From: Johannes Thumshirn <johannes.thumshirn@wdc.com>
To: socketpair@gmail.com
Cc: David Sterba <dsterba@suse.com>, Chris Mason <mason@kernel.org>,
Naohiro Aota <naohiro.aota@wdc.com>,
linux-btrfs@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3] btrfs: zoned: reset leaked zones on read-write mount
Date: Tue, 17 Nov 2026 12:46:39 +0100 [thread overview]
Message-ID: <avw98epZjlOLcu8g@mayhem.fritz.box> (raw)
In-Reply-To: <20261102-b4-btrfs-zoned-reset-leaked-zones-v3-1-7979bda654ff@gmail.com>
On Mon, Nov 02, 2026 at 08:35:04PM +0500, Mark Korenberg wrote:
> If a transaction commit is aborted after the zones of a new block group
> were written but before the superblock was, the chunk item and the dev
> extent are never persisted and the zones leak. The chunk allocator skips
> non-empty zones, so they are never reused, open ones keep counting
> against the active zone limit, and btrfs check reports them as
> unreferenced.
Sorry for the delay and thanks for picking this up.
Please mention that this needs btrfs-progs v7.3 or later, older versions
don't check for leaked zones.
[...]
> Testing: virtme-ng VM, zoned null_blk with 64 MiB zones,
> CONFIG_PROVE_LOCKING, this patch plus debug logging of every decision:
> - not zoned filesystem: nothing done
> - fresh filesystem: the first mount resets zone 3, which mkfs.btrfs
> (btrfs-progs 7.1) leaves written outside any chunk; nothing to do on
> the next mount, btrfs check clean
That one is fixed in btrfs-progs v7.3 by commit dcc41ad6 ("btrfs-progs:
zoned: refresh stale cached zone state in btrfs_reset_dev_zone()").
mkfs didn't reset the zone of a removed temporary chunk because the
cached zone state was stale.
[...]
> + /*
> + * Chunks are only removed under reclaim_bgs_lock, and their zones must
> + * not be reset before the removal is committed. A relocating repair,
> + * started by a read error during mount, may already have removed one in
> + * the running transaction, so commit it before looking for leaked zones.
> + */
> + mutex_lock(&fs_info->reclaim_bgs_lock);
> + ret = btrfs_commit_current_transaction(fs_info->tree_root);
On a filesystem without free active zones, this commit can be the one
that aborts with -EAGAIN, before the reset could free anything. In my
tests there was never a running transaction at this point, so it didn't
matter. Can you skip the commit when nothing is running, or add a
sentence to the changelog why it is safe here?
With that
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Btw, I tested v3 on top of for-next (146946479b45) and told Claude to write an
fstests [1] case for it (zloop with dm-flakey on top, power failure after direct
I/O to new zones). It fails on for-next and passes with your patch, I'll send
it to fstests once you've send a v4.
Thanks,
Johannes
[1]
diff --git a/tests/generic/815 b/tests/generic/815
new file mode 100755
index 00000000..f5969e1a
--- /dev/null
+++ b/tests/generic/815
@@ -0,0 +1,78 @@
+#! /bin/bash
+# SPDX-License-Identifier: GPL-2.0
+# Copyright (c) 2026 Western Digital Corporation. All Rights Reserved.
+#
+# FS QA Test 815
+#
+# Simulate a power failure on a zoned device after data has been written to
+# newly allocated zones, but before the allocation itself was committed. The
+# zones written that way are not referenced by the filesystem. Check that the
+# filesystem is consistent after the next read-write mount, i.e. that it
+# doesn't leave these zones written and active.
+#
+. ./common/preamble
+_begin_fstest auto quick zone log
+
+_cleanup()
+{
+ cd /
+ [ -n "$mnt" ] && _unmount $mnt 2>/dev/null
+ [ -n "$flakey" ] && _cleanup_flakey_dev $flakey
+ _destroy_zloop $zloop
+ rm -r -f $tmp.*
+}
+
+. ./common/zoned
+. ./common/dmflakey
+
+_fixed_by_fs_commit btrfs XXXXXXXXXXXX \
+ "btrfs: zoned: reset leaked zones on read-write mount"
+
+_require_scratch_size $((4 * 1024 * 1024)) # kB
+_require_block_device $SCRATCH_DEV
+_require_zloop
+_require_dm_target flakey
+_require_command "$BLKZONE_PROG" blkzone
+
+_scratch_mkfs > /dev/null 2>&1
+_scratch_mount
+
+mnt=$SCRATCH_MNT/mnt
+mkdir -p $mnt
+zloop=$(_create_zloop $SCRATCH_MNT/zloop 64 4)
+
+flakey=zflakey-test.$seq
+_init_flakey_dev $flakey $zloop
+dev=/dev/mapper/$flakey
+
+_try_mkfs_dev $dev >> $seqres.full 2>&1 || \
+ _notrun "cannot mkfs zoned filesystem"
+_mount $dev $mnt
+sync
+
+# Direct I/O puts the data on the device right away, but the allocation of
+# the zones it went to is only persisted with the next commit.
+for i in $(seq 4); do
+ $XFS_IO_PROG -f -d -c "pwrite -q -b 1m 0 96m" $mnt/file$i >> $seqres.full 2>&1
+done
+
+echo "zones before power failure:" >> $seqres.full
+$BLKZONE_PROG report $zloop | grep -cvE "\((em|nw)\)" >> $seqres.full
+
+_load_flakey_dev $flakey $zloop $FLAKEY_DROP_WRITES
+_unmount $mnt
+_load_flakey_dev $flakey $zloop $FLAKEY_ALLOW_WRITES
+
+# The read-write mount is expected to clean up after the power failure.
+_mount $dev $mnt
+_unmount $mnt
+
+echo "zones after recovery:" >> $seqres.full
+$BLKZONE_PROG report $zloop | grep -cvE "\((em|nw)\)" >> $seqres.full
+
+_check_dev_fs $dev
+
+echo "Silence is golden"
+
+# success, all done
+_exit 0
diff --git a/tests/generic/815.out b/tests/generic/815.out
new file mode 100644
index 00000000..6ea462f3
--- /dev/null
+++ b/tests/generic/815.out
@@ -0,0 +1,2 @@
+QA output created by 815
+Silence is golden
diff --git a/common/dmflakey b/common/dmflakey
index 6ac666f5..9bba8958 100644
--- a/common/dmflakey
+++ b/common/dmflakey
@@ -12,6 +12,47 @@ export FLAKEY_NAME="flakey-test.$seq"
export FLAKEY_RTNAME="flakey-rttest.$seq"
export FLAKEY_LOGNAME="flakey-logtest.$seq"
+_flakey_table()
+{
+ local size=`blockdev --getsz $1`
+
+ case "$2" in
+ "$FLAKEY_DROP_WRITES")
+ echo "0 $size flakey $1 0 0 180 1 drop_writes"
+ ;;
+ "$FLAKEY_ERROR_WRITES")
+ echo "0 $size flakey $1 0 0 180 1 error_writes"
+ ;;
+ *)
+ echo "0 $size flakey $1 0 180 0"
+ ;;
+ esac
+}
+
+_init_flakey_dev()
+{
+ _dmsetup_create $1 --table "`_flakey_table $2`" || \
+ _fatal "failed to create flakey device $1"
+}
+
+_load_flakey_dev()
+{
+ local suspend_opt="--nolockfs"
+
+ [ $# -gt 3 ] && [ $4 -eq 1 ] && suspend_opt=""
+
+ $DMSETUP_PROG suspend $suspend_opt $1 || _fatal "failed to suspend $1"
+ _flakey_table $2 $3 | $DMSETUP_PROG load $1 || \
+ _fatal "failed to load table into $1"
+ $DMSETUP_PROG resume $1 || _fatal "failed to resume $1"
+}
+
+_cleanup_flakey_dev()
+{
+ $DMSETUP_PROG resume $1 &> /dev/null
+ _dmsetup_remove $1
+}
+
_init_flakey()
{
# Scratch device
@@ -24,12 +65,10 @@ _init_flakey()
# Already set up; recreate tables
local backing_dev="$NON_FLAKEY_DEV"
fi
- local BLK_DEV_SIZE=`blockdev --getsz $backing_dev`
- FLAKEY_TABLE="0 $BLK_DEV_SIZE flakey $backing_dev 0 180 0"
- FLAKEY_TABLE_DROP="0 $BLK_DEV_SIZE flakey $backing_dev 0 0 180 1 drop_writes"
- FLAKEY_TABLE_ERROR="0 $BLK_DEV_SIZE flakey $backing_dev 0 0 180 1 error_writes"
- _dmsetup_create $FLAKEY_NAME --table "$FLAKEY_TABLE" || \
- _fatal "failed to create flakey device"
+ FLAKEY_TABLE=`_flakey_table $backing_dev`
+ FLAKEY_TABLE_DROP=`_flakey_table $backing_dev $FLAKEY_DROP_WRITES`
+ FLAKEY_TABLE_ERROR=`_flakey_table $backing_dev $FLAKEY_ERROR_WRITES`
+ _init_flakey_dev $FLAKEY_NAME $backing_dev
# Realtime device
if [ -n "$SCRATCH_RTDEV" ]; then
@@ -42,12 +81,10 @@ _init_flakey()
# Already set up; recreate tables
local backing_dev="$NON_FLAKEY_RTDEV"
fi
- local BLK_DEV_SIZE=`blockdev --getsz $backing_dev`
- FLAKEY_RTTABLE="0 $BLK_DEV_SIZE flakey $backing_dev 0 180 0"
- FLAKEY_RTTABLE_DROP="0 $BLK_DEV_SIZE flakey $backing_dev 0 0 180 1 drop_writes"
- FLAKEY_RTTABLE_ERROR="0 $BLK_DEV_SIZE flakey $backing_dev 0 0 180 1 error_writes"
- _dmsetup_create $FLAKEY_RTNAME --table "$FLAKEY_RTTABLE" || \
- _fatal "failed to create flakey rt device"
+ FLAKEY_RTTABLE=`_flakey_table $backing_dev`
+ FLAKEY_RTTABLE_DROP=`_flakey_table $backing_dev $FLAKEY_DROP_WRITES`
+ FLAKEY_RTTABLE_ERROR=`_flakey_table $backing_dev $FLAKEY_ERROR_WRITES`
+ _init_flakey_dev $FLAKEY_RTNAME $backing_dev
fi
# External log device
@@ -61,12 +98,10 @@ _init_flakey()
# Already set up; recreate tables
local backing_dev="$NON_FLAKEY_LOGDEV"
fi
- local BLK_DEV_SIZE=`blockdev --getsz $backing_dev`
- FLAKEY_LOGTABLE="0 $BLK_DEV_SIZE flakey $backing_dev 0 180 0"
- FLAKEY_LOGTABLE_DROP="0 $BLK_DEV_SIZE flakey $backing_dev 0 0 180 1 drop_writes"
- FLAKEY_LOGTABLE_ERROR="0 $BLK_DEV_SIZE flakey $backing_dev 0 0 180 1 error_writes"
- _dmsetup_create $FLAKEY_LOGNAME --table "$FLAKEY_LOGTABLE" || \
- _fatal "failed to create flakey log device"
+ FLAKEY_LOGTABLE=`_flakey_table $backing_dev`
+ FLAKEY_LOGTABLE_DROP=`_flakey_table $backing_dev $FLAKEY_DROP_WRITES`
+ FLAKEY_LOGTABLE_ERROR=`_flakey_table $backing_dev $FLAKEY_ERROR_WRITES`
+ _init_flakey_dev $FLAKEY_LOGNAME $backing_dev
fi
}
@@ -80,9 +115,9 @@ _cleanup_flakey()
_unmount $SCRATCH_MNT > /dev/null 2>&1
- test -n "$NON_FLAKEY_DEV" && _dmsetup_remove $FLAKEY_NAME
- test -n "$NON_FLAKEY_LOGDEV" && _dmsetup_remove $FLAKEY_LOGNAME
- test -n "$NON_FLAKEY_RTDEV" && _dmsetup_remove $FLAKEY_RTNAME
+ test -n "$NON_FLAKEY_DEV" && _cleanup_flakey_dev $FLAKEY_NAME
+ test -n "$NON_FLAKEY_LOGDEV" && _cleanup_flakey_dev $FLAKEY_LOGNAME
+ test -n "$NON_FLAKEY_RTDEV" && _cleanup_flakey_dev $FLAKEY_RTNAME
SCRATCH_DEV="$NON_FLAKEY_DEV"
unset NON_FLAKEY_DEV
prev parent reply other threads:[~2026-11-17 11:51 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-11-02 15:35 [PATCH v3] btrfs: zoned: reset leaked zones on read-write mount Mark Korenberg
2026-11-02 15:35 ` Mark Korenberg via B4 Relay
2026-11-17 6:54 ` Mark Korenberg
2026-11-17 11:46 ` Johannes Thumshirn [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=avw98epZjlOLcu8g@mayhem.fritz.box \
--to=johannes.thumshirn@wdc.com \
--cc=dsterba@suse.com \
--cc=linux-btrfs@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mason@kernel.org \
--cc=naohiro.aota@wdc.com \
--cc=socketpair@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.