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: <18042cd3-c1fe-a54a-8682-2b5d1523c3d9@redhat.com>
Date:   Fri, 11 Feb 2022 12:21:18 +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: [PATCH 05/12] KVM: MMU: avoid NULL-pointer dereference on page
 freeing bugs
Content-Language: en-US
To:     Sean Christopherson <seanjc@google.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        vkuznets@redhat.com, mlevitsk@redhat.com, dmatlack@google.com
References: <20220209170020.1775368-1-pbonzini@redhat.com>
 <20220209170020.1775368-6-pbonzini@redhat.com> <YgWsoKskWnahgR8j@google.com>
From:   Paolo Bonzini <pbonzini@redhat.com>
In-Reply-To: <YgWsoKskWnahgR8j@google.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 2/11/22 01:24, Sean Christopherson wrote:
>>   
>>   	sp = to_shadow_page(*root_hpa & PT64_BASE_ADDR_MASK);
>> +	if (WARN_ON(!sp))
>
> Should this be KVM_BUG_ON()?  I.e. when you triggered these, would continuing on
> potentially corrupt guest data, or was it truly benign-ish?

It only triggered on the mode_switch SVM unit test (with npt=0); so, in 
a very small test which just hung after the bug.  The WARN however was 
the 10-minute difference between rmmod and reboot...

I didn't use KVM_BUG_ON because we're in a pretty deep call stack 
(kvm_mmu_new_pgd, itself called from nested vmentry) and all sort of 
stuff will happen before bailing out.  My mental model is to use 
KVM_BUG_ON in situations for which error propagation is possible and clean.

Paolo