[PATCH] mm/damon/paddr: fix pin page problem

[PATCH] mm/damon/paddr: fix pin page problem

[Andrew Yang]

From: "andrew.yang" <andrew.yang@mediatek.com>

damon_get_page() would always increase page _refcount and
isolate_lru_page() would increase page _refcount if the page's lru
flag is set.

If a unevictable page isolated successfully, there will be two more
_refcount. The one from isolate_lru_page() will be decreased in
putback_lru_page(), but the other one from damon_get_page() will be
left behind. This causes a pin page.

Whatever the case, the _refcount from damon_get_page() should be
decreased.

Signed-off-by: andrew.yang <andrew.yang@mediatek.com>
---
 mm/damon/paddr.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/damon/paddr.c b/mm/damon/paddr.c
index e1a4315c4be6..56d8abd08fb1 100644
--- a/mm/damon/paddr.c
+++ b/mm/damon/paddr.c
@@ -223,8 +223,8 @@ static unsigned long damon_pa_pageout(struct damon_region *r)
 			putback_lru_page(page);
 		} else {
 			list_add(&page->lru, &page_list);
-			put_page(page);
 		}
+		put_page(page);
 	}
 	applied = reclaim_pages(&page_list);
 	cond_resched();
-- 
2.18.0

Re: [PATCH] mm/damon/paddr: fix pin page problem

[David Hildenbrand]

On 21.02.23 10:03, Andrew Yang wrote:
> From: "andrew.yang" <andrew.yang@mediatek.com>

Nit: I'd suggest a different subject

"mm/damon/paddr: fix missing put_page()"


Fixes: ?
Cc: stable ?


I think a patch from Willy is queued that converts this code into folios.

> 
> damon_get_page() would always increase page _refcount and
> isolate_lru_page() would increase page _refcount if the page's lru
> flag is set.
> 
> If a unevictable page isolated successfully, there will be two more
> _refcount. The one from isolate_lru_page() will be decreased in
> putback_lru_page(), but the other one from damon_get_page() will be
> left behind. This causes a pin page.
> 
> Whatever the case, the _refcount from damon_get_page() should be
> decreased.
> 
> Signed-off-by: andrew.yang <andrew.yang@mediatek.com>
> ---
>   mm/damon/paddr.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/damon/paddr.c b/mm/damon/paddr.c
> index e1a4315c4be6..56d8abd08fb1 100644
> --- a/mm/damon/paddr.c
> +++ b/mm/damon/paddr.c
> @@ -223,8 +223,8 @@ static unsigned long damon_pa_pageout(struct damon_region *r)
>   			putback_lru_page(page);
>   		} else {
>   			list_add(&page->lru, &page_list);
> -			put_page(page);
>   		}
> +		put_page(page);
>   	}
>   	applied = reclaim_pages(&page_list);
>   	cond_resched();

-- 
Thanks,

David / dhildenb

Re: [PATCH] mm/damon/paddr: fix pin page problem

[SeongJae Park]

Hi Andrew,


On Tue, 21 Feb 2023 17:03:13 +0800 Andrew Yang <andrew.yang@mediatek.com> wrote:

> From: "andrew.yang" <andrew.yang@mediatek.com>
> 
> damon_get_page() would always increase page _refcount and
> isolate_lru_page() would increase page _refcount if the page's lru
> flag is set.
> 
> If a unevictable page isolated successfully, there will be two more
> _refcount. The one from isolate_lru_page() will be decreased in
> putback_lru_page(), but the other one from damon_get_page() will be
> left behind. This causes a pin page.
> 
> Whatever the case, the _refcount from damon_get_page() should be
> decreased.

Thank you for finding this issue!  I think the David suggested subject[1] is
better, though.

I think we could add below Fixes: and Cc: tags?

Fixes: 57223ac29584 ("mm/damon/paddr: support the pageout scheme")
Cc: <stable@vger.kernel.org> # 5.16.x

> 
> Signed-off-by: andrew.yang <andrew.yang@mediatek.com>
> ---
>  mm/damon/paddr.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/damon/paddr.c b/mm/damon/paddr.c
> index e1a4315c4be6..56d8abd08fb1 100644
> --- a/mm/damon/paddr.c
> +++ b/mm/damon/paddr.c
> @@ -223,8 +223,8 @@ static unsigned long damon_pa_pageout(struct damon_region *r)
>  			putback_lru_page(page);
>  		} else {
>  			list_add(&page->lru, &page_list);
> -			put_page(page);
>  		}
> +		put_page(page);

Seems your patch is not based on mm-unstable tree[2].  Could you please rebase
on it?

Also, let's remove the braces for the single statements[3].

[1] https://lore.kernel.org/damon/1b3e8e88-ed5c-7302-553f-4ddb3400d466@redhat.com/
[2] https://docs.kernel.org/next/mm/damon/maintainer-profile.html#scm-trees
[3] https://docs.kernel.org/process/coding-style.html?highlight=coding+style#placing-braces-and-spaces


Thanks,
SJ

>  	}
>  	applied = reclaim_pages(&page_list);
>  	cond_resched();
> -- 
> 2.18.0