From: Miao Xie <miaox@cn.fujitsu.com>
To: Nick Piggin <npiggin@suse.de>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>,
David Rientjes <rientjes@google.com>,
Paul Menage <menage@google.com>,
Linux-Kernel <linux-kernel@vger.kernel.org>,
Linux-MM <linux-mm@kvack.org>
Subject: Re: [PATCH 1/4] cpuset: fix the problem that cpuset_mem_spread_node() returns an offline node(was: Re: [regression] cpuset,mm: update tasks' mems_allowed in time (58568d2))
Date: Thu, 04 Mar 2010 17:31:19 +0800 [thread overview]
Message-ID: <4B8F7DE7.1050705@cn.fujitsu.com> (raw)
In-Reply-To: <20100304032209.GM8653@laptop>
on 2010-3-4 11:22, Nick Piggin wrote:
...
>> + /*
>> + * After current->mems_allowed is set to a new value, current will
>> + * allocate new pages for the migrating memory region. So we must
>> + * ensure that update of current->mems_allowed have been completed
>> + * by this moment.
>> + */
>> + smp_wmb();
>> do_migrate_pages(mm, from, to, MPOL_MF_MOVE_ALL);
>>
>> guarantee_online_mems(task_cs(tsk),&tsk->mems_allowed);
>> +
>> + /*
>> + * After doing migrate pages, current will allocate new pages for
>> + * itself not the other tasks. So we must ensure that update of
>> + * current->mems_allowed have been completed by this moment.
>> + */
>> + smp_wmb();
>
> The comments don't really make sense. A task always sees its own
> memory operations in program order. You keep saying *current* allocates
> pages so *current*->mems_allowed must be updated. This doesn't make
> sense. Do you mean to say tsk->?
>
> Secondly, memory ordering operations do not ensure anything is
> completed. They only ensure ordering. So to make sense to use them,
> you generally need corresponding barriers in other code that can
> run concurrently.
>
> So you need to comment what is being ordered (ie. at least 2 memory
> operations). And what other code might be running that requires this
> ordering.
>
> You need to comment to all these sites and operations. Sprinkling of
> memory barriers just gets unmaintainable.
My thought is wrong.
I thought the kernel might call do_migrate_pages() before updating
->mems_allowed, so I used smp_wmb() to ensure this order.
In fact, this problem which I worried can't occur, so these smp_wmb()
is unnecessary.
Thanks!
Miao
--
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>
prev parent reply other threads:[~2010-03-04 9:31 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-03-03 10:44 Miao Xie
2010-03-04 3:22 ` Nick Piggin
2010-03-04 9:31 ` Miao Xie [this message]
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=4B8F7DE7.1050705@cn.fujitsu.com \
--to=miaox@cn.fujitsu.com \
--cc=lee.schermerhorn@hp.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=menage@google.com \
--cc=npiggin@suse.de \
--cc=rientjes@google.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