From mboxrd@z Thu Jan  1 00:00:00 1970
From: Marcelo Tosatti <mtosatti@redhat.com>
Subject: Re: [patch 3/5] KVM: MMU: notifiers support for pinned sptes
Date: Thu, 19 Jun 2014 15:28:25 -0300
Message-ID: <20140619182825.GB32410@amt.cnet>
References: <20140618231203.846608908@amt.cnet>
 <20140618231521.648087161@amt.cnet>
 <20140619064850.GB10948@minantech.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: kvm@vger.kernel.org, ak@linux.intel.com, pbonzini@redhat.com,
	xiaoguangrong@linux.vnet.ibm.com, avi@cloudius-systems.com
To: Gleb Natapov <gleb@kernel.org>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mx1.redhat.com ([209.132.183.28]:13967 "EHLO mx1.redhat.com"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S934039AbaFSTK4 (ORCPT <rfc822;kvm@vger.kernel.org>);
	Thu, 19 Jun 2014 15:10:56 -0400
Content-Disposition: inline
In-Reply-To: <20140619064850.GB10948@minantech.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>

On Thu, Jun 19, 2014 at 09:48:50AM +0300, Gleb Natapov wrote:
> On Wed, Jun 18, 2014 at 08:12:06PM -0300, mtosatti@redhat.com wrote:
> > Request KVM_REQ_MMU_RELOAD when deleting sptes from MMU notifiers.
> > 
> > Keep pinned sptes intact if page aging.
> > 
> > Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
> > 
> > ---
> >  arch/x86/kvm/mmu.c |   71 ++++++++++++++++++++++++++++++++++++++++++++++-------
> >  1 file changed, 62 insertions(+), 9 deletions(-)
> > 
> > Index: kvm.pinned-sptes/arch/x86/kvm/mmu.c
> > ===================================================================
> > --- kvm.pinned-sptes.orig/arch/x86/kvm/mmu.c	2014-06-18 17:28:24.339435654 -0300
> > +++ kvm.pinned-sptes/arch/x86/kvm/mmu.c	2014-06-18 17:29:32.510225755 -0300
> > @@ -1184,6 +1184,42 @@
> >  		kvm_flush_remote_tlbs(vcpu->kvm);
> >  }
> >  
> > +static void ack_flush(void *_completed)
> > +{
> > +}
> > +
> > +static void mmu_reload_pinned_vcpus(struct kvm *kvm)
> > +{
> > +	int i, cpu, me;
> > +	cpumask_var_t cpus;
> > +	struct kvm_vcpu *vcpu;
> > +	unsigned int req = KVM_REQ_MMU_RELOAD;
> > +
> > +	zalloc_cpumask_var(&cpus, GFP_ATOMIC);
> > +
> > +	me = get_cpu();
> > +	kvm_for_each_vcpu(i, vcpu, kvm) {
> > +		if (list_empty(&vcpu->arch.pinned_mmu_pages))
> > +			continue;
> > +		kvm_make_request(req, vcpu);
> > +		cpu = vcpu->cpu;
> > +
> > +		/* Set ->requests bit before we read ->mode */
> > +		smp_mb();
> > +
> > +		if (cpus != NULL && cpu != -1 && cpu != me &&
> > +		      kvm_vcpu_exiting_guest_mode(vcpu) != OUTSIDE_GUEST_MODE)
> > +			cpumask_set_cpu(cpu, cpus);
> > +	}
> > +	if (unlikely(cpus == NULL))
> > +		smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
> > +	else if (!cpumask_empty(cpus))
> > +		smp_call_function_many(cpus, ack_flush, NULL, 1);
> > +	put_cpu();
> > +	free_cpumask_var(cpus);
> > +	return;
> > +}
> This is a c&p of make_all_cpus_request(), the only difference is checking
> of vcpu->arch.pinned_mmu_pages.  You can add make_some_cpus_request(..., bool (*predicate)(struct kvm_vcpu *))
> to kvm_main.c and rewrite make_all_cpus_request() to use it instead.

Half-way through it i decided it was better to c&p.

Can change make_all_cpus_request() though if it makes more sense to you.