diff mbox series

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

Message ID 20230502235622.3652586-1-ackerleytng@google.com (mailing list archive)
State Mainlined, archived
Headers show
Series fs: hugetlbfs: Set vma policy only when needed for allocating folio | expand

Commit Message

Ackerley Tng May 2, 2023, 11:56 p.m. UTC 
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(-)

Comments

Mike Kravetz May 3, 2023, 3:11 a.m. UTC | #1 
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>
diff mbox series

Patch

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)) {