Re: [PATCH 7.2 v4 09/12] mm/truncate: use folio_split() in truncate_inode_partial_folio()

[David Hildenbrand (Arm)]

On 4/24/26 17:12, Zi Yan wrote:
> On 24 Apr 2026, at 11:07, Zi Yan wrote:
> 
>> On 24 Apr 2026, at 8:54, David Hildenbrand (Arm) wrote:
>>
>>>
>>> In general looks good. Just one question:
>>>
>>> folio_check_splittable() also rejects SPLIT_TYPE_NON_UNIFORM / new_order for
>>> swapcache pages.
>>
>> Since swapcache only supports PMD order or 0 order yet.
>>
>>>
>>> It's confusing, but truncate would never stumble over such folios, correct?
>>
>> Right.
> 
> BTW, sashiko had a similar question when it reviewed v2[1]. It also complained
> about when the split fails, the entire folio remains in memory. But that is the
> consequence of failed split, unless we want to keep retrying the split.

Retry isn't guaranteed to make progress, so it's the current expected behavior.

-- 
Cheers,

David