On Tue, Jan 23, 2024 at 3:37 PM Boris Burkov <boris@xxxxxx> wrote:
>
> This test was reproducing a bug triggered by creating a subvolume qgroup
> before creating the subvolume itself with a snapshot.
>
> The kernel patch:
> btrfs: forbid creating subvol qgroups
>
> explicitly prevents that and makes it fail with EINVAL. I could "fix"
> this test by expecting the EINVAL message in the output, but at that
> point it would simply be a test that creating a subvolume and
> snapshotting it works with qgroups, which is adequately tested by other
> tests which focus on accurately measuring shared/exclusive usage in
> various snapshot/reflink scenarios. To avoid confusion, I think it is
> best to simply delete this test.
>
> Signed-off-by: Boris Burkov
> ---
> tests/btrfs/303 | 77 ---------------------------------------------
> tests/btrfs/303.out | 2 --
> 2 files changed, 79 deletions(-)
> delete mode 100755 tests/btrfs/303
> delete mode 100644 tests/btrfs/303.out
>
> diff --git a/tests/btrfs/303 b/tests/btrfs/303
> deleted file mode 100755
> index 410460af5..000000000
> --- a/tests/btrfs/303
> +++ /dev/null
> @@ -1,77 +0,0 @@
> -#! /bin/bash
> -# SPDX-License-Identifier: GPL-2.0
> -# Copyright (C) 2023 SUSE Linux Products GmbH. All Rights Reserved.
> -#
> -# FS QA Test 303
> -#
> -# A regression test to make sure snapshot creation won't cause transaction
> -# abort if there is already an existing qgroup.
> -#
> -. ./common/preamble
> -_begin_fstest auto quick snapshot subvol qgroup
> -
> -. ./common/filter
> -
> -_supported_fs btrfs
> -_require_scratch
> -
> -_fixed_by_kernel_commit xxxxxxxxxxxx \
> - "btrfs: do not abort transaction if there is already an existing qgroup"
> -
> -_scratch_mkfs >> $seqres.full 2>&1 || _fail "mkfs failed"
> -_scratch_mount
> -
> -# Create the first subvolume and get its id.
> -# This subvolume id should not change no matter if there is an existing
> -# qgroup for it.
> -$BTRFS_UTIL_PROG subvolume create "$SCRATCH_MNT/subvol" >> $seqres.full
> -$BTRFS_UTIL_PROG subvolume snapshot "$SCRATCH_MNT/subvol" \
> - "$SCRATCH_MNT/snapshot">> $seqres.full
> -
> -init_subvolid=$(_btrfs_get_subvolid "$SCRATCH_MNT" "snapshot")
> -
> -if [ -z "$init_subvolid" ]; then
> - _fail "Unable to get the subvolid of the first snapshot"
> -fi
> -
> -echo "Subvolumeid: ${init_subvolid}" >> $seqres.full
> -
> -_scratch_unmount
> -
> -# Re-create the fs, as btrfs won't reuse the subvolume id.
> -_scratch_mkfs >> $seqres.full 2>&1 || _fail "2nd mkfs failed"
> -_scratch_mount
> -
> -$BTRFS_UTIL_PROG quota enable "$SCRATCH_MNT" >> $seqres.full
> -_qgroup_rescan $SCRATCH_MNT >> $seqres.full
> -
> -# Create a qgroup for the first subvolume, this would make the later
> -# subvolume creation to find an existing qgroup, and abort transaction.
> -$BTRFS_UTIL_PROG qgroup create 0/"$init_subvolid" "$SCRATCH_MNT" >> $seqres.full
> -
> -# Now create the first snapshot, which should have the same subvolid no matter
> -# if the quota is enabled.
> -$BTRFS_UTIL_PROG subvolume create "$SCRATCH_MNT/subvol" >> $seqres.full
> -$BTRFS_UTIL_PROG subvolume snapshot "$SCRATCH_MNT/subvol" \
> - "$SCRATCH_MNT/snapshot" >> $seqres.full
> -
> -# Either the snapshot create failed and transaction is aborted thus no
> -# snapshot here, or we should be able to create the snapshot.
> -new_subvolid=$(_btrfs_get_subvolid "$SCRATCH_MNT" "snapshot")
> -
> -echo "Subvolumeid: ${new_subvolid}" >> $seqres.full
> -
> -if [ -z "$new_subvolid" ]; then
> - _fail "Unable to get the subvolid of the first snapshot"
> -fi
> -
> -# Make sure the subvolumeid for the first snapshot didn't change.
> -if [ "$new_subvolid" -ne "$init_subvolid" ]; then
> - _fail "Subvolumeid for the first snapshot changed, has ${new_subvolid} expect ${init_subvolid}"
> -fi
> -
> -echo "Silence is golden"
> -
> -# success, all done
> -status=0
> -exit
> diff --git a/tests/btrfs/303.out b/tests/btrfs/303.out
> deleted file mode 100644
> index d48808e60..000000000
> --- a/tests/btrfs/303.out
> +++ /dev/null
> @@ -1,2 +0,0 @@
> -QA output created by 303
> -Silence is golden
> --
> 2.43.0
>
>
Seems reasonable.
Reviewed-by: Neal Gompa <neal@xxxxxxxxx>
--
真実はいつも一つ！/ Always, there's only one truth!
