[libvirt] [PATCH] qemu: call endjob in RevertToSnapshot

[Jincheng Miao]

On 08/27/2014 07:52 PM, Eric Blake wrote:
> On 08/27/2014 03:20 AM, Jincheng Miao wrote:
>> On 08/27/2014 11:02 AM, Eric Blake wrote:
>>> On 08/26/2014 08:10 PM, Jincheng Miao wrote:
>>>> In qemuDomainRevertToSnapshot(), it will check snap->def->state.
>>>> But when the state is PMSUSPENDED/NOSTATE/BLOCKED, it forgets to
>>>> call qemuDomainObjEndJob.
>>>>
>>>> Signed-off-by: Jincheng Miao <jmiao at redhat.com>
>>>> ---
>>>>    src/qemu/qemu_driver.c |    4 ++--
>>>>    1 files changed, 2 insertions(+), 2 deletions(-)
>>> This looks like the fix for
>>> https://bugzilla.redhat.com/show_bug.cgi?id=1134154 - thanks for chasing
>>> that!
>> This patch fix the part of that bug, but it doesn't touch the race
>> condition:
>> Why the snap->def->state is NOSTATE? It should be RUNNING.
>> The only reasonable explanation is that snapshot object is free()ed.
>>
>> IMHO, the SnapshotObjList should be clocked during snapshot job running.
> Okay, I see what you are saying; we need to make sure two parallel
> snapshot operations can't stomp on one another (probably by making them
> all guarded by the vm job lock).

Yes, wrapper the snapshot operations by vm job lock good choice,
and I will check some other functions, and compose a patch for it.


> I hope to have time to look into a fix
> before 1.2.8; but as it has been a long-standing issue, it's not a new
> regression in this release if I miss that goal (my priority today is
> getting 1.2.8 APIs to be feature-complete).
>