[PATCH 0/2] mm/gup: Cleanup memfd_pin_folios()

[PATCH 0/2] mm/gup: Cleanup memfd_pin_folios()

[Vishal Moola (Oracle)]

A couple straightforward cleanups to memfd_pin_folios() found through
code inspection. Saves 124 bytes of kernel text overall and makes the
code more readable.

Vishal Moola (Oracle) (2):
  mm/gup: Remove unnecessary check in memfd_pin_folios()
  mm/gup: Remove page_folio() in memfd_pin_folios()

 mm/gup.c | 17 ++---------------
 1 file changed, 2 insertions(+), 15 deletions(-)

-- 
2.49.0

[PATCH 1/2] mm/gup: Remove unnecessary check in memfd_pin_folios()

[Vishal Moola (Oracle)]

Commit 89c1905d9c14 ("mm/gup: introduce memfd_pin_folios() for pinning memfd folios")
checks if filemap_get_folios_contig() returned duplicate folios to
prevent multiple attempts at pinning the same folio.

Commit 8ab1b1602396 ("mm: fix filemap_get_folios_contig returning batches of identical folios")
ensures that filemap_get_folios_contig() returns a batch of distinct folios.

We can remove the duplicate folio check to simplify the code and save
58 bytes of text.

Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
---
 mm/gup.c | 15 +--------------
 1 file changed, 1 insertion(+), 14 deletions(-)

diff --git a/mm/gup.c b/mm/gup.c
index f32168339390..1fb8f3b9a493 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -3589,7 +3589,7 @@ long memfd_pin_folios(struct file *memfd, loff_t start, loff_t end,
 {
 	unsigned int flags, nr_folios, nr_found;
 	unsigned int i, pgshift = PAGE_SHIFT;
-	pgoff_t start_idx, end_idx, next_idx;
+	pgoff_t start_idx, end_idx;
 	struct folio *folio = NULL;
 	struct folio_batch fbatch;
 	struct hstate *h;
@@ -3639,19 +3639,7 @@ long memfd_pin_folios(struct file *memfd, loff_t start, loff_t end,
 				folio = NULL;
 			}
 
-			next_idx = 0;
 			for (i = 0; i < nr_found; i++) {
-				/*
-				 * As there can be multiple entries for a
-				 * given folio in the batch returned by
-				 * filemap_get_folios_contig(), the below
-				 * check is to ensure that we pin and return a
-				 * unique set of folios between start and end.
-				 */
-				if (next_idx &&
-				    next_idx != folio_index(fbatch.folios[i]))
-					continue;
-
 				folio = page_folio(&fbatch.folios[i]->page);
 
 				if (try_grab_folio(folio, 1, FOLL_PIN)) {
@@ -3664,7 +3652,6 @@ long memfd_pin_folios(struct file *memfd, loff_t start, loff_t end,
 					*offset = offset_in_folio(folio, start);
 
 				folios[nr_folios] = folio;
-				next_idx = folio_next_index(folio);
 				if (++nr_folios == max_folios)
 					break;
 			}
-- 
2.49.0

[PATCH 2/2] mm/gup: Remove page_folio() in memfd_pin_folios()

[Vishal Moola (Oracle)]

We can get the folio directly from the folio batch, so remove the
unnecessary page_folio() call.

Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
---
 mm/gup.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/gup.c b/mm/gup.c
index 1fb8f3b9a493..795fd94f379d 100644
--- a/mm/gup.c
+++ b/mm/gup.c
@@ -3640,7 +3640,7 @@ long memfd_pin_folios(struct file *memfd, loff_t start, loff_t end,
 			}
 
 			for (i = 0; i < nr_found; i++) {
-				folio = page_folio(&fbatch.folios[i]->page);
+				folio = fbatch.folios[i];
 
 				if (try_grab_folio(folio, 1, FOLL_PIN)) {
 					folio_batch_release(&fbatch);
-- 
2.49.0

Re: [PATCH 1/2] mm/gup: Remove unnecessary check in memfd_pin_folios()

[David Hildenbrand]

On 30.04.25 03:00, Vishal Moola (Oracle) wrote:
> Commit 89c1905d9c14 ("mm/gup: introduce memfd_pin_folios() for pinning memfd folios")
> checks if filemap_get_folios_contig() returned duplicate folios to
> prevent multiple attempts at pinning the same folio.
> 
> Commit 8ab1b1602396 ("mm: fix filemap_get_folios_contig returning batches of identical folios")
> ensures that filemap_get_folios_contig() returns a batch of distinct folios.
> 
> We can remove the duplicate folio check to simplify the code and save
> 58 bytes of text.
> 
> Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
> ---

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers,

David / dhildenb

Re: [PATCH 2/2] mm/gup: Remove page_folio() in memfd_pin_folios()

[David Hildenbrand]

On 30.04.25 03:00, Vishal Moola (Oracle) wrote:
> We can get the folio directly from the folio batch, so remove the
> unnecessary page_folio() call.
> 
> Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
> ---
>   mm/gup.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/gup.c b/mm/gup.c
> index 1fb8f3b9a493..795fd94f379d 100644
> --- a/mm/gup.c
> +++ b/mm/gup.c
> @@ -3640,7 +3640,7 @@ long memfd_pin_folios(struct file *memfd, loff_t start, loff_t end,
>   			}
>   
>   			for (i = 0; i < nr_found; i++) {
> -				folio = page_folio(&fbatch.folios[i]->page);
> +				folio = fbatch.folios[i];
>   
>   				if (try_grab_folio(folio, 1, FOLL_PIN)) {
>   					folio_batch_release(&fbatch);

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers,

David / dhildenb

RE: [PATCH 1/2] mm/gup: Remove unnecessary check in memfd_pin_folios()

[Kasireddy, Vivek]

Hi Vishal,

> Subject: [PATCH 1/2] mm/gup: Remove unnecessary check in
> memfd_pin_folios()
> 
> Commit 89c1905d9c14 ("mm/gup: introduce memfd_pin_folios() for pinning
> memfd folios")
> checks if filemap_get_folios_contig() returned duplicate folios to
> prevent multiple attempts at pinning the same folio.
> 
> Commit 8ab1b1602396 ("mm: fix filemap_get_folios_contig returning batches
> of identical folios")
> ensures that filemap_get_folios_contig() returns a batch of distinct folios.
> 
> We can remove the duplicate folio check to simplify the code and save
> 58 bytes of text.
> 
> Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
> ---
>  mm/gup.c | 15 +--------------
>  1 file changed, 1 insertion(+), 14 deletions(-)
> 
> diff --git a/mm/gup.c b/mm/gup.c
> index f32168339390..1fb8f3b9a493 100644
> --- a/mm/gup.c
> +++ b/mm/gup.c
> @@ -3589,7 +3589,7 @@ long memfd_pin_folios(struct file *memfd, loff_t
> start, loff_t end,
>  {
>  	unsigned int flags, nr_folios, nr_found;
>  	unsigned int i, pgshift = PAGE_SHIFT;
> -	pgoff_t start_idx, end_idx, next_idx;
> +	pgoff_t start_idx, end_idx;
>  	struct folio *folio = NULL;
>  	struct folio_batch fbatch;
>  	struct hstate *h;
> @@ -3639,19 +3639,7 @@ long memfd_pin_folios(struct file *memfd, loff_t
> start, loff_t end,
>  				folio = NULL;
>  			}
> 
> -			next_idx = 0;
>  			for (i = 0; i < nr_found; i++) {
> -				/*
> -				 * As there can be multiple entries for a
> -				 * given folio in the batch returned by
> -				 * filemap_get_folios_contig(), the below
> -				 * check is to ensure that we pin and return a
> -				 * unique set of folios between start and end.
> -				 */
> -				if (next_idx &&
> -				    next_idx != folio_index(fbatch.folios[i]))
> -					continue;
At that time, I believed that filemap_get_folios_contig() returning duplicate
folios was by design and not a result of a bug and hence added this workaround,
which is no longer needed in light of your fix.
Anyway, LGTM.

Acked-by: Vivek Kasireddy <vivek.kasireddy@intel.com>

Thanks,
Vivek

> -
>  				folio = page_folio(&fbatch.folios[i]->page);
> 
>  				if (try_grab_folio(folio, 1, FOLL_PIN)) {
> @@ -3664,7 +3652,6 @@ long memfd_pin_folios(struct file *memfd, loff_t
> start, loff_t end,
>  					*offset = offset_in_folio(folio, start);
> 
>  				folios[nr_folios] = folio;
> -				next_idx = folio_next_index(folio);
>  				if (++nr_folios == max_folios)
>  					break;
>  			}
> --
> 2.49.0
> 

RE: [PATCH 2/2] mm/gup: Remove page_folio() in memfd_pin_folios()

[Kasireddy, Vivek]

Hi Vishal,

> Subject: [PATCH 2/2] mm/gup: Remove page_folio() in memfd_pin_folios()
> 
> We can get the folio directly from the folio batch, so remove the
> unnecessary page_folio() call.
> 
> Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
> ---
>  mm/gup.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/gup.c b/mm/gup.c
> index 1fb8f3b9a493..795fd94f379d 100644
> --- a/mm/gup.c
> +++ b/mm/gup.c
> @@ -3640,7 +3640,7 @@ long memfd_pin_folios(struct file *memfd, loff_t
> start, loff_t end,
>  			}
> 
>  			for (i = 0; i < nr_found; i++) {
> -				folio = page_folio(&fbatch.folios[i]->page);
> +				folio = fbatch.folios[i];
I don't recall why I chose to extract the folio in a roundabout way. LGTM.

Acked-by: Vivek Kasireddy <vivek.kasireddy@intel.com>

Thanks,
Vivek

> 
>  				if (try_grab_folio(folio, 1, FOLL_PIN)) {
>  					folio_batch_release(&fbatch);
> --
> 2.49.0