Re: [Qemu-discuss] KVM on Mac OS X host

[Programmingkid]

On Aug 21, 2015, at 5:53 PM, Alexander Graf wrote:

> 
> 
>> Am 21.08.2015 um 12:08 schrieb Programmingkid <address@hidden>:
>> 
>> 
>>> On Aug 21, 2015, at 2:57 PM, Peter Maydell wrote:
>>> 
>>>> On 19 August 2015 at 20:40, Programmingkid <address@hidden> wrote:
>>>> 
>>>>> On Aug 19, 2015, at 3:25 PM, Peter Maydell wrote:
>>>>> 
>>>>>> On 18 August 2015 at 14:12, Programmingkid <address@hidden> wrote:
>>>>>> Do you think this should be a Google Summer of Code project?
>>>>> 
>>>>> I guess it's probably a size that would work for a GSoC
>>>>> project with a suitably capable and enthusiastic student.
>>>>> 
>>>>> Are you asking because you would want to do it as a GSoC
>>>>> project yourself, or because you'd like to see it on the list
>>>>> so somebody might apply and volunteer to do it?
>>>> 
>>>> I would do it myself but I don't know too much about hypervisors,
>>>> so I'm hoping there is someone else out there who is far more
>>>> qualified to do it. This will probably take a GSOC student to do it.
>>>> Maybe I will add this to the 2016 GSOC page if it exists yet.
>>> 
>>> I spoke to Alex Graf, who said he'd had a look into the OSX
>>> hypervisor API. Apparently the amount of work required to
>>> support this would be more than my initial glance at the API
>>> suggested, so it's really too large to be a GSOC project.
>>> Also a major problem is that it would be a large amount of
>>> code which would need to be maintained into the future by
>>> somebody, which makes a GSOC project a poor fit -- GSOC
>>> students often don't stick around after the end of the summer,
>>> so there wouldn't be a maintainer for the code.
>> 
>> Do you think it would be possible to use VirtualBox's hypervisor with QEMU 
>> on Mac OS X, and/or Linux, and/or Windows?
> 
> Yes and no. The virtualbox kernel interface consists of 2 ioctls:
> 
>  - copy code into executable kernel space
>  - jump to offset in that code in kernel mode
> 
> So all of the logic of vbox is implemented in an additional blob that needs 
> to get loaded on demand. There are no guarantees for stable interfaces to 
> that blob whatsoever.

Do you think there is any hope for hypervisor support in Mac OS X for QEMU?