[PATCH] mm: move the easily assessable conditions forward

Huan Yang posted 1 patch 1 year, 5 months ago
mm/migrate.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)
[PATCH] mm: move the easily assessable conditions forward
Posted by Huan Yang 1 year, 5 months ago
Current try_to_map_unused_to_zeropage try to use shared zero page to
save some memory of sub page.

If forbids zeropage, no need to do anything rather than attempting to
assess wthether to use it afterwards.

Signed-off-by: Huan Yang <link@vivo.com>
---
 mm/migrate.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/mm/migrate.c b/mm/migrate.c
index 6e32098ac2dc..d71cc4ff190f 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -192,6 +192,9 @@ static bool try_to_map_unused_to_zeropage(struct page_vma_mapped_walk *pvmw,
 	VM_BUG_ON_PAGE(!PageLocked(page), page);
 	VM_BUG_ON_PAGE(pte_present(*pvmw->pte), page);
 
+	if (mm_forbids_zeropage(pvmw->vma->vm_mm))
+		return false;
+
 	if (PageMlocked(page) || (pvmw->vma->vm_flags & VM_LOCKED))
 		return false;
 
@@ -204,7 +207,7 @@ static bool try_to_map_unused_to_zeropage(struct page_vma_mapped_walk *pvmw,
 	contains_data = memchr_inv(addr, 0, PAGE_SIZE);
 	kunmap_local(addr);
 
-	if (contains_data || mm_forbids_zeropage(pvmw->vma->vm_mm))
+	if (contains_data)
 		return false;
 
 	newpte = pte_mkspecial(pfn_pte(my_zero_pfn(pvmw->address),

base-commit: edd1ec2e3a9f5de7fb267a3af73e4f00e7e052b7
-- 
2.45.2
Re: [PATCH] mm: move the easily assessable conditions forward
Posted by Andrew Morton 1 year, 5 months ago
On Thu, 15 Aug 2024 16:31:01 +0800 Huan Yang <link@vivo.com> wrote:

> Current try_to_map_unused_to_zeropage try to use shared zero page to
> save some memory of sub page.
> 
> If forbids zeropage, no need to do anything rather than attempting to
> assess wthether to use it afterwards.
> 
> ...
>
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -192,6 +192,9 @@ static bool try_to_map_unused_to_zeropage(struct page_vma_mapped_walk *pvmw,
>  	VM_BUG_ON_PAGE(!PageLocked(page), page);
>  	VM_BUG_ON_PAGE(pte_present(*pvmw->pte), page);
>  
> +	if (mm_forbids_zeropage(pvmw->vma->vm_mm))
> +		return false;
> +
>  	if (PageMlocked(page) || (pvmw->vma->vm_flags & VM_LOCKED))
>  		return false;
>  
> @@ -204,7 +207,7 @@ static bool try_to_map_unused_to_zeropage(struct page_vma_mapped_walk *pvmw,
>  	contains_data = memchr_inv(addr, 0, PAGE_SIZE);
>  	kunmap_local(addr);
>  
> -	if (contains_data || mm_forbids_zeropage(pvmw->vma->vm_mm))
> +	if (contains_data)
>  		return false;
>  

Looks sensible.  I'll add it as a fixup to "mm: remap unused subpages to shared zeropage when splitting isolated thp".