Linux-mm Archive on lore.kernel.org
 help / color / Atom feed
From: "Li Xinhai" <lixinhai.lxh@gmail.com>
To: yang.shi <yang.shi@linux.alibaba.com>,
	 "Mike Kravetz" <mike.kravetz@oracle.com>,
	 "linux-mm@kvack.org" <linux-mm@kvack.org>
Cc: akpm <akpm@linux-foundation.org>,  mhocko <mhocko@suse.com>,
	 n-horiguchi <n-horiguchi@ah.jp.nec.com>
Subject: Re: [PATCH 2/2] mm/mempolicy: Skip walking HUGETLB vma if MPOL_MF_STRICT is specified alone
Date: Wed, 15 Jan 2020 15:36:27 +0800
Message-ID: <2020011515362520135446@gmail.com> (raw)
In-Reply-To: <253e9110-4ffd-e9ba-feec-48ce899af057@linux.alibaba.com>

On 2020-01-15 at 13:23 Yang Shi wrote:
>
>
>On 1/14/20 8:28 PM, Mike Kravetz wrote:
>> On 1/14/20 5:24 PM, Yang Shi wrote:
>>>
>>> On 1/14/20 5:07 PM, Mike Kravetz wrote:
>>>> On 1/14/20 6:09 AM, Li Xinhai wrote:
>>>>> Add cc to
>>>>> Yang Shi <yang.shi@linux.alibaba.com>
>>>>> Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
>>>>> , who has been worked on this part
>>>>>
>>>>> On 2020-01-14 at 17:16 Li Xinhai wrote:
>>>>>> Checking MPOL_MF_STRICT is ignored for HUGETLB vma according to mbind man
>>>>>> page:
>>>>>>
>>>>>> Notes
>>>>>> MPOL_MF_STRICT is ignored on huge page mappings.
>>>>>>
>>>>>> If MPOL_MF_STRICT is specified alone without any MOVE flag, we should
>>>>>> indicate, from test_walk, that walking this vma should be skipped even if
>>>>>> there are misplaced pages.
>>>>>>
>>>>>> Signed-off-by: Li Xinhai <lixinhai.lxh@gmail.com>
>>>>>> Cc: Michal Hocko <mhocko@suse.com>
>>>>>> Cc: Mike Kravetz <mike.kravetz@oracle.com>
>>>> I do not necessarily disagree with the change.  However, this has made me
>>>> question a couple things:
>>>> 1) Why does the man page say MPOL_MF_STRICT is ignored on huge page mappings?
>>>>      - Is that leftover from the the days when huge page migration was not
>>>>        supported?
>>>>      - Is it just because huge page migration is more likely to fail than
>>>>        base page migration.
>>>> 2) Does the mbind code function properly when unable to migrate a huge page
>>>>      MPOL_MF_STRICT is set?  A quick look at the code looks like it returns
>>>>      EIO. 
for question (2),
look into queue_pages_hugetlb(), the misplaced page would not
cause -EIO reported, for both STRICT set alone and STRICT set with MOVE*;
that means STRICT been effectively ignored during isolation phase.

In unmap and move phase, -EIO is reported if failed to move page and
STRICT is set.

>>> I don't know the answer about question #1 I didn't dig into the history. The queue_pages_hugetlb() returns 0 unconditionally, I think this is what "MPOL_MF_STRICT is ignored on huge page mappings" means in code.
>>>
>>> It would return -EIO for base pages or THP as what the manpage describes.
>>>
>> I was thinking about a migration failure after isolation.  This block of
>> code in do_mbind() after queue_pages_range() and mbind_range().
>>
>>          if (!err) {
>>                  int nr_failed = 0;
>>
>>                  if (!list_empty(&pagelist)) {
>>                          WARN_ON_ONCE(flags & MPOL_MF_LAZY);
>>                          nr_failed = migrate_pages(&pagelist, new_page, NULL,
>>                                  start, MIGRATE_SYNC, MR_MEMPOLICY_MBIND);
>>                          if (nr_failed)
>>                                  putback_movable_pages(&pagelist);
>>                  }
>>
>>                  if ((ret > 0) || (nr_failed && (flags & MPOL_MF_STRICT)))
>>                          err = -EIO;
>
>Hmm.. I agree this part in man page does look ambiguous. We may assume
>"MPOL_MF_STRICT is ignored on huge page mappings." implies if
>MPOL_MF_STRICT is specified alone? If MOVE flag is specified it should
>return -EIO if some pages could not be moved as what the man page describes.
> 
It looks to me that current code has no feasible way to ignore STRICT
flag for hugetlb page when failure happen in  unmap&move phase,
because mbind is about to handle multiple vma(i.e., hugetlb vma mixed with
other vma) in one call.

>I don't know what the intention was at the first place. We may have to
>dig into the history.
>
>>
>
>

  reply index

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-01-14  9:16 [PATCH 1/2] mm/mempolicy: Checking hstate for hugetlbfs page in vma_migratable Li Xinhai
2020-01-14  9:16 ` [PATCH 2/2] mm/mempolicy: Skip walking HUGETLB vma if MPOL_MF_STRICT is specified alone Li Xinhai
2020-01-14 14:09   ` Li Xinhai
2020-01-14 18:27     ` Yang Shi
2020-01-15  1:07     ` Mike Kravetz
2020-01-15  1:24       ` Yang Shi
2020-01-15  4:28         ` Mike Kravetz
2020-01-15  5:23           ` Yang Shi
2020-01-15  7:36             ` Li Xinhai [this message]
2020-01-15 17:16               ` Yang Shi
2020-01-15 21:07       ` Mike Kravetz
2020-01-15 21:30         ` Yang Shi
2020-01-15 21:45           ` Mike Kravetz
2020-01-15 21:59             ` Yang Shi
2020-01-16  8:07               ` HORIGUCHI NAOYA(堀口 直也)
2020-01-16 15:32                 ` Li Xinhai
2020-01-16  7:59         ` Michal Hocko
2020-01-16 19:22           ` Mike Kravetz
2020-01-17  2:32             ` Yang Shi
2020-01-17  2:38             ` Li Xinhai
2020-01-17  7:57               ` Michal Hocko
2020-01-17 12:05                 ` Li Xinhai
2020-01-17 15:20                   ` Michal Hocko
2020-01-17 15:46                     ` Li Xinhai
2020-01-20 12:45                       ` Michal Hocko
2020-01-21 14:15                         ` Li Xinhai
2020-01-21 14:53                           ` Michal Hocko
2020-01-22 13:55                             ` Li Xinhai
2020-01-14 19:12 ` [PATCH 1/2] mm/mempolicy: Checking hstate for hugetlbfs page in vma_migratable Mike Kravetz
2020-01-15  1:25 ` Andrew Morton

Reply instructions:

You may reply publically 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=2020011515362520135446@gmail.com \
    --to=lixinhai.lxh@gmail.com \
    --cc=akpm@linux-foundation.org \
    --cc=linux-mm@kvack.org \
    --cc=mhocko@suse.com \
    --cc=mike.kravetz@oracle.com \
    --cc=n-horiguchi@ah.jp.nec.com \
    --cc=yang.shi@linux.alibaba.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

Linux-mm Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-mm/0 linux-mm/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-mm linux-mm/ https://lore.kernel.org/linux-mm \
		linux-mm@kvack.org
	public-inbox-index linux-mm

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kvack.linux-mm


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git