lkml.org 
[lkml]   [2010]   [Sep]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    /
    Date
    From
    SubjectRe: [RFC][PATCH] big continuous memory allocator v2
    On Tue, 7 Sep 2010 01:37:27 -0700
    Minchan Kim <minchan.kim@gmail.com> wrote:

    > Nice cleanup.
    > There are some comments in below.
    >
    > On Mon, Sep 6, 2010 at 7:45 PM, KAMEZAWA Hiroyuki
    > <kamezawa.hiroyu@jp.fujitsu.com> wrote:
    > >
    > > This is a page allcoator based on memory migration/hotplug code.
    > > passed some small tests, and maybe easier to read than previous one.
    > >
    > > ==
    > > From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
    > >
    > > This patch as a memory allocator for contiguous memory larger than MAX_ORDER.
    > >
    > >  alloc_contig_pages(hint, size, node);
    >
    > I have thought this patch is to be good for dumb device drivers which
    > want big contiguous
    > memory. So if some device driver want big memory and they can tolerate
    > latency or fail,
    > this is good solution, I think.
    > And some device driver can't tolerate fail, they have to use MOVABLE zone.
    >
    > For it, I hope we have a option like ALLOC_FIXED(like MAP_FIXED).
    > That's because embedded people wanted to aware BANK of memory.
    > So if they get free page which they don't want, it can be pointless.
    >
    Okay.


    > In addition, I hope it can support CROSS_ZONE migration mode.
    > Most of small system can't support swap system. So if we can't migrate
    > anon pages into other zones, external fragment problem still happens.
    >
    Now, this code migrates pages to somewhere, including crossing zone, node etc..
    (because it just use GFP_HIGHUSER_MOVABLE)

    > I think reclaim(ex, discard file-backed pages) can become one option to prevent
    > the problem. But it's more cost so we can support it by calling mode.
    > (But it could be trivial since caller should know this function is very cost)
    >

    > ex) alloc_contig_pages(hint, size, node, ALLOC_FIXED|ALLOC_RECLAIM);
    >

    This migration's page allocation code will cause memory reclaim and
    kswapd wakeup if memory is in short. But hmm, there are no codes as

    reclaim_memory_within(start, end).

    But I guess if there are LRU pages within the range which cannot be migrated,
    they can't be dropped. In another consideration,

    shrink_slab_within(start, end)
    will be able to make success-rate better. (and this is good for memory hotplug, too)

    I'll start from adding ALLOC_FIXED.

    Thanks,
    -Kame

    --
    To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
    the body of a message to majordomo@vger.kernel.org
    More majordomo info at http://vger.kernel.org/majordomo-info.html
    Please read the FAQ at http://www.tux.org/lkml/

    \
     
     \ /
      Last update: 2010-09-07 10:55    [W:0.033 / U:59.252 seconds]
    ©2003-2016 Jasper Spaans. hosted at Digital OceanAdvertise on this site