Re: [PATCH v1 02/12] Add and use d_for_each_positive_child family of iterators

On Mon, 3 Aug 2026 at 03:38, NeilBrown <[email protected]> wrote:

> coda_flag_children() doesn't need rcu_read_lock() as holding any
> spinlock prevents an RCU critical section from finishing.

I think prep changes should go in separate patches.

The conversion should be an equivalent transformation.

> nfs_clear_verifier_directory() is changed slightly so the parent lock is
> dropped and retaken between clearing the verifier on the parent and on
> all the children.  I think this is safe.

This can also be a prep patch.

> get_next_postive_dentry() in autofs now holds the parent lock a bit
> less.  The only non-trivial code that is no locked any more is
>    parent = p->d_parent
> This cannot need the lock as autofs doesn't support rename and
> a ref is held on a descendent of parent, so ->d_parent cannot
> change.  And if it could, the lock wouldn't help.

And this too.

> diff --git a/fs/autofs/expire.c b/fs/autofs/expire.c
> index 5c2d459e1e48..7b0a22629415 100644
> --- a/fs/autofs/expire.c
> +++ b/fs/autofs/expire.c
> @@ -70,12 +70,9 @@ static int autofs_mount_busy(struct vfsmount *mnt,
>         return status;
>  }
>
> -/* p->d_lock held */
>  static struct dentry *positive_after(struct dentry *p, struct dentry *child)

Move this primitive to libfs.c?

> --- a/fs/ceph/mds_client.c
> +++ b/fs/ceph/mds_client.c
> @@ -2182,14 +2182,10 @@ static bool drop_negative_children(struct dentry *dentry)
>         if (!d_is_dir(dentry))
>                 goto out;
>
> -       spin_lock(&dentry->d_lock);
> -       hlist_for_each_entry(child, &dentry->d_children, d_sib) {
> -               if (d_really_is_positive(child)) {
> -                       all_negative = false;
> -                       break;
> -               }
> +       d_for_each_positive_child(child, dentry) {
> +               all_negative = false;
> +               break;
>         }
> -       spin_unlock(&dentry->d_lock);

simple_empty()?

Thanks,
Miklos