[lkml]   [2011]   [Nov]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH 2/9] mm: alloc_contig_freed_pages() added
On Sun, Oct 23, 2011 at 09:05:05PM -0700, Michal Nazarewicz wrote:
> > On Thu, Oct 06, 2011 at 03:54:42PM +0200, Marek Szyprowski wrote:
> >> This commit introduces alloc_contig_freed_pages() function
> >> which allocates (ie. removes from buddy system) free pages
> >> in range. Caller has to guarantee that all pages in range
> >> are in buddy system.
> On Tue, 18 Oct 2011 05:21:09 -0700, Mel Gorman <> wrote:
> > Straight away, I'm wondering why you didn't use
> > mm/compaction.c#isolate_freepages()
> Does the below look like a step in the right direction?
> It basically moves isolate_freepages_block() to page_alloc.c (changing

For the purposes of review, have a separate patch for moving
isolate_freepages_block to another file that does not alter the
function in any way. When the function is updated in a follow-on patch,
it'll be far easier to see what has changed.

page_isolation.c may also be a better fit than page_alloc.c

As it is, the patch for isolate_freepages_block is almost impossible
to read because it's getting munged with existing code that is already
in page_alloc.c . About all I caught from it is that scannedp does
not have a type. It defaults to unsigned int but it's unnecessarily

> it name to isolate_freepages_range()) and changes it so that depending
> on arguments it treats holes (either invalid PFN or non-free page) as
> errors so that CMA can use it.

I haven't actually read the function because it's too badly mixed with
page_alloc.c code but this description fits what I'm looking for.

> It also accepts a range rather then just assuming a single pageblock
> thus the change moves range calculation in compaction.c from
> isolate_freepages_block() up to isolate_freepages().
> The change also modifies spilt_free_page() so that it does not try to
> change pageblock's migrate type if current migrate type is ISOLATE or
> CMA.

This is fine. Later, the flags that determine what happens to pageblocks
may be placed in compact_control.

> ---
> include/linux/mm.h | 1 -
> include/linux/page-isolation.h | 4 +-
> mm/compaction.c | 73 +++--------------------
> mm/internal.h | 5 ++
> mm/page_alloc.c | 128 +++++++++++++++++++++++++---------------
> 5 files changed, 95 insertions(+), 116 deletions(-)

I confess I didn't read closely because of the mess in page_alloc.c but
the intent seems fine. Hopefully there will be a new version of CMA
posted that will be easier to review.


Mel Gorman

 \ /
  Last update: 2011-11-01 16:07    [W:0.119 / U:40.104 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site