From: Andrew Morton <akpm@linux-foundation.org>
To: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: kosaki.motohiro@gmail.com, mingo@kernel.org, hpa@zytor.com,
srivatsa.bhat@linux.vnet.ibm.com, linux-kernel@vger.kernel.org,
x86@kernel.org, linux-mm@kvack.org
Subject: Re: [Bug fix PATCH v2] numa, cpu hotplug: Change links of CPU and node when changing node number by onlining CPU
Date: Mon, 22 Apr 2013 17:34:59 -0700 [thread overview]
Message-ID: <20130422173459.487fa3e6.akpm@linux-foundation.org> (raw)
In-Reply-To: <5175D01E.5000302@jp.fujitsu.com>
On Tue, 23 Apr 2013 09:04:46 +0900 Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com> wrote:
> 2013/04/23 7:35, Andrew Morton wrote:
> > On Fri, 19 Apr 2013 14:23:23 +0900 Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com> wrote:
> >
> >> When booting x86 system contains memoryless node, node numbers of CPUs
> >> on memoryless node were changed to nearest online node number by
> >> init_cpu_to_node() because the node is not online.
> >>
> >> ...
> >>
> >> If we hot add memory to memoryless node and offine/online all CPUs on
> >> the node, node numbers of these CPUs are changed to correct node numbers
> >> by srat_detect_node() because the node become online.
> >
> > OK, here's a dumb question.
> >
> > At boot time the CPUs are assigned to the "nearest online node" rather
> > than to their real memoryless node. The patch arranges for those CPUs
> > to still be assigned to the "nearest online node" _after_ some memory
> > is hot-added to their real node. Correct?
>
> Yes. For changing node number of CPUs safely, we should offline CPUs.
>
> >
> > Would it not be better to fix this by assigning those CPUs to their real,
> > memoryless node right at the initial boot? Or is there something in
> > the kernel which makes cpus-on-a-memoryless-node not work correctly?
> >
>
> I think assigning CPUs to real node is better. But current Linux's node
> strongly depend on memory. Thus if we just create cpus-on-a-memoryless-node,
> the kernel cannot work correctly.
hm, why. I'd have thought that if we tell the kernel something like
"this node has one zone, the size of which is zero bytes" then a
surprising amount of the existing code will Just Work.
What goes wrong?
--
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:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2013-04-23 0:35 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-04-19 5:23 [Bug fix PATCH v2] numa, cpu hotplug: Change links of CPU and node when changing node number by onlining CPU Yasuaki Ishimatsu
2013-04-22 22:35 ` Andrew Morton
2013-04-23 0:04 ` Yasuaki Ishimatsu
2013-04-23 0:34 ` Andrew Morton [this message]
2013-04-23 1:24 ` Yasuaki Ishimatsu
2013-04-23 16:06 ` Andi Kleen
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=20130422173459.487fa3e6.akpm@linux-foundation.org \
--to=akpm@linux-foundation.org \
--cc=hpa@zytor.com \
--cc=isimatu.yasuaki@jp.fujitsu.com \
--cc=kosaki.motohiro@gmail.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mingo@kernel.org \
--cc=srivatsa.bhat@linux.vnet.ibm.com \
--cc=x86@kernel.org \
/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).