Re: [PATCH v2] seq_file: fix clang warning for NULL pointer arithmetic

[Christoph Hellwig]

> diff --git a/fs/kernfs/file.c b/fs/kernfs/file.c
> index f277d023ebcd..eafeb8bf4fe4 100644
> --- a/fs/kernfs/file.c
> +++ b/fs/kernfs/file.c
> @@ -121,10 +121,10 @@ static void *kernfs_seq_start(struct seq_file *sf, loff_t *ppos)
>  		return next;
>  	} else {
>  		/*
> -		 * The same behavior and code as single_open().  Returns
> -		 * !NULL if pos is at the beginning; otherwise, NULL.
> +		 * The same behavior and code as single_open().  Continues
> +		 * if pos is at the beginning; otherwise, EOF.
>  		 */
> -		return NULL + !*ppos;
> +		return *ppos ? SEQ_OPEN_SINGLE : SEQ_OPEN_EOF;

Why the somewhat obsfucating unary expression instead of a good
old if?

e.g.

		return next;
	}
	if (*ppos)
		retun SEQ_OPEN_SINGLE;
	return NULL;


>  		++*ppos;
> -		return NULL;
> +		return SEQ_OPEN_EOF;

I don't think SEQ_OPEN_EOF is all that useful.  NULL is the documented
end case already.

> diff --git a/include/linux/seq_file.h b/include/linux/seq_file.h
> index 813614d4b71f..26f0758b6551 100644
> --- a/include/linux/seq_file.h
> +++ b/include/linux/seq_file.h
> @@ -37,6 +37,9 @@ struct seq_operations {
>  
>  #define SEQ_SKIP 1
>  
> +#define SEQ_OPEN_EOF	(void *)0
> +#define SEQ_OPEN_SINGLE	(void *)1

I think SEQ_OPEN_SINGLE also wants a comment documenting it.
AFAICS the reason for it is that ->start needs to return something
non-NULL for the seq_file code to make progress, and there is nothing
better for the single_open case.