Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 9 Apr 2021 17:12:11 +0300
From:   "Kirill A. Shutemov" <kirill@shutemov.name>
To:     David Hildenbrand <david@redhat.com>
Cc:     Dave Hansen <dave.hansen@linux.intel.com>,
        Andy Lutomirski <luto@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>,
        Sean Christopherson <seanjc@google.com>,
        Jim Mattson <jmattson@google.com>,
        David Rientjes <rientjes@google.com>,
        "Edgecombe, Rick P" <rick.p.edgecombe@intel.com>,
        "Kleen, Andi" <andi.kleen@intel.com>,
        "Yamahata, Isaku" <isaku.yamahata@intel.com>, x86@kernel.org,
        kvm@vger.kernel.org, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Oscar Salvador <osalvador@suse.de>,
        Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Subject: Re: [RFCv1 7/7] KVM: unmap guest memory using poisoned pages
Message-ID: <20210409141211.wfbyzflj7ygtx7ex@box.shutemov.name>
References: <20210402152645.26680-1-kirill.shutemov@linux.intel.com>
 <20210402152645.26680-8-kirill.shutemov@linux.intel.com>
 <5e934d94-414c-90de-c58e-34456e4ab1cf@redhat.com>
 <20210409133347.r2uf3u5g55pp27xn@box>
 <5ef83789-ffa5-debd-9ea2-50d831262237@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <5ef83789-ffa5-debd-9ea2-50d831262237@redhat.com>
Precedence: bulk

On Fri, Apr 09, 2021 at 03:50:42PM +0200, David Hildenbrand wrote:
> > > 3. Allow selected users to still grab the pages (esp. KVM to fault them into
> > > the page tables).
> > 
> > As long as fault leads to non-present PTEs we are fine. Usespace still may
> > want to mlock() some of guest memory. There's no reason to prevent this.
> 
> I'm curious, even get_user_pages() will lead to a present PTE as is, no? So
> that will need modifications I assume. (although I think it fundamentally
> differs to the way get_user_pages() works - trigger a fault first, then
> lookup the PTE in the page tables).

For now, the patch has two step poisoning: first fault in, on the add to
shadow PTE -- poison. By the time VM has chance to use the page it's
poisoned and unmapped from the host userspace.

-- 
 Kirill A. Shutemov