On Fri, Aug 10, 2018 at 03:00:27PM -0700, Andrei Vagin wrote: > From: Andrei Vagin <avagin@xxxxxxxxx> > > Currently, a reconfigure call-back is implemented only for the proc file > system. This test creates a new mount and pid namespace and then we can > create a new proc mount instance and be sure that a host configuration > will not be affected. > > Signed-off-by: Andrei Vagin <avagin@xxxxxxxxx> > --- > tools/testing/selftests/fsopen/Makefile | 5 + > tools/testing/selftests/fsopen/config | 2 + > tools/testing/selftests/fsopen/fsopen.c | 120 ++++++++++++++++++++++++ > 3 files changed, 127 insertions(+) > create mode 100644 tools/testing/selftests/fsopen/Makefile > create mode 100644 tools/testing/selftests/fsopen/config > create mode 100644 tools/testing/selftests/fsopen/fsopen.c can we get all this test stuff put into fstests? And then add a config option to the fstests mount commands so that they use the internal mount command to exercise the new API rather than use the existing mount binary? That way every filesystem developer /will/ exercise the new mount API in lots of interesting and unusual ways that are expected to work (or not work, as fstests exercises failure cases, too) in their day-to-day testing and development. This will flush out the bugs and problems much, much faster than a small selftest in the kernel tree that doesn't really interact with the bulk of the code that runs when filesystems are mounted/remounted. Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
