Re: [Qemu-devel] [PATCH V4 5/5] block: make all steps in qmp_transaction() as callback

[Wenchao Xia]

于 2013-5-3 22:46, Eric Blake 写道:

On 05/01/2013 08:26 PM, Wenchao Xia wrote:

Make it easier to add other operations to qmp_transaction() by using
callbacks, with external snapshots serving as an example implementation
of the callbacks.

Signed-off-by: Wenchao Xia <xiaw...@linux.vnet.ibm.com>
---
  blockdev.c |   92 ++++++++++++++++++++++++++++++++++++++++++++++-------------
  1 files changed, 71 insertions(+), 21 deletions(-)

+/*
+ * This structure must be arranged as first member in parent type, assuming

To me, "parent type" seems like the wrong relationship - in C++, the
parent type is the smaller type, and the child type is the larger type
that includes the parent type as its first member.  That is,
BlkTransationStates IS the parent type, and the comment would read
better as "first member in child type".

+ * that compiler will also arrange it to the same address with parent instance.
+ * Later it will be used in free().
+ */
+struct BlkTransactionStates {
+    BlockdevAction *action;
+    const BdrvActionOps *ops;
+    QSIMPLEQ_ENTRY(BlkTransactionStates) entry;
+};

      /* Now we are going to do the actual pivot.  Everything up to this point
       * is reversible, but we are committed at this point */

Is this comment still correct, now that things are extensible, or should
you s/pivot/commit/?

If you do respin, I agree that s/rollback/abort/ might be a nicer name
for that particular callback.  But I'm also quite okay with using the
patch as-is without a respin (while I pointed out two possible comment
wording changes, they don't make or break the patch for me).

Reviewed-by: Eric Blake <ebl...@redhat.com>

  I'll respin to fix comments and use "abort" as v5, thanks for your
reviewing!

--
Best Regards

Wenchao Xia