[v2] btrfs: Test proper interaction between skip_balance and paused balance

Commit Message

Nikolay Borisov Nov. 1, 2021, 11:54 a.m. UTC

Ensure a device can be added to a filesystem that has a paused balance
operation and has been mounted with the 'skip_balance' mount option

Signed-off-by: Nikolay Borisov <[e-mail]>
---

Changes since v1:

 * Removed dependence on exclusive_operation sysfs
 * Extended the test to ensure all exclusive ops are forbidden apart from device
 add.
 * Also ensures exclusive ops are still forbidden following device add

 tests/btrfs/049     | 72 +++++++++++++++++++++++++++++++++++++++++++++
 tests/btrfs/049.out |  1 +
 2 files changed, 73 insertions(+)
 create mode 100755 tests/btrfs/049

--
2.17.1

Comments

Anand Jain Nov. 2, 2021, 3:49 a.m. UTC | #1

>   * Extended the test to ensure all exclusive ops are forbidden apart from device
>   add. >   * Also ensures exclusive ops are still forbidden following device add


> +_begin_fstest quick balance auto
> +
> +# real QA test starts here
> +
> +_supported_fs btrfs
> +_require_scratch_dev_pool 3
> +
> +external_device=$(echo $SCRATCH_DEV_POOL | $AWK_PROG '{print $3}')

  Pls use _spare_dev_get().
  Because _scratch_dev_pool_get() can change devices it gets.

> +
> +_scratch_dev_pool_get 2
> +
> +_scratch_pool_mkfs >/dev/null
> +_scratch_mount
> +
> +check_exclusive_ops()
> +{
> +	$BTRFS_UTIL_PROG device remove 2 $SCRATCH_MNT &>/dev/null
> +	[ $? -ne 0 ] || _fail "Successfully removed device"
> +	$BTRFS_UTIL_PROG filesystem resize -5m $SCRATCH_MNT &> /dev/null
> +	[ $? -ne 0 ] || _fail "Successfully resized filesystem"
> +	$BTRFS_UTIL_PROG replace start -B 2 $external_device $SCRATCH_MNT &> /dev/null
> +	[ $? -ne 0 ] || _fail "Successfully replaced device"

  Also, add the exclusive ops swap activate.
  We had a couple of bugs with excl-ops with swap earlier.

> +}
> +
> +uuid=$(findmnt -n -o UUID $SCRATCH_MNT)
> +
> +# Create some files on the so that balance doesn't complete instantly
> +args=`_scale_fsstress_args -z \
> +	-f write=10 -f creat=10 \
> +	-n 1000 -p 2 -d $SCRATCH_MNT/stress_dir`
> +echo "Run fsstress $args" >>$seqres.full
> +$FSSTRESS_PROG $args >/dev/null 2>&1
> +
> +# Start and pause balance to ensure it will be restored on remount
> +echo "Start balance" >>$seqres.full
> +_run_btrfs_balance_start --bg "$SCRATCH_MNT"
> +$BTRFS_UTIL_PROG balance pause "$SCRATCH_MNT"

Here it depends on the synchronizations between the balance start and 
the pause.
Pls, use btrfs balance status to check if the pause is successful.
Then call _fail("Couldn't pause balance on this system") if it isn't. 
Which is better than letting the further of the commands to fail.

Patch

diff --git a/tests/btrfs/049 b/tests/btrfs/049
new file mode 100755
index 000000000000..dc3c75646d63
--- /dev/null
+++ b/tests/btrfs/049
@@ -0,0 +1,72 @@ 
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2021 SUSE Linux Products GmbH.  All Rights Reserved.
+#
+# FS QA Test 049
+#
+# Ensure that it's possible to add a device when we have a paused balance
+# and the filesystem is mounted with skip_balance. The issue is fixed by a patch
+# titled "btrfs: allow device add if balance is paused"
+#
+. ./common/preamble
+_begin_fstest quick balance auto
+
+# real QA test starts here
+
+_supported_fs btrfs
+_require_scratch_dev_pool 3
+
+external_device=$(echo $SCRATCH_DEV_POOL | $AWK_PROG '{print $3}')
+
+_scratch_dev_pool_get 2
+
+_scratch_pool_mkfs >/dev/null
+_scratch_mount
+
+check_exclusive_ops()
+{
+	$BTRFS_UTIL_PROG device remove 2 $SCRATCH_MNT &>/dev/null
+	[ $? -ne 0 ] || _fail "Successfully removed device"
+	$BTRFS_UTIL_PROG filesystem resize -5m $SCRATCH_MNT &> /dev/null
+	[ $? -ne 0 ] || _fail "Successfully resized filesystem"
+	$BTRFS_UTIL_PROG replace start -B 2 $external_device $SCRATCH_MNT &> /dev/null
+	[ $? -ne 0 ] || _fail "Successfully replaced device"
+}
+
+uuid=$(findmnt -n -o UUID $SCRATCH_MNT)
+
+# Create some files on the so that balance doesn't complete instantly
+args=`_scale_fsstress_args -z \
+	-f write=10 -f creat=10 \
+	-n 1000 -p 2 -d $SCRATCH_MNT/stress_dir`
+echo "Run fsstress $args" >>$seqres.full
+$FSSTRESS_PROG $args >/dev/null 2>&1
+
+# Start and pause balance to ensure it will be restored on remount
+echo "Start balance" >>$seqres.full
+_run_btrfs_balance_start --bg "$SCRATCH_MNT"
+$BTRFS_UTIL_PROG balance pause "$SCRATCH_MNT"
+
+# exclusive ops should be blocked on manual pause of balance
+check_exclusive_ops
+
+# balance is now placed in paused state during mount
+_scratch_cycle_mount "skip_balance"
+
+# exclusive ops should be blocked on balance pause due to 'skip_balance'
+check_exclusive_ops
+
+# Device add is the only allowed operation
+$BTRFS_UTIL_PROG device add -K -f $external_device "$SCRATCH_MNT"
+
+# exclusive ops should still be blocked on account that balance is still paused
+check_exclusive_ops
+
+# Should be possible to resume balance after device add
+$BTRFS_UTIL_PROG balance resume "$SCRATCH_MNT" &>/dev/null
+[ $? -eq 0 ] || _fail "Couldn't resume balance after device add"
+
+_scratch_dev_pool_put
+echo "Silence is golden"
+status=0
+exit
diff --git a/tests/btrfs/049.out b/tests/btrfs/049.out
index cb0061b33ff0..c69568ad9323 100644
--- a/tests/btrfs/049.out
+++ b/tests/btrfs/049.out
@@ -1 +1,2 @@ 
 QA output created by 049
+Silence is golden