[Patch v2 0/3] unify PMD scan results and remove redundant cleanup

[Patch v2 0/3] unify PMD scan results and remove redundant cleanup

[Wei Yang]

This small series addresses two minor cleanup opportunities in the hugepage
collapse logic.

The initial motivation arose during a code review of madvise_collapse(),
where it was noted that the function was missing a handler for
SCAN_PMD_NONE. This oversight exposed the inconsistent handling of
SCAN_PMD_NULL and SCAN_PMD_NONE.

Since both scan results are functionally identical (they indicate the
absence of a PTE table), the primary patch unifies them into a single,
clearer identifier, SCAN_NO_PTE_TABLE.

The series also takes the opportunity to remove a redundant clearing of
the struct collapse_control.

---
v1: http://lkml.kernel.org/r/20251112020031.25350-1-richard.weiyang@gmail.com

Wei Yang (3):
  mm/khugepaged: remove redundant clearing of struct collapse_control
  mm/khugepaged: continue to collapse on SCAN_PMD_NONE
  mm/khugepaged: unify SCAN_PMD_NONE and SCAN_PMD_NULL into
    SCAN_NO_PTE_TABLE

 include/trace/events/huge_memory.h |  3 +--
 mm/khugepaged.c                    | 24 ++++++++++--------------
 2 files changed, 11 insertions(+), 16 deletions(-)

-- 
2.34.1

[Patch v2 1/3] mm/khugepaged: remove redundant clearing of struct collapse_control

[Wei Yang]

The structure struct collapse_control is being unnecessarily cleared
twice during the huge page collapse process.

Both hpage_collapse_scan_file() and hpage_collapse_scan_pmd() currently
perform a clear operation on this structure.

Remove the redundant clear operation.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
---
 mm/khugepaged.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 7e8cb181d5bd..1fc8986a28b3 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -2803,8 +2803,6 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
 			hend = min(hend, vma->vm_end & HPAGE_PMD_MASK);
 		}
 		mmap_assert_locked(mm);
-		memset(cc->node_load, 0, sizeof(cc->node_load));
-		nodes_clear(cc->alloc_nmask);
 		if (!vma_is_anonymous(vma)) {
 			struct file *file = get_file(vma->vm_file);
 			pgoff_t pgoff = linear_page_index(vma, addr);
-- 
2.34.1

[Patch v2 2/3] mm/khugepaged: continue to collapse on SCAN_PMD_NONE

[Wei Yang]

SCAN_PMD_NONE means current pmd is empty, but we can still continue
collapse next pmd range.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Lance Yang <lance.yang@linux.dev>
Reviewed-by: Dev Jain <dev.jain@arm.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
---
 mm/khugepaged.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 1fc8986a28b3..2ee5048b764e 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -2833,6 +2833,7 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
 			goto handle_result;
 		/* Whitelisted set of results where continuing OK */
 		case SCAN_PMD_NULL:
+		case SCAN_PMD_NONE:
 		case SCAN_PTE_NON_PRESENT:
 		case SCAN_PTE_UFFD_WP:
 		case SCAN_LACK_REFERENCED_PAGE:
-- 
2.34.1

[Patch v2 3/3] mm/khugepaged: unify SCAN_PMD_NONE and SCAN_PMD_NULL into SCAN_NO_PTE_TABLE

[Wei Yang]

The current hugepage collapse scan results include two separate values,
SCAN_PMD_NONE and SCAN_PMD_NULL, which are handled identically by
the consuming code.

To reduce confusion and improve long-term maintenance, this commit
merges these two functionally equivalent states into a single, clearer
identifier: SCAN_NO_PTE_TABLE

Suggested-by: "David Hildenbrand (Red Hat)" <david@kernel.org>
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
---
 include/trace/events/huge_memory.h |  3 +--
 mm/khugepaged.c                    | 23 ++++++++++-------------
 2 files changed, 11 insertions(+), 15 deletions(-)

diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
index dd94d14a2427..4cde53b45a85 100644
--- a/include/trace/events/huge_memory.h
+++ b/include/trace/events/huge_memory.h
@@ -10,8 +10,7 @@
 #define SCAN_STATUS							\
 	EM( SCAN_FAIL,			"failed")			\
 	EM( SCAN_SUCCEED,		"succeeded")			\
-	EM( SCAN_PMD_NULL,		"pmd_null")			\
-	EM( SCAN_PMD_NONE,		"pmd_none")			\
+	EM( SCAN_NO_PTE_TABLE,		"no_pte_table")			\
 	EM( SCAN_PMD_MAPPED,		"page_pmd_mapped")		\
 	EM( SCAN_EXCEED_NONE_PTE,	"exceed_none_pte")		\
 	EM( SCAN_EXCEED_SWAP_PTE,	"exceed_swap_pte")		\
diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 2ee5048b764e..40f9d5939aa5 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -30,8 +30,7 @@
 enum scan_result {
 	SCAN_FAIL,
 	SCAN_SUCCEED,
-	SCAN_PMD_NULL,
-	SCAN_PMD_NONE,
+	SCAN_NO_PTE_TABLE,
 	SCAN_PMD_MAPPED,
 	SCAN_EXCEED_NONE_PTE,
 	SCAN_EXCEED_SWAP_PTE,
@@ -934,7 +933,7 @@ static inline int check_pmd_state(pmd_t *pmd)
 	pmd_t pmde = pmdp_get_lockless(pmd);
 
 	if (pmd_none(pmde))
-		return SCAN_PMD_NONE;
+		return SCAN_NO_PTE_TABLE;
 
 	/*
 	 * The folio may be under migration when khugepaged is trying to
@@ -944,11 +943,11 @@ static inline int check_pmd_state(pmd_t *pmd)
 	if (pmd_is_migration_entry(pmde))
 		return SCAN_PMD_MAPPED;
 	if (!pmd_present(pmde))
-		return SCAN_PMD_NULL;
+		return SCAN_NO_PTE_TABLE;
 	if (pmd_trans_huge(pmde))
 		return SCAN_PMD_MAPPED;
 	if (pmd_bad(pmde))
-		return SCAN_PMD_NULL;
+		return SCAN_NO_PTE_TABLE;
 	return SCAN_SUCCEED;
 }
 
@@ -958,7 +957,7 @@ static int find_pmd_or_thp_or_none(struct mm_struct *mm,
 {
 	*pmd = mm_find_pmd(mm, address);
 	if (!*pmd)
-		return SCAN_PMD_NULL;
+		return SCAN_NO_PTE_TABLE;
 
 	return check_pmd_state(*pmd);
 }
@@ -1013,7 +1012,7 @@ static int __collapse_huge_page_swapin(struct mm_struct *mm,
 			pte = pte_offset_map_ro_nolock(mm, pmd, addr, &ptl);
 			if (!pte) {
 				mmap_read_unlock(mm);
-				result = SCAN_PMD_NULL;
+				result = SCAN_NO_PTE_TABLE;
 				goto out;
 			}
 		}
@@ -1187,7 +1186,7 @@ static int collapse_huge_page(struct mm_struct *mm, unsigned long address,
 						      &compound_pagelist);
 		spin_unlock(pte_ptl);
 	} else {
-		result = SCAN_PMD_NULL;
+		result = SCAN_NO_PTE_TABLE;
 	}
 
 	if (unlikely(result != SCAN_SUCCEED)) {
@@ -1270,7 +1269,7 @@ static int hpage_collapse_scan_pmd(struct mm_struct *mm,
 	nodes_clear(cc->alloc_nmask);
 	pte = pte_offset_map_lock(mm, pmd, start_addr, &ptl);
 	if (!pte) {
-		result = SCAN_PMD_NULL;
+		result = SCAN_NO_PTE_TABLE;
 		goto out;
 	}
 
@@ -1544,8 +1543,7 @@ int collapse_pte_mapped_thp(struct mm_struct *mm, unsigned long addr,
 	switch (result) {
 	case SCAN_SUCCEED:
 		break;
-	case SCAN_PMD_NULL:
-	case SCAN_PMD_NONE:
+	case SCAN_NO_PTE_TABLE:
 		/*
 		 * All pte entries have been removed and pmd cleared.
 		 * Skip all the pte checks and just update the pmd mapping.
@@ -2832,8 +2830,7 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
 			mmap_read_unlock(mm);
 			goto handle_result;
 		/* Whitelisted set of results where continuing OK */
-		case SCAN_PMD_NULL:
-		case SCAN_PMD_NONE:
+		case SCAN_NO_PTE_TABLE:
 		case SCAN_PTE_NON_PRESENT:
 		case SCAN_PTE_UFFD_WP:
 		case SCAN_LACK_REFERENCED_PAGE:
-- 
2.34.1

Re: [Patch v2 1/3] mm/khugepaged: remove redundant clearing of struct collapse_control

[Dev Jain]

On 14/11/25 8:30 am, Wei Yang wrote:
> The structure struct collapse_control is being unnecessarily cleared
> twice during the huge page collapse process.
>
> Both hpage_collapse_scan_file() and hpage_collapse_scan_pmd() currently
> perform a clear operation on this structure.
>
> Remove the redundant clear operation.
>
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
>   mm/khugepaged.c | 2 --
>   1 file changed, 2 deletions(-)
>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 7e8cb181d5bd..1fc8986a28b3 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -2803,8 +2803,6 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
>   			hend = min(hend, vma->vm_end & HPAGE_PMD_MASK);
>   		}
>   		mmap_assert_locked(mm);
> -		memset(cc->node_load, 0, sizeof(cc->node_load));
> -		nodes_clear(cc->alloc_nmask);
>   		if (!vma_is_anonymous(vma)) {
>   			struct file *file = get_file(vma->vm_file);
>   			pgoff_t pgoff = linear_page_index(vma, addr);

Reviewed-by: Dev Jain <dev.jain@arm.com>

Re: [Patch v2 3/3] mm/khugepaged: unify SCAN_PMD_NONE and SCAN_PMD_NULL into SCAN_NO_PTE_TABLE

[Dev Jain]

On 14/11/25 8:30 am, Wei Yang wrote:
> The current hugepage collapse scan results include two separate values,
> SCAN_PMD_NONE and SCAN_PMD_NULL, which are handled identically by
> the consuming code.
>
> To reduce confusion and improve long-term maintenance, this commit
> merges these two functionally equivalent states into a single, clearer
> identifier: SCAN_NO_PTE_TABLE
>
> Suggested-by: "David Hildenbrand (Red Hat)" <david@kernel.org>
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
>   

Reviewed-by: Dev Jain <dev.jain@arm.com>

Re: [Patch v2 1/3] mm/khugepaged: remove redundant clearing of struct collapse_control

[David Hildenbrand (Red Hat)]

On 14.11.25 04:00, Wei Yang wrote:
> The structure struct collapse_control is being unnecessarily cleared
> twice during the huge page collapse process.
> 
> Both hpage_collapse_scan_file() and hpage_collapse_scan_pmd() currently
> perform a clear operation on this structure.
> 
> Remove the redundant clear operation.
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>

Right, and the structure does not seem to get used in between. Thanks!

Acked-by: David Hildenbrand (Red Hat) <david@kernel.org>

-- 
Cheers

David

Re: [Patch v2 2/3] mm/khugepaged: continue to collapse on SCAN_PMD_NONE

[David Hildenbrand (Red Hat)]

On 14.11.25 04:00, Wei Yang wrote:
> SCAN_PMD_NONE means current pmd is empty, but we can still continue
> collapse next pmd range.
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> Reviewed-by: Lance Yang <lance.yang@linux.dev>
> Reviewed-by: Dev Jain <dev.jain@arm.com>
> Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> ---

Acked-by: David Hildenbrand (Red Hat) <david@kernel.org>

-- 
Cheers

David

Re: [Patch v2 3/3] mm/khugepaged: unify SCAN_PMD_NONE and SCAN_PMD_NULL into SCAN_NO_PTE_TABLE

[David Hildenbrand (Red Hat)]

On 14.11.25 04:00, Wei Yang wrote:
> The current hugepage collapse scan results include two separate values,
> SCAN_PMD_NONE and SCAN_PMD_NULL, which are handled identically by
> the consuming code.
> 
> To reduce confusion and improve long-term maintenance, this commit
> merges these two functionally equivalent states into a single, clearer
> identifier: SCAN_NO_PTE_TABLE
> 
> Suggested-by: "David Hildenbrand (Red Hat)" <david@kernel.org>
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---

Acked-by: David Hildenbrand (Red Hat) <david@kernel.org>

-- 
Cheers

David

Re: [Patch v2 1/3] mm/khugepaged: remove redundant clearing of struct collapse_control

[Baolin Wang]

On 2025/11/14 11:00, Wei Yang wrote:
> The structure struct collapse_control is being unnecessarily cleared
> twice during the huge page collapse process.
> 
> Both hpage_collapse_scan_file() and hpage_collapse_scan_pmd() currently
> perform a clear operation on this structure.
> 
> Remove the redundant clear operation.
> 
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---

LGTM.
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>

Re: [Patch v2 3/3] mm/khugepaged: unify SCAN_PMD_NONE and SCAN_PMD_NULL into SCAN_NO_PTE_TABLE

[Baolin Wang]

On 2025/11/14 11:00, Wei Yang wrote:
> The current hugepage collapse scan results include two separate values,
> SCAN_PMD_NONE and SCAN_PMD_NULL, which are handled identically by
> the consuming code.
> 
> To reduce confusion and improve long-term maintenance, this commit
> merges these two functionally equivalent states into a single, clearer
> identifier: SCAN_NO_PTE_TABLE
> 
> Suggested-by: "David Hildenbrand (Red Hat)" <david@kernel.org>
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---

Looks cleaner. Thanks.
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>

Re: [Patch v2 1/3] mm/khugepaged: remove redundant clearing of struct collapse_control

[Nico Pache]

On Thu, Nov 13, 2025 at 8:01 PM Wei Yang <richard.weiyang@gmail.com> wrote:
>
> The structure struct collapse_control is being unnecessarily cleared
> twice during the huge page collapse process.
>
> Both hpage_collapse_scan_file() and hpage_collapse_scan_pmd() currently
> perform a clear operation on this structure.
>
> Remove the redundant clear operation.

Nice! The more we can simplify/unify madvise_collapse/collapse the better :)

LGTM!

Reviewed-by: Nico Pache <npache@redhat.com>

>
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
>  mm/khugepaged.c | 2 --
>  1 file changed, 2 deletions(-)
>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 7e8cb181d5bd..1fc8986a28b3 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -2803,8 +2803,6 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
>                         hend = min(hend, vma->vm_end & HPAGE_PMD_MASK);
>                 }
>                 mmap_assert_locked(mm);
> -               memset(cc->node_load, 0, sizeof(cc->node_load));
> -               nodes_clear(cc->alloc_nmask);
>                 if (!vma_is_anonymous(vma)) {
>                         struct file *file = get_file(vma->vm_file);
>                         pgoff_t pgoff = linear_page_index(vma, addr);
> --
> 2.34.1
>

Re: [Patch v2 3/3] mm/khugepaged: unify SCAN_PMD_NONE and SCAN_PMD_NULL into SCAN_NO_PTE_TABLE

[Nico Pache]

On Thu, Nov 13, 2025 at 8:00 PM Wei Yang <richard.weiyang@gmail.com> wrote:
>
> The current hugepage collapse scan results include two separate values,
> SCAN_PMD_NONE and SCAN_PMD_NULL, which are handled identically by
> the consuming code.
>
> To reduce confusion and improve long-term maintenance, this commit
> merges these two functionally equivalent states into a single, clearer
> identifier: SCAN_NO_PTE_TABLE

Nice cleanup!

Reviewed-by: Nico Pache <npache@redhat.com>

>
> Suggested-by: "David Hildenbrand (Red Hat)" <david@kernel.org>
> Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
> ---
>  include/trace/events/huge_memory.h |  3 +--
>  mm/khugepaged.c                    | 23 ++++++++++-------------
>  2 files changed, 11 insertions(+), 15 deletions(-)
>
> diff --git a/include/trace/events/huge_memory.h b/include/trace/events/huge_memory.h
> index dd94d14a2427..4cde53b45a85 100644
> --- a/include/trace/events/huge_memory.h
> +++ b/include/trace/events/huge_memory.h
> @@ -10,8 +10,7 @@
>  #define SCAN_STATUS                                                    \
>         EM( SCAN_FAIL,                  "failed")                       \
>         EM( SCAN_SUCCEED,               "succeeded")                    \
> -       EM( SCAN_PMD_NULL,              "pmd_null")                     \
> -       EM( SCAN_PMD_NONE,              "pmd_none")                     \
> +       EM( SCAN_NO_PTE_TABLE,          "no_pte_table")                 \
>         EM( SCAN_PMD_MAPPED,            "page_pmd_mapped")              \
>         EM( SCAN_EXCEED_NONE_PTE,       "exceed_none_pte")              \
>         EM( SCAN_EXCEED_SWAP_PTE,       "exceed_swap_pte")              \
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 2ee5048b764e..40f9d5939aa5 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -30,8 +30,7 @@
>  enum scan_result {
>         SCAN_FAIL,
>         SCAN_SUCCEED,
> -       SCAN_PMD_NULL,
> -       SCAN_PMD_NONE,
> +       SCAN_NO_PTE_TABLE,
>         SCAN_PMD_MAPPED,
>         SCAN_EXCEED_NONE_PTE,
>         SCAN_EXCEED_SWAP_PTE,
> @@ -934,7 +933,7 @@ static inline int check_pmd_state(pmd_t *pmd)
>         pmd_t pmde = pmdp_get_lockless(pmd);
>
>         if (pmd_none(pmde))
> -               return SCAN_PMD_NONE;
> +               return SCAN_NO_PTE_TABLE;
>
>         /*
>          * The folio may be under migration when khugepaged is trying to
> @@ -944,11 +943,11 @@ static inline int check_pmd_state(pmd_t *pmd)
>         if (pmd_is_migration_entry(pmde))
>                 return SCAN_PMD_MAPPED;
>         if (!pmd_present(pmde))
> -               return SCAN_PMD_NULL;
> +               return SCAN_NO_PTE_TABLE;
>         if (pmd_trans_huge(pmde))
>                 return SCAN_PMD_MAPPED;
>         if (pmd_bad(pmde))
> -               return SCAN_PMD_NULL;
> +               return SCAN_NO_PTE_TABLE;
>         return SCAN_SUCCEED;
>  }
>
> @@ -958,7 +957,7 @@ static int find_pmd_or_thp_or_none(struct mm_struct *mm,
>  {
>         *pmd = mm_find_pmd(mm, address);
>         if (!*pmd)
> -               return SCAN_PMD_NULL;
> +               return SCAN_NO_PTE_TABLE;
>
>         return check_pmd_state(*pmd);
>  }
> @@ -1013,7 +1012,7 @@ static int __collapse_huge_page_swapin(struct mm_struct *mm,
>                         pte = pte_offset_map_ro_nolock(mm, pmd, addr, &ptl);
>                         if (!pte) {
>                                 mmap_read_unlock(mm);
> -                               result = SCAN_PMD_NULL;
> +                               result = SCAN_NO_PTE_TABLE;
>                                 goto out;
>                         }
>                 }
> @@ -1187,7 +1186,7 @@ static int collapse_huge_page(struct mm_struct *mm, unsigned long address,
>                                                       &compound_pagelist);
>                 spin_unlock(pte_ptl);
>         } else {
> -               result = SCAN_PMD_NULL;
> +               result = SCAN_NO_PTE_TABLE;
>         }
>
>         if (unlikely(result != SCAN_SUCCEED)) {
> @@ -1270,7 +1269,7 @@ static int hpage_collapse_scan_pmd(struct mm_struct *mm,
>         nodes_clear(cc->alloc_nmask);
>         pte = pte_offset_map_lock(mm, pmd, start_addr, &ptl);
>         if (!pte) {
> -               result = SCAN_PMD_NULL;
> +               result = SCAN_NO_PTE_TABLE;
>                 goto out;
>         }
>
> @@ -1544,8 +1543,7 @@ int collapse_pte_mapped_thp(struct mm_struct *mm, unsigned long addr,
>         switch (result) {
>         case SCAN_SUCCEED:
>                 break;
> -       case SCAN_PMD_NULL:
> -       case SCAN_PMD_NONE:
> +       case SCAN_NO_PTE_TABLE:
>                 /*
>                  * All pte entries have been removed and pmd cleared.
>                  * Skip all the pte checks and just update the pmd mapping.
> @@ -2832,8 +2830,7 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
>                         mmap_read_unlock(mm);
>                         goto handle_result;
>                 /* Whitelisted set of results where continuing OK */
> -               case SCAN_PMD_NULL:
> -               case SCAN_PMD_NONE:
> +               case SCAN_NO_PTE_TABLE:
>                 case SCAN_PTE_NON_PRESENT:
>                 case SCAN_PTE_UFFD_WP:
>                 case SCAN_LACK_REFERENCED_PAGE:
> --
> 2.34.1
>