Re: [Qemu-devel] [PATCH v2 2/3] qapi: Add "detect-zeroes" option to drive-mirror

[Eric Blake]

On 06/08/2015 04:34 AM, Fam Zheng wrote:
> The new optional flag defaults to true, in which case, mirror job would
> check the read sectors and use sparse write if they are zero.  Otherwise
> data will be fully copied.

Is that a different default than in qemu 2.3?  That's okay, but I need
to figure out how to probe if qemu supports the new flag for libvirt to
know to set it.  I'm hoping Markus' work on introspection might save the
day...

> 
> Signed-off-by: Fam Zheng <address@hidden>
> ---
>  blockdev.c           | 26 +++++++++++++++++++++++++-
>  hmp.c                |  2 +-
>  qapi/block-core.json |  4 +++-
>  qmp-commands.hx      |  4 +++-
>  4 files changed, 32 insertions(+), 4 deletions(-)
> 
> diff --git a/blockdev.c b/blockdev.c
> index 5ad6960..3d008a2 100644
> --- a/blockdev.c
> +++ b/blockdev.c
> @@ -2622,6 +2622,7 @@ void qmp_drive_mirror(const char *device, const char 
> *target,
>                        bool has_on_source_error, BlockdevOnError 
> on_source_error,
>                        bool has_on_target_error, BlockdevOnError 
> on_target_error,
>                        bool has_unmap, bool unmap,
> +                      bool has_detect_zeroes, bool detect_zeroes,

Again, as I mentioned on 1/3, I think a tri-state enum might be easier
to use than two competing bools.  In fact, it might be more than
tri-state.  What are our possibilities?

1. We want the dest to be fully allocated, regardless of the source
being sparse
2. We want the dest to be as sparse as possible, regardless of the
source being fully allocated (or at least being unable to tell us about
holes)
3. We want the dest to mirror the sparseness of the host, but only where
that is efficient (if the source reads holes, make a hole in the dest)
4. Any other modes?

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

signature.asc
Description: OpenPGP digital signature