Skip to contentSkip to conversation

KVM: VMX: Set VMENTER_L1D_FLUSH_NOT_REQUIRED if !X86_BUG_L1TF

PATCH Applied Waiman Long 3 participants 5 messages Aug 26, 2019 – Sep 27, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: VMX: Set VMENTER_L1D_FLUSH_NOT_REQUIRED if !X86_BUG_L1TF

The l1tf_vmx_mitigation is only set to VMENTER_L1D_FLUSH_NOT_REQUIRED when the ARCH_CAPABILITIES MSR indicates that L1D flush is not required. However, if the CPU is not affected by L1TF, l1tf_vmx_mitigation will still be set to VMENTER_L1D_FLUSH_AUTO. This is certainly not the best option for a !X86_BUG_L1TF CPU.

So force l1tf_vmx_mitigation to VMENTER_L1D_FLUSH_NOT_REQUIRED to make it more explicit in case users are checking the vmentry_l1d_flush parameter.

---
 arch/x86/kvm/vmx/vmx.c | 2 ++
 1 file changed, 2 insertions(+)
arch/x86/kvm/vmx/vmx.c +2 -0
index 42ed3faa6af8..a00ce3d6bbfd 100644
@@ -7896,6 +7896,8 @@ static int __init vmx_init(void)
 			vmx_exit();
 			return r;
 		}
+	} else {
+		l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED;
 	}
 
 #ifdef CONFIG_KEXEC_CORE
Signature · 1 line
2.18.1
#1 lore raw
Quoted 19 lines · Waiman Long

The l1tf_vmx_mitigation is only set to VMENTER_L1D_FLUSH_NOT_REQUIRED when the ARCH_CAPABILITIES MSR indicates that L1D flush is not required. However, if the CPU is not affected by L1TF, l1tf_vmx_mitigation will still be set to VMENTER_L1D_FLUSH_AUTO. This is certainly not the best option for a !X86_BUG_L1TF CPU.

So force l1tf_vmx_mitigation to VMENTER_L1D_FLUSH_NOT_REQUIRED to make it more explicit in case users are checking the vmentry_l1d_flush parameter.

---
 arch/x86/kvm/vmx/vmx.c | 2 ++
 1 file changed, 2 insertions(+)
arch/x86/kvm/vmx/vmx.c +0 -0
index 42ed3faa6af8..a00ce3d6bbfd 100644

arch/x86/kvm/vmx/vmx.c static int __init vmx_init(void) · line 7903

Quoted 8 lines
arch/x86/kvm/vmx/vmx.c +2 -0
@@ -7896,6 +7896,8 @@ static int __init vmx_init(void)
 			vmx_exit();
 			return r;
 		}
+	} else {
+		l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED;
 	}
 
 #ifdef CONFIG_KEXEC_CORE

Ping. Any comment on that one?

Cheers,
Longman

maintainer
#2 lore raw
Quoted 27 lines · Waiman Long
Quoted 26 lines · Waiman Long

The l1tf_vmx_mitigation is only set to VMENTER_L1D_FLUSH_NOT_REQUIRED when the ARCH_CAPABILITIES MSR indicates that L1D flush is not required. However, if the CPU is not affected by L1TF, l1tf_vmx_mitigation will still be set to VMENTER_L1D_FLUSH_AUTO. This is certainly not the best option for a !X86_BUG_L1TF CPU.

So force l1tf_vmx_mitigation to VMENTER_L1D_FLUSH_NOT_REQUIRED to make it more explicit in case users are checking the vmentry_l1d_flush parameter.

---
 arch/x86/kvm/vmx/vmx.c | 2 ++
 1 file changed, 2 insertions(+)
arch/x86/kvm/vmx/vmx.c +2 -0
index 42ed3faa6af8..a00ce3d6bbfd 100644
@@ -7896,6 +7896,8 @@ static int __init vmx_init(void)
 			vmx_exit();
 			return r;
 		}
+	} else {
+		l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED;
 	}
 
Quoted 3 lines

arch/x86/kvm/vmx/vmx.c static int __init vmx_init(void) · line 7903

Quoted 1 line
arch/x86/kvm/vmx/vmx.c +0 -0
 #ifdef CONFIG_KEXEC_CORE

Ping. Any comment on that one?

I'd move that logic with the if (boot_cpu_has(X86_BUG_L1TF)) check inside vmx_setup_l1d_flush() so that I have this:

        if (!boot_cpu_has_bug(X86_BUG_L1TF)) {
                l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED;
                return 0;
        }

	if (!enable_ept) {
		...

	}

inside the function and outside am left with:

	r = vmx_setup_l1d_flush(vmentry_l1d_flush_param);
        if (r) {
		vmx_exit();
                return r;
	}

only. This way I'm concentrating the whole l1tf_vmx_mitigation picking apart in one place.

Also, note that X86_BUG flags are checked with boot_cpu_has_bug() even if it boils down to the same thing now.

Thx.

Signature · 4 lines
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
maintainer
#3 lore raw
Quoted 15 lines · Borislav Petkov

I'd move that logic with the if (boot_cpu_has(X86_BUG_L1TF)) check inside vmx_setup_l1d_flush() so that I have this:

        if (!boot_cpu_has_bug(X86_BUG_L1TF)) {
                l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED;
                return 0;
        }

	if (!enable_ept) {
		...

	}

inside the function and outside am left with:

Quoted 8 lines
	r = vmx_setup_l1d_flush(vmentry_l1d_flush_param);
        if (r) {
		vmx_exit();
                return r;
	}

only. This way I'm concentrating the whole l1tf_vmx_mitigation picking apart in one place.

Right you are, I'm sending v2.

Paolo

maintainer Applied
#4 lore raw
In reply to Waiman Long
Quoted 20 lines · Waiman Long

The l1tf_vmx_mitigation is only set to VMENTER_L1D_FLUSH_NOT_REQUIRED when the ARCH_CAPABILITIES MSR indicates that L1D flush is not required. However, if the CPU is not affected by L1TF, l1tf_vmx_mitigation will still be set to VMENTER_L1D_FLUSH_AUTO. This is certainly not the best option for a !X86_BUG_L1TF CPU.

So force l1tf_vmx_mitigation to VMENTER_L1D_FLUSH_NOT_REQUIRED to make it more explicit in case users are checking the vmentry_l1d_flush parameter.

---
 arch/x86/kvm/vmx/vmx.c | 2 ++
 1 file changed, 2 insertions(+)
arch/x86/kvm/vmx/vmx.c +0 -0
index 42ed3faa6af8..a00ce3d6bbfd 100644

arch/x86/kvm/vmx/vmx.c static int __init vmx_init(void) · line 7903

Quoted 8 lines
arch/x86/kvm/vmx/vmx.c +2 -0
@@ -7896,6 +7896,8 @@ static int __init vmx_init(void)
 			vmx_exit();
 			return r;
 		}
+	} else {
+		l1tf_vmx_mitigation = VMENTER_L1D_FLUSH_NOT_REQUIRED;
 	}
 
 #ifdef CONFIG_KEXEC_CORE

Queued (for -rc2), thanks.

Paolo