[libvirt] [PATCH v2 1/3] qemu: Move allow reboot check setting

[John Ferlan]

On 11/6/18 7:48 AM, Michal Privoznik wrote:
> On 11/06/2018 01:28 PM, John Ferlan wrote:
>>
>>
>> On 11/6/18 4:38 AM, Michal Privoznik wrote:
>>> On 11/01/2018 05:04 PM, John Ferlan wrote:
>>>> Checking and setting the priv->allowReboot can be done before we start
>>>> processing the job. A subsequent patch will make use of the value to
>>>> make decisions in the error label, so we need to have it set properly.
>>>>
>>>> Signed-off-by: John Ferlan <jferlan at redhat.com>
>>>> ---
>>>>  src/qemu/qemu_process.c | 8 ++++----
>>>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>>>
>>>> diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c
>>>> index 9cf971808c..5232f761af 100644
>>>> --- a/src/qemu/qemu_process.c
>>>> +++ b/src/qemu/qemu_process.c
>>>> @@ -7767,6 +7767,10 @@ qemuProcessReconnect(void *opaque)
>>>>      cfg = virQEMUDriverGetConfig(driver);
>>>>      priv = obj->privateData;
>>>>  
>>>> +    /* If we are connecting to a guest started by old libvirt there is no
>>>> +     * allowReboot in status XML and we need to initialize it. */
>>>> +    qemuProcessPrepareAllowReboot(obj);
>>>
>>> I'm not quite sure why this happens outside of job. It doesn't look like
>>> it has to.
>>>
>>
>> Is there a reason in your opinion it needs to occur inside a job? It is
>> a void function.
> 
> The type of the return value doesn't matter.
> qemuProcessPrepareAllowReboot() changes private data and that is
> potentially dangerous if done outside modify job (even though the @vm is
> locked at this point so I guess it is not that dangerous after all).
> 

Does that mean we need to cull the code looking for everywhere in the
code where @priv data is modified outside a job? and start a job if so?
I thought jobs were more related to monitor interactions rather than
@priv modification.

The answer still doesn't make sense to me.

John

>>
>> It's moved to prior to the first "goto error" because of patch3 which
>> would call qemuDomainIsUsingNoShutdown which checks priv->allowReboot
>> which is possibly set in *AllowReboot. Without that move, the code would
>> need to be reworked, which is fine, but understanding the reason why I
>> wrote things the way I did is important, IMO. I can add a comment to
>> "warn" the next person trying to move it that the error: logic uses the
>> ->allowReboot value.
>>
>> The allowReboot alteration has nothing to do with a job AFAICT and
>> whether on error: there is a job or not. Perhaps no different to what
>> qemuDomainObjRestoreJob is doing by using @priv fields for @oldjob.
> 
> Yeah, but RestoreJob is special - we can't call it after job is
> acquired, we want to save currently running job to a temp variable so
> that we can start a new job.
> 
>>
>> Although looking at and quickly thinking about the code now, I wonder if
>> the virQEMUDriverGetCapabilities and goto error should be inside a job
>> too since error would then call qemuProcessStop without being in a job.
> 
> Ooops, yes.
> 
>>
>> If this is dropped then logic in patch3 needs to be altered in order to
>> account for jobStarted = true... I think that got too messy when I first
>> thought about it.
> 
> Michal
>