Re: [PATCH v5 09/11] KVM: X86: Add vendor callbacks for writing the TSC multiplier
From: Sean Christopherson <seanjc@google.com>
Date: 2021-06-01 15:23:04
Also in:
lkml
Subsystem:
kernel virtual machine for x86 (kvm/x86), the rest, x86 architecture (32-bit and 64-bit) · Maintainers:
Sean Christopherson, Paolo Bonzini, Linus Torvalds, Thomas Gleixner, Ingo Molnar, Borislav Petkov, Dave Hansen
On Fri, May 28, 2021, Ilias Stamatis wrote:
quoted hunk ↗ jump to hunk
@@ -10343,8 +10360,6 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu) else vcpu->arch.mp_state = KVM_MP_STATE_UNINITIALIZED; - kvm_set_tsc_khz(vcpu, max_tsc_khz); - r = kvm_mmu_create(vcpu); if (r < 0) return r;@@ -10443,6 +10458,7 @@ void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu) if (mutex_lock_killable(&vcpu->mutex)) return; vcpu_load(vcpu); + kvm_set_tsc_khz(vcpu, max_tsc_khz);
It would be safer to move the call within kvm_arch_vcpu_create(), not to postcreate(). postcreate() runs after the vCPU file descriptor is exposed to userspace, e.g. KVM_GET_TSC_KHZ could theoretically read the wrong value. I don't know if there's anything more sinister that could be done, but I also don't see any reason to find out :-)
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index bc6160db94b1..08f6b8f82506 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c@@ -10291,8 +10291,6 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu) else vcpu->arch.mp_state = KVM_MP_STATE_UNINITIALIZED; - kvm_set_tsc_khz(vcpu, max_tsc_khz); - r = kvm_mmu_create(vcpu); if (r < 0) return r;
@@ -10360,6 +10358,7 @@ int kvm_arch_vcpu_create(struct kvm_vcpu *vcpu) vcpu->arch.msr_platform_info = MSR_PLATFORM_INFO_CPUID_FAULT; kvm_vcpu_mtrr_init(vcpu); vcpu_load(vcpu); + kvm_set_tsc_khz(vcpu, max_tsc_khz); kvm_vcpu_reset(vcpu, false); kvm_init_mmu(vcpu, false); vcpu_put(vcpu);
kvm_synchronize_tsc(vcpu, 0); vcpu_put(vcpu); -- 2.17.1