Re: [PATCH] qemu-{img,nbd}: Don't report zeroed cluster as a hole

[Eric Blake <eblake@redhat.com>]

On Sat, Jun 12, 2021 at 12:23:06AM +0300, Nir Soffer wrote:
> > Otherwise, you do have a point: "depth":1 in isolation is ambiguous
> > between "not allocated anywhere in this 1-element chain" and
> > "allocated at the first backing file in this chain of length 2 or
> > more".  At which point you can indeed use "qemu-img info" to determine
> > the backing chain depth.  How painful is that extra step?  Does it
> > justify the addition of a new optional "backing":true to any portion
> > of the file that was beyond the end of the chain (and omit that line
> > for all other regions, rather than printing "backing":false)?
> 
> Dealing with depth: N + 1 is not that painful, but also not great.
> 
> I think it is worth a little more effort, and it will save time in the long term
> for users and for developers. Better APIs need simpler and shorter
> documentation and require less support.
> 
> I'm not sure about backing: false, maybe absent: true to match libnbd?

In the patch [1], I did "backing":true if the cluster was not found in
the chain, and omitted the bool altogether when the cluster is
present.  If we like the name "absent":true better than
"backing":true, that's an easy change.  The libnbd change for nbdinfo
to report 'absent' instead of 'unallocated' has not yet been released,
so we have some leeway on naming choices.

[1] https://lists.gnu.org/archive/html/qemu-devel/2021-06/msg03067.html

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3266
Virtualization:  qemu.org | libvirt.org