Re: [PATCH 0/7] KVM: Kernel-based Virtual Machine

[Avi Kivity <avi@qumranet.com>]

Arnd Bergmann wrote:
> On Sunday 22 October 2006 20:41, Avi Kivity wrote:
>   
>>> Ok, but if you radically change the kernel<->user API, doesn't that mean
>>> you have to upgrade in the same way?
>>>       
>> No, why? I'd just upgrade the userspace.  Am I misunderstanding you?
>>     
>
> If you change the kernel interface, you also have to change the kernel
> itself, at least if you introduce new syscalls.
>
>   

But I don't have to upgrade all my software to 64 bit [but 32-bit 
emulation solves that].

Still, an upgrade to the next 32-bit kernel could be seen as less 
threatening.


>>> The 32 bit emulation mode in x86_64
>>> is actually pretty complete, so it probably boils down to a kernel
>>> upgrade for you, without having to touch any of the user space.
>>>  
>>>       
>> For me personally, I don't mind.  I don't know about others.
>>     
>
> I'd really love to see your code in get into the mainline kernel,
> but I'd consider 32 bit host support an unnecessary burden for
> long-term maintenance. Maybe you could maintain the 32 bit version
> out of tree as long as there is still interest? I would expect that
> at least the point where it works out of the box on x86_64
> distros is when it becomes completely obsolete.
>   

One of my motivations was to get testers who run 32-bit for historical 
or flash plugin reasons.

If there is a consensus that it should be dropped, though, I'll drop 
it.  I certainly didn't have any fun getting it to run.

-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.