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;
Message-ID: <58a9410a-62f2-5a35-2ce3-f5203912465e@grsecurity.net>
Date:   Wed, 5 Apr 2023 20:31:50 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101
 Thunderbird/102.9.0
Subject: Re: [PATCH] KVM: x86/mmu: Refresh CR0.WP prior to checking for
 emulated permission faults
Content-Language: en-US, de-DE
To:     Sean Christopherson <seanjc@google.com>
Cc:     Paolo Bonzini <pbonzini@redhat.com>, kvm@vger.kernel.org,
        linux-kernel@vger.kernel.org
References: <20230405002608.418442-1-seanjc@google.com>
 <a1ed2308-b521-14c0-a118-19c1afffd1d6@grsecurity.net>
 <ZC2HfNLa7sLewJtl@google.com>
From:   Mathias Krause <minipli@grsecurity.net>
In-Reply-To: <ZC2HfNLa7sLewJtl@google.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk


On 05.04.23 16:36, Sean Christopherson wrote:
> On Wed, Apr 05, 2023, Mathias Krause wrote:
>> On 05.04.23 02:26, Sean Christopherson wrote:
>>> +void __kvm_mmu_refresh_passthrough_bits(struct kvm_vcpu *vcpu,
>>> +					struct kvm_mmu *mmu)
>>> +{
>>> +	const bool cr0_wp = kvm_is_cr0_bit_set(vcpu, X86_CR0_WP);
>>> +
>>> +	BUILD_BUG_ON((KVM_MMU_CR0_ROLE_BITS & KVM_POSSIBLE_CR0_GUEST_BITS) != X86_CR0_WP);
>>
>>> +	BUILD_BUG_ON((KVM_MMU_CR4_ROLE_BITS & KVM_POSSIBLE_CR4_GUEST_BITS));
>>
>> Just curious, this should assert that we don't run into similar issues
>> if we make more bits of CR4 guest owned?
> 
> Yes?  I'm not sure what you're asking.  BUILD_BUG_ON() is a just more flexible
> version of stiatic_assert(); it only requires that the inputs be compile-time
> constants, not purely "static>
> he above throws an error at compile-time if there is new overlap between the
> CR{0,4} MMU role bits and the possible guest-owned bits.  E.g. adding SMEP to the
> possible guest-owned CR4 bits yields:

Yes, I was just asking about the reasoning behind it as it, obviously,
isn't a problem with the current code. But I ran into it while doing
backports, so thanks for adding it :D

> 
> [...]