From mboxrd@z Thu Jan  1 00:00:00 1970
From: Avi Kivity <avi@redhat.com>
Subject: Re: KVM: MMU: zero caches before entering mmu_lock protected section
Date: Thu, 08 Jan 2009 20:35:56 +0200
Message-ID: <4966478C.7060608@redhat.com>
References: <20090108182617.GA7250@amt.cnet>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Cc: kvm@vger.kernel.org
To: Marcelo Tosatti <mtosatti@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx2.redhat.com ([66.187.237.31]:33876 "EHLO mx2.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1754333AbZAHSfi (ORCPT <rfc822;kvm@vger.kernel.org>);
	Thu, 8 Jan 2009 13:35:38 -0500
Received: from int-mx2.corp.redhat.com (int-mx2.corp.redhat.com [172.16.27.26])
	by mx2.redhat.com (8.13.8/8.13.8) with ESMTP id n08IZcMj006837
	for <kvm@vger.kernel.org>; Thu, 8 Jan 2009 13:35:38 -0500
In-Reply-To: <20090108182617.GA7250@amt.cnet>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

Marcelo Tosatti wrote:
> Clean the pre-allocated cache pages before entering mmu_lock region.
> This is safe since the caches are per-vcpu.
>
> Smaller chunks are already zeroed by kmem_cache_zalloc.
>
> ~= 0.90% reduction in system time with AIM7 on RHEL3 / 4-vcpu.
>
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
>
> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
> index 10bdb2a..823d0cd 100644
> --- a/arch/x86/kvm/mmu.c
> +++ b/arch/x86/kvm/mmu.c
> @@ -301,7 +301,7 @@ static int mmu_topup_memory_cache_page(struct kvm_mmu_memory_cache *cache,
>  	if (cache->nobjs >= min)
>  		return 0;
>  	while (cache->nobjs < ARRAY_SIZE(cache->objects)) {
> -		page = alloc_page(GFP_KERNEL);
> +		page = alloc_page(GFP_KERNEL|__GFP_ZERO);
>  		if (!page)
>  			return -ENOMEM;
>  		set_page_private(page, 0);
> @@ -352,7 +352,6 @@ static void *mmu_memory_cache_alloc(struct kvm_mmu_memory_cache *mc,
>  
>  	BUG_ON(!mc->nobjs);
>  	p = mc->objects[--mc->nobjs];
> -	memset(p, 0, size);
>  	return p;
>  }
>  
>   

I think we can drop the memset altogether, since we will clear the page 
in ->prefetch_page() anyway.


-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.