Re: [PATCH 3/4] xfs: detect time limits from filesystem

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Thu, Oct 29, 2020 at 12:47:32PM +0200, Amir Goldstein wrote:
> On Wed, Oct 28, 2020 at 10:24 PM Darrick J. Wong
> <darrick.wong@oracle.com> wrote:
> >
> > From: Darrick J. Wong <darrick.wong@oracle.com>
> >
> > Teach fstests to extract timestamp limits of a filesystem using the new
> > xfs_db timelimit command.
> >
> > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> > ---
> >  common/rc         |    2 +-
> >  common/xfs        |   14 ++++++++++++++
> >  tests/xfs/911     |   44 ++++++++++++++++++++++++++++++++++++++++++++
> >  tests/xfs/911.out |   15 +++++++++++++++
> >  tests/xfs/group   |    1 +
> >  5 files changed, 75 insertions(+), 1 deletion(-)
> >  create mode 100755 tests/xfs/911
> >  create mode 100644 tests/xfs/911.out
> >
> >
> > diff --git a/common/rc b/common/rc
> > index 41f93047..162d957a 100644
> > --- a/common/rc
> > +++ b/common/rc
> > @@ -2029,7 +2029,7 @@ _filesystem_timestamp_range()
> >                 echo "0 $u32max"
> >                 ;;
> >         xfs)
> > -               echo "$s32min $s32max"
> > +               _xfs_timestamp_range "$device"
> >                 ;;
> >         btrfs)
> >                 echo "$s64min $s64max"
> > diff --git a/common/xfs b/common/xfs
> > index e548a0a1..19ccee03 100644
> > --- a/common/xfs
> > +++ b/common/xfs
> > @@ -994,3 +994,17 @@ _require_xfs_scratch_inobtcount()
> >                 _notrun "inobtcount not supported by scratch filesystem type: $FSTYP"
> >         _scratch_unmount
> >  }
> > +
> > +_xfs_timestamp_range()
> > +{
> > +       local use_db=0
> > +       local dbprog="$XFS_DB_PROG $device"

Heh, device isn't defined, I'll fix that.

> > +       test "$device" = "$SCRATCH_DEV" && dbprog=_scratch_xfs_db
> > +
> > +       $dbprog -f -c 'help timelimit' | grep -v -q 'not found' && use_db=1
> > +       if [ $use_db -eq 0 ]; then
> > +               echo "-$((1<<31)) $(((1<<31)-1))"
> 
> This embodies an assumption that the tested filesystem does not have
> bigtime enabled if xfs_db tool is not uptodate.

If the xfs_db tool doesn't support the timelimit command then it doesn't
support formatting with bigtime.  I don't think it's reasonable to
expect to be able to run fstests on a test filesystem that xfsprogs
doesn't support.  Hence it's fine to output the old limits if the
timelimit command doesn't exist.

> Maybe it makes sense, but it may be safer to return "-1 -1" and not_run
> generic/402 if xfs_db is not uptodate, perhaps with an extra message
> hinting the user to upgrade xfs_db.

TBH it boggles my mind that there *still* is no way to ask the kernel
for the supported timestamp range of a mounted filesystem.  The
timelimit command and this mess in fstests was supposed to be a
temporary workaround that would (in my ideal world) have become
unnecessary before this landed, but ... ugh.

--D

> Thanks,
> Amir.