Re: [RFC, PATCH 09/12] KVM: TDX: Preallocate PAMT pages to be used in page fault path

["Huang, Kai" <kai.huang@intel.com>]

On Fri, 2025-05-02 at 16:08 +0300, Kirill A. Shutemov wrote:
> Preallocate a page to be used in the link_external_spt() and
> set_external_spte() paths.
> 
> In the worst-case scenario, handling a page fault might require a
> tdx_nr_pamt_pages() pages for each page table level.
> 
> Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> ---
>  arch/x86/include/asm/kvm_host.h |  2 ++
>  arch/x86/kvm/mmu/mmu.c          | 10 ++++++++++
>  2 files changed, 12 insertions(+)
> 
> diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> index 91958c55f918..a5661499a176 100644
> --- a/arch/x86/include/asm/kvm_host.h
> +++ b/arch/x86/include/asm/kvm_host.h
> @@ -849,6 +849,8 @@ struct kvm_vcpu_arch {
>  	 */
>  	struct kvm_mmu_memory_cache mmu_external_spt_cache;
>  
> +	struct kvm_mmu_memory_cache pamt_page_cache;
> +
>  	/*
>  	 * QEMU userspace and the guest each have their own FPU state.
>  	 * In vcpu_run, we switch between the user and guest FPU contexts.
> diff --git a/arch/x86/kvm/mmu/mmu.c b/arch/x86/kvm/mmu/mmu.c
> index a284dce227a0..7bfa0dc50440 100644
> --- a/arch/x86/kvm/mmu/mmu.c
> +++ b/arch/x86/kvm/mmu/mmu.c
> @@ -616,6 +616,15 @@ static int mmu_topup_memory_caches(struct kvm_vcpu *vcpu, bool maybe_indirect)
>  		if (r)
>  			return r;
>  	}
> +
> +	if (vcpu->kvm->arch.vm_type == KVM_X86_TDX_VM) {
> +		int nr = tdx_nr_pamt_pages(tdx_get_sysinfo());
> +		r = kvm_mmu_topup_memory_cache(&vcpu->arch.pamt_page_cache,
> +					       nr * PT64_ROOT_MAX_LEVEL);
> +		if (r)
> +			return r;
> +	}
> +
>  	return kvm_mmu_topup_memory_cache(&vcpu->arch.mmu_page_header_cache,
>  					  PT64_ROOT_MAX_LEVEL);
>  }
> @@ -626,6 +635,7 @@ static void mmu_free_memory_caches(struct kvm_vcpu *vcpu)
>  	kvm_mmu_free_memory_cache(&vcpu->arch.mmu_shadow_page_cache);
>  	kvm_mmu_free_memory_cache(&vcpu->arch.mmu_shadowed_info_cache);
>  	kvm_mmu_free_memory_cache(&vcpu->arch.mmu_external_spt_cache);
> +	kvm_mmu_free_memory_cache(&vcpu->arch.pamt_page_cache);
>  	kvm_mmu_free_memory_cache(&vcpu->arch.mmu_page_header_cache);
>  }
>  

IIUC, this patch can be avoided if we create an actual kmem_cache for
mmu_external_spt_cache with an actual 'ctor' where we simply call
tdx_alloc_page() as replied to the previous patch.