Re: [patch] add an iterator index in struct pagevec

[Andrew Morton <akpm@osdl.org>]

On Mon, 4 Dec 2006 21:21:31 -0800
"Chen, Kenneth W" <kenneth.w.chen@intel.com> wrote:

> pagevec is never expected to be more than PAGEVEC_SIZE, I think a
> unsigned char is enough to count them.  This patch makes nr, cold
> to be unsigned char

Is that on the right side of the speed/space tradeoff?

> and also adds an iterator index. With that,
> the size can be even bumped up by 1 to 15.
> 
> Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
> 
> 
> diff -Nurp linux-2.6.19/include/linux/pagevec.h linux-2.6.19.ken/include/linux/pagevec.h
> --- linux-2.6.19/include/linux/pagevec.h	2006-11-29 13:57:37.000000000 -0800
> +++ linux-2.6.19.ken/include/linux/pagevec.h	2006-12-04 19:18:21.000000000 -0800
> @@ -8,15 +8,16 @@
>  #ifndef _LINUX_PAGEVEC_H
>  #define _LINUX_PAGEVEC_H
>  
> -/* 14 pointers + two long's align the pagevec structure to a power of two */
> -#define PAGEVEC_SIZE	14
> +/* 15 pointers + 3 char's align the pagevec structure to a power of two */
> +#define PAGEVEC_SIZE	15
>  
>  struct page;
>  struct address_space;
>  
>  struct pagevec {
> -	unsigned long nr;
> -	unsigned long cold;
> +	unsigned char nr;
> +	unsigned char cold;
> +	unsigned char idx;
>  	struct page *pages[PAGEVEC_SIZE];
>  };
>  

I'd have thought that pagevec_init() would want to be involved in this, no?

I must say I'm a bit skeptical about the need for this.  But I haven't
looked closely at the blockdev-specific dio code yet.