diff mbox

[v5,5/6] arm64: mm: Implement 4 levels of translation tables

Message ID 000501cf64e5$d92ae870$8b80b950$@samsung.com
State New, archived
Headers show

Commit Message

??? May 1, 2014, 2:34 a.m. UTC
This patch implements 4 levels of translation tables since 3 levels
of page tables with 4KB pages cannot support 40-bit physical address
space described in [1] due to the following issue.

It is a restriction that kernel logical memory map with 4KB + 3 levels
(0xffffffc000000000-0xffffffffffffffff) cannot cover RAM region from
544GB to 1024GB in [1]. Specifically, ARM64 kernel fails to create
mapping for this region in map_mem function since __phys_to_virt for
this region reaches to address overflow.

If SoC design follows the document, [1], over 32GB RAM would be placed
from 544GB. Even 64GB system is supposed to use the region from 544GB
to 576GB for only 32GB RAM. Naturally, it would reach to enable 4 levels
of page tables to avoid hacking __virt_to_phys and __phys_to_virt.

However, it is recommended 4 levels of page table should be only enabled
if memory map is too sparse or there is about 512GB RAM.

References
----------
[1]: Principles of ARM Memory Maps, White Paper, Issue C

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Steve Capper <steve.capper@linaro.org>
Signed-off-by: Jungseok Lee <jays.lee@samsung.com>
Reviewed-by: Sungjinn Chung <sungjinn.chung@samsung.com>
---
 arch/arm64/Kconfig                     |    8 ++++++
 arch/arm64/include/asm/memblock.h      |    6 +++++
 arch/arm64/include/asm/page.h          |    4 ++-
 arch/arm64/include/asm/pgalloc.h       |   20 ++++++++++++++
 arch/arm64/include/asm/pgtable-hwdef.h |    6 +++--
 arch/arm64/include/asm/pgtable.h       |   45 +++++++++++++++++++++++++++++++
 arch/arm64/include/asm/tlb.h           |    9 +++++++
 arch/arm64/kernel/head.S               |   46 +++++++++++++++++++++++++-------
 arch/arm64/kernel/traps.c              |    5 ++++
 arch/arm64/mm/fault.c                  |    1 +
 arch/arm64/mm/mmu.c                    |   16 ++++++++---
 11 files changed, 150 insertions(+), 16 deletions(-)

Comments

Christoffer Dall May 6, 2014, 10:49 a.m. UTC | #1
On Thu, May 01, 2014 at 11:34:16AM +0900, Jungseok Lee wrote:
> This patch implements 4 levels of translation tables since 3 levels
> of page tables with 4KB pages cannot support 40-bit physical address
> space described in [1] due to the following issue.
> 
> It is a restriction that kernel logical memory map with 4KB + 3 levels
> (0xffffffc000000000-0xffffffffffffffff) cannot cover RAM region from
> 544GB to 1024GB in [1]. Specifically, ARM64 kernel fails to create
> mapping for this region in map_mem function since __phys_to_virt for
> this region reaches to address overflow.
> 
> If SoC design follows the document, [1], over 32GB RAM would be placed
> from 544GB. Even 64GB system is supposed to use the region from 544GB
> to 576GB for only 32GB RAM. Naturally, it would reach to enable 4 levels
> of page tables to avoid hacking __virt_to_phys and __phys_to_virt.
> 
> However, it is recommended 4 levels of page table should be only enabled
> if memory map is too sparse or there is about 512GB RAM.

Who recommends this then?  This paragraph just confuses me.

> 
> References
> ----------
> [1]: Principles of ARM Memory Maps, White Paper, Issue C
> 
> Cc: Catalin Marinas <catalin.marinas@arm.com>
> Cc: Steve Capper <steve.capper@linaro.org>
> Signed-off-by: Jungseok Lee <jays.lee@samsung.com>
> Reviewed-by: Sungjinn Chung <sungjinn.chung@samsung.com>
> ---
>  arch/arm64/Kconfig                     |    8 ++++++
>  arch/arm64/include/asm/memblock.h      |    6 +++++
>  arch/arm64/include/asm/page.h          |    4 ++-
>  arch/arm64/include/asm/pgalloc.h       |   20 ++++++++++++++
>  arch/arm64/include/asm/pgtable-hwdef.h |    6 +++--
>  arch/arm64/include/asm/pgtable.h       |   45 +++++++++++++++++++++++++++++++
>  arch/arm64/include/asm/tlb.h           |    9 +++++++
>  arch/arm64/kernel/head.S               |   46 +++++++++++++++++++++++++-------
>  arch/arm64/kernel/traps.c              |    5 ++++
>  arch/arm64/mm/fault.c                  |    1 +
>  arch/arm64/mm/mmu.c                    |   16 ++++++++---
>  11 files changed, 150 insertions(+), 16 deletions(-)
> 
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index b438540..3e49671 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -182,12 +182,17 @@ config ARM64_VA_BITS_42
>  	bool "42-bit"
>  	depends on ARM64_64K_PAGES
>  
> +config ARM64_VA_BITS_48
> +	bool "48-bit"
> +	depends on ARM64_4K_PAGES
> +
>  endchoice
>  
>  config ARM64_VA_BITS
>  	int
>  	default 39 if ARM64_VA_BITS_39
>  	default 42 if ARM64_VA_BITS_42
> +	default 48 if ARM64_VA_BITS_48
>  
>  config ARM64_2_LEVELS
>  	def_bool y if ARM64_64K_PAGES && ARM64_VA_BITS_42
> @@ -195,6 +200,9 @@ config ARM64_2_LEVELS
>  config ARM64_3_LEVELS
>  	def_bool y if ARM64_4K_PAGES && ARM64_VA_BITS_39
>  
> +config ARM64_4_LEVELS
> +	def_bool y if ARM64_4K_PAGES && ARM64_VA_BITS_48
> +
>  config CPU_BIG_ENDIAN
>         bool "Build big-endian kernel"
>         help
> diff --git a/arch/arm64/include/asm/memblock.h b/arch/arm64/include/asm/memblock.h
> index 6afeed2..e4ac8bf 100644
> --- a/arch/arm64/include/asm/memblock.h
> +++ b/arch/arm64/include/asm/memblock.h
> @@ -16,6 +16,12 @@
>  #ifndef __ASM_MEMBLOCK_H
>  #define __ASM_MEMBLOCK_H
>  
> +#ifndef CONFIG_ARM64_4_LEVELS
> +#define MEMBLOCK_INITIAL_LIMIT	PGDIR_SIZE
> +#else
> +#define MEMBLOCK_INITIAL_LIMIT	PUD_SIZE
> +#endif
> +
>  extern void arm64_memblock_init(void);
>  
>  #endif
> diff --git a/arch/arm64/include/asm/page.h b/arch/arm64/include/asm/page.h
> index 268e53d..83b5289 100644
> --- a/arch/arm64/include/asm/page.h
> +++ b/arch/arm64/include/asm/page.h
> @@ -35,8 +35,10 @@
>  
>  #ifdef CONFIG_ARM64_2_LEVELS
>  #include <asm/pgtable-2level-types.h>
> -#else
> +#elif defined(CONFIG_ARM64_3_LEVELS)
>  #include <asm/pgtable-3level-types.h>
> +#else
> +#include <asm/pgtable-4level-types.h>
>  #endif
>  
>  extern void __cpu_clear_user_page(void *p, unsigned long user);
> diff --git a/arch/arm64/include/asm/pgalloc.h b/arch/arm64/include/asm/pgalloc.h
> index 4829837..8d745fa 100644
> --- a/arch/arm64/include/asm/pgalloc.h
> +++ b/arch/arm64/include/asm/pgalloc.h
> @@ -26,6 +26,26 @@
>  
>  #define check_pgt_cache()		do { } while (0)
>  
> +#ifdef CONFIG_ARM64_4_LEVELS
> +
> +static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
> +{
> +	return (pud_t *)get_zeroed_page(GFP_KERNEL | __GFP_REPEAT);
> +}
> +
> +static inline void pud_free(struct mm_struct *mm, pud_t *pud)
> +{
> +	BUG_ON((unsigned long)pud & (PAGE_SIZE-1));
> +	free_page((unsigned long)pud);
> +}
> +
> +static inline void pgd_populate(struct mm_struct *mm, pgd_t *pgd, pud_t *pud)
> +{
> +	set_pgd(pgd, __pgd(__pa(pud) | PUD_TYPE_TABLE));
> +}
> +
> +#endif  /* CONFIG_ARM64_4_LEVELS */
> +
>  #ifndef CONFIG_ARM64_2_LEVELS
>  
>  static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
> diff --git a/arch/arm64/include/asm/pgtable-hwdef.h b/arch/arm64/include/asm/pgtable-hwdef.h
> index 9cd86c6..ba30053 100644
> --- a/arch/arm64/include/asm/pgtable-hwdef.h
> +++ b/arch/arm64/include/asm/pgtable-hwdef.h
> @@ -18,8 +18,10 @@
>  
>  #ifdef CONFIG_ARM64_2_LEVELS
>  #include <asm/pgtable-2level-hwdef.h>
> -#else
> +#elif defined(CONFIG_ARM64_3_LEVELS)
>  #include <asm/pgtable-3level-hwdef.h>
> +#else
> +#include <asm/pgtable-4level-hwdef.h>
>  #endif
>  
>  /*
> @@ -27,7 +29,7 @@
>   *
>   * Level 1 descriptor (PUD).
>   */
> -
> +#define PUD_TYPE_TABLE		(_AT(pudval_t, 3) << 0)
>  #define PUD_TABLE_BIT		(_AT(pgdval_t, 1) << 1)
>  
>  /*
> diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
> index a64ce5e..b27cc426 100644
> --- a/arch/arm64/include/asm/pgtable.h
> +++ b/arch/arm64/include/asm/pgtable.h
> @@ -35,7 +35,11 @@
>   * VMALLOC and SPARSEMEM_VMEMMAP ranges.
>   */
>  #define VMALLOC_START		(UL(0xffffffffffffffff) << VA_BITS)
> +#ifndef CONFIG_ARM64_4_LEVELS
>  #define VMALLOC_END		(PAGE_OFFSET - UL(0x400000000) - SZ_64K)
> +#else
> +#define VMALLOC_END		(PAGE_OFFSET - UL(0x40000000000) - SZ_64K)
> +#endif
>  
>  #define vmemmap			((struct page *)(VMALLOC_END + SZ_64K))
>  
> @@ -44,12 +48,16 @@
>  #ifndef __ASSEMBLY__
>  extern void __pte_error(const char *file, int line, unsigned long val);
>  extern void __pmd_error(const char *file, int line, unsigned long val);
> +extern void __pud_error(const char *file, int line, unsigned long val);
>  extern void __pgd_error(const char *file, int line, unsigned long val);
>  
>  #define pte_ERROR(pte)		__pte_error(__FILE__, __LINE__, pte_val(pte))
>  #ifndef CONFIG_ARM64_2_LEVELS
>  #define pmd_ERROR(pmd)		__pmd_error(__FILE__, __LINE__, pmd_val(pmd))
>  #endif
> +#ifdef CONFIG_ARM64_4_LEVELS
> +#define pud_ERROR(pud)		__pud_error(__FILE__, __LINE__, pud_val(pud))
> +#endif
>  #define pgd_ERROR(pgd)		__pgd_error(__FILE__, __LINE__, pgd_val(pgd))
>  
>  /*
> @@ -344,6 +352,30 @@ static inline pmd_t *pud_page_vaddr(pud_t pud)
>  
>  #endif	/* CONFIG_ARM64_2_LEVELS */
>  
> +#ifdef CONFIG_ARM64_4_LEVELS
> +
> +#define pgd_none(pgd)		(!pgd_val(pgd))
> +#define pgd_bad(pgd)		(!(pgd_val(pgd) & 2))
> +#define pgd_present(pgd)	(pgd_val(pgd))
> +
> +static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
> +{
> +	*pgdp = pgd;
> +	dsb();
> +}
> +
> +static inline void pgd_clear(pgd_t *pgdp)
> +{
> +	set_pgd(pgdp, __pgd(0));
> +}
> +
> +static inline pud_t *pgd_page_vaddr(pgd_t pgd)
> +{
> +	return __va(pgd_val(pgd) & PHYS_MASK & (s32)PAGE_MASK);
> +}
> +
> +#endif  /* CONFIG_ARM64_4_LEVELS */
> +
>  /* to find an entry in a page-table-directory */
>  #define pgd_index(addr)		(((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
>  
> @@ -352,6 +384,14 @@ static inline pmd_t *pud_page_vaddr(pud_t pud)
>  /* to find an entry in a kernel page-table-directory */
>  #define pgd_offset_k(addr)	pgd_offset(&init_mm, addr)
>  
> +#ifdef CONFIG_ARM64_4_LEVELS
> +#define pud_index(addr)		(((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1))
> +static inline pud_t *pud_offset(pgd_t *pgd, unsigned long addr)
> +{
> +	return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(addr);
> +}
> +#endif
> +
>  /* Find an entry in the second-level page table.. */
>  #ifndef CONFIG_ARM64_2_LEVELS
>  #define pmd_index(addr)		(((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))
> @@ -380,8 +420,13 @@ static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
>  extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
>  extern pgd_t idmap_pg_dir[PTRS_PER_PGD];
>  
> +#ifdef CONFIG_ARM64_4_LEVELS
> +#define SWAPPER_DIR_SIZE	(4 * PAGE_SIZE)
> +#define IDMAP_DIR_SIZE		(3 * PAGE_SIZE)
> +#else
>  #define SWAPPER_DIR_SIZE	(3 * PAGE_SIZE)
>  #define IDMAP_DIR_SIZE		(2 * PAGE_SIZE)
> +#endif
>  
>  /*
>   * Encode and decode a swap entry:
> diff --git a/arch/arm64/include/asm/tlb.h b/arch/arm64/include/asm/tlb.h
> index bc19101..086112b 100644
> --- a/arch/arm64/include/asm/tlb.h
> +++ b/arch/arm64/include/asm/tlb.h
> @@ -100,6 +100,15 @@ static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmdp,
>  }
>  #endif
>  
> +#ifdef CONFIG_ARM64_4_LEVELS
> +static inline void __pud_free_tlb(struct mmu_gather *tlb, pmd_t *pudp,
> +				  unsigned long addr)
> +{
> +	tlb_add_flush(tlb, addr);
> +	tlb_remove_page(tlb, virt_to_page(pudp));
> +}
> +#endif
> +
>  static inline void __tlb_remove_pmd_tlb_entry(struct mmu_gather *tlb, pmd_t *pmdp,
>  						unsigned long address)
>  {
> diff --git a/arch/arm64/kernel/head.S b/arch/arm64/kernel/head.S
> index 0fd5650..03ec424 100644
> --- a/arch/arm64/kernel/head.S
> +++ b/arch/arm64/kernel/head.S
> @@ -37,8 +37,9 @@
>  
>  /*
>   * swapper_pg_dir is the virtual address of the initial page table. We place
> - * the page tables 3 * PAGE_SIZE below KERNEL_RAM_VADDR. The idmap_pg_dir has
> - * 2 pages and is placed below swapper_pg_dir.
> + * the page tables 3 * PAGE_SIZE (2 or 3 levels) or 4 * PAGE_SIZE (4 levels)
> + * below KERNEL_RAM_VADDR. The idmap_pg_dir has 2 pages (2 or 3 levels) or
> + * 3 pages (4 levels) and is placed below swapper_pg_dir.
>   */
>  #define KERNEL_RAM_VADDR	(PAGE_OFFSET + TEXT_OFFSET)
>  
> @@ -46,8 +47,13 @@
>  #error KERNEL_RAM_VADDR must start at 0xXXX80000
>  #endif
>  
> +#ifdef CONFIG_ARM64_4_LEVELS
> +#define SWAPPER_DIR_SIZE	(4 * PAGE_SIZE)
> +#define IDMAP_DIR_SIZE		(3 * PAGE_SIZE)
> +#else
>  #define SWAPPER_DIR_SIZE	(3 * PAGE_SIZE)
>  #define IDMAP_DIR_SIZE		(2 * PAGE_SIZE)
> +#endif
>  
>  	.globl	swapper_pg_dir
>  	.equ	swapper_pg_dir, KERNEL_RAM_VADDR - SWAPPER_DIR_SIZE
> @@ -370,16 +376,38 @@ ENDPROC(__calc_phys_offset)
>  	.quad	PAGE_OFFSET
>  
>  /*
> + * Macro to populate the PUD for the corresponding block entry in the next
> + * level (tbl) for the given virtual address in case of 4levels.
> + */

With the relatively high number of parameters to this macro, I feel it
would be helpful to state in the comment that this actually modifies \tbl
and returns a value in \pud.

> +	.macro	create_pud_entry, pgd, tbl, virt, pud, tmp1, tmp2
> +#ifdef CONFIG_ARM64_4_LEVELS
> +	add	\tbl, \tbl, #PAGE_SIZE		// bump tbl 1 page up.
> +						// to make room for pud
> +	add	\pud, \pgd, #PAGE_SIZE		// pgd points to pud which
> +						// follows pgd
> +	lsr	\tmp1, \virt, #PUD_SHIFT
> +	and	\tmp1, \tmp1, #PTRS_PER_PUD - 1	// PUD index
> +	orr	\tmp2, \tbl, #3			// PUD entry table type
> +	str	\tmp2, [\pud, \tmp1, lsl #3]
> +#else
> +	mov	\pud, \tbl
> +#endif
> +	.endm
> +
> +/*
>   * Macro to populate the PGD for the corresponding block entry in the next
>   * level (tbl) for the given virtual address.

Then this should be changed to be: "populate the PGD (and possibly PUD)"

>   *
> - * Preserves:	pgd, tbl, virt
> - * Corrupts:	tmp1, tmp2
> + * Preserves:	pgd, virt
> + * Corrupts:	tmp1, tmp2, tmp3
> + * Returns:	tbl -> page where block mappings can be placed
> + *	(changed to make room for pud with 4levels, preserved otherwise)
>   */
> -	.macro	create_pgd_entry, pgd, tbl, virt, tmp1, tmp2
> +	.macro	create_pgd_entry, pgd, tbl, virt, tmp1, tmp2, tmp3
> +	create_pud_entry \pgd, \tbl, \virt, \tmp3, \tmp1, \tmp2
>  	lsr	\tmp1, \virt, #PGDIR_SHIFT
>  	and	\tmp1, \tmp1, #PTRS_PER_PGD - 1	// PGD index
> -	orr	\tmp2, \tbl, #3			// PGD entry table type
> +	orr	\tmp2, \tmp3, #3		// PGD entry table type
>  	str	\tmp2, [\pgd, \tmp1, lsl #3]
>  	.endm
>  
> @@ -444,7 +472,7 @@ __create_page_tables:
>  	add	x0, x25, #PAGE_SIZE		// section table address
>  	ldr	x3, =KERNEL_START
>  	add	x3, x3, x28			// __pa(KERNEL_START)
> -	create_pgd_entry x25, x0, x3, x5, x6
> +	create_pgd_entry x25, x0, x3, x1, x5, x6
>  	ldr	x6, =KERNEL_END
>  	mov	x5, x3				// __pa(KERNEL_START)
>  	add	x6, x6, x28			// __pa(KERNEL_END)
> @@ -455,7 +483,7 @@ __create_page_tables:
>  	 */
>  	add	x0, x26, #PAGE_SIZE		// section table address
>  	mov	x5, #PAGE_OFFSET
> -	create_pgd_entry x26, x0, x5, x3, x6
> +	create_pgd_entry x26, x0, x5, x1, x3, x6
>  	ldr	x6, =KERNEL_END
>  	mov	x3, x24				// phys offset
>  	create_block_map x0, x7, x3, x5, x6
> @@ -481,7 +509,7 @@ __create_page_tables:
>  	 */
>  	ldr	x5, =FIXADDR_TOP		// Fixed mapping virtual address
>  	add	x0, x26, #2 * PAGE_SIZE		// section table address
> -	create_pgd_entry x26, x0, x5, x6, x7
> +	create_pgd_entry x26, x0, x5, x1, x6, x7
>  
>  	/*
>  	 * Since the page tables have been populated with non-cacheable
> diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
> index 268ce96..237757d 100644
> --- a/arch/arm64/kernel/traps.c
> +++ b/arch/arm64/kernel/traps.c
> @@ -336,6 +336,11 @@ void __pmd_error(const char *file, int line, unsigned long val)
>  	pr_crit("%s:%d: bad pmd %016lx.\n", file, line, val);
>  }
>  
> +void __pud_error(const char *file, int line, unsigned long val)
> +{
> +	pr_crit("%s:%d: bad pud %016lx.\n", file, line, val);
> +}
> +
>  void __pgd_error(const char *file, int line, unsigned long val)
>  {
>  	pr_crit("%s:%d: bad pgd %016lx.\n", file, line, val);
> diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
> index c23751b..ed4a343 100644
> --- a/arch/arm64/mm/fault.c
> +++ b/arch/arm64/mm/fault.c
> @@ -61,6 +61,7 @@ void show_pte(struct mm_struct *mm, unsigned long addr)
>  			break;
>  
>  		pud = pud_offset(pgd, addr);
> +		printk(", *pud=%016llx", pud_val(*pud));
>  		if (pud_none(*pud) || pud_bad(*pud))
>  			break;
>  
> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> index 6b7e895..4d29332 100644
> --- a/arch/arm64/mm/mmu.c
> +++ b/arch/arm64/mm/mmu.c
> @@ -32,6 +32,7 @@
>  #include <asm/setup.h>
>  #include <asm/sizes.h>
>  #include <asm/tlb.h>
> +#include <asm/memblock.h>
>  #include <asm/mmu_context.h>
>  
>  #include "mm.h"
> @@ -222,9 +223,15 @@ static void __init alloc_init_pmd(pud_t *pud, unsigned long addr,
>  static void __init alloc_init_pud(pgd_t *pgd, unsigned long addr,
>  				  unsigned long end, unsigned long phys)
>  {
> -	pud_t *pud = pud_offset(pgd, addr);
> +	pud_t *pud;
>  	unsigned long next;
>  
> +	if (pgd_none(*pgd) || pgd_bad(*pgd)) {
> +		pud = early_alloc(PTRS_PER_PUD * sizeof(pud_t));
> +		pgd_populate(&init_mm, pgd, pud);
> +	}
> +
> +	pud = pud_offset(pgd, addr);
>  	do {
>  		next = pud_addr_end(addr, end);
>  		alloc_init_pmd(pud, addr, next, phys);
> @@ -271,10 +278,11 @@ static void __init map_mem(void)
>  	 * memory addressable from the initial direct kernel mapping.
>  	 *
>  	 * The initial direct kernel mapping, located at swapper_pg_dir,
> -	 * gives us PGDIR_SIZE memory starting from PHYS_OFFSET (which must be
> -	 * aligned to 2MB as per Documentation/arm64/booting.txt).
> +	 * gives us PGDIR_SIZE (2 and 3 levels) or PUD_SIZE (4 levels) memory
> +	 * starting from PHYS_OFFSET (which must be aligned to 2MB as per
> +	 * Documentation/arm64/booting.txt).
>  	 */
> -	limit = PHYS_OFFSET + PGDIR_SIZE;
> +	limit = PHYS_OFFSET + MEMBLOCK_INITIAL_LIMIT;
>  	memblock_set_current_limit(limit);
>  
>  	/* map all the memory banks */
> -- 
> 1.7.10.4
> 
> 

Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
--
To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Steve Capper May 6, 2014, 12:01 p.m. UTC | #2
On Thu, May 01, 2014 at 11:34:16AM +0900, Jungseok Lee wrote:
> This patch implements 4 levels of translation tables since 3 levels
> of page tables with 4KB pages cannot support 40-bit physical address
> space described in [1] due to the following issue.
> 
> It is a restriction that kernel logical memory map with 4KB + 3 levels
> (0xffffffc000000000-0xffffffffffffffff) cannot cover RAM region from
> 544GB to 1024GB in [1]. Specifically, ARM64 kernel fails to create
> mapping for this region in map_mem function since __phys_to_virt for
> this region reaches to address overflow.
> 
> If SoC design follows the document, [1], over 32GB RAM would be placed
> from 544GB. Even 64GB system is supposed to use the region from 544GB
> to 576GB for only 32GB RAM. Naturally, it would reach to enable 4 levels
> of page tables to avoid hacking __virt_to_phys and __phys_to_virt.
> 
> However, it is recommended 4 levels of page table should be only enabled
> if memory map is too sparse or there is about 512GB RAM.
> 

Hi Jungseok,
One comment below:

[ ... ]

> diff --git a/arch/arm64/include/asm/tlb.h b/arch/arm64/include/asm/tlb.h
> index bc19101..086112b 100644
> --- a/arch/arm64/include/asm/tlb.h
> +++ b/arch/arm64/include/asm/tlb.h
> @@ -100,6 +100,15 @@ static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmdp,
>  }
>  #endif
>  
> +#ifdef CONFIG_ARM64_4_LEVELS
> +static inline void __pud_free_tlb(struct mmu_gather *tlb, pmd_t *pudp,
> +				  unsigned long addr)

The second parameter needs to be a pointer to pud_t ?
(this fires up a warning with STRICT_MM_TYPECHECKS).

With that and Christoffer's feedback about expanding the comments on
create_pud_entry addressed:

Reviewed-by: Steve Capper <steve.capper@linaro.org>

Cheers,
??? May 7, 2014, 4:22 a.m. UTC | #3
On Tuesday, May 06, 2014 7:49 PM, Christoffer Dall wrote:
> On Thu, May 01, 2014 at 11:34:16AM +0900, Jungseok Lee wrote:
> > This patch implements 4 levels of translation tables since 3 levels of
> > page tables with 4KB pages cannot support 40-bit physical address
> > space described in [1] due to the following issue.
> >
> > It is a restriction that kernel logical memory map with 4KB + 3 levels
> > (0xffffffc000000000-0xffffffffffffffff) cannot cover RAM region from
> > 544GB to 1024GB in [1]. Specifically, ARM64 kernel fails to create
> > mapping for this region in map_mem function since __phys_to_virt for
> > this region reaches to address overflow.
> >
> > If SoC design follows the document, [1], over 32GB RAM would be placed
> > from 544GB. Even 64GB system is supposed to use the region from 544GB
> > to 576GB for only 32GB RAM. Naturally, it would reach to enable 4
> > levels of page tables to avoid hacking __virt_to_phys and __phys_to_virt.
> >
> > However, it is recommended 4 levels of page table should be only
> > enabled if memory map is too sparse or there is about 512GB RAM.
> 
> Who recommends this then?  This paragraph just confuses me.

It is a paraphrase of Catalin's comment:
"I agree, we should only enable 4-levels of page table if we have close
to 512GB of RAM or the range is too sparse but I would actually push
back on the hardware guys to keep it tighter."

The above message comes from the discussion on 4 levels.
http://www.spinics.net/linux/lists/arm-kernel/msg319055.html

> >
> > References
> > ----------
> > [1]: Principles of ARM Memory Maps, White Paper, Issue C
> >
> > Cc: Catalin Marinas <catalin.marinas@arm.com>
> > Cc: Steve Capper <steve.capper@linaro.org>
> > Signed-off-by: Jungseok Lee <jays.lee@samsung.com>
> > Reviewed-by: Sungjinn Chung <sungjinn.chung@samsung.com>
> > ---
> >  arch/arm64/Kconfig                     |    8 ++++++
> >  arch/arm64/include/asm/memblock.h      |    6 +++++
> >  arch/arm64/include/asm/page.h          |    4 ++-
> >  arch/arm64/include/asm/pgalloc.h       |   20 ++++++++++++++
> >  arch/arm64/include/asm/pgtable-hwdef.h |    6 +++--
> >  arch/arm64/include/asm/pgtable.h       |   45 +++++++++++++++++++++++++++++++
> >  arch/arm64/include/asm/tlb.h           |    9 +++++++
> >  arch/arm64/kernel/head.S               |   46 +++++++++++++++++++++++++-------
> >  arch/arm64/kernel/traps.c              |    5 ++++
> >  arch/arm64/mm/fault.c                  |    1 +
> >  arch/arm64/mm/mmu.c                    |   16 ++++++++---
> >  11 files changed, 150 insertions(+), 16 deletions(-)
> >

[ ... ]

> > diff --git a/arch/arm64/kernel/head.S b/arch/arm64/kernel/head.S index
> > 0fd5650..03ec424 100644
> > --- a/arch/arm64/kernel/head.S
> > +++ b/arch/arm64/kernel/head.S
> > @@ -37,8 +37,9 @@
> >
> >  /*
> >   * swapper_pg_dir is the virtual address of the initial page table.
> > We place
> > - * the page tables 3 * PAGE_SIZE below KERNEL_RAM_VADDR. The
> > idmap_pg_dir has
> > - * 2 pages and is placed below swapper_pg_dir.
> > + * the page tables 3 * PAGE_SIZE (2 or 3 levels) or 4 * PAGE_SIZE (4
> > + levels)
> > + * below KERNEL_RAM_VADDR. The idmap_pg_dir has 2 pages (2 or 3
> > + levels) or
> > + * 3 pages (4 levels) and is placed below swapper_pg_dir.
> >   */
> >  #define KERNEL_RAM_VADDR	(PAGE_OFFSET + TEXT_OFFSET)
> >
> > @@ -46,8 +47,13 @@
> >  #error KERNEL_RAM_VADDR must start at 0xXXX80000  #endif
> >
> > +#ifdef CONFIG_ARM64_4_LEVELS
> > +#define SWAPPER_DIR_SIZE	(4 * PAGE_SIZE)
> > +#define IDMAP_DIR_SIZE		(3 * PAGE_SIZE)
> > +#else
> >  #define SWAPPER_DIR_SIZE	(3 * PAGE_SIZE)
> >  #define IDMAP_DIR_SIZE		(2 * PAGE_SIZE)
> > +#endif
> >
> >  	.globl	swapper_pg_dir
> >  	.equ	swapper_pg_dir, KERNEL_RAM_VADDR - SWAPPER_DIR_SIZE
> > @@ -370,16 +376,38 @@ ENDPROC(__calc_phys_offset)
> >  	.quad	PAGE_OFFSET
> >
> >  /*
> > + * Macro to populate the PUD for the corresponding block entry in the
> > + next
> > + * level (tbl) for the given virtual address in case of 4levels.
> > + */
> 
> With the relatively high number of parameters to this macro, I feel it would be helpful to state in
> the comment that this actually modifies \tbl and returns a value in \pud.

Okay, I will add it.

> > +	.macro	create_pud_entry, pgd, tbl, virt, pud, tmp1, tmp2
> > +#ifdef CONFIG_ARM64_4_LEVELS
> > +	add	\tbl, \tbl, #PAGE_SIZE		// bump tbl 1 page up.
> > +						// to make room for pud
> > +	add	\pud, \pgd, #PAGE_SIZE		// pgd points to pud which
> > +						// follows pgd
> > +	lsr	\tmp1, \virt, #PUD_SHIFT
> > +	and	\tmp1, \tmp1, #PTRS_PER_PUD - 1	// PUD index
> > +	orr	\tmp2, \tbl, #3			// PUD entry table type
> > +	str	\tmp2, [\pud, \tmp1, lsl #3]
> > +#else
> > +	mov	\pud, \tbl
> > +#endif
> > +	.endm
> > +
> > +/*
> >   * Macro to populate the PGD for the corresponding block entry in the next
> >   * level (tbl) for the given virtual address.
> 
> Then this should be changed to be: "populate the PGD (and possibly PUD)"

Okay. I will revise it.

> >   *
> > - * Preserves:	pgd, tbl, virt
> > - * Corrupts:	tmp1, tmp2
> > + * Preserves:	pgd, virt
> > + * Corrupts:	tmp1, tmp2, tmp3
> > + * Returns:	tbl -> page where block mappings can be placed
> > + *	(changed to make room for pud with 4levels, preserved otherwise)
> >   */
> > -	.macro	create_pgd_entry, pgd, tbl, virt, tmp1, tmp2
> > +	.macro	create_pgd_entry, pgd, tbl, virt, tmp1, tmp2, tmp3
> > +	create_pud_entry \pgd, \tbl, \virt, \tmp3, \tmp1, \tmp2
> >  	lsr	\tmp1, \virt, #PGDIR_SHIFT
> >  	and	\tmp1, \tmp1, #PTRS_PER_PGD - 1	// PGD index
> > -	orr	\tmp2, \tbl, #3			// PGD entry table type
> > +	orr	\tmp2, \tmp3, #3		// PGD entry table type
> >  	str	\tmp2, [\pgd, \tmp1, lsl #3]
> >  	.endm
> >
> > @@ -444,7 +472,7 @@ __create_page_tables:
> >  	add	x0, x25, #PAGE_SIZE		// section table address
> >  	ldr	x3, =KERNEL_START
> >  	add	x3, x3, x28			// __pa(KERNEL_START)
> > -	create_pgd_entry x25, x0, x3, x5, x6
> > +	create_pgd_entry x25, x0, x3, x1, x5, x6
> >  	ldr	x6, =KERNEL_END
> >  	mov	x5, x3				// __pa(KERNEL_START)
> >  	add	x6, x6, x28			// __pa(KERNEL_END)
> > @@ -455,7 +483,7 @@ __create_page_tables:
> >  	 */
> >  	add	x0, x26, #PAGE_SIZE		// section table address
> >  	mov	x5, #PAGE_OFFSET
> > -	create_pgd_entry x26, x0, x5, x3, x6
> > +	create_pgd_entry x26, x0, x5, x1, x3, x6
> >  	ldr	x6, =KERNEL_END
> >  	mov	x3, x24				// phys offset
> >  	create_block_map x0, x7, x3, x5, x6
> > @@ -481,7 +509,7 @@ __create_page_tables:
> >  	 */
> >  	ldr	x5, =FIXADDR_TOP		// Fixed mapping virtual address
> >  	add	x0, x26, #2 * PAGE_SIZE		// section table address
> > -	create_pgd_entry x26, x0, x5, x6, x7
> > +	create_pgd_entry x26, x0, x5, x1, x6, x7
> >
> >  	/*
> >  	 * Since the page tables have been populated with non-cacheable diff
> > --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c index
> > 268ce96..237757d 100644
> > --- a/arch/arm64/kernel/traps.c
> > +++ b/arch/arm64/kernel/traps.c
> > @@ -336,6 +336,11 @@ void __pmd_error(const char *file, int line, unsigned long val)
> >  	pr_crit("%s:%d: bad pmd %016lx.\n", file, line, val);  }
> >
> > +void __pud_error(const char *file, int line, unsigned long val) {
> > +	pr_crit("%s:%d: bad pud %016lx.\n", file, line, val); }
> > +
> >  void __pgd_error(const char *file, int line, unsigned long val)  {
> >  	pr_crit("%s:%d: bad pgd %016lx.\n", file, line, val); diff --git
> > a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c index c23751b..ed4a343
> > 100644
> > --- a/arch/arm64/mm/fault.c
> > +++ b/arch/arm64/mm/fault.c
> > @@ -61,6 +61,7 @@ void show_pte(struct mm_struct *mm, unsigned long addr)
> >  			break;
> >
> >  		pud = pud_offset(pgd, addr);
> > +		printk(", *pud=%016llx", pud_val(*pud));
> >  		if (pud_none(*pud) || pud_bad(*pud))
> >  			break;
> >
> > diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c index
> > 6b7e895..4d29332 100644
> > --- a/arch/arm64/mm/mmu.c
> > +++ b/arch/arm64/mm/mmu.c
> > @@ -32,6 +32,7 @@
> >  #include <asm/setup.h>
> >  #include <asm/sizes.h>
> >  #include <asm/tlb.h>
> > +#include <asm/memblock.h>
> >  #include <asm/mmu_context.h>
> >
> >  #include "mm.h"
> > @@ -222,9 +223,15 @@ static void __init alloc_init_pmd(pud_t *pud,
> > unsigned long addr,  static void __init alloc_init_pud(pgd_t *pgd, unsigned long addr,
> >  				  unsigned long end, unsigned long phys)  {
> > -	pud_t *pud = pud_offset(pgd, addr);
> > +	pud_t *pud;
> >  	unsigned long next;
> >
> > +	if (pgd_none(*pgd) || pgd_bad(*pgd)) {
> > +		pud = early_alloc(PTRS_PER_PUD * sizeof(pud_t));
> > +		pgd_populate(&init_mm, pgd, pud);
> > +	}
> > +
> > +	pud = pud_offset(pgd, addr);
> >  	do {
> >  		next = pud_addr_end(addr, end);
> >  		alloc_init_pmd(pud, addr, next, phys); @@ -271,10 +278,11 @@ static
> > void __init map_mem(void)
> >  	 * memory addressable from the initial direct kernel mapping.
> >  	 *
> >  	 * The initial direct kernel mapping, located at swapper_pg_dir,
> > -	 * gives us PGDIR_SIZE memory starting from PHYS_OFFSET (which must be
> > -	 * aligned to 2MB as per Documentation/arm64/booting.txt).
> > +	 * gives us PGDIR_SIZE (2 and 3 levels) or PUD_SIZE (4 levels) memory
> > +	 * starting from PHYS_OFFSET (which must be aligned to 2MB as per
> > +	 * Documentation/arm64/booting.txt).
> >  	 */
> > -	limit = PHYS_OFFSET + PGDIR_SIZE;
> > +	limit = PHYS_OFFSET + MEMBLOCK_INITIAL_LIMIT;
> >  	memblock_set_current_limit(limit);
> >
> >  	/* map all the memory banks */
> > --
> > 1.7.10.4
> >
> >
> 
> Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>

Thanks for comment and review!

- Jungseok Lee

--
To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
??? May 7, 2014, 4:27 a.m. UTC | #4
On Tuesday, May 06, 2014 9:02 PM, Steve Capper wrote:
> On Thu, May 01, 2014 at 11:34:16AM +0900, Jungseok Lee wrote:
> > This patch implements 4 levels of translation tables since 3 levels of
> > page tables with 4KB pages cannot support 40-bit physical address
> > space described in [1] due to the following issue.
> >
> > It is a restriction that kernel logical memory map with 4KB + 3 levels
> > (0xffffffc000000000-0xffffffffffffffff) cannot cover RAM region from
> > 544GB to 1024GB in [1]. Specifically, ARM64 kernel fails to create
> > mapping for this region in map_mem function since __phys_to_virt for
> > this region reaches to address overflow.
> >
> > If SoC design follows the document, [1], over 32GB RAM would be placed
> > from 544GB. Even 64GB system is supposed to use the region from 544GB
> > to 576GB for only 32GB RAM. Naturally, it would reach to enable 4
> > levels of page tables to avoid hacking __virt_to_phys and __phys_to_virt.
> >
> > However, it is recommended 4 levels of page table should be only
> > enabled if memory map is too sparse or there is about 512GB RAM.
> >
> 
> Hi Jungseok,
> One comment below:

Hi Steve.

> [ ... ]
> 
> > diff --git a/arch/arm64/include/asm/tlb.h
> > b/arch/arm64/include/asm/tlb.h index bc19101..086112b 100644
> > --- a/arch/arm64/include/asm/tlb.h
> > +++ b/arch/arm64/include/asm/tlb.h
> > @@ -100,6 +100,15 @@ static inline void __pmd_free_tlb(struct
> > mmu_gather *tlb, pmd_t *pmdp,  }  #endif
> >
> > +#ifdef CONFIG_ARM64_4_LEVELS
> > +static inline void __pud_free_tlb(struct mmu_gather *tlb, pmd_t *pudp,
> > +				  unsigned long addr)
> 
> The second parameter needs to be a pointer to pud_t ?
> (this fires up a warning with STRICT_MM_TYPECHECKS).

You're right. My fault...

> With that and Christoffer's feedback about expanding the comments on create_pud_entry addressed:

Okay. I will add it.

> Reviewed-by: Steve Capper <steve.capper@linaro.org>

Thanks for review!

- Jungseok Lee

--
To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Christoffer Dall May 7, 2014, 8:13 a.m. UTC | #5
On Wed, May 07, 2014 at 01:22:50PM +0900, Jungseok Lee wrote:
> On Tuesday, May 06, 2014 7:49 PM, Christoffer Dall wrote:
> > On Thu, May 01, 2014 at 11:34:16AM +0900, Jungseok Lee wrote:
> > > This patch implements 4 levels of translation tables since 3 levels of
> > > page tables with 4KB pages cannot support 40-bit physical address
> > > space described in [1] due to the following issue.
> > >
> > > It is a restriction that kernel logical memory map with 4KB + 3 levels
> > > (0xffffffc000000000-0xffffffffffffffff) cannot cover RAM region from
> > > 544GB to 1024GB in [1]. Specifically, ARM64 kernel fails to create
> > > mapping for this region in map_mem function since __phys_to_virt for
> > > this region reaches to address overflow.
> > >
> > > If SoC design follows the document, [1], over 32GB RAM would be placed
> > > from 544GB. Even 64GB system is supposed to use the region from 544GB
> > > to 576GB for only 32GB RAM. Naturally, it would reach to enable 4
> > > levels of page tables to avoid hacking __virt_to_phys and __phys_to_virt.
> > >
> > > However, it is recommended 4 levels of page table should be only
> > > enabled if memory map is too sparse or there is about 512GB RAM.
> > 
> > Who recommends this then?  This paragraph just confuses me.
> 
> It is a paraphrase of Catalin's comment:
> "I agree, we should only enable 4-levels of page table if we have close
> to 512GB of RAM or the range is too sparse but I would actually push
> back on the hardware guys to keep it tighter."
> 
> The above message comes from the discussion on 4 levels.
> http://www.spinics.net/linux/lists/arm-kernel/msg319055.html
> 
Hmm, I find it just confusing in the commit message and think it's more
clear if you take it out.  The only relevant bits here are that if you
want to be able to address an address map of a certain size, you need
this feature.  It should be enabled on all systems that have such
address maps to take advantage of the available hardware - arguing to
hardware vendors on which systems to produce is fine, but a Linux commit
message doesn't have to iterate on it - IMHO.

-Christoffer
--
To unsubscribe from this list: send the line "unsubscribe linux-samsung-soc" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index b438540..3e49671 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -182,12 +182,17 @@  config ARM64_VA_BITS_42
 	bool "42-bit"
 	depends on ARM64_64K_PAGES
 
+config ARM64_VA_BITS_48
+	bool "48-bit"
+	depends on ARM64_4K_PAGES
+
 endchoice
 
 config ARM64_VA_BITS
 	int
 	default 39 if ARM64_VA_BITS_39
 	default 42 if ARM64_VA_BITS_42
+	default 48 if ARM64_VA_BITS_48
 
 config ARM64_2_LEVELS
 	def_bool y if ARM64_64K_PAGES && ARM64_VA_BITS_42
@@ -195,6 +200,9 @@  config ARM64_2_LEVELS
 config ARM64_3_LEVELS
 	def_bool y if ARM64_4K_PAGES && ARM64_VA_BITS_39
 
+config ARM64_4_LEVELS
+	def_bool y if ARM64_4K_PAGES && ARM64_VA_BITS_48
+
 config CPU_BIG_ENDIAN
        bool "Build big-endian kernel"
        help
diff --git a/arch/arm64/include/asm/memblock.h b/arch/arm64/include/asm/memblock.h
index 6afeed2..e4ac8bf 100644
--- a/arch/arm64/include/asm/memblock.h
+++ b/arch/arm64/include/asm/memblock.h
@@ -16,6 +16,12 @@ 
 #ifndef __ASM_MEMBLOCK_H
 #define __ASM_MEMBLOCK_H
 
+#ifndef CONFIG_ARM64_4_LEVELS
+#define MEMBLOCK_INITIAL_LIMIT	PGDIR_SIZE
+#else
+#define MEMBLOCK_INITIAL_LIMIT	PUD_SIZE
+#endif
+
 extern void arm64_memblock_init(void);
 
 #endif
diff --git a/arch/arm64/include/asm/page.h b/arch/arm64/include/asm/page.h
index 268e53d..83b5289 100644
--- a/arch/arm64/include/asm/page.h
+++ b/arch/arm64/include/asm/page.h
@@ -35,8 +35,10 @@ 
 
 #ifdef CONFIG_ARM64_2_LEVELS
 #include <asm/pgtable-2level-types.h>
-#else
+#elif defined(CONFIG_ARM64_3_LEVELS)
 #include <asm/pgtable-3level-types.h>
+#else
+#include <asm/pgtable-4level-types.h>
 #endif
 
 extern void __cpu_clear_user_page(void *p, unsigned long user);
diff --git a/arch/arm64/include/asm/pgalloc.h b/arch/arm64/include/asm/pgalloc.h
index 4829837..8d745fa 100644
--- a/arch/arm64/include/asm/pgalloc.h
+++ b/arch/arm64/include/asm/pgalloc.h
@@ -26,6 +26,26 @@ 
 
 #define check_pgt_cache()		do { } while (0)
 
+#ifdef CONFIG_ARM64_4_LEVELS
+
+static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
+{
+	return (pud_t *)get_zeroed_page(GFP_KERNEL | __GFP_REPEAT);
+}
+
+static inline void pud_free(struct mm_struct *mm, pud_t *pud)
+{
+	BUG_ON((unsigned long)pud & (PAGE_SIZE-1));
+	free_page((unsigned long)pud);
+}
+
+static inline void pgd_populate(struct mm_struct *mm, pgd_t *pgd, pud_t *pud)
+{
+	set_pgd(pgd, __pgd(__pa(pud) | PUD_TYPE_TABLE));
+}
+
+#endif  /* CONFIG_ARM64_4_LEVELS */
+
 #ifndef CONFIG_ARM64_2_LEVELS
 
 static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
diff --git a/arch/arm64/include/asm/pgtable-hwdef.h b/arch/arm64/include/asm/pgtable-hwdef.h
index 9cd86c6..ba30053 100644
--- a/arch/arm64/include/asm/pgtable-hwdef.h
+++ b/arch/arm64/include/asm/pgtable-hwdef.h
@@ -18,8 +18,10 @@ 
 
 #ifdef CONFIG_ARM64_2_LEVELS
 #include <asm/pgtable-2level-hwdef.h>
-#else
+#elif defined(CONFIG_ARM64_3_LEVELS)
 #include <asm/pgtable-3level-hwdef.h>
+#else
+#include <asm/pgtable-4level-hwdef.h>
 #endif
 
 /*
@@ -27,7 +29,7 @@ 
  *
  * Level 1 descriptor (PUD).
  */
-
+#define PUD_TYPE_TABLE		(_AT(pudval_t, 3) << 0)
 #define PUD_TABLE_BIT		(_AT(pgdval_t, 1) << 1)
 
 /*
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index a64ce5e..b27cc426 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -35,7 +35,11 @@ 
  * VMALLOC and SPARSEMEM_VMEMMAP ranges.
  */
 #define VMALLOC_START		(UL(0xffffffffffffffff) << VA_BITS)
+#ifndef CONFIG_ARM64_4_LEVELS
 #define VMALLOC_END		(PAGE_OFFSET - UL(0x400000000) - SZ_64K)
+#else
+#define VMALLOC_END		(PAGE_OFFSET - UL(0x40000000000) - SZ_64K)
+#endif
 
 #define vmemmap			((struct page *)(VMALLOC_END + SZ_64K))
 
@@ -44,12 +48,16 @@ 
 #ifndef __ASSEMBLY__
 extern void __pte_error(const char *file, int line, unsigned long val);
 extern void __pmd_error(const char *file, int line, unsigned long val);
+extern void __pud_error(const char *file, int line, unsigned long val);
 extern void __pgd_error(const char *file, int line, unsigned long val);
 
 #define pte_ERROR(pte)		__pte_error(__FILE__, __LINE__, pte_val(pte))
 #ifndef CONFIG_ARM64_2_LEVELS
 #define pmd_ERROR(pmd)		__pmd_error(__FILE__, __LINE__, pmd_val(pmd))
 #endif
+#ifdef CONFIG_ARM64_4_LEVELS
+#define pud_ERROR(pud)		__pud_error(__FILE__, __LINE__, pud_val(pud))
+#endif
 #define pgd_ERROR(pgd)		__pgd_error(__FILE__, __LINE__, pgd_val(pgd))
 
 /*
@@ -344,6 +352,30 @@  static inline pmd_t *pud_page_vaddr(pud_t pud)
 
 #endif	/* CONFIG_ARM64_2_LEVELS */
 
+#ifdef CONFIG_ARM64_4_LEVELS
+
+#define pgd_none(pgd)		(!pgd_val(pgd))
+#define pgd_bad(pgd)		(!(pgd_val(pgd) & 2))
+#define pgd_present(pgd)	(pgd_val(pgd))
+
+static inline void set_pgd(pgd_t *pgdp, pgd_t pgd)
+{
+	*pgdp = pgd;
+	dsb();
+}
+
+static inline void pgd_clear(pgd_t *pgdp)
+{
+	set_pgd(pgdp, __pgd(0));
+}
+
+static inline pud_t *pgd_page_vaddr(pgd_t pgd)
+{
+	return __va(pgd_val(pgd) & PHYS_MASK & (s32)PAGE_MASK);
+}
+
+#endif  /* CONFIG_ARM64_4_LEVELS */
+
 /* to find an entry in a page-table-directory */
 #define pgd_index(addr)		(((addr) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 
@@ -352,6 +384,14 @@  static inline pmd_t *pud_page_vaddr(pud_t pud)
 /* to find an entry in a kernel page-table-directory */
 #define pgd_offset_k(addr)	pgd_offset(&init_mm, addr)
 
+#ifdef CONFIG_ARM64_4_LEVELS
+#define pud_index(addr)		(((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1))
+static inline pud_t *pud_offset(pgd_t *pgd, unsigned long addr)
+{
+	return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(addr);
+}
+#endif
+
 /* Find an entry in the second-level page table.. */
 #ifndef CONFIG_ARM64_2_LEVELS
 #define pmd_index(addr)		(((addr) >> PMD_SHIFT) & (PTRS_PER_PMD - 1))
@@ -380,8 +420,13 @@  static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
 extern pgd_t swapper_pg_dir[PTRS_PER_PGD];
 extern pgd_t idmap_pg_dir[PTRS_PER_PGD];
 
+#ifdef CONFIG_ARM64_4_LEVELS
+#define SWAPPER_DIR_SIZE	(4 * PAGE_SIZE)
+#define IDMAP_DIR_SIZE		(3 * PAGE_SIZE)
+#else
 #define SWAPPER_DIR_SIZE	(3 * PAGE_SIZE)
 #define IDMAP_DIR_SIZE		(2 * PAGE_SIZE)
+#endif
 
 /*
  * Encode and decode a swap entry:
diff --git a/arch/arm64/include/asm/tlb.h b/arch/arm64/include/asm/tlb.h
index bc19101..086112b 100644
--- a/arch/arm64/include/asm/tlb.h
+++ b/arch/arm64/include/asm/tlb.h
@@ -100,6 +100,15 @@  static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmdp,
 }
 #endif
 
+#ifdef CONFIG_ARM64_4_LEVELS
+static inline void __pud_free_tlb(struct mmu_gather *tlb, pmd_t *pudp,
+				  unsigned long addr)
+{
+	tlb_add_flush(tlb, addr);
+	tlb_remove_page(tlb, virt_to_page(pudp));
+}
+#endif
+
 static inline void __tlb_remove_pmd_tlb_entry(struct mmu_gather *tlb, pmd_t *pmdp,
 						unsigned long address)
 {
diff --git a/arch/arm64/kernel/head.S b/arch/arm64/kernel/head.S
index 0fd5650..03ec424 100644
--- a/arch/arm64/kernel/head.S
+++ b/arch/arm64/kernel/head.S
@@ -37,8 +37,9 @@ 
 
 /*
  * swapper_pg_dir is the virtual address of the initial page table. We place
- * the page tables 3 * PAGE_SIZE below KERNEL_RAM_VADDR. The idmap_pg_dir has
- * 2 pages and is placed below swapper_pg_dir.
+ * the page tables 3 * PAGE_SIZE (2 or 3 levels) or 4 * PAGE_SIZE (4 levels)
+ * below KERNEL_RAM_VADDR. The idmap_pg_dir has 2 pages (2 or 3 levels) or
+ * 3 pages (4 levels) and is placed below swapper_pg_dir.
  */
 #define KERNEL_RAM_VADDR	(PAGE_OFFSET + TEXT_OFFSET)
 
@@ -46,8 +47,13 @@ 
 #error KERNEL_RAM_VADDR must start at 0xXXX80000
 #endif
 
+#ifdef CONFIG_ARM64_4_LEVELS
+#define SWAPPER_DIR_SIZE	(4 * PAGE_SIZE)
+#define IDMAP_DIR_SIZE		(3 * PAGE_SIZE)
+#else
 #define SWAPPER_DIR_SIZE	(3 * PAGE_SIZE)
 #define IDMAP_DIR_SIZE		(2 * PAGE_SIZE)
+#endif
 
 	.globl	swapper_pg_dir
 	.equ	swapper_pg_dir, KERNEL_RAM_VADDR - SWAPPER_DIR_SIZE
@@ -370,16 +376,38 @@  ENDPROC(__calc_phys_offset)
 	.quad	PAGE_OFFSET
 
 /*
+ * Macro to populate the PUD for the corresponding block entry in the next
+ * level (tbl) for the given virtual address in case of 4levels.
+ */
+	.macro	create_pud_entry, pgd, tbl, virt, pud, tmp1, tmp2
+#ifdef CONFIG_ARM64_4_LEVELS
+	add	\tbl, \tbl, #PAGE_SIZE		// bump tbl 1 page up.
+						// to make room for pud
+	add	\pud, \pgd, #PAGE_SIZE		// pgd points to pud which
+						// follows pgd
+	lsr	\tmp1, \virt, #PUD_SHIFT
+	and	\tmp1, \tmp1, #PTRS_PER_PUD - 1	// PUD index
+	orr	\tmp2, \tbl, #3			// PUD entry table type
+	str	\tmp2, [\pud, \tmp1, lsl #3]
+#else
+	mov	\pud, \tbl
+#endif
+	.endm
+
+/*
  * Macro to populate the PGD for the corresponding block entry in the next
  * level (tbl) for the given virtual address.
  *
- * Preserves:	pgd, tbl, virt
- * Corrupts:	tmp1, tmp2
+ * Preserves:	pgd, virt
+ * Corrupts:	tmp1, tmp2, tmp3
+ * Returns:	tbl -> page where block mappings can be placed
+ *	(changed to make room for pud with 4levels, preserved otherwise)
  */
-	.macro	create_pgd_entry, pgd, tbl, virt, tmp1, tmp2
+	.macro	create_pgd_entry, pgd, tbl, virt, tmp1, tmp2, tmp3
+	create_pud_entry \pgd, \tbl, \virt, \tmp3, \tmp1, \tmp2
 	lsr	\tmp1, \virt, #PGDIR_SHIFT
 	and	\tmp1, \tmp1, #PTRS_PER_PGD - 1	// PGD index
-	orr	\tmp2, \tbl, #3			// PGD entry table type
+	orr	\tmp2, \tmp3, #3		// PGD entry table type
 	str	\tmp2, [\pgd, \tmp1, lsl #3]
 	.endm
 
@@ -444,7 +472,7 @@  __create_page_tables:
 	add	x0, x25, #PAGE_SIZE		// section table address
 	ldr	x3, =KERNEL_START
 	add	x3, x3, x28			// __pa(KERNEL_START)
-	create_pgd_entry x25, x0, x3, x5, x6
+	create_pgd_entry x25, x0, x3, x1, x5, x6
 	ldr	x6, =KERNEL_END
 	mov	x5, x3				// __pa(KERNEL_START)
 	add	x6, x6, x28			// __pa(KERNEL_END)
@@ -455,7 +483,7 @@  __create_page_tables:
 	 */
 	add	x0, x26, #PAGE_SIZE		// section table address
 	mov	x5, #PAGE_OFFSET
-	create_pgd_entry x26, x0, x5, x3, x6
+	create_pgd_entry x26, x0, x5, x1, x3, x6
 	ldr	x6, =KERNEL_END
 	mov	x3, x24				// phys offset
 	create_block_map x0, x7, x3, x5, x6
@@ -481,7 +509,7 @@  __create_page_tables:
 	 */
 	ldr	x5, =FIXADDR_TOP		// Fixed mapping virtual address
 	add	x0, x26, #2 * PAGE_SIZE		// section table address
-	create_pgd_entry x26, x0, x5, x6, x7
+	create_pgd_entry x26, x0, x5, x1, x6, x7
 
 	/*
 	 * Since the page tables have been populated with non-cacheable
diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
index 268ce96..237757d 100644
--- a/arch/arm64/kernel/traps.c
+++ b/arch/arm64/kernel/traps.c
@@ -336,6 +336,11 @@  void __pmd_error(const char *file, int line, unsigned long val)
 	pr_crit("%s:%d: bad pmd %016lx.\n", file, line, val);
 }
 
+void __pud_error(const char *file, int line, unsigned long val)
+{
+	pr_crit("%s:%d: bad pud %016lx.\n", file, line, val);
+}
+
 void __pgd_error(const char *file, int line, unsigned long val)
 {
 	pr_crit("%s:%d: bad pgd %016lx.\n", file, line, val);
diff --git a/arch/arm64/mm/fault.c b/arch/arm64/mm/fault.c
index c23751b..ed4a343 100644
--- a/arch/arm64/mm/fault.c
+++ b/arch/arm64/mm/fault.c
@@ -61,6 +61,7 @@  void show_pte(struct mm_struct *mm, unsigned long addr)
 			break;
 
 		pud = pud_offset(pgd, addr);
+		printk(", *pud=%016llx", pud_val(*pud));
 		if (pud_none(*pud) || pud_bad(*pud))
 			break;
 
diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index 6b7e895..4d29332 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -32,6 +32,7 @@ 
 #include <asm/setup.h>
 #include <asm/sizes.h>
 #include <asm/tlb.h>
+#include <asm/memblock.h>
 #include <asm/mmu_context.h>
 
 #include "mm.h"
@@ -222,9 +223,15 @@  static void __init alloc_init_pmd(pud_t *pud, unsigned long addr,
 static void __init alloc_init_pud(pgd_t *pgd, unsigned long addr,
 				  unsigned long end, unsigned long phys)
 {
-	pud_t *pud = pud_offset(pgd, addr);
+	pud_t *pud;
 	unsigned long next;
 
+	if (pgd_none(*pgd) || pgd_bad(*pgd)) {
+		pud = early_alloc(PTRS_PER_PUD * sizeof(pud_t));
+		pgd_populate(&init_mm, pgd, pud);
+	}
+
+	pud = pud_offset(pgd, addr);
 	do {
 		next = pud_addr_end(addr, end);
 		alloc_init_pmd(pud, addr, next, phys);
@@ -271,10 +278,11 @@  static void __init map_mem(void)
 	 * memory addressable from the initial direct kernel mapping.
 	 *
 	 * The initial direct kernel mapping, located at swapper_pg_dir,
-	 * gives us PGDIR_SIZE memory starting from PHYS_OFFSET (which must be
-	 * aligned to 2MB as per Documentation/arm64/booting.txt).
+	 * gives us PGDIR_SIZE (2 and 3 levels) or PUD_SIZE (4 levels) memory
+	 * starting from PHYS_OFFSET (which must be aligned to 2MB as per
+	 * Documentation/arm64/booting.txt).
 	 */
-	limit = PHYS_OFFSET + PGDIR_SIZE;
+	limit = PHYS_OFFSET + MEMBLOCK_INITIAL_LIMIT;
 	memblock_set_current_limit(limit);
 
 	/* map all the memory banks */