From: Filipe Manana <fdmanana@xxxxxxxx> Test case btrfs/228 creates a subvolume and then calls the dump-tree command from btrfs-progs. The tree dumping accesses the device directly and therefore can only see committed metadata - this used to work because subvolume creation used to commit the transaction that was used to create the subvolume, however it is no longer the case after a recent patch that currently is only on the btrfs integration branch "misc-next". That patch has the following subject: "btrfs: don't commit transaction for every subvol create" So explicitly sync the filesystem before calling the dump-tree command, commenting why we do it. This way the test works before and after that patch, for any kernel release. Signed-off-by: Filipe Manana <fdmanana@xxxxxxxx> --- tests/btrfs/228 | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/tests/btrfs/228 b/tests/btrfs/228 index fde623fc..5ef1dfd7 100755 --- a/tests/btrfs/228 +++ b/tests/btrfs/228 @@ -28,6 +28,11 @@ _scratch_mount $BTRFS_UTIL_PROG subvolume create $SCRATCH_MNT/newvol >> $seqres.full 2>&1 \ || _fail "couldn't create subvol" +# Subvolume creation used to commit the transaction used to create it, but after +# the patch "btrfs: don't commit transaction for every subvol create", that +# changed, so sync the fs to commit any open transaction. +$BTRFS_UTIL_PROG filesystem sync $SCRATCH_MNT + $BTRFS_UTIL_PROG inspect-internal dump-tree -t1 $SCRATCH_DEV \ | grep -q "256 ROOT_ITEM" || _fail "First subvol with id 256 doesn't exist" -- 2.35.1
