* [PATCH 1/3] mm: define generic pudp_invalidate() implementation
2026-09-09 9:00 [PATCH 0/3] arm64: Add support for PUD THP Jinjiang Tu
@ 2026-09-09 9:00 ` Jinjiang Tu
2026-09-09 10:24 ` David Hildenbrand (Arm)
2026-09-09 9:00 ` [PATCH 2/3] arm64: simplify pmd_special and pmd_mkspecial Jinjiang Tu
2026-09-09 9:00 ` [PATCH 3/3] arm64: Add support for PUD THP Jinjiang Tu
2 siblings, 1 reply; 6+ messages in thread
From: Jinjiang Tu @ 2026-09-09 9:00 UTC (permalink / raw)
To: akpm, david, ljs, catalin.marinas, will, mark.rutland,
ryan.roberts, peterx, dave.jiang, linux-arm-kernel, linux-mm
Cc: wangkefeng.wang, sunnanyong, tujinjiang
x86 and riscv have identical pudp_invalidate() implementations. Move it to
mm/pgtable-generic.c with the __HAVE_ARCH_PUDP_INVALIDATE override guard.
PowerPC keeps its own implementation. In the later patch, arm64 will use
the generic implementation too.
No functional change intended.
Signed-off-by: Jinjiang Tu <tujinjiang@huawei.com>
---
arch/powerpc/include/asm/book3s/64/pgtable.h | 1 +
arch/riscv/include/asm/pgtable.h | 3 ---
arch/riscv/mm/pgtable.c | 9 ---------
arch/x86/include/asm/pgtable.h | 3 ---
arch/x86/mm/pgtable.c | 12 ------------
include/linux/pgtable.h | 7 +++++++
mm/pgtable-generic.c | 14 ++++++++++++++
7 files changed, 22 insertions(+), 27 deletions(-)
diff --git a/arch/powerpc/include/asm/book3s/64/pgtable.h b/arch/powerpc/include/asm/book3s/64/pgtable.h
index f4db7d7fbd5c..55d33c5537f1 100644
--- a/arch/powerpc/include/asm/book3s/64/pgtable.h
+++ b/arch/powerpc/include/asm/book3s/64/pgtable.h
@@ -1419,6 +1419,7 @@ static inline pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm,
#define __HAVE_ARCH_PMDP_INVALIDATE
extern pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address,
pmd_t *pmdp);
+#define __HAVE_ARCH_PUDP_INVALIDATE
extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
pud_t *pudp);
diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
index 40b1ed4f3ea8..327fa378032b 100644
--- a/arch/riscv/include/asm/pgtable.h
+++ b/arch/riscv/include/asm/pgtable.h
@@ -1159,9 +1159,6 @@ static inline pud_t pud_mkinvalid(pud_t pud)
return __pud(pud_val(pud) & ~(_PAGE_PRESENT | _PAGE_PROT_NONE));
}
-extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
- pud_t *pudp);
-
static inline pud_t pud_modify(pud_t pud, pgprot_t newprot)
{
return pte_pud(pte_modify(pud_pte(pud), newprot));
diff --git a/arch/riscv/mm/pgtable.c b/arch/riscv/mm/pgtable.c
index 9c4427d0b187..156e385e7db4 100644
--- a/arch/riscv/mm/pgtable.c
+++ b/arch/riscv/mm/pgtable.c
@@ -154,15 +154,6 @@ pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
return pmd;
}
-pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
- pud_t *pudp)
-{
- VM_WARN_ON_ONCE(!pud_present(*pudp));
- pud_t old = pudp_establish(vma, address, pudp, pud_mkinvalid(*pudp));
-
- flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
- return old;
-}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
pte_t pte_mkwrite(pte_t pte, struct vm_area_struct *vma)
diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
index d5f4917c1edc..97051c02385e 100644
--- a/arch/x86/include/asm/pgtable.h
+++ b/arch/x86/include/asm/pgtable.h
@@ -1387,9 +1387,6 @@ static inline pud_t pudp_establish(struct vm_area_struct *vma,
extern pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma,
unsigned long address, pmd_t *pmdp);
-pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
- pud_t *pudp);
-
/*
* Page table pages are page-aligned. The lower half of the top
* level is used for userspace and the top half for the kernel.
diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c
index cb03f5a2b243..0945908631dc 100644
--- a/arch/x86/mm/pgtable.c
+++ b/arch/x86/mm/pgtable.c
@@ -519,18 +519,6 @@ pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address,
}
#endif
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && \
- defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
-pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
- pud_t *pudp)
-{
- VM_WARN_ON_ONCE(!pud_present(*pudp));
- pud_t old = pudp_establish(vma, address, pudp, pud_mkinvalid(*pudp));
- flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
- return old;
-}
-#endif
-
/**
* reserve_top_address - Reserve a hole in the top of the kernel address space
* @reserve: Size of hole to reserve
diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
index 8c093c119e5a..32173005b087 100644
--- a/include/linux/pgtable.h
+++ b/include/linux/pgtable.h
@@ -1318,6 +1318,13 @@ extern pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma,
unsigned long address, pmd_t *pmdp);
#endif
+#ifndef __HAVE_ARCH_PUDP_INVALIDATE
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
+ pud_t *pudp);
+#endif
+#endif
+
#ifndef __HAVE_ARCH_PTE_SAME
static inline int pte_same(pte_t pte_a, pte_t pte_b)
{
diff --git a/mm/pgtable-generic.c b/mm/pgtable-generic.c
index b91b1a98029c..ac1175d9a59a 100644
--- a/mm/pgtable-generic.c
+++ b/mm/pgtable-generic.c
@@ -217,6 +217,20 @@ pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address,
}
#endif
+#ifndef __HAVE_ARCH_PUDP_INVALIDATE
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
+ pud_t *pudp)
+{
+ VM_WARN_ON_ONCE(!pud_present(*pudp));
+ pud_t old = pudp_establish(vma, address, pudp, pud_mkinvalid(*pudp));
+
+ flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
+ return old;
+}
+#endif
+#endif
+
#ifndef pmdp_collapse_flush
pmd_t pmdp_collapse_flush(struct vm_area_struct *vma, unsigned long address,
pmd_t *pmdp)
--
2.43.0
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH 1/3] mm: define generic pudp_invalidate() implementation
2026-09-09 9:00 ` [PATCH 1/3] mm: define generic pudp_invalidate() implementation Jinjiang Tu
@ 2026-09-09 10:24 ` David Hildenbrand (Arm)
2026-09-11 2:28 ` Jinjiang Tu
0 siblings, 1 reply; 6+ messages in thread
From: David Hildenbrand (Arm) @ 2026-09-09 10:24 UTC (permalink / raw)
To: Jinjiang Tu, akpm, ljs, catalin.marinas, will, mark.rutland,
ryan.roberts, peterx, dave.jiang, linux-arm-kernel, linux-mm
Cc: wangkefeng.wang, sunnanyong
On 9/9/26 11:00, Jinjiang Tu wrote:
> x86 and riscv have identical pudp_invalidate() implementations. Move it to
> mm/pgtable-generic.c with the __HAVE_ARCH_PUDP_INVALIDATE override guard.
> PowerPC keeps its own implementation. In the later patch, arm64 will use
> the generic implementation too.
>
> No functional change intended.
>
> Signed-off-by: Jinjiang Tu <tujinjiang@huawei.com>
> ---
> arch/powerpc/include/asm/book3s/64/pgtable.h | 1 +
> arch/riscv/include/asm/pgtable.h | 3 ---
> arch/riscv/mm/pgtable.c | 9 ---------
> arch/x86/include/asm/pgtable.h | 3 ---
> arch/x86/mm/pgtable.c | 12 ------------
> include/linux/pgtable.h | 7 +++++++
> mm/pgtable-generic.c | 14 ++++++++++++++
> 7 files changed, 22 insertions(+), 27 deletions(-)
>
> diff --git a/arch/powerpc/include/asm/book3s/64/pgtable.h b/arch/powerpc/include/asm/book3s/64/pgtable.h
> index f4db7d7fbd5c..55d33c5537f1 100644
> --- a/arch/powerpc/include/asm/book3s/64/pgtable.h
> +++ b/arch/powerpc/include/asm/book3s/64/pgtable.h
> @@ -1419,6 +1419,7 @@ static inline pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm,
> #define __HAVE_ARCH_PMDP_INVALIDATE
> extern pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address,
> pmd_t *pmdp);
> +#define __HAVE_ARCH_PUDP_INVALIDATE
> extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
> pud_t *pudp);
>
> diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
> index 40b1ed4f3ea8..327fa378032b 100644
> --- a/arch/riscv/include/asm/pgtable.h
> +++ b/arch/riscv/include/asm/pgtable.h
> @@ -1159,9 +1159,6 @@ static inline pud_t pud_mkinvalid(pud_t pud)
> return __pud(pud_val(pud) & ~(_PAGE_PRESENT | _PAGE_PROT_NONE));
> }
>
> -extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
> - pud_t *pudp);
> -
> static inline pud_t pud_modify(pud_t pud, pgprot_t newprot)
> {
> return pte_pud(pte_modify(pud_pte(pud), newprot));
> diff --git a/arch/riscv/mm/pgtable.c b/arch/riscv/mm/pgtable.c
> index 9c4427d0b187..156e385e7db4 100644
> --- a/arch/riscv/mm/pgtable.c
> +++ b/arch/riscv/mm/pgtable.c
> @@ -154,15 +154,6 @@ pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
> return pmd;
> }
>
> -pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
> - pud_t *pudp)
> -{
> - VM_WARN_ON_ONCE(!pud_present(*pudp));
> - pud_t old = pudp_establish(vma, address, pudp, pud_mkinvalid(*pudp));
> -
> - flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
> - return old;
> -}
> #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>
> pte_t pte_mkwrite(pte_t pte, struct vm_area_struct *vma)
> diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
> index d5f4917c1edc..97051c02385e 100644
> --- a/arch/x86/include/asm/pgtable.h
> +++ b/arch/x86/include/asm/pgtable.h
> @@ -1387,9 +1387,6 @@ static inline pud_t pudp_establish(struct vm_area_struct *vma,
> extern pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma,
> unsigned long address, pmd_t *pmdp);
>
> -pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
> - pud_t *pudp);
> -
> /*
> * Page table pages are page-aligned. The lower half of the top
> * level is used for userspace and the top half for the kernel.
> diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c
> index cb03f5a2b243..0945908631dc 100644
> --- a/arch/x86/mm/pgtable.c
> +++ b/arch/x86/mm/pgtable.c
> @@ -519,18 +519,6 @@ pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address,
> }
> #endif
>
> -#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && \
> - defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
> -pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
> - pud_t *pudp)
> -{
> - VM_WARN_ON_ONCE(!pud_present(*pudp));
> - pud_t old = pudp_establish(vma, address, pudp, pud_mkinvalid(*pudp));
> - flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
> - return old;
> -}
> -#endif
> -
> /**
> * reserve_top_address - Reserve a hole in the top of the kernel address space
> * @reserve: Size of hole to reserve
> diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
> index 8c093c119e5a..32173005b087 100644
> --- a/include/linux/pgtable.h
> +++ b/include/linux/pgtable.h
> @@ -1318,6 +1318,13 @@ extern pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma,
> unsigned long address, pmd_t *pmdp);
> #endif
>
> +#ifndef __HAVE_ARCH_PUDP_INVALIDATE
> +#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
> +extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
> + pud_t *pudp);
No "extern" please.
Staring at this file, I do wonder why we are not smarter about combining blocks. E.g., here simply:
diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
index e3c8ab96941c5..8c9fe3bf580d4 100644
--- a/include/linux/pgtable.h
+++ b/include/linux/pgtable.h
@@ -1229,9 +1229,9 @@ static inline void pmdp_set_wrprotect(struct mm_struct *mm,
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
#endif
-#ifndef __HAVE_ARCH_PUDP_SET_WRPROTECT
-#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+#ifndef __HAVE_ARCH_PUDP_SET_WRPROTECT
static inline void pudp_set_wrprotect(struct mm_struct *mm,
unsigned long address, pud_t *pudp)
{
@@ -1245,9 +1245,13 @@ static inline void pudp_set_wrprotect(struct mm_struct *mm,
{
BUILD_BUG();
}
-#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
-#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
#endif
+#ifndef __HAVE_ARCH_PUDP_INVALIDATE
+pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
+ pud_t *pudp);
+#endif
+#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
+#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
> +#endif
> +#endif
> +
> #ifndef __HAVE_ARCH_PTE_SAME
> static inline int pte_same(pte_t pte_a, pte_t pte_b)
> {
> diff --git a/mm/pgtable-generic.c b/mm/pgtable-generic.c
> index b91b1a98029c..ac1175d9a59a 100644
> --- a/mm/pgtable-generic.c
> +++ b/mm/pgtable-generic.c
> @@ -217,6 +217,20 @@ pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address,
> }
> #endif
>
> +#ifndef __HAVE_ARCH_PUDP_INVALIDATE
> +#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
I'd revert these and similarly see if they can be combined somehow (likely combining with
pudp_huge_clear_flush is harder as we are quite inconsistent with it, gluing it to
__HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH).
Which raises the question whether we could also piggy-back on __HAVE_ARCH_PMDP_SET_WRPROTECT
(I suspect no without a closer look.)
--
Cheers,
David
^ permalink raw reply related [flat|nested] 6+ messages in thread* Re: [PATCH 1/3] mm: define generic pudp_invalidate() implementation
2026-09-09 10:24 ` David Hildenbrand (Arm)
@ 2026-09-11 2:28 ` Jinjiang Tu
0 siblings, 0 replies; 6+ messages in thread
From: Jinjiang Tu @ 2026-09-11 2:28 UTC (permalink / raw)
To: David Hildenbrand (Arm), akpm, ljs, catalin.marinas, will,
mark.rutland, ryan.roberts, peterx, dave.jiang, linux-arm-kernel,
linux-mm
Cc: wangkefeng.wang, sunnanyong
在 2026/9/9 18:24, David Hildenbrand (Arm) 写道:
> On 9/9/26 11:00, Jinjiang Tu wrote:
>> x86 and riscv have identical pudp_invalidate() implementations. Move it to
>> mm/pgtable-generic.c with the __HAVE_ARCH_PUDP_INVALIDATE override guard.
>> PowerPC keeps its own implementation. In the later patch, arm64 will use
>> the generic implementation too.
>>
>> No functional change intended.
>>
>> Signed-off-by: Jinjiang Tu <tujinjiang@huawei.com>
>> ---
>> arch/powerpc/include/asm/book3s/64/pgtable.h | 1 +
>> arch/riscv/include/asm/pgtable.h | 3 ---
>> arch/riscv/mm/pgtable.c | 9 ---------
>> arch/x86/include/asm/pgtable.h | 3 ---
>> arch/x86/mm/pgtable.c | 12 ------------
>> include/linux/pgtable.h | 7 +++++++
>> mm/pgtable-generic.c | 14 ++++++++++++++
>> 7 files changed, 22 insertions(+), 27 deletions(-)
>>
>> diff --git a/arch/powerpc/include/asm/book3s/64/pgtable.h b/arch/powerpc/include/asm/book3s/64/pgtable.h
>> index f4db7d7fbd5c..55d33c5537f1 100644
>> --- a/arch/powerpc/include/asm/book3s/64/pgtable.h
>> +++ b/arch/powerpc/include/asm/book3s/64/pgtable.h
>> @@ -1419,6 +1419,7 @@ static inline pgtable_t pgtable_trans_huge_withdraw(struct mm_struct *mm,
>> #define __HAVE_ARCH_PMDP_INVALIDATE
>> extern pmd_t pmdp_invalidate(struct vm_area_struct *vma, unsigned long address,
>> pmd_t *pmdp);
>> +#define __HAVE_ARCH_PUDP_INVALIDATE
>> extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
>> pud_t *pudp);
>>
>> diff --git a/arch/riscv/include/asm/pgtable.h b/arch/riscv/include/asm/pgtable.h
>> index 40b1ed4f3ea8..327fa378032b 100644
>> --- a/arch/riscv/include/asm/pgtable.h
>> +++ b/arch/riscv/include/asm/pgtable.h
>> @@ -1159,9 +1159,6 @@ static inline pud_t pud_mkinvalid(pud_t pud)
>> return __pud(pud_val(pud) & ~(_PAGE_PRESENT | _PAGE_PROT_NONE));
>> }
>>
>> -extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
>> - pud_t *pudp);
>> -
>> static inline pud_t pud_modify(pud_t pud, pgprot_t newprot)
>> {
>> return pte_pud(pte_modify(pud_pte(pud), newprot));
>> diff --git a/arch/riscv/mm/pgtable.c b/arch/riscv/mm/pgtable.c
>> index 9c4427d0b187..156e385e7db4 100644
>> --- a/arch/riscv/mm/pgtable.c
>> +++ b/arch/riscv/mm/pgtable.c
>> @@ -154,15 +154,6 @@ pmd_t pmdp_collapse_flush(struct vm_area_struct *vma,
>> return pmd;
>> }
>>
>> -pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
>> - pud_t *pudp)
>> -{
>> - VM_WARN_ON_ONCE(!pud_present(*pudp));
>> - pud_t old = pudp_establish(vma, address, pudp, pud_mkinvalid(*pudp));
>> -
>> - flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
>> - return old;
>> -}
>> #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>>
>> pte_t pte_mkwrite(pte_t pte, struct vm_area_struct *vma)
>> diff --git a/arch/x86/include/asm/pgtable.h b/arch/x86/include/asm/pgtable.h
>> index d5f4917c1edc..97051c02385e 100644
>> --- a/arch/x86/include/asm/pgtable.h
>> +++ b/arch/x86/include/asm/pgtable.h
>> @@ -1387,9 +1387,6 @@ static inline pud_t pudp_establish(struct vm_area_struct *vma,
>> extern pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma,
>> unsigned long address, pmd_t *pmdp);
>>
>> -pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
>> - pud_t *pudp);
>> -
>> /*
>> * Page table pages are page-aligned. The lower half of the top
>> * level is used for userspace and the top half for the kernel.
>> diff --git a/arch/x86/mm/pgtable.c b/arch/x86/mm/pgtable.c
>> index cb03f5a2b243..0945908631dc 100644
>> --- a/arch/x86/mm/pgtable.c
>> +++ b/arch/x86/mm/pgtable.c
>> @@ -519,18 +519,6 @@ pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address,
>> }
>> #endif
>>
>> -#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && \
>> - defined(CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD)
>> -pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
>> - pud_t *pudp)
>> -{
>> - VM_WARN_ON_ONCE(!pud_present(*pudp));
>> - pud_t old = pudp_establish(vma, address, pudp, pud_mkinvalid(*pudp));
>> - flush_pud_tlb_range(vma, address, address + HPAGE_PUD_SIZE);
>> - return old;
>> -}
>> -#endif
>> -
>> /**
>> * reserve_top_address - Reserve a hole in the top of the kernel address space
>> * @reserve: Size of hole to reserve
>> diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
>> index 8c093c119e5a..32173005b087 100644
>> --- a/include/linux/pgtable.h
>> +++ b/include/linux/pgtable.h
>> @@ -1318,6 +1318,13 @@ extern pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma,
>> unsigned long address, pmd_t *pmdp);
>> #endif
>>
>> +#ifndef __HAVE_ARCH_PUDP_INVALIDATE
>> +#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
>> +extern pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
>> + pud_t *pudp);
> No "extern" please.
Will update it.
>
> Staring at this file, I do wonder why we are not smarter about combining blocks. E.g., here simply:'
Do you mean merging all #ifdef CONFIG_TRANSPARENT_HUGEPAGE blocks? There are many of them scattered
across include/linux/pgtable.h and arch/arm64/include/asm/pgtable.h. They are intentionally kept
separate to group the PTE, PMD, and PUD level variants of the same operation together. For example,
ptep_set_access_flags, pmdp_set_access_flags, and pudp_set_access_flags are placed adjacently.
>
> diff --git a/include/linux/pgtable.h b/include/linux/pgtable.h
> index e3c8ab96941c5..8c9fe3bf580d4 100644
> --- a/include/linux/pgtable.h
> +++ b/include/linux/pgtable.h
> @@ -1229,9 +1229,9 @@ static inline void pmdp_set_wrprotect(struct mm_struct *mm,
> }
> #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
> #endif
> -#ifndef __HAVE_ARCH_PUDP_SET_WRPROTECT
> -#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
> #ifdef CONFIG_TRANSPARENT_HUGEPAGE
> +#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
> +#ifndef __HAVE_ARCH_PUDP_SET_WRPROTECT
> static inline void pudp_set_wrprotect(struct mm_struct *mm,
> unsigned long address, pud_t *pudp)
> {
> @@ -1245,9 +1245,13 @@ static inline void pudp_set_wrprotect(struct mm_struct *mm,
> {
> BUILD_BUG();
> }
> -#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
> -#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
> #endif
> +#ifndef __HAVE_ARCH_PUDP_INVALIDATE
> +pud_t pudp_invalidate(struct vm_area_struct *vma, unsigned long address,
> + pud_t *pudp);
> +#endif
> +#endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
> +#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>
>> +#endif
>> +#endif
>> +
>> #ifndef __HAVE_ARCH_PTE_SAME
>> static inline int pte_same(pte_t pte_a, pte_t pte_b)
>> {
>> diff --git a/mm/pgtable-generic.c b/mm/pgtable-generic.c
>> index b91b1a98029c..ac1175d9a59a 100644
>> --- a/mm/pgtable-generic.c
>> +++ b/mm/pgtable-generic.c
>> @@ -217,6 +217,20 @@ pmd_t pmdp_invalidate_ad(struct vm_area_struct *vma, unsigned long address,
>> }
>> #endif
>>
>> +#ifndef __HAVE_ARCH_PUDP_INVALIDATE
>> +#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
> I'd revert these and similarly see if they can be combined somehow (likely combining with
> pudp_huge_clear_flush is harder as we are quite inconsistent with it, gluing it to
> __HAVE_ARCH_PMDP_HUGE_CLEAR_FLUSH).
>
> Which raises the question whether we could also piggy-back on __HAVE_ARCH_PMDP_SET_WRPROTECT
> (I suspect no without a closer look.)
>
^ permalink raw reply [flat|nested] 6+ messages in thread
* [PATCH 2/3] arm64: simplify pmd_special and pmd_mkspecial
2026-09-09 9:00 [PATCH 0/3] arm64: Add support for PUD THP Jinjiang Tu
2026-09-09 9:00 ` [PATCH 1/3] mm: define generic pudp_invalidate() implementation Jinjiang Tu
@ 2026-09-09 9:00 ` Jinjiang Tu
2026-09-09 9:00 ` [PATCH 3/3] arm64: Add support for PUD THP Jinjiang Tu
2 siblings, 0 replies; 6+ messages in thread
From: Jinjiang Tu @ 2026-09-09 9:00 UTC (permalink / raw)
To: akpm, david, ljs, catalin.marinas, will, mark.rutland,
ryan.roberts, peterx, dave.jiang, linux-arm-kernel, linux-mm
Cc: wangkefeng.wang, sunnanyong, tujinjiang
Use pte_special() and pte_mkspecial() helper to simplify pmd_special()
and pmd_mkspecial(), and kill unused set_pmd_bit() and clear_pmd_bit()
helper.
Signed-off-by: Jinjiang Tu <tujinjiang@huawei.com>
---
arch/arm64/include/asm/pgtable.h | 19 ++-----------------
1 file changed, 2 insertions(+), 17 deletions(-)
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 6000905a2e86..427ea3d2bac7 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -243,18 +243,6 @@ static inline pte_t set_pte_bit(pte_t pte, pgprot_t prot)
return pte;
}
-static inline pmd_t clear_pmd_bit(pmd_t pmd, pgprot_t prot)
-{
- pmd_val(pmd) &= ~pgprot_val(prot);
- return pmd;
-}
-
-static inline pmd_t set_pmd_bit(pmd_t pmd, pgprot_t prot)
-{
- pmd_val(pmd) |= pgprot_val(prot);
- return pmd;
-}
-
static inline pte_t pte_mkwrite_novma(pte_t pte)
{
pte = set_pte_bit(pte, __pgprot(PTE_WRITE));
@@ -619,11 +607,8 @@ static inline pmd_t pmd_mkhuge(pmd_t pmd)
}
#ifdef CONFIG_ARCH_SUPPORTS_PMD_PFNMAP
-#define pmd_special(pte) (!!((pmd_val(pte) & PTE_SPECIAL)))
-static inline pmd_t pmd_mkspecial(pmd_t pmd)
-{
- return set_pmd_bit(pmd, __pgprot(PTE_SPECIAL));
-}
+#define pmd_special(pmd) pte_special(pmd_pte(pmd))
+#define pmd_mkspecial(pmd) pte_pmd(pte_mkspecial(pmd_pte(pmd)))
#endif
#define __pmd_to_phys(pmd) __pte_to_phys(pmd_pte(pmd))
--
2.43.0
^ permalink raw reply related [flat|nested] 6+ messages in thread* [PATCH 3/3] arm64: Add support for PUD THP
2026-09-09 9:00 [PATCH 0/3] arm64: Add support for PUD THP Jinjiang Tu
2026-09-09 9:00 ` [PATCH 1/3] mm: define generic pudp_invalidate() implementation Jinjiang Tu
2026-09-09 9:00 ` [PATCH 2/3] arm64: simplify pmd_special and pmd_mkspecial Jinjiang Tu
@ 2026-09-09 9:00 ` Jinjiang Tu
2 siblings, 0 replies; 6+ messages in thread
From: Jinjiang Tu @ 2026-09-09 9:00 UTC (permalink / raw)
To: akpm, david, ljs, catalin.marinas, will, mark.rutland,
ryan.roberts, peterx, dave.jiang, linux-arm-kernel, linux-mm
Cc: wangkefeng.wang, sunnanyong, tujinjiang
Add the necessary page table functions to deal with PUD THP, this enables
PUD THP for DAX and pfnmap.
I tested with CONFIG_DEBUG_VM_PGTABLE and CONFIG_PAGE_TABLE_CHECK
enabled. I wrote a test driver that invokes vmf_insert_pfn_pud() in
the huge_fault callback. Page table setup works as expected with no
errors, and userspace read/write access to the PUD-sized mapping is
verified correct.
Signed-off-by: Jinjiang Tu <tujinjiang@huawei.com>
---
arch/arm64/Kconfig | 1 +
arch/arm64/include/asm/pgtable.h | 52 ++++++++++++++++++++++++++++++++
2 files changed, 53 insertions(+)
diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index b5a51b0ef944..8a7ca1863595 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -173,6 +173,7 @@ config ARM64
select HAVE_ARCH_THREAD_STRUCT_WHITELIST
select HAVE_ARCH_TRACEHOOK
select HAVE_ARCH_TRANSPARENT_HUGEPAGE
+ select HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD if PAGE_SIZE_4KB
select HAVE_ARCH_VMAP_STACK
select HAVE_ARM_SMCCC
select HAVE_ASM_MODVERSIONS
diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h
index 427ea3d2bac7..fdc338970934 100644
--- a/arch/arm64/include/asm/pgtable.h
+++ b/arch/arm64/include/asm/pgtable.h
@@ -611,6 +611,11 @@ static inline pmd_t pmd_mkhuge(pmd_t pmd)
#define pmd_mkspecial(pmd) pte_pmd(pte_mkspecial(pmd_pte(pmd)))
#endif
+#ifdef CONFIG_ARCH_SUPPORTS_PUD_PFNMAP
+#define pud_special(pud) pte_special(pud_pte(pud))
+#define pud_mkspecial(pud) pte_pud(pte_mkspecial(pud_pte(pud)))
+#endif
+
#define __pmd_to_phys(pmd) __pte_to_phys(pmd_pte(pmd))
#define __phys_to_pmd_val(phys) __phys_to_pte_val(phys)
#define pmd_pfn(pmd) ((__pmd_to_phys(pmd) & PMD_MASK) >> PAGE_SHIFT)
@@ -619,8 +624,15 @@ static inline pmd_t pmd_mkhuge(pmd_t pmd)
#define pud_young(pud) pte_young(pud_pte(pud))
#define pud_mkyoung(pud) pte_pud(pte_mkyoung(pud_pte(pud)))
#define pud_mkwrite_novma(pud) pte_pud(pte_mkwrite_novma(pud_pte(pud)))
+#define pud_mkwrite(pud) pud_mkwrite_novma(pud)
#define pud_mkvalid_k(pud) pte_pud(pte_mkvalid_k(pud_pte(pud)))
#define pud_write(pud) pte_write(pud_pte(pud))
+#define pud_dirty(pud) pte_dirty(pud_pte(pud))
+#define pud_wrprotect(pud) pte_pud(pte_wrprotect(pud_pte(pud)))
+#define pud_mkdirty(pud) pte_pud(pte_mkdirty(pud_pte(pud)))
+#define pud_mkold(pud) pte_pud(pte_mkold(pud_pte(pud)))
+#define pud_mkinvalid(pud) pte_pud(pte_mkinvalid(pud_pte(pud)))
+#define pud_mkclean(pud) pte_pud(pte_mkclean(pud_pte(pud)))
static inline pud_t pud_mkhuge(pud_t pud)
{
@@ -871,6 +883,13 @@ static inline bool pud_leaf(pud_t pud)
#define pud_user(pud) pte_user(pud_pte(pud))
#define pud_user_exec(pud) pte_user_exec(pud_pte(pud))
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+static inline int pud_trans_huge(pud_t pud)
+{
+ return pud_present(pud) && !pud_table(__pud(pud_val(pud) | PTE_VALID));
+}
+#endif
+
static inline bool pgtable_l4_enabled(void);
static inline void set_pud(pud_t *pudp, pud_t pud)
@@ -1237,6 +1256,11 @@ static inline pmd_t pmd_modify(pmd_t pmd, pgprot_t newprot)
return pte_pmd(pte_modify(pmd_pte(pmd), newprot));
}
+static inline pud_t pud_modify(pud_t pud, pgprot_t newprot)
+{
+ return pte_pud(pte_modify(pud_pte(pud), newprot));
+}
+
extern int __ptep_set_access_flags_anysz(struct vm_area_struct *vma,
unsigned long address, pte_t *ptep,
pte_t entry, int dirty,
@@ -1261,6 +1285,16 @@ static inline int pmdp_set_access_flags(struct vm_area_struct *vma,
}
#endif
+#ifdef CONFIG_ARCH_SUPPORTS_PUD_PFNMAP
+static inline int pudp_set_access_flags(struct vm_area_struct *vma,
+ unsigned long address, pud_t *pudp,
+ pud_t entry, int dirty)
+{
+ return __ptep_set_access_flags_anysz(vma, address, (pte_t *)pudp,
+ pud_pte(entry), dirty, PUD_SIZE);
+}
+#endif
+
#ifdef CONFIG_PAGE_TABLE_CHECK
static inline bool pte_user_accessible_page(struct mm_struct *mm, unsigned long addr, pte_t pte)
{
@@ -1335,6 +1369,14 @@ static inline bool pmdp_test_and_clear_young(struct vm_area_struct *vma,
}
#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_ARCH_HAS_NONLEAF_PMD_YOUNG */
+#ifdef CONFIG_ARCH_SUPPORTS_PUD_PFNMAP
+static inline bool pudp_test_and_clear_young(struct vm_area_struct *vma,
+ unsigned long address, pud_t *pudp)
+{
+ return __ptep_test_and_clear_young(vma, address, (pte_t *)pudp);
+}
+#endif
+
static inline pte_t __ptep_get_and_clear_anysz(struct mm_struct *mm,
unsigned long address,
pte_t *ptep,
@@ -1497,6 +1539,15 @@ static inline pmd_t pmdp_establish(struct vm_area_struct *vma,
}
#endif
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
+static inline pud_t pudp_establish(struct vm_area_struct *vma,
+ unsigned long address, pud_t *pudp, pud_t pud)
+{
+ page_table_check_pud_set(vma->vm_mm, address, pudp, pud);
+ return __pud(xchg_relaxed(&pud_val(*pudp), pud_val(pud)));
+}
+#endif
+
/*
* Encode and decode a swap entry:
* bits 0-1: present (must be zero)
@@ -1571,6 +1622,7 @@ static inline void update_mmu_cache_range(struct vm_fault *vmf,
#define update_mmu_cache(vma, addr, ptep) \
update_mmu_cache_range(NULL, vma, addr, ptep, 1)
#define update_mmu_cache_pmd(vma, address, pmd) do { } while (0)
+#define update_mmu_cache_pud(vma, address, pud) do { } while (0)
#ifdef CONFIG_ARM64_PA_BITS_52
#define phys_to_ttbr(addr) (((addr) | ((addr) >> 46)) & TTBR_BADDR_MASK_52)
--
2.43.0
^ permalink raw reply related [flat|nested] 6+ messages in thread