[PATCH] ARM: KVM: Handle loads from I/O addresses to PC

[PATCH] ARM: KVM: Handle loads from I/O addresses to PC

[Christoffer Dall]

If a VM does something like this:

ldr     pc, [r0]        @ r0 is an I/O address

Then simply let that value load into the PC on kvm_handle_mmio_return
overwriting the PC value set from kvm_skip_instr.

Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>
---
 arch/arm/kvm/mmio.c |    6 ------
 1 file changed, 6 deletions(-)

diff --git a/arch/arm/kvm/mmio.c b/arch/arm/kvm/mmio.c
index 72a12f2..b8e06b7 100644
--- a/arch/arm/kvm/mmio.c
+++ b/arch/arm/kvm/mmio.c
@@ -86,12 +86,6 @@ static int decode_hsr(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
 	sign_extend = kvm_vcpu_dabt_issext(vcpu);
 	rt = kvm_vcpu_dabt_get_rd(vcpu);
 
-	if (kvm_vcpu_reg_is_pc(vcpu, rt)) {
-		/* IO memory trying to read/write pc */
-		kvm_inject_pabt(vcpu, kvm_vcpu_get_hfar(vcpu));
-		return 1;
-	}
-
 	mmio->is_write = is_write;
 	mmio->phys_addr = fault_ipa;
 	mmio->len = len;
-- 
1.7.9.5

[PATCH] ARM: KVM: Handle loads from I/O addresses to PC

[Marc Zyngier]

On 12/05/13 19:58, Christoffer Dall wrote:
> If a VM does something like this:
> 
> ldr     pc, [r0]        @ r0 is an I/O address
> 
> Then simply let that value load into the PC on kvm_handle_mmio_return
> overwriting the PC value set from kvm_skip_instr.
> 
> Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>

I've had a slightly more complete patch in my tree for a while (look at
bef0ef01ae75 on the kvm-arm64/kvm-3.10-rc0 branch).

> ---
>  arch/arm/kvm/mmio.c |    6 ------
>  1 file changed, 6 deletions(-)
> 
> diff --git a/arch/arm/kvm/mmio.c b/arch/arm/kvm/mmio.c
> index 72a12f2..b8e06b7 100644
> --- a/arch/arm/kvm/mmio.c
> +++ b/arch/arm/kvm/mmio.c
> @@ -86,12 +86,6 @@ static int decode_hsr(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
>  	sign_extend = kvm_vcpu_dabt_issext(vcpu);
>  	rt = kvm_vcpu_dabt_get_rd(vcpu);
>  
> -	if (kvm_vcpu_reg_is_pc(vcpu, rt)) {
> -		/* IO memory trying to read/write pc */
> -		kvm_inject_pabt(vcpu, kvm_vcpu_get_hfar(vcpu));
> -		return 1;
> -	}
> -
>  	mmio->is_write = is_write;
>  	mmio->phys_addr = fault_ipa;
>  	mmio->len = len;
> 

Otherwise looks good to me.

	M.
-- 
Jazz is not dead. It just smells funny...

[PATCH] ARM: KVM: Handle loads from I/O addresses to PC

[Catalin Marinas]

On Sun, May 12, 2013 at 07:58:55PM +0100, Christoffer Dall wrote:
> If a VM does something like this:
> 
> ldr     pc, [r0]        @ r0 is an I/O address
> 
> Then simply let that value load into the PC on kvm_handle_mmio_return
> overwriting the PC value set from kvm_skip_instr.
> 
> Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>
> ---
>  arch/arm/kvm/mmio.c |    6 ------
>  1 file changed, 6 deletions(-)
> 
> diff --git a/arch/arm/kvm/mmio.c b/arch/arm/kvm/mmio.c
> index 72a12f2..b8e06b7 100644
> --- a/arch/arm/kvm/mmio.c
> +++ b/arch/arm/kvm/mmio.c
> @@ -86,12 +86,6 @@ static int decode_hsr(struct kvm_vcpu *vcpu, phys_addr_t fault_ipa,
>  	sign_extend = kvm_vcpu_dabt_issext(vcpu);
>  	rt = kvm_vcpu_dabt_get_rd(vcpu);
>  
> -	if (kvm_vcpu_reg_is_pc(vcpu, rt)) {
> -		/* IO memory trying to read/write pc */
> -		kvm_inject_pabt(vcpu, kvm_vcpu_get_hfar(vcpu));
> -		return 1;

I haven't checked Marc's patch yet but I would get rid of
kvm_vcpu_reg_is_pc() as well, that's the only place where it is used.

-- 
Catalin