[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v3 1/1] xen: move TLB-flush filtering out into populate_physmap during vm creation

On 09/09/16 10:46, Jan Beulich wrote:
>>>> On 09.09.16 at 11:34, <andrew.cooper3@xxxxxxxxxx> wrote:
>> On 09/09/16 08:56, Jan Beulich wrote:
>>>>>> On 09.09.16 at 07:37, <dongli.zhang@xxxxxxxxxx> wrote:
>>>>>> Did you go through and check that there is nothing this information
>>>>>> can already get derived from? I can't immediately point you at
>>>>>> anything, but it feels like there should. 
>>>>> Indeed. And if there isn't and we need to do add our own flagging,
>>>>> isn't there a better way and place where to put it (e.g., what Juergen
>>>>> and Andrew are hinting at)?
>>>> I prefer that to derive whether guest boot finishes is neither limited to a
>>>> specific domain type (pv, hvm, pvhvm or pvh) nor a specific arch (i386, 
>>>> x86_64,
>>>> arm32, arm64 or more in the future). Ideally, I would have one or a combo 
>>>> of
>>>> fields belong to "struct domain" but there isn't.
>>> Of course.
>>>> Can we use the field in vcpu[0]?
>>> I don't think so: What if another vCPU was scheduled to run first?
>> Use the XEN_DOMCTL_unpausedomain hypercall.  It must be called by
>> toolstacks on all architectures to complete domain construction.
> I think we had settled on this one already. The question now being
> discussed is whether a new field in the domain structure is needed,
> or whether the information he's after can be derived from already
> existing fields.

If the information were derivable from other means, there would be no
need to alter XEN_DOMCTL_unpausedomain in the first place.

I can't think of any option other than to add something new.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.