Web lists-archives.com

[PATCH 4.4 196/266] KVM: x86: SVM: Call x86_spec_ctrl_set_guest/host() with interrupts disabled

From: Thomas Gleixner <tglx@xxxxxxxxxxxxx>

commit 024d83cadc6b2af027e473720f3c3da97496c318 upstream.

Mikhail reported the following lockdep splat:

WARNING: possible irq lock inversion dependency detected
CPU 0/KVM/10284 just changed the state of lock:
  000000000d538a88 (&st->lock){+...}, at:

but this lock was taken by another, HARDIRQ-safe lock
in the past:


   and interrupts could create inverse lock ordering between them.

Possible interrupt unsafe locking scenario:

    CPU0                    CPU1
    ----                    ----
     *** DEADLOCK ***

The code path which connects those locks is:


In svm_vcpu_run() speculative_store_bypass_update() is called with
interupts enabled via x86_virt_spec_ctrl_set_guest/host().

This is actually a false positive, because GIF=0 so interrupts are
disabled even if IF=1; however, we can easily move the invocations of
x86_virt_spec_ctrl_set_guest/host() into the interrupt disabled region to
cure it, and it's a good idea to keep the GIF=0/IF=1 area as small
and self-contained as possible.

Fixes: 1f50ddb4f418 ("x86/speculation: Handle HT correctly on AMD")
Reported-by: Mikhail Gavrilov <mikhail.v.gavrilov@xxxxxxxxx>
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Tested-by: Mikhail Gavrilov <mikhail.v.gavrilov@xxxxxxxxx>
Cc: Joerg Roedel <joro@xxxxxxxxxx>
Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
Cc: Radim Krčmář <rkrcmar@xxxxxxxxxx>
Cc: Matthew Wilcox <willy@xxxxxxxxxxxxx>
Cc: Borislav Petkov <bp@xxxxxxx>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
Cc: Tom Lendacky <thomas.lendacky@xxxxxxx>
Cc: kvm@xxxxxxxxxxxxxxx
Cc: x86@xxxxxxxxxx
Signed-off-by: Paolo Bonzini <pbonzini@xxxxxxxxxx>
Signed-off-by: Ben Hutchings <ben@xxxxxxxxxxxxxxx>
Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
 arch/x86/kvm/svm.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -3928,8 +3928,6 @@ static void svm_vcpu_run(struct kvm_vcpu
-	local_irq_enable();
 	 * If this vCPU has touched SPEC_CTRL, restore the guest's value if
 	 * it's non-zero. Since vmentry is serialising on affected CPUs, there
@@ -3938,6 +3936,8 @@ static void svm_vcpu_run(struct kvm_vcpu
 	x86_spec_ctrl_set_guest(svm->spec_ctrl, svm->virt_spec_ctrl);
+	local_irq_enable();
 	asm volatile (
 		"push %%" _ASM_BP "; \n\t"
 		"mov %c[rbx](%[svm]), %%" _ASM_BX " \n\t"
@@ -4060,12 +4060,12 @@ static void svm_vcpu_run(struct kvm_vcpu
 	if (!msr_write_intercepted(vcpu, MSR_IA32_SPEC_CTRL))
 		svm->spec_ctrl = native_read_msr(MSR_IA32_SPEC_CTRL);
-	x86_spec_ctrl_restore_host(svm->spec_ctrl, svm->virt_spec_ctrl);
+	x86_spec_ctrl_restore_host(svm->spec_ctrl, svm->virt_spec_ctrl);
 	vcpu->arch.cr2 = svm->vmcb->save.cr2;
 	vcpu->arch.regs[VCPU_REGS_RAX] = svm->vmcb->save.rax;
 	vcpu->arch.regs[VCPU_REGS_RSP] = svm->vmcb->save.rsp;