTUAL_DISK=/var/tmp/disk
-log_must $DD if=/dev/urandom of=$VIRTUAL_DISK bs=1M count=64
+log_must $MKFILE $MINVDEVSIZE $VIRTUAL_DISK
 log_must $ZPOOL create $TESTPOOL2 $VIRTUAL_DISK
 log_must poolexists $TESTPOOL2
 

diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_get/zfs_get_004_pos.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_get/zfs_get_004_pos.ksh
@@ -26,7 +26,7 @@
 #
 
 #
-# Copyright (c) 2012 by Delphix. All rights reserved.
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
 #
 
 . $STF_SUITE/include/libtest.shlib
@@ -107,9 +107,9 @@ allds="$fs $clone $fssnap $volsnap"
 
 #create pool and datasets to guarantee testing under multiple pools and datasets.
 file=$TESTDIR1/poolfile
-typeset -i FILESIZE=104857600    #100M
-(( DFILESIZE = FILESIZE * 2 ))   # double of FILESIZE
-typeset -i VOLSIZE=10485760      #10M
+typeset FILESIZE=$MINVDEVSIZE
+(( DFILESIZE = $FILESIZE * 2 ))
+typeset -i VOLSIZE=10485760
 availspace=$(get_prop available $TESTPOOL)
 typeset -i i=0
 

diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_rename/zfs_rename_005_neg.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_rename/zfs_rename_005_neg.ksh
@@ -24,6 +24,11 @@
 # Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
 # Use is subject to license terms.
 #
+
+#
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+#
+
 . $STF_SUITE/include/libtest.shlib
 . $STF_SUITE/tests/functional/cli_root/zfs_rename/zfs_rename.kshlib
 
@@ -63,8 +68,7 @@ log_assert "'zfs rename' should fail while datasets are within different pool."
 
 additional_setup
 
-typeset FILESIZE=64m
-log_must $MKFILE $FILESIZE $TESTDIR/$TESTFILE1
+log_must $MKFILE $MINVDEVSIZE $TESTDIR/$TESTFILE1
 create_pool $TESTPOOL1 $TESTDIR/$TESTFILE1
 
 for src in ${src_dataset[@]} ; do

diff --git a/tests/zfs-tests/tests/functional/cli_root/zfs_snapshot/zfs_snapshot_008_neg.ksh b/tests/zfs-tests/tests/functional/cli_root/zfs_snapshot/zfs_snapshot_008_neg.ksh
@@ -21,7 +21,7 @@
 #
 
 #
-# Copyright (c) 2012 by Delphix. All rights reserved.
+# Copyright (c) 2012, 2015 by Delphix. All rights reserved.
 #
 
 . $STF_SUITE/tests/functional/cli_root/zfs_snapshot/zfs_snapshot.cfg
@@ -57,8 +57,8 @@ function cleanup
 log_assert "'zfs snapshot pool1@snap1 pool2@snap2' should fail since snapshots are in different pools."
 log_onexit cleanup
 
-log_must $MKFILE 64m $SNAPDEV1
-log_must $MKFILE 64m $SNAPDEV2
+log_must $MKFILE $MINVDEVSIZE $SNAPDEV1
+log_must $MKFILE $MINVDEVSIZE $SNAPDEV2
 
 log_must $ZPOOL create $SNAPPOOL1 $SNAPDEV1
 log_must $ZPOOL create $SNAPPOOL2 $SNAPDEV2
-Original file line number
+Diff line change
@@ Expand Up / @@ -27,6 +27,10 @@ @@
     # Copyright 2015 Nexenta Systems, Inc.
     #
+    #
+    # Copyright (c) 2012, 2015 by Delphix. All rights reserved.
+    #
     . $STF_SUITE/include/libtest.shlib
     #
@@ Expand Down @@