2015-02-06 14:51:13

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 00/19] expose page table levels as Kconfig option

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

v2:
- powerpc: s/64K_PAGES/PPC_64K_PAGES/;
- s390: fix typo s/64BI/64BIT/;
- ia64: fix default for IA64_PAGE_SIZE_64KB;
- x86: s/PAGETABLE_LEVELS/CONFIG_PGTABLE_LEVELS/ include/trace/events/xen.h;

Kirill A. Shutemov (19):
alpha: expose number of page table levels on Kconfig level
arm64: expose number of page table levels on Kconfig level
arm: expose number of page table levels on Kconfig level
frv: mark PUD and PMD folded
ia64: expose number of page table levels on Kconfig level
m32r: mark PMD folded
m68k: mark PMD folded and expose number of page table levels
mips: expose number of page table levels on Kconfig level
mn10300: mark PUD and PMD folded
parisc: expose number of page table levels on Kconfig level
powerpc: expose number of page table levels on Kconfig level
s390: expose number of page table levels
sh: expose number of page table levels
sparc: expose number of page table levels
tile: expose number of page table levels
um: expose number of page table levels
x86: expose number of page table levels on Kconfig level
mm: define default PGTABLE_LEVELS to two
mm: do not add nr_pmds into mm_struct if PMD is folded

arch/Kconfig | 4 ++++
arch/alpha/Kconfig | 4 ++++
arch/arm/Kconfig | 5 +++++
arch/arm64/Kconfig | 14 +++++++-------
arch/arm64/include/asm/kvm_mmu.h | 4 ++--
arch/arm64/include/asm/page.h | 4 ++--
arch/arm64/include/asm/pgalloc.h | 8 ++++----
arch/arm64/include/asm/pgtable-hwdef.h | 6 +++---
arch/arm64/include/asm/pgtable-types.h | 12 ++++++------
arch/arm64/include/asm/pgtable.h | 8 ++++----
arch/arm64/include/asm/tlb.h | 4 ++--
arch/arm64/mm/mmu.c | 4 ++--
arch/frv/include/asm/pgtable.h | 2 ++
arch/ia64/Kconfig | 18 +++++-------------
arch/ia64/include/asm/page.h | 4 ++--
arch/ia64/include/asm/pgalloc.h | 4 ++--
arch/ia64/include/asm/pgtable.h | 12 ++++++------
arch/ia64/kernel/ivt.S | 12 ++++++------
arch/ia64/kernel/machine_kexec.c | 4 ++--
arch/m32r/include/asm/pgtable-2level.h | 1 +
arch/m68k/Kconfig | 4 ++++
arch/m68k/include/asm/pgtable_mm.h | 2 ++
arch/mips/Kconfig | 5 +++++
arch/mn10300/include/asm/pgtable.h | 2 ++
arch/parisc/Kconfig | 5 +++++
arch/parisc/include/asm/pgalloc.h | 2 +-
arch/parisc/include/asm/pgtable.h | 17 ++++++++---------
arch/parisc/kernel/entry.S | 4 ++--
arch/parisc/kernel/head.S | 4 ++--
arch/parisc/mm/init.c | 2 +-
arch/powerpc/Kconfig | 6 ++++++
arch/s390/Kconfig | 5 +++++
arch/s390/include/asm/pgtable.h | 2 ++
arch/sh/Kconfig | 4 ++++
arch/sparc/Kconfig | 4 ++++
arch/tile/Kconfig | 5 +++++
arch/um/Kconfig.um | 5 +++++
arch/x86/Kconfig | 6 ++++++
arch/x86/include/asm/paravirt.h | 8 ++++----
arch/x86/include/asm/paravirt_types.h | 8 ++++----
arch/x86/include/asm/pgalloc.h | 8 ++++----
arch/x86/include/asm/pgtable-2level_types.h | 1 -
arch/x86/include/asm/pgtable-3level_types.h | 2 --
arch/x86/include/asm/pgtable.h | 8 ++++----
arch/x86/include/asm/pgtable_64_types.h | 1 -
arch/x86/include/asm/pgtable_types.h | 4 ++--
arch/x86/kernel/paravirt.c | 6 +++---
arch/x86/mm/pgtable.c | 14 +++++++-------
arch/x86/xen/mmu.c | 14 +++++++-------
include/asm-generic/pgtable.h | 5 +++++
include/linux/mm_types.h | 2 ++
include/trace/events/xen.h | 2 +-
52 files changed, 183 insertions(+), 118 deletions(-)

--
2.1.4


2015-02-06 14:56:51

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 01/19] alpha: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Richard Henderson <[email protected]>
Cc: Ivan Kokshaysky <[email protected]>
Cc: Matt Turner <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/alpha/Kconfig | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/arch/alpha/Kconfig b/arch/alpha/Kconfig
index b7ff9a318c31..bf9e9d3b3792 100644
--- a/arch/alpha/Kconfig
+++ b/arch/alpha/Kconfig
@@ -76,6 +76,10 @@ config GENERIC_ISA_DMA
bool
default y

+config PGTABLE_LEVELS
+ int
+ default 3
+
source "init/Kconfig"
source "kernel/Kconfig.freezer"

--
2.1.4

2015-02-06 14:51:19

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 02/19] arm64: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

ARM64_PGTABLE_LEVELS is renamed to PGTABLE_LEVELS and defined before
sourcing init/Kconfig: arch/Kconfig will define default value and it's
sourced from init/Kconfig.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Acked-by: Catalin Marinas <[email protected]>
Cc: Will Deacon <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/arm64/Kconfig | 14 +++++++-------
arch/arm64/include/asm/kvm_mmu.h | 4 ++--
arch/arm64/include/asm/page.h | 4 ++--
arch/arm64/include/asm/pgalloc.h | 8 ++++----
arch/arm64/include/asm/pgtable-hwdef.h | 6 +++---
arch/arm64/include/asm/pgtable-types.h | 12 ++++++------
arch/arm64/include/asm/pgtable.h | 8 ++++----
arch/arm64/include/asm/tlb.h | 4 ++--
arch/arm64/mm/mmu.c | 4 ++--
9 files changed, 32 insertions(+), 32 deletions(-)

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 1b8e97331ffb..3f2fba996bc2 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -143,6 +143,13 @@ config KERNEL_MODE_NEON
config FIX_EARLYCON_MEM
def_bool y

+config PGTABLE_LEVELS
+ int
+ default 2 if ARM64_64K_PAGES && ARM64_VA_BITS_42
+ default 3 if ARM64_64K_PAGES && ARM64_VA_BITS_48
+ default 3 if ARM64_4K_PAGES && ARM64_VA_BITS_39
+ default 4 if ARM64_4K_PAGES && ARM64_VA_BITS_48
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
@@ -413,13 +420,6 @@ config ARM64_VA_BITS
default 42 if ARM64_VA_BITS_42
default 48 if ARM64_VA_BITS_48

-config ARM64_PGTABLE_LEVELS
- int
- default 2 if ARM64_64K_PAGES && ARM64_VA_BITS_42
- default 3 if ARM64_64K_PAGES && ARM64_VA_BITS_48
- default 3 if ARM64_4K_PAGES && ARM64_VA_BITS_39
- default 4 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/kvm_mmu.h b/arch/arm64/include/asm/kvm_mmu.h
index 6458b5373142..19660692efec 100644
--- a/arch/arm64/include/asm/kvm_mmu.h
+++ b/arch/arm64/include/asm/kvm_mmu.h
@@ -161,12 +161,12 @@ static inline bool kvm_s2pmd_readonly(pmd_t *pmd)
/*
* If we are concatenating first level stage-2 page tables, we would have less
* than or equal to 16 pointers in the fake PGD, because that's what the
- * architecture allows. In this case, (4 - CONFIG_ARM64_PGTABLE_LEVELS)
+ * architecture allows. In this case, (4 - CONFIG_PGTABLE_LEVELS)
* represents the first level for the host, and we add 1 to go to the next
* level (which uses contatenation) for the stage-2 tables.
*/
#if PTRS_PER_S2_PGD <= 16
-#define KVM_PREALLOC_LEVEL (4 - CONFIG_ARM64_PGTABLE_LEVELS + 1)
+#define KVM_PREALLOC_LEVEL (4 - CONFIG_PGTABLE_LEVELS + 1)
#else
#define KVM_PREALLOC_LEVEL (0)
#endif
diff --git a/arch/arm64/include/asm/page.h b/arch/arm64/include/asm/page.h
index 22b16232bd60..8fc8fa280e92 100644
--- a/arch/arm64/include/asm/page.h
+++ b/arch/arm64/include/asm/page.h
@@ -36,9 +36,9 @@
* for more information).
*/
#ifdef CONFIG_ARM64_64K_PAGES
-#define SWAPPER_PGTABLE_LEVELS (CONFIG_ARM64_PGTABLE_LEVELS)
+#define SWAPPER_PGTABLE_LEVELS (CONFIG_PGTABLE_LEVELS)
#else
-#define SWAPPER_PGTABLE_LEVELS (CONFIG_ARM64_PGTABLE_LEVELS - 1)
+#define SWAPPER_PGTABLE_LEVELS (CONFIG_PGTABLE_LEVELS - 1)
#endif

#define SWAPPER_DIR_SIZE (SWAPPER_PGTABLE_LEVELS * PAGE_SIZE)
diff --git a/arch/arm64/include/asm/pgalloc.h b/arch/arm64/include/asm/pgalloc.h
index e20df38a8ff3..76420568d66a 100644
--- a/arch/arm64/include/asm/pgalloc.h
+++ b/arch/arm64/include/asm/pgalloc.h
@@ -28,7 +28,7 @@

#define PGALLOC_GFP (GFP_KERNEL | __GFP_NOTRACK | __GFP_REPEAT | __GFP_ZERO)

-#if CONFIG_ARM64_PGTABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2

static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
{
@@ -46,9 +46,9 @@ static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
set_pud(pud, __pud(__pa(pmd) | PMD_TYPE_TABLE));
}

-#endif /* CONFIG_ARM64_PGTABLE_LEVELS > 2 */
+#endif /* CONFIG_PGTABLE_LEVELS > 2 */

-#if CONFIG_ARM64_PGTABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3

static inline pud_t *pud_alloc_one(struct mm_struct *mm, unsigned long addr)
{
@@ -66,7 +66,7 @@ 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_PGTABLE_LEVELS > 3 */
+#endif /* CONFIG_PGTABLE_LEVELS > 3 */

extern pgd_t *pgd_alloc(struct mm_struct *mm);
extern void pgd_free(struct mm_struct *mm, pgd_t *pgd);
diff --git a/arch/arm64/include/asm/pgtable-hwdef.h b/arch/arm64/include/asm/pgtable-hwdef.h
index 5f930cc9ea83..80f3d241cff8 100644
--- a/arch/arm64/include/asm/pgtable-hwdef.h
+++ b/arch/arm64/include/asm/pgtable-hwdef.h
@@ -21,7 +21,7 @@
/*
* PMD_SHIFT determines the size a level 2 page table entry can map.
*/
-#if CONFIG_ARM64_PGTABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
#define PMD_SHIFT ((PAGE_SHIFT - 3) * 2 + 3)
#define PMD_SIZE (_AC(1, UL) << PMD_SHIFT)
#define PMD_MASK (~(PMD_SIZE-1))
@@ -31,7 +31,7 @@
/*
* PUD_SHIFT determines the size a level 1 page table entry can map.
*/
-#if CONFIG_ARM64_PGTABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
#define PUD_SHIFT ((PAGE_SHIFT - 3) * 3 + 3)
#define PUD_SIZE (_AC(1, UL) << PUD_SHIFT)
#define PUD_MASK (~(PUD_SIZE-1))
@@ -42,7 +42,7 @@
* PGDIR_SHIFT determines the size a top-level page table entry can map
* (depending on the configuration, this level can be 0, 1 or 2).
*/
-#define PGDIR_SHIFT ((PAGE_SHIFT - 3) * CONFIG_ARM64_PGTABLE_LEVELS + 3)
+#define PGDIR_SHIFT ((PAGE_SHIFT - 3) * CONFIG_PGTABLE_LEVELS + 3)
#define PGDIR_SIZE (_AC(1, UL) << PGDIR_SHIFT)
#define PGDIR_MASK (~(PGDIR_SIZE-1))
#define PTRS_PER_PGD (1 << (VA_BITS - PGDIR_SHIFT))
diff --git a/arch/arm64/include/asm/pgtable-types.h b/arch/arm64/include/asm/pgtable-types.h
index ca9df80af896..2b1bd7e52c3b 100644
--- a/arch/arm64/include/asm/pgtable-types.h
+++ b/arch/arm64/include/asm/pgtable-types.h
@@ -38,13 +38,13 @@ typedef struct { pteval_t pte; } pte_t;
#define pte_val(x) ((x).pte)
#define __pte(x) ((pte_t) { (x) } )

-#if CONFIG_ARM64_PGTABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
typedef struct { pmdval_t pmd; } pmd_t;
#define pmd_val(x) ((x).pmd)
#define __pmd(x) ((pmd_t) { (x) } )
#endif

-#if CONFIG_ARM64_PGTABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
typedef struct { pudval_t pud; } pud_t;
#define pud_val(x) ((x).pud)
#define __pud(x) ((pud_t) { (x) } )
@@ -64,13 +64,13 @@ typedef pteval_t pte_t;
#define pte_val(x) (x)
#define __pte(x) (x)

-#if CONFIG_ARM64_PGTABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
typedef pmdval_t pmd_t;
#define pmd_val(x) (x)
#define __pmd(x) (x)
#endif

-#if CONFIG_ARM64_PGTABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
typedef pudval_t pud_t;
#define pud_val(x) (x)
#define __pud(x) (x)
@@ -86,9 +86,9 @@ typedef pteval_t pgprot_t;

#endif /* STRICT_MM_TYPECHECKS */

-#if CONFIG_ARM64_PGTABLE_LEVELS == 2
+#if CONFIG_PGTABLE_LEVELS == 2
#include <asm-generic/pgtable-nopmd.h>
-#elif CONFIG_ARM64_PGTABLE_LEVELS == 3
+#elif CONFIG_PGTABLE_LEVELS == 3
#include <asm-generic/pgtable-nopud.h>
#endif

diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index a26a5740afc4..9fb699405eb9 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -376,7 +376,7 @@ static inline pte_t *pmd_page_vaddr(pmd_t pmd)
*/
#define mk_pte(page,prot) pfn_pte(page_to_pfn(page),prot)

-#if CONFIG_ARM64_PGTABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2

#define pmd_ERROR(pmd) __pmd_error(__FILE__, __LINE__, pmd_val(pmd))

@@ -411,9 +411,9 @@ static inline pmd_t *pmd_offset(pud_t *pud, unsigned long addr)

#define pud_page(pud) pfn_to_page(__phys_to_pfn(pud_val(pud) & PHYS_MASK))

-#endif /* CONFIG_ARM64_PGTABLE_LEVELS > 2 */
+#endif /* CONFIG_PGTABLE_LEVELS > 2 */

-#if CONFIG_ARM64_PGTABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3

#define pud_ERROR(pud) __pud_error(__FILE__, __LINE__, pud_val(pud))

@@ -447,7 +447,7 @@ static inline pud_t *pud_offset(pgd_t *pgd, unsigned long addr)

#define pgd_page(pgd) pfn_to_page(__phys_to_pfn(pgd_val(pgd) & PHYS_MASK))

-#endif /* CONFIG_ARM64_PGTABLE_LEVELS > 3 */
+#endif /* CONFIG_PGTABLE_LEVELS > 3 */

#define pgd_ERROR(pgd) __pgd_error(__FILE__, __LINE__, pgd_val(pgd))

diff --git a/arch/arm64/include/asm/tlb.h b/arch/arm64/include/asm/tlb.h
index c028fe37456f..66399cf6b490 100644
--- a/arch/arm64/include/asm/tlb.h
+++ b/arch/arm64/include/asm/tlb.h
@@ -52,7 +52,7 @@ static inline void __pte_free_tlb(struct mmu_gather *tlb, pgtable_t pte,
tlb_remove_entry(tlb, pte);
}

-#if CONFIG_ARM64_PGTABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmdp,
unsigned long addr)
{
@@ -60,7 +60,7 @@ static inline void __pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmdp,
}
#endif

-#if CONFIG_ARM64_PGTABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
static inline void __pud_free_tlb(struct mmu_gather *tlb, pud_t *pudp,
unsigned long addr)
{
diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index c6daaf6c6f97..79e01163a981 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -550,10 +550,10 @@ void vmemmap_free(unsigned long start, unsigned long end)
#endif /* CONFIG_SPARSEMEM_VMEMMAP */

static pte_t bm_pte[PTRS_PER_PTE] __page_aligned_bss;
-#if CONFIG_ARM64_PGTABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
static pmd_t bm_pmd[PTRS_PER_PMD] __page_aligned_bss;
#endif
-#if CONFIG_ARM64_PGTABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
static pud_t bm_pud[PTRS_PER_PUD] __page_aligned_bss;
#endif

--
2.1.4

2015-02-06 14:56:54

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 03/19] arm: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Russell King <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/arm/Kconfig | 5 +++++
1 file changed, 5 insertions(+)

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index 9f1f09a2bc9b..16f20b2b37e3 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -286,6 +286,11 @@ config GENERIC_BUG
def_bool y
depends on BUG

+config PGTABLE_LEVELS
+ int
+ default 3 if ARM_LPAE
+ default 2
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
--
2.1.4

2015-02-06 14:51:45

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 04/19] frv: mark PUD and PMD folded

Core mm expects __PAGETABLE_{PUD,PMD}_FOLDED to be defined if these page
table levels folded.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: David Howells <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/frv/include/asm/pgtable.h | 2 ++
1 file changed, 2 insertions(+)

diff --git a/arch/frv/include/asm/pgtable.h b/arch/frv/include/asm/pgtable.h
index 93bcf2abd1a1..07d7a7ef8bd5 100644
--- a/arch/frv/include/asm/pgtable.h
+++ b/arch/frv/include/asm/pgtable.h
@@ -123,12 +123,14 @@ extern unsigned long empty_zero_page;
#define PGDIR_MASK (~(PGDIR_SIZE - 1))
#define PTRS_PER_PGD 64

+#define __PAGETABLE_PUD_FOLDED
#define PUD_SHIFT 26
#define PTRS_PER_PUD 1
#define PUD_SIZE (1UL << PUD_SHIFT)
#define PUD_MASK (~(PUD_SIZE - 1))
#define PUE_SIZE 256

+#define __PAGETABLE_PMD_FOLDED
#define PMD_SHIFT 26
#define PMD_SIZE (1UL << PMD_SHIFT)
#define PMD_MASK (~(PMD_SIZE - 1))
--
2.1.4

2015-02-06 14:51:24

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 05/19] ia64: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

We need to define PGTABLE_LEVELS before sourcing init/Kconfig:
arch/Kconfig will define default value and it's sourced from init/Kconfig.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Tony Luck <[email protected]>
Cc: Fenghua Yu <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/ia64/Kconfig | 18 +++++-------------
arch/ia64/include/asm/page.h | 4 ++--
arch/ia64/include/asm/pgalloc.h | 4 ++--
arch/ia64/include/asm/pgtable.h | 12 ++++++------
arch/ia64/kernel/ivt.S | 12 ++++++------
arch/ia64/kernel/machine_kexec.c | 4 ++--
6 files changed, 23 insertions(+), 31 deletions(-)

diff --git a/arch/ia64/Kconfig b/arch/ia64/Kconfig
index 074e52bf815c..4f9a6661491b 100644
--- a/arch/ia64/Kconfig
+++ b/arch/ia64/Kconfig
@@ -1,3 +1,8 @@
+config PGTABLE_LEVELS
+ int "Page Table Levels" if !IA64_PAGE_SIZE_64KB
+ range 3 4 if !IA64_PAGE_SIZE_64KB
+ default 3
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
@@ -286,19 +291,6 @@ config IA64_PAGE_SIZE_64KB

endchoice

-choice
- prompt "Page Table Levels"
- default PGTABLE_3
-
-config PGTABLE_3
- bool "3 Levels"
-
-config PGTABLE_4
- depends on !IA64_PAGE_SIZE_64KB
- bool "4 Levels"
-
-endchoice
-
if IA64_HP_SIM
config HZ
default 32
diff --git a/arch/ia64/include/asm/page.h b/arch/ia64/include/asm/page.h
index 1f1bf144fe62..ec48bb9f95e1 100644
--- a/arch/ia64/include/asm/page.h
+++ b/arch/ia64/include/asm/page.h
@@ -173,7 +173,7 @@ get_order (unsigned long size)
*/
typedef struct { unsigned long pte; } pte_t;
typedef struct { unsigned long pmd; } pmd_t;
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
typedef struct { unsigned long pud; } pud_t;
#endif
typedef struct { unsigned long pgd; } pgd_t;
@@ -182,7 +182,7 @@ get_order (unsigned long size)

# define pte_val(x) ((x).pte)
# define pmd_val(x) ((x).pmd)
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
# define pud_val(x) ((x).pud)
#endif
# define pgd_val(x) ((x).pgd)
diff --git a/arch/ia64/include/asm/pgalloc.h b/arch/ia64/include/asm/pgalloc.h
index 5767cdfc08db..f5e70e961948 100644
--- a/arch/ia64/include/asm/pgalloc.h
+++ b/arch/ia64/include/asm/pgalloc.h
@@ -32,7 +32,7 @@ static inline void pgd_free(struct mm_struct *mm, pgd_t *pgd)
quicklist_free(0, NULL, pgd);
}

-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
static inline void
pgd_populate(struct mm_struct *mm, pgd_t * pgd_entry, pud_t * pud)
{
@@ -49,7 +49,7 @@ static inline void pud_free(struct mm_struct *mm, pud_t *pud)
quicklist_free(0, NULL, pud);
}
#define __pud_free_tlb(tlb, pud, address) pud_free((tlb)->mm, pud)
-#endif /* CONFIG_PGTABLE_4 */
+#endif /* CONFIG_PGTABLE_LEVELS == 4 */

static inline void
pud_populate(struct mm_struct *mm, pud_t * pud_entry, pmd_t * pmd)
diff --git a/arch/ia64/include/asm/pgtable.h b/arch/ia64/include/asm/pgtable.h
index 7b6f8801df57..9f3ed9ee8f13 100644
--- a/arch/ia64/include/asm/pgtable.h
+++ b/arch/ia64/include/asm/pgtable.h
@@ -99,7 +99,7 @@
#define PMD_MASK (~(PMD_SIZE-1))
#define PTRS_PER_PMD (1UL << (PTRS_PER_PTD_SHIFT))

-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
/*
* Definitions for second level:
*
@@ -117,7 +117,7 @@
*
* PGDIR_SHIFT determines what a first-level page table entry can map.
*/
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
#define PGDIR_SHIFT (PUD_SHIFT + (PTRS_PER_PTD_SHIFT))
#else
#define PGDIR_SHIFT (PMD_SHIFT + (PTRS_PER_PTD_SHIFT))
@@ -180,7 +180,7 @@
#define __S111 __pgprot(__ACCESS_BITS | _PAGE_PL_3 | _PAGE_AR_RWX)

#define pgd_ERROR(e) printk("%s:%d: bad pgd %016lx.\n", __FILE__, __LINE__, pgd_val(e))
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
#define pud_ERROR(e) printk("%s:%d: bad pud %016lx.\n", __FILE__, __LINE__, pud_val(e))
#endif
#define pmd_ERROR(e) printk("%s:%d: bad pmd %016lx.\n", __FILE__, __LINE__, pmd_val(e))
@@ -281,7 +281,7 @@ extern unsigned long VMALLOC_END;
#define pud_page_vaddr(pud) ((unsigned long) __va(pud_val(pud) & _PFN_MASK))
#define pud_page(pud) virt_to_page((pud_val(pud) + PAGE_OFFSET))

-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
#define pgd_none(pgd) (!pgd_val(pgd))
#define pgd_bad(pgd) (!ia64_phys_addr_valid(pgd_val(pgd)))
#define pgd_present(pgd) (pgd_val(pgd) != 0UL)
@@ -384,7 +384,7 @@ pgd_offset (const struct mm_struct *mm, unsigned long address)
here. */
#define pgd_offset_gate(mm, addr) pgd_offset_k(addr)

-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
/* Find an entry in the second-level page table.. */
#define pud_offset(dir,addr) \
((pud_t *) pgd_page_vaddr(*(dir)) + (((addr) >> PUD_SHIFT) & (PTRS_PER_PUD - 1)))
@@ -586,7 +586,7 @@ extern struct page *zero_page_memmap_ptr;
#define __HAVE_ARCH_PGD_OFFSET_GATE


-#ifndef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 3
#include <asm-generic/pgtable-nopud.h>
#endif
#include <asm-generic/pgtable.h>
diff --git a/arch/ia64/kernel/ivt.S b/arch/ia64/kernel/ivt.S
index 18e794a57248..e42bf7a913f3 100644
--- a/arch/ia64/kernel/ivt.S
+++ b/arch/ia64/kernel/ivt.S
@@ -146,7 +146,7 @@ ENTRY(vhpt_miss)
(p6) dep r17=r18,r19,3,(PAGE_SHIFT-3) // r17=pgd_offset for region 5
(p7) dep r17=r18,r17,3,(PAGE_SHIFT-6) // r17=pgd_offset for region[0-4]
cmp.eq p7,p6=0,r21 // unused address bits all zeroes?
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
shr.u r28=r22,PUD_SHIFT // shift pud index into position
#else
shr.u r18=r22,PMD_SHIFT // shift pmd index into position
@@ -155,7 +155,7 @@ ENTRY(vhpt_miss)
ld8 r17=[r17] // get *pgd (may be 0)
;;
(p7) cmp.eq p6,p7=r17,r0 // was pgd_present(*pgd) == NULL?
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
dep r28=r28,r17,3,(PAGE_SHIFT-3) // r28=pud_offset(pgd,addr)
;;
shr.u r18=r22,PMD_SHIFT // shift pmd index into position
@@ -222,13 +222,13 @@ ENTRY(vhpt_miss)
*/
ld8 r25=[r21] // read *pte again
ld8 r26=[r17] // read *pmd again
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
ld8 r19=[r28] // read *pud again
#endif
cmp.ne p6,p7=r0,r0
;;
cmp.ne.or.andcm p6,p7=r26,r20 // did *pmd change
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
cmp.ne.or.andcm p6,p7=r19,r29 // did *pud change
#endif
mov r27=PAGE_SHIFT<<2
@@ -476,7 +476,7 @@ ENTRY(nested_dtlb_miss)
(p6) dep r17=r18,r19,3,(PAGE_SHIFT-3) // r17=pgd_offset for region 5
(p7) dep r17=r18,r17,3,(PAGE_SHIFT-6) // r17=pgd_offset for region[0-4]
cmp.eq p7,p6=0,r21 // unused address bits all zeroes?
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
shr.u r18=r22,PUD_SHIFT // shift pud index into position
#else
shr.u r18=r22,PMD_SHIFT // shift pmd index into position
@@ -487,7 +487,7 @@ ENTRY(nested_dtlb_miss)
(p7) cmp.eq p6,p7=r17,r0 // was pgd_present(*pgd) == NULL?
dep r17=r18,r17,3,(PAGE_SHIFT-3) // r17=p[u|m]d_offset(pgd,addr)
;;
-#ifdef CONFIG_PGTABLE_4
+#if CONFIG_PGTABLE_LEVELS == 4
(p7) ld8 r17=[r17] // get *pud (may be 0)
shr.u r18=r22,PMD_SHIFT // shift pmd index into position
;;
diff --git a/arch/ia64/kernel/machine_kexec.c b/arch/ia64/kernel/machine_kexec.c
index 5151a649c96b..b72cd7a07222 100644
--- a/arch/ia64/kernel/machine_kexec.c
+++ b/arch/ia64/kernel/machine_kexec.c
@@ -156,9 +156,9 @@ void arch_crash_save_vmcoreinfo(void)
VMCOREINFO_OFFSET(node_memblk_s, start_paddr);
VMCOREINFO_OFFSET(node_memblk_s, size);
#endif
-#ifdef CONFIG_PGTABLE_3
+#if CONFIG_PGTABLE_LEVELS == 3
VMCOREINFO_CONFIG(PGTABLE_3);
-#elif defined(CONFIG_PGTABLE_4)
+#elif CONFIG_PGTABLE_LEVELS == 4
VMCOREINFO_CONFIG(PGTABLE_4);
#endif
}
--
2.1.4

2015-02-06 14:56:56

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 06/19] m32r: mark PMD folded

Core mm expects __PAGETABLE_{PUD,PMD}_FOLDED to be defined if these page
table levels folded.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/m32r/include/asm/pgtable-2level.h | 1 +
1 file changed, 1 insertion(+)

diff --git a/arch/m32r/include/asm/pgtable-2level.h b/arch/m32r/include/asm/pgtable-2level.h
index 8fd8ee70266a..421e6ba3a173 100644
--- a/arch/m32r/include/asm/pgtable-2level.h
+++ b/arch/m32r/include/asm/pgtable-2level.h
@@ -13,6 +13,7 @@
* the M32R is two-level, so we don't really have any
* PMD directory physically.
*/
+#define __PAGETABLE_PMD_FOLDED
#define PMD_SHIFT 22
#define PTRS_PER_PMD 1

--
2.1.4

2015-02-06 14:56:50

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 07/19] m68k: mark PMD folded and expose number of page table levels

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Core mm expects __PAGETABLE_{PUD,PMD}_FOLDED to be defined if these page
table levels folded.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Geert Uytterhoeven <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/m68k/Kconfig | 4 ++++
arch/m68k/include/asm/pgtable_mm.h | 2 ++
2 files changed, 6 insertions(+)

diff --git a/arch/m68k/Kconfig b/arch/m68k/Kconfig
index 87b7c7581b1d..2dd8f63bfbbb 100644
--- a/arch/m68k/Kconfig
+++ b/arch/m68k/Kconfig
@@ -67,6 +67,10 @@ config HZ
default 1000 if CLEOPATRA
default 100

+config PGTABLE_LEVELS
+ default 2 if SUN3 || COLDFIRE
+ default 3
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
diff --git a/arch/m68k/include/asm/pgtable_mm.h b/arch/m68k/include/asm/pgtable_mm.h
index 28a145bfbb71..35ed4a9981ae 100644
--- a/arch/m68k/include/asm/pgtable_mm.h
+++ b/arch/m68k/include/asm/pgtable_mm.h
@@ -54,10 +54,12 @@
*/
#ifdef CONFIG_SUN3
#define PTRS_PER_PTE 16
+#define __PAGETABLE_PMD_FOLDED
#define PTRS_PER_PMD 1
#define PTRS_PER_PGD 2048
#elif defined(CONFIG_COLDFIRE)
#define PTRS_PER_PTE 512
+#define __PAGETABLE_PMD_FOLDED
#define PTRS_PER_PMD 1
#define PTRS_PER_PGD 1024
#else
--
2.1.4

2015-02-06 14:56:53

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 08/19] mips: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Ralf Baechle <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/mips/Kconfig | 5 +++++
1 file changed, 5 insertions(+)

diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
index 843713c05b79..2eea81795139 100644
--- a/arch/mips/Kconfig
+++ b/arch/mips/Kconfig
@@ -2535,6 +2535,11 @@ config STACKTRACE_SUPPORT
bool
default y

+config PGTABLE_LEVELS
+ int
+ default 3 if 64BIT && !PAGE_SIZE_64KB
+ default 2
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
--
2.1.4

2015-02-06 14:56:49

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 09/19] mn10300: mark PUD and PMD folded

Core mm expects __PAGETABLE_{PUD,PMD}_FOLDED to be defined if these page
table levels folded.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: David Howells <[email protected]>
Cc: Koichi Yasutake <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/mn10300/include/asm/pgtable.h | 2 ++
1 file changed, 2 insertions(+)

diff --git a/arch/mn10300/include/asm/pgtable.h b/arch/mn10300/include/asm/pgtable.h
index afab728ab65e..96d3f9deb59c 100644
--- a/arch/mn10300/include/asm/pgtable.h
+++ b/arch/mn10300/include/asm/pgtable.h
@@ -56,7 +56,9 @@ extern void paging_init(void);
#define PGDIR_SHIFT 22
#define PTRS_PER_PGD 1024
#define PTRS_PER_PUD 1 /* we don't really have any PUD physically */
+#define __PAGETABLE_PUD_FOLDED
#define PTRS_PER_PMD 1 /* we don't really have any PMD physically */
+#define __PAGETABLE_PMD_FOLDED
#define PTRS_PER_PTE 1024

#define PGD_SIZE PAGE_SIZE
--
2.1.4

2015-02-06 14:51:53

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 10/19] parisc: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: "James E.J. Bottomley" <[email protected]>
Cc: Helge Deller <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---
arch/parisc/Kconfig | 5 +++++
arch/parisc/include/asm/pgalloc.h | 2 +-
arch/parisc/include/asm/pgtable.h | 17 ++++++++---------
arch/parisc/kernel/entry.S | 4 ++--
arch/parisc/kernel/head.S | 4 ++--
arch/parisc/mm/init.c | 2 +-
6 files changed, 19 insertions(+), 15 deletions(-)

diff --git a/arch/parisc/Kconfig b/arch/parisc/Kconfig
index 1554a6f2a5bb..1899d65b283c 100644
--- a/arch/parisc/Kconfig
+++ b/arch/parisc/Kconfig
@@ -103,6 +103,11 @@ config ARCH_MAY_HAVE_PC_FDC
depends on BROKEN
default y

+config PGTABLE_LEVELS
+ int
+ default 3 if 64BIT && PARISC_PAGE_SIZE_4KB
+ default 2
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
diff --git a/arch/parisc/include/asm/pgalloc.h b/arch/parisc/include/asm/pgalloc.h
index f213f5b4c423..55ad8be9b7f3 100644
--- a/arch/parisc/include/asm/pgalloc.h
+++ b/arch/parisc/include/asm/pgalloc.h
@@ -51,7 +51,7 @@ static inline void pgd_free(struct mm_struct *mm, pgd_t *pgd)
free_pages((unsigned long)pgd, PGD_ALLOC_ORDER);
}

-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3

/* Three Level Page Table Support for pmd's */

diff --git a/arch/parisc/include/asm/pgtable.h b/arch/parisc/include/asm/pgtable.h
index 8c966b2270aa..0a183756d6ec 100644
--- a/arch/parisc/include/asm/pgtable.h
+++ b/arch/parisc/include/asm/pgtable.h
@@ -68,13 +68,11 @@ extern void purge_tlb_entries(struct mm_struct *, unsigned long);
#define KERNEL_INITIAL_ORDER 24 /* 0 to 1<<24 = 16MB */
#define KERNEL_INITIAL_SIZE (1 << KERNEL_INITIAL_ORDER)

-#if defined(CONFIG_64BIT) && defined(CONFIG_PARISC_PAGE_SIZE_4KB)
-#define PT_NLEVELS 3
+#if CONFIG_PGTABLE_LEVELS == 3
#define PGD_ORDER 1 /* Number of pages per pgd */
#define PMD_ORDER 1 /* Number of pages per pmd */
#define PGD_ALLOC_ORDER 2 /* first pgd contains pmd */
#else
-#define PT_NLEVELS 2
#define PGD_ORDER 1 /* Number of pages per pgd */
#define PGD_ALLOC_ORDER PGD_ORDER
#endif
@@ -93,9 +91,10 @@ extern void purge_tlb_entries(struct mm_struct *, unsigned long);
#define PMD_SHIFT (PLD_SHIFT + BITS_PER_PTE)
#define PMD_SIZE (1UL << PMD_SHIFT)
#define PMD_MASK (~(PMD_SIZE-1))
-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
#define BITS_PER_PMD (PAGE_SHIFT + PMD_ORDER - BITS_PER_PMD_ENTRY)
#else
+#define __PAGETABLE_PMD_FOLDED
#define BITS_PER_PMD 0
#endif
#define PTRS_PER_PMD (1UL << BITS_PER_PMD)
@@ -276,7 +275,7 @@ extern unsigned long *empty_zero_page;
#define pgd_flag(x) (pgd_val(x) & PxD_FLAG_MASK)
#define pgd_address(x) ((unsigned long)(pgd_val(x) &~ PxD_FLAG_MASK) << PxD_VALUE_SHIFT)

-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
/* The first entry of the permanent pmd is not there if it contains
* the gateway marker */
#define pmd_none(x) (!pmd_val(x) || pmd_flag(x) == PxD_FLAG_ATTACHED)
@@ -286,7 +285,7 @@ extern unsigned long *empty_zero_page;
#define pmd_bad(x) (!(pmd_flag(x) & PxD_FLAG_VALID))
#define pmd_present(x) (pmd_flag(x) & PxD_FLAG_PRESENT)
static inline void pmd_clear(pmd_t *pmd) {
-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
if (pmd_flag(*pmd) & PxD_FLAG_ATTACHED)
/* This is the entry pointing to the permanent pmd
* attached to the pgd; cannot clear it */
@@ -298,7 +297,7 @@ static inline void pmd_clear(pmd_t *pmd) {



-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
#define pgd_page_vaddr(pgd) ((unsigned long) __va(pgd_address(pgd)))
#define pgd_page(pgd) virt_to_page((void *)pgd_page_vaddr(pgd))

@@ -308,7 +307,7 @@ static inline void pmd_clear(pmd_t *pmd) {
#define pgd_bad(x) (!(pgd_flag(x) & PxD_FLAG_VALID))
#define pgd_present(x) (pgd_flag(x) & PxD_FLAG_PRESENT)
static inline void pgd_clear(pgd_t *pgd) {
-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
if(pgd_flag(*pgd) & PxD_FLAG_ATTACHED)
/* This is the permanent pmd attached to the pgd; cannot
* free it */
@@ -392,7 +391,7 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)

/* Find an entry in the second-level page table.. */

-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
#define pmd_offset(dir,address) \
((pmd_t *) pgd_page_vaddr(*(dir)) + (((address)>>PMD_SHIFT) & (PTRS_PER_PMD-1)))
#else
diff --git a/arch/parisc/kernel/entry.S b/arch/parisc/kernel/entry.S
index e8f07dd28401..2cd5fb8798a8 100644
--- a/arch/parisc/kernel/entry.S
+++ b/arch/parisc/kernel/entry.S
@@ -398,7 +398,7 @@
* can address up to 1TB
*/
.macro L2_ptep pmd,pte,index,va,fault
-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
extru \va,31-ASM_PMD_SHIFT,ASM_BITS_PER_PMD,\index
#else
# if defined(CONFIG_64BIT)
@@ -436,7 +436,7 @@
* all ILP32 processes and all the kernel for machines with
* under 4GB of memory) */
.macro L3_ptep pgd,pte,index,va,fault
-#if PT_NLEVELS == 3 /* we might have a 2-Level scheme, e.g. with 16kb page size */
+#if CONFIG_PGTABLE_LEVELS == 3 /* we might have a 2-Level scheme, e.g. with 16kb page size */
extrd,u \va,63-ASM_PGDIR_SHIFT,ASM_BITS_PER_PGD,\index
copy %r0,\pte
extrd,u,*= \va,63-ASM_PGDIR_SHIFT,64-ASM_PGDIR_SHIFT,%r0
diff --git a/arch/parisc/kernel/head.S b/arch/parisc/kernel/head.S
index d4dc588c0dc1..e7d64527aff9 100644
--- a/arch/parisc/kernel/head.S
+++ b/arch/parisc/kernel/head.S
@@ -74,7 +74,7 @@ $bss_loop:
mtctl %r4,%cr24 /* Initialize kernel root pointer */
mtctl %r4,%cr25 /* Initialize user root pointer */

-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
/* Set pmd in pgd */
load32 PA(pmd0),%r5
shrd %r5,PxD_VALUE_SHIFT,%r3
@@ -97,7 +97,7 @@ $bss_loop:
stw %r3,0(%r4)
ldo (PAGE_SIZE >> PxD_VALUE_SHIFT)(%r3),%r3
addib,> -1,%r1,1b
-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
ldo ASM_PMD_ENTRY_SIZE(%r4),%r4
#else
ldo ASM_PGD_ENTRY_SIZE(%r4),%r4
diff --git a/arch/parisc/mm/init.c b/arch/parisc/mm/init.c
index 0bef864264c0..849226b0f21e 100644
--- a/arch/parisc/mm/init.c
+++ b/arch/parisc/mm/init.c
@@ -34,7 +34,7 @@
extern int data_start;
extern void parisc_kernel_start(void); /* Kernel entry point in head.S */

-#if PT_NLEVELS == 3
+#if CONFIG_PGTABLE_LEVELS == 3
/* NOTE: This layout exactly conforms to the hybrid L2/L3 page table layout
* with the first pmd adjacent to the pgd and below it. gcc doesn't actually
* guarantee that global objects will be laid out in memory in the same order
--
2.1.4

2015-02-06 14:51:51

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 11/19] powerpc: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Michael Ellerman <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/powerpc/Kconfig | 6 ++++++
1 file changed, 6 insertions(+)

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 22b0940494bb..91ad76f30d18 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
@@ -297,6 +297,12 @@ config ZONE_DMA32
bool
default y if PPC64

+config PGTABLE_LEVELS
+ int
+ default 2 if !PPC64
+ default 3 if PPC_64K_PAGES
+ default 4
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
--
2.1.4

2015-02-06 14:54:41

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 12/19] s390: expose number of page table levels

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Core mm expects __PAGETABLE_{PUD,PMD}_FOLDED to be defined if these page
table levels folded.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Martin Schwidefsky <[email protected]>
Cc: Heiko Carstens <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/s390/Kconfig | 5 +++++
arch/s390/include/asm/pgtable.h | 2 ++
2 files changed, 7 insertions(+)

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 373cd5badf1c..f6aebcb7a0f8 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -155,6 +155,11 @@ config S390
config SCHED_OMIT_FRAME_POINTER
def_bool y

+config PGTABLE_LEVELS
+ int
+ default 4 if 64BIT
+ default 2
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h
index fbb5ee3ae57c..e08ec38f8c6e 100644
--- a/arch/s390/include/asm/pgtable.h
+++ b/arch/s390/include/asm/pgtable.h
@@ -91,7 +91,9 @@ extern unsigned long zero_page_mask;
*/
#define PTRS_PER_PTE 256
#ifndef CONFIG_64BIT
+#define __PAGETABLE_PUD_FOLDED
#define PTRS_PER_PMD 1
+#define __PAGETABLE_PMD_FOLDED
#define PTRS_PER_PUD 1
#else /* CONFIG_64BIT */
#define PTRS_PER_PMD 2048
--
2.1.4

2015-02-06 14:56:48

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 13/19] sh: expose number of page table levels

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: [email protected]
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/sh/Kconfig | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/arch/sh/Kconfig b/arch/sh/Kconfig
index eb4ef274ae9b..50057fed819d 100644
--- a/arch/sh/Kconfig
+++ b/arch/sh/Kconfig
@@ -162,6 +162,10 @@ config NEED_DMA_MAP_STATE
config NEED_SG_DMA_LENGTH
def_bool y

+config PGTABLE_LEVELS
+ default 3 if X2TLB
+ default 2
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
--
2.1.4

2015-02-06 14:55:13

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 14/19] sparc: expose number of page table levels

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: "David S. Miller" <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/sparc/Kconfig | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/arch/sparc/Kconfig b/arch/sparc/Kconfig
index 96ac69c5eba0..cb06f5433e12 100644
--- a/arch/sparc/Kconfig
+++ b/arch/sparc/Kconfig
@@ -143,6 +143,10 @@ config GENERIC_ISA_DMA
config ARCH_SUPPORTS_DEBUG_PAGEALLOC
def_bool y if SPARC64

+config PGTABLE_LEVELS
+ default 4 if 64BIT
+ default 3
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
--
2.1.4

2015-02-06 14:55:14

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 15/19] tile: expose number of page table levels

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Acked-by: Chris Metcalf <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/tile/Kconfig | 5 +++++
1 file changed, 5 insertions(+)

diff --git a/arch/tile/Kconfig b/arch/tile/Kconfig
index 7cca41842a9e..0142d578b5a8 100644
--- a/arch/tile/Kconfig
+++ b/arch/tile/Kconfig
@@ -147,6 +147,11 @@ config ARCH_DEFCONFIG
default "arch/tile/configs/tilepro_defconfig" if !TILEGX
default "arch/tile/configs/tilegx_defconfig" if TILEGX

+config PGTABLE_LEVELS
+ int
+ default 3 if 64BIT
+ default 2
+
source "init/Kconfig"

source "kernel/Kconfig.freezer"
--
2.1.4

2015-02-06 14:51:25

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 16/19] um: expose number of page table levels

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Acked-by: Richard Weinberger <[email protected]>
Cc: Jeff Dike <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/um/Kconfig.um | 5 +++++
1 file changed, 5 insertions(+)

diff --git a/arch/um/Kconfig.um b/arch/um/Kconfig.um
index a7520c90f62d..5dbfe3d9107c 100644
--- a/arch/um/Kconfig.um
+++ b/arch/um/Kconfig.um
@@ -155,3 +155,8 @@ config MMAPPER

config NO_DMA
def_bool y
+
+config PGTABLE_LEVELS
+ int
+ default 3 if 3_LEVEL_PGTABLES
+ default 2
--
2.1.4

2015-02-06 14:51:23

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 17/19] x86: expose number of page table levels on Kconfig level

We would want to use number of page table level to define mm_struct.
Let's expose it as CONFIG_PGTABLE_LEVELS.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---

I've implemented accounting for pmd page tables as we have for pte (see
mm->nr_ptes). It's requires a new counter in mm_struct: mm->nr_pmds.

But the feature doesn't make any sense if an architecture has PMD level
folded and it would be nice get rid of the counter in this case.

The problem is that we cannot use __PAGETABLE_PMD_FOLDED in
<linux/mm_types.h> due to circular dependencies:

<linux/mm_types> -> <asm/pgtable.h> -> <linux/mm_types.h>

In most cases <asm/pgtable.h> wants <linux/mm_types.h> to get definition
of struct page and struct vm_area_struct. I've tried to split mm_struct
into separate header file to be able to user <asm/pgtable.h> there.

But it doesn't fly on some architectures, like ARM: it wants mm_struct
<asm/pgtable.h> to implement tlb flushing. I don't see how to fix it without
massive de-inlining or coverting a lot for inline functions to macros.

This is other approach: expose number of page tables in use via Kconfig
and use it in <linux/mm_types.h> instead of __PAGETABLE_PMD_FOLDED from
<asm/pgtable.h>.

---
arch/x86/Kconfig | 6 ++++++
arch/x86/include/asm/paravirt.h | 8 ++++----
arch/x86/include/asm/paravirt_types.h | 8 ++++----
arch/x86/include/asm/pgalloc.h | 8 ++++----
arch/x86/include/asm/pgtable-2level_types.h | 1 -
arch/x86/include/asm/pgtable-3level_types.h | 2 --
arch/x86/include/asm/pgtable.h | 8 ++++----
arch/x86/include/asm/pgtable_64_types.h | 1 -
arch/x86/include/asm/pgtable_types.h | 4 ++--
arch/x86/kernel/paravirt.c | 6 +++---
arch/x86/mm/pgtable.c | 14 +++++++-------
arch/x86/xen/mmu.c | 14 +++++++-------
include/trace/events/xen.h | 2 +-
13 files changed, 42 insertions(+), 40 deletions(-)

diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig
index e5c87b21658e..d4327c71d64f 100644
--- a/arch/x86/Kconfig
+++ b/arch/x86/Kconfig
@@ -279,6 +279,12 @@ config ARCH_SUPPORTS_UPROBES
config FIX_EARLYCON_MEM
def_bool y

+config PGTABLE_LEVELS
+ int
+ default 4 if X86_64
+ default 3 if X86_PAE
+ default 2
+
source "init/Kconfig"
source "kernel/Kconfig.freezer"

diff --git a/arch/x86/include/asm/paravirt.h b/arch/x86/include/asm/paravirt.h
index 32444ae939ca..7ced2aaab829 100644
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -545,7 +545,7 @@ static inline void set_pmd(pmd_t *pmdp, pmd_t pmd)
PVOP_VCALL2(pv_mmu_ops.set_pmd, pmdp, val);
}

-#if PAGETABLE_LEVELS >= 3
+#if CONFIG_PGTABLE_LEVELS >= 3
static inline pmd_t __pmd(pmdval_t val)
{
pmdval_t ret;
@@ -585,7 +585,7 @@ static inline void set_pud(pud_t *pudp, pud_t pud)
PVOP_VCALL2(pv_mmu_ops.set_pud, pudp,
val);
}
-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
static inline pud_t __pud(pudval_t val)
{
pudval_t ret;
@@ -636,9 +636,9 @@ static inline void pud_clear(pud_t *pudp)
set_pud(pudp, __pud(0));
}

-#endif /* PAGETABLE_LEVELS == 4 */
+#endif /* CONFIG_PGTABLE_LEVELS == 4 */

-#endif /* PAGETABLE_LEVELS >= 3 */
+#endif /* CONFIG_PGTABLE_LEVELS >= 3 */

#ifdef CONFIG_X86_PAE
/* Special-case pte-setting operations for PAE, which can't update a
diff --git a/arch/x86/include/asm/paravirt_types.h b/arch/x86/include/asm/paravirt_types.h
index 7549b8b369e4..f7b0b5c112f2 100644
--- a/arch/x86/include/asm/paravirt_types.h
+++ b/arch/x86/include/asm/paravirt_types.h
@@ -294,7 +294,7 @@ struct pv_mmu_ops {
struct paravirt_callee_save pgd_val;
struct paravirt_callee_save make_pgd;

-#if PAGETABLE_LEVELS >= 3
+#if CONFIG_PGTABLE_LEVELS >= 3
#ifdef CONFIG_X86_PAE
void (*set_pte_atomic)(pte_t *ptep, pte_t pteval);
void (*pte_clear)(struct mm_struct *mm, unsigned long addr,
@@ -308,13 +308,13 @@ struct pv_mmu_ops {
struct paravirt_callee_save pmd_val;
struct paravirt_callee_save make_pmd;

-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
struct paravirt_callee_save pud_val;
struct paravirt_callee_save make_pud;

void (*set_pgd)(pgd_t *pudp, pgd_t pgdval);
-#endif /* PAGETABLE_LEVELS == 4 */
-#endif /* PAGETABLE_LEVELS >= 3 */
+#endif /* CONFIG_PGTABLE_LEVELS == 4 */
+#endif /* CONFIG_PGTABLE_LEVELS >= 3 */

struct pv_lazy_ops lazy_mode;

diff --git a/arch/x86/include/asm/pgalloc.h b/arch/x86/include/asm/pgalloc.h
index c4412e972bbd..bf7f8b55b0f9 100644
--- a/arch/x86/include/asm/pgalloc.h
+++ b/arch/x86/include/asm/pgalloc.h
@@ -77,7 +77,7 @@ static inline void pmd_populate(struct mm_struct *mm, pmd_t *pmd,

#define pmd_pgtable(pmd) pmd_page(pmd)

-#if PAGETABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long addr)
{
struct page *page;
@@ -116,7 +116,7 @@ static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
}
#endif /* CONFIG_X86_PAE */

-#if PAGETABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
static inline void pgd_populate(struct mm_struct *mm, pgd_t *pgd, pud_t *pud)
{
paravirt_alloc_pud(mm, __pa(pud) >> PAGE_SHIFT);
@@ -142,7 +142,7 @@ static inline void __pud_free_tlb(struct mmu_gather *tlb, pud_t *pud,
___pud_free_tlb(tlb, pud);
}

-#endif /* PAGETABLE_LEVELS > 3 */
-#endif /* PAGETABLE_LEVELS > 2 */
+#endif /* CONFIG_PGTABLE_LEVELS > 3 */
+#endif /* CONFIG_PGTABLE_LEVELS > 2 */

#endif /* _ASM_X86_PGALLOC_H */
diff --git a/arch/x86/include/asm/pgtable-2level_types.h b/arch/x86/include/asm/pgtable-2level_types.h
index daacc23e3fb9..392576433e77 100644
--- a/arch/x86/include/asm/pgtable-2level_types.h
+++ b/arch/x86/include/asm/pgtable-2level_types.h
@@ -17,7 +17,6 @@ typedef union {
#endif /* !__ASSEMBLY__ */

#define SHARED_KERNEL_PMD 0
-#define PAGETABLE_LEVELS 2

/*
* traditional i386 two-level paging structure:
diff --git a/arch/x86/include/asm/pgtable-3level_types.h b/arch/x86/include/asm/pgtable-3level_types.h
index 1bd5876c8649..bcc89625ebe5 100644
--- a/arch/x86/include/asm/pgtable-3level_types.h
+++ b/arch/x86/include/asm/pgtable-3level_types.h
@@ -24,8 +24,6 @@ typedef union {
#define SHARED_KERNEL_PMD 1
#endif

-#define PAGETABLE_LEVELS 3
-
/*
* PGDIR_SHIFT determines what a top-level page table entry can map
*/
diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index 9d0ade00923e..f89d6c9943ea 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -554,7 +554,7 @@ static inline unsigned long pages_to_mb(unsigned long npg)
return npg >> (20 - PAGE_SHIFT);
}

-#if PAGETABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
static inline int pud_none(pud_t pud)
{
return native_pud_val(pud) == 0;
@@ -597,9 +597,9 @@ static inline int pud_large(pud_t pud)
{
return 0;
}
-#endif /* PAGETABLE_LEVELS > 2 */
+#endif /* CONFIG_PGTABLE_LEVELS > 2 */

-#if PAGETABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
static inline int pgd_present(pgd_t pgd)
{
return pgd_flags(pgd) & _PAGE_PRESENT;
@@ -636,7 +636,7 @@ static inline int pgd_none(pgd_t pgd)
{
return !native_pgd_val(pgd);
}
-#endif /* PAGETABLE_LEVELS > 3 */
+#endif /* CONFIG_PGTABLE_LEVELS > 3 */

#endif /* __ASSEMBLY__ */

diff --git a/arch/x86/include/asm/pgtable_64_types.h b/arch/x86/include/asm/pgtable_64_types.h
index 602b6028c5b6..e6844dfb4471 100644
--- a/arch/x86/include/asm/pgtable_64_types.h
+++ b/arch/x86/include/asm/pgtable_64_types.h
@@ -20,7 +20,6 @@ typedef struct { pteval_t pte; } pte_t;
#endif /* !__ASSEMBLY__ */

#define SHARED_KERNEL_PMD 0
-#define PAGETABLE_LEVELS 4

/*
* PGDIR_SHIFT determines what a top-level page table entry can map
diff --git a/arch/x86/include/asm/pgtable_types.h b/arch/x86/include/asm/pgtable_types.h
index 8c7c10802e9c..78f0c8cbe316 100644
--- a/arch/x86/include/asm/pgtable_types.h
+++ b/arch/x86/include/asm/pgtable_types.h
@@ -234,7 +234,7 @@ static inline pgdval_t pgd_flags(pgd_t pgd)
return native_pgd_val(pgd) & PTE_FLAGS_MASK;
}

-#if PAGETABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
typedef struct { pudval_t pud; } pud_t;

static inline pud_t native_make_pud(pmdval_t val)
@@ -255,7 +255,7 @@ static inline pudval_t native_pud_val(pud_t pud)
}
#endif

-#if PAGETABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
typedef struct { pmdval_t pmd; } pmd_t;

static inline pmd_t native_make_pmd(pmdval_t val)
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index 548d25f00c90..c614dd492f5f 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -443,7 +443,7 @@ struct pv_mmu_ops pv_mmu_ops = {
.ptep_modify_prot_start = __ptep_modify_prot_start,
.ptep_modify_prot_commit = __ptep_modify_prot_commit,

-#if PAGETABLE_LEVELS >= 3
+#if CONFIG_PGTABLE_LEVELS >= 3
#ifdef CONFIG_X86_PAE
.set_pte_atomic = native_set_pte_atomic,
.pte_clear = native_pte_clear,
@@ -454,13 +454,13 @@ struct pv_mmu_ops pv_mmu_ops = {
.pmd_val = PTE_IDENT,
.make_pmd = PTE_IDENT,

-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
.pud_val = PTE_IDENT,
.make_pud = PTE_IDENT,

.set_pgd = native_set_pgd,
#endif
-#endif /* PAGETABLE_LEVELS >= 3 */
+#endif /* CONFIG_PGTABLE_LEVELS >= 3 */

.pte_val = PTE_IDENT,
.pgd_val = PTE_IDENT,
diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c
index 7b22adaad4f1..9885592b9fa7 100644
--- a/arch/x86/mm/pgtable.c
+++ b/arch/x86/mm/pgtable.c
@@ -58,7 +58,7 @@ void ___pte_free_tlb(struct mmu_gather *tlb, struct page *pte)
tlb_remove_page(tlb, pte);
}

-#if PAGETABLE_LEVELS > 2
+#if CONFIG_PGTABLE_LEVELS > 2
void ___pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd)
{
struct page *page = virt_to_page(pmd);
@@ -74,14 +74,14 @@ void ___pmd_free_tlb(struct mmu_gather *tlb, pmd_t *pmd)
tlb_remove_page(tlb, page);
}

-#if PAGETABLE_LEVELS > 3
+#if CONFIG_PGTABLE_LEVELS > 3
void ___pud_free_tlb(struct mmu_gather *tlb, pud_t *pud)
{
paravirt_release_pud(__pa(pud) >> PAGE_SHIFT);
tlb_remove_page(tlb, virt_to_page(pud));
}
-#endif /* PAGETABLE_LEVELS > 3 */
-#endif /* PAGETABLE_LEVELS > 2 */
+#endif /* CONFIG_PGTABLE_LEVELS > 3 */
+#endif /* CONFIG_PGTABLE_LEVELS > 2 */

static inline void pgd_list_add(pgd_t *pgd)
{
@@ -117,9 +117,9 @@ static void pgd_ctor(struct mm_struct *mm, pgd_t *pgd)
/* If the pgd points to a shared pagetable level (either the
ptes in non-PAE, or shared PMD in PAE), then just copy the
references from swapper_pg_dir. */
- if (PAGETABLE_LEVELS == 2 ||
- (PAGETABLE_LEVELS == 3 && SHARED_KERNEL_PMD) ||
- PAGETABLE_LEVELS == 4) {
+ if (CONFIG_PGTABLE_LEVELS == 2 ||
+ (CONFIG_PGTABLE_LEVELS == 3 && SHARED_KERNEL_PMD) ||
+ CONFIG_PGTABLE_LEVELS == 4) {
clone_pgd_range(pgd + KERNEL_PGD_BOUNDARY,
swapper_pg_dir + KERNEL_PGD_BOUNDARY,
KERNEL_PGD_PTRS);
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c
index adca9e2b6553..65083ad63b6f 100644
--- a/arch/x86/xen/mmu.c
+++ b/arch/x86/xen/mmu.c
@@ -502,7 +502,7 @@ __visible pmd_t xen_make_pmd(pmdval_t pmd)
}
PV_CALLEE_SAVE_REGS_THUNK(xen_make_pmd);

-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
__visible pudval_t xen_pud_val(pud_t pud)
{
return pte_mfn_to_pfn(pud.pud);
@@ -589,7 +589,7 @@ static void xen_set_pgd(pgd_t *ptr, pgd_t val)

xen_mc_issue(PARAVIRT_LAZY_MMU);
}
-#endif /* PAGETABLE_LEVELS == 4 */
+#endif /* CONFIG_PGTABLE_LEVELS == 4 */

/*
* (Yet another) pagetable walker. This one is intended for pinning a
@@ -1628,7 +1628,7 @@ static void xen_release_pmd(unsigned long pfn)
xen_release_ptpage(pfn, PT_PMD);
}

-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
static void xen_alloc_pud(struct mm_struct *mm, unsigned long pfn)
{
xen_alloc_ptpage(mm, pfn, PT_PUD);
@@ -2046,7 +2046,7 @@ static void __init xen_post_allocator_init(void)
pv_mmu_ops.set_pte = xen_set_pte;
pv_mmu_ops.set_pmd = xen_set_pmd;
pv_mmu_ops.set_pud = xen_set_pud;
-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
pv_mmu_ops.set_pgd = xen_set_pgd;
#endif

@@ -2056,7 +2056,7 @@ static void __init xen_post_allocator_init(void)
pv_mmu_ops.alloc_pmd = xen_alloc_pmd;
pv_mmu_ops.release_pte = xen_release_pte;
pv_mmu_ops.release_pmd = xen_release_pmd;
-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
pv_mmu_ops.alloc_pud = xen_alloc_pud;
pv_mmu_ops.release_pud = xen_release_pud;
#endif
@@ -2122,14 +2122,14 @@ static const struct pv_mmu_ops xen_mmu_ops __initconst = {
.make_pmd = PV_CALLEE_SAVE(xen_make_pmd),
.pmd_val = PV_CALLEE_SAVE(xen_pmd_val),

-#if PAGETABLE_LEVELS == 4
+#if CONFIG_PGTABLE_LEVELS == 4
.pud_val = PV_CALLEE_SAVE(xen_pud_val),
.make_pud = PV_CALLEE_SAVE(xen_make_pud),
.set_pgd = xen_set_pgd_hyper,

.alloc_pud = xen_alloc_pmd_init,
.release_pud = xen_release_pmd_init,
-#endif /* PAGETABLE_LEVELS == 4 */
+#endif /* CONFIG_PGTABLE_LEVELS == 4 */

.activate_mm = xen_activate_mm,
.dup_mmap = xen_dup_mmap,
diff --git a/include/trace/events/xen.h b/include/trace/events/xen.h
index d06b6da5c1e3..bce990f5a35d 100644
--- a/include/trace/events/xen.h
+++ b/include/trace/events/xen.h
@@ -224,7 +224,7 @@ TRACE_EVENT(xen_mmu_pmd_clear,
TP_printk("pmdp %p", __entry->pmdp)
);

-#if PAGETABLE_LEVELS >= 4
+#if CONFIG_PGTABLE_LEVELS >= 4

TRACE_EVENT(xen_mmu_set_pud,
TP_PROTO(pud_t *pudp, pud_t pudval),
--
2.1.4

2015-02-06 14:54:39

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 18/19] mm: define default PGTABLE_LEVELS to two

By this time all architectures which support more than two page table
levels should be covered. This patch add default definiton of
PGTABLE_LEVELS equal 2.

We also add assert to detect inconsistence between CONFIG_PGTABLE_LEVELS
and __PAGETABLE_PMD_FOLDED/__PAGETABLE_PUD_FOLDED.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---
arch/Kconfig | 4 ++++
include/asm-generic/pgtable.h | 5 +++++
2 files changed, 9 insertions(+)

diff --git a/arch/Kconfig b/arch/Kconfig
index 05d7a8a458d5..a9c95d36ba70 100644
--- a/arch/Kconfig
+++ b/arch/Kconfig
@@ -484,6 +484,10 @@ config HAVE_IRQ_EXIT_ON_IRQ_STACK
This spares a stack switch and improves cache usage on softirq
processing.

+config PGTABLE_LEVELS
+ int
+ default 2
+
#
# ABI hall of shame
#
diff --git a/include/asm-generic/pgtable.h b/include/asm-generic/pgtable.h
index 4d46085c1b90..1f9f5da6828f 100644
--- a/include/asm-generic/pgtable.h
+++ b/include/asm-generic/pgtable.h
@@ -7,6 +7,11 @@
#include <linux/mm_types.h>
#include <linux/bug.h>

+#if 4 - defined(__PAGETABLE_PUD_FOLDED) - defined(__PAGETABLE_PMD_FOLDED) != \
+ CONFIG_PGTABLE_LEVELS
+#error CONFIG_PGTABLE_LEVELS is not consistent with __PAGETABLE_{PUD,PMD}_FOLDED
+#endif
+
/*
* On almost all architectures and configurations, 0 can be used as the
* upper ceiling to free_pgtables(): on many architectures it has the same
--
2.1.4

2015-02-06 14:51:21

by Kirill A. Shutemov

[permalink] [raw]
Subject: [PATCHv2, RESEND 19/19] mm: do not add nr_pmds into mm_struct if PMD is folded

CONFIG_PGTABLE_LEVELS is now available on every architecture and we can
use it to check if we need to add nr_pmds into mm_struct.

Signed-off-by: Kirill A. Shutemov <[email protected]>
Tested-by: Guenter Roeck <[email protected]>
---
include/linux/mm_types.h | 2 ++
1 file changed, 2 insertions(+)

diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 199a03aab8dc..590630eb59ba 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -364,7 +364,9 @@ struct mm_struct {
atomic_t mm_users; /* How many users with user space? */
atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */
atomic_long_t nr_ptes; /* PTE page table pages */
+#if CONFIG_PGTABLE_LEVELS > 2
atomic_long_t nr_pmds; /* PMD page table pages */
+#endif
int map_count; /* number of VMAs */

spinlock_t page_table_lock; /* Protects page tables and some counters */
--
2.1.4

2015-02-06 16:26:28

by Richard Henderson

[permalink] [raw]
Subject: Re: [PATCHv2, RESEND 01/19] alpha: expose number of page table levels on Kconfig level

On 02/06/2015 06:50 AM, Kirill A. Shutemov wrote:
> We would want to use number of page table level to define mm_struct.
> Let's expose it as CONFIG_PGTABLE_LEVELS.
>
> Signed-off-by: Kirill A. Shutemov <[email protected]>
> Cc: Richard Henderson <[email protected]>
> Cc: Ivan Kokshaysky <[email protected]>
> Cc: Matt Turner <[email protected]>
> Tested-by: Guenter Roeck <[email protected]>

Acked-by: Richard Henderson <[email protected]>


r~