KVM: SVM: fix tsc scaling cache logic

SVM uses a per-cpu variable to cache the current value of the
tsc scaling multiplier msr on each cpu.

Commit 1ab9287add5e2
("KVM: X86: Add vendor callbacks for writing the TSC multiplier")
broke this caching logic.

Refactor the code so that all TSC scaling multiplier writes go through
a single function which checks and updates the cache.

This fixes the following scenario:

1. A CPU runs a guest with some tsc scaling ratio.

2. New guest with different tsc scaling ratio starts on this CPU
   and terminates almost immediately.

   This ensures that the short running guest had set the tsc scaling ratio just
   once when it was set via KVM_SET_TSC_KHZ. Due to the bug,
   the per-cpu cache is not updated.

3. The original guest continues to run, it doesn't restore the msr
   value back to its own value, because the cache matches,
   and thus continues to run with a wrong tsc scaling ratio.

Fixes: 1ab9287add5e2 ("KVM: X86: Add vendor callbacks for writing the TSC multiplier")
Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
Message-Id: <20220606181149.103072-1-mlevitsk@redhat.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

diff --git a/arch/x86/kvm/svm/nested.c b/arch/x86/kvm/svm/nested.c
index bed5e1692cef0209457697cd96dc401139a6167f..3361258640a27e15f3ab04f46e6daedbfb09079a 100644 (file)
--- a/arch/x86/kvm/svm/nested.c
+++ b/arch/x86/kvm/svm/nested.c
@@ -982,7 +982,7 @@ int nested_svm_vmexit(struct vcpu_svm *svm)
        if (svm->tsc_ratio_msr != kvm_default_tsc_scaling_ratio) {
                WARN_ON(!svm->tsc_scaling_enabled);
                vcpu->arch.tsc_scaling_ratio = vcpu->arch.l1_tsc_scaling_ratio;
-               svm_write_tsc_multiplier(vcpu, vcpu->arch.tsc_scaling_ratio);
+               __svm_write_tsc_multiplier(vcpu->arch.tsc_scaling_ratio);
        }
 
        svm->nested.ctl.nested_cr3 = 0;
@@ -1387,7 +1387,7 @@ void nested_svm_update_tsc_ratio_msr(struct kvm_vcpu *vcpu)
        vcpu->arch.tsc_scaling_ratio =
                kvm_calc_nested_tsc_multiplier(vcpu->arch.l1_tsc_scaling_ratio,
                                               svm->tsc_ratio_msr);
-       svm_write_tsc_multiplier(vcpu, vcpu->arch.tsc_scaling_ratio);
+       __svm_write_tsc_multiplier(vcpu->arch.tsc_scaling_ratio);
 }
 
 /* Inverse operation of nested_copy_vmcb_control_to_cache(). asid is copied too. */

diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
index 63880b33ce370801eda69e9758b69cf838c06df2..478e6ee81d8838b43a8972608769c98cc9c6560a 100644 (file)
--- a/arch/x86/kvm/svm/svm.c
+++ b/arch/x86/kvm/svm/svm.c
@@ -465,11 +465,24 @@ static int has_svm(void)
        return 1;
 }
 
+void __svm_write_tsc_multiplier(u64 multiplier)
+{
+       preempt_disable();
+
+       if (multiplier == __this_cpu_read(current_tsc_ratio))
+               goto out;
+
+       wrmsrl(MSR_AMD64_TSC_RATIO, multiplier);
+       __this_cpu_write(current_tsc_ratio, multiplier);
+out:
+       preempt_enable();
+}
+
 static void svm_hardware_disable(void)
 {
        /* Make sure we clean up behind us */
        if (tsc_scaling)
-               wrmsrl(MSR_AMD64_TSC_RATIO, SVM_TSC_RATIO_DEFAULT);
+               __svm_write_tsc_multiplier(SVM_TSC_RATIO_DEFAULT);
 
        cpu_svm_disable();
 
@@ -515,8 +528,7 @@ static int svm_hardware_enable(void)
                 * Set the default value, even if we don't use TSC scaling
                 * to avoid having stale value in the msr
                 */
-               wrmsrl(MSR_AMD64_TSC_RATIO, SVM_TSC_RATIO_DEFAULT);
-               __this_cpu_write(current_tsc_ratio, SVM_TSC_RATIO_DEFAULT);
+               __svm_write_tsc_multiplier(SVM_TSC_RATIO_DEFAULT);
        }
 
 
@@ -999,11 +1011,12 @@ static void svm_write_tsc_offset(struct kvm_vcpu *vcpu, u64 offset)
        vmcb_mark_dirty(svm->vmcb, VMCB_INTERCEPTS);
 }
 
-void svm_write_tsc_multiplier(struct kvm_vcpu *vcpu, u64 multiplier)
+static void svm_write_tsc_multiplier(struct kvm_vcpu *vcpu, u64 multiplier)
 {
-       wrmsrl(MSR_AMD64_TSC_RATIO, multiplier);
+       __svm_write_tsc_multiplier(multiplier);
 }
 
+
 /* Evaluate instruction intercepts that depend on guest CPUID features. */
 static void svm_recalc_instruction_intercepts(struct kvm_vcpu *vcpu,
                                              struct vcpu_svm *svm)
@@ -1363,13 +1376,8 @@ static void svm_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
                sev_es_prepare_switch_to_guest(hostsa);
        }
 
-       if (tsc_scaling) {
-               u64 tsc_ratio = vcpu->arch.tsc_scaling_ratio;
-               if (tsc_ratio != __this_cpu_read(current_tsc_ratio)) {
-                       __this_cpu_write(current_tsc_ratio, tsc_ratio);
-                       wrmsrl(MSR_AMD64_TSC_RATIO, tsc_ratio);
-               }
-       }
+       if (tsc_scaling)
+               __svm_write_tsc_multiplier(vcpu->arch.tsc_scaling_ratio);
 
        if (likely(tsc_aux_uret_slot >= 0))
                kvm_set_user_return_msr(tsc_aux_uret_slot, svm->tsc_aux, -1ull);

diff --git a/arch/x86/kvm/svm/svm.h b/arch/x86/kvm/svm/svm.h
index 45a87b2a8b3c47637a27eb0049ba411c22cfa5a9..29d6fd205a4946a1bca4ad7fa015111670979672 100644 (file)
--- a/arch/x86/kvm/svm/svm.h
+++ b/arch/x86/kvm/svm/svm.h
@@ -590,7 +590,7 @@ int nested_svm_check_exception(struct vcpu_svm *svm, unsigned nr,
                               bool has_error_code, u32 error_code);
 int nested_svm_exit_special(struct vcpu_svm *svm);
 void nested_svm_update_tsc_ratio_msr(struct kvm_vcpu *vcpu);
-void svm_write_tsc_multiplier(struct kvm_vcpu *vcpu, u64 multiplier);
+void __svm_write_tsc_multiplier(u64 multiplier);
 void nested_copy_vmcb_control_to_cache(struct vcpu_svm *svm,
                                       struct vmcb_control_area *control);
 void nested_copy_vmcb_save_to_cache(struct vcpu_svm *svm,