xfs: convert dquot iterator map allocation to kmalloc_objs()

[PATCH v2] xfs: convert dquot iterator map allocation to kmalloc_objs()

Posted by William Theesfeld 5 days, 15 hours ago

v1 converted the open-coded kmalloc(N * sizeof(*p), ...) in
xfs_qm_reset_dqcounts_buf() to kmalloc_array().  Per review feedback
from Carlos Maiolino on the v1 thread, switch instead to the
kmalloc_objs() helper, which is the preferred form for new XFS code
and matches the style of recent conversions elsewhere in fs/xfs/.

While touching this, also drop the __GFP_NOFAIL flag and handle the
allocation failure explicitly by returning -ENOMEM.  The function
already has callers that propagate errors out of quotacheck setup,
so an early -ENOMEM is harmless and avoids relying on the allocator
to retry forever on memory pressure.

No functional change for the success path.

Signed-off-by: William Theesfeld <william@theesfeld.net>
---
 fs/xfs/xfs_qm.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c
index 7bd15d9e7..83e8d80bc 100644
--- a/fs/xfs/xfs_qm.c
+++ b/fs/xfs/xfs_qm.c
@@ -1170,8 +1170,9 @@ xfs_qm_reset_dqcounts_buf(
 	if (qip->i_nblocks == 0)
 		return 0;
 
-	map = kmalloc(XFS_DQITER_MAP_SIZE * sizeof(*map),
-			GFP_KERNEL | __GFP_NOFAIL);
+	map = kmalloc_objs(*map, XFS_DQITER_MAP_SIZE, GFP_KERNEL);
+	if (!map)
+		return -ENOMEM;
 
 	lblkno = 0;
 	maxlblkcnt = XFS_B_TO_FSB(mp, mp->m_super->s_maxbytes);
-- 
2.54.0

Re: [PATCH v2] xfs: convert dquot iterator map allocation to kmalloc_objs()

Posted by Carlos Maiolino 5 days, 13 hours ago

On Tue, Jun 02, 2026 at 07:25:08AM -0400, William Theesfeld wrote:
> v1 converted the open-coded kmalloc(N * sizeof(*p), ...) in
> xfs_qm_reset_dqcounts_buf() to kmalloc_array().  Per review feedback
> from Carlos Maiolino on the v1 thread, switch instead to the

Me? Wasn't Christoph?

> kmalloc_objs() helper, which is the preferred form for new XFS code
> and matches the style of recent conversions elsewhere in fs/xfs/.
> 
> While touching this, also drop the __GFP_NOFAIL flag and handle the
> allocation failure explicitly by returning -ENOMEM.  The function
> already has callers that propagate errors out of quotacheck setup,
> so an early -ENOMEM is harmless and avoids relying on the allocator
> to retry forever on memory pressure.
> 
> No functional change for the success path.
> 
> Signed-off-by: William Theesfeld <william@theesfeld.net>
> ---
>  fs/xfs/xfs_qm.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/xfs/xfs_qm.c b/fs/xfs/xfs_qm.c
> index 7bd15d9e7..83e8d80bc 100644
> --- a/fs/xfs/xfs_qm.c
> +++ b/fs/xfs/xfs_qm.c
> @@ -1170,8 +1170,9 @@ xfs_qm_reset_dqcounts_buf(
>  	if (qip->i_nblocks == 0)
>  		return 0;
>  
> -	map = kmalloc(XFS_DQITER_MAP_SIZE * sizeof(*map),
> -			GFP_KERNEL | __GFP_NOFAIL);
> +	map = kmalloc_objs(*map, XFS_DQITER_MAP_SIZE, GFP_KERNEL);
> +	if (!map)
> +		return -ENOMEM;
>  
>  	lblkno = 0;
>  	maxlblkcnt = XFS_B_TO_FSB(mp, mp->m_super->s_maxbytes);
> -- 
> 2.54.0
>