Re: [PATCH 1/3] x86/bugs: Use VM_CLEAR_CPU_BUFFERS in VMX as well

[Brendan Jackman <jackmanb@google.com>]

On Wed Oct 29, 2025 at 9:26 PM UTC, Pawan Gupta wrote:
> TSA mitigation:
>
>   d8010d4ba43e ("x86/bugs: Add a Transient Scheduler Attacks mitigation")
>
> introduced VM_CLEAR_CPU_BUFFERS for guests on AMD CPUs. Currently on Intel
> CLEAR_CPU_BUFFERS is being used for guests which has a much broader scope
> (kernel->user also).
>
> Make mitigations on Intel consistent with TSA. This would help handling the
> guest-only mitigations better in future.
>
> Signed-off-by: Pawan Gupta <pawan.kumar.gupta@linux.intel.com>
> ---
>  arch/x86/kernel/cpu/bugs.c | 9 +++++++--
>  arch/x86/kvm/vmx/vmenter.S | 3 ++-
>  2 files changed, 9 insertions(+), 3 deletions(-)
>
> diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c
> index d7fa03bf51b4517c12cc68e7c441f7589a4983d1..6d00a9ea7b4f28da291114a7a096b26cc129b57e 100644
> --- a/arch/x86/kernel/cpu/bugs.c
> +++ b/arch/x86/kernel/cpu/bugs.c
> @@ -194,7 +194,7 @@ DEFINE_STATIC_KEY_FALSE(switch_mm_cond_l1d_flush);
>  
>  /*
>   * Controls CPU Fill buffer clear before VMenter. This is a subset of
> - * X86_FEATURE_CLEAR_CPU_BUF, and should only be enabled when KVM-only
> + * X86_FEATURE_CLEAR_CPU_BUF_VM, and should only be enabled when KVM-only
>   * mitigation is required.
>   */

So if I understand correctly with this patch the aim is:

X86_FEATURE_CLEAR_CPU_BUF means verw before exit to usermode

X86_FEATURE_CLEAR_CPU_BUF_VM means unconditional verw before VM Enter

cpu_buf_vm_clear[_mmio_only] means verw before VM Enter for
MMIO-capable guests.

Since this is being cleaned up can we also:

- Update the definition of X86_FEATURE_CLEAR_CPU_BUF in cpufeatures.h to
  say what context it applies to (now it's specifically exit to user)

- Clear up how verw_clear_cpu_buf_mitigation_selected relates to these
  two flags. Thinking aloud here... it looks like this is set:

  - If MDS mitigations are on, meaning both flags are set

  - If TAA mitigations are on, meaning both flags are set

  - If MMIO mitigations are on, and the CPU has MDS or TAA. In this case
    both flags are set, but this causality is messier.

  - If RFDS mitigations are on and supported, meaning both flags are set

  So if I'm reading this correctly whenever
  verw_clear_cpu_buf_mitigation_selected we should expect both flags
  enabled. So I think all that's needed is to add a reference to
  X86_FEATURE_CLEAR_CPU_BUF_VM to the comment?

I think we also need to update the assertion of vmx->disable_fb_clear?

Anyway thanks this seems like a very clear improvement to me.