QEMU’s always been confused over what a snapshot is: Is it the overlay?
Is it the backing image?

Confusion is rarely a good thing.  I can’t think of any objective reason
why the overlay would be a snapshot.  A snapshot is something that does
not change over time; the overlay does.

(I suppose historically the reason is that “Taking an overlay” makes no
sense, so the operations are called “Taking a snapshot”.  Somehow, this
meaning carried over to the new file that is created during that
operation; if “Creating a snapshot” creates a file, that file must be
the snapshot, right?  Well, no, it isn’t.)

Let’s fix this as best as we can.  Better Nate than lever.


v2:
- Don’t break the iotests for a change
  (kept Eric’s R-b, because it felt like the right thing to do)


git backport-diff against v1:

Key:
[----] : patches are identical
[####] : number of functional differences between upstream/downstream patch
[down] : patch is downstream-only
The flags [FC] indicate (F)unctional and (C)ontextual differences, respectively

001/2:[----] [--] 'qapi/block-core: Overlays are not snapshots'
002/2:[0010] [FC] 'blockdev: Overlays are not snapshots'


Max Reitz (2):
  qapi/block-core: Overlays are not snapshots
  blockdev: Overlays are not snapshots

 qapi/block-core.json       | 20 ++++++++++----------
 blockdev.c                 | 10 +++++-----
 tests/qemu-iotests/085.out | 10 +++++-----
 3 files changed, 20 insertions(+), 20 deletions(-)

-- 
2.21.0

On Mon 03 Jun 2019 10:22:34 PM CEST, Max Reitz wrote:
> QEMU’s always been confused over what a snapshot is: Is it the overlay?
> Is it the backing image?
>
> Confusion is rarely a good thing.  I can’t think of any objective reason
> why the overlay would be a snapshot.  A snapshot is something that does
> not change over time; the overlay does.
>
> (I suppose historically the reason is that “Taking an overlay” makes no
> sense, so the operations are called “Taking a snapshot”.  Somehow, this
> meaning carried over to the new file that is created during that
> operation; if “Creating a snapshot” creates a file, that file must be
> the snapshot, right?  Well, no, it isn’t.)
>
> Let’s fix this as best as we can.  Better Nate than lever.

Reviewed-by: Alberto Garcia <berto@igalia.com>

Berto

On 03.06.19 22:22, Max Reitz wrote:
> QEMU’s always been confused over what a snapshot is: Is it the overlay?
> Is it the backing image?
> 
> Confusion is rarely a good thing.  I can’t think of any objective reason
> why the overlay would be a snapshot.  A snapshot is something that does
> not change over time; the overlay does.
> 
> (I suppose historically the reason is that “Taking an overlay” makes no
> sense, so the operations are called “Taking a snapshot”.  Somehow, this
> meaning carried over to the new file that is created during that
> operation; if “Creating a snapshot” creates a file, that file must be
> the snapshot, right?  Well, no, it isn’t.)
> 
> Let’s fix this as best as we can.  Better Nate than lever.

Applied to my block branch.

Max

On 6/3/19 4:22 PM, Max Reitz wrote:
> QEMU’s always been confused over what a snapshot is: Is it the overlay?
> Is it the backing image?
> 
> Confusion is rarely a good thing.  I can’t think of any objective reason
> why the overlay would be a snapshot.  A snapshot is something that does
> not change over time; the overlay does.
> 
> (I suppose historically the reason is that “Taking an overlay” makes no
> sense, so the operations are called “Taking a snapshot”.  Somehow, this
> meaning carried over to the new file that is created during that
> operation; if “Creating a snapshot” creates a file, that file must be
> the snapshot, right?  Well, no, it isn’t.)
> 
> Let’s fix this as best as we can.  Better Nate than lever.
> 
> 
> v2:
> - Don’t break the iotests for a change
>   (kept Eric’s R-b, because it felt like the right thing to do)
> 
> 
> git backport-diff against v1:
> 
> Key:
> [----] : patches are identical
> [####] : number of functional differences between upstream/downstream patch
> [down] : patch is downstream-only
> The flags [FC] indicate (F)unctional and (C)ontextual differences, respectively
> 
> 001/2:[----] [--] 'qapi/block-core: Overlays are not snapshots'
> 002/2:[0010] [FC] 'blockdev: Overlays are not snapshots'
> 
> 
> Max Reitz (2):
>   qapi/block-core: Overlays are not snapshots
>   blockdev: Overlays are not snapshots
> 
>  qapi/block-core.json       | 20 ++++++++++----------
>  blockdev.c                 | 10 +++++-----
>  tests/qemu-iotests/085.out | 10 +++++-----
>  3 files changed, 20 insertions(+), 20 deletions(-)
> 

Makes good sense to me.

There are only 3,283 things named "snapshot" in QEMU so one less is
probably not the worst.

Reviewed-by: John Snow <jsnow@redhat.com>