1. Patchew
  2. linux
  3. View series

[PATCH v3] mm/khugepaged: avoid underflow in madvise_collapse for sub-PMD MADV_COLLAPSE

Wandun Chen posted 1 patch 1 month ago 
mm/khugepaged.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)
[PATCH v3] mm/khugepaged: avoid underflow in madvise_collapse for sub-PMD MADV_COLLAPSE
Posted by Wandun Chen 1 month ago 
From: Chen Wandun <chenwandun@lixiang.com>

madvise_collapse() computes the THP-aligned window:

    hstart = ALIGN(start, HPAGE_PMD_SIZE); 	/* round up */
    hend = ALIGN_DOWN(end, HPAGE_PMD_SIZE);	/* round down */

The following case will cause hstart > hend, and result in underflow
in the return statement, avoid it by returning zero early when
hstart > hend. The return value is due to input is valid to madvise(),
and there is nothing to collapse.

    madvise(PMD-aligned + PAGE_SIZE, PAGE_SIZE, MADV_COLLAPSE);

In addition, kmalloc_obj(), mmgrab() and lru_add_drain_all() are
unnecessary when hstart == hend, so skip these operations by
returning early too.

Signed-off-by: Chen Wandun <chenwandun@lixiang.com>
---
v2 --> v3:
- Return 0 when hstart > hend, suggested by David and Lorenzo.

v1 --> v2:
- Rebase and resolve code conflict.
- Return -EINVAL when hstart > hend, suggested by Lorenzo.
- Drop Fixes tag, suggested by David and Lorenzo.
- Updated commit message to be more explicit, suggested by Lorenzo.
---
 mm/khugepaged.c | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index 28a843f30b32..fd7e893c998d 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -2837,6 +2837,12 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
 	if (!thp_vma_allowable_order(vma, vma->vm_flags, TVA_FORCED_COLLAPSE, PMD_ORDER))
 		return -EINVAL;
 
+	hstart = ALIGN(start, HPAGE_PMD_SIZE);
+	hend = ALIGN_DOWN(end, HPAGE_PMD_SIZE);
+
+	if (hstart >= hend)
+		return 0;
+
 	cc = kmalloc_obj(*cc);
 	if (!cc)
 		return -ENOMEM;
@@ -2846,9 +2852,6 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
 	mmgrab(mm);
 	lru_add_drain_all();
 
-	hstart = ALIGN(start, HPAGE_PMD_SIZE);
-	hend = ALIGN_DOWN(end, HPAGE_PMD_SIZE);
-
 	for (addr = hstart; addr < hend; addr += HPAGE_PMD_SIZE) {
 		enum scan_result result = SCAN_FAIL;
 
-- 
2.43.0
Re: [PATCH v3] mm/khugepaged: avoid underflow in madvise_collapse for sub-PMD MADV_COLLAPSE
Posted by Lorenzo Stoakes 1 month ago 
On Wed, May 13, 2026 at 01:54:28PM +0800, Wandun Chen wrote:
> From: Chen Wandun <chenwandun@lixiang.com>
>
> madvise_collapse() computes the THP-aligned window:
>
>     hstart = ALIGN(start, HPAGE_PMD_SIZE); 	/* round up */
>     hend = ALIGN_DOWN(end, HPAGE_PMD_SIZE);	/* round down */
>
> The following case will cause hstart > hend, and result in underflow
> in the return statement, avoid it by returning zero early when
> hstart > hend. The return value is due to input is valid to madvise(),
> and there is nothing to collapse.
>
>     madvise(PMD-aligned + PAGE_SIZE, PAGE_SIZE, MADV_COLLAPSE);
>
> In addition, kmalloc_obj(), mmgrab() and lru_add_drain_all() are
> unnecessary when hstart == hend, so skip these operations by
> returning early too.
>
> Signed-off-by: Chen Wandun <chenwandun@lixiang.com>

LGTM, so:

Reviewed-by: Lorenzo Stoakes <ljs@kernel.org>

Cheers, Lorenzo

> ---
> v2 --> v3:
> - Return 0 when hstart > hend, suggested by David and Lorenzo.
>
> v1 --> v2:
> - Rebase and resolve code conflict.
> - Return -EINVAL when hstart > hend, suggested by Lorenzo.
> - Drop Fixes tag, suggested by David and Lorenzo.
> - Updated commit message to be more explicit, suggested by Lorenzo.
> ---
>  mm/khugepaged.c | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index 28a843f30b32..fd7e893c998d 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -2837,6 +2837,12 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
>  	if (!thp_vma_allowable_order(vma, vma->vm_flags, TVA_FORCED_COLLAPSE, PMD_ORDER))
>  		return -EINVAL;
>
> +	hstart = ALIGN(start, HPAGE_PMD_SIZE);
> +	hend = ALIGN_DOWN(end, HPAGE_PMD_SIZE);
> +
> +	if (hstart >= hend)
> +		return 0;
> +
>  	cc = kmalloc_obj(*cc);
>  	if (!cc)
>  		return -ENOMEM;
> @@ -2846,9 +2852,6 @@ int madvise_collapse(struct vm_area_struct *vma, unsigned long start,
>  	mmgrab(mm);
>  	lru_add_drain_all();
>
> -	hstart = ALIGN(start, HPAGE_PMD_SIZE);
> -	hend = ALIGN_DOWN(end, HPAGE_PMD_SIZE);
> -
>  	for (addr = hstart; addr < hend; addr += HPAGE_PMD_SIZE) {
>  		enum scan_result result = SCAN_FAIL;
>
> --
> 2.43.0
>
Re: [PATCH v3] mm/khugepaged: avoid underflow in madvise_collapse for sub-PMD MADV_COLLAPSE
Posted by David Hildenbrand (Arm) 1 month ago 
On 5/13/26 07:54, Wandun Chen wrote:
> From: Chen Wandun <chenwandun@lixiang.com>
> 
> madvise_collapse() computes the THP-aligned window:
> 
>     hstart = ALIGN(start, HPAGE_PMD_SIZE); 	/* round up */
>     hend = ALIGN_DOWN(end, HPAGE_PMD_SIZE);	/* round down */
> 
> The following case will cause hstart > hend, and result in underflow
> in the return statement, avoid it by returning zero early when
> hstart > hend. The return value is due to input is valid to madvise(),
> and there is nothing to collapse.
> 
>     madvise(PMD-aligned + PAGE_SIZE, PAGE_SIZE, MADV_COLLAPSE);
> 
> In addition, kmalloc_obj(), mmgrab() and lru_add_drain_all() are
> unnecessary when hstart == hend, so skip these operations by
> returning early too.
> 
> Signed-off-by: Chen Wandun <chenwandun@lixiang.com>
> ---

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

-- 
Cheers,

David

Patchew 2.1-devel-ea86937

© 2016 - 2026 Red Hat, Inc.