Re: [PATCH 3/5] KVM: Shrink struct kvm_mmu_memory_cache

[Sean Christopherson <seanjc@xxxxxxxxxx>]

On Mon, Feb 13, 2023, Mathias Krause wrote:
> Move the 'capacity' member around to make use of the padding hole on 64
> bit systems instead of introducing yet another one.
> 
> This allows us to save 8 bytes per instance for 64 bit builds of which,
> e.g., x86's struct kvm_vcpu_arch has a few.
> 
> Signed-off-by: Mathias Krause <minipli@xxxxxxxxxxxxxx>
> ---
>  include/linux/kvm_types.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/include/linux/kvm_types.h b/include/linux/kvm_types.h
> index 76de36e56cdf..8e4f8fa31457 100644
> --- a/include/linux/kvm_types.h
> +++ b/include/linux/kvm_types.h
> @@ -92,10 +92,10 @@ struct gfn_to_pfn_cache {
>   */
>  struct kvm_mmu_memory_cache {
>  	int nobjs;
> +	int capacity;
>  	gfp_t gfp_zero;
>  	gfp_t gfp_custom;
>  	struct kmem_cache *kmem_cache;
> -	int capacity;
>  	void **objects;

If we touch this, I vote to do a more aggressive cleanup and place nobjs next
to objects, e.g.

struct kvm_mmu_memory_cache {
	gfp_t gfp_zero;
	gfp_t gfp_custom;
	struct kmem_cache *kmem_cache;
	int capacity;
	int nobjs;
	void **objects;
};