diff mbox series

[2/3] damon: Convert damon_pa_mark_accessed_or_deactivate() to use folios

Message ID 20221207002158.418789-3-vishal.moola@gmail.com
State New
Headers
Series Convert deactivate_page() to deactivate_folio() |

Commit Message

Vishal Moola Dec. 7, 2022, 12:21 a.m. UTC 
  This change replaces 2 calls to compound_head() with one. This is in
preparation for the conversion of deactivate_page() to
deactivate_folio().

Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
---
 mm/damon/paddr.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

Comments

SeongJae Park Dec. 7, 2022, 12:56 a.m. UTC | #1 
Hi Vishal,


Thank you for this patch.

Nit pick.  Could we please replace 'damon:' on the subject with 'mm/damon:' to
look more consistent with other DAMON patches?

On Tue, 6 Dec 2022 16:21:57 -0800 "Vishal Moola (Oracle)" <vishal.moola@gmail.com> wrote:

> This change replaces 2 calls to compound_head() with one.

I guess you mean _compound_head() calls in page_folio() that called from
mark_page_accessed() and folio_mark_accessed().  However, deactivate_page()
calls page_folio() anyway, so this patch doesn't reduce the number of calls to
one but keep the number, correct?  Am I missing something?  If I'm not, I'd
like to clean up the wording.

> This is in preparation for the conversion of deactivate_page() to
> deactivate_folio().

I think folio_deactivate() might be a more consistent naming.  What do you
think?

Also, you may keep the above sentence if you implement folio_deactivate()
first, update deactivate_page() calls to use folio_deactivate(), and finally
remove deactivate_page() definition.

I don't really care about the name and the sequence, though.

> 
> Signed-off-by: Vishal Moola (Oracle) <vishal.moola@gmail.com>
> ---
>  mm/damon/paddr.c | 11 +++++++----
>  1 file changed, 7 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/damon/paddr.c b/mm/damon/paddr.c
> index e1a4315c4be6..73548bc82297 100644
> --- a/mm/damon/paddr.c
> +++ b/mm/damon/paddr.c
> @@ -238,15 +238,18 @@ static inline unsigned long damon_pa_mark_accessed_or_deactivate(
>  
>  	for (addr = r->ar.start; addr < r->ar.end; addr += PAGE_SIZE) {
>  		struct page *page = damon_get_page(PHYS_PFN(addr));
> +		struct folio *folio;
>  
>  		if (!page)
>  			continue;
> +		folio = page_folio(page);

One _compound_head() call here,

> +
>  		if (mark_accessed)
> -			mark_page_accessed(page);
> +			folio_mark_accessed(folio);
>  		else
> -			deactivate_page(page);
> -		put_page(page);
> -		applied++;
> +			deactivate_page(&folio->page);

And second _compound_head() call here.

> +		folio_put(folio);
> +		applied += folio_nr_pages(folio);
>  	}
>  	return applied * PAGE_SIZE;
>  }
> -- 
> 2.38.1
> 


Thanks,
SJ
SeongJae Park Dec. 7, 2022, 1:04 a.m. UTC | #2 
On Wed, 7 Dec 2022 00:56:24 +0000 SeongJae Park <sj@kernel.org> wrote:

> Hi Vishal,
> 
> 
> Thank you for this patch.
> 
> Nit pick.  Could we please replace 'damon:' on the subject with 'mm/damon:' to
> look more consistent with other DAMON patches?
> 
> On Tue, 6 Dec 2022 16:21:57 -0800 "Vishal Moola (Oracle)" <vishal.moola@gmail.com> wrote:
> 
> > This change replaces 2 calls to compound_head() with one.
> 
> I guess you mean _compound_head() calls in page_folio() that called from
> mark_page_accessed() and folio_mark_accessed().

s/folio_mark_accessed()/deactivate_page()/

Sorry for the typo.


Thanks,
SJ

[...]
Vishal Moola Dec. 7, 2022, 2:43 a.m. UTC | #3 
On Tue, Dec 6, 2022 at 4:56 PM SeongJae Park <sj@kernel.org> wrote:
>
> Hi Vishal,
>
>
> Thank you for this patch.
>
> Nit pick.  Could we please replace 'damon:' on the subject with 'mm/damon:' to
> look more consistent with other DAMON patches?

Thanks for looking over this patch! I'll make the subject mm/damon in a v2.

> On Tue, 6 Dec 2022 16:21:57 -0800 "Vishal Moola (Oracle)" <vishal.moola@gmail.com> wrote:
>
> > This change replaces 2 calls to compound_head() with one.
>
> I guess you mean _compound_head() calls in page_folio() that called from
> mark_page_accessed() and folio_mark_accessed().  However, deactivate_page()
> calls page_folio() anyway, so this patch doesn't reduce the number of calls to
> one but keep the number, correct?  Am I missing something?  If I'm not, I'd
> like to clean up the wording.

The 2 calls I was referring to were from mark_page_accessed() and put_page().
As you've noticed, deactivate_page() still calls page_folio() here :).

> > This is in preparation for the conversion of deactivate_page() to
> > deactivate_folio().
>
> I think folio_deactivate() might be a more consistent naming.  What do you
> think?

I do like the name folio_deactivate() better than deactivate_folio(), I'll
change that in v2 as well.
diff mbox series

Patch

diff --git a/mm/damon/paddr.c b/mm/damon/paddr.c
index e1a4315c4be6..73548bc82297 100644
--- a/mm/damon/paddr.c
+++ b/mm/damon/paddr.c
@@ -238,15 +238,18 @@  static inline unsigned long damon_pa_mark_accessed_or_deactivate(
 
 	for (addr = r->ar.start; addr < r->ar.end; addr += PAGE_SIZE) {
 		struct page *page = damon_get_page(PHYS_PFN(addr));
+		struct folio *folio;
 
 		if (!page)
 			continue;
+		folio = page_folio(page);
+
 		if (mark_accessed)
-			mark_page_accessed(page);
+			folio_mark_accessed(folio);
 		else
-			deactivate_page(page);
-		put_page(page);
-		applied++;
+			deactivate_page(&folio->page);
+		folio_put(folio);
+		applied += folio_nr_pages(folio);
 	}
 	return applied * PAGE_SIZE;
 }