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;
To:     Alexander Graf <graf@amazon.de>
Cc:     "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "kvm@vger.kernel.org" <kvm@vger.kernel.org>,
        Aaron Lewis <aaronlewis@google.com>,
        Peter Xu <peterx@redhat.com>,
        Sean Christopherson <sean.j.christopherson@intel.com>
References: <20201019170519.1855564-1-pbonzini@redhat.com>
 <618E2129-7AB5-4F0D-A6C9-E782937FE935@amazon.de>
 <c9dd6726-2783-2dfd-14d1-5cec6f69f051@redhat.com>
 <bce2aee1-bfac-0640-066b-068fa5f12cf8@amazon.de>
 <6edd5e08-92c2-40ff-57be-37b92d1ca2bc@redhat.com>
 <47eb1a4a-d015-b573-d773-e34e578ad753@amazon.de>
From:   Paolo Bonzini <pbonzini@redhat.com>
Subject: Re: [PATCH] KVM: VMX: Forbid userspace MSR filters for x2APIC
Message-ID: <ac2be818-04c8-6027-870c-184148e511ef@redhat.com>
Date:   Tue, 20 Oct 2020 13:49:31 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.3.1
MIME-Version: 1.0
In-Reply-To: <47eb1a4a-d015-b573-d773-e34e578ad753@amazon.de>
Content-Type: text/plain; charset=utf-8
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 20/10/20 12:52, Alexander Graf wrote:
>>
>> Yes, but the idea is that x2apic registers are always allowed, even
>> overriding default_allow, and therefore it makes no sense to have them
>> in a range.  The patch is only making things fail early for userspace,
>> the policy is defined by Sean's patch.
> 
> I don't think we should fail on the following:
> 
> {
>     default_allow: false,
>     ranges: [
>         {
>             flags: KVM_MSR_FILTER_READ,
>             nmsrs: 4096,
>             base: 0,
>             bitmap: { 1, 1, 1, 1, [...] },
>         },
>         {
>             flags: KVM_MSR_FILTER_READ,
>             nmsrs: 4096,
>             base: 0xc0000000,
>             bitmap: { 1, 1, 1, 1, [...] },
>         },
>     ],
> }
> 
> as a way to say "everything in normal ranges is allowed, the rest please
> deflect". Or even just to set default policies with less ranges.
> 
> Or to say it differently: Why can't we just check explicitly after
> setting up all filter lists whether x2apic MSRs are *denied*? If so,
> clear the filter and return -EINVAL.

Hmm, if you start looking at the bitmaps setting up default-deny
policies correctly is almost impossible :/ because you'd have to ensure
that you have at least one range covering the x2apic MSRs.  I'll just
document that x2APIC MSRs ignore the filter.

Paolo