[PATCH] arm64: KVM: Add 48bit PA support for EL2 translations

[Marc Zyngier]

Tirumalesh,

On Fri, 5 Feb 2016 11:28:16 -0800
<tchalamarla at caviumnetworks.com> wrote:

> From: Tirumalesh Chalamarla <tchalamarla at caviumnetworks.com>
> 
> Systems where RAM is at higher address and MultiNode systems

What influence a "MultiNode" (for whatever definition of node you have)
system has on the memory map?

> whith higher address RAM, 39 bit max PA for EL2 translations
> is not sufficient.

I read it as 40 bits, not 39.

> 
> This patch sets max possible PA(48bit) for EL2 translations.

From the ARM ARM:

"If {I}PS is programmed to a value larger than the implemented physical
address size, then the PE behaves as if programmed with the implemented
physical address size, but software must not rely on this behavior.
That is, the output address size is never larger than the implemented
physical address size."

> Signed-off-by: Tirumalesh Chalamarla <tchalamarla at caviumnetworks.com>
> ---
>  arch/arm64/include/asm/kvm_arm.h | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm64/include/asm/kvm_arm.h b/arch/arm64/include/asm/kvm_arm.h
> index 738a95f..3fc38a4 100644
> --- a/arch/arm64/include/asm/kvm_arm.h
> +++ b/arch/arm64/include/asm/kvm_arm.h
> @@ -98,7 +98,12 @@
>  #define TCR_EL2_RES1	((1 << 31) | (1 << 23))
>  #define TCR_EL2_TBI	(1 << 20)
>  #define TCR_EL2_PS	(7 << 16)
> +#define TCR_EL2_PS_32B	(0 << 16)
> +#define TCR_EL2_PS_36B	(1 << 16)
>  #define TCR_EL2_PS_40B	(2 << 16)
> +#define TCR_EL2_PS_42B	(3 << 16)
> +#define TCR_EL2_PS_44B	(4 << 16)
> +#define TCR_EL2_PS_48B	(5 << 16)
>  #define TCR_EL2_TG0	(1 << 14)
>  #define TCR_EL2_SH0	(3 << 12)
>  #define TCR_EL2_ORGN0	(3 << 10)
> @@ -107,7 +112,10 @@
>  #define TCR_EL2_MASK	(TCR_EL2_TG0 | TCR_EL2_SH0 | \
>  			 TCR_EL2_ORGN0 | TCR_EL2_IRGN0 | TCR_EL2_T0SZ)
>  
> -#define TCR_EL2_FLAGS	(TCR_EL2_RES1 | TCR_EL2_PS_40B)
> +/*
> + * Assign the highest possible PS.
> + */
> +#define TCR_EL2_FLAGS	(TCR_EL2_RES1 | TCR_EL2_PS_48B)
>  
>  /* VTCR_EL2 Registers bits */
>  #define VTCR_EL2_RES1		(1 << 31)

So this is a NAK. If you want to address this problem, do it properly
(hint: look at how we deal with VTCR_EL2).

Thanks,

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