[PATCH v2 3/3] mm: rmap: support batched unmapping for file large folios

[Baolin Wang]

On 2025/12/16 18:54, Lorenzo Stoakes wrote:
> On Tue, Dec 16, 2025 at 02:22:11PM +0800, Baolin Wang wrote:
>>
>>
>> On 2025/12/16 14:13, Barry Song wrote:
>>>>>>
>>>>>> diff --git a/mm/rmap.c b/mm/rmap.c
>>>>>> index ec232165c47d..4c9d5777c8da 100644
>>>>>> --- a/mm/rmap.c
>>>>>> +++ b/mm/rmap.c
>>>>>> @@ -1855,9 +1855,10 @@ static inline unsigned int folio_unmap_pte_batch(struct folio *folio,
>>>>>>        end_addr = pmd_addr_end(addr, vma->vm_end);
>>>>>>        max_nr = (end_addr - addr) >> PAGE_SHIFT;
>>>>>>
>>>>>> -    /* We only support lazyfree batching for now ... */
>>>>>> -    if (!folio_test_anon(folio) || folio_test_swapbacked(folio))
>>>>>> +    /* We only support lazyfree or file folios batching for now ... */
>>>>>> +    if (folio_test_anon(folio) && folio_test_swapbacked(folio))
>>>>>
>>>>> Why is it now ok to support file-backed batched unmapping when it wasn't in
>>>>> Barry's series (see [0])? You don't seem to be justifying this?
>>>>
>>>> Barry's series[0] is merely aimed at optimizing lazyfree anonymous large
>>>> folios and does not continue to optimize anonymous large folios or
>>>> file-backed large folios at that point.
>>>
>>> Yep. At that time, I didn’t have an Android machine with a filesystem
>>> that supported large folios, so I focused on lazyfree. But I
>>> agree that lazyfree anon folios and file folios are quite
>>> similar.
>>>
>>>>
>>>> Subsequently, Barry sent out a new patch (see [1]) to optimize anonymous
>>>> large folios. As for file-backed large folios, the batched unmapping
>>>> support is relatively simple, since we only need to clear the PTE
>>>> entries for file-backed large folios.
>>>
>>> Yep. It is actually quite straightforward to go from lazyfree
>>> anon folios to file folios. Swap-backed anon folios are much
>>> more tricky, though.
>>
>> Agree. Thanks Barry for reviewing and confirming.
> 
> OK that makes me less concerned, but you do need to put some more justification
> in the commit message.

Sure. Will do.