Re: [PATCH RFC fstests] generic: add a test for multigrain timestamps

[Christoph Hellwig <hch@infradead.org>]

On Wed, Oct 28, 2026 at 01:20:24PM -0400, Jeff Layton wrote:
> On Wed, 2026-10-28 at 08:54 -0700, Christoph Hellwig wrote:
> > On Tue, Oct 27, 2026 at 09:58:50AM -0400, Jeff Layton wrote:
> > > Ensure that the mtime and ctime apparently change, even when there are
> > > multiple writes in quick succession. Since this is opt-in for in-kernel
> > > filesystems and there is no way to test for this feature in userland,
> > > limit it to just ext4, f2fs, btrfs, tmpfs and xfs for now.
> > > 
> > > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > > ---
> > > I'd like to start testing this to make sure it doesn't break, but it's
> > > not universally supported on all filesystems. Is an explicit allowlist
> > > like this the right approach?
> > 
> > Can we please expose the fact that multi-grain timestamp are supported
> > from the kernel?  This seems like it would be just one bit we'd need
> > in sttx or something similar.
> 
> I'd been resisting that, but maybe it's a good idea. I'll see what I
> can come up with. We can probably just add a flag to stx_attributes?

Yes, ever since STATX_ATTR_DAX/STATX_ATTR_WRITE_ATOMIC these don't
seem to be tied to actual inode flags any more.