Re: [PATCH 6.12.y] KVM: x86: Always initialize EFER reserved bits on vendor initialization

[Sean Christopherson <seanjc@google.com>]

On Fri, Sep 11, 2026, Sasha Levin wrote:
> From: Yosry Ahmed <yosry@kernel.org>
> 
> [ Upstream commit a4b8acd428ba3f03952ef309d7f438dbc357557b ]
> 
> EFER reserved bits are statically initialized, and do not reset if a
> vendor module is re-loaded. For example, loading kvm_amd with nested=1
> removes EFER.SVME (and potentially EFER.LMSLE) from the reserved bits.
> Reloading kvm_amd with nested=0 does not add them back, allowing
> userspace to set EFER.SVME with nested=0.
> 
> Re-initializing EFER reserved bits before configuring them on vendor
> initialization.
> 
> Cc: stable@vger.kernel.org
> Signed-off-by: Yosry Ahmed <yosry@kernel.org>
> Link: https://patch.msgid.link/20260713181020.2735367-4-yosry@kernel.org
> Signed-off-by: Sean Christopherson <seanjc@google.com>
> [ placed EFER initialization directly in x86.c’s kvm_x86_vendor_init() because this branch lacks kvm_setup_efer_caps() ]
> Signed-off-by: Sasha Levin <sashal@kernel.org>
> ---
>  arch/x86/kvm/x86.c | 21 +++++++++++----------
>  1 file changed, 11 insertions(+), 10 deletions(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 9e7c9e3edb5b4..84f8120ac31f8 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -108,16 +108,7 @@ EXPORT_SYMBOL_GPL(kvm_host);
>  #define emul_to_vcpu(ctxt) \
>  	((struct kvm_vcpu *)(ctxt)->vcpu)
>  
> -/* EFER defaults:
> - * - enable syscall per default because its emulated by KVM
> - * - enable LME and LMA per default on 64 bit KVM
> - */
> -#ifdef CONFIG_X86_64
> -static
> -u64 __read_mostly efer_reserved_bits = ~((u64)(EFER_SCE | EFER_LME | EFER_LMA));
> -#else
> -static u64 __read_mostly efer_reserved_bits = ~((u64)EFER_SCE);
> -#endif
> +static u64 __read_mostly efer_reserved_bits;
>  
>  static u64 __read_mostly cr4_reserved_bits = CR4_RESERVED_BITS;
>  
> @@ -9818,6 +9809,16 @@ int kvm_x86_vendor_init(struct kvm_x86_init_ops *ops)
>  	if (boot_cpu_has(X86_FEATURE_ARCH_CAPABILITIES))
>  		rdmsrl(MSR_IA32_ARCH_CAPABILITIES, kvm_host.arch_capabilities);
>  
> +	/* EFER defaults:
> +	 * - enable syscall per default because its emulated by KVM
> +	 * - enable LME and LMA per default on 64 bit KVM
> +	 */
> +#ifdef CONFIG_X86_64
> +	efer_reserved_bits = ~((u64)(EFER_SCE | EFER_LME | EFER_LMA));
> +#else
> +	efer_reserved_bits = ~((u64)EFER_SCE);
> +#endif

This is going to conflict with "KVM: x86: Disallow EFER.LME and EFER.LMA if long
mode is not supported"[*], it probably makes sense to send a version on top of
that, both to avoid a conflict when applying and to make it obvious 6.12.y will
have the same patch ordering as upstream (though that probably doesn't matter in
this case).

[*] https://lore.kernel.org/all/20260911120502.489632-1-sashal@kernel.org

> +
>  	r = ops->hardware_setup();
>  	if (r != 0)
>  		goto out_mmu_exit;
> -- 
> 2.53.0
>