[PATCH] fs: hugetlbfs: Set vma policy only when needed for allocating folio

[PATCH] fs: hugetlbfs: Set vma policy only when needed for allocating folio

[Ackerley Tng]

Calling hugetlb_set_vma_policy() later avoids setting the vma policy
and then dropping it on a page cache hit.

Signed-off-by: Ackerley Tng <ackerleytng@google.com>
---
 fs/hugetlbfs/inode.c | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
index f640cff1bbce..380b1cd6c93f 100644
--- a/fs/hugetlbfs/inode.c
+++ b/fs/hugetlbfs/inode.c
@@ -833,9 +833,6 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
 			break;
 		}
 
-		/* Set numa allocation policy based on index */
-		hugetlb_set_vma_policy(&pseudo_vma, inode, index);
-
 		/* addr is the offset within the file (zero based) */
 		addr = index * hpage_size;
 
@@ -846,7 +843,6 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
 		/* See if already present in mapping to avoid alloc/free */
 		if (filemap_has_folio(mapping, index)) {
 			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
-			hugetlb_drop_vma_policy(&pseudo_vma);
 			continue;
 		}
 
@@ -858,6 +854,7 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
 		 * folios in these areas, we need to consume the reserves
 		 * to keep reservation accounting consistent.
 		 */
+		hugetlb_set_vma_policy(&pseudo_vma, inode, index);
 		folio = alloc_hugetlb_folio(&pseudo_vma, addr, 0);
 		hugetlb_drop_vma_policy(&pseudo_vma);
 		if (IS_ERR(folio)) {
-- 
2.40.1.495.gc816e09b53d-goog

Re: [PATCH] fs: hugetlbfs: Set vma policy only when needed for allocating folio

[Mike Kravetz]

On 05/02/23 23:56, Ackerley Tng wrote:
> Calling hugetlb_set_vma_policy() later avoids setting the vma policy
> and then dropping it on a page cache hit.
> 
> Signed-off-by: Ackerley Tng <ackerleytng@google.com>
> ---
>  fs/hugetlbfs/inode.c | 5 +----
>  1 file changed, 1 insertion(+), 4 deletions(-)

I suspect you noticed this when debugging the issue with using
page_cache_next_miss.  :)

Thanks!

Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
-- 
Mike Kravetz

> 
> diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c
> index f640cff1bbce..380b1cd6c93f 100644
> --- a/fs/hugetlbfs/inode.c
> +++ b/fs/hugetlbfs/inode.c
> @@ -833,9 +833,6 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
>  			break;
>  		}
>  
> -		/* Set numa allocation policy based on index */
> -		hugetlb_set_vma_policy(&pseudo_vma, inode, index);
> -
>  		/* addr is the offset within the file (zero based) */
>  		addr = index * hpage_size;
>  
> @@ -846,7 +843,6 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
>  		/* See if already present in mapping to avoid alloc/free */
>  		if (filemap_has_folio(mapping, index)) {
>  			mutex_unlock(&hugetlb_fault_mutex_table[hash]);
> -			hugetlb_drop_vma_policy(&pseudo_vma);
>  			continue;
>  		}
>  
> @@ -858,6 +854,7 @@ static long hugetlbfs_fallocate(struct file *file, int mode, loff_t offset,
>  		 * folios in these areas, we need to consume the reserves
>  		 * to keep reservation accounting consistent.
>  		 */
> +		hugetlb_set_vma_policy(&pseudo_vma, inode, index);
>  		folio = alloc_hugetlb_folio(&pseudo_vma, addr, 0);
>  		hugetlb_drop_vma_policy(&pseudo_vma);
>  		if (IS_ERR(folio)) {
> -- 
> 2.40.1.495.gc816e09b53d-goog
>