[Qemu-devel] [PATCH v2 0/5] Support streaming to an intermediate layer

[Alberto Garcia <berto@igalia.com>]

This is a new version of the patches that add support for streaming
to any intermediate layer. You can check the previous version here:

   https://lists.gnu.org/archive/html/qemu-devel/2015-02/msg04116.html

I followed Kevin's idea to put the ownership of the block job directly
on the node that receives the data instead of the root node.

There's a few things that I'm not completely sure of and will
certainly generate some debate, but I decided to send the code anyway
so the actual changes can be seen. Note that this depends on the "Add
bdrv_get_device_or_node_name()" patchset I sent last week.

Here are the changes:

1) The 'top' parameter of block-stream disappears. 'device' also
   accepts a node name now, which is used to specify the node where
   the data will be written.

2) Block jobs can now be owned by any node. This implies:

   - The block-job-* commands now also accept a node name in the
     'device' parameter. I decided not to add a separate 'node-name'
     parameter, following what we agreed with the 'block-stream'
     command.

   - The BlockJobInfo type and BLOCK_JOB_* events will report the node
     name in the 'device' field if the node does not have a device
     name. It seems that we had an agreement for the BlockJobInfo case
     so I decided to follow the same approach for these events.

     However, in the BLOCK_IMAGE_CORRUPTED case the preferred solution
     was to add a new 'node-name' field, so I guess we might want to
     do the same here?

   - query-block-jobs now searches the whole tree, not just the root
     nodes.

3) Operations are blocked in all intermediate nodes during the job. If
   we have a chain [A] -> [B] -> [C] -> [D] -> [E] and we stream from
   [B] to [E], then [C] and [D] will also be blocked during the job.

   Since [C] and [D] will be removed from the chain after the job is
   finished I understand that we don't want to perform any operation
   with them.

4) As a consequence of 3), op blockers are also checked in all
   intermediate nodes (not just in the topmost one) before starting a
   streaming operation. I'm currently checking BLOCK_OP_TYPE_STREAM,
   but maybe I should use a different op for the intermediate nodes
   since what I'm going to do there is removing them from the chain?

I think those are the most important changes. Any feedback is welcome!

Berto

v2:
- The 'block-stream' command does not have a 'node-name' parameter
  anymore and reuses 'device' for that purpose.
- Block jobs can now be owned by any intermediate node, and not just
  by the ones at the root. query-block-jobs is updated to reflect that
  change.
- The 'device' parameter of all 'block-job-*' commands can now take a
  node name.
- The BlockJobInfo type and all BLOCK_JOB_* events report the node
  name in the 'device' field if the node does not have a device name.
- All intermediate nodes are blocked (and checked for blockers) during
  the streaming operation.

Alberto Garcia (5):
  block: allow block jobs in any arbitrary node
  block: never cancel a streaming job without running stream_complete()
  block: Support streaming to an intermediate layer
  block: Add QMP support for streaming to an intermediate layer
  docs: Document how to stream to an intermediate layer

 block.c                   |  4 +++-
 block/mirror.c            |  5 +++--
 block/stream.c            | 47 ++++++++++++++++++++++++++++++++++++++++++-----
 blockdev.c                | 47 +++++++++++++++++++++++++----------------------
 blockjob.c                | 17 +++++++++--------
 docs/live-block-ops.txt   | 32 ++++++++++++++++++++------------
 docs/qmp/qmp-events.txt   |  8 ++++----
 include/qapi/qmp/qerror.h |  3 ---
 qapi/block-core.json      | 28 ++++++++++++++++------------
 9 files changed, 122 insertions(+), 69 deletions(-)

-- 
2.1.4