Re: [PATCH v4 0/2] fix MADV_COLLAPSE issue if THP settings are disabled

[David Hildenbrand <david@redhat.com>]

On 25.06.25 12:02, Baolin Wang wrote:
> 
> 
> On 2025/6/25 17:31, Lorenzo Stoakes wrote:
>> On Wed, Jun 25, 2025 at 04:52:03PM +0800, Baolin Wang wrote:
>>>
>>>
>>> On 2025/6/25 16:37, Lorenzo Stoakes wrote:
>>>> Yeah maybe the best way is to just have another tunable for this?
>>>>
>>>> /sys/kernel/mm/transparent_hugepage/disable_collapse perhaps?
>>>>
>>>> What do you think Hugh, Baolin?
>>>
>>> I think it's not necessary to find a way to disable madvise_collapse.
>>> Essentially, it's a conflict between the semantics of madvise_collapse and
>>> the '/sys/kernel/mm/transparent_hugepage/enabled' interface. We should reach
>>> a consensus on the semantics first:
>>>
>>> Semantic 1: madv_collapse() should ignore any THP system settings, meaning
>>> we need to update the 'never' semantics in
>>> '/sys/kernel/mm/transparent_hugepage/enabled', which would only disable page
>>> fault and khugepaged, not including madvise_collapse. If we agree on this,
>>> then the 'never' for per-sized mTHP would have the same semantics, i.e.,
>>> when I set 64K mTHP to 'always' and 2M mTHP to 'never', madvise_collapse
>>> would still allow the collapse of 2M THP. We should document this clearly in
>>> case users still want 64K mTHP from madvise_collapse.
>>
>> Right yeah, I mean this is in effect how things are now. So the task is
>> documentation.
>>
>>>
>>>
>>> Semantic 2: madv_collapse() needs to respect THP system settings, which is
>>> what my patch does. Never means never, and we would need to update the
>>> documentation of madv_collapse() to make it clearer.
>>
>> Yes, and indeed this is the choice.
>>
>> I think, as David said, it comes down to whether we have a legit use case that
>> truly relies on this.
>>
>>>> (One side note on PMD-sized MADV_COLLAPSE - this is basically completely
>>>> useless for 64 KB page size arm64 systems where PMD's are 512 MB :)
>>>>
>>>> Thoughts Baolin?
>>>
>>> We should not collapse 512MB THP on 64K pagesize kernel. So seems
>>> madv_collapse() can not work on 64K pagesize kernel.
>>
>> Well I don't think anything would prevent this now right? So MADV_COLLAPSE is
>> pretty problematic on 64K pagesize kernels in general.
> 
> Yes, I don't mean it will prevent madvise_collapse(), just as you said
> that it could be problematic (it's horrible to try to collapse 512MB).

Well, assume you have a VM at that is 2 GiB and could use 4 THPs. It's 
stupid, but there might be some selected use cases where it's not 
completely stupid.

-- 
Cheers,

David / dhildenb