linux-kernel - Re: [PATCH v3 2/2] KVM: VMX: untangle VMXON revision

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <6ce77a63-0505-36f8-3a92-f0f6b275fed1@oracle.com>
Date:   Fri, 6 Mar 2020 17:34:10 -0800
From:   Krish Sadhukhan <krish.sadhukhan@...cle.com>
To:     Sean Christopherson <sean.j.christopherson@...el.com>
Cc:     Vitaly Kuznetsov <vkuznets@...hat.com>,
        Paolo Bonzini <pbonzini@...hat.com>,
        Jim Mattson <jmattson@...gle.com>,
        Wanpeng Li <wanpengli@...cent.com>, kvm@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 2/2] KVM: VMX: untangle VMXON revision_id setting when
 using eVMCS


On 3/6/20 4:28 PM, Sean Christopherson wrote:
> On Fri, Mar 06, 2020 at 03:57:25PM -0800, Krish Sadhukhan wrote:
>> On 3/6/20 3:07 PM, Sean Christopherson wrote:
>>> On Fri, Mar 06, 2020 at 02:20:13PM -0800, Krish Sadhukhan wrote:
>>>>> @@ -2599,7 +2607,7 @@ void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs)
>>>>>   int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs)
>>>>>   {
>>>>> -	loaded_vmcs->vmcs = alloc_vmcs(false);
>>>>> +	loaded_vmcs->vmcs = alloc_vmcs(VMCS_REGION);
>>>>>   	if (!loaded_vmcs->vmcs)
>>>>>   		return -ENOMEM;
>>>>> @@ -2652,25 +2660,13 @@ static __init int alloc_vmxon_regions(void)
>>>>>   	for_each_possible_cpu(cpu) {
>>>>>   		struct vmcs *vmcs;
>>>>> -		vmcs = alloc_vmcs_cpu(false, cpu, GFP_KERNEL);
>>>>> +		/* The VMXON region is really just a special type of VMCS. */
>>>> Not sure if this is the right way to correlate the two.
>>>>
>>>> AFAIU, the SDM calls VMXON region as a memory area that holds the VMCS data
>>>> structure and it calls VMCS the data structure that is used by software to
>>>> switch between VMX root-mode and not-root-mode. So VMXON is a memory area
>>>> whereas VMCS is the structure of the data that resides in that memory area.
>>>>
>>>> So if we follow this interpretation, your enum should rather look like,
>>>>
>>>> enum vmcs_type {
>>>> +    VMCS,
>>>> +    EVMCS,
>>>> +    SHADOW_VMCS
>>> No (to the EVMCS suggestion), because this allocation needs to happen for
>>> !eVMCS.  The SDM never explictly calls the VMXON region a VMCS, but it's
>>> just being coy.  E.g. VMCLEAR doesn't fail if you point it at random
>>> memory, but point it at the VMXON region and it yells.
>>>
>>> We could call it VMXON_VMCS if that helps.
>> Are you saying,
>>
>> + enum vmcs_type {
>> +     VMXON_REGION,
>> +     VMXON_VMCS,
>> +     SHADOW_VMCS_REGION,
>> +};
>>
>> ?
>>
>> In that case, "VMXON_REGION" and "VMXON_VMCS" are no different according to
>> your explanation.
>    enum vmcs_type {
> 	VMXON_VMCS,
> 	VMCS,
> 	SHADOW_VMCS,
>    };


It looks reasonable.


>
> alloc_vmcs_cpu() does more than just allocate the memory, it also
> initializes the data structure, e.g. "allocate and initalize a VMXON VMCS",
>
>>>   The SDM does call the memory
>>> allocation for regular VMCSes a "VMCS region":
>>>
>>>    A logical processor associates a region in memory with each VMCS. This
>>>    region is called the VMCS region.
>>>
>>> I don't think I've ever heard anyone differentiate that two though, i.e.
>>> VMCS is used colloquially to mean both the data structure itself and the
>>> memory region containing the data structure.
>>>
>>>>> +		vmcs = alloc_vmcs_cpu(VMXON_REGION, cpu, GFP_KERNEL);
>>>>>   		if (!vmcs) {
>>>>>   			free_vmxon_regions();
>>>>>   			return -ENOMEM;
>>>>>   		}
>>>>> -		/*
>>>>> -		 * When eVMCS is enabled, alloc_vmcs_cpu() sets
>>>>> -		 * vmcs->revision_id to KVM_EVMCS_VERSION instead of
>>>>> -		 * revision_id reported by MSR_IA32_VMX_BASIC.
>>>>> -		 *
>>>>> -		 * However, even though not explicitly documented by
>>>>> -		 * TLFS, VMXArea passed as VMXON argument should
>>>>> -		 * still be marked with revision_id reported by
>>>>> -		 * physical CPU.
>>>>> -		 */
>>>>> -		if (static_branch_unlikely(&enable_evmcs))
>>>>> -			vmcs->hdr.revision_id = vmcs_config.revision_id;
>>>>> -
>>>>>   		per_cpu(vmxarea, cpu) = vmcs;
>>>>>   	}
>>>>>   	return 0;
>>>>> diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
>>>>> index e64da06c7009..a5eb92638ac2 100644
>>>>> --- a/arch/x86/kvm/vmx/vmx.h
>>>>> +++ b/arch/x86/kvm/vmx/vmx.h
>>>>> @@ -489,16 +489,22 @@ static inline struct pi_desc *vcpu_to_pi_desc(struct kvm_vcpu *vcpu)
>>>>>   	return &(to_vmx(vcpu)->pi_desc);
>>>>>   }
>>>>> -struct vmcs *alloc_vmcs_cpu(bool shadow, int cpu, gfp_t flags);
>>>>> +enum vmcs_type {
>>>>> +	VMXON_REGION,
>>>>> +	VMCS_REGION,
>>>>> +	SHADOW_VMCS_REGION,
>>>>> +};
>>>>> +
>>>>> +struct vmcs *alloc_vmcs_cpu(enum vmcs_type type, int cpu, gfp_t flags);
>>>>>   void free_vmcs(struct vmcs *vmcs);
>>>>>   int alloc_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
>>>>>   void free_loaded_vmcs(struct loaded_vmcs *loaded_vmcs);
>>>>>   void loaded_vmcs_init(struct loaded_vmcs *loaded_vmcs);
>>>>>   void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs);
>>>>> -static inline struct vmcs *alloc_vmcs(bool shadow)
>>>>> +static inline struct vmcs *alloc_vmcs(enum vmcs_type type)
>>>>>   {
>>>>> -	return alloc_vmcs_cpu(shadow, raw_smp_processor_id(),
>>>>> +	return alloc_vmcs_cpu(type, raw_smp_processor_id(),
>>>>>   			      GFP_KERNEL_ACCOUNT);
>>>>>   }