Re: [PATCH] mm/huge_memory.c: introduce folio_split_unmapped

[Wei Yang <richard.weiyang@gmail.com>]

On Fri, Nov 14, 2025 at 02:30:03PM +1100, Balbir Singh wrote:
>On 11/14/25 14:21, Wei Yang wrote:
>> On Fri, Nov 14, 2025 at 12:22:28PM +1100, Balbir Singh wrote:
>> [...]
>>> @@ -4079,6 +4091,36 @@ static int __folio_split(struct folio *folio, unsigned int new_order,
>>> 	return ret;
>>> }
>>>
>>> +/*
>>> + * This function is a helper for splitting folios that have already been unmapped.
>>> + * The use case is that the device or the CPU can refuse to migrate THP pages in
>>> + * the middle of migration, due to allocation issues on either side
>>> + *
>>> + * The high level code is copied from __folio_split, since the pages are anonymous
>>> + * and are already isolated from the LRU, the code has been simplified to not
>>> + * burden __folio_split with unmapped sprinkled into the code.
>>> + *
>>> + * None of the split folios are unlocked
>>> + */
>>> +int folio_split_unmapped(struct folio *folio, unsigned int new_order)
>>> +{
>>> +	int extra_pins, ret = 0;
>>> +
>>> +	VM_WARN_ON_FOLIO(folio_mapped(folio), folio);
>>> +	VM_WARN_ON_ONCE_FOLIO(!folio_test_locked(folio), folio);
>>> +	VM_WARN_ON_ONCE_FOLIO(!folio_test_large(folio), folio);
>>> +
>> 
>> Compared with original logic, we did check folio_split_supported() and
>> check whether new_order is supported for the file system.
>> 
>> Currently folio_split_unmapped() only pass 0 as new_order, which looks good.
>> But for a generic helper, it looks reasonable to do the check, IMHO.
>> 
>
>This is meant to be used in the middle of a migration where the src/dst do 
>no agree on the folio_order() due to allocation issues. When mTHP support
>is added to device migration, order support will be added and checked.
>FYI: This routines supports just anonymous pages ATM
>

OK, I don't see these assumptions. Not sure it would be abused.

Maybe a comment would help? Or remove the new_order now? We can add it when it
is truly used.

>>> +	if (!can_split_folio(folio, 1, &extra_pins))
>>> +		return -EAGAIN;
>>> +
>>> +	local_irq_disable();
>>> +	ret = __folio_freeze_and_split_unmapped(folio, new_order, &folio->page, NULL,
>>> +						NULL, false, NULL, SPLIT_TYPE_UNIFORM,
>>> +						0, extra_pins);
>>> +	local_irq_enable();
>>> +	return ret;
>>> +}
>>> +
>>> /*
>>>  * This function splits a large folio into smaller folios of order @new_order.
>>>  * @page can point to any page of the large folio to split. The split operation
>> 
>> 
>
>Balbir

-- 
Wei Yang
Help you, Help me