2021-07-27 21:11:53

by Paolo Bonzini

Subject: [PATCH v3] KVM: x86: accept userspace interrupt only if no event is injected

Once an exception has been injected, any side effects related to
the exception (such as setting CR2 or DR6) have been taked place.
Therefore, once KVM sets the VM-entry interruption information
field or the AMD EVENTINJ field, the next VM-entry must deliver that
exception.

Pending interrupts are processed after injected exceptions, so
in theory it would not be a problem to use KVM_INTERRUPT when
an injected exception is present. However, DOSEMU is using
run->ready_for_interrupt_injection to detect interrupt windows
and then using KVM_SET_SREGS/KVM_SET_REGS to inject the
interrupt manually. For this to work, the interrupt window
must be delayed after the completion of the previous event
injection.

Cc: [email protected]
Reported-by: Stas Sergeev <[email protected]>
Tested-by: Stas Sergeev <[email protected]>
Fixes: 71cc849b7093 ("KVM: x86: Fix split-irqchip vs interrupt injection window request")
Signed-off-by: Paolo Bonzini <[email protected]>
---
arch/x86/kvm/x86.c | 13 +++++++++++--
1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 4116567f3d44..e5d5c5ed7dd4 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -4358,8 +4358,17 @@ static int kvm_cpu_accept_dm_intr(struct kvm_vcpu *vcpu)

static int kvm_vcpu_ready_for_interrupt_injection(struct kvm_vcpu *vcpu)
{
- return kvm_arch_interrupt_allowed(vcpu) &&
- kvm_cpu_accept_dm_intr(vcpu);
+ /*
+ * Do not cause an interrupt window exit if an exception
+ * is pending or an event needs reinjection; userspace
+ * might want to inject the interrupt manually using KVM_SET_REGS
+ * or KVM_SET_SREGS. For that to work, we must be at an
+ * instruction boundary and with no events half-injected.
+ */
+ return (kvm_arch_interrupt_allowed(vcpu) &&
+ kvm_cpu_accept_dm_intr(vcpu) &&
+ !kvm_event_needs_reinjection(vcpu) &&
+ !vcpu->arch.exception.pending);
}

static int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu,
--
2.27.0

2021-07-28 23:13:06

by Sean Christopherson

Subject: Re: [PATCH v3] KVM: x86: accept userspace interrupt only if no event is injected

On Tue, Jul 27, 2021, Paolo Bonzini wrote:
> Once an exception has been injected, any side effects related to
> the exception (such as setting CR2 or DR6) have been taked place.
> Therefore, once KVM sets the VM-entry interruption information
> field or the AMD EVENTINJ field, the next VM-entry must deliver that
> exception.
>
> Pending interrupts are processed after injected exceptions, so
> in theory it would not be a problem to use KVM_INTERRUPT when
> an injected exception is present. However, DOSEMU is using
> run->ready_for_interrupt_injection to detect interrupt windows
> and then using KVM_SET_SREGS/KVM_SET_REGS to inject the
> interrupt manually. For this to work, the interrupt window
> must be delayed after the completion of the previous event
> injection.
>
> Cc: [email protected]
> Reported-by: Stas Sergeev <[email protected]>
> Tested-by: Stas Sergeev <[email protected]>
> Fixes: 71cc849b7093 ("KVM: x86: Fix split-irqchip vs interrupt injection window request")
> Signed-off-by: Paolo Bonzini <[email protected]>
> ---
> arch/x86/kvm/x86.c | 13 +++++++++++--
> 1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 4116567f3d44..e5d5c5ed7dd4 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -4358,8 +4358,17 @@ static int kvm_cpu_accept_dm_intr(struct kvm_vcpu *vcpu)
>
> static int kvm_vcpu_ready_for_interrupt_injection(struct kvm_vcpu *vcpu)
> {
> - return kvm_arch_interrupt_allowed(vcpu) &&
> - kvm_cpu_accept_dm_intr(vcpu);
> + /*
> + * Do not cause an interrupt window exit if an exception
> + * is pending or an event needs reinjection; userspace
> + * might want to inject the interrupt manually using KVM_SET_REGS
> + * or KVM_SET_SREGS. For that to work, we must be at an
> + * instruction boundary and with no events half-injected.
> + */
> + return (kvm_arch_interrupt_allowed(vcpu) &&

Ha, adding a '(' is one way to fix the indentation.

Reviewed-by: Sean Christopherson <[email protected]>

> + kvm_cpu_accept_dm_intr(vcpu) &&
> + !kvm_event_needs_reinjection(vcpu) &&
> + !vcpu->arch.exception.pending);
> }
>
> static int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu,
> --
> 2.27.0
>

2021-08-06 23:59:32

by stsp

Subject: Re: [PATCH v3] KVM: x86: accept userspace interrupt only if no event is injected

28.07.2021 00:09, Paolo Bonzini пишет:
> Once an exception has been injected, any side effects related to
> the exception (such as setting CR2 or DR6) have been taked place.
> Therefore, once KVM sets the VM-entry interruption information
> field or the AMD EVENTINJ field, the next VM-entry must deliver that
> exception.
>
> Pending interrupts are processed after injected exceptions, so
> in theory it would not be a problem to use KVM_INTERRUPT when
> an injected exception is present. However, DOSEMU is using
> run->ready_for_interrupt_injection to detect interrupt windows
> and then using KVM_SET_SREGS/KVM_SET_REGS to inject the
> interrupt manually. For this to work, the interrupt window
> must be delayed after the completion of the previous event
> injection.
>
> Cc: [email protected]
> Reported-by: Stas Sergeev <[email protected]>
> Tested-by: Stas Sergeev <[email protected]>
Acked-by: [email protected]