Re: [PATCH v10 2/8] mm/huge_memory: add two new (not yet used) functions for folio_split()

[Zi Yan <ziy@nvidia.com>]

On 10 Mar 2025, at 13:00, Matthew Wilcox wrote:

> On Mon, Mar 10, 2025 at 12:42:06PM -0400, Zi Yan wrote:
>>> Because of the “Careful” comment. But new_folio->* should be fine,
>>> since it is the same as new_head. So I probably can replace all
>>> new_head with new_folio except those VM_BUG_ON_PAGE checks?
>
> Why not also the VM_BUG_ON_PAGE check?  I mean:
>
>> @@ -3364,8 +3364,8 @@ static void __split_folio_to_order(struct folio *folio, int old_order,
>>  		/* ->mapping in first and second tail page is replaced by other uses */
>>  		VM_BUG_ON_PAGE(new_nr_pages > 2 && new_head->mapping != TAIL_MAPPING,
>>  			       new_head);
>
> 		VM_BUG_ON_PAGE(new_nr_pages > 2 && new_folio->mapping != TAIL_MAPPING, new_head);

We are checking new_folio but dump new_head, so it can cause some confusion.
But it might not be that bad.
>
> (or we could just ditch the assert entirely; it's not all that useful)

I am open to that.

>
>> -		new_head->mapping = head->mapping;
>> -		new_head->index = head->index + index;
>> +		new_folio->mapping = head->mapping;
>> +		new_folio->index = head->index + index;
>
> 	new_folio->mapping = folio->mapping
> 	new_folio->index = folio->index +index;
>
> (um, and that index + index looks weird; better name might be just 'i')

OK. Let me make the changes you suggested and fold it to Hugh’s fix patch,
before Andrew picks that up.

Best Regards,
Yan, Zi