Skip to contentSkip to conversation

fstests: Add lustre support

PATCH In review Xiao Yang 4 participants 9 messages Nov 18, 2025 – Nov 19, 2025

lore
Share & tools
lore

Xiao Yang

author
#0 lore raw

PATCH 1/1 fstests: Add lustre support

Add basic lustre support so that we can continuously improve lustre by fstests.

---
 README                 |  2 ++
 README.config-sections |  7 +++++++
 check                  |  3 ++-
 common/config          |  5 +++++
 common/rc              | 29 ++++++++++++++++++++++++++++-
 5 files changed, 44 insertions(+), 2 deletions(-)
README +2 -0
index 9e9afe3c..9e3ff13f 100644
@@ -87,6 +87,8 @@ L4: Active support from the fs list, has lots of own cases.
 +------------+-------+---------------------------------------------------------+
 | 9p         |  L1   | N/A                                                     |
 +------------+-------+---------------------------------------------------------+
+| lustre     |  L1   | N/A                                                     |
++------------+-------+---------------------------------------------------------+
 
 _______________________
 BUILDING THE FSQA SUITE
README.config-sections +7 -0
index a42d9d7b..18af445f 100644
@@ -135,3 +135,10 @@ TEST_DEV=%example.com:xfstest.test
 TEST_DIR=/mnt/xfstest.test
 SCRATCH_DEV=%example.com:xfstest.scratch
 SCRATCH_MNT=/mnt/xfstest.scratch
+
+[lustre]
+FSTYP=lustre
+TEST_DIR=/mnt/test
+TEST_DEV=192.168.1.1@tcp:/test
+SCRATCH_MNT=/mnt/scratch
+SCRATCH_DEV=192.168.1.1@tcp:/scratch
check +2 -1
index ce7eacb7..a5d43d98 100755
@@ -70,6 +70,7 @@ check options
     -pvfs2		test PVFS2
     -tmpfs		test TMPFS
     -ubifs		test ubifs
+    -lustre             test lustre
     -l			line mode diff
     -udiff		show unified diff (default)
     -n			show me, do not run tests
@@ -277,7 +278,7 @@ while [ $# -gt 0 ]; do
 	case "$1" in
 	-\? | -h | --help) usage ;;
 
-	-nfs|-afs|-glusterfs|-cifs|-9p|-fuse|-virtiofs|-pvfs2|-tmpfs|-ubifs)
+	-nfs|-afs|-glusterfs|-cifs|-9p|-fuse|-virtiofs|-pvfs2|-tmpfs|-ubifs|-lustre)
 		FSTYP="${1:1}"
 		;;
 	-overlay)
common/config +5 -0
index 1420e35d..9d3c36e8 100644
@@ -394,6 +394,9 @@ _common_mount_opts()
 	ubifs)
 		echo $UBIFS_MOUNT_OPTIONS
 		;;
+	lustre)
+		echo $LUSTRE_MOUNT_OPTIONS
+		;;
 	*)
 		;;
 	esac
@@ -539,6 +542,8 @@ _source_specific_fs()
 		[ "$UBIUPDATEVOL_PROG" = "" ] && _fatal "ubiupdatevol not found"
 		. ./common/ubifs
 		;;
+	lustre)
+		;;
 	esac
 }
 
common/rc +28 -1
index 8fd7876a..7f75963b 100644
@@ -867,6 +867,9 @@ _test_mkfs()
     pvfs2)
 	# do nothing for pvfs2
 	;;
+    lustre)
+	# do nothing for lustre
+	;;
     udf)
         $MKFS_UDF_PROG $MKFS_OPTIONS $* $TEST_DEV > /dev/null
 	;;
@@ -981,7 +984,7 @@ _scratch_mkfs()
 	local mkfs_status
 
 	case $FSTYP in
-	nfs*|afs|cifs|ceph|overlay|glusterfs|pvfs2|9p|fuse|virtiofs)
+	nfs*|afs|cifs|ceph|overlay|glusterfs|pvfs2|9p|fuse|virtiofs|lustre)
 		# unable to re-create this fstyp, just remove all files in
 		# $SCRATCH_MNT to avoid EEXIST caused by the leftover files
 		# created in previous runs
@@ -2064,6 +2067,15 @@ _require_scratch_nocheck()
 			_notrun "this test requires a valid \$SCRATCH_MNT"
 		fi
 		;;
+	lustre)
+		echo $SCRATCH_DEV | grep -q "@.*:/" > /dev/null 2>&1
+		if [ -z "$SCRATCH_DEV" -o "$?" != "0" ]; then
+			_notrun "this test requires a valid \$SCRATCH_DEV"
+		fi
+		if [ ! -d "$SCRATCH_MNT" ]; then
+			_notrun "this test requires a valid \$SCRATCH_MNT"
+		fi
+		;;
 	*)
 		 if [ -z "$SCRATCH_DEV" -o "`_is_block_dev "$SCRATCH_DEV"`" = "" ]
 		 then
@@ -2288,6 +2300,15 @@ _require_test()
 			_notrun "this test requires a valid \$TEST_DIR"
 		fi
 		;;
+	lustre)
+		echo $TEST_DEV | grep -q "@.*:/" > /dev/null 2>&1
+		if [ -z "$TEST_DEV" -o "$?" != "0" ]; then
+			_notrun "this test requires a valid \$TEST_DEV"
+		fi
+		if [ ! -d "$TEST_DIR" ]; then
+			_notrun "this test requires a valid \$TEST_DIR"
+		fi
+		;;
 	*)
 		 if [ -z "$TEST_DEV" ] || [ "`_is_block_dev "$TEST_DEV"`" = "" ]
 		 then
@@ -3658,6 +3679,9 @@ _check_test_fs()
 	;;
     pvfs2)
 	;;
+    lustre)
+	# no way to check consistency for lustre
+	;;
     udf)
 	# do nothing for now
 	;;
@@ -3720,6 +3744,9 @@ _check_scratch_fs()
 	;;
     pvfs2)
 	;;
+    lustre)
+	# no way to check consistency for lustre
+	;;
     btrfs)
 	_check_btrfs_filesystem $device
 	;;
Signature · 1 line
2.43.7
#1 lore raw
Quoted 2 lines · Xiao Yang

Add basic lustre support so that we can continuously improve lustre by fstests.

xfstests is only supposed to test upstream file systems and not perpetual out of tree things like Lustre. Community is a giving and taking.

#2 lore raw
Quoted 6 lines
Quoted 2 lines

Add basic lustre support so that we can continuously improve lustre by fstests.

xfstests is only supposed to test upstream file systems and not perpetual out of tree things like Lustre. Community is a giving and taking.

For better or worse, xfstests does support some out-of-tree filesystems like gluster and bcachefs (both listed in the support matrix in the README, which also suggests that " Firstly, there's not hard restriction about which filesystem can use fstests. Any filesystem can give fstests a try"). Lustre support wouldn't seem that out of place, even if it takes a while for Lustre to make its way in-tree.

Tim Day

author
#3 lore raw
Quoted 9 lines · Timothy

For better or worse, xfstests does support some out-of-tree filesystems like gluster and bcachefs (both listed in the support matrix in the README, which also suggests that " Firstly, there's not hard restriction about which filesystem can use fstests. Any filesystem can give fstests a try"). Lustre support wouldn't seem that out of place, even if it takes a while for Lustre to make its way in-tree.

Tim Day

Agreed. According to the README, I think xfstests can also support lustre. We can improve lustre through xfstests and give various feedback to xfstests.

Best Regards,
Xiao Yang

#4 lore raw
In reply to Day, Timothy
Quoted 9 lines · Timothy
Quoted 6 lines
Quoted 2 lines

Add basic lustre support so that we can continuously improve lustre by fstests.

xfstests is only supposed to test upstream file systems and not perpetual out of tree things like Lustre. Community is a giving and taking.

For better or worse, xfstests does support some out-of-tree filesystems like gluster

gluster is a userspace file system using fuse.

Quoted 1 line

and bcachefs (both listed in the support

bcachefs was upstream, but yes this should be cleaned up.

#5 lore raw
In reply to Xiao Yang
Quoted 2 lines

Add basic lustre support so that we can continuously improve lustre by fstests.

fstests accepts a wildcard filesystem type. I _think_ the wildcard assumes the filesystem is a disk filesystem. But could we improve that enough that it would work for Lustre? The changes required for Lustre aren't huge, so that seems plausible.

Tim Day

author
#6 lore raw
Quoted 7 lines · Timothy

fstests accepts a wildcard filesystem type. I_think_ the wildcard assumes the filesystem is a disk filesystem. But could we improve that enough that it would work for Lustre? The changes required for Lustre aren't huge, so that seems plausible.

Tim Day

Hello, Tim

Thanks for your reply.
I think this patch can make xfstests support lustre.
Would you like to improve the patch or add other features? If so, could you provide some more details?

Best Regards,
Xiao Yang

#7 lore raw
In reply to Xiao Yang
Quoted 5 lines · Xiao Yang

Add basic lustre support so that we can continuously improve lustre by fstests.

---

Hi Yang,

As you can see, the fstests community doesn't like to spend time on maintaining fstests for out-of-tree fs, especially downstream-only filesystems. I'm not a dictator. If we accept to support someone filesystem and its patches, that means we'd like to spend time to review patches for it, think about it when update fstests, fix issues on it, and so on. So I'd like to respect the willing of fstests community.

For bcachefs, it was in-tree, and we know its contributors are still trying to keep it up to date. So I'd like to keep its supporting.

For glusterfs, it's an upstream open-source filesystem, it's base on FUSE of linux kernel. It can't be a kernel module, but it's a pure upstream fs at least. Actually I brought in it for testing FUSE at beginning.

So do you do this for *upstreaming* the lustre filesystem (any active source code link?) or just for using xfstests downstream? If it's the latter, I'd suggest you to fork xfstests, and maintain this patch downstream.

I noticed that there was a discussion about Lustre filesystem upstreaming:
https://lore.kernel.org/linux-fsdevel/[email protected]/T/#med943c38d27ddac1b68e9db37d6b0f2803ee6133

But I haven't found it's online. So if the "lustre upstreaming" project get an offical version be acked by mainline linux, fstests willing to support its later developing/testing. At your early developing stage, I think you can fork and maintain a downstream xfstests temporarily.

Thanks,
Zorro

Quoted 157 lines
 README                 |  2 ++
 README.config-sections |  7 +++++++
 check                  |  3 ++-
 common/config          |  5 +++++
 common/rc              | 29 ++++++++++++++++++++++++++++-
 5 files changed, 44 insertions(+), 2 deletions(-)
README +2 -0
index 9e9afe3c..9e3ff13f 100644
@@ -87,6 +87,8 @@ L4: Active support from the fs list, has lots of own cases.
 +------------+-------+---------------------------------------------------------+
 | 9p         |  L1   | N/A                                                     |
 +------------+-------+---------------------------------------------------------+
+| lustre     |  L1   | N/A                                                     |
++------------+-------+---------------------------------------------------------+
 
 _______________________
 BUILDING THE FSQA SUITE
README.config-sections +7 -0
index a42d9d7b..18af445f 100644
@@ -135,3 +135,10 @@ TEST_DEV=%example.com:xfstest.test
 TEST_DIR=/mnt/xfstest.test
 SCRATCH_DEV=%example.com:xfstest.scratch
 SCRATCH_MNT=/mnt/xfstest.scratch
+
+[lustre]
+FSTYP=lustre
+TEST_DIR=/mnt/test
+TEST_DEV=192.168.1.1@tcp:/test
+SCRATCH_MNT=/mnt/scratch
+SCRATCH_DEV=192.168.1.1@tcp:/scratch
check +2 -1
index ce7eacb7..a5d43d98 100755
@@ -70,6 +70,7 @@ check options
     -pvfs2		test PVFS2
     -tmpfs		test TMPFS
     -ubifs		test ubifs
+    -lustre             test lustre
     -l			line mode diff
     -udiff		show unified diff (default)
     -n			show me, do not run tests
@@ -277,7 +278,7 @@ while [ $# -gt 0 ]; do
 	case "$1" in
 	-\? | -h | --help) usage ;;
 
-	-nfs|-afs|-glusterfs|-cifs|-9p|-fuse|-virtiofs|-pvfs2|-tmpfs|-ubifs)
+	-nfs|-afs|-glusterfs|-cifs|-9p|-fuse|-virtiofs|-pvfs2|-tmpfs|-ubifs|-lustre)
 		FSTYP="${1:1}"
 		;;
 	-overlay)
common/config +5 -0
index 1420e35d..9d3c36e8 100644
@@ -394,6 +394,9 @@ _common_mount_opts()
 	ubifs)
 		echo $UBIFS_MOUNT_OPTIONS
 		;;
+	lustre)
+		echo $LUSTRE_MOUNT_OPTIONS
+		;;
 	*)
 		;;
 	esac
@@ -539,6 +542,8 @@ _source_specific_fs()
 		[ "$UBIUPDATEVOL_PROG" = "" ] && _fatal "ubiupdatevol not found"
 		. ./common/ubifs
 		;;
+	lustre)
+		;;
 	esac
 }
 
common/rc +28 -1
index 8fd7876a..7f75963b 100644
@@ -867,6 +867,9 @@ _test_mkfs()
     pvfs2)
 	# do nothing for pvfs2
 	;;
+    lustre)
+	# do nothing for lustre
+	;;
     udf)
         $MKFS_UDF_PROG $MKFS_OPTIONS $* $TEST_DEV > /dev/null
 	;;
@@ -981,7 +984,7 @@ _scratch_mkfs()
 	local mkfs_status
 
 	case $FSTYP in
-	nfs*|afs|cifs|ceph|overlay|glusterfs|pvfs2|9p|fuse|virtiofs)
+	nfs*|afs|cifs|ceph|overlay|glusterfs|pvfs2|9p|fuse|virtiofs|lustre)
 		# unable to re-create this fstyp, just remove all files in
 		# $SCRATCH_MNT to avoid EEXIST caused by the leftover files
 		# created in previous runs
@@ -2064,6 +2067,15 @@ _require_scratch_nocheck()
 			_notrun "this test requires a valid \$SCRATCH_MNT"
 		fi
 		;;
+	lustre)
+		echo $SCRATCH_DEV | grep -q "@.*:/" > /dev/null 2>&1
+		if [ -z "$SCRATCH_DEV" -o "$?" != "0" ]; then
+			_notrun "this test requires a valid \$SCRATCH_DEV"
+		fi
+		if [ ! -d "$SCRATCH_MNT" ]; then
+			_notrun "this test requires a valid \$SCRATCH_MNT"
+		fi
+		;;
 	*)
 		 if [ -z "$SCRATCH_DEV" -o "`_is_block_dev "$SCRATCH_DEV"`" = "" ]
 		 then
@@ -2288,6 +2300,15 @@ _require_test()
 			_notrun "this test requires a valid \$TEST_DIR"
 		fi
 		;;
+	lustre)
+		echo $TEST_DEV | grep -q "@.*:/" > /dev/null 2>&1
+		if [ -z "$TEST_DEV" -o "$?" != "0" ]; then
+			_notrun "this test requires a valid \$TEST_DEV"
+		fi
+		if [ ! -d "$TEST_DIR" ]; then
+			_notrun "this test requires a valid \$TEST_DIR"
+		fi
+		;;
 	*)
 		 if [ -z "$TEST_DEV" ] || [ "`_is_block_dev "$TEST_DEV"`" = "" ]
 		 then
@@ -3658,6 +3679,9 @@ _check_test_fs()
 	;;
     pvfs2)
 	;;
+    lustre)
+	# no way to check consistency for lustre
+	;;
     udf)
 	# do nothing for now
 	;;
@@ -3720,6 +3744,9 @@ _check_scratch_fs()
 	;;
     pvfs2)
 	;;
+    lustre)
+	# no way to check consistency for lustre
+	;;
     btrfs)
 	_check_btrfs_filesystem $device
 	;;
Signature · 1 line
2.43.7
author
#8 lore raw
Quoted 23 lines · Zorro Lang

Hi Yang,

As you can see, the fstests community doesn't like to spend time on maintaining fstests for out-of-tree fs, especially downstream-only filesystems. I'm not a dictator. If we accept to support someone filesystem and its patches, that means we'd like to spend time to review patches for it, think about it when update fstests, fix issues on it, and so on. So I'd like to respect the willing of fstests community.

For bcachefs, it was in-tree, and we know its contributors are still trying to keep it up to date. So I'd like to keep its supporting.

For glusterfs, it's an upstream open-source filesystem, it's base on FUSE of linux kernel. It can't be a kernel module, but it's a pure upstream fs at least. Actually I brought in it for testing FUSE at beginning.

So do you do this for*upstreaming* the lustre filesystem (any active source code link?) or just for using xfstests downstream? If it's the latter, I'd suggest you to fork xfstests, and maintain this patch downstream.

I noticed that there was a discussion about Lustre filesystem upstreaming:
https://lore.kernel.org/linux-fsdevel/[email protected]/T/#med943c38d27ddac1b68e9db37d6b0f2803ee6133

Quoted 7 lines

But I haven't found it's online. So if the "lustre upstreaming" project get an offical version be acked by mainline linux, fstests willing to support its later developing/testing. At your early developing stage, I think you can fork and maintain a downstream xfstests temporarily.

Thanks,
Zorro

Hi zorro

Thanks for your explanation.
OK, I will fork xfstests and maintain it downstream for now.

Best Regards,
Xiao Yang