mm/page_alloc.c: remove hand-coded get_order()

mm/page_alloc.c: remove hand-coded get_order()

[Pavel Machek]

__get_free_pages() is strange interface:

It has underscores, yet get_free_pages() does not exists.

It returns long when most people need pointer.

(And it takes order, when many people want to pass size).

What about creating void *get_free_pages(flags, order) version, then
slowly converting users to it?
									Pavel

---

Remove hand-coded get_order() from page_alloc.c.

Signed-off-by: Pavel Machek <pavel@suse.cz>

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 402a504..c48aa45 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -4339,9 +4339,7 @@ void *__init alloc_large_system_hash(con
 		else if (hashdist)
 			table = __vmalloc(size, GFP_ATOMIC, PAGE_KERNEL);
 		else {
-			unsigned long order;
-			for (order = 0; ((1UL << order) << PAGE_SHIFT) < size; order++)
-				;
+			unsigned long order = get_order(size);
 			table = (void*) __get_free_pages(GFP_ATOMIC, order);
 			/*
 			 * If bucketsize is not a power-of-two, we may free


-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: mm/page_alloc.c: remove hand-coded get_order()

[Andrew Morton]

On Tue, 15 Apr 2008 11:23:27 +0200 Pavel Machek <pavel@ucw.cz> wrote:

> 
> __get_free_pages() is strange interface:
> 
> It has underscores, yet get_free_pages() does not exists.
> 
> It returns long when most people need pointer.

It is a bit odd.

> (And it takes order, when many people want to pass size).
> 
> What about creating void *get_free_pages(flags, order) version,

Sounds sensible.  It matches free_pages().

> then slowly converting users to it?

alas, poor me.

> 
> ---
> 
> Remove hand-coded get_order() from page_alloc.c.
> 
> Signed-off-by: Pavel Machek <pavel@suse.cz>
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 402a504..c48aa45 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -4339,9 +4339,7 @@ void *__init alloc_large_system_hash(con
>  		else if (hashdist)
>  			table = __vmalloc(size, GFP_ATOMIC, PAGE_KERNEL);
>  		else {
> -			unsigned long order;
> -			for (order = 0; ((1UL << order) << PAGE_SHIFT) < size; order++)
> -				;
> +			unsigned long order = get_order(size);
>  			table = (void*) __get_free_pages(GFP_ATOMIC, order);
>  			/*
>  			 * If bucketsize is not a power-of-two, we may free
> 

OK.  get_order() does round up, doesn't it?  We seem to have forgotten to
document that rather important detail.

Re: mm/page_alloc.c: remove hand-coded get_order()

[Andrew Morton]

On Tue, 15 Apr 2008 02:35:05 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:

> > What about creating void *get_free_pages(flags, order) version,
> 
> Sounds sensible.  It matches free_pages().

No it doesn't.  free_pages() takes unsigned long.

Re: mm/page_alloc.c: remove hand-coded get_order()

[Andi Kleen]

Pavel Machek <pavel@ucw.cz> writes:
>
> What about creating void *get_free_pages(flags, order) version, then
> slowly converting users to it?

Then we would also need a free_pages() that takes void *, but unfortunately
free_pages() already has other semantics.

-Andi

Re: mm/page_alloc.c: remove hand-coded get_order()

[Christoph Lameter]

On Tue, 15 Apr 2008, Andi Kleen wrote:

> Pavel Machek <pavel@ucw.cz> writes:
> >
> > What about creating void *get_free_pages(flags, order) version, then
> > slowly converting users to it?
> 
> Then we would also need a free_pages() that takes void *, but unfortunately
> free_pages() already has other semantics.

But it would be more consistent.