[PATCH v3 8/8] ARM: NOMMU: Support MPU in XIP configuration

Alexandre Torgue alexandre.torgue at st.com
Mon Sep 25 05:54:30 PDT 2017


Hi Vlad,

On 09/25/2017 11:05 AM, Vladimir Murzin wrote:
> Currently, there is assumption in early MPU setup code that kernel
> image is located in RAM, which is obviously not true for XIP. To run
> code from ROM we need to make sure that it is covered by MPU. However,
> due to we allocate regions (semi-)dynamically we can run into issue of
> trimming region we are running from in case ROM spawns several MPU
> regions. To help deal with that we enforce minimum alignments for start
> end end of XIP address space as 1MB and 128Kb correspondingly.
> 
> Signed-off-by: Vladimir Murzin <vladimir.murzin at arm.com>


Thanks for the XIP fix.

Tested-by: Alexandre TORGUE <alexandre.torgue at st.com>

regards
Alex


> ---
>   arch/arm/Kconfig-nommu            |  2 +-
>   arch/arm/include/asm/mpu.h        |  3 ++-
>   arch/arm/kernel/head-nommu.S      | 20 ++++++++++++++++
>   arch/arm/kernel/vmlinux-xip.lds.S | 23 ++++++++++++++++++
>   arch/arm/mm/pmsa-v7.c             | 49 +++++++++++++++++++++++++++++++++++----
>   5 files changed, 90 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/arm/Kconfig-nommu b/arch/arm/Kconfig-nommu
> index 930e000..0fad7d9 100644
> --- a/arch/arm/Kconfig-nommu
> +++ b/arch/arm/Kconfig-nommu
> @@ -52,7 +52,7 @@ config REMAP_VECTORS_TO_RAM
>   
>   config ARM_MPU
>          bool 'Use the ARM v7 PMSA Compliant MPU'
> -       depends on !XIP_KERNEL && (CPU_V7 || CPU_V7M)
> +       depends on CPU_V7 || CPU_V7M
>          default y if CPU_V7
>          help
>            Some ARM systems without an MMU have instead a Memory Protection
> diff --git a/arch/arm/include/asm/mpu.h b/arch/arm/include/asm/mpu.h
> index 5db37a6..56ec026 100644
> --- a/arch/arm/include/asm/mpu.h
> +++ b/arch/arm/include/asm/mpu.h
> @@ -41,6 +41,7 @@
>   #endif
>   
>   /* Access permission bits of ACR (only define those that we use)*/
> +#define MPU_AP_PL1RO_PL0NA	(0x5 << 8)
>   #define MPU_AP_PL1RW_PL0RW	(0x3 << 8)
>   #define MPU_AP_PL1RW_PL0R0	(0x2 << 8)
>   #define MPU_AP_PL1RW_PL0NA	(0x1 << 8)
> @@ -49,7 +50,7 @@
>   #define MPU_PROBE_REGION	0
>   #define MPU_BG_REGION		1
>   #define MPU_RAM_REGION		2
> -#define MPU_VECTORS_REGION	3
> +#define MPU_ROM_REGION		3
>   
>   /* Maximum number of regions Linux is interested in */
>   #define MPU_MAX_REGIONS		16
> diff --git a/arch/arm/kernel/head-nommu.S b/arch/arm/kernel/head-nommu.S
> index 8a0718f..2ce4455 100644
> --- a/arch/arm/kernel/head-nommu.S
> +++ b/arch/arm/kernel/head-nommu.S
> @@ -279,6 +279,26 @@ M_CLASS(ldr    r0, [r12, #MPU_TYPE])
>   	setup_region r0, r5, r6, MPU_INSTR_SIDE r12	@ 0x0, BG region, enabled
>   2:	isb
>   
> +#ifdef CONFIG_XIP_KERNEL
> +	set_region_nr r0, #MPU_ROM_REGION, r12
> +	isb
> +
> +	ldr	r5,=(MPU_AP_PL1RO_PL0NA | MPU_RGN_NORMAL)
> +
> +	ldr	r0, =CONFIG_XIP_PHYS_ADDR		@ ROM start
> +	ldr     r6, =(_exiprom)				@ ROM end
> +	sub	r6, r6, r0				@ Minimum size of region to map
> +	clz	r6, r6					@ Region size must be 2^N...
> +	rsb	r6, r6, #31				@ ...so round up region size
> +	lsl	r6, r6, #MPU_RSR_SZ			@ Put size in right field
> +	orr	r6, r6, #(1 << MPU_RSR_EN)		@ Set region enabled bit
> +
> +	setup_region r0, r5, r6, MPU_DATA_SIDE, r12	@ XIP_PHYS_ADDR, shared, enabled
> +	beq	3f					@ Memory-map not unified
> +	setup_region r0, r5, r6, MPU_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'
> diff --git a/arch/arm/kernel/vmlinux-xip.lds.S b/arch/arm/kernel/vmlinux-xip.lds.S
> index 8265b11..c6f7056 100644
> --- a/arch/arm/kernel/vmlinux-xip.lds.S
> +++ b/arch/arm/kernel/vmlinux-xip.lds.S
> @@ -6,6 +6,8 @@
>   /* No __ro_after_init data in the .rodata section - which will always be ro */
>   #define RO_AFTER_INIT_DATA
>   
> +#include <linux/sizes.h>
> +
>   #include <asm-generic/vmlinux.lds.h>
>   #include <asm/cache.h>
>   #include <asm/thread_info.h>
> @@ -205,6 +207,9 @@ SECTIONS
>   	PERCPU_SECTION(L1_CACHE_BYTES)
>   #endif
>   
> +#ifdef CONFIG_ARM_MPU
> +	. = ALIGN(SZ_128K);
> +#endif
>   	_exiprom = .;			/* End of XIP ROM area */
>   	__data_loc = ALIGN(4);		/* location in binary */
>   	. = PAGE_OFFSET + TEXT_OFFSET;
> @@ -322,3 +327,21 @@ ASSERT((__arch_info_end - __arch_info_begin), "no machine record defined")
>    */
>   ASSERT(__hyp_idmap_text_end - (__hyp_idmap_text_start & PAGE_MASK) <= PAGE_SIZE,
>   	"HYP init code too big or misaligned")
> +
> +#ifdef CONFIG_ARM_MPU
> +/*
> + * Due to PMSAv7 restriction on base address and size we have to
> + * enforce minimal alignment restrictions. It was seen that weaker
> + * alignment restriction on _xiprom will likely force XIP address
> + * space spawns multiple MPU regions thus it is likely we run in
> + * situation when we are reprogramming MPU region we run on with
> + * something which doesn't cover reprogramming code itself, so as soon
> + * as we update MPU settings we'd immediately try to execute straight
> + * from background region which is XN.
> + * It seem that alignment in 1M should suit most users.
> + * _exiprom is aligned as 1/8 of 1M so can be covered by subregion
> + * disable
> + */
> +ASSERT(!(_xiprom & (SZ_1M - 1)), "XIP start address may cause MPU programming issues")
> +ASSERT(!(_exiprom & (SZ_128K - 1)), "XIP end address may cause MPU programming issues")
> +#endif
> diff --git a/arch/arm/mm/pmsa-v7.c b/arch/arm/mm/pmsa-v7.c
> index c1f1fc7..27434d7 100644
> --- a/arch/arm/mm/pmsa-v7.c
> +++ b/arch/arm/mm/pmsa-v7.c
> @@ -7,9 +7,11 @@
>   #include <linux/bitops.h>
>   #include <linux/memblock.h>
>   
> +#include <asm/cacheflush.h>
>   #include <asm/cp15.h>
>   #include <asm/cputype.h>
>   #include <asm/mpu.h>
> +#include <asm/sections.h>
>   
>   #include "mm.h"
>   
> @@ -20,6 +22,9 @@ struct region {
>   };
>   
>   static struct region __initdata mem[MPU_MAX_REGIONS];
> +#ifdef CONFIG_XIP_KERNEL
> +static struct region __initdata xip[MPU_MAX_REGIONS];
> +#endif
>   
>   static unsigned int __initdata mpu_min_region_order;
>   static unsigned int __initdata mpu_max_regions;
> @@ -229,7 +234,6 @@ static int __init allocate_region(phys_addr_t base, phys_addr_t size,
>   /* MPU initialisation functions */
>   void __init adjust_lowmem_bounds_mpu(void)
>   {
> -	phys_addr_t phys_offset = PHYS_OFFSET;
>   	phys_addr_t  specified_mem_size, total_mem_size = 0;
>   	struct memblock_region *reg;
>   	bool first = true;
> @@ -256,8 +260,19 @@ void __init adjust_lowmem_bounds_mpu(void)
>   	/* ... and one for vectors */
>   	mem_max_regions--;
>   #endif
> +
> +#ifdef CONFIG_XIP_KERNEL
> +	/* plus some regions to cover XIP ROM */
> +	num = allocate_region(CONFIG_XIP_PHYS_ADDR, __pa(_exiprom) - CONFIG_XIP_PHYS_ADDR,
> +			      mem_max_regions, xip);
> +
> +	mem_max_regions -= num;
> +#endif
> +
>   	for_each_memblock(memory, reg) {
>   		if (first) {
> +			phys_addr_t phys_offset = PHYS_OFFSET;
> +
>   			/*
>   			 * Initially only use memory continuous from
>   			 * PHYS_OFFSET */
> @@ -355,7 +370,7 @@ static int __init __mpu_min_region_order(void)
>   
>   static int __init mpu_setup_region(unsigned int number, phys_addr_t start,
>   				   unsigned int size_order, unsigned int properties,
> -				   unsigned int subregions)
> +				   unsigned int subregions, bool need_flush)
>   {
>   	u32 size_data;
>   
> @@ -374,6 +389,9 @@ static int __init mpu_setup_region(unsigned int number, phys_addr_t start,
>   	size_data = ((size_order - 1) << MPU_RSR_SZ) | 1 << MPU_RSR_EN;
>   	size_data |= subregions << MPU_RSR_SD;
>   
> +	if (need_flush)
> +		flush_cache_all();
> +
>   	dsb(); /* Ensure all previous data accesses occur with old mappings */
>   	rgnr_write(number);
>   	isb();
> @@ -416,7 +434,28 @@ void __init mpu_setup(void)
>   	/* Background */
>   	err |= mpu_setup_region(region++, 0, 32,
>   				MPU_ACR_XN | MPU_RGN_STRONGLY_ORDERED | MPU_AP_PL1RW_PL0NA,
> -				0);
> +				0, false);
> +
> +#ifdef CONFIG_XIP_KERNEL
> +	/* ROM */
> +	for (i = 0; i < ARRAY_SIZE(xip); i++) {
> +		/*
> +                 * In case we overwrite RAM region we set earlier in
> +                 * head-nommu.S (which is cachable) all subsequent
> +                 * data access till we setup RAM bellow would be done
> +                 * with BG region (which is uncachable), thus we need
> +                 * to clean and invalidate cache.
> +		 */
> +		bool need_flush = region == MPU_RAM_REGION;
> +
> +		if (!xip[i].size)
> +			continue;
> +
> +		err |= mpu_setup_region(region++, xip[i].base, ilog2(xip[i].size),
> +					MPU_AP_PL1RO_PL0NA | MPU_RGN_NORMAL,
> +					xip[i].subreg, need_flush);
> +	}
> +#endif
>   
>   	/* RAM */
>   	for (i = 0; i < ARRAY_SIZE(mem); i++) {
> @@ -425,14 +464,14 @@ void __init mpu_setup(void)
>   
>   		err |= mpu_setup_region(region++, mem[i].base, ilog2(mem[i].size),
>   					MPU_AP_PL1RW_PL0RW | MPU_RGN_NORMAL,
> -					mem[i].subreg);
> +					mem[i].subreg, false);
>   	}
>   
>   	/* Vectors */
>   #ifndef CONFIG_CPU_V7M
>   	err |= mpu_setup_region(region++, vectors_base, ilog2(2 * PAGE_SIZE),
>   				MPU_AP_PL1RW_PL0NA | MPU_RGN_NORMAL,
> -				0);
> +				0, false);
>   #endif
>   	if (err) {
>   		panic("MPU region initialization failure! %d", err);
> 



More information about the linux-arm-kernel mailing list