From: Yu Xu <xuyu@linux.alibaba.com>
To: Hugh Dickins <hughd@google.com>
Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org,
akpm@linux-foundation.org, gavin.dg@linux.alibaba.com,
Greg Thelen <gthelen@google.com>, Wei Xu <weixugc@google.com>,
Matthew Wilcox <willy@infradead.org>,
Nicholas Piggin <npiggin@gmail.com>,
Vlastimil Babka <vbabka@suse.cz>
Subject: Re: [PATCH] mm, thp: relax migration wait when failed to get tail page
Date: Mon, 7 Jun 2021 15:24:41 +0800 [thread overview]
Message-ID: <6c4e0df7-1f06-585f-d113-f38db6c819b5@linux.alibaba.com> (raw)
In-Reply-To: <alpine.LSU.2.11.2106020831590.6388@eggly.anvils>
On 6/2/21 11:57 PM, Hugh Dickins wrote:
> On Wed, 2 Jun 2021, Yu Xu wrote:
>> On 6/2/21 12:55 AM, Hugh Dickins wrote:
>>> On Wed, 2 Jun 2021, Xu Yu wrote:
>>>
>>>> We notice that hung task happens in a conner but practical scenario when
>>>> CONFIG_PREEMPT_NONE is enabled, as follows.
>>>>
>>>> Process 0 Process 1 Process
>>>> 2..Inf
>>>> split_huge_page_to_list
>>>> unmap_page
>>>> split_huge_pmd_address
>>>> __migration_entry_wait(head)
>>>> __migration_entry_wait(tail)
>>>> remap_page (roll back)
>>>> remove_migration_ptes
>>>> rmap_walk_anon
>>>> cond_resched
>>>>
>>>> Where __migration_entry_wait(tail) is occurred in kernel space, e.g.,
>>>> copy_to_user, which will immediately fault again without rescheduling,
>>>> and thus occupy the cpu fully.
>>>>
>>>> When there are too many processes performing __migration_entry_wait on
>>>> tail page, remap_page will never be done after cond_resched.
>>>>
>>>> This relaxes __migration_entry_wait on tail page, thus gives remap_page
>>>> a chance to complete.
>>>>
>>>> Signed-off-by: Gang Deng <gavin.dg@linux.alibaba.com>
>>>> Signed-off-by: Xu Yu <xuyu@linux.alibaba.com>
>>>
>>> Well caught: you're absolutely right that there's a bug there.
>>> But isn't cond_resched() just papering over the real bug, and
>>> what it should do is a "page = compound_head(page);" before the
>>> get_page_unless_zero()? How does that work out in your testing?
>>
>> compound_head works. The patched kernel is alive for hours under
>> our reproducer, which usually makes the vanilla kernel hung after
>> tens of minutes at most.
>
> Oh, that's good news, thanks.
>
> (It's still likely that a well-placed cond_resched() somewhere in
> mm/gup.c would also be a good idea, but none of us have yet got
> around to identifying where.)
We neither. If really have to do it outside of __migration_entry_wait,
return value of __migration_entry_wait is needed, and many related
functions have to updated, which may be undesirable.
>
>>
>> If we use compound_head, the behavior of __migration_entry_wait(tail)
>> changes from "retry fault" to "prevent THP from being split". Is that
>> right? Then which is preferred? If it were me, I would prefer "retry
>> fault".
>
> As Matthew remarked, you are asking very good questions, and split
> migration entries are difficult to think about. But I believe you'll
> find it works out okay.
>
> The point of *put_and_* wait_on_page_locked() is that it does drop
> the page reference you acquired with get_page_unless_zero, as soon
> as the page is on the wait queue, before actually waiting.
>
> So splitting the THP is only prevented for a brief interval. Now,
> it's true that if there are very many tasks faulting on portions
> of the huge page, in that interval between inserting the migration
> entries and freezing the huge page's refcount to 0, they can reduce
> the chance of splitting considerably. But that's not an excuse for
> for doing get_page_unless_zero() on the wrong thing, as it was doing.
We finally come to your solution, i.e., compound_head.
In that case, who should resend the compound_head patch to this issue?
shall we do with your s.o.b?
>
> Hugh
>
--
Thanks,
Yu
next prev parent reply other threads:[~2021-06-07 7:24 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-06-01 16:31 [PATCH] mm, thp: relax migration wait when failed to get tail page Xu Yu
2021-06-01 16:49 ` Matthew Wilcox
2021-06-01 16:55 ` Hugh Dickins
2021-06-01 17:11 ` Matthew Wilcox
2021-06-01 19:10 ` Hugh Dickins
2021-06-01 20:27 ` Matthew Wilcox
2021-06-02 3:27 ` Yu Xu
2021-06-02 11:58 ` Matthew Wilcox
2021-06-02 12:59 ` Yu Xu
2021-06-02 13:20 ` Yu Xu
2021-06-02 15:57 ` Hugh Dickins
2021-06-07 7:24 ` Yu Xu [this message]
2021-06-08 4:44 ` Hugh Dickins
2021-06-08 5:43 ` Yu Xu
2021-06-08 6:53 ` Hugh Dickins
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=6c4e0df7-1f06-585f-d113-f38db6c819b5@linux.alibaba.com \
--to=xuyu@linux.alibaba.com \
--cc=akpm@linux-foundation.org \
--cc=gavin.dg@linux.alibaba.com \
--cc=gthelen@google.com \
--cc=hughd@google.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=npiggin@gmail.com \
--cc=vbabka@suse.cz \
--cc=weixugc@google.com \
--cc=willy@infradead.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).