Re: [PATCH v1 2/5] exfat: dirty all new pages when extending valid_size

[Chi Zhiling <chizhiling@163.com>]

On 9/30/26 2:16 PM, Chi Zhiling wrote:
> On 9/29/26 6:26 PM, Yuezhang.Mo@sony.com wrote:
>>> +               if (folio->mapping == mapping) {
>>>                          folio_mark_dirty(folio);
>>> +                       pos = next;
>>
>> Setting 'pos' under this condition may result in an infinite loop.
> 
> If we set pos unconditionally, I'm concerned that we might miss marking 
> a folio as dirty, which could expose stale data.
> 
> I think it would be better to switch to __filemap_get_folio() in the 
> next version. It can directly return a locked folio, so we don't need to 
> check whether the folio has been truncated.
> 
> folio = __filemap_get_folio(mapping, index, FGP_WRITEBEGIN, 0);

Sorry, this change is incorrect.

I took another look at it, and I think this loop will not block. 
Reacquiring the folio is the correct approach, and do_read_cache_folio() 
performs the check in the same way.

> 
>>
>>> +               }
>>>                  folio_unlock(folio);
>>>                  folio_put(folio);
>>> -               pos = next;
>>>          }
>>>
>>>         return 0;
>>
>