Re: [PATCH -mm 1/5] list.h: add list_singleton

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 14 Mar 2008 16:40:36 -0400
Masami Hiramatsu <mhiramat@redhat.com> wrote:

> Add list_singleton to check a list has just one entry.
> 
> list_singleton is useful to check whether a list_head which
> have been temporarily allocated for listing objects can be
> released or not.
> 
> Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com>
> ---
>  include/linux/list.h |    9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> Index: 2.6.25-rc5-mm1/include/linux/list.h
> ===================================================================
> --- 2.6.25-rc5-mm1.orig/include/linux/list.h
> +++ 2.6.25-rc5-mm1/include/linux/list.h
> @@ -211,6 +211,15 @@ static inline int list_empty_careful(con
>  	return (next == head) && (next == head->prev);
>  }
> 
> +/**
> + * list_singleton - tests whether a list has just one entry.
> + * @head: the list to test.
> + */
> +static inline int list_singleton(const struct list_head *head)
> +{
> +	return !list_empty(head) && (head->next == head->prev);
> +}
> +

This hurts my brain.

If your usage pattern is:

struct foo {
	...
	struct list_head bar_list;	/* A list of `struct bar's */
};

struct bar {
	struct list_head list;		/* Attached to foo.bar_list */
	...
};

then yes, list_singleton() makes sense.

But in other usage patterns it does not:

struct foo {
	struct bar *bar_list;
	...
};

struct bar {
	struct list_head list;		/* All the other bars go here */
	...
};

In the second case, emptiness is signified by foo.bar_list==NULL.  And in
this case, code which does

	if (foo->bar_list && list_singleton(&foo->bar_list->list))

will fail if there is a single item on the list!

The second usage pattern is uncommon and list_empty() also returns
misleading answers when list_heads are used this way.

So I guess we can proceed with your list_singleton(), but I'd just like to
flag this possible confusion, see what people think..