[PATCH v2 00/13] perf: KVM: Fix, optimize, and clean up callbacks

[Paolo Bonzini]

On 20/09/21 15:40, Marc Zyngier wrote:
>>> At least not before we
>>> declare the arm64 single kernel image policy to be obsolete.
>>
>> --verbose please.:)   I am sure you're right, but I don't understand
>> the link between the two.
>
> To start making KVM/arm64 modular, you'd have to build it such as
> there is no support for the nVHE hypervisor anymore. Which would mean
> two different configs (one that can only work with VHE, and one for
> the rest) and contradicts the current single kernel image policy.

Ah okay, I interpreted the policy as "it's possible to build a single 
kernel image but it would be possible to build an image for a subset of 
the features as well".

In that case you could have one config that can work either with or 
without VHE (and supports y/n) and one config that can only work with 
VHE (and supports y/m/n).  The code to enter VHE EL2 would of course 
always be builtin.

> It is bad enough that we have to support 3 sets of page sizes.
> Doubling the validation space for the sake of being able to unload KVM
> seems a dubious prospect.

It's not even a configuration that matches kconfig very well, since it 
does have a way to build something *only as a module*, but not a way to 
build something only as built-in.

That said, if you had the possibility to unload/reload KVM, you'll 
quickly become unable to live without it. :)

Paolo