From: Rajeev Tapadia <rtapadia730@gmail.com>

All callers of alloc_bitmap() hold a transaction handle, so GFP_NOFS is
needed to avoid deadlocks on recursion. Update the comment and drop the
stale TODO.

Signed-off-by: Rajeev Tapadia <rtapadia730@gmail.com>
---
Change log:
As per previous review the change is not required. So just removing the
stale TODO.

 fs/btrfs/free-space-tree.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/fs/btrfs/free-space-tree.c b/fs/btrfs/free-space-tree.c
index dad0b492a663..bb8ca7b679be 100644
--- a/fs/btrfs/free-space-tree.c
+++ b/fs/btrfs/free-space-tree.c
@@ -165,11 +165,9 @@ static unsigned long *alloc_bitmap(u32 bitmap_size)
 
 	/*
 	 * GFP_NOFS doesn't work with kvmalloc(), but we really can't recurse
-	 * into the filesystem as the free space bitmap can be modified in the
-	 * critical section of a transaction commit.
-	 *
-	 * TODO: push the memalloc_nofs_{save,restore}() to the caller where we
-	 * know that recursion is unsafe.
+	 * into the filesystem here. All callers hold a transaction handle
+	 * open, so if a GFP_KERNEL allocation recurses into the filesystem
+	 * and triggers a transaction commit, we would deadlock.
 	 */
 	nofs_flag = memalloc_nofs_save();
 	ret = kvzalloc(bitmap_rounded_size, GFP_KERNEL);
-- 
2.51.0

On Fri, Oct 3, 2025 at 2:31 PM <rtapadia730@gmail.com> wrote:
>
> From: Rajeev Tapadia <rtapadia730@gmail.com>
>
> All callers of alloc_bitmap() hold a transaction handle, so GFP_NOFS is
> needed to avoid deadlocks on recursion. Update the comment and drop the
> stale TODO.
>
> Signed-off-by: Rajeev Tapadia <rtapadia730@gmail.com>

Reviewed-by: Filipe Manana <fdmanana@suse.com>

Thanks.
I pushed it to the for-next branch:

https://github.com/btrfs/linux/commits/for-next/

> ---
> Change log:
> As per previous review the change is not required. So just removing the
> stale TODO.
>
>  fs/btrfs/free-space-tree.c | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/fs/btrfs/free-space-tree.c b/fs/btrfs/free-space-tree.c
> index dad0b492a663..bb8ca7b679be 100644
> --- a/fs/btrfs/free-space-tree.c
> +++ b/fs/btrfs/free-space-tree.c
> @@ -165,11 +165,9 @@ static unsigned long *alloc_bitmap(u32 bitmap_size)
>
>         /*
>          * GFP_NOFS doesn't work with kvmalloc(), but we really can't recurse
> -        * into the filesystem as the free space bitmap can be modified in the
> -        * critical section of a transaction commit.
> -        *
> -        * TODO: push the memalloc_nofs_{save,restore}() to the caller where we
> -        * know that recursion is unsafe.
> +        * into the filesystem here. All callers hold a transaction handle
> +        * open, so if a GFP_KERNEL allocation recurses into the filesystem
> +        * and triggers a transaction commit, we would deadlock.
>          */
>         nofs_flag = memalloc_nofs_save();
>         ret = kvzalloc(bitmap_rounded_size, GFP_KERNEL);
> --
> 2.51.0
>