Re: [PATCH] btrfs: test creating a large number of snapshots of a received subvolume

[Boris Burkov <boris@bur.io>]

On Thu, Feb 26, 2026 at 02:34:24PM +0000, fdmanana@kernel.org wrote:
> From: Filipe Manana <fdmanana@suse.com>
> 
> Test that we can create a large number of snapshots of a received
> subvolume without triggering a transaction abort due to leaf item
> overflow. Also check that we are able to delete the snapshots and use
> the last one for an incremental send/receive despite an item overflow
> when updating the uuid tree to insert a BTRFS_UUID_KEY_RECEIVED_SUBVOL
> item.
> 
> This exercises a bug fixed by the following kernel patch:
> 
>   "btrfs: fix transaction abort when snapshotting received subvolumes"
> 
> Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Boris Burkov <boris@bur.io>
> ---
>  tests/btrfs/345     | 73 +++++++++++++++++++++++++++++++++++++++++++++
>  tests/btrfs/345.out |  2 ++
>  2 files changed, 75 insertions(+)
>  create mode 100755 tests/btrfs/345
>  create mode 100644 tests/btrfs/345.out
> 
> diff --git a/tests/btrfs/345 b/tests/btrfs/345
> new file mode 100755
> index 00000000..75cc3067
> --- /dev/null
> +++ b/tests/btrfs/345
> @@ -0,0 +1,73 @@
> +#! /bin/bash
> +# SPDX-License-Identifier: GPL-2.0
> +# Copyright (c) 2026 SUSE S.A.  All Rights Reserved.
> +#
> +# FS QA Test 345
> +#
> +# Test that we can create a large number of snapshots of a received subvolume
> +# without triggering a transaction abort due to leaf item overflow. Also check
> +# that we are able to delete the snapshots and use the last one for an
> +# incremental send/receive despite an item overflow when updating the uuid tree
> +# to insert a BTRFS_UUID_KEY_RECEIVED_SUBVOL item.
> +#
> +. ./common/preamble
> +_begin_fstest auto quick subvol send snapshot
> +
> +_require_scratch
> +_require_btrfs_support_sectorsize 4096
> +_require_btrfs_command "property"
> +
> +_fixed_by_kernel_commit xxxxxxxxxxxx \
> +	"btrfs: fix transaction abort when snapshotting received subvolumes"
> +
> +# Use a 4K node/leaf size to make the test faster.
> +_scratch_mkfs -n 4K >> $seqres.full 2>&1 || _fail "mkfs failed"
> +_scratch_mount
> +
> +# Create a subvolume.
> +_btrfs subvolume create $SCRATCH_MNT/sv
> +touch $SCRATCH_MNT/sv/foo
> +
> +# Turn the subvolume to RO mode so that we can send it.
> +_btrfs property set $SCRATCH_MNT/sv ro true
> +
> +# Send the subvolume and receive it. Our received version of the subvolume
> +# (in $SCRATCH_MNT/snaps/sv) will have a non-NULL received UUID field in its
> +# root item.
> +mkdir $SCRATCH_MNT/snaps
> +_btrfs send -f $SCRATCH_MNT/send.stream $SCRATCH_MNT/sv
> +_btrfs receive -f $SCRATCH_MNT/send.stream $SCRATCH_MNT/snaps
> +
> +# Now snapshot the received the subvolume a lot of times and check we are able
> +# to snapshot and that we don't trigger a transaction abort (will trigger a
> +# warning and dump a stack trace in dmesg/syslog).
> +total=$(( 1000 * LOAD_FACTOR ))
> +for ((i = 1; i <= $total; i++)); do
> +    _btrfs subvolume snapshot -r $SCRATCH_MNT/snaps/sv $SCRATCH_MNT/snaps/sv_$i
> +done
> +
> +# Create a snapshot based on the last snapshot, add a file to it, and turn it
> +# to RO so that we can used it for a send operation.
> +last_snap="${SCRATCH_MNT}/snaps/sv_${total}"
> +_btrfs subvolume snapshot $last_snap $SCRATCH_MNT/snaps/sv_last_as_parent
> +echo -n "hello world" > $SCRATCH_MNT/snaps/sv_last_as_parent/bar
> +_btrfs property set $SCRATCH_MNT/snaps/sv_last_as_parent ro true
> +
> +# Now we attempt to send and receive that snapshot, verify that it works even
> +# if the creation of the snapshot $last_snap was not able to add a
> +# BTRFS_UUID_KEY_RECEIVED_SUBVOL item to the uuid tree due to leaf overflow.
> +_btrfs send -f $SCRATCH_MNT/inc_send.stream -p $last_snap \
> +       $SCRATCH_MNT/snaps/sv_last_as_parent
> +_btrfs receive -f $SCRATCH_MNT/inc_send.stream $SCRATCH_MNT/
> +
> +# Verify the received snapshot has the new file with the right content.
> +diff $SCRATCH_MNT/snaps/sv_last_as_parent/bar $SCRATCH_MNT/sv_last_as_parent/bar
> +
> +# Now check that we are able to delete all the snapshots too.
> +for ((i = 1; i <= $total; i++)); do
> +    _btrfs subvolume delete $SCRATCH_MNT/snaps/sv_$i
> +done
> +
> +# success, all done
> +echo "Silence is golden"
> +_exit 0
> diff --git a/tests/btrfs/345.out b/tests/btrfs/345.out
> new file mode 100644
> index 00000000..11ef4e7e
> --- /dev/null
> +++ b/tests/btrfs/345.out
> @@ -0,0 +1,2 @@
> +QA output created by 345
> +Silence is golden
> -- 
> 2.47.2
>