Skip to contentSkip to conversation

fstests: btrfs: check qgroup doesn't crash when beyond limit

PATCH Superseded Qu Wenruo 4 participants 7 messages Nov 11, 2020 – Jan 11, 2021 Btrfs file system

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Qu Wenruo

author
#0 lore raw

PATCH 1/1 fstests: btrfs: check qgroup doesn't crash when beyond limit

There is a bug that, when btrfs is beyond qgroup limit, touching a file could crash btrfs.

Such beyond limit situation needs to be intentionally created, e.g. writing 1GiB file, then limit the subvolume to 512 MiB.
As current qgroup works pretty well at preventing us from reaching the limit.

This makes existing qgroup test cases unable to detect it.

The regression is introduced by commit c53e9653605d ("btrfs: qgroup: try to flush qgroup space when we get -EDQUOT"), and the fix is titled "btrfs: qgroup: don't commit transaction when we have already
hold a transaction handler"

---
 tests/btrfs/154     | 62 +++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/154.out |  2 ++
 tests/btrfs/group   |  1 +
 3 files changed, 65 insertions(+)
 create mode 100755 tests/btrfs/154
 create mode 100644 tests/btrfs/154.out
tests/btrfs/154 new file +62 -0
new file mode 100755
index 00000000..2a65d182
@@ -0,0 +1,62 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 SUSE Linux Products GmbH. All Rights Reserved.
+#
+# FS QA Test 154
+#
+# Test if btrfs qgroup would crash if we're modifying the fs
+# after exceeding the limit
+#
+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 btrfs
+
+# Need at least 2GiB
+_require_scratch_size $((2 * 1024 * 1024))
+_scratch_mkfs > /dev/null 2>&1
+
+_scratch_mount
+
+_pwrite_byte 0xcd 0 1G $SCRATCH_MNT/file >> $seqres.full
+
+# Make sure the data reach disk so later qgroup scan can see it
+sync
+
+$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT
+$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full
+
+# Set the limit to just 512MiB, which is way below the existing usage
+$BTRFS_UTIL_PROG qgroup limit  512M $SCRATCH_MNT $SCRATCH_MNT
+
+# Touch above file, if kernel not patched, it will trigger an ASSERT()
+#
+# Even for patched kernel, we will still get EDQUOT error, but that
+# is expected behavior.
+touch $SCRATCH_MNT/file 2>&1 | _filter_scratch
+
+# success, all done
+status=0
+exit
tests/btrfs/154.out new file +2 -0
new file mode 100644
index 00000000..b526c3f3
@@ -0,0 +1,2 @@
+QA output created by 154
+touch: setting times of 'SCRATCH_MNT/file': Disk quota exceeded
tests/btrfs/group +1 -0
index d18450c7..c491e339 100644
@@ -156,6 +156,7 @@
 151 auto quick volume
 152 auto quick metadata qgroup send
 153 auto quick qgroup limit
+154 auto quick qgroup limit
 155 auto quick send
 156 auto quick trim balance
 157 auto quick raid
Signature · 1 line
2.28.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 11 lines · Qu Wenruo

There is a bug that, when btrfs is beyond qgroup limit, touching a file could crash btrfs.

Such beyond limit situation needs to be intentionally created, e.g. writing 1GiB file, then limit the subvolume to 512 MiB.
As current qgroup works pretty well at preventing us from reaching the limit.

This makes existing qgroup test cases unable to detect it.

Quoted 7 lines

The regression is introduced by commit c53e9653605d ("btrfs: qgroup: try to flush qgroup space when we get -EDQUOT"), and the fix is titled "btrfs: qgroup: don't commit transaction when we have already
hold a transaction handler"

Looks good, just one comment below.

Quoted 56 lines
---
 tests/btrfs/154     | 62 +++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/154.out |  2 ++
 tests/btrfs/group   |  1 +
 3 files changed, 65 insertions(+)
 create mode 100755 tests/btrfs/154
 create mode 100644 tests/btrfs/154.out
tests/btrfs/154 new file +42 -0
new file mode 100755
index 00000000..2a65d182
@@ -0,0 +1,62 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 SUSE Linux Products GmbH. All Rights Reserved.
+#
+# FS QA Test 154
+#
+# Test if btrfs qgroup would crash if we're modifying the fs
+# after exceeding the limit
+#
+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 btrfs
+
+# Need at least 2GiB
+_require_scratch_size $((2 * 1024 * 1024))
+_scratch_mkfs > /dev/null 2>&1
+
+_scratch_mount
+

tests/btrfs/154 line 52

Quoted 10 lines
tests/btrfs/154 new file +10 -0
+_pwrite_byte 0xcd 0 1G $SCRATCH_MNT/file >> $seqres.full
+
+# Make sure the data reach disk so later qgroup scan can see it
+sync
+
+$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT
+$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full
+
+# Set the limit to just 512MiB, which is way below the existing usage
+$BTRFS_UTIL_PROG qgroup limit  512M $SCRATCH_MNT $SCRATCH_MNT

$SCRATCH_MNT twice by mistake, though the command still works and the test still reproduces the issue.

Eryu can probably remove one occurrence when picking this patch.

Thanks.

Quoted 33 lines
tests/btrfs/154 new file +10 -0
+
+# Touch above file, if kernel not patched, it will trigger an ASSERT()
+#
+# Even for patched kernel, we will still get EDQUOT error, but that
+# is expected behavior.
+touch $SCRATCH_MNT/file 2>&1 | _filter_scratch
+
+# success, all done
+status=0
+exit
tests/btrfs/154.out new file +2 -0
new file mode 100644
index 00000000..b526c3f3
@@ -0,0 +1,2 @@
+QA output created by 154
+touch: setting times of 'SCRATCH_MNT/file': Disk quota exceeded
tests/btrfs/group +1 -1
index d18450c7..c491e339 100644
@@ -156,6 +156,7 @@
 151 auto quick volume
 152 auto quick metadata qgroup send
 153 auto quick qgroup limit
+154 auto quick qgroup limit
 155 auto quick send
 156 auto quick trim balance
 157 auto quick raid
--

2.28.0

Signature · 3 lines
Filipe David Manana,

“Whether you think you can, or you think you can't — you're right.”
#2 lore raw
Quoted 85 lines · Filipe Manana
Quoted 11 lines · Qu Wenruo

There is a bug that, when btrfs is beyond qgroup limit, touching a file could crash btrfs.

Such beyond limit situation needs to be intentionally created, e.g. writing 1GiB file, then limit the subvolume to 512 MiB.
As current qgroup works pretty well at preventing us from reaching the limit.

This makes existing qgroup test cases unable to detect it.

Quoted 7 lines

The regression is introduced by commit c53e9653605d ("btrfs: qgroup: try to flush qgroup space when we get -EDQUOT"), and the fix is titled "btrfs: qgroup: don't commit transaction when we have already
hold a transaction handler"

Looks good, just one comment below.

Quoted 61 lines
---
 tests/btrfs/154     | 62 +++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/154.out |  2 ++
 tests/btrfs/group   |  1 +
 3 files changed, 65 insertions(+)
 create mode 100755 tests/btrfs/154
 create mode 100644 tests/btrfs/154.out
tests/btrfs/154 new file +47 -0
new file mode 100755
index 00000000..2a65d182
@@ -0,0 +1,62 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2020 SUSE Linux Products GmbH. All Rights Reserved.
+#
+# FS QA Test 154
+#
+# Test if btrfs qgroup would crash if we're modifying the fs
+# after exceeding the limit
+#
+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 btrfs
+
+# Need at least 2GiB
+_require_scratch_size $((2 * 1024 * 1024))
+_scratch_mkfs > /dev/null 2>&1
+
+_scratch_mount
+
+_pwrite_byte 0xcd 0 1G $SCRATCH_MNT/file >> $seqres.full
+
+# Make sure the data reach disk so later qgroup scan can see it
+sync
+
Quoted 8 lines

tests/btrfs/154 line 52

Quoted 5 lines
tests/btrfs/154 new file +5 -0
+$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT
+$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full
+
+# Set the limit to just 512MiB, which is way below the existing usage
+$BTRFS_UTIL_PROG qgroup limit  512M $SCRATCH_MNT $SCRATCH_MNT

$SCRATCH_MNT twice by mistake, though the command still works and the test still reproduces the issue.

Nope, that's the expected behavior.

Btrfs qgroup limit <size> <path>|<qgroupid> <path>

The first path is to determine qgroupid, while the last path is to determine the fs.

In this particular case, since we're limit the 0/5 qgroup, it's also the as the mount point, thus we specific it twice.

Thanks,
Qu

Quoted 40 lines

Eryu can probably remove one occurrence when picking this patch.

Thanks.

Quoted 33 lines
tests/btrfs/154 new file +10 -0
+
+# Touch above file, if kernel not patched, it will trigger an ASSERT()
+#
+# Even for patched kernel, we will still get EDQUOT error, but that
+# is expected behavior.
+touch $SCRATCH_MNT/file 2>&1 | _filter_scratch
+
+# success, all done
+status=0
+exit
tests/btrfs/154.out new file +2 -0
new file mode 100644
index 00000000..b526c3f3
@@ -0,0 +1,2 @@
+QA output created by 154
+touch: setting times of 'SCRATCH_MNT/file': Disk quota exceeded
tests/btrfs/group +1 -1
index d18450c7..c491e339 100644
@@ -156,6 +156,7 @@
 151 auto quick volume
 152 auto quick metadata qgroup send
 153 auto quick qgroup limit
+154 auto quick qgroup limit
 155 auto quick send
 156 auto quick trim balance
 157 auto quick raid
--

2.28.0

#3 lore raw
Quoted 11 lines · Qu Wenruo
Quoted 8 lines

tests/btrfs/154 line 52

Quoted 5 lines

+$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT +$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full

tests/btrfs/154 new file +3 -0
+
+# Set the limit to just 512MiB, which is way below the existing usage
+$BTRFS_UTIL_PROG qgroup limit  512M $SCRATCH_MNT $SCRATCH_MNT

$SCRATCH_MNT twice by mistake, though the command still works and the test still reproduces the issue.

Nope, that's the expected behavior.

Quoted 7 lines

Btrfs qgroup limit <size> <path>|<qgroupid> <path>

The first path is to determine qgroupid, while the last path is to determine the fs.

In this particular case, since we're limit the 0/5 qgroup, it's also the as the mount point, thus we specific it twice.

So why didn't you specify 0/5 so it's clear?

#4 lore raw
Quoted 14 lines · David Sterba
Quoted 13 lines · Qu Wenruo
Quoted 8 lines

tests/btrfs/154 line 52

Quoted 5 lines

+$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT +$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full

tests/btrfs/154 new file +3 -0
+
+# Set the limit to just 512MiB, which is way below the existing usage
+$BTRFS_UTIL_PROG qgroup limit  512M $SCRATCH_MNT $SCRATCH_MNT

$SCRATCH_MNT twice by mistake, though the command still works and the test still reproduces the issue.

Nope, that's the expected behavior.

Btrfs qgroup limit <size> <path>|<qgroupid> <path>

Quoted 8 lines
Quoted 5 lines

The first path is to determine qgroupid, while the last path is to determine the fs.

In this particular case, since we're limit the 0/5 qgroup, it's also the as the mount point, thus we specific it twice.

So why didn't you specify 0/5 so it's clear?

Oh no, my brain just shorted, and forgot that it's 0/5 fixed for fs tree.

0/5 is indeed much better.

Thanks,
Qu

#5 lore raw
Quoted 21 lines · Qu Wenruo
Quoted 17 lines · David Sterba
Quoted 16 lines · Qu Wenruo
Quoted 8 lines

tests/btrfs/154 line 52

Quoted 5 lines

+$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT +$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full

tests/btrfs/154 new file +3 -0
+
+# Set the limit to just 512MiB, which is way below the existing usage
+$BTRFS_UTIL_PROG qgroup limit  512M $SCRATCH_MNT $SCRATCH_MNT

$SCRATCH_MNT twice by mistake, though the command still works and the test still reproduces the issue.

Nope, that's the expected behavior.

Btrfs qgroup limit <size> <path>|<qgroupid> <path>

The first path is to determine qgroupid, while the last path is to determine the fs.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

In this particular case, since we're limit the 0/5 qgroup, it's also the as the mount point, thus we specific it twice.

So why didn't you specify 0/5 so it's clear?

Oh no, my brain just shorted, and forgot that it's 0/5 fixed for fs tree.

0/5 is indeed much better.

Any reason this wasn't merged? What happened?

Thanks.

Quoted 4 lines

Thanks,
Qu

Signature · 3 lines
Filipe David Manana,

“Whether you think you can, or you think you can't — you're right.”
#6 lore raw
Quoted 25 lines · Filipe Manana
Quoted 24 lines · Qu Wenruo
Quoted 20 lines · David Sterba
Quoted 18 lines · Qu Wenruo
Quoted 8 lines

tests/btrfs/154 line 52

Quoted 5 lines

+$BTRFS_UTIL_PROG quota enable $SCRATCH_MNT +$BTRFS_UTIL_PROG quota rescan -w $SCRATCH_MNT >> $seqres.full

tests/btrfs/154 new file +3 -0
+
+# Set the limit to just 512MiB, which is way below the existing usage
+$BTRFS_UTIL_PROG qgroup limit  512M $SCRATCH_MNT $SCRATCH_MNT

$SCRATCH_MNT twice by mistake, though the command still works and the test still reproduces the issue.

Nope, that's the expected behavior.

Btrfs qgroup limit <size> <path>|<qgroupid> <path>

The first path is to determine qgroupid, while the last path is to determine the fs.

In this particular case, since we're limit the 0/5 qgroup, it's also the as the mount point, thus we specific it twice.

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

So why didn't you specify 0/5 so it's clear?

Oh no, my brain just shorted, and forgot that it's 0/5 fixed for fs tree.

0/5 is indeed much better.

Any reason this wasn't merged? What happened?

Thanks for the reminder.

Let me just resend with 0/5 to replace the $SCRATCH_MNT.

Thanks,
Qu

Quoted 9 lines

Thanks.

Quoted 4 lines

Thanks,
Qu