[PATCH 0/4] Introduce PMSAv8 memory protection unit

[PATCH 0/4] Introduce PMSAv8 memory protection unit

[Vladimir Murzin]

Hi,

This series adds support for PMSAv8 MPU defined by ARMv8R/M
architecture.

Changelog:

    RFC -> v1
        - Rebased on v4.16-rc1
	- Added Tested-by for v7M bits from Andras

Thanks!

Vladimir Murzin (4):
  ARM: NOMMU: Reorganise __setup_mpu
  ARM: NOMMU: Postpone MPU activation till __after_proc_init
  ARM: NOMMU: Make _stext and _end meet PMSAv8 alignment restrictions
  ARM: NOMMU: Support PMSAv8 MPU

 arch/arm/include/asm/mpu.h    |  52 ++++++-
 arch/arm/include/asm/v7m.h    |   8 ++
 arch/arm/kernel/asm-offsets.c |   2 +
 arch/arm/kernel/head-nommu.S  | 180 +++++++++++++++++++++----
 arch/arm/kernel/vmlinux.lds.S |   8 ++
 arch/arm/mm/Makefile          |   2 +-
 arch/arm/mm/nommu.c           |   6 +
 arch/arm/mm/pmsa-v8.c         | 307 ++++++++++++++++++++++++++++++++++++++++++
 8 files changed, 538 insertions(+), 27 deletions(-)
 create mode 100644 arch/arm/mm/pmsa-v8.c

-- 
2.0.0

[PATCH 1/4] ARM: NOMMU: Reorganise __setup_mpu

[Vladimir Murzin]

Currently, we have mixed code placement between .head.text and .text
depends on configuration we are building:

_text                           M       R(UP)   R(SMP)
======================================================
 __setup_mpu                    __HEAD  __HEAD  text
 __after_proc_init              __HEAD  __HEAD  text
 __mmap_switched                text    text    text

We are going to support another variant of MPU which is different to
PMSAv7 in sense overlapping MPU regions are not allowed, so this patch
makes boundaries between these sections precise and consistent:

_text                           M       R(UP)   R(SMP)
======================================================
 __setup_mpu                    __HEAD  __HEAD  __HEAD
 __after_proc_init              text    text    text
 __mmap_switched                text    text    text

Additionally, it paves a path to postpone MPU activation till
__after_proc_init where we do set SCTLR anyway and can return
directly to __mmap_switched.

Tested-by: Szemz? Andr?s <sza@esh.hu>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
---
 arch/arm/kernel/head-nommu.S | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)

diff --git a/arch/arm/kernel/head-nommu.S b/arch/arm/kernel/head-nommu.S
index 0d17187..aaa25a6 100644
--- a/arch/arm/kernel/head-nommu.S
+++ b/arch/arm/kernel/head-nommu.S
@@ -75,8 +75,8 @@ ENTRY(stext)
 	ldr	r12, [r10, #PROCINFO_INITFUNC]
 	add	r12, r12, r10
 	ret	r12
-1:	bl	__after_proc_init
-	b	__mmap_switched
+1:	ldr	lr, =__mmap_switched
+	b	__after_proc_init
 ENDPROC(stext)
 
 #ifdef CONFIG_SMP
@@ -123,6 +123,7 @@ __secondary_data:
 /*
  * Set the Control Register and Read the process ID.
  */
+	.text
 __after_proc_init:
 #ifdef CONFIG_CPU_CP15
 	/*
@@ -202,6 +203,7 @@ ENDPROC(__after_proc_init)
  *
  * r6: Value to be written to DRSR (and IRSR if required) for PMSAv7_RAM_REGION
 */
+	__HEAD
 
 ENTRY(__setup_mpu)
 
@@ -301,6 +303,7 @@ ENDPROC(__setup_pmsa_v7)
  * r6: pointer at mpu_rgn_info
  */
 
+	.text
 ENTRY(__secondary_setup_mpu)
 	/* Use MPU region info supplied by __cpu_up */
 	ldr	r6, [r7]			@ get secondary_data.mpu_rgn_info
-- 
2.0.0

[PATCH 2/4] ARM: NOMMU: Postpone MPU activation till __after_proc_init

[Vladimir Murzin]

This patch postpone MPU activation till __after_proc_init (which is
placed in .text section) rather than doing it in __setup_mpu. It
allows us ignore used-only-once .head.text section while programming
PMSAv8 MPU (for PMSAv7 it stays covered anyway).

Tested-by: Szemz? Andr?s <sza@esh.hu>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
---
 arch/arm/kernel/head-nommu.S | 45 ++++++++++++++++++++++----------------------
 1 file changed, 22 insertions(+), 23 deletions(-)

diff --git a/arch/arm/kernel/head-nommu.S b/arch/arm/kernel/head-nommu.S
index aaa25a6..482936a 100644
--- a/arch/arm/kernel/head-nommu.S
+++ b/arch/arm/kernel/head-nommu.S
@@ -125,11 +125,24 @@ __secondary_data:
  */
 	.text
 __after_proc_init:
+#ifdef CONFIG_ARM_MPU
+M_CLASS(movw	r12, #:lower16:BASEADDR_V7M_SCB)
+M_CLASS(movt	r12, #:upper16:BASEADDR_V7M_SCB)
+M_CLASS(ldr	r3, [r12, 0x50])
+AR_CLASS(mrc	p15, 0, r3, c0, c1, 4)          @ Read ID_MMFR0
+	and	r3, r3, #(MMFR0_PMSA)           @ PMSA field
+	teq	r3, #(MMFR0_PMSAv7)             @ PMSA v7
+#endif
 #ifdef CONFIG_CPU_CP15
 	/*
 	 * CP15 system control register value returned in r0 from
 	 * the CPU init function.
 	 */
+
+#ifdef CONFIG_ARM_MPU
+	biceq	r0, r0, #CR_BR			@ Disable the 'default mem-map'
+	orreq	r0, r0, #CR_M			@ Set SCTRL.M (MPU on)
+#endif
 #if defined(CONFIG_ALIGNMENT_TRAP) && __LINUX_ARM_ARCH__ < 6
 	orr	r0, r0, #CR_A
 #else
@@ -145,7 +158,15 @@ __after_proc_init:
 	bic	r0, r0, #CR_I
 #endif
 	mcr	p15, 0, r0, c1, c0, 0		@ write control reg
+	isb
 #elif defined (CONFIG_CPU_V7M)
+#ifdef CONFIG_ARM_MPU
+	ldreq	r3, [r12, MPU_CTRL]
+	biceq	r3, #MPU_CTRL_PRIVDEFENA
+	orreq	r3, #MPU_CTRL_ENABLE
+	streq	r3, [r12, MPU_CTRL]
+	isb
+#endif
 	/* For V7M systems we want to modify the CCR similarly to the SCTLR */
 #ifdef CONFIG_CPU_DCACHE_DISABLE
 	bic	r0, r0, #V7M_SCB_CCR_DC
@@ -156,9 +177,7 @@ __after_proc_init:
 #ifdef CONFIG_CPU_ICACHE_DISABLE
 	bic	r0, r0, #V7M_SCB_CCR_IC
 #endif
-	movw	r3, #:lower16:(BASEADDR_V7M_SCB + V7M_SCB_CCR)
-	movt	r3, #:upper16:(BASEADDR_V7M_SCB + V7M_SCB_CCR)
-	str	r0, [r3]
+	str	r0, [r12, V7M_SCB_CCR]
 #endif /* CONFIG_CPU_CP15 elif CONFIG_CPU_V7M */
 	ret	lr
 ENDPROC(__after_proc_init)
@@ -282,19 +301,6 @@ M_CLASS(ldr    r0, [r12, #MPU_TYPE])
 	setup_region r0, r5, r6, PMSAv7_INSTR_SIDE, r12	@ XIP_PHYS_ADDR, shared, enabled
 3:	isb
 #endif
-
-	/* Enable the MPU */
-AR_CLASS(mrc	p15, 0, r0, c1, c0, 0)		@ Read SCTLR
-AR_CLASS(bic	r0, r0, #CR_BR)			@ Disable the 'default mem-map'
-AR_CLASS(orr	r0, r0, #CR_M)			@ Set SCTRL.M (MPU on)
-AR_CLASS(mcr	p15, 0, r0, c1, c0, 0)		@ Enable MPU
-
-M_CLASS(ldr	r0, [r12, #MPU_CTRL])
-M_CLASS(bic	r0, #MPU_CTRL_PRIVDEFENA)
-M_CLASS(orr	r0, #MPU_CTRL_ENABLE)
-M_CLASS(str	r0, [r12, #MPU_CTRL])
-	isb
-
 	ret	lr
 ENDPROC(__setup_pmsa_v7)
 
@@ -352,13 +358,6 @@ ENTRY(__secondary_setup_pmsa_v7)
 	cmp	r4, #0
 	bgt	1b
 
-	/* Enable the MPU */
-	mrc	p15, 0, r0, c1, c0, 0		@ Read SCTLR
-	bic	r0, r0, #CR_BR			@ Disable the 'default mem-map'
-	orr	r0, r0, #CR_M			@ Set SCTRL.M (MPU on)
-	mcr	p15, 0, r0, c1, c0, 0		@ Enable MPU
-	isb
-
 	ret	lr
 ENDPROC(__secondary_setup_pmsa_v7)
 
-- 
2.0.0

[PATCH 3/4] ARM: NOMMU: Make _stext and _end meet PMSAv8 alignment restrictions

[Vladimir Murzin]

PMSAv8 require addresses to be either 32 (for M-class) or 64 bytes
(for R-class) aligned. To keep it simple align to 64 bytes always.

For XIP we already have alignment restrictions in place from PMSAv7,
but they can be potentially relaxed for PMSAv8.

Tested-by: Szemz? Andr?s <sza@esh.hu>
Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
---
 arch/arm/kernel/vmlinux.lds.S | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/arch/arm/kernel/vmlinux.lds.S b/arch/arm/kernel/vmlinux.lds.S
index 84a1ae3..edcf07b 100644
--- a/arch/arm/kernel/vmlinux.lds.S
+++ b/arch/arm/kernel/vmlinux.lds.S
@@ -8,6 +8,8 @@
 #include "vmlinux-xip.lds.S"
 #else
 
+#include <linux/sizes.h>
+
 #include <asm-generic/vmlinux.lds.h>
 #include <asm/cache.h>
 #include <asm/thread_info.h>
@@ -102,6 +104,9 @@ SECTIONS
 	. = ALIGN(1<<SECTION_SHIFT);
 #endif
 
+#ifdef CONFIG_ARM_MPU
+	. = ALIGN(SZ_64);
+#endif
 	.text : {			/* Real text segment		*/
 		_stext = .;		/* Text and read-only data	*/
 			IDMAP_TEXT
@@ -295,6 +300,9 @@ SECTIONS
 #endif
 
 	BSS_SECTION(0, 0, 0)
+#ifdef CONFIG_ARM_MPU
+	. = ALIGN(SZ_64);
+#endif
 	_end = .;
 
 	STABS_DEBUG
-- 
2.0.0

[PATCH 4/4] ARM: NOMMU: Support PMSAv8 MPU

[Vladimir Murzin]

ARMv8R/M architecture defines new memory protection scheme - PMSAv8
which is not compatible with PMSAv7.

Key differences to PMSAv7 are:
 - Region geometry is defined by base and limit addresses
 - Addresses need to be either 32 or 64 byte aligned
 - No region priority due to overlapping regions are not allowed
 - It is unified, i.e. no distinction between data/instruction regions
 - Memory attributes are controlled via MAIR

This patch implements support for PMSAv8 MPU defined by ARMv8R/M
architecture.

Signed-off-by: Vladimir Murzin <vladimir.murzin@arm.com>
---
 arch/arm/include/asm/mpu.h    |  52 ++++++-
 arch/arm/include/asm/v7m.h    |   8 ++
 arch/arm/kernel/asm-offsets.c |   2 +
 arch/arm/kernel/head-nommu.S  | 132 ++++++++++++++++++
 arch/arm/mm/Makefile          |   2 +-
 arch/arm/mm/nommu.c           |   6 +
 arch/arm/mm/pmsa-v8.c         | 307 ++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 505 insertions(+), 4 deletions(-)
 create mode 100644 arch/arm/mm/pmsa-v8.c

diff --git a/arch/arm/include/asm/mpu.h b/arch/arm/include/asm/mpu.h
index fbde275..5e088c8 100644
--- a/arch/arm/include/asm/mpu.h
+++ b/arch/arm/include/asm/mpu.h
@@ -12,6 +12,7 @@
 /* ID_MMFR0 data relevant to MPU */
 #define MMFR0_PMSA		(0xF << 4)
 #define MMFR0_PMSAv7		(3 << 4)
+#define MMFR0_PMSAv8		(4 << 4)
 
 /* MPU D/I Size Register fields */
 #define PMSAv7_RSR_SZ		1
@@ -47,12 +48,43 @@
 #define PMSAv7_AP_PL1RW_PL0R0	(0x2 << 8)
 #define PMSAv7_AP_PL1RW_PL0NA	(0x1 << 8)
 
+#define PMSAv8_BAR_XN		1
+
+#define PMSAv8_LAR_EN		1
+#define PMSAv8_LAR_IDX(n)	(((n) & 0x7) << 1)
+
+
+#define PMSAv8_AP_PL1RW_PL0NA	(0 << 1)
+#define PMSAv8_AP_PL1RW_PL0RW	(1 << 1)
+#define PMSAv8_AP_PL1RO_PL0RO	(3 << 1)
+
+#ifdef CONFIG_SMP
+#define PMSAv8_RGN_SHARED	(3 << 3) // inner sharable
+#else
+#define PMSAv8_RGN_SHARED	(0 << 3)
+#endif
+
+#define PMSAv8_RGN_DEVICE_nGnRnE	0
+#define PMSAv8_RGN_NORMAL		1
+
+#define PMSAv8_MAIR(attr, mt)	((attr) << ((mt) * 8))
+
+#ifdef CONFIG_CPU_V7M
+#define PMSAv8_MINALIGN		32
+#else
+#define PMSAv8_MINALIGN		64
+#endif
+
 /* For minimal static MPU region configurations */
 #define PMSAv7_PROBE_REGION	0
 #define PMSAv7_BG_REGION	1
 #define PMSAv7_RAM_REGION	2
 #define PMSAv7_ROM_REGION	3
 
+/* Fixed for PMSAv8 only */
+#define PMSAv8_XIP_REGION	0
+#define PMSAv8_KERNEL_REGION	1
+
 /* Maximum number of regions Linux is interested in */
 #define MPU_MAX_REGIONS	16
 
@@ -63,9 +95,18 @@
 
 struct mpu_rgn {
 	/* Assume same attributes for d/i-side  */
-	u32 drbar;
-	u32 drsr;
-	u32 dracr;
+	union {
+		u32 drbar;   /* PMSAv7 */
+		u32 prbar;   /* PMSAv8 */
+	};
+	union {
+		u32 drsr;   /* PMSAv7 */
+		u32 prlar;  /* PMSAv8 */
+	};
+	union {
+		u32 dracr;  /* PMSAv7 */
+		u32 unused; /* not used in PMSAv8 */
+	};
 };
 
 struct mpu_rgn_info {
@@ -76,10 +117,15 @@ extern struct mpu_rgn_info mpu_rgn_info;
 
 #ifdef CONFIG_ARM_MPU
 extern void __init pmsav7_adjust_lowmem_bounds(void);
+extern void __init pmsav8_adjust_lowmem_bounds(void);
+
 extern void __init pmsav7_setup(void);
+extern void __init pmsav8_setup(void);
 #else
 static inline void pmsav7_adjust_lowmem_bounds(void) {};
+static inline void pmsav8_adjust_lowmem_bounds(void) {};
 static inline void pmsav7_setup(void) {};
+static inline void pmsav8_setup(void) {};
 #endif
 
 #endif /* __ASSEMBLY__ */
diff --git a/arch/arm/include/asm/v7m.h b/arch/arm/include/asm/v7m.h
index aba49e0..187ccf6 100644
--- a/arch/arm/include/asm/v7m.h
+++ b/arch/arm/include/asm/v7m.h
@@ -68,6 +68,14 @@
 #define PMSAv7_RBAR		0x9c
 #define PMSAv7_RASR		0xa0
 
+#define PMSAv8_RNR		0x98
+#define PMSAv8_RBAR		0x9c
+#define PMSAv8_RLAR		0xa0
+#define PMSAv8_RBAR_A(n)	(PMSAv8_RBAR + 8*(n))
+#define PMSAv8_RLAR_A(n)	(PMSAv8_RLAR + 8*(n))
+#define PMSAv8_MAIR0		0xc0
+#define PMSAv8_MAIR1		0xc4
+
 /* Cache opeartions */
 #define	V7M_SCB_ICIALLU		0x250	/* I-cache invalidate all to PoU */
 #define	V7M_SCB_ICIMVAU		0x258	/* I-cache invalidate by MVA to PoU */
diff --git a/arch/arm/kernel/asm-offsets.c b/arch/arm/kernel/asm-offsets.c
index 250a985..27c5381 100644
--- a/arch/arm/kernel/asm-offsets.c
+++ b/arch/arm/kernel/asm-offsets.c
@@ -197,6 +197,8 @@ int main(void)
   DEFINE(MPU_RGN_DRBAR,	offsetof(struct mpu_rgn, drbar));
   DEFINE(MPU_RGN_DRSR,	offsetof(struct mpu_rgn, drsr));
   DEFINE(MPU_RGN_DRACR,	offsetof(struct mpu_rgn, dracr));
+  DEFINE(MPU_RGN_PRBAR,	offsetof(struct mpu_rgn, prbar));
+  DEFINE(MPU_RGN_PRLAR,	offsetof(struct mpu_rgn, prlar));
 #endif
   return 0; 
 }
diff --git a/arch/arm/kernel/head-nommu.S b/arch/arm/kernel/head-nommu.S
index 482936a..cdc1177 100644
--- a/arch/arm/kernel/head-nommu.S
+++ b/arch/arm/kernel/head-nommu.S
@@ -132,6 +132,25 @@ M_CLASS(ldr	r3, [r12, 0x50])
 AR_CLASS(mrc	p15, 0, r3, c0, c1, 4)          @ Read ID_MMFR0
 	and	r3, r3, #(MMFR0_PMSA)           @ PMSA field
 	teq	r3, #(MMFR0_PMSAv7)             @ PMSA v7
+	beq	1f
+	teq	r3, #(MMFR0_PMSAv8)		@ PMSA v8
+	/*
+	 * Memory region attributes for PMSAv8:
+	 *
+	 *   n = AttrIndx[2:0]
+	 *                      n       MAIR
+	 *   DEVICE_nGnRnE      000     00000000
+	 *   NORMAL             001     11111111
+	 */
+	ldreq	r3, =PMSAv8_MAIR(0x00, PMSAv8_RGN_DEVICE_nGnRnE) | \
+		     PMSAv8_MAIR(0xff, PMSAv8_RGN_NORMAL)
+AR_CLASS(mcreq	p15, 0, r3, c10, c2, 0)		@ MAIR 0
+M_CLASS(streq	r3, [r12, #PMSAv8_MAIR0])
+	moveq	r3, #0
+AR_CLASS(mcreq	p15, 0, r3, c10, c2, 1)		@ MAIR 1
+M_CLASS(streq	r3, [r12, #PMSAv8_MAIR1])
+
+1:
 #endif
 #ifdef CONFIG_CPU_CP15
 	/*
@@ -235,6 +254,8 @@ M_CLASS(ldr	r0, [r12, 0x50])
 	and	r0, r0, #(MMFR0_PMSA)		@ PMSA field
 	teq	r0, #(MMFR0_PMSAv7)		@ PMSA v7
 	beq	__setup_pmsa_v7
+	teq	r0, #(MMFR0_PMSAv8)		@ PMSA v8
+	beq	__setup_pmsa_v8
 
 	ret	lr
 ENDPROC(__setup_mpu)
@@ -304,6 +325,88 @@ M_CLASS(ldr    r0, [r12, #MPU_TYPE])
 	ret	lr
 ENDPROC(__setup_pmsa_v7)
 
+ENTRY(__setup_pmsa_v8)
+	mov	r0, #0
+AR_CLASS(mcr	p15, 0, r0, c6, c2, 1)		@ PRSEL
+M_CLASS(str	r0, [r12, #PMSAv8_RNR])
+	isb
+
+#ifdef CONFIG_XIP_KERNEL
+	ldr	r5, =CONFIG_XIP_PHYS_ADDR		@ ROM start
+	ldr     r6, =(_exiprom)				@ ROM end
+	sub	r6, r6, #1
+	bic	r6, r6, #(PMSAv8_MINALIGN - 1)
+
+	orr	r5, r5, #(PMSAv8_AP_PL1RW_PL0NA | PMSAv8_RGN_SHARED)
+	orr	r6, r6, #(PMSAv8_LAR_IDX(PMSAv8_RGN_NORMAL) | PMSAv8_LAR_EN)
+
+AR_CLASS(mcr	p15, 0, r5, c6, c8, 0)			@ PRBAR0
+AR_CLASS(mcr	p15, 0, r6, c6, c8, 1)			@ PRLAR0
+M_CLASS(str	r5, [r12, #PMSAv8_RBAR_A(0)])
+M_CLASS(str	r6, [r12, #PMSAv8_RLAR_A(0)])
+#endif
+
+	ldr	r5, =KERNEL_START
+	ldr	r6, =KERNEL_END
+	sub	r6, r6, #1
+	bic	r6, r6, #(PMSAv8_MINALIGN - 1)
+
+	orr	r5, r5, #(PMSAv8_AP_PL1RW_PL0NA | PMSAv8_RGN_SHARED)
+	orr	r6, r6, #(PMSAv8_LAR_IDX(PMSAv8_RGN_NORMAL) | PMSAv8_LAR_EN)
+
+AR_CLASS(mcr	p15, 0, r5, c6, c8, 4)			@ PRBAR1
+AR_CLASS(mcr	p15, 0, r6, c6, c8, 5)			@ PRLAR1
+M_CLASS(str	r5, [r12, #PMSAv8_RBAR_A(1)])
+M_CLASS(str	r6, [r12, #PMSAv8_RLAR_A(1)])
+
+	/* Setup Background: 0x0 - min(KERNEL_START, XIP_PHYS_ADDR) */
+#ifdef CONFIG_XIP_KERNEL
+	ldr	r6, =KERNEL_START
+	ldr	r5, =CONFIG_XIP_PHYS_ADDR
+	cmp	r6, r5
+	movcs	r6, r5
+#else
+	ldr	r6, =KERNEL_START
+#endif
+	cmp	r6, #0
+	beq	1f
+
+	mov	r5, #0
+	sub	r6, r6, #1
+	bic	r6, r6, #(PMSAv8_MINALIGN - 1)
+
+	orr	r5, r5, #(PMSAv8_AP_PL1RW_PL0NA | PMSAv8_RGN_SHARED | PMSAv8_BAR_XN)
+	orr	r6, r6, #(PMSAv8_LAR_IDX(PMSAv8_RGN_DEVICE_nGnRnE) | PMSAv8_LAR_EN)
+
+AR_CLASS(mcr	p15, 0, r5, c6, c9, 0)			@ PRBAR2
+AR_CLASS(mcr	p15, 0, r6, c6, c9, 1)			@ PRLAR2
+M_CLASS(str	r5, [r12, #PMSAv8_RBAR_A(2)])
+M_CLASS(str	r6, [r12, #PMSAv8_RLAR_A(2)])
+
+1:
+	/* Setup Background: max(KERNEL_END, _exiprom) - 0xffffffff */
+#ifdef CONFIG_XIP_KERNEL
+	ldr	r5, =KERNEL_END
+	ldr	r6, =(_exiprom)
+	cmp	r5, r6
+	movcc	r5, r6
+#else
+	ldr	r5, =KERNEL_END
+#endif
+	mov	r6, #0xffffffff
+	bic	r6, r6, #(PMSAv8_MINALIGN - 1)
+
+	orr	r5, r5, #(PMSAv8_AP_PL1RW_PL0NA | PMSAv8_RGN_SHARED | PMSAv8_BAR_XN)
+	orr	r6, r6, #(PMSAv8_LAR_IDX(PMSAv8_RGN_DEVICE_nGnRnE) | PMSAv8_LAR_EN)
+
+AR_CLASS(mcr	p15, 0, r5, c6, c9, 4)			@ PRBAR3
+AR_CLASS(mcr	p15, 0, r6, c6, c9, 5)			@ PRLAR3
+M_CLASS(str	r5, [r12, #PMSAv8_RBAR_A(3)])
+M_CLASS(str	r6, [r12, #PMSAv8_RLAR_A(3)])
+
+	ret	lr
+ENDPROC(__setup_pmsa_v8)
+
 #ifdef CONFIG_SMP
 /*
  * r6: pointer@mpu_rgn_info
@@ -319,6 +422,8 @@ ENTRY(__secondary_setup_mpu)
 	and	r0, r0, #(MMFR0_PMSA)		@ PMSA field
 	teq	r0, #(MMFR0_PMSAv7)		@ PMSA v7
 	beq	__secondary_setup_pmsa_v7
+	teq	r0, #(MMFR0_PMSAv8)		@ PMSA v8
+	beq	__secondary_setup_pmsa_v8
  	b	__error_p
 ENDPROC(__secondary_setup_mpu)
 
@@ -361,6 +466,33 @@ ENTRY(__secondary_setup_pmsa_v7)
 	ret	lr
 ENDPROC(__secondary_setup_pmsa_v7)
 
+ENTRY(__secondary_setup_pmsa_v8)
+	ldr	r4, [r6, #MPU_RNG_INFO_USED]
+#ifndef CONFIG_XIP_KERNEL
+	add	r4, r4, #1
+#endif
+	mov	r5, #MPU_RNG_SIZE
+	add	r3, r6, #MPU_RNG_INFO_RNGS
+	mla	r3, r4, r5, r3
+
+1:
+	sub	r3, r3, #MPU_RNG_SIZE
+	sub	r4, r4, #1
+
+	mcr	p15, 0, r4, c6, c2, 1		@ PRSEL
+	isb
+
+	ldr	r5, [r3, #MPU_RGN_PRBAR]
+	ldr	r6, [r3, #MPU_RGN_PRLAR]
+
+	mcr	p15, 0, r5, c6, c3, 0		@ PRBAR
+	mcr	p15, 0, r6, c6, c3, 1           @ PRLAR
+
+	cmp	r4, #0
+	bgt	1b
+
+	ret	lr
+ENDPROC(__secondary_setup_pmsa_v8)
 #endif /* CONFIG_SMP */
 #endif /* CONFIG_ARM_MPU */
 #include "head-common.S"
diff --git a/arch/arm/mm/Makefile b/arch/arm/mm/Makefile
index 9dbb849..d19b209 100644
--- a/arch/arm/mm/Makefile
+++ b/arch/arm/mm/Makefile
@@ -10,7 +10,7 @@ obj-$(CONFIG_MMU)		+= fault-armv.o flush.o idmap.o ioremap.o \
 
 ifneq ($(CONFIG_MMU),y)
 obj-y				+= nommu.o
-obj-$(CONFIG_ARM_MPU)		+= pmsa-v7.o
+obj-$(CONFIG_ARM_MPU)		+= pmsa-v7.o pmsa-v8.o
 endif
 
 obj-$(CONFIG_ARM_PTDUMP_CORE)	+= dump.o
diff --git a/arch/arm/mm/nommu.c b/arch/arm/mm/nommu.c
index edbaa47..5dd6c58 100644
--- a/arch/arm/mm/nommu.c
+++ b/arch/arm/mm/nommu.c
@@ -107,6 +107,9 @@ static void __init adjust_lowmem_bounds_mpu(void)
 	case MMFR0_PMSAv7:
 		pmsav7_adjust_lowmem_bounds();
 		break;
+	case MMFR0_PMSAv8:
+		pmsav8_adjust_lowmem_bounds();
+		break;
 	default:
 		break;
 	}
@@ -120,6 +123,9 @@ static void __init mpu_setup(void)
 	case MMFR0_PMSAv7:
 		pmsav7_setup();
 		break;
+	case MMFR0_PMSAv8:
+		pmsav8_setup();
+		break;
 	default:
 		break;
 	}
diff --git a/arch/arm/mm/pmsa-v8.c b/arch/arm/mm/pmsa-v8.c
new file mode 100644
index 0000000..617a83d
--- /dev/null
+++ b/arch/arm/mm/pmsa-v8.c
@@ -0,0 +1,307 @@
+/*
+ * Based on linux/arch/arm/pmsa-v7.c
+ *
+ * ARM PMSAv8 supporting functions.
+ */
+
+#include <linux/memblock.h>
+#include <linux/range.h>
+
+#include <asm/cp15.h>
+#include <asm/cputype.h>
+#include <asm/mpu.h>
+
+#include <asm/memory.h>
+#include <asm/sections.h>
+
+#include "mm.h"
+
+#ifndef CONFIG_CPU_V7M
+
+#define PRSEL	__ACCESS_CP15(c6, 0, c2, 1)
+#define PRBAR	__ACCESS_CP15(c6, 0, c3, 0)
+#define PRLAR	__ACCESS_CP15(c6, 0, c3, 1)
+
+static inline u32 prlar_read(void)
+{
+	return read_sysreg(PRLAR);
+}
+
+static inline u32 prbar_read(void)
+{
+	return read_sysreg(PRBAR);
+}
+
+static inline void prsel_write(u32 v)
+{
+	write_sysreg(v, PRSEL);
+}
+
+static inline void prbar_write(u32 v)
+{
+	write_sysreg(v, PRBAR);
+}
+
+static inline void prlar_write(u32 v)
+{
+	write_sysreg(v, PRLAR);
+}
+#else
+
+static inline u32 prlar_read(void)
+{
+	return readl_relaxed(BASEADDR_V7M_SCB + PMSAv8_RLAR);
+}
+
+static inline u32 prbar_read(void)
+{
+	return readl_relaxed(BASEADDR_V7M_SCB + PMSAv8_RBAR);
+}
+
+static inline void prsel_write(u32 v)
+{
+	writel_relaxed(v, BASEADDR_V7M_SCB + PMSAv8_RNR);
+}
+
+static inline void prbar_write(u32 v)
+{
+	writel_relaxed(v, BASEADDR_V7M_SCB + PMSAv8_RBAR);
+}
+
+static inline void prlar_write(u32 v)
+{
+	writel_relaxed(v, BASEADDR_V7M_SCB + PMSAv8_RLAR);
+}
+
+#endif
+
+static struct range __initdata io[MPU_MAX_REGIONS];
+static struct range __initdata mem[MPU_MAX_REGIONS];
+
+static unsigned int __initdata mpu_max_regions;
+
+static __init bool is_region_fixed(int number)
+{
+	switch (number) {
+	case PMSAv8_XIP_REGION:
+	case PMSAv8_KERNEL_REGION:
+		return true;
+	default:
+		return false;
+	}
+}
+
+void __init pmsav8_adjust_lowmem_bounds(void)
+{
+	phys_addr_t mem_end;
+	struct memblock_region *reg;
+	bool first = true;
+
+	for_each_memblock(memory, reg) {
+		if (first) {
+			phys_addr_t phys_offset = PHYS_OFFSET;
+
+			/*
+			 * Initially only use memory continuous from
+			 * PHYS_OFFSET */
+			if (reg->base != phys_offset)
+				panic("First memory bank must be contiguous from PHYS_OFFSET");
+			mem_end = reg->base + reg->size;
+			first = false;
+		} else {
+			/*
+			 * memblock auto merges contiguous blocks, remove
+			 * all blocks afterwards in one go (we can't remove
+			 * blocks separately while iterating)
+			 */
+			pr_notice("Ignoring RAM after %pa, memory at %pa ignored\n",
+				  &mem_end, &reg->base);
+			memblock_remove(reg->base, 0 - reg->base);
+			break;
+		}
+	}
+}
+
+static int __init __mpu_max_regions(void)
+{
+	static int max_regions;
+	u32 mpuir;
+
+	if (max_regions)
+		return max_regions;
+
+	mpuir = read_cpuid_mputype();
+
+	max_regions  = (mpuir & MPUIR_DREGION_SZMASK) >> MPUIR_DREGION;
+
+	return max_regions;
+}
+
+static int __init __pmsav8_setup_region(unsigned int number, u32 bar, u32 lar)
+{
+	if (number > mpu_max_regions
+	    || number >= MPU_MAX_REGIONS)
+		return -ENOENT;
+
+	dsb();
+	prsel_write(number);
+	isb();
+	prbar_write(bar);
+	prlar_write(lar);
+
+	mpu_rgn_info.rgns[number].prbar = bar;
+	mpu_rgn_info.rgns[number].prlar = lar;
+
+	mpu_rgn_info.used++;
+
+	return 0;
+}
+
+static int __init pmsav8_setup_ram(unsigned int number, phys_addr_t start,phys_addr_t end)
+{
+	u32 bar, lar;
+
+	if (is_region_fixed(number))
+		return -EINVAL;
+
+	bar = start;
+	lar = (end - 1) & ~(PMSAv8_MINALIGN - 1);;
+
+	bar |= PMSAv8_AP_PL1RW_PL0RW | PMSAv8_RGN_SHARED;
+	lar |= PMSAv8_LAR_IDX(PMSAv8_RGN_NORMAL) | PMSAv8_LAR_EN;
+
+	return __pmsav8_setup_region(number, bar, lar);
+}
+
+static int __init pmsav8_setup_io(unsigned int number, phys_addr_t start,phys_addr_t end)
+{
+	u32 bar, lar;
+
+	if (is_region_fixed(number))
+		return -EINVAL;
+
+	bar = start;
+	lar = (end - 1) & ~(PMSAv8_MINALIGN - 1);;
+
+	bar |= PMSAv8_AP_PL1RW_PL0RW | PMSAv8_RGN_SHARED | PMSAv8_BAR_XN;
+	lar |= PMSAv8_LAR_IDX(PMSAv8_RGN_DEVICE_nGnRnE) | PMSAv8_LAR_EN;
+
+	return __pmsav8_setup_region(number, bar, lar);
+}
+
+static int __init pmsav8_setup_fixed(unsigned int number, phys_addr_t start,phys_addr_t end)
+{
+	u32 bar, lar;
+
+	if (!is_region_fixed(number))
+		return -EINVAL;
+
+	bar = start;
+	lar = (end - 1) & ~(PMSAv8_MINALIGN - 1);
+
+	bar |= PMSAv8_AP_PL1RW_PL0NA | PMSAv8_RGN_SHARED;
+	lar |= PMSAv8_LAR_IDX(PMSAv8_RGN_NORMAL) | PMSAv8_LAR_EN;
+
+	prsel_write(number);
+	isb();
+
+	if (prbar_read() != bar || prlar_read() != lar)
+		return -EINVAL;
+
+	/* Reserved region was set up early, we just need a record for secondaries */
+	mpu_rgn_info.rgns[number].prbar = bar;
+	mpu_rgn_info.rgns[number].prlar = lar;
+
+	mpu_rgn_info.used++;
+
+	return 0;
+}
+
+#ifndef CONFIG_CPU_V7M
+static int __init pmsav8_setup_vector(unsigned int number, phys_addr_t start,phys_addr_t end)
+{
+	u32 bar, lar;
+
+	if (number == PMSAv8_KERNEL_REGION)
+		return -EINVAL;
+
+	bar = start;
+	lar = (end - 1) & ~(PMSAv8_MINALIGN - 1);
+
+	bar |= PMSAv8_AP_PL1RW_PL0NA | PMSAv8_RGN_SHARED;
+	lar |= PMSAv8_LAR_IDX(PMSAv8_RGN_NORMAL) | PMSAv8_LAR_EN;
+
+	return __pmsav8_setup_region(number, bar, lar);
+}
+#endif
+
+void __init pmsav8_setup(void)
+{
+	int i, err = 0;
+	int region = PMSAv8_KERNEL_REGION;
+
+	/* How many regions are supported ? */
+	mpu_max_regions = __mpu_max_regions();
+
+	/* RAM: single chunk of memory */
+	add_range(mem,  ARRAY_SIZE(mem), 0,  memblock.memory.regions[0].base,
+		  memblock.memory.regions[0].base + memblock.memory.regions[0].size);
+
+	/* IO: cover full 4G range */
+	add_range(io, ARRAY_SIZE(io), 0, 0, 0xffffffff);
+
+	/* RAM and IO: exclude kernel */
+	subtract_range(mem, ARRAY_SIZE(mem), __pa(KERNEL_START), __pa(KERNEL_END));
+	subtract_range(io, ARRAY_SIZE(io),  __pa(KERNEL_START), __pa(KERNEL_END));
+
+#ifdef CONFIG_XIP_KERNEL
+	/* RAM and IO: exclude xip */
+	subtract_range(mem, ARRAY_SIZE(mem), CONFIG_XIP_PHYS_ADDR, __pa(_exiprom));
+	subtract_range(io, ARRAY_SIZE(io), CONFIG_XIP_PHYS_ADDR, __pa(_exiprom));
+#endif
+
+#ifndef CONFIG_CPU_V7M
+	/* RAM and IO: exclude vectors */
+	subtract_range(mem, ARRAY_SIZE(mem),  vectors_base, vectors_base + 2 * PAGE_SIZE);
+	subtract_range(io, ARRAY_SIZE(io),  vectors_base, vectors_base + 2 * PAGE_SIZE);
+#endif
+	/* IO: exclude RAM */
+	for (i = 0; i < ARRAY_SIZE(mem); i++)
+		subtract_range(io, ARRAY_SIZE(io), mem[i].start, mem[i].end);
+
+	/* Now program MPU */
+
+#ifdef CONFIG_XIP_KERNEL
+	/* ROM */
+	err |= pmsav8_setup_fixed(PMSAv8_XIP_REGION, CONFIG_XIP_PHYS_ADDR, __pa(_exiprom));
+#endif
+	/* Kernel */
+	err |= pmsav8_setup_fixed(region++, __pa(KERNEL_START), __pa(KERNEL_END));
+
+
+	/* IO */
+	for (i = 0; i < ARRAY_SIZE(io); i++) {
+		if (!io[i].end)
+			continue;
+
+		err |= pmsav8_setup_io(region++, io[i].start, io[i].end);
+	}
+
+	/* RAM */
+	for (i = 0; i < ARRAY_SIZE(mem); i++) {
+		if (!mem[i].end)
+			continue;
+
+		err |= pmsav8_setup_ram(region++, mem[i].start, mem[i].end);
+	}
+
+	/* Vectors */
+#ifndef CONFIG_CPU_V7M
+	err |= pmsav8_setup_vector(region++, vectors_base, vectors_base + 2 * PAGE_SIZE);
+#endif
+	if (err)
+		pr_warn("MPU region initialization failure! %d", err);
+	else
+		pr_info("Using ARM PMSAv8 Compliant MPU. Used %d of %d regions\n",
+			mpu_rgn_info.used, mpu_max_regions);
+}
-- 
2.0.0