Skip to contentSkip to conversation

btrfs: add a test for umount racing mount

PATCH Superseded Boris Burkov 4 participants 11 messages Jul 10, 2020 – Jul 21, 2020 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 btrfs: add a test for umount racing mount

Test if dirtying many inodes, which delays the umount in evict_inodes, then umounting and quickly mounting again causes the mount to fail.

This race is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
 tests/btrfs/215     | 52 +++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/215.out |  2 ++
 tests/btrfs/group   |  1 +
 3 files changed, 55 insertions(+)
 create mode 100755 tests/btrfs/215
 create mode 100644 tests/btrfs/215.out
tests/btrfs/215 new file +52 -0
new file mode 100755
index 00000000..b142c2d6
@@ -0,0 +1,52 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 YOUR NAME HERE.  All Rights Reserved.
+#
+# FS QA Test 215
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_test
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+	do
+		  dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+		  done
+		  _scratch_unmount&
+		  _scratch_mount
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/btrfs/215.out new file +2 -0
new file mode 100644
index 00000000..0a11773b
@@ -0,0 +1,2 @@
+QA output created by 215
+Silence is golden
tests/btrfs/group +1 -0
index 505665b5..dda0763e 100644
@@ -217,3 +217,4 @@
 212 auto balance dangerous
 213 auto balance dangerous
 214 auto quick send snapshot
+215 auto quick
Signature · 1 line
2.24.1
#1 lore raw
Quoted 21 lines · Boris Burkov

Test if dirtying many inodes, which delays the umount in evict_inodes, then umounting and quickly mounting again causes the mount to fail.

This race is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
  tests/btrfs/215     | 52 +++++++++++++++++++++++++++++++++++++++++++++
  tests/btrfs/215.out |  2 ++
  tests/btrfs/group   |  1 +
  3 files changed, 55 insertions(+)

create mode 100755 tests/btrfs/215
create mode 100644 tests/btrfs/215.out

tests/btrfs/215 new file +0 -0
new file mode 100755
index 00000000..b142c2d6

tests/btrfs/215 line 3

Quoted 3 lines
tests/btrfs/215 new file +3 -0
@@ -0,0 +1,52 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 YOUR NAME HERE.  All Rights Reserved.

Facebook should go in "YOUR NAME HERE".

Quoted 22 lines
tests/btrfs/215 new file +22 -0
+#
+# FS QA Test 215
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks

tests/btrfs/215 line 35

Quoted 10 lines
tests/btrfs/215 new file +10 -0
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic

This isn't btrfs specific, so I'd put it in tests/generic/<whatever>

tests/btrfs/215 line 47

Quoted 12 lines
tests/btrfs/215 new file +12 -0
+_supported_os Linux
+_require_test
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+	do
+		  dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+		  done
+		  _scratch_unmount&
+		  _scratch_mount
+

Looks like your tabbing got carried away here, and add a space between _scratch_unmount and the &.

Quoted 18 lines
tests/btrfs/215 new file +5 -0
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/btrfs/215.out new file +2 -0
new file mode 100644
index 00000000..0a11773b
@@ -0,0 +1,2 @@
+QA output created by 215
+Silence is golden
tests/btrfs/group +0 -0
index 505665b5..dda0763e 100644

tests/btrfs/group line 220

Quoted 4 lines
tests/btrfs/group +1 -0
@@ -217,3 +217,4 @@
  212 auto balance dangerous
  213 auto balance dangerous
  214 auto quick send snapshot
+215 auto quick

Spacing appears to be off here too. Thanks,

Josef

#2 lore raw

PATCH v2 1/1 generic: add a test for umount racing mount

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
 tests/generic/603     | 52 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/603.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 55 insertions(+)
 create mode 100755 tests/generic/603
 create mode 100644 tests/generic/603.out
tests/generic/603 new file +52 -0
new file mode 100755
index 00000000..e67899cb
@@ -0,0 +1,52 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 603
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_test
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+done
+_scratch_unmount &
+_scratch_mount
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/603.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 603
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 600 auto quick quota
 601 auto quick quota
 602 auto quick encrypt
+603 auto quick
Signature · 1 line
2.24.1
Reviewed-by on v2 1/1
#3 lore raw

Re: [PATCH v2] generic: add a test for umount racing mount

Quoted 7 lines · Boris Burkov

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

Thanks,

Josef

#4 lore raw

Re: [PATCH v2] generic: add a test for umount racing mount

In reply to Boris Burkov
Quoted 48 lines · Boris Burkov

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
 tests/generic/603     | 52 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/603.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 55 insertions(+)
 create mode 100755 tests/generic/603
 create mode 100644 tests/generic/603.out
tests/generic/603 new file +27 -0
new file mode 100755
index 00000000..e67899cb
@@ -0,0 +1,52 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 603
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter

tests/generic/603 line 37

Quoted 10 lines
tests/generic/603 new file +10 -0
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_test

It should be _require_scratch, due to you never use TEST_DIR below, but you used $SCRATCH_MNT.

tests/generic/603 line 45

Quoted 8 lines
tests/generic/603 new file +8 -0
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+done
+_scratch_unmount &

I have a question, can we make sure the this umount operation won't affect next running case?

Quoted 27 lines
tests/generic/603 new file +7 -0
+_scratch_mount
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/603.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 603
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 600 auto quick quota
 601 auto quick quota
 602 auto quick encrypt
+603 auto quick
Signature · 1 line
2.24.1
#5 lore raw

PATCH v3 1/1 generic: add a test for umount racing mount

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
 tests/generic/603     | 53 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/603.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 56 insertions(+)
 create mode 100755 tests/generic/603
 create mode 100644 tests/generic/603.out
tests/generic/603 new file +53 -0
new file mode 100755
index 00000000..8e9a80e6
@@ -0,0 +1,53 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 603
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_scratch
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+done
+_scratch_unmount &
+_scratch_mount
+wait
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/603.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 603
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 600 auto quick quota
 601 auto quick quota
 602 auto quick encrypt
+603 auto quick
Signature · 1 line
2.24.1
Reviewed-by on v3 1/1
#6 lore raw

Re: [PATCH v3] generic: add a test for umount racing mount

Quoted 8 lines · Boris Burkov

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---

If Eryu won't tend to change dd to XFS_IO_PROG, this version is good to me:)

Thanks,
Zorro

Quoted 86 lines
 tests/generic/603     | 53 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/603.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 56 insertions(+)
 create mode 100755 tests/generic/603
 create mode 100644 tests/generic/603.out
tests/generic/603 new file +53 -0
new file mode 100755
index 00000000..8e9a80e6
@@ -0,0 +1,53 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 603
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_scratch
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+done
+_scratch_unmount &
+_scratch_mount
+wait
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/603.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 603
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 600 auto quick quota
 601 auto quick quota
 602 auto quick encrypt
+603 auto quick
Signature · 1 line
2.24.1
#7 lore raw

Re: [PATCH v3] generic: add a test for umount racing mount

In reply to Boris Burkov
Quoted 48 lines · Boris Burkov

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
 tests/generic/603     | 53 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/603.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 56 insertions(+)
 create mode 100755 tests/generic/603
 create mode 100644 tests/generic/603.out
tests/generic/603 new file +27 -0
new file mode 100755
index 00000000..8e9a80e6
@@ -0,0 +1,53 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 603
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter

tests/generic/603 line 37

Quoted 10 lines
tests/generic/603 new file +10 -0
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_scratch

Ah, it's fixed in v3, so ignore my first comment in the reply to v2 :)

Thanks,
Eryu

Quoted 35 lines
tests/generic/603 new file +16 -0
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+done
+_scratch_unmount &
+_scratch_mount
+wait
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/603.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 603
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 600 auto quick quota
 601 auto quick quota
 602 auto quick encrypt
+603 auto quick
Signature · 1 line
2.24.1
#8 lore raw

PATCH v4 1/1 generic: add a test for umount racing mount

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
- dd to XFS_IO_PROG
- 1M writes to 4k writes

 tests/generic/603     | 53 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/603.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 56 insertions(+)
 create mode 100755 tests/generic/603
 create mode 100644 tests/generic/603.out
tests/generic/603 new file +53 -0
new file mode 100755
index 00000000..90f0d1d3
@@ -0,0 +1,53 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 603
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_scratch
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	$XFS_IO_PROG -c "pwrite 0 4K" $SCRATCH_MNT/$i >/dev/null 2>&1
+done
+_scratch_unmount &
+_scratch_mount
+wait
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/603.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 603
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 600 auto quick quota
 601 auto quick quota
 602 auto quick encrypt
+603 auto quick
Signature · 1 line
2.24.1
Reviewed-by on v4 1/1
#9 lore raw

Re: [PATCH v4] generic: add a test for umount racing mount

Quoted 8 lines · Boris Burkov

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---

This version looks good to me,

Quoted 89 lines

- dd to XFS_IO_PROG
- 1M writes to 4k writes

 tests/generic/603     | 53 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/603.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 56 insertions(+)
 create mode 100755 tests/generic/603
 create mode 100644 tests/generic/603.out
tests/generic/603 new file +53 -0
new file mode 100755
index 00000000..90f0d1d3
@@ -0,0 +1,53 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 603
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_scratch
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	$XFS_IO_PROG -c "pwrite 0 4K" $SCRATCH_MNT/$i >/dev/null 2>&1
+done
+_scratch_unmount &
+_scratch_mount
+wait
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/603.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 603
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 600 auto quick quota
 601 auto quick quota
 602 auto quick encrypt
+603 auto quick
Signature · 1 line
2.24.1
#10 lore raw

Re: [PATCH v2] generic: add a test for umount racing mount

In reply to Boris Burkov

[ cc'ed Amir for failure on overlayfs ]

Quoted 48 lines · Boris Burkov

Test if dirtying many inodes (which can delay umount) then unmounting and quickly mounting again causes the mount to fail.

A race, which breaks the test in btrfs, is fixed by the patch:
"btrfs: fix mount failure caused by race with umount"

---
 tests/generic/604     | 52 +++++++++++++++++++++++++++++++++++++++++++
 tests/generic/604.out |  2 ++
 tests/generic/group   |  1 +
 3 files changed, 55 insertions(+)
 create mode 100755 tests/generic/604
 create mode 100644 tests/generic/604.out
tests/generic/604 new file +27 -0
new file mode 100755
index 00000000..e67899cb
@@ -0,0 +1,52 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 Facebook  All Rights Reserved.
+#
+# FS QA Test 604
+#
+# Evicting dirty inodes can take a long time during umount.
+# Check that a new mount racing with such a delayed umount succeeds.
+#
+seq=`basename $0`
+seqres=$RESULT_DIR/$seq
+echo "QA output created by $seq"
+
+here=`pwd`
+tmp=/tmp/$$
+status=1	# failure is the default!
+trap "_cleanup; exit \$status" 0 1 2 3 15
+
+_cleanup()
+{
+	cd /
+	rm -f $tmp.*
+}
+
+# get standard environment, filters and checks
+. ./common/rc
+. ./common/filter

tests/generic/604 line 37

Quoted 10 lines
tests/generic/604 new file +10 -0
+
+# remove previous $seqres.full before test
+rm -f $seqres.full
+
+# real QA test starts here
+
+# Modify as appropriate.
+_supported_fs generic
+_supported_os Linux
+_require_test

Test takes use of scratch dev, but required test dev here.

tests/generic/604 line 44

Quoted 7 lines
tests/generic/604 new file +7 -0
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+for i in $(seq 0 500)
+do
+	dd if=/dev/zero of="$SCRATCH_MNT/$i" bs=1M count=1 > /dev/null 2>&1
+done

The kernel patch describes that it only needs to make a bunch of inodes dirty, so is it really necessary to write 500M data to the fs? Does writing less files work (e.g. 50)? Or does writing less data work (e.g. 4k file)?

Also, fstests perfers code style like

for i in $(seq 0 500); do

	$XFS_IO_PROG -c "pwrite 0 1M" $SCRATCH_MNT/$i >/dev/null 2>&1

done

tests/generic/604 line 46

Quoted 2 lines
tests/generic/604 new file +2 -0
+_scratch_unmount &
+_scratch_mount

xfs and ext4 both passed this test, but overlayfs always fails the test. I'm not sure if this is a valid test for overlay? Or it's just that overlayfs should be fixed? Amir, any comments here?

Thanks,
Eryu

Quoted 25 lines
tests/generic/604 new file +6 -0
+
+echo "Silence is golden"
+
+# success, all done
+status=0
+exit
tests/generic/604.out new file +2 -0
new file mode 100644
index 00000000..6810da89
@@ -0,0 +1,2 @@
+QA output created by 604
+Silence is golden
tests/generic/group +1 -0
index d9ab9a31..c0ace35b 100644
@@ -605,3 +605,4 @@
 601 auto quick quota
 602 auto quick encrypt
 603 auto quick quota
+604 auto quick
Signature · 1 line
2.24.1