Re: [PATCH v4 1/2] KVM: VMX: Cleanup VMX basic information defines and usages

[Huang, Kai]

On Wed, 2024-01-17 at 16:34 +0000, Li, Xin3 wrote:
> > > +#define VMX_BASIC_FEATURES_MASK			\
> > > +	(VMX_BASIC_DUAL_MONITOR_TREATMENT |	\
> > > +	 VMX_BASIC_INOUT |			\
> > > +	 VMX_BASIC_TRUE_CTLS)
> > > +
> > > +#define VMX_BASIC_RESERVED_BITS			\
> > > +	(GENMASK_ULL(63, 56) | GENMASK_ULL(47, 45) | BIT_ULL(31))
> > 
> > When we add a new feature (e.g., in CET series, bit 56 is added), the above
> > two macros need to be modified.
> > 
> > Would it be better to use a macro for bits exempt from the bitwise check below
> > e.g.,
> > 
> > #define VMX_BASIC_MULTI_BITS_FEATURES_MASK
> > 
> > 	(GENMASK_ULL(53, 50) | GENMASK_ULL(44, 32) | GENMASK_ULL(30, 0))
> > 
> > and do
> > 	if (!is_bitwise_subset(vmx_basic, data,
> > 			       ~VMX_BASIC_MULTI_BITS_FEATURES_MASK)
> > 
> > then we don't need to change the macro when adding new features.
> 
> Sounds a good idea to me, and just need to add comments about why.
> 
> > 
> > > +
> > > static int vmx_restore_vmx_basic(struct vcpu_vmx *vmx, u64 data)
> > > {
> > > -	const u64 feature_and_reserved =
> > > -		/* feature (except bit 48; see below) */
> > > -		BIT_ULL(49) | BIT_ULL(54) | BIT_ULL(55) |
> > > -		/* reserved */
> > > -		BIT_ULL(31) | GENMASK_ULL(47, 45) | GENMASK_ULL(63, 56);
> > > 	u64 vmx_basic = vmcs_config.nested.basic;
> > > 
> > > -	if (!is_bitwise_subset(vmx_basic, data, feature_and_reserved))
> > > +	static_assert(!(VMX_BASIC_FEATURES_MASK &
> > VMX_BASIC_RESERVED_BITS));
> > > +
> > > +	if (!is_bitwise_subset(vmx_basic, data,
> > > +			       VMX_BASIC_FEATURES_MASK |
> > VMX_BASIC_RESERVED_BITS))
> > > 		return -EINVAL;
> > > 
> > > 	/*
> > > 	 * KVM does not emulate a version of VMX that constrains physical
> > > 	 * addresses of VMX structures (e.g. VMCS) to 32-bits.
> > > 	 */
> > > -	if (data & BIT_ULL(48))
> > > +	if (data & VMX_BASIC_32BIT_PHYS_ADDR_ONLY)
> > > 		return -EINVAL;
> > 
> > Side topic:
> > 
> > Actually, there is no need to handle bit 48 as a special case. If we add bit 48
> > to VMX_BASIC_FEATURES_MASK, the bitwise check will fail if bit 48 of @data is 1.
> 
> Good point!  This is also what you suggested above.
> 

Please try to avoid mixing things together in one patch.  If you want to do
above, could you please do it in a separate patch so that can be reviewed
separately?

E.g., people who have reviewed or acked this patch may not be interested in the
new (logically separate) things.