Re: [PATCH v2 2/3] KVM: VMX: validate individual bits of guest MSR_IA32_FEATURE_CONTROL

From: Haozhong Zhang <haozhong.zhang@intel.com>
To: Paolo Bonzini <pbonzini@redhat.com>
Cc: kvm@vger.kernel.org, rkrcmar@redhat.com,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, "H . Peter Anvin" <hpa@zytor.com>,
	x86@kernel.org, linux-kernel@vger.kernel.org,
	Gleb Natapov <gleb@kernel.org>, Boris Petkov <bp@suse.de>,
	Tony Luck <tony.luck@intel.com>,
	Andi Kleen <andi.kleen@intel.com>,
	Ashok Raj <ashok.raj@intel.com>
Subject: Re: [PATCH v2 2/3] KVM: VMX: validate individual bits of guest MSR_IA32_FEATURE_CONTROL
Date: Thu, 16 Jun 2016 19:16:55 +0800	[thread overview]
Message-ID: <20160616111655.ltcgle2cju2fksrx@hz-desktop> (raw)
In-Reply-To: <088f9fcb-a648-05d4-530a-790cb084bd09@redhat.com>

On 06/16/16 12:01, Paolo Bonzini wrote:
> 
> 
> On 16/06/2016 08:05, Haozhong Zhang wrote:
> > +/*
> > + * FEATURE_CONTROL_LOCKED is added/removed automatically by
> > + * feature_control_valid_bits_add/del(), so it's not included here.
> > + */
> > +#define FEATURE_CONTROL_MAX_VALID_BITS \
> > +	FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX
> > +
> > +static void feature_control_valid_bits_add(struct kvm_vcpu *vcpu, uint64_t bits)
> > +{
> > +	ASSERT(!(bits & ~FEATURE_CONTROL_MAX_VALID_BITS));
> 
> The KVM-specific ASSERT macro should go.  You can use WARN_ON.
>

will change to WARN_ON.

> However, I think FEATURE_CONTROL_LOCKED should always be writable.  If
> you change that, it's simpler to just do |= and &= in the caller.
>

These two functions (add/del) are to prevent callers from forgetting
setting/clearing FEATURE_CONTROL_LOCKED in
msr_ia32_feature_control_valid_bits: it should be set if any feature
bit is set, and be cleared if all feature bits are cleared. The second
rule could relaxed as we can always present MSR_IA32_FEATURE_CONTROL
to guest.

I'm okey to let callers take care for the locked bit.

> > +	to_vmx(vcpu)->msr_ia32_feature_control_valid_bits |=
> > +		bits | FEATURE_CONTROL_LOCKED;
> > +}
> > +
> > +static void feature_control_valid_bits_del(struct kvm_vcpu *vcpu, uint64_t bits)
> > +{
> > +	uint64_t *valid_bits =
> > +		&to_vmx(vcpu)->msr_ia32_feature_control_valid_bits;
> > +	ASSERT(!(bits & ~FEATURE_CONTROL_MAX_VALID_BITS));
> > +	*valid_bits &= ~bits;
> > +	if (!(*valid_bits & ~FEATURE_CONTROL_LOCKED))
> > +		*valid_bits = 0;
> > +}
> > +
> >  #define VMCS12_OFFSET(x) offsetof(struct vmcs12, x)
> >  #define FIELD(number, name)	[number] = VMCS12_OFFSET(name)
> >  #define FIELD64(number, name)	[number] = VMCS12_OFFSET(name), \
> > @@ -2864,6 +2897,14 @@ static int vmx_get_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata)
> >  	return 0;
> >  }
> >  
> > +static inline bool vmx_feature_control_msr_valid(struct kvm_vcpu *vcpu,
> > +						 uint64_t val)
> > +{
> > +	uint64_t valid_bits = to_vmx(vcpu)->msr_ia32_feature_control_valid_bits;
> > +
> > +	return valid_bits && !(val & ~valid_bits);
> > +}
> >  /*
> >   * Reads an msr value (of 'msr_index') into 'pdata'.
> >   * Returns 0 on success, non-0 otherwise.
> > @@ -2906,7 +2947,7 @@ static int vmx_get_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
> >  		msr_info->data = vmcs_read64(GUEST_BNDCFGS);
> >  		break;
> >  	case MSR_IA32_FEATURE_CONTROL:
> > -		if (!nested_vmx_allowed(vcpu))
> > +		if (!vmx_feature_control_msr_valid(vcpu, 0))
> 
> You can remove this if completely in patch 1.  It's okay to make the MSR
> always available.
>

But then it also allows all bits to be set by guests, even though some
features are not available. Though this problem already exists before
Patch 1, I prefer to leave it as was in Patch 1 and fix it here.

Haozhong