On Thu, Apr 10, 2025, Pawan Gupta wrote: > On Thu, Apr 10, 2025 at 01:39:28PM -0700, Sean Christopherson wrote: > > On Thu, Apr 10, 2025, Pawan Gupta wrote: > > > The static key mmio_stale_data_clear controls the KVM-only mitigation for > > > MMIO Stale Data vulnerability. Rename it to reflect its purpose. > > > > > > No functional change. > > > > > > Signed-off-by: Pawan Gupta <pawan.kumar.gupta@xxxxxxxxxxxxxxx> > > > --- > > > arch/x86/include/asm/nospec-branch.h | 2 +- > > > arch/x86/kernel/cpu/bugs.c | 16 ++++++++++------ > > > arch/x86/kvm/vmx/vmx.c | 2 +- > > > 3 files changed, 12 insertions(+), 8 deletions(-) > > > > > > diff --git a/arch/x86/include/asm/nospec-branch.h b/arch/x86/include/asm/nospec-branch.h > > > index 8a5cc8e70439e10aab4eeb5b0f5e116cf635b43d..c0474e2b741737dad129159adf3b5fc056b6097c 100644 > > > --- a/arch/x86/include/asm/nospec-branch.h > > > +++ b/arch/x86/include/asm/nospec-branch.h > > > @@ -561,7 +561,7 @@ DECLARE_STATIC_KEY_FALSE(mds_idle_clear); > > > > > > DECLARE_STATIC_KEY_FALSE(switch_mm_cond_l1d_flush); > > > > > > -DECLARE_STATIC_KEY_FALSE(mmio_stale_data_clear); > > > +DECLARE_STATIC_KEY_FALSE(cpu_buf_vm_clear); > > > > Could we tack on "if_mmio" or something? E.g. cpu_buf_vm_clear_if_mmio. FWIW, > > I don't love that name, so if anyone can come up with something better... > > Keeping it generic has an advantage that it plays nicely with "Attack vector > controls" series: > > https://lore.kernel.org/lkml/20250310164023.779191-1-david.kaplan@xxxxxxx/ > > The idea being to allow mitigations to be enabled/disabled based on > user-defined threat model. MDS/TAA mitigations may be able to take > advantage this KVM-only control. > > > I like the idea of tying the static key back to X86_FEATURE_CLEAR_CPU_BUF, but > > when looking at just the usage in KVM, "cpu_buf_vm_clear" doesn't provide any > > hints as to when/why KVM needs to clear buffers. > > Thats fair, can we cover that with a comment like below: > > --- > diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c > index c79720aad3df..cddad4a6eb46 100644 > --- a/arch/x86/kvm/vmx/vmx.c > +++ b/arch/x86/kvm/vmx/vmx.c > @@ -7358,6 +7358,10 @@ static noinstr void vmx_vcpu_enter_exit(struct kvm_vcpu *vcpu, > * mitigation for MDS is done late in VMentry and is still > * executed in spite of L1D Flush. This is because an extra VERW > * should not matter much after the big hammer L1D Flush. > + * > + * cpu_buf_vm_clear is used when system is not vulnerable to MDS/TAA, > + * but is affected by MMIO Stale Data that only needs mitigation > + * against a rogue guest. Would this be accurate? * Clear CPU buffers if the vCPU has access to host MMIO, and the CPU * is NOT vulnerable to MSD/TAA but is affected by MMIO Stale Data, as * indicated by cpu_buf_vm_clear. If the CPU is vulnerable to MDS/TAA, * CPU buffers will be cleared just before every VM-Entry. > */ > if (static_branch_unlikely(&vmx_l1d_should_flush)) > vmx_l1d_flush(vcpu);
