* [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset
@ 2013-12-16 9:32 Jan Kiszka
2013-12-17 13:25 ` Paolo Bonzini
` (2 more replies)
0 siblings, 3 replies; 7+ messages in thread
From: Jan Kiszka @ 2013-12-16 9:32 UTC (permalink / raw)
To: Gleb Natapov, Paolo Bonzini; +Cc: kvm
As long as we do not expose all the VMX related states to user space,
there is no way to properly reset a VCPU when VMX is enabled. Emulate
this for now by catching host-side clearings of the feature control MSR.
This allows to reboot a VM while it is running some hypervisor code.
Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
---
Better ideas? Or continue to leave it as it is?
arch/x86/kvm/vmx.c | 35 +++++++++++++++++++++++++++++++++++
1 file changed, 35 insertions(+)
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index f90320b..da04247 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -2455,6 +2455,8 @@ static int vmx_get_vmx_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata)
return 1;
}
+static void vmx_reset_nested(struct kvm_vcpu *vcpu);
+
static int vmx_set_vmx_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
{
u32 msr_index = msr_info->index;
@@ -2470,6 +2472,12 @@ static int vmx_set_vmx_msr(struct kvm_vcpu *vcpu, struct msr_data *msr_info)
& FEATURE_CONTROL_LOCKED)
return 0;
to_vmx(vcpu)->nested.msr_ia32_feature_control = data;
+ /*
+ * Detect reset and allow to leave VMX mode this way until we
+ * expose all related states to user space.
+ */
+ if (host_initialized && data == 0)
+ vmx_reset_nested(vcpu);
return 1;
}
@@ -8487,6 +8495,33 @@ static void nested_vmx_vmexit(struct kvm_vcpu *vcpu)
vmx->nested.sync_shadow_vmcs = true;
}
+static void vmx_reset_nested(struct kvm_vcpu *vcpu)
+{
+ struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
+ struct vcpu_vmx *vmx = to_vmx(vcpu);
+
+ if (!vmx->nested.vmxon)
+ return;
+
+ if (is_guest_mode(vcpu)) {
+ vmcs12->host_cr0 = X86_CR0_NW | X86_CR0_CD | X86_CR0_ET;
+ vmcs12->host_cr3 = 0;
+ vmcs12->host_cr4 = 0;
+ vmcs12->host_rsp = 0;
+ vmcs12->vm_exit_controls = 0;
+ nested_vmx_vmexit(vcpu);
+ }
+
+ free_nested(vmx);
+
+ /*
+ * If we were in guest mode, the reset state user space wrote so far
+ * is now inconsistent. If we were in host mode, some state update may
+ * have been rejected. So simply repeat the reset her.
+ */
+ vmx_vcpu_reset(vcpu);
+}
+
/*
* L1's failure to enter L2 is a subset of a normal exit, as explained in
* 23.7 "VM-entry failures during or after loading guest state" (this also
--
1.8.1.1.298.ge7eed54
^ permalink raw reply related [flat|nested] 7+ messages in thread* Re: [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset
2013-12-16 9:32 [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset Jan Kiszka
@ 2013-12-17 13:25 ` Paolo Bonzini
2013-12-17 14:40 ` Jan Kiszka
2013-12-30 20:02 ` Marcelo Tosatti
2013-12-30 20:02 ` Marcelo Tosatti
2 siblings, 1 reply; 7+ messages in thread
From: Paolo Bonzini @ 2013-12-17 13:25 UTC (permalink / raw)
To: Jan Kiszka; +Cc: Gleb Natapov, kvm
Il 16/12/2013 10:32, Jan Kiszka ha scritto:
> As long as we do not expose all the VMX related states to user space,
> there is no way to properly reset a VCPU when VMX is enabled. Emulate
> this for now by catching host-side clearings of the feature control MSR.
> This allows to reboot a VM while it is running some hypervisor code.
>
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> ---
>
> Better ideas? Or continue to leave it as it is?
The final vmx_vcpu_reset is the only really ugly part, but it is
_really_ ugly... Can you modify QEMU to restore MSRs first, and reduce
vmx_reset_nested to just
if (is_guest_mode(vcpu))
nested_vmx_vmexit(vcpu);
free_nested(vmx);
?
Paolo
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset
2013-12-17 13:25 ` Paolo Bonzini
@ 2013-12-17 14:40 ` Jan Kiszka
2013-12-17 14:43 ` Paolo Bonzini
0 siblings, 1 reply; 7+ messages in thread
From: Jan Kiszka @ 2013-12-17 14:40 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: Gleb Natapov, kvm
On 2013-12-17 14:25, Paolo Bonzini wrote:
> Il 16/12/2013 10:32, Jan Kiszka ha scritto:
>> As long as we do not expose all the VMX related states to user space,
>> there is no way to properly reset a VCPU when VMX is enabled. Emulate
>> this for now by catching host-side clearings of the feature control MSR.
>> This allows to reboot a VM while it is running some hypervisor code.
>>
>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>> ---
>>
>> Better ideas? Or continue to leave it as it is?
>
> The final vmx_vcpu_reset is the only really ugly part, but it is
> _really_ ugly... Can you modify QEMU to restore MSRs first, and reduce
> vmx_reset_nested to just
>
> if (is_guest_mode(vcpu))
> nested_vmx_vmexit(vcpu);
>
> free_nested(vmx);
>
> ?
Well, I could make setting of MSR_IA32_FEATURE_CONTROL to 0 an official
"clear VMX" interface. Then QEMU would have to issue this MSR set
request before doing any other CPU state manipulation. Is that what you
have in mind?
Jan
--
Siemens AG, Corporate Technology, CT RTC ITP SES-DE
Corporate Competence Center Embedded Linux
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset
2013-12-17 14:40 ` Jan Kiszka
@ 2013-12-17 14:43 ` Paolo Bonzini
0 siblings, 0 replies; 7+ messages in thread
From: Paolo Bonzini @ 2013-12-17 14:43 UTC (permalink / raw)
To: Jan Kiszka; +Cc: Gleb Natapov, kvm
Il 17/12/2013 15:40, Jan Kiszka ha scritto:
>> > The final vmx_vcpu_reset is the only really ugly part, but it is
>> > _really_ ugly... Can you modify QEMU to restore MSRs first, and reduce
>> > vmx_reset_nested to just
>> >
>> > if (is_guest_mode(vcpu))
>> > nested_vmx_vmexit(vcpu);
>> >
>> > free_nested(vmx);
>> >
>> > ?
> Well, I could make setting of MSR_IA32_FEATURE_CONTROL to 0 an official
> "clear VMX" interface. Then QEMU would have to issue this MSR set
> request before doing any other CPU state manipulation. Is that what you
> have in mind?
Yes, that was the idea.
Paolo
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset
2013-12-16 9:32 [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset Jan Kiszka
2013-12-17 13:25 ` Paolo Bonzini
@ 2013-12-30 20:02 ` Marcelo Tosatti
2013-12-30 20:05 ` Marcelo Tosatti
2013-12-30 20:02 ` Marcelo Tosatti
2 siblings, 1 reply; 7+ messages in thread
From: Marcelo Tosatti @ 2013-12-30 20:02 UTC (permalink / raw)
To: Jan Kiszka; +Cc: Gleb Natapov, Paolo Bonzini, kvm
On Mon, Dec 16, 2013 at 10:32:34AM +0100, Jan Kiszka wrote:
> As long as we do not expose all the VMX related states to user space,
> there is no way to properly reset a VCPU when VMX is enabled. Emulate
> this for now by catching host-side clearings of the feature control MSR.
> This allows to reboot a VM while it is running some hypervisor code.
>
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> ---
>
> Better ideas? Or continue to leave it as it is?
The hidden backport might be problematic... How about an explicit
notification to leave guest mode?
Such as a new flag to KVM_SET_REGS (or even a separate ioctl?).
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset
2013-12-30 20:02 ` Marcelo Tosatti
@ 2013-12-30 20:05 ` Marcelo Tosatti
0 siblings, 0 replies; 7+ messages in thread
From: Marcelo Tosatti @ 2013-12-30 20:05 UTC (permalink / raw)
To: Jan Kiszka; +Cc: Gleb Natapov, Paolo Bonzini, kvm
On Mon, Dec 30, 2013 at 06:02:17PM -0200, Marcelo Tosatti wrote:
> On Mon, Dec 16, 2013 at 10:32:34AM +0100, Jan Kiszka wrote:
> > As long as we do not expose all the VMX related states to user space,
> > there is no way to properly reset a VCPU when VMX is enabled. Emulate
> > this for now by catching host-side clearings of the feature control MSR.
> > This allows to reboot a VM while it is running some hypervisor code.
> >
> > Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> > ---
> >
> > Better ideas? Or continue to leave it as it is?
>
> The hidden backport might be problematic... How about an explicit
> notification to leave guest mode?
> Such as a new flag to KVM_SET_REGS (or even a separate ioctl?).
>
Nevermind.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset
2013-12-16 9:32 [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset Jan Kiszka
2013-12-17 13:25 ` Paolo Bonzini
2013-12-30 20:02 ` Marcelo Tosatti
@ 2013-12-30 20:02 ` Marcelo Tosatti
2 siblings, 0 replies; 7+ messages in thread
From: Marcelo Tosatti @ 2013-12-30 20:02 UTC (permalink / raw)
To: Jan Kiszka; +Cc: Gleb Natapov, Paolo Bonzini, kvm
On Mon, Dec 16, 2013 at 10:32:34AM +0100, Jan Kiszka wrote:
> As long as we do not expose all the VMX related states to user space,
> there is no way to properly reset a VCPU when VMX is enabled. Emulate
> this for now by catching host-side clearings of the feature control MSR.
> This allows to reboot a VM while it is running some hypervisor code.
>
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> ---
>
> Better ideas? Or continue to leave it as it is?
So basically reset is broken, with nested virt?
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2013-12-30 20:06 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-12-16 9:32 [RFC][PATCH] KVM: nVMX: Leave VMX mode on apparent CPU reset Jan Kiszka
2013-12-17 13:25 ` Paolo Bonzini
2013-12-17 14:40 ` Jan Kiszka
2013-12-17 14:43 ` Paolo Bonzini
2013-12-30 20:02 ` Marcelo Tosatti
2013-12-30 20:05 ` Marcelo Tosatti
2013-12-30 20:02 ` Marcelo Tosatti
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox