Re: [RFC 00/10] Process-local memory allocations for hiding KVM secrets

[Dave Hansen]

On 6/13/19 12:27 AM, Alexander Graf wrote:
>> Where's the context-switching code? Did I just miss it?
> 
> I'm not sure I understand the question. With this mechanism, the global
> linear map pages are just not present anymore, so there is no context
> switching needed. For the process local memory, the page table is
> already mm local, so we don't need to do anything special during context
> switch, no?

Thanks for explaining, I was just confused.

Andy reminded me when comparing it to the LDT area: since this area is
per-mm/pgd and we context switch that *obviously* there's no extra work
to do at context switch time, as long as the area is marked non-Global.