From mboxrd@z Thu Jan  1 00:00:00 1970
From: Marcelo Tosatti <mtosatti@redhat.com>
Subject: Re: [patch 4/4] KVM: MMU: pinned sps are not candidates for deletion.
Date: Tue, 9 Sep 2014 12:41:27 -0300
Message-ID: <20140909154127.GC4153@amt.cnet>
References: <20140709191250.408928362@amt.cnet>
 <20140709191611.349481802@amt.cnet>
 <1C17FA2A-5EFB-46D8-A105-CCC31B4BCFC6@linux.vnet.ibm.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: kvm@vger.kernel.org, ak@linux.intel.com, pbonzini@redhat.com,
	gleb@kernel.org, avi.kivity@gmail.com
To: Xiao Guangrong <xiaoguangrong.eric@gmail.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:53719 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1757123AbaIIPlt (ORCPT <rfc822;kvm@vger.kernel.org>);
	Tue, 9 Sep 2014 11:41:49 -0400
Content-Disposition: inline
In-Reply-To: <1C17FA2A-5EFB-46D8-A105-CCC31B4BCFC6@linux.vnet.ibm.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On Tue, Jul 22, 2014 at 05:59:42AM +0800, Xiao Guangrong wrote:
> 
> On Jul 10, 2014, at 3:12 AM, mtosatti@redhat.com wrote:
> 
> > Skip pinned shadow pages when selecting pages to zap.
> 
> It seems there is no way to prevent changing pinned spte on
> zap-all path?

Xiao,

The way would be to reload remote mmus, forcing the vcpu to exit,
zapping a page, then vcpu will pagefault any necessary page via
kvm_mmu_pin_pages.

kvm_mmu_invalidate_zap_all_pages does: 

- spin_lock(mmu_lock)
- kvm_reload_remote_mmus
...
- spin_unlock(mmu_lock)

So its OK to change pinned spte on zap all path.

> I am thing if we could move pinned spte to another list (eg. pinned_shadow_pages)
> instead of active list so that it can not be touched by any other free paths. 
> Your idea?

As mentioned it above, it is ok to zap pinned sptes as long w 
reload remote mmus request is performed.

That said, you still consider a separate list?