Re: [PATCH 12/12] xfs/650: require a real SCRATCH_RTDEV

["Darrick J. Wong" <djwong@kernel.org>]

On Wed, Dec 10, 2025 at 11:45:49AM -0800, Darrick J. Wong wrote:
> On Wed, Dec 10, 2025 at 06:46:58AM +0100, Christoph Hellwig wrote:
> > Require a real SCRATCH_RTDEV instead of faking one up using a loop
> > device, as otherwise the options specified in MKFS_OPTIONS might
> > not actually work the configuration.
> > 
> > Signed-off-by: Christoph Hellwig <hch@lst.de>
> > ---
> >  tests/xfs/650 | 63 +++++----------------------------------------------
> 
> Er... what test is xfs/650?  There isn't one in for-next.

Aha, I missed that an earlier patch created it. :(

The one downside to not injecting a rt device here is that now the only
testing for the actual bug is if you happen to have rt enabled.  That
used to be a concern of mine, but maybe between you, me, Meta, and the
kdevops folks there's enough now.

Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>

--D

> 
> >  1 file changed, 6 insertions(+), 57 deletions(-)
> > 
> > diff --git a/tests/xfs/650 b/tests/xfs/650
> > index d8f70539665f..418a1e7aae7c 100755
> > --- a/tests/xfs/650
> > +++ b/tests/xfs/650
> > @@ -9,21 +9,11 @@
> >  # bunmapi"). On XFS without the fixes, truncate will hang forever.
> >  #
> >  . ./common/preamble
> > -_begin_fstest auto prealloc preallocrw
> > -
> > -# Override the default cleanup function.
> > -_cleanup()
> > -{
> > -	_scratch_unmount &>/dev/null
> > -	[ -n "$loop_dev" ] && _destroy_loop_device $loop_dev
> > -	cd /
> > -	rm -f $tmp.*
> > -	rm -f "$TEST_DIR/$seq"
> > -}
> > +_begin_fstest auto prealloc preallocrw realtime
> >  
> >  . ./common/filter
> >  
> > -_require_scratch_nocheck
> > +_require_realtime
> >  _require_xfs_io_command "falloc"
> >  
> >  maxextlen=$((0x1fffff))
> > @@ -31,51 +21,11 @@ bs=4096
> >  rextsize=4
> >  filesz=$(((maxextlen + 1) * bs))
> >  
> > -must_disable_feature() {
> > -	local feat="$1"
> > -
> > -	# If mkfs doesn't know about the feature, we don't need to disable it
> > -	$MKFS_XFS_PROG --help 2>&1 | grep -q "${feat}=0" || return 1
> > -
> > -	# If turning the feature on works, we don't need to disable it
> > -	_scratch_mkfs_xfs_supported -m "${feat}=1" "${disabled_features[@]}" \
> > -		> /dev/null 2>&1 && return 1
> > -
> > -	# Otherwise mkfs knows of the feature and formatting with it failed,
> > -	# so we do need to mask it.
> > -	return 0
> > -}
> > -
> > -extra_options=""
> > -# Set up the realtime device to reproduce the bug.
> > +_scratch_mkfs \
> > +	-d agsize=$(((maxextlen + 1) * bs / 2)),rtinherit=1 \
> > +	-r extsize=$((bs * rextsize)) \
> > +	>>$seqres.full 2>&1
> >  
> > -# If we don't have a realtime device, set up a loop device on the test
> > -# filesystem.
> > -if [[ $USE_EXTERNAL != yes || -z $SCRATCH_RTDEV ]]; then
> > -	_require_test
> > -	loopsz="$((filesz + (1 << 26)))"
> > -	_require_fs_space "$TEST_DIR" $((loopsz / 1024))
> > -	$XFS_IO_PROG -c "truncate $loopsz" -f "$TEST_DIR/$seq"
> > -	loop_dev="$(_create_loop_device "$TEST_DIR/$seq")"
> > -	USE_EXTERNAL=yes
> > -	SCRATCH_RTDEV="$loop_dev"
> > -	disabled_features=()
> > -
> > -	# disable reflink if not supported by realtime devices
> > -	must_disable_feature reflink &&
> > -		disabled_features=(-m reflink=0)
> > -
> > -	# disable rmap if not supported by realtime devices
> > -	must_disable_feature rmapbt &&
> > -		disabled_features+=(-m rmapbt=0)
> > -fi
> > -extra_options="$extra_options -r extsize=$((bs * rextsize))"
> > -extra_options="$extra_options -d agsize=$(((maxextlen + 1) * bs / 2)),rtinherit=1"
> > -
> > -_scratch_mkfs $extra_options "${disabled_features[@]}" >>$seqres.full 2>&1
> > -_try_scratch_mount >>$seqres.full 2>&1 || \
> > -	_notrun "mount failed, kernel doesn't support realtime?"
> > -_scratch_unmount
> >  _scratch_mount
> >  _require_fs_space "$SCRATCH_MNT" $((filesz / 1024))
> >  
> > @@ -108,7 +58,6 @@ $XFS_IO_PROG -c "pwrite -b 1M -W 0 $(((maxextlen + 2 - rextsize) * bs))" \
> >  # Truncate the extents.
> >  $XFS_IO_PROG -c "truncate 0" -c fsync "$SCRATCH_MNT/file"
> >  
> > -# We need to do this before the loop device gets torn down.
> >  _scratch_unmount
> >  _check_scratch_fs
> >  
> > -- 
> > 2.47.3
> > 
> > 
>