Re: [Qemu-devel] [PATCH 3/3] block: mirror - zero unallocated target sectors when zero init not present

[Kevin Wolf <kwolf@redhat.com>]

Am 30.09.2015 um 18:06 hat Paolo Bonzini geschrieben:
> 
> 
> On 30/09/2015 18:02, Jeff Cody wrote:
> > As a general rule for blockjobs, I disagree.
> > 
> > Right away, there is a key difference: we don't know that the image is
> > (or should be) empty.
> 
> Not necessarily empty.  sync='top' && mode='existing' &&
> !target->backing_file, for example, makes sense if the target is a copy
> of source->bs.
> 
> In fact, commit of the active layer is almost exactly a mode='existing'
> drive-mirror operation.
> 
> But if you use mode == 'existing', and don't provide an image that
> follows the rules, it's garbage-in garbage-out.  The sequence of
> operation makes sense, but the resulting image does not.

Yes, that's the point I was trying to make. The behaviour of our block
jobs doesn't depend on the contents of the target image, and I don't see
a future block job where this would change.

So if the user thinks it makes sense to start with a non-empty image and
keep some non-zero data in places that the block job didn't touch,
that's their choice (and as you mentioned, there are examples where it
does indeed make sense).

The block job code need not care about that, it just does its job
without ever looking at the contents of the target image.

Kevin