Re: [PATCH 3/6] mm/page_io: use swap entries directly in zeromap helpers

[Tal Zussman]

On 8/10/26 8:55 AM, Lorenzo Stoakes (ARM) wrote:
> On Thu, Aug 06, 2026 at 06:55:35PM -0400, Tal Zussman wrote:
>> Increment swp_entry_t::val directly instead of recomputing each entry
>> with page_swap_entry(). This removes the last struct page usage in
>> page_io.c and saves one call to compound_head() per page.
>>
>> Signed-off-by: Tal Zussman <tz2294@xxxxxxxxxxxx>
> 
> Lord I just looked at page_swap_entry() and... yeah. This is an improvement
> thanks :)
> 
> I haven't looked at the rest of the series so maybe you already did it but I
> wonder if we could remove it altogether?
> 

Indeed we can - I have ~7 more patches which do exactly that, but they're more
wide-ranging (rmap, zswap, huge_memory, swapfile, arm64 MTE), so I kept them
separate from this.

I believe Andrew is planning to pick this up post-rc1, so I'll probably send
them out then since they rely on some of these changes.

> Should we rename page_io.c to folio_io.c now? :P
> 
> Anyway LGTM so:
> 
> Reviewed-by: Lorenzo Stoakes (ARM) <ljs@xxxxxxxxxx>
> 

Thanks!

>> ---
>>  mm/page_io.c | 8 ++++----
>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/mm/page_io.c b/mm/page_io.c
>> index 43b4383a4729..c6d0949b3f1e 100644
>> --- a/mm/page_io.c
>> +++ b/mm/page_io.c
>> @@ -159,7 +159,7 @@ static void swap_zeromap_folio_set(struct folio *folio)
>>  	struct obj_cgroup *objcg = get_obj_cgroup_from_folio(folio);
>>  	int nr_pages = folio_nr_pages(folio);
>>  	struct swap_cluster_info *ci;
>> -	swp_entry_t entry;
>> +	swp_entry_t entry = folio->swap;
>>  	unsigned int i;
>>
>>  	VM_WARN_ON_ONCE_FOLIO(!folio_test_swapcache(folio), folio);
>> @@ -167,8 +167,8 @@ static void swap_zeromap_folio_set(struct folio *folio)
>>
>>  	ci = swap_cluster_get_and_lock(folio);
>>  	for (i = 0; i < folio_nr_pages(folio); i++) {
>> -		entry = page_swap_entry(folio_page(folio, i));
>>  		__swap_table_set_zero(ci, swp_cluster_offset(entry));
>> +		entry.val++;
>>  	}
>>  	swap_cluster_unlock(ci);
>>
>> @@ -182,7 +182,7 @@ static void swap_zeromap_folio_set(struct folio *folio)
>>  static void swap_zeromap_folio_clear(struct folio *folio)
>>  {
>>  	struct swap_cluster_info *ci;
>> -	swp_entry_t entry;
>> +	swp_entry_t entry = folio->swap;
>>  	unsigned int i;
>>
>>  	VM_WARN_ON_ONCE_FOLIO(!folio_test_swapcache(folio), folio);
>> @@ -190,8 +190,8 @@ static void swap_zeromap_folio_clear(struct folio *folio)
>>
>>  	ci = swap_cluster_get_and_lock(folio);
>>  	for (i = 0; i < folio_nr_pages(folio); i++) {
>> -		entry = page_swap_entry(folio_page(folio, i));
>>  		__swap_table_clear_zero(ci, swp_cluster_offset(entry));
>> +		entry.val++;
>>  	}
>>  	swap_cluster_unlock(ci);
>>  }
>>
>> --
>> 2.39.5
>>
> 
> --
> Cheers, Lorenzo
>