Re: [PATCH v10 21/28] KVM: nVMX: Handle FRED VMCS fields in nested VMX context

[Binbin Wu]

On 9/12/2026 5:36 AM, Sohil Mehta wrote:
> From: "Xin Li (Intel)" <xin@xxxxxxxxx>
> 
> Extend nested VMX context management to include FRED-related VMCS fields,
> enabling proper handling of FRED state during nested virtualization.

No strong preference, but for me this is a bit too thin to describe the
overall change of this patch.

> 
> Because KVM always sets SECONDARY_VM_EXIT_SAVE_IA32_FRED, FRED MSRs are
> always saved to vmcs02.  However an L1 VMM may choose to clear this bit,
> i.e., not to save FRED MSRs to vmcs12.  This is not a problem when the L1
> VMM sets SECONDARY_VM_EXIT_LOAD_IA32_FRED, as KVM then immediately loads
> host FRED MSRs of vmcs12 to guest FRED MSRs of vmcs01.  However if the L1
> VMM clears SECONDARY_VM_EXIT_LOAD_IA32_FRED, KVM should retain FRED MSRs
> to run the L1 VMM.
> 
> To propagate guest FRED MSRs from vmcs02 to vmcs01, save them in
> sync_vmcs02_to_vmcs12() regardless of whether
> SECONDARY_VM_EXIT_SAVE_IA32_FRED is set in vmcs12.  Then, use the saved
> values to set guest FRED MSRs in vmcs01 within load_vmcs12_host_state()
> when !nested_cpu_load_host_fred_state().
> 
> Signed-off-by: Xin Li (Intel) <xin@xxxxxxxxx>
> Signed-off-by: Sohil Mehta <sohil.mehta@xxxxxxxxx>
> ---
> v10:
>  - Simplify nested FRED MSRs update with read/write helpers (Chao).
>  - Add other FRED MSRS to MSR bitmap nested filtering (Chao).
>  - Add {ORIGINAL,INJECTED}_EVENT_DATA to shadow fields in a separate
>    patch (Sean).
>  - Add a new parameter to load_vmcs12_host_state() to indicate the call
>    is from a failed VM entry.
>  - Update exception status member variable nested to is_nested (Sean).
> 
> v9:
>  - Rebase to kvm-x86/next.
>  - Guard FRED state save/restore with guest_cpu_cap_has(vcpu,
>    X86_FEATURE_FRED) (syzbot & Chao).
> ---

[...]

> diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
> index 14fd3f57d5c1..cc1e4038cd44 100644
> --- a/arch/x86/kvm/vmx/vmx.h
> +++ b/arch/x86/kvm/vmx/vmx.h
> @@ -182,6 +182,29 @@ struct nested_vmx {
>  	u64 pre_vmenter_ssp;
>  	u64 pre_vmenter_ssp_tbl;
>  
> +	/*
> +	 * Used to snapshot FRED MSRs that may NOT be saved to vmcs12 as specified
> +	 * in the VM-Exit controls of vmcs12 configured by L1 VMM.
> +	 *
> +	 * FRED MSRs are *always* saved into vmcs02 because KVM always sets
> +	 * SECONDARY_VM_EXIT_SAVE_IA32_FRED.  However an L1 VMM may choose to clear
> +	 * this bit, resulting in FRED MSRs not being propagated to vmcs12 from
> +	 * vmcs02.  When the L1 VMM sets SECONDARY_VM_EXIT_LOAD_IA32_FRED, this is
> +	 * not a problem, since KVM then immediately loads the host FRED MSRs of
> +	 * vmcs12 to the guest FRED MSRs of vmcs01.
> +	 *
> +	 * But if the L1 VMM clears SECONDARY_VM_EXIT_LOAD_IA32_FRED, KVM should
> +	 * retain the FRED MSRs, i.e., propagate the guest FRED MSRs of vmcs02 to
> +	 * the guest FRED MSRs of vmcs01.
> +	 *
> +	 * This structure stores guest FRED MSRs that an L1 VMM opts not to save
> +	 * during VM-Exits from L2 to L1.  These MSRs may still be retained for
> +	 * running the L1 VMM if SECONDARY_VM_EXIT_LOAD_IA32_FRED is cleared in
> +	 * vmcs12.
> +	 */
> +	struct vmcs_fred_msrs pre_vmenter_fred_msrs;> +	struct vmcs_fred_msrs at_vmexit_fred_msrs;

I understand that the two fields are both belong to FRED, but the comment blocks
are for at_vmexit_fred_msrs, not for pre_vmenter_fred_msrs.

Should pre_vmenter_fred_msrs be put together with the other pre_vmenter_xxx,
since the comments above the pre_vmenter_xxx explains why these fields are
needed.
> +
>  	u16 vpid02;
>  	u16 last_vpid;
>