Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Thu, 15 Dec 2022 00:11:15 +0000
From:   Sean Christopherson <seanjc@google.com>
To:     Robert Hoo <robert.hu@linux.intel.com>
Cc:     Paolo Bonzini <pbonzini@redhat.com>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org, Greg Thelen <gthelen@google.com>,
        David Matlack <dmatlack@google.com>,
        Ben Gardon <bgardon@google.com>,
        Mingwei Zhang <mizhang@google.com>
Subject: Re: [PATCH 3/5] KVM: x86/mmu: Re-check under lock that TDP MMU SP
 hugepage is disallowed
Message-ID: <Y5pmI6xwJhvrWXJ8@google.com>
References: <20221213033030.83345-1-seanjc@google.com>
 <20221213033030.83345-4-seanjc@google.com>
 <247fcfc6de8ec08d0667de125e707046dce903fc.camel@linux.intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <247fcfc6de8ec08d0667de125e707046dce903fc.camel@linux.intel.com>
Precedence: bulk

On Wed, Dec 14, 2022, Robert Hoo wrote:
> On Tue, 2022-12-13 at 03:30 +0000, Sean Christopherson wrote:
> > ---
> >  arch/x86/kvm/mmu/tdp_mmu.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > diff --git a/arch/x86/kvm/mmu/tdp_mmu.c b/arch/x86/kvm/mmu/tdp_mmu.c
> > index e2e197d41780..fd4ae99790d7 100644
> > --- a/arch/x86/kvm/mmu/tdp_mmu.c
> > +++ b/arch/x86/kvm/mmu/tdp_mmu.c
> > @@ -1203,7 +1203,8 @@ int kvm_tdp_mmu_map(struct kvm_vcpu *vcpu,
> > struct kvm_page_fault *fault)
> >  		if (fault->huge_page_disallowed &&
> >  		    fault->req_level >= iter.level) {
> >  			spin_lock(&kvm->arch.tdp_mmu_pages_lock);
> > -			track_possible_nx_huge_page(kvm, sp);
> > +			if (sp->nx_huge_page_disallowed)
> > +				track_possible_nx_huge_page(kvm, sp);
> >  			spin_unlock(&kvm->arch.tdp_mmu_pages_lock);
> >  		}
> >  	}
> 
> Is this possible?
> The aforementioned situation happened, i.e. before above hunk
> track_possible_nx_huge_page(), the sp is zapped by some other task,
> tdp_mmu_unlink_sp() --> untrack_possible_nx_huge_page(kvm, sp):

It's possible for untrack_possible_nx_huge_page() to be called before the above
snippet, but the stat won't be decremented in that case since the page won't be on
the list of possible NX huge pages, i.e. list_empty() will be true.

  void untrack_possible_nx_huge_page(struct kvm *kvm, struct kvm_mmu_page *sp)
  {
	if (list_empty(&sp->possible_nx_huge_page_link))
		return;

	--kvm->stat.nx_lpage_splits;

And by not calling track_possible_nx_huge_page() (this bug fix), nx_lpage_splits
won't be incorrectly incremented.

> 
> --kvm->stat.nx_lpage_splits;
> 
> But looks like the stat for this sp hasn't been increased yet.
>