[PATCH v3] mm/page_vma_mapped_walk: Use ptep_get_lockless() for lockless access

[PATCH v3] mm/page_vma_mapped_walk: Use ptep_get_lockless() for lockless access

[Alexander Gordeev]

Switch from ptep_get() to ptep_get_lockless() accessor for
PTE reads when no lock is taken.

Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
---
 mm/page_vma_mapped.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
index a4d52fdb3056..2ccbabfb2cc1 100644
--- a/mm/page_vma_mapped.c
+++ b/mm/page_vma_mapped.c
@@ -41,7 +41,7 @@ static bool map_pte(struct page_vma_mapped_walk *pvmw, pmd_t *pmdvalp,
 	if (!pvmw->pte)
 		return false;
 
-	ptent = ptep_get(pvmw->pte);
+	ptent = ptep_get_lockless(pvmw->pte);
 
 	if (pte_none(ptent)) {
 		return false;
@@ -183,6 +183,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
 	struct mm_struct *mm = vma->vm_mm;
 	unsigned long end;
 	spinlock_t *ptl;
+	pte_t pteval;
 	pgd_t *pgd;
 	p4d_t *p4d;
 	pud_t *pud;
@@ -310,7 +311,11 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
 				goto restart;
 			}
 			pvmw->pte++;
-		} while (pte_none(ptep_get(pvmw->pte)));
+			if (!pvmw->ptl)
+				pteval = ptep_get_lockless(pvmw->pte);
+			else
+				pteval = ptep_get(pvmw->pte);
+		} while (pte_none(pteval));
 
 		if (!pvmw->ptl) {
 			spin_lock(ptl);
-- 
2.51.0

Re: [PATCH v3] mm/page_vma_mapped_walk: Use ptep_get_lockless() for lockless access

[Wei Yang]

On Thu, May 28, 2026 at 09:55:07AM +0200, Alexander Gordeev wrote:
>Switch from ptep_get() to ptep_get_lockless() accessor for
>PTE reads when no lock is taken.
>
>Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
>---
> mm/page_vma_mapped.c | 9 +++++++--
> 1 file changed, 7 insertions(+), 2 deletions(-)
>
>diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
>index a4d52fdb3056..2ccbabfb2cc1 100644
>--- a/mm/page_vma_mapped.c
>+++ b/mm/page_vma_mapped.c
>@@ -41,7 +41,7 @@ static bool map_pte(struct page_vma_mapped_walk *pvmw, pmd_t *pmdvalp,
> 	if (!pvmw->pte)
> 		return false;
> 
>-	ptent = ptep_get(pvmw->pte);
>+	ptent = ptep_get_lockless(pvmw->pte);
> 
> 	if (pte_none(ptent)) {
> 		return false;
>@@ -183,6 +183,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
> 	struct mm_struct *mm = vma->vm_mm;
> 	unsigned long end;
> 	spinlock_t *ptl;
>+	pte_t pteval;
> 	pgd_t *pgd;
> 	p4d_t *p4d;
> 	pud_t *pud;
>@@ -310,7 +311,11 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
> 				goto restart;
> 			}
> 			pvmw->pte++;
>-		} while (pte_none(ptep_get(pvmw->pte)));
>+			if (!pvmw->ptl)
>+				pteval = ptep_get_lockless(pvmw->pte);
>+			else
>+				pteval = ptep_get(pvmw->pte);
>+		} while (pte_none(pteval));

Use ptep_get_lockless() when lock is not taken.

LGTM, thanks

Reviewed-by: Wei Yang <richard.weiyang@gmail.com>

> 
> 		if (!pvmw->ptl) {
> 			spin_lock(ptl);
>-- 
>2.51.0

-- 
Wei Yang
Help you, Help me

Re: [PATCH v3] mm/page_vma_mapped_walk: Use ptep_get_lockless() for lockless access

[Lorenzo Stoakes]

On Thu, May 28, 2026 at 09:55:07AM +0200, Alexander Gordeev wrote:
> Switch from ptep_get() to ptep_get_lockless() accessor for
> PTE reads when no lock is taken.

This is an insufficient commit message that's just saying what you're doing,
which isn't very useful.

Please explain why you're doing this and what the benefit is please.

Thanks, Lorenzo

>
> Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
> ---
>  mm/page_vma_mapped.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
> index a4d52fdb3056..2ccbabfb2cc1 100644
> --- a/mm/page_vma_mapped.c
> +++ b/mm/page_vma_mapped.c
> @@ -41,7 +41,7 @@ static bool map_pte(struct page_vma_mapped_walk *pvmw, pmd_t *pmdvalp,
>  	if (!pvmw->pte)
>  		return false;
>
> -	ptent = ptep_get(pvmw->pte);
> +	ptent = ptep_get_lockless(pvmw->pte);
>
>  	if (pte_none(ptent)) {
>  		return false;
> @@ -183,6 +183,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
>  	struct mm_struct *mm = vma->vm_mm;
>  	unsigned long end;
>  	spinlock_t *ptl;
> +	pte_t pteval;
>  	pgd_t *pgd;
>  	p4d_t *p4d;
>  	pud_t *pud;
> @@ -310,7 +311,11 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
>  				goto restart;
>  			}
>  			pvmw->pte++;
> -		} while (pte_none(ptep_get(pvmw->pte)));
> +			if (!pvmw->ptl)
> +				pteval = ptep_get_lockless(pvmw->pte);
> +			else
> +				pteval = ptep_get(pvmw->pte);
> +		} while (pte_none(pteval));
>
>  		if (!pvmw->ptl) {
>  			spin_lock(ptl);
> --
> 2.51.0
>

Re: [PATCH v3] mm/page_vma_mapped_walk: Use ptep_get_lockless() for lockless access

[Oscar Salvador (SUSE)]

On Thu, May 28, 2026 at 12:24:35PM +0100, Lorenzo Stoakes wrote:
> On Thu, May 28, 2026 at 09:55:07AM +0200, Alexander Gordeev wrote:
> > Switch from ptep_get() to ptep_get_lockless() accessor for
> > PTE reads when no lock is taken.
> 
> This is an insufficient commit message that's just saying what you're doing,
> which isn't very useful.
> 
> Please explain why you're doing this and what the benefit is please.

Maybe something like

"When not holding the lock, there is a chance that the pte gets modified
 under our feet, so we need to use the lockless API to make sure that
 the entries remain consistent during the read."

Only arm64 makes use of it though, due to the contpte stuff and being
unable to do it atomically.



-- 
Oscar Salvador
SUSE Labs

Re: [PATCH v3] mm/page_vma_mapped_walk: Use ptep_get_lockless() for lockless access

[Alexander Gordeev]

On Thu, May 28, 2026 at 04:14:03PM +0200, Oscar Salvador (SUSE) wrote:
> On Thu, May 28, 2026 at 12:24:35PM +0100, Lorenzo Stoakes wrote:
> > On Thu, May 28, 2026 at 09:55:07AM +0200, Alexander Gordeev wrote:
> > > Switch from ptep_get() to ptep_get_lockless() accessor for
> > > PTE reads when no lock is taken.
> > 
> > This is an insufficient commit message that's just saying what you're doing,
> > which isn't very useful.
> > 
> > Please explain why you're doing this and what the benefit is please.
> 
> Maybe something like
> 
> "When not holding the lock, there is a chance that the pte gets modified
>  under our feet, so we need to use the lockless API to make sure that
>  the entries remain consistent during the read."

I would gladly use this, if it makes sense for everyone.

@Lorenzo?

> Only arm64 makes use of it though, due to the contpte stuff and being
> unable to do it atomically.

Thanks!

> -- 
> Oscar Salvador
> SUSE Labs

Re: [PATCH v3] mm/page_vma_mapped_walk: Use ptep_get_lockless() for lockless access

[Oscar Salvador (SUSE)]

On Thu, May 28, 2026 at 09:55:07AM +0200, Alexander Gordeev wrote:
> Switch from ptep_get() to ptep_get_lockless() accessor for
> PTE reads when no lock is taken.
> 
> Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>

I see Andrew has picked up the patch with the amended changelog, so with
that:

Reviewed-by: Oscar Salvador (SUSE) <osalvador@kernel.org>

> ---
>  mm/page_vma_mapped.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/page_vma_mapped.c b/mm/page_vma_mapped.c
> index a4d52fdb3056..2ccbabfb2cc1 100644
> --- a/mm/page_vma_mapped.c
> +++ b/mm/page_vma_mapped.c
> @@ -41,7 +41,7 @@ static bool map_pte(struct page_vma_mapped_walk *pvmw, pmd_t *pmdvalp,
>  	if (!pvmw->pte)
>  		return false;
>  
> -	ptent = ptep_get(pvmw->pte);
> +	ptent = ptep_get_lockless(pvmw->pte);
>  
>  	if (pte_none(ptent)) {
>  		return false;
> @@ -183,6 +183,7 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
>  	struct mm_struct *mm = vma->vm_mm;
>  	unsigned long end;
>  	spinlock_t *ptl;
> +	pte_t pteval;
>  	pgd_t *pgd;
>  	p4d_t *p4d;
>  	pud_t *pud;
> @@ -310,7 +311,11 @@ bool page_vma_mapped_walk(struct page_vma_mapped_walk *pvmw)
>  				goto restart;
>  			}
>  			pvmw->pte++;
> -		} while (pte_none(ptep_get(pvmw->pte)));
> +			if (!pvmw->ptl)
> +				pteval = ptep_get_lockless(pvmw->pte);
> +			else
> +				pteval = ptep_get(pvmw->pte);
> +		} while (pte_none(pteval));
>  
>  		if (!pvmw->ptl) {
>  			spin_lock(ptl);
> -- 
> 2.51.0
> 

-- 
Oscar Salvador
SUSE Labs