It might be useful to compress images during block-stream;
this way the user can merge compressed images of a backing chain and
the result will remain compressed.

Anton Nefedov (4):
  qcow2: reject unaligned offsets in write compressed
  block: support compressed write for copy-on-read
  block-stream: add compress option
  iotest 030: add compressed block-stream test

Pavel Butsykin (1):
  qcow2: multiple clusters write compressed

 qapi/block-core.json       |  4 +++
 include/block/block_int.h  |  4 ++-
 block/io.c                 | 30 +++++++++++++++----
 block/qcow2.c              | 73 +++++++++++++++++++++++++++++++++++-----------
 block/stream.c             | 16 +++++++---
 blockdev.c                 | 13 ++++++++-
 hmp.c                      |  2 ++
 block/trace-events         |  2 +-
 hmp-commands.hx            |  4 +--
 tests/qemu-iotests/030     | 69 ++++++++++++++++++++++++++++++++++++++++++-
 tests/qemu-iotests/030.out |  4 +--
 11 files changed, 186 insertions(+), 35 deletions(-)

-- 
2.7.4

On Tue, 11/14 13:16, Anton Nefedov wrote:
> It might be useful to compress images during block-stream;
> this way the user can merge compressed images of a backing chain and
> the result will remain compressed.

I haven't looked at the patches yet so maybe the answer is obvious, but still:
would the "compress" option be still necessary if what we have is
blockdev-stream instead?

Fam

On 16/11/2017 6:26 AM, Fam Zheng wrote:
> On Tue, 11/14 13:16, Anton Nefedov wrote:
>> It might be useful to compress images during block-stream;
>> this way the user can merge compressed images of a backing chain and
>> the result will remain compressed.
> 
> I haven't looked at the patches yet so maybe the answer is obvious, but still:
> would the "compress" option be still necessary if what we have is
> blockdev-stream instead?
> 
> Fam
> 

Sorry, I can't see blockdev-stream available now. How that would be
different? Would it imply the blocks are pulled up to the top (active)
image?

Or did you mean block-commit? The option might be useful there as well,
and needs to be implemented separately

/Anton