Re: [RFC][PATCH] unify all architecture PAGE_SIZE definitions

[Dave Hansen <haveblue@us.ibm.com>]

On Thu, 2006-08-24 at 17:14 -0700, William Lee Irwin III wrote:
> With CONFIG_PAGE_SHIFT there doesn't need to be a CONFIG_PAGE_SIZE at
> all; just let #define PAGE_SIZE (1 << PAGE_SHIFT) handle it. 

I think it does that:

> +#define PAGE_SHIFT      CONFIG_PAGE_SHIFT
> +#define PAGE_SIZE       (ASM_CONST(1) << PAGE_SHIFT)

Is it the Kconfig names themselves which you don't like?  I guess it is
a bit silly to have CONFIG_PAGE_SIZE_X, defined to be number Y, which is
then calculated back to being X again.  But, we were really already
doing that (in three places):

> -#if defined(CONFIG_SPARC64_PAGE_SIZE_8KB)
> -#define PAGE_SHIFT   13
> -#elif defined(CONFIG_SPARC64_PAGE_SIZE_64KB)
> -#define PAGE_SHIFT   16
> -#elif defined(CONFIG_SPARC64_PAGE_SIZE_512KB)
> -#define PAGE_SHIFT   19
> -#elif defined(CONFIG_SPARC64_PAGE_SIZE_4MB)
> -#define PAGE_SHIFT   22
> -#else
> -#error No page size specified in kernel configuration
> -#endif

But, unless users are willing to live with CONFIG_SMALLEST_PAGE,
CONFIG_SLIGHTLY_LESS_SMALL_PAGE, CONFIG_MEDIUM_SIZE_PAGE, etc... I'm not
sure what other names we can use.  Does it seem any less objectionable
to think of the Kconfig options as just the names that we _present_ to
users?  

-- Dave