| Message ID | 511F72B8.6060305@web.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 2013-02-16 12:51, Jan Kiszka wrote: > From: Jan Kiszka <jan.kiszka@siemens.com> > > The comment was wrong: enable_irq_window might be called after > prepare_vmcs02 when we left L2 to prepare IRQ injecting for L1. Same for > NMIs. > > Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com> > --- > arch/x86/kvm/vmx.c | 10 +++------- > 1 files changed, 3 insertions(+), 7 deletions(-) > > diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c > index 81a25de..68a045ae 100644 > --- a/arch/x86/kvm/vmx.c > +++ b/arch/x86/kvm/vmx.c > @@ -6113,14 +6113,10 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu) > case EXIT_REASON_TRIPLE_FAULT: > return 1; > case EXIT_REASON_PENDING_INTERRUPT: > + return nested_cpu_has(vmcs12, CPU_BASED_VIRTUAL_INTR_PENDING); > case EXIT_REASON_NMI_WINDOW: > - /* > - * prepare_vmcs02() set the CPU_BASED_VIRTUAL_INTR_PENDING bit > - * (aka Interrupt Window Exiting) only when L1 turned it on, > - * so if we got a PENDING_INTERRUPT exit, this must be for L1. > - * Same for NMI Window Exiting. > - */ > - return 1; > + return vmcs12->pin_based_vm_exec_control & > + PIN_BASED_NMI_EXITING; > case EXIT_REASON_TASK_SWITCH: > return 1; > case EXIT_REASON_CPUID: > No, this is wrong. I first wrote a patch that ignored enable_irq_window when the guest is not interested in external IRQs. But then I thought that wasn't correct. I tend to believe now my first idea was better. Need to rethink. Jan
On Sat, Feb 16, 2013, Jan Kiszka wrote about "Re: [PATCH] KVM: nVMX: Fix injection of PENDING_INTERRUPT and NMI_WINDOW exits to L1": > No, this is wrong. I first wrote a patch that ignored enable_irq_window > when the guest is not interested in external IRQs. But then I thought > that wasn't correct. I tend to believe now my first idea was better. If you want to make the !PIN_BASED_EXT_INTR_MASK case work correctly, please also see: https://bugzilla.kernel.org/show_bug.cgi?id=53711 Nadav.
On 2013-02-16 15:46, Nadav Har'El wrote: > On Sat, Feb 16, 2013, Jan Kiszka wrote about "Re: [PATCH] KVM: nVMX: Fix injection of PENDING_INTERRUPT and NMI_WINDOW exits to L1": >> No, this is wrong. I first wrote a patch that ignored enable_irq_window >> when the guest is not interested in external IRQs. But then I thought >> that wasn't correct. I tend to believe now my first idea was better. > > If you want to make the !PIN_BASED_EXT_INTR_MASK case work correctly, > please also see: > > https://bugzilla.kernel.org/show_bug.cgi?id=53711 Yes, I already realized that there is more broken. Trying to find a way to fix this ATM. That specifically means deciding when to execute vmx_complete/cancel_interrupts and when not. This patch seems correct now, just insufficient to make direct injection from L0 to L2 work. Jan
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c index 81a25de..68a045ae 100644 --- a/arch/x86/kvm/vmx.c +++ b/arch/x86/kvm/vmx.c @@ -6113,14 +6113,10 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu) case EXIT_REASON_TRIPLE_FAULT: return 1; case EXIT_REASON_PENDING_INTERRUPT: + return nested_cpu_has(vmcs12, CPU_BASED_VIRTUAL_INTR_PENDING); case EXIT_REASON_NMI_WINDOW: - /* - * prepare_vmcs02() set the CPU_BASED_VIRTUAL_INTR_PENDING bit - * (aka Interrupt Window Exiting) only when L1 turned it on, - * so if we got a PENDING_INTERRUPT exit, this must be for L1. - * Same for NMI Window Exiting. - */ - return 1; + return vmcs12->pin_based_vm_exec_control & + PIN_BASED_NMI_EXITING; case EXIT_REASON_TASK_SWITCH: return 1; case EXIT_REASON_CPUID: