On Wed, Feb 21, 2024 at 07:35:24AM +0100, Christoph Hellwig wrote: > generic/449 tests of xattr behavior when we run out of disk space for > xattrs which are on the main device, but _scratch_mkfs_sized will control > the size of the RT device. > > Skip it when a RT device is in used, as otherwise it won't test what it > is supposed to while taking a long time to fill the unrestricted data > device. > > Signed-off-by: Christoph Hellwig <hch@xxxxxx> > --- > tests/generic/449 | 11 ++++++----- > 1 file changed, 6 insertions(+), 5 deletions(-) > > diff --git a/tests/generic/449 b/tests/generic/449 > index 2b77a6a49..4269703f6 100755 > --- a/tests/generic/449 > +++ b/tests/generic/449 > @@ -24,14 +24,15 @@ _require_test > _require_acls > _require_attrs trusted > > +# This is a test of xattr behavior when we run out of disk space for xattrs, > +# but _scratch_mkfs_sized will control the size of the RT device. Skip it > +# when a RT device is in used, as otherwise it won't test what it is supposed > +# to while taking a long time to fill the unrestricted data device > +_require_no_realtime > + > _scratch_mkfs_sized $((256 * 1024 * 1024)) >> $seqres.full 2>&1 Odd... this test only takes ~50s on my rt testing rig. _scratch_mkfs_sized should restrict the size of both the data device and the rt volume to 256M, right? Looking at tot, it sets "-d size=$fssize" and "-r size=$fssize", so I don't think I understand what's going on here. > _scratch_mount || _fail "mount failed" > > -# This is a test of xattr behavior when we run out of disk space for xattrs, > -# so make sure the pwrite goes to the data device and not the rt volume. > -test "$FSTYP" = "xfs" && \ > - _xfs_force_bdev data $SCRATCH_MNT Shouldn't this ^^^^ be sufficient to cause $TFILE and all the xattrs to be allocated from the data device? Or are they ending up on the rt volume? <confused> --D > - > TFILE=$SCRATCH_MNT/testfile.$seq > > # Create the test file and choose its permissions > -- > 2.39.2 > >