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;
Sender: Paolo Bonzini <paolo.bonzini@gmail.com>
Message-ID: <7ccb16e5-579e-b3d9-cedc-305152ef9b8f@redhat.com>
Date:   Fri, 11 Feb 2022 17:06:00 +0100
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.5.0
Subject: Re: [RFC PATCH 4/6] KVM: X86: Introduce role.level_promoted
Content-Language: en-US
To:     Sean Christopherson <seanjc@google.com>,
        Lai Jiangshan <jiangshanlai@gmail.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Lai Jiangshan <laijs@linux.alibaba.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>, x86@kernel.org,
        "H. Peter Anvin" <hpa@zytor.com>
References: <20211210092508.7185-1-jiangshanlai@gmail.com>
 <20211210092508.7185-5-jiangshanlai@gmail.com> <YdTGuoh2vYPsRcu+@google.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
In-Reply-To: <YdTGuoh2vYPsRcu+@google.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 1/4/22 23:14, Sean Christopherson wrote:
> Alternatively, should we mark passthrough shadow pages as direct=1?  That would
> naturally handle this code, and for things like reexecute_instruction()'s usage
> of kvm_mmu_unprotect_page(), I don't think passthrough shadow pages should be
> considered indirect, e.g. zapping them won't help and the shadow page can't become
> unsync.

So the main difference between direct and passthrough shadow pages is that
passthrough pages can have indirect children.  A direct page maps the
page at sp->gfn, while a passthrough page maps the page _table_ at
sp->gfn.

Is this correct?

If so, I think there is a difference between a passthrough page that
maps a level-2 page from level-4, and a passthrough page that maps a
level-3 page from level-4.  This means that a single bit in the role
is not enough.

One way to handle this could be to have a single field "direct_levels"
that subsumes both "direct" and "passthrough".  direct && !passthrough
would correspond to "direct_levels == level", while !direct && !passthrough
would correspond to "direct_levels == 0".

Paolo