On 05/05/20 15:41, Thomas Gleixner wrote:
> Context tracking for KVM happens way too early in the vcpu_run()
> code. Anything after guest_enter_irqoff() and before guest_exit_irqoff()
> cannot use RCU and should also be not instrumented.
>
> The current way of doing this covers way too much code. Move it closer to
> the actual vmenter/exit code.
>
> Signed-off-by: Thomas Gleixner <t...@linutronix.de>
> Cc: Paolo Bonzini <pbonz...@redhat.com>
> Cc: Sean Christopherson <sean.j.christopher...@intel.com>
> ---
> arch/x86/kvm/svm/svm.c | 16 ++++++++++++++++
> arch/x86/kvm/vmx/vmx.c | 10 ++++++++++
> arch/x86/kvm/x86.c | 2 --
> 3 files changed, 26 insertions(+), 2 deletions(-)
>
> --- a/arch/x86/kvm/svm/svm.c
> +++ b/arch/x86/kvm/svm/svm.c
> @@ -3330,6 +3330,14 @@ static void svm_vcpu_run(struct kvm_vcpu
> */
> x86_spec_ctrl_set_guest(svm->spec_ctrl, svm->virt_spec_ctrl);
>
> + /*
> + * Tell context tracking that this CPU is about to enter guest
> + * mode. This has to be after x86_spec_ctrl_set_guest() because
> + * that can take locks (lockdep needs RCU) and calls into world and
> + * some more.
> + */
> + guest_enter_irqoff();
> +
> __svm_vcpu_run(svm->vmcb_pa, (unsigned long *)&svm->vcpu.arch.regs);
>
> #ifdef CONFIG_X86_64
> @@ -3340,6 +3348,14 @@ static void svm_vcpu_run(struct kvm_vcpu
> loadsegment(gs, svm->host.gs);
> #endif
> #endif
> + /*
> + * Tell context tracking that this CPU is back.
> + *
> + * This needs to be done before the below as native_read_msr()
> + * contains a tracepoint and x86_spec_ctrl_restore_host() calls
> + * into world and some more.
> + */
> + guest_exit_irqoff();
>
> /*
> * We do not use IBRS in the kernel. If this vCPU has used the
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -6603,6 +6603,11 @@ static void vmx_vcpu_run(struct kvm_vcpu
> */
> x86_spec_ctrl_set_guest(vmx->spec_ctrl, 0);
>
> + /*
> + * Tell context tracking that this CPU is about to enter guest mode.
> + */
> + guest_enter_irqoff();
> +
> /* L1D Flush includes CPU buffer clear to mitigate MDS */
> if (static_branch_unlikely(&vmx_l1d_should_flush))
> vmx_l1d_flush(vcpu);
> @@ -6618,6 +6623,11 @@ static void vmx_vcpu_run(struct kvm_vcpu
> vcpu->arch.cr2 = read_cr2();
>
> /*
> + * Tell context tracking that this CPU is back.
> + */
> + guest_exit_irqoff();
> +
> + /*
> * We do not use IBRS in the kernel. If this vCPU has used the
> * SPEC_CTRL MSR it may have left it on; save the value and
> * turn it off. This is much more efficient than blindly adding
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -8404,7 +8404,6 @@ static int vcpu_enter_guest(struct kvm_v
> }
>
> trace_kvm_entry(vcpu->vcpu_id);
> - guest_enter_irqoff();
>
> fpregs_assert_state_consistent();
> if (test_thread_flag(TIF_NEED_FPU_LOAD))
> @@ -8467,7 +8466,6 @@ static int vcpu_enter_guest(struct kvm_v
> local_irq_disable();
> kvm_after_interrupt(vcpu);
>
> - guest_exit_irqoff();
> if (lapic_in_kernel(vcpu)) {
> s64 delta = vcpu->arch.apic->lapic_timer.advance_expire_delta;
> if (delta != S64_MIN) {
>
Acked-by: Paolo Bonzini <pbonz...@redhat.com>
