From: Andi Kleen <ak@suse.de>
To: Ray Bryant <raybry@sgi.com>
Cc: Andi Kleen <ak@suse.de>, Paul Jackson <pj@sgi.com>,
ak@muc.de, raybry@austin.rr.com, linux-mm@kvack.org,
linux-kernel@vger.kernel.org, Dave Hansen <haveblue@us.ibm.com>
Subject: Re: [RFC 2.6.11-rc2-mm2 0/7] mm: manual page migration -- overview II
Date: Mon, 21 Feb 2005 13:10:11 +0100 [thread overview]
Message-ID: <20050221121010.GC17667@wotan.suse.de> (raw)
In-Reply-To: <42199EE8.9090101@sgi.com>
On Mon, Feb 21, 2005 at 02:42:16AM -0600, Ray Bryant wrote:
> All,
>
> Just an update on the idea of migrating a process without suspending
> it.
>
> The hard part of the problem here is to make sure that the page_migrate()
> system call sees all of the pages to migrate. If the process that is
> being migrated can still allocate pages, then the page_migrate() call
> may miss some of the pages.
I would do an easy 95% solution:
When process has default process policy set temporarily a prefered policy
with the new node
[this won't work with multiple nodes though, so you have to decide on one
or stop the process if that is unacceptable]
>
> One way to solve this problem is to force the process to start allocating
> pages on the new nodes before calling page_migrate(). There are a couple
> of subcases:
>
> (1) For memory mapped files with a non-DEFAULT associated memory policy,
> one can use mbind() to fixup the memory policy. (This assumes the
> Steve Longerbeam patches are applied, as I understand things).
I would just ignore them. If user space wants it can handle it,
but it's probably not worth it.
> (1) could be handled as part of the page_migrate() system call --
> make one pass through the address space searching for mempolicy()
> data structures, and updating them as necessary. Then make a second
> pass through and do the migrations. Any new allocations will then
> be done under the new mempolicy, so they won't be missed. But this
> still gets us into trouble if the old and new node lists are not
> disjoint.
I wouldn't bother fixing up VMA policies.
> This doesn't handle anonymous memory or mapped files associated with
> the DEFAULT policy. A way around that would be to add a target cpu_id
[...]
I would set temporarily a prefered policy as mentioned above.
That only handles a single node, but you solution is not better.
-Andi
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>
next prev parent reply other threads:[~2005-02-21 12:10 UTC|newest]
Thread overview: 103+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-02-12 3:25 [RFC 2.6.11-rc2-mm2 0/7] mm: manual page migration -- overview Ray Bryant
2005-02-12 3:25 ` [RFC 2.6.11-rc2-mm2 1/7] mm: manual page migration -- cleanup 1 Ray Bryant
2005-02-12 3:25 ` [RFC 2.6.11-rc2-mm2 2/7] mm: manual page migration -- cleanup 2 Ray Bryant
2005-02-12 3:25 ` [RFC 2.6.11-rc2-mm2 3/7] mm: manual page migration -- cleanup 3 Ray Bryant
2005-02-12 3:26 ` [RFC 2.6.11-rc2-mm2 4/7] mm: manual page migration -- cleanup 4 Ray Bryant
2005-02-12 3:26 ` [RFC 2.6.11-rc2-mm2 5/7] mm: manual page migration -- cleanup 5 Ray Bryant
2005-02-12 3:26 ` [RFC 2.6.11-rc2-mm2 6/7] mm: manual page migration -- add node_map arg to try_to_migrate_pages() Ray Bryant
2005-02-12 3:26 ` [RFC 2.6.11-rc2-mm2 7/7] mm: manual page migration -- sys_page_migrate Ray Bryant
2005-02-12 8:08 ` Paul Jackson
2005-02-12 12:34 ` Arjan van de Ven
2005-02-12 14:48 ` Andi Kleen
2005-02-12 20:51 ` Paul Jackson
2005-02-12 21:04 ` Dave Hansen
2005-02-12 21:44 ` Paul Jackson
2005-02-14 13:52 ` Robin Holt
2005-02-14 18:50 ` Dave Hansen
2005-02-14 22:01 ` Robin Holt
2005-02-14 22:22 ` Dave Hansen
2005-02-15 10:50 ` Robin Holt
2005-02-15 15:38 ` Paul Jackson
2005-02-15 18:39 ` Dave Hansen
2005-02-15 18:54 ` Ray Bryant
2005-02-15 15:49 ` Paul Jackson
2005-02-15 16:21 ` Robin Holt
2005-02-15 16:35 ` Paul Jackson
2005-02-15 18:59 ` Robin Holt
2005-02-15 20:54 ` Dave Hansen
2005-02-15 21:58 ` Peter Chubb
2005-02-15 22:10 ` Paul Jackson
2005-02-15 22:51 ` Robin Holt
2005-02-15 23:00 ` Paul Jackson
2005-02-15 23:21 ` Ray Bryant
2005-02-15 23:51 ` Martin J. Bligh
2005-02-16 0:38 ` Ray Bryant
2005-02-16 0:44 ` Andi Kleen
2005-02-16 0:54 ` Martin J. Bligh
2005-02-16 10:02 ` Andi Kleen
2005-02-16 15:21 ` Martin J. Bligh
2005-02-16 15:49 ` Paul Jackson
2005-02-16 16:08 ` Andi Kleen
2005-02-16 16:55 ` Martin J. Bligh
2005-02-16 23:35 ` Ray Bryant
2005-02-16 0:50 ` Martin J. Bligh
2005-02-15 15:40 ` Paul Jackson
2005-02-12 11:17 ` [RFC 2.6.11-rc2-mm2 0/7] mm: manual page migration -- overview Andi Kleen
2005-02-12 12:12 ` Robin Holt
2005-02-14 19:18 ` Andi Kleen
2005-02-15 1:02 ` Steve Longerbeam
2005-02-12 15:54 ` Marcelo Tosatti
2005-02-12 16:18 ` Marcelo Tosatti
2005-02-12 21:29 ` Andi Kleen
2005-02-14 16:38 ` Robin Holt
2005-02-14 19:15 ` Andi Kleen
2005-02-14 23:49 ` Ray Bryant
2005-02-15 3:16 ` Paul Jackson
2005-02-15 9:14 ` Ray Bryant
2005-02-15 15:21 ` Paul Jackson
2005-02-15 0:29 ` Ray Bryant
2005-02-15 11:05 ` Robin Holt
2005-02-15 17:44 ` Ray Bryant
2005-02-15 11:53 ` Andi Kleen
2005-02-15 12:15 ` Robin Holt
2005-02-15 15:07 ` Paul Jackson
2005-02-15 15:11 ` Paul Jackson
2005-02-15 18:16 ` Ray Bryant
2005-02-15 18:24 ` Andi Kleen
2005-02-15 12:14 ` [RFC 2.6.11-rc2-mm2 0/7] mm: manual page migration -- overview II Andi Kleen
2005-02-15 18:38 ` Ray Bryant
2005-02-15 21:48 ` Andi Kleen
2005-02-15 22:37 ` Paul Jackson
2005-02-16 3:44 ` Ray Bryant
2005-02-17 23:54 ` Andi Kleen
2005-02-18 8:38 ` Ray Bryant
2005-02-18 13:02 ` Andi Kleen
2005-02-18 16:18 ` Paul Jackson
2005-02-18 16:20 ` Paul Jackson
2005-02-18 16:22 ` Paul Jackson
2005-02-18 16:25 ` Paul Jackson
2005-02-19 1:01 ` Ray Bryant
2005-02-20 21:49 ` Andi Kleen
2005-02-20 22:30 ` Paul Jackson
2005-02-20 22:35 ` Andi Kleen
2005-02-21 1:50 ` Paul Jackson
2005-02-21 7:39 ` Ray Bryant
2005-02-21 7:29 ` Ray Bryant
2005-02-21 9:57 ` Andi Kleen
2005-02-21 12:02 ` Paul Jackson
2005-02-21 8:42 ` Ray Bryant
2005-02-21 12:10 ` Andi Kleen [this message]
2005-02-21 17:12 ` Ray Bryant
2005-02-22 18:03 ` Andi Kleen
2005-02-23 3:33 ` Ray Bryant
2005-02-22 6:40 ` Ray Bryant
2005-02-22 18:01 ` Andi Kleen
2005-02-22 18:45 ` Ray Bryant
2005-02-22 18:49 ` Andi Kleen
2005-02-26 18:22 ` Ray Bryant
2005-02-22 22:04 ` Ray Bryant
2005-02-22 6:44 ` Ray Bryant
2005-02-21 4:20 ` Ray Bryant
2005-02-18 16:58 ` Ray Bryant
2005-02-18 17:02 ` Ray Bryant
2005-02-18 17:11 ` Ray Bryant
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=20050221121010.GC17667@wotan.suse.de \
--to=ak@suse.de \
--cc=ak@muc.de \
--cc=haveblue@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=pj@sgi.com \
--cc=raybry@austin.rr.com \
--cc=raybry@sgi.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