Re: [PATCH 09/10 RFC] KVM: x86: MMU: Move parent_pte handling from kvm_mmu_get_page() to link_shadow_page()

From: Paolo Bonzini <pbonzini@redhat.com>
To: Takuya Yoshikawa <yoshikawa_takuya_b1@lab.ntt.co.jp>
Cc: kvm@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 09/10 RFC] KVM: x86: MMU: Move parent_pte handling from kvm_mmu_get_page() to link_shadow_page()
Date: Thu, 12 Nov 2015 15:27:26 +0100	[thread overview]
Message-ID: <5644A1CE.1040906@redhat.com> (raw)
In-Reply-To: <20151112205656.33f23effece7e6914c3e2646@lab.ntt.co.jp>

On 12/11/2015 12:56, Takuya Yoshikawa wrote:
> diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
> index 9d21b44..f414ca6 100644
> --- a/arch/x86/kvm/paging_tmpl.h
> +++ b/arch/x86/kvm/paging_tmpl.h
> @@ -598,7 +598,7 @@ static int FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
>  			goto out_gpte_changed;
>  
>  		if (sp)
> -			link_shadow_page(it.sptep, sp, PT_GUEST_ACCESSED_MASK);
> +			link_shadow_page(vcpu, it.sptep, sp, PT_GUEST_ACCESSED_MASK);
>  	}
>  

Here I think you can remove completely the

	if (sp)
		kvm_mmu_put_page(sp, it.sptep);

later in FNAME(fetch).  Apart from this nit, it's okay.

On to kvm_mmu_get_page...

        if (!direct) {
                if (rmap_write_protect(vcpu, gfn))
                        kvm_flush_remote_tlbs(vcpu->kvm);
                if (level > PT_PAGE_TABLE_LEVEL && need_sync)
                        kvm_sync_pages(vcpu, gfn);

This seems fishy.

need_sync is set if sp->unsync, but then the parents have not been
unsynced yet.

On the other hand, all calls to kvm_mmu_get_page except for the
roots are followed by link_shadow_page...  Perhaps if parent_pte != NULL
you can call link_shadow_page directly from kvm_mmu_get_page.  The call
would go before the "if (!direct)" and it would subsume all the existing
calls.

We could probably also warn if

	(parent_pte == NULL)
		!= (level == vcpu->arch.mmu.root_level)

in kvm_mmu_get_page.

Paolo