arm64: 48-bit Physical Address support
cover arm64: 48-bit Physical Address support
From: Radha Mohan Chintakuntla <[email protected]>
This patch adds support for ARMv8 platforms that have 48-bit Physical Address implemented. The number of levels for translating VA to PA has not changed.
Radha Mohan Chintakuntla (1):
arm64: Add support for 48-bit Physical Address
arch/arm64/Kconfig | 3 +++ arch/arm64/include/asm/kvm_arm.h | 9 ++++++++- arch/arm64/include/asm/pgtable-hwdef.h | 15 +++++++++++++++ arch/arm64/mm/proc.S | 2 +- 4 files changed, 27 insertions(+), 2 deletions(-)
RFC PATCH 1/1 arm64: Add support for 48-bit Physical Address
From: Radha Mohan Chintakuntla <[email protected]>
Added support for ARMv8 platforms that have 48-bit physical address implemented. There is no change in the VA bits and levels of translation.
--- arch/arm64/Kconfig | 3 +++ arch/arm64/include/asm/kvm_arm.h | 9 ++++++++- arch/arm64/include/asm/pgtable-hwdef.h | 15 +++++++++++++++ arch/arm64/mm/proc.S | 2 +- 4 files changed, 27 insertions(+), 2 deletions(-)
arch/arm64/Kconfig +3 -0
@@ -106,6 +106,9 @@ config IOMMU_HELPER config KERNEL_MODE_NEON def_bool y +config ARCH_SUPPORTS_48BIT_PA + def_bool n + source "init/Kconfig" source "kernel/Kconfig.freezer"
arch/arm64/include/asm/kvm_arm.h +8 -1
@@ -107,6 +107,7 @@/* VTCR_EL2 Registers bits */ #define VTCR_EL2_PS_MASK (7 << 16)#define VTCR_EL2_PS_40B (2 << 16) +#define VTCR_EL2_PS_48B (5 << 16)#define VTCR_EL2_TG0_MASK (1 << 14)#define VTCR_EL2_TG0_4K (0 << 14)#define VTCR_EL2_TG0_64K (1 << 14)
@@ -121,6 +122,12 @@#define VTCR_EL2_T0SZ_MASK 0x3f#define VTCR_EL2_T0SZ_40B 24 +#ifdef CONFIG_ARCH_SUPPORTS_48BIT_PA +#define VTCR_EL2_PS_BITS VTCR_EL2_PS_48B +#else +#define VTCR_EL2_PS_BITS VTCR_EL2_PS_40B +#endif +#ifdef CONFIG_ARM64_64K_PAGES /* * Stage2 translation configuration:
@@ -129,7 +136,7 @@* 64kB pages (TG0 = 1) * 2 level page tables (SL = 1) */ -#define VTCR_EL2_FLAGS (VTCR_EL2_PS_40B | VTCR_EL2_TG0_64K | \ +#define VTCR_EL2_FLAGS (VTCR_EL2_PS_BITS | VTCR_EL2_TG0_64K | \ VTCR_EL2_SH0_INNER | VTCR_EL2_ORGN0_WBWA | \ VTCR_EL2_IRGN0_WBWA | VTCR_EL2_SL0_LVL1 | \ VTCR_EL2_T0SZ_40B)
arch/arm64/include/asm/pgtable-hwdef.h +15 -0
@@ -99,10 +99,18 @@#define PMD_HYP PMD_SECT_USER#define PTE_HYP PTE_USER +#ifdef CONFIG_ARCH_SUPPORTS_48BIT_PA +/* +* 48-bit physical address supported. +*/ +#define PHYS_MASK_SHIFT (48) +#else /* * 40-bit physical address supported. */ #define PHYS_MASK_SHIFT (40) +#endif +#define PHYS_MASK ((UL(1) << PHYS_MASK_SHIFT) - 1)/*
@@ -123,7 +131,14 @@#define TCR_TG0_64K (UL(1) << 14)#define TCR_TG1_64K (UL(1) << 30)#define TCR_IPS_40BIT (UL(2) << 32) +#define TCR_IPS_48BIT (UL(5) << 32)#define TCR_ASID16 (UL(1) << 36)#define TCR_TBI0 (UL(1) << 37) +#ifdef CONFIG_ARCH_SUPPORTS_48BIT_PA +#define TCR_IPS_BITS TCR_IPS_48BIT +#else +#define TCR_IPS_BITS TCR_IPS_40BIT +#endif +#endif
arch/arm64/mm/proc.S +1 -1
@@ -215,7 +215,7 @@ ENTRY(__cpu_setup) * Set/prepare TCR and TTBR. We use 512GB (39-bit) address range for * both user and kernel. */ - ldr x10, =TCR_TxSZ(VA_BITS) | TCR_FLAGS | TCR_IPS_40BIT | \ + ldr x10, =TCR_TxSZ(VA_BITS) | TCR_FLAGS | TCR_IPS_BITS | \ TCR_ASID16 | TCR_TBI0 | (1 << 31) #ifdef CONFIG_ARM64_64K_PAGES orr x10, x10, TCR_TG0_64K
Signature · 1 line
1.7.1
RFC PATCH 1/1 arm64: Add support for 48-bit Physical Address
Hi Radha,
Quoted 5 lines · On 02/26/2014 12:58 AM, mohun106 at gmail.com wrote:
From: Radha Mohan Chintakuntla <[email protected]>
Added support for ARMv8 platforms that have 48-bit physical address implemented. There is no change in the VA bits and levels of translation.
Is there benefit to using build-time constants for this information? Why not look up the physical address range at run time, so that a single kernel binary could run on any system?
Thanks,
Christopher
Signature · 3 lines
Employee of Qualcomm Innovation Center, Inc. Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted by the Linux Foundation.
Catalin Marinas
maintainerRFC PATCH 1/1 arm64: Add support for 48-bit Physical Address
Quoted 10 lines · Christopher Covington
Quoted 5 lines · On 02/26/2014 12:58 AM, mohun106 at gmail.com wrote:
From: Radha Mohan Chintakuntla <[email protected]>
Added support for ARMv8 platforms that have 48-bit physical address implemented. There is no change in the VA bits and levels of translation.
Is there benefit to using build-time constants for this information? Why not look up the physical address range at run time, so that a single kernel binary could run on any system?
I agree, we don't need a Kconfig option for this, just check the available (intermediate) physical address bits at boot-time (ID_AA64MMFR0_EL1) and program the TCR_ELx registers accordingly.
(for whatever reason, I got the same patches in private and I replied with the same but they haven't been changed before appearing on the list)
Signature · 1 line
Catalin
RFC PATCH 1/1 arm64: Add support for 48-bit Physical Address
Quoted 8 lines · Catalin Marinas
Quoted 7 lines · Christopher Covington
Quoted 5 lines · On 02/26/2014 12:58 AM, mohun106 at gmail.com wrote:
From: Radha Mohan Chintakuntla <[email protected]>
Added support for ARMv8 platforms that have 48-bit physical address implemented. There is no change in the VA bits and levels of translation.
Quoted 8 lines
Quoted 3 lines
Is there benefit to using build-time constants for this information? Why not look up the physical address range at run time, so that a single kernel binary could run on any system?
I agree, we don't need a
Kconfigoption for this, just check the available (intermediate) physical address bits at boot-time (ID_AA64MMFR0_EL1) and program theTCR_ELxregisters accordingly.
OK, I will redo with the proposed changes.
Quoted 4 lines
(for whatever reason, I got the same patches in private and I replied with the same but they haven't been changed before appearing on the list)
Oops, I did miss that mail.
regards,
Radha
Quoted 2 lines
--
Catalin