linux-kernel - Re: [PATCH 4/7] migrate_pages(): fix failure counting for THP subpages retrying

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <7b98cabf-9a8f-2dcb-a410-9a8f9611ae88@linux.alibaba.com>
Date:   Mon, 27 Jun 2022 13:56:57 +0800
From:   Baolin Wang <baolin.wang@...ux.alibaba.com>
To:     "Huang, Ying" <ying.huang@...el.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org, Zi Yan <ziy@...dia.com>,
        Yang Shi <shy828301@...il.com>
Subject: Re: [PATCH 4/7] migrate_pages(): fix failure counting for THP
 subpages retrying



On 6/27/2022 12:23 PM, Huang, Ying wrote:
> Baolin Wang <baolin.wang@...ux.alibaba.com> writes:
> 
>> On 6/27/2022 9:46 AM, Huang, Ying wrote:
>>> Baolin Wang <baolin.wang@...ux.alibaba.com> writes:
>>>
>>>> On 6/24/2022 10:53 AM, Huang Ying wrote:
>>>>> If THP is failed to be migrated for -ENOSYS and -ENOMEM, the THP will
>>>>> be split into thp_split_pages, and after other pages are migrated,
>>>>> pages in thp_split_pages will be migrated with no_subpage_counting ==
>>>>> true, because its failure have been counted already.  If some pages in
>>>>> thp_split_pages are retried during migration, we should not count
>>>>> their failure if no_subpage_counting == true too.  This is done this
>>>>> patch to fix the failure counting for THP subpages retrying.
>>>>
>>>> Good catch. Totally agree with you. It seems we can move the condition
>>>> into -EAGAIN case like other cases did?
>>>>
>>>> diff --git a/mm/migrate.c b/mm/migrate.c
>>>> index 1ece23d80bc4..491c2d07402b 100644
>>>> --- a/mm/migrate.c
>>>> +++ b/mm/migrate.c
>>>> @@ -1463,7 +1463,7 @@ int migrate_pages(struct list_head *from,
>>>> new_page_t get_new_page,
>>>>                           case -EAGAIN:
>>>>                                   if (is_thp)
>>>>                                           thp_retry++;
>>>> -                               else
>>>> +                               else if (!no_subpage_counting)
>>>>                                           retry++;
>>>>                                   break;
>>> This has another effect except fixing the failure counting.  That
>>> is,
>>> the split subpages of THP will not be retried for 10 times for -EAGAIN.
>>
>> Ah, yes.
>>
>>> TBH, I think that we should do that.  But because this has some behavior
>>
>> OK. So you afraid that 10 times retry for each subpage of THP will
>> waste lots of time?
> 
> I just think that it's unnecessary.  We have already regarded the
> migration as failed.  And for the worst case, we will try 512 * 10 =
> 5120 times in total.
> 
>>> change, it's better to be done in a separate patch?  Do you have
>>> interest to do that on top of this patchset?
>>
>> Sure. I can send a patch which can be folded into your series. Is this
>> OK for you?
>>
>> By the way, if I do like I said, the patch 4 can be avoided.
> 
> I tend to keep both.  [4/7] is just a fix.  You patch will introduce the
> behavior change.  And your patch needn't to be folded into this series.
> You can send it and push it separately.

OK. Thanks.