Message-ID: <op.v3fufkaw3l0zgt@mpn-glaptop>
Date:	Sun, 16 Oct 2011 11:39:10 +0200
From:	"Michal Nazarewicz" <mina86@...a86.com>
To:	"Andrew Morton" <akpm@...ux-foundation.org>
Cc:	"Marek Szyprowski" <m.szyprowski@...sung.com>,
	linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
	linux-media@...r.kernel.org, linux-mm@...ck.org,
	linaro-mm-sig@...ts.linaro.org,
	"Kyungmin Park" <kyungmin.park@...sung.com>,
	"Russell King" <linux@....linux.org.uk>,
	"KAMEZAWA Hiroyuki" <kamezawa.hiroyu@...fujitsu.com>,
	"Ankita Garg" <ankita@...ibm.com>,
	"Daniel Walker" <dwalker@...eaurora.org>,
	"Mel Gorman" <mel@....ul.ie>, "Arnd Bergmann" <arnd@...db.de>,
	"Jesse Barker" <jesse.barker@...aro.org>,
	"Jonathan Corbet" <corbet@....net>,
	"Shariq Hasnain" <shariq.hasnain@...aro.org>,
	"Chunsang Jeong" <chunsang.jeong@...aro.org>,
	"Dave Hansen" <dave@...ux.vnet.ibm.com>
Subject: Re: [PATCH 2/9] mm: alloc_contig_freed_pages() added

> On Sun, 16 Oct 2011 10:01:36 +0200 "Michal Nazarewicz" wrote:
>> Still, as I think of it now, maybe alloc_contig_free_range() would be
>> better?

On Sun, 16 Oct 2011 10:31:16 +0200, Andrew Morton wrote:
> Nope.  Of *course* the pages were free.  Otherwise we couldn't
> (re)allocate them.  I still think the "free" part is redundant.

Makes sense.

> What could be improved is the "alloc" part.  This really isn't an
> allocation operation.  The pages are being removed from buddy then
> moved into the free arena of a different memory manager from where they
> will _later_ be "allocated".

Not quite.  After alloc_contig_range() returns, the pages are passed with
no further processing to the caller.  Ie. the area is not later split into
several parts nor kept in CMA's pool unused.

alloc_contig_freed_pages() is a little different since it must be called on
a buddy page boundary and may return more then requested (because of the way
buddy system merges buddies) so there is a little processing after it returns
(namely freeing of the excess pages).

> So we should move away from the alloc/free naming altogether for this
> operation and think up new terms.  How about "claim" and "release"?
> claim_contig_pages, claim_contig_range, release_contig_pages, etc?
> Or we could use take/return.

Personally, I'm not convinced about changing the names of alloc_contig_range()
and free_contig_pages() but I see merit in changing alloc_contig_freed_pages()
to something else.

Since at the moment, it's used only by alloc_contig_range(), I'd lean
towards removing it from page-isolation.h, marking as static and renaming
to __alloc_contig_range().

> Also, if we have no expectation that anything apart from CMA will use
> these interfaces (?), the names could/should be prefixed with "cma_".

In Kamezawa's original patchset, he used those for a bit different
approach (IIRC, Kamezawa's patchset introduced a function that scanned memory
and tried to allocate contiguous memory where it could), so I can imagine that
someone will make use of those functions.  It may be used in any situation
where a range of pages is either free (ie. in buddy system) or movable and
one wants to allocate them for some reason.

-- 
Best regards,                                         _     _
.o. | Liege of Serenely Enlightened Majesty of      o' \,=./ `o
..o | Computer Science,  Michal "mina86" Nazarewicz    (o o)
ooo +--<mina86@...a86.com>---<mina86@...ber.org>---ooO--(_)--Ooo--
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists