On Mon, Jan 29, 2018 at 07:06:23PM +0200, Amir Goldstein wrote:
> On Mon, Jan 29, 2018 at 4:33 PM, Eryu Guan <eguan@xxxxxxxxxx> wrote:
> > On Sun, Jan 28, 2018 at 11:22:32AM +0200, Amir Goldstein wrote:
> >> Some overlayfs features must be checked together, because they cannot
> >> be enabled without a dependent feature (e.g. nfs_export and index).
> >>
> >> Signed-off-by: Amir Goldstein <amir73il@xxxxxxxxx>
> >> ---
> >> common/overlay | 23 ++++++++++++++++++++---
> >> common/rc | 15 +++++++++------
> >> 2 files changed, 29 insertions(+), 9 deletions(-)
> >>
> >> diff --git a/common/overlay b/common/overlay
> >> index 1da4ab1..38fe994 100644
> >> --- a/common/overlay
> >> +++ b/common/overlay
> >> @@ -138,9 +138,6 @@ _require_scratch_overlay_feature()
> >> [ "$default" = Y ] || [ "$default" = N ] || \
> >> _notrun "feature '${feature}' not supported by ${FSTYP}"
> >>
> >> - _scratch_mkfs > /dev/null 2>&1
> >> - _scratch_mount -o ${feature}=on || \
> >> - _notrun "${FSTYP} feature '${feature}' cannot be enabled on ${SCRATCH_DEV}"
> >
> > So _require_scratch_overlay_feature can't be used by any test directly
> > now and should only be called by _require_scratch_overlay_features.
> >
> > IMHO, the name _require_scratch_overlay_feature becomes misleading after
> > this change, it implies it can be called by tests directly. I'd rename
> > it to a more proper name (which I can't think of any..) with double
> > underscores at the beginning to imply it's kind of "static", e.g. the
> > internal helpers like "__populate_create_dir" in common/populate.
>
> __check_one_scratch_overlay_feature
"one" seems redundant, given the ending "feature".
>
> >
> > Another way to go is just open-code this in
> > _require_scratch_overlay_features().
> >
> >> # Check options to be sure. For example, Overlayfs will fallback to
> >> # index=off if underlying fs does not support file handles.
> >> # Overlayfs only displays mount option if it differs from the default.
> >> @@ -149,5 +146,25 @@ _require_scratch_overlay_feature()
> >> ( [ "$default" = Y ] && ! _fs_options $SCRATCH_DEV | grep -q "${feature}=off" )) && \
> >> touch $SCRATCH_MNT/foo 2>/dev/null ) || \
> >> _notrun "${FSTYP} feature '${feature}' cannot be enabled on ${SCRATCH_DEV}"
> >> +}
> >> +
> >> +# Require a set of overlayfs features
> >> +_require_scratch_overlay_features()
> >> +{
> >> + local features=( $* )
> >> + local opts="rw"
> >> +
> >> + for feature in ${features[*]}; do
> >> + opts+=",${feature}=on"
> >> + done
> >> +
> >> + _scratch_mkfs > /dev/null 2>&1
> >> + _scratch_mount -o $opts || \
> >> + _notrun "overlay options '$opts' cannot be enabled on ${SCRATCH_DEV}"
> >> +
> >> + for feature in ${features[*]}; do
> >> + _require_scratch_overlay_feature ${feature}
> >> + done
> >> +
> >> _scratch_unmount
> >> }
> >> diff --git a/common/rc b/common/rc
> >> index 77a4eb4..3b81061 100644
> >> --- a/common/rc
> >> +++ b/common/rc
> >> @@ -3658,22 +3658,25 @@ _get_fs_sysfs_attr()
> >> cat /sys/fs/${FSTYP}/${dname}/${attr}
> >> }
> >>
> >> -# Generic test for specific filesystem feature.
> >> +# Generic test for specific filesystem features.
> >> # Currently only implemented to test overlayfs features.
> >> -_require_scratch_feature()
> >> +_require_scratch_features()
> >> {
> >> - local feature=$1
> >> -
> >> case "$FSTYP" in
> >> overlay)
> >> - _require_scratch_overlay_feature ${feature}
> >> + _require_scratch_overlay_features $*
> >> ;;
> >> *)
> >> - _fail "Test for feature '${feature}' of ${FSTYP} is not implemented"
> >> + _fail "Test for features '$*' of ${FSTYP} is not implemented"
> >> ;;
> >> esac
> >> }
> >>
> >> +_require_scratch_feature()
> >> +{
> >> + _require_scratch_features $1
> >> +}
> >> +
> >
> > Hmm, I don't think this is necessary. Just drop all the plural variants
> > and teach _require_scratch_overlay_feature to accept & test multiple
> > features, and add some good comments.
> >
>
> I see what you are saying, but I rather call a function called
> _require_scratch_overlay_features from the tests that require plural features
> I can leave _require_scratch_feature for the singular feature case though,
> until another filesystem needs the plural variant.
Yeah, I can live with that :)
Thanks,
Eryu
--
To unsubscribe from this list: send the line "unsubscribe linux-unionfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
