linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de>
To: Andrea Arcangeli <andrea@suse.de>
Cc: Goswin von Brederlow <brederlo@informatik.uni-tuebingen.de>,
	Andrew Morton <akpm@linux-foundation.org>,
	Joern Engel <joern@logfs.org>,
	Nick Piggin <nickpiggin@yahoo.com.au>,
	Christoph Lameter <clameter@sgi.com>,
	torvalds@linux-foundation.org, linux-fsdevel@vger.kernel.org,
	linux-kernel@vger.kernel.org, Christoph Hellwig <hch@lst.de>,
	Mel Gorman <mel@skynet.ie>,
	William Lee Irwin III <wli@holomorphy.com>,
	David Chinner <dgc@sgi.com>, Jens Axboe <jens.axboe@oracle.com>,
	Badari Pulavarty <pbadari@gmail.com>,
	Maxim Levitsky <maximlevitsky@gmail.com>,
	Fengguang Wu <fengguang.wu@gmail.com>,
	swin wang <wangswin@gmail.com>,
	totty.lu@gmail.com, hugh@veritas.com
Subject: Re: [00/41] Large Blocksize Support V7 (adds memmap support)
Date: Sun, 16 Sep 2007 15:54:56 +0200	[thread overview]
Message-ID: <87d4wi3ebz.fsf@informatik.uni-tuebingen.de> (raw)
In-Reply-To: <20070915223032.GA6708@v2.random> (Andrea Arcangeli's message of "Sun, 16 Sep 2007 00:30:32 +0200")

Andrea Arcangeli <andrea@suse.de> writes:

> On Sat, Sep 15, 2007 at 10:14:44PM +0200, Goswin von Brederlow wrote:
>> - Userspace allocates a lot of memory in those slabs.
>
> If with slabs you mean slab/slub, I can't follow, there has never been
> a single byte of userland memory allocated there since ever the slab
> existed in linux.

This and other comments in your reply show me that you completly
misunderstood what I was talking about.

Look at
http://www.skynet.ie/~mel/anti-frag/2007-02-28/page_type_distribution.jpg

The red dots (pinned) are dentries, page tables, kernel stacks,
whatever kernel stuff, right?

The green dots (movable) are mostly userspace pages being mapped
there, right?

What I was refering too is that because movable objects (green dots)
aren't moved out of a mixed group (the boxes) when some unmovable
object needs space all the groups become mixed over time. That means
the unmovable objects are spread out over all the ram and the buddy
system can't recombine regions when unmovable objects free them. There
will nearly always be some movable objects in the other buddy. The
system of having unmovable and movable groups breaks down and becomes
useless.


I'm assuming here that we want the possibility of larger order pages
for unmovable objects (large continiuos regions for DMA for example)
than the smallest order user space gets (or any movable object). If
mmap() still works on 4k page bounaries then those will fragment all
regions into 4k chunks in the worst case.

Obviously if userspace has a minimum order of 64k chunks then it will
never break any region smaller than 64k chunks and will never cause a
fragmentation catastroph. I know that is verry roughly your aproach
(make order 0 bigger), and I like it, but it has some limits as to how
big you can make it. I don't think my system with 1GB ram would work
so well with 2MB order 0 pages. But I wasn't refering to that but to
the picture.

MfG
        Goswin

  reply	other threads:[~2007-09-16 13:55 UTC|newest]

Thread overview: 187+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2007-09-11  6:03 [00/41] Large Blocksize Support V7 (adds memmap support) Christoph Lameter
2007-09-10 18:52 ` Nick Piggin
2007-09-11 12:05   ` Andrea Arcangeli
2007-09-11 20:03     ` Christoph Lameter
2007-09-11 12:12   ` Jörn Engel
2007-09-10 21:13     ` Nick Piggin
2007-09-11 16:02       ` Goswin von Brederlow
2007-09-11 20:07     ` Christoph Lameter
2007-09-11 20:29       ` Jörn Engel
2007-09-11 20:41         ` Christoph Lameter
2007-09-11 23:26           ` Andrea Arcangeli
2007-09-12  0:04             ` Christoph Lameter
2007-09-12  8:20               ` Andrea Arcangeli
2007-09-15  8:44     ` Andrew Morton
2007-09-15 12:14       ` Goswin von Brederlow
2007-09-15 15:51         ` Andrea Arcangeli
2007-09-15 20:14           ` Goswin von Brederlow
2007-09-15 22:30             ` Andrea Arcangeli
2007-09-16 13:54               ` Goswin von Brederlow [this message]
2007-09-16 15:08                 ` Andrea Arcangeli
2007-09-16 21:08                   ` Mel Gorman
2007-09-16 22:48                     ` Goswin von Brederlow
2007-09-17  9:30                       ` Mel Gorman
2007-09-16 17:46               ` Jörn Engel
2007-09-16 18:15                 ` Linus Torvalds
2007-09-16 18:21                   ` Jörn Engel
2007-09-16 18:44                     ` Linus Torvalds
2007-09-16 22:51                       ` Goswin von Brederlow
2007-09-23 17:44                       ` Jörn Engel
2007-09-16 22:06                 ` Goswin von Brederlow
2007-09-16 22:40                   ` Jörn Engel
2007-09-16 18:15           ` Mel Gorman
2007-09-16 18:50             ` Andrea Arcangeli
2007-09-16 20:54               ` Mel Gorman
2007-09-16 21:31                 ` Andrea Arcangeli
2007-09-17 10:13                   ` Mel Gorman
2007-09-23  5:50                     ` Goswin von Brederlow
2007-09-16 22:56               ` Goswin von Brederlow
2007-09-18 19:31                 ` Andrea Arcangeli
2007-09-23  6:56                   ` Goswin von Brederlow
2007-09-24 15:39                     ` Andrea Arcangeli
2007-09-16 18:13         ` Mel Gorman
2007-09-16  9:03           ` Nick Piggin
2007-09-17 22:00             ` Christoph Lameter
2007-09-18  0:11               ` Nick Piggin
2007-09-18 20:36                 ` Christoph Lameter
2007-09-18 10:00               ` Mel Gorman
2007-09-18 10:49                 ` Jörn Engel
2007-09-18 12:31                 ` David Chinner
2007-09-16 21:58           ` Goswin von Brederlow
2007-09-17 10:03             ` Mel Gorman
2007-09-23  6:22               ` Goswin von Brederlow
2007-09-24 12:32                 ` Kyle Moffett
2007-09-16 17:53       ` Jörn Engel
2007-09-16 21:31         ` Mel Gorman
2007-09-17 22:03         ` Christoph Lameter
2007-09-11 15:36   ` Mel Gorman
2007-09-11  1:44     ` Nick Piggin
2007-09-11 20:11       ` Christoph Lameter
2007-09-11  4:53         ` Nick Piggin
2007-09-11 20:42           ` Christoph Lameter
2007-09-11  5:30             ` Nick Piggin
2007-09-11 21:41               ` Christoph Lameter
2007-09-11  6:06                 ` Nick Piggin
2007-09-11 21:52                   ` Christoph Lameter
2007-09-11 18:07                     ` Nick Piggin
2007-09-12 23:06                       ` Christoph Lameter
2007-09-13 20:51                         ` Nick Piggin
2007-09-14 17:52                           ` Christoph Lameter
2007-09-16  8:22                             ` Nick Piggin
2007-09-17 22:05                               ` Christoph Lameter
2007-09-18  0:10                                 ` Nick Piggin
2007-09-18 20:42                                   ` Christoph Lameter
2007-09-17 11:10                         ` Bernd Schmidt
2007-09-17 22:10                           ` Christoph Lameter
2007-09-14 16:10                       ` Goswin von Brederlow
2007-09-14 17:42                         ` Mel Gorman
2007-09-15  0:31                           ` Goswin von Brederlow
2007-09-16 21:16                             ` Mel Gorman
2007-09-16 22:38                               ` Goswin von Brederlow
2007-09-17  8:57                                 ` Mel Gorman
2007-09-23  6:49                                   ` Goswin von Brederlow
2007-09-11 20:53       ` Mel Gorman
2007-09-11  6:00         ` Nick Piggin
2007-09-11 21:48           ` Christoph Lameter
2007-09-11  6:17             ` Nick Piggin
2007-09-12  0:00               ` Christoph Lameter
2007-09-12  2:46                 ` Nick Piggin
2007-09-12 23:17                   ` Christoph Lameter
2007-09-13  9:40                     ` Mel Gorman
2007-09-14  2:38                       ` Christoph Lameter
2007-09-13 21:20                     ` Nick Piggin
2007-09-14 18:08                       ` Christoph Lameter
2007-09-14 18:15                         ` Christoph Lameter
2007-09-15  0:33                           ` Goswin von Brederlow
2007-09-16  8:53                         ` Nick Piggin
2007-09-17 22:21                           ` Christoph Lameter
2007-09-18  1:16                             ` Nick Piggin
2007-09-18 18:30                               ` Linus Torvalds
2007-09-18 17:53                                 ` Nick Piggin
2007-09-18 19:18                                 ` Andrea Arcangeli
2007-09-18 19:44                                   ` Linus Torvalds
2007-09-19  0:58                                     ` Nathan Scott
2007-09-19  1:06                                       ` Linus Torvalds
2007-09-19  2:45                                         ` Nathan Scott
2007-09-19  5:09                                         ` David Chinner
2007-09-19  9:41                                           ` Alex Tomas
2007-09-19 14:04                                           ` Andrea Arcangeli
2007-09-20  1:38                                             ` David Chinner
2007-09-20 14:54                                               ` Andrea Arcangeli
2007-09-20 18:11                                                 ` Christoph Lameter
2007-09-20 18:07                                               ` Christoph Lameter
2007-09-21 20:41                                                 ` Hugh Dickins
2007-09-24 21:13                                                   ` Christoph Lameter
2007-09-28  2:46                                               ` Nick Piggin
2007-09-19  3:41                                     ` Rene Herman
2007-09-19  3:50                                       ` Linus Torvalds
2007-09-19  4:26                                         ` Rene Herman
2007-09-19  4:33                                           ` Linus Torvalds
2007-09-19  4:56                                             ` Rene Herman
2007-09-11 21:54             ` Mel Gorman
2007-09-12 14:29             ` Martin J. Bligh
2007-09-12  1:49           ` David Chinner
2007-09-11 15:27             ` Nick Piggin
2007-09-13  1:49               ` David Chinner
2007-09-12 17:23                 ` Nick Piggin
2007-09-13 13:03                   ` David Chinner
2007-09-13  2:01                     ` Nick Piggin
2007-09-13 20:48                       ` Nick Piggin
2007-09-17  4:07                         ` David Chinner
2007-09-16 21:13                           ` Nick Piggin
2007-09-12  2:01             ` Nick Piggin
2007-09-11 21:35         ` Christoph Lameter
2007-09-11 16:47     ` Andrea Arcangeli
2007-09-11 18:31       ` Mel Gorman
2007-09-11  2:26         ` Nick Piggin
2007-09-11 18:25           ` Maxim Levitsky
2007-09-11  3:05             ` Nick Piggin
2007-09-11 21:03           ` Mel Gorman
2007-09-11 19:20         ` Andrea Arcangeli
2007-09-11 20:19           ` Jörn Engel
2007-09-11 20:13       ` Christoph Lameter
2007-09-11 20:01   ` Christoph Lameter
2007-09-11  4:43     ` Nick Piggin
2007-09-11  5:17     ` Nick Piggin
2007-09-11 21:27       ` Mel Gorman
2007-09-11  6:03 ` [01/41] Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user Christoph Lameter
2007-09-11  6:03 ` [02/41] Define functions for page cache handling Christoph Lameter
2007-09-11  6:03 ` [03/41] Use page_cache_xxx functions in mm/filemap.c Christoph Lameter
2007-09-11  6:03 ` [04/41] Use page_cache_xxx in mm/page-writeback.c Christoph Lameter
2007-09-11  6:03 ` [05/41] Use page_cache_xxx in mm/truncate.c Christoph Lameter
2007-09-11  6:03 ` [06/41] Use page_cache_xxx in mm/rmap.c Christoph Lameter
2007-09-11  6:03 ` [07/41] Use page_cache_xxx in mm/filemap_xip.c Christoph Lameter
2007-09-11  6:03 ` [08/41] Use page_cache_xxx in mm/migrate.c Christoph Lameter
2007-09-11  6:03 ` [09/41] Use page_cache_xxx in fs/libfs.c Christoph Lameter
2007-09-11  6:04 ` [10/41] Use page_cache_xxx in fs/sync Christoph Lameter
2007-09-11  6:04 ` [11/41] Use page_cache_xxx in fs/buffer.c Christoph Lameter
2007-09-11  6:04 ` [12/41] Use page_cache_xxx in mm/mpage.c Christoph Lameter
2007-09-11  6:04 ` [13/41] Use page_cache_xxx in mm/fadvise.c Christoph Lameter
2007-09-11  6:04 ` [14/41] Use page_cache_xxx in fs/splice.c Christoph Lameter
2007-09-11  6:04 ` [15/41] Use page_cache_xxx in ext2 Christoph Lameter
2007-09-11  6:04 ` [16/41] Use page_cache_xxx in fs/ext3 Christoph Lameter
2007-09-11  6:04 ` [17/41] Use page_cache_xxx in fs/ext4 Christoph Lameter
2007-09-11  6:04 ` [18/41] Use page_cache_xxx in fs/reiserfs Christoph Lameter
2007-09-11  6:04 ` [19/41] Use page_cache_xxx for fs/xfs Christoph Lameter
2007-09-11  6:04 ` [20/41] Use page_cache_xxx in drivers/block/rd.c Christoph Lameter
2007-09-11  6:04 ` [21/41] compound pages: Better PageHead/PageTail handling Christoph Lameter
2007-09-11  6:04 ` [22/41] compound pages: Add new support functions Christoph Lameter
2007-09-11  6:04 ` [23/41] compound pages: vmstat support Christoph Lameter
2007-09-11  6:04 ` [24/41] compound pages: Use new compound vmstat functions in SLUB Christoph Lameter
2007-09-11  6:04 ` [25/41] compound pages: Allow use of get_page_unless_zero with compound pages Christoph Lameter
2007-09-11  6:04 ` [26/41] compound pages: Allow freeing of compound pages via pagevec Christoph Lameter
2007-09-11  6:04 ` [27/41] Large page order operations, zeroing and flushing Christoph Lameter
2007-09-11  6:04 ` [28/41] Futex: Fix PAGE SIZE assumption Christoph Lameter
2007-09-11  6:04 ` [29/41] Fix up reclaim counters Christoph Lameter
2007-09-11  6:04 ` [30/41] Add VM_BUG_ONs to check for correct page order Christoph Lameter
2007-09-11  6:04 ` [31/41] Large Blocksize: Core piece Christoph Lameter
2007-09-11  6:04 ` [32/41] Readahead changes to support large blocksize Christoph Lameter
2007-09-11  6:04 ` [33/41] Large blocksize support in ramfs Christoph Lameter
2007-09-11  6:04 ` [34/41] Large blocksize support for XFS Christoph Lameter
2007-09-11  6:04 ` [35/41] Reiserfs: Fix up mapping_set_gfp_mask() Christoph Lameter
2007-09-11  6:04 ` [36/41] 64k block size support for Ext2/3/4 Christoph Lameter
2007-09-11  6:04 ` [37/41] ext2: fix rec_len overflow for 64KB block size Christoph Lameter
2007-09-11  6:04 ` [38/41] ext3: fix rec_len overflow with " Christoph Lameter
2007-09-11  6:04 ` [39/41] ext4: fix rec_len overflow for " Christoph Lameter
2007-09-11  6:04 ` [40/41] Do not use f_mapping in simple_prepare_write() Christoph Lameter
2007-09-11  6:04 ` [41/41] Mmap support using pte PAGE_SIZE mappings Christoph Lameter

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87d4wi3ebz.fsf@informatik.uni-tuebingen.de \
    --to=brederlo@informatik.uni-tuebingen.de \
    --cc=akpm@linux-foundation.org \
    --cc=andrea@suse.de \
    --cc=clameter@sgi.com \
    --cc=dgc@sgi.com \
    --cc=fengguang.wu@gmail.com \
    --cc=hch@lst.de \
    --cc=hugh@veritas.com \
    --cc=jens.axboe@oracle.com \
    --cc=joern@logfs.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=maximlevitsky@gmail.com \
    --cc=mel@skynet.ie \
    --cc=nickpiggin@yahoo.com.au \
    --cc=pbadari@gmail.com \
    --cc=torvalds@linux-foundation.org \
    --cc=totty.lu@gmail.com \
    --cc=wangswin@gmail.com \
    --cc=wli@holomorphy.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).