Message-ID: <533f37e9-81bf-4fa2-9b72-12cdcb1edb3f@redhat.com>
Date: Thu, 25 Apr 2024 12:19:32 +0200
From: David Hildenbrand <david@...hat.com>
To: Ryan Roberts <ryan.roberts@....com>,
 Baolin Wang <baolin.wang@...ux.alibaba.com>, akpm@...ux-foundation.org,
 hughd@...gle.com
Cc: willy@...radead.org, wangkefeng.wang@...wei.com, 21cnbao@...il.com,
 ying.huang@...el.com, shy828301@...il.com, ziy@...dia.com,
 linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH 0/5] add mTHP support for anonymous share pages

On 25.04.24 12:17, Ryan Roberts wrote:
> On 25/04/2024 10:50, Baolin Wang wrote:
>>
>>
>> On 2024/4/25 17:20, David Hildenbrand wrote:
>>> On 25.04.24 11:05, Baolin Wang wrote:
>>>>
>>>>
>>>> On 2024/4/25 16:57, David Hildenbrand wrote:
>>>>> On 25.04.24 10:46, Ryan Roberts wrote:
>>>>>> On 25/04/2024 09:26, David Hildenbrand wrote:
>>>>>>> On 25.04.24 10:17, Ryan Roberts wrote:
>>>>>>>> On 25/04/2024 07:20, Baolin Wang wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On 2024/4/24 22:20, Ryan Roberts wrote:
>>>>>>>>>> On 24/04/2024 14:49, Baolin Wang wrote:
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> On 2024/4/24 18:01, Ryan Roberts wrote:
>>>>>>>>>>>> On 24/04/2024 10:55, Baolin Wang wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> On 2024/4/24 16:26, Ryan Roberts wrote:
>>>>>>>>>>>>>> On 24/04/2024 07:55, Baolin Wang wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> On 2024/4/23 18:41, Ryan Roberts wrote:
>>>>>>>>>>>>>>>> On 22/04/2024 08:02, Baolin Wang wrote:
>>>>>>>>>>>>>>>>> Anonymous pages have already been supported for multi-size
>>>>>>>>>>>>>>>>> (mTHP)
>>>>>>>>>>>>>>>>> allocation
>>>>>>>>>>>>>>>>> through commit 19eaf44954df, that can allow THP to be
>>>>>>>>>>>>>>>>> configured
>>>>>>>>>>>>>>>>> through the
>>>>>>>>>>>>>>>>> sysfs interface located at
>>>>>>>>>>>>>>>>> '/sys/kernel/mm/transparent_hugepage/hugepage-XXkb/enabled'.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> However, the anonymous shared pages will ignore the
>>>>>>>>>>>>>>>>> anonymous mTHP rule
>>>>>>>>>>>>>>>>> configured through the sysfs interface, and can only use
>>>>>>>>>>>>>>>>> the PMD-mapped
>>>>>>>>>>>>>>>>> THP, that is not reasonable. Many implement anonymous page
>>>>>>>>>>>>>>>>> sharing
>>>>>>>>>>>>>>>>> through
>>>>>>>>>>>>>>>>> mmap(MAP_SHARED | MAP_ANONYMOUS), especially in database usage
>>>>>>>>>>>>>>>>> scenarios,
>>>>>>>>>>>>>>>>> therefore, users expect to apply an unified mTHP strategy
>>>>>>>>>>>>>>>>> for anonymous
>>>>>>>>>>>>>>>>> pages,
>>>>>>>>>>>>>>>>> also including the anonymous shared pages, in order to
>>>>>>>>>>>>>>>>> enjoy the
>>>>>>>>>>>>>>>>> benefits of
>>>>>>>>>>>>>>>>> mTHP. For example, lower latency than PMD-mapped THP,
>>>>>>>>>>>>>>>>> smaller memory
>>>>>>>>>>>>>>>>> bloat
>>>>>>>>>>>>>>>>> than PMD-mapped THP, contiguous PTEs on ARM architecture to
>>>>>>>>>>>>>>>>> reduce TLB
>>>>>>>>>>>>>>>>> miss
>>>>>>>>>>>>>>>>> etc.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> This sounds like a very useful addition!
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Out of interest, can you point me at any workloads (and
>>>>>>>>>>>>>>>> off-the-shelf
>>>>>>>>>>>>>>>> benchmarks
>>>>>>>>>>>>>>>> for those workloads) that predominantly use shared anon memory?
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> As far as I know, some database related workloads make
>>>>>>>>>>>>>>> extensive use of
>>>>>>>>>>>>>>> shared
>>>>>>>>>>>>>>> anonymous page, such as PolarDB[1] in our Alibaba fleet, or
>>>>>>>>>>>>>>> MySQL likely
>>>>>>>>>>>>>>> also
>>>>>>>>>>>>>>> uses shared anonymous memory. And I still need to do some
>>>>>>>>>>>>>>> investigation to
>>>>>>>>>>>>>>> measure the performance.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> [1] https://github.com/ApsaraDB/PolarDB-for-PostgreSQL
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Thanks for the pointer!
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> The primary strategy is that, the use of huge pages for
>>>>>>>>>>>>>>>>> anonymous shared
>>>>>>>>>>>>>>>>> pages
>>>>>>>>>>>>>>>>> still follows the global control determined by the mount
>>>>>>>>>>>>>>>>> option "huge="
>>>>>>>>>>>>>>>>> parameter
>>>>>>>>>>>>>>>>> or the sysfs interface at
>>>>>>>>>>>>>>>>> '/sys/kernel/mm/transparent_hugepage/shmem_enabled'.
>>>>>>>>>>>>>>>>> The utilization of mTHP is allowed only when the global
>>>>>>>>>>>>>>>>> 'huge' switch is
>>>>>>>>>>>>>>>>> enabled.
>>>>>>>>>>>>>>>>> Subsequently, the mTHP sysfs interface
>>>>>>>>>>>>>>>>> (/sys/kernel/mm/transparent_hugepage/hugepage-XXkb/enabled)
>>>>>>>>>>>>>>>>> is checked to determine the mTHP size that can be used for
>>>>>>>>>>>>>>>>> large folio
>>>>>>>>>>>>>>>>> allocation
>>>>>>>>>>>>>>>>> for these anonymous shared pages.
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> I'm not sure about this proposed control mechanism; won't it
>>>>>>>>>>>>>>>> break
>>>>>>>>>>>>>>>> compatibility? I could be wrong, but I don't think shmem's
>>>>>>>>>>>>>>>> use of THP
>>>>>>>>>>>>>>>> used to
>>>>>>>>>>>>>>>> depend upon the value of
>>>>>>>>>>>>>>>> /sys/kernel/mm/transparent_hugepage/enabled?
>>>>>>>>>>>>>>>> So it
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Yes, I realized this after more testing.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> doesn't make sense to me that we now depend upon the
>>>>>>>>>>>>>>>> /sys/kernel/mm/transparent_hugepage/hugepage-XXkb/enabled
>>>>>>>>>>>>>>>> values
>>>>>>>>>>>>>>>> (which by
>>>>>>>>>>>>>>>> default disables all sizes except 2M, which is set to
>>>>>>>>>>>>>>>> "inherit" from
>>>>>>>>>>>>>>>> /sys/kernel/mm/transparent_hugepage/enabled).
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> The other problem is that shmem_enabled has a different set
>>>>>>>>>>>>>>>> of options
>>>>>>>>>>>>>>>> (always/never/within_size/advise/deny/force) to enabled
>>>>>>>>>>>>>>>> (always/madvise/never)
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Perhaps it would be cleaner to do the same trick we did for
>>>>>>>>>>>>>>>> enabled;
>>>>>>>>>>>>>>>> Introduce
>>>>>>>>>>>>>>>> /mm/transparent_hugepage/hugepage-XXkb/shmem_enabled, which
>>>>>>>>>>>>>>>> can have all
>>>>>>>>>>>>>>>> the
>>>>>>>>>>>>>>>> same values as the top-level
>>>>>>>>>>>>>>>> /sys/kernel/mm/transparent_hugepage/shmem_enabled,
>>>>>>>>>>>>>>>> plus the additional "inherit" option. By default all sizes
>>>>>>>>>>>>>>>> will be set to
>>>>>>>>>>>>>>>> "never" except 2M, which is set to "inherit".
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Sounds good to me. But I do not want to copy all same values
>>>>>>>>>>>>>>> from
>>>>>>>>>>>>>>> top-level
>>>>>>>>>>>>>>> '/sys/kernel/mm/transparent_hugepage/shmem_enabled':
>>>>>>>>>>>>>>> always within_size advise never deny force
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> For mTHP's shmem_enabled interface, we can just keep below
>>>>>>>>>>>>>>> values:
>>>>>>>>>>>>>>> always within_size advise never
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Cause when checking if mTHP can be used for anon shmem,
>>>>>>>>>>>>>>> 'deny' is equal to
>>>>>>>>>>>>>>> 'never', and 'force' is equal to 'always'.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> I'll admit it wasn't completely clear to me after reading the
>>>>>>>>>>>>>> docs, but my
>>>>>>>>>>>>>> rough
>>>>>>>>>>>>>> understanding is:
>>>>>>>>>>>>>>
>>>>>>>>>>>>>>         - /sys/kernel/mm/transparent_hugepage/shmem_enabled
>>>>>>>>>>>>>> controls
>>>>>>>>>>>>>>           mmap(SHARED|ANON) allocations (mostly; see rule 3)
>>>>>>>>>>>>>>         - huge=... controls tmpfs allocations
>>>>>>>>>>>>>>         - deny and force in shmem_enabled are equivalent to
>>>>>>>>>>>>>> never and
>>>>>>>>>>>>>> always for
>>>>>>>>>>>>>>           mmap(SHARED|ANON) but additionally override all tmpfs
>>>>>>>>>>>>>> mounts so they
>>>>>>>>>>>>>> act as
>>>>>>>>>>>>>>           if they were mounted with huge=never or huge=always
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Is that correct? If so, then I think it still makes sense to
>>>>>>>>>>>>>> support
>>>>>>>>>>>>>> per-size
>>>>>>>>>>>>>
>>>>>>>>>>>>> Correct.
>>>>>>>>>>>>>
>>>>>>>>>>>>>> deny/force. Certainly if a per-size control is set to
>>>>>>>>>>>>>> "inherit" and the
>>>>>>>>>>>>>> top-level control is set to deny or force, you would need that
>>>>>>>>>>>>>> to mean
>>>>>>>>>>>>>> something.
>>>>>>>>>>>>>
>>>>>>>>>>>>> IMHO, the
>>>>>>>>>>>>> '/mm/transparent_hugepage/hugepage-XXkb/shmem_enabled' interface
>>>>>>>>>>>>> should only control the anonymous shmem. And 'huge=' controls
>>>>>>>>>>>>> tmpfs
>>>>>>>>>>>>> allocation,
>>>>>>>>>>>>> so we should not use anonymous control to override tmpfs
>>>>>>>>>>>>> control, which
>>>>>>>>>>>>> seems a
>>>>>>>>>>>>> little mess?
>>>>>>>>>>>>
>>>>>>>>>>>> I agree it would be cleaner to only handle mmap(SHARED|ANON)
>>>>>>>>>>>> here, and leave
>>>>>>>>>>>> the
>>>>>>>>>>>> tmpfs stuff for another time. But my point is that
>>>>>>>>>>>> /mm/transparent_hugepage/shmem_enabled already interferes with
>>>>>>>>>>>> tmpfs if the
>>>>>>>>>>>> value is deny or force. So if you have:
>>>>>>>>>>>>
>>>>>>>>>>>> echo deny > /mm/transparent_hugepage/shmem_enabled
>>>>>>>>>>>
>>>>>>>>>>> IIUC, this global control will cause shmem_is_huge() to always
>>>>>>>>>>> return
>>>>>>>>>>> false, so
>>>>>>>>>>> no matter how
>>>>>>>>>>> '/mm/transparent_hugepage/hugepage-xxxkB/shmem_enabled' is set,
>>>>>>>>>>> anonymous shmem will not use mTHP. No?
>>>>>>>>>>
>>>>>>>>>> No, that's not how
>>>>>>>>>> '/mm/transparent_hugepage/hugepage-xxxkB/enabled' works, and
>>>>>>>>>> I think '/mm/transparent_hugepage/hugepage-xxxkB/shmem_enabled'
>>>>>>>>>> should follow
>>>>>>>>>> the established pattern.
>>>>>>>>>>
>>>>>>>>>> For anon-private, each size is controlled by its
>>>>>>>>>> /mm/transparent_hugepage/hugepage-xxxkB/enabled value. Unless that
>>>>>>>>>> value is
>>>>>>>>>> "inherit", in which case the value in
>>>>>>>>>> /mm/transparent_hugepage/enabled is used
>>>>>>>>>> for that size.
>>>>>>>>>>
>>>>>>>>>> That approach enables us to 1) maintain back-compat and 2) control
>>>>>>>>>> each size
>>>>>>>>>> independently
>>>>>>>>>>
>>>>>>>>>> 1) is met because the default is that all sizes are initially set
>>>>>>>>>> to "never",
>>>>>>>>>> except the PMD-size (e.g.
>>>>>>>>>> /mm/transparent_hugepage/hugepage-2048kB/enabled)
>>>>>>>>>> which is initially set to inherit. So any mTHP unaware SW can
>>>>>>>>>> still modify
>>>>>>>>>> /mm/transparent_hugepage/enabled and it will still only apply to
>>>>>>>>>> PMD size.
>>>>>>>>>>
>>>>>>>>>> 2) is met because mTHP aware SW can come along and e.g. enable the
>>>>>>>>>> 64K size
>>>>>>>>>> (echo always > /mm/transparent_hugepage/hugepage-64kB/enabled)
>>>>>>>>>> without
>>>>>>>>>> having to
>>>>>>>>>> modify the value in /mm/transparent_hugepage/enabled.
>>>>>>>>>
>>>>>>>>> Thanks for explanation. Initially, I want to make
>>>>>>>>> ‘/mm/transparent_hugepage/shmem_enabled’ be a global control for
>>>>>>>>> huge page, but
>>>>>>>>> I think it should follow the same strategy as anon mTHP as you said.
>>>>>>>>>
>>>>>>>>>>>> echo inherit > /mm/transparent_hugepage/hugepage-64kB/shmem_enabled
>>>>>>>>>>>>
>>>>>>>>>>>> What does that mean?
>>>>>>>>>>
>>>>>>>>>> So I think /mm/transparent_hugepage/hugepage-xxxkB/shmem_enabled
>>>>>>>>>> will need to
>>>>>>>>>> support the deny and force values. When applied to non-PMD sizes,
>>>>>>>>>> "deny" can
>>>>>>>>>> just be a noop for now, because there was no way to configure a
>>>>>>>>>> tmpfs mount for
>>>>>>>>>> non-PMD size THP in the first place. But I'm not sure what to do
>>>>>>>>>> with "force"?
>>>>>>>>>
>>>>>>>>> OK. And I also prefer that "force" should be a noop too, since anon
>>>>>>>>> shmem
>>>>>>>>> control should not configure tmpfs huge page allocation.
>>>>>>>>
>>>>>>>> I guess technically they won't be noops, but (for the non-PMD-sizes)
>>>>>>>> "force"
>>>>>>>> will be an alias for "always" and "deny" will be an alias for "never"?
>>>>>>>>
>>>>>>>> I was just a bit concerned about later changing that behavior to
>>>>>>>> also impact
>>>>>>>> tmpfs once tmpfs supports mTHP; could that cause breaks? But
>>>>>>>> thinking about it,
>>>>>>>> I don't see that as a problem.
>>>>>>>
>>>>>>> Is the question what should happen if we "inherit" "force" or if someone
>>>>>>> specifies "force" for a mTP size explicitly?
>>>>>>
>>>>>> Well I think it amounts to the same thing; there isn't much point in
>>>>>> forbidding
>>>>>> "force" to be set directly because it can still be set indirectly through
>>>>>> "inherit". We can't forbid indirectly setting it, because "inherit"
>>>>>> could be set
>>>>>> first, then the top-level shmem_enabled changed to "force" after - and we
>>>>>> wouldn't want to fail that.
>>>>>
>>>>> The default for PMD should be "inherit", for the other mTHP sizes it
>>>>> should be "never".
>>>>>
>>>>> So we should fail if:
>>>>> * Setting top-level to "force" when any non-PMD size is "inherit"
>>>>> * Setting "inherit" of a non-PMD size when the top-level is force
>>>>
>>>> IMO, for tmpfs this is true, but for anon shmem, this 2 cases should not
>>>> fail.
>>>
>>> If force does not apply to an mTHP size, it should fail if it would get
>>> inherited. Until it applies and we enable it.
>>>
>>> I'm still confused about all the toggles here, so could be I am missing
>>> something.
>>
>> Yes, this is a little messy:(
>>
>> After thinking more, considering that 'force' is used to override the tmpfs
>> mount option, and 'inherit' will inherit the global setting. Your suggestion
>> will make the logic eary to understand (though it is valid for anon shmem mTHP
>> allocations, which are not part of this scenario), Ryan, what do you think?
> 
> I was thinking that ever returning an error code when trying to set "force" for
> /mm/transparent_hugepage/shmem_enabled would be weird behavior and to be avoided
> if possible.
> 
> But as always, David makes a clear and compelling case. I'm happy to go with his
> suggestion.

I'm afraid "as always" is far from true :D

> 
> I wonder if we should do the same for "deny"? The argument I previously made was
> that non-PMD-size mTHP is always effectively denied while tmpfs doesn't support
> mTHP, so we can allow "deny" and for now it would just be an alias for "never".
> The semantics don't change when you introduce tmpfs support.
> 
> Personally I lean towards allowing "deny" from day 1. David?

If it has clear semantics that can be fulfilled as of today, sure, we 
can have it.

-- 
Cheers,

David / dhildenb

Powered by blists - more mailing lists