On Mon, Jul 20, 2015 at 15:49:46 -0600, Eric Blake wrote:
On 07/20/2015 02:48 PM, John Ferlan wrote:
>
>
> On 07/15/2015 12:33 PM, Peter Krempa wrote:
>> This series refactors the block job command code so that it is handled by
>> separate functions rather than one mega-method. Additionally this series then
>> fixes the routine for waiting for a block job.
>>
>> As it's perhaps obvious from the patches the API design for the block job
APIs
>> is rather unfortunate for users.
>>
>> This series contains a few patches that already were submitted and partially
>> reviewed previously.
>>
>
> Noted a couple of minor things, but looks good otherwise.
and I pointed out a few more
>
> ACK series,
Concur. This makes virsh look a lot nicer, and the big benefit is that
it tries to address the data race bug where live block-commit --pivot
could read status too early and fail (probably because qemu was
reporting info.end of 0 before actually getting into the main loop of
the job).
Qemu was actually reporting 100% completion (non-zero) for a brief
moment before the event is delivered. If you try to pivot the job at
that point you'll get the error message too.
Peter