[PATCH v2] arm64: KVM: Only skip MMIO insn once

[PATCH v2] arm64: KVM: Only skip MMIO insn once

[Andrew Jones]

If after an MMIO exit to userspace a VCPU is immediately run with an
immediate_exit request, such as when a signal is delivered or an MMIO
emulation completion is needed, then the VCPU completes the MMIO
emulation and immediately returns to userspace. As the exit_reason
does not get changed from KVM_EXIT_MMIO in these cases we have to
be careful not to complete the MMIO emulation again, when the VCPU is
eventually run again, because the emulation does an instruction skip
(and doing too many skips would be a waste of guest code :-) We need
to use additional VCPU state to track if the emulation is complete.
As luck would have it, we already have 'mmio_needed', which even
appears to be used in this way by other architectures already.

Fixes: 0d640732dbeb ("arm64: KVM: Skip MMIO insn after emulation")
Signed-off-by: Andrew Jones <drjones@redhat.com>
---
v2: move mmio_needed use closer to other mmio state use [maz]

 virt/kvm/arm/mmio.c | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/virt/kvm/arm/mmio.c b/virt/kvm/arm/mmio.c
index a8a6a0c883f1..6af5c91337f2 100644
--- a/virt/kvm/arm/mmio.c
+++ b/virt/kvm/arm/mmio.c
@@ -86,6 +86,12 @@ int kvm_handle_mmio_return(struct kvm_vcpu *vcpu, struct kvm_run *run)
 	unsigned int len;
 	int mask;
 
+	/* Detect an already handled MMIO return */
+	if (unlikely(!vcpu->mmio_needed))
+		return 0;
+
+	vcpu->mmio_needed = 0;
+
 	if (!run->mmio.is_write) {
 		len = run->mmio.len;
 		if (len > sizeof(unsigned long))
@@ -188,6 +194,7 @@ int io_mem_abort(struct kvm_vcpu *vcpu, struct kvm_run *run,
 	run->mmio.is_write	= is_write;
 	run->mmio.phys_addr	= fault_ipa;
 	run->mmio.len		= len;
+	vcpu->mmio_needed	= 1;
 
 	if (!ret) {
 		/* We handled the access successfully in the kernel. */
-- 
2.18.1

Re: [PATCH v2] arm64: KVM: Only skip MMIO insn once

[Mark Rutland]

On Thu, Aug 22, 2019 at 01:03:05PM +0200, Andrew Jones wrote:
> If after an MMIO exit to userspace a VCPU is immediately run with an
> immediate_exit request, such as when a signal is delivered or an MMIO
> emulation completion is needed, then the VCPU completes the MMIO
> emulation and immediately returns to userspace. As the exit_reason
> does not get changed from KVM_EXIT_MMIO in these cases we have to
> be careful not to complete the MMIO emulation again, when the VCPU is
> eventually run again, because the emulation does an instruction skip
> (and doing too many skips would be a waste of guest code :-) We need
> to use additional VCPU state to track if the emulation is complete.
> As luck would have it, we already have 'mmio_needed', which even
> appears to be used in this way by other architectures already.
> 
> Fixes: 0d640732dbeb ("arm64: KVM: Skip MMIO insn after emulation")
> Signed-off-by: Andrew Jones <drjones@redhat.com>

Ouch; sorry about this!

I haven't dug too deeply, but from the commit message, the below makes
sense to me. FWIW:

Acked-by: Mark Rutland <mark.rutland@arm.com>

Mark.

> ---
> v2: move mmio_needed use closer to other mmio state use [maz]
> 
>  virt/kvm/arm/mmio.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/virt/kvm/arm/mmio.c b/virt/kvm/arm/mmio.c
> index a8a6a0c883f1..6af5c91337f2 100644
> --- a/virt/kvm/arm/mmio.c
> +++ b/virt/kvm/arm/mmio.c
> @@ -86,6 +86,12 @@ int kvm_handle_mmio_return(struct kvm_vcpu *vcpu, struct kvm_run *run)
>  	unsigned int len;
>  	int mask;
>  
> +	/* Detect an already handled MMIO return */
> +	if (unlikely(!vcpu->mmio_needed))
> +		return 0;
> +
> +	vcpu->mmio_needed = 0;
> +
>  	if (!run->mmio.is_write) {
>  		len = run->mmio.len;
>  		if (len > sizeof(unsigned long))
> @@ -188,6 +194,7 @@ int io_mem_abort(struct kvm_vcpu *vcpu, struct kvm_run *run,
>  	run->mmio.is_write	= is_write;
>  	run->mmio.phys_addr	= fault_ipa;
>  	run->mmio.len		= len;
> +	vcpu->mmio_needed	= 1;
>  
>  	if (!ret) {
>  		/* We handled the access successfully in the kernel. */
> -- 
> 2.18.1
> 

Re: [PATCH v2] arm64: KVM: Only skip MMIO insn once

[Marc Zyngier]

On 22/08/2019 12:03, Andrew Jones wrote:
> If after an MMIO exit to userspace a VCPU is immediately run with an
> immediate_exit request, such as when a signal is delivered or an MMIO
> emulation completion is needed, then the VCPU completes the MMIO
> emulation and immediately returns to userspace. As the exit_reason
> does not get changed from KVM_EXIT_MMIO in these cases we have to
> be careful not to complete the MMIO emulation again, when the VCPU is
> eventually run again, because the emulation does an instruction skip
> (and doing too many skips would be a waste of guest code :-) We need
> to use additional VCPU state to track if the emulation is complete.
> As luck would have it, we already have 'mmio_needed', which even
> appears to be used in this way by other architectures already.
> 
> Fixes: 0d640732dbeb ("arm64: KVM: Skip MMIO insn after emulation")
> Signed-off-by: Andrew Jones <drjones@redhat.com>
> ---
> v2: move mmio_needed use closer to other mmio state use [maz]
> 
>  virt/kvm/arm/mmio.c | 7 +++++++
>  1 file changed, 7 insertions(+)

Applied with Mark's Ack.

Thanks,

	M.
-- 
Jazz is not dead, it just smells funny...