2019-05-24 19:34:55

by Ira Weiny

Subject: [PATCH RFC] mm/swap: make release_pages() and put_pages() match

From: Ira Weiny <[email protected]>

RFC I have no idea if this is correct or not. But looking at
release_pages() I see a call to both __ClearPageActive() and
__ClearPageWaiters() while in __page_cache_release() I do not.

Is this a bug which needs to be fixed? Did I miss clearing active
somewhere else in the call chain of put_page?

This was found via code inspection while determining if release_pages()
and the new put_user_pages() could be interchangeable.

Signed-off-by: Ira Weiny <[email protected]>
---
mm/swap.c | 1 +
1 file changed, 1 insertion(+)

diff --git a/mm/swap.c b/mm/swap.c
index 3a75722e68a9..9d0432baddb0 100644
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -69,6 +69,7 @@ static void __page_cache_release(struct page *page)
del_page_from_lru_list(page, lruvec, page_off_lru(page));
spin_unlock_irqrestore(&pgdat->lru_lock, flags);
}
+ __ClearPageActive(page);
__ClearPageWaiters(page);
mem_cgroup_uncharge(page);
}
--
2.20.1

2019-05-24 19:47:41

by Shakeel Butt

Subject: Re: [PATCH RFC] mm/swap: make release_pages() and put_pages() match

On Fri, May 24, 2019 at 12:33 PM <[email protected]> wrote:
>
> From: Ira Weiny <[email protected]>
>
> RFC I have no idea if this is correct or not. But looking at
> release_pages() I see a call to both __ClearPageActive() and
> __ClearPageWaiters() while in __page_cache_release() I do not.
>
> Is this a bug which needs to be fixed? Did I miss clearing active
> somewhere else in the call chain of put_page?
>
> This was found via code inspection while determining if release_pages()
> and the new put_user_pages() could be interchangeable.
>
> Signed-off-by: Ira Weiny <[email protected]>
> ---
> mm/swap.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/mm/swap.c b/mm/swap.c
> index 3a75722e68a9..9d0432baddb0 100644
> --- a/mm/swap.c
> +++ b/mm/swap.c
> @@ -69,6 +69,7 @@ static void __page_cache_release(struct page *page)
> del_page_from_lru_list(page, lruvec, page_off_lru(page));

see page_off_lru(page) above which clear active bit.

> spin_unlock_irqrestore(&pgdat->lru_lock, flags);
> }
> + __ClearPageActive(page);
> __ClearPageWaiters(page);
> mem_cgroup_uncharge(page);
> }
> --
> 2.20.1
>

2019-05-24 20:10:07

by Ira Weiny

Subject: RE: [PATCH RFC] mm/swap: make release_pages() and put_pages() match

>
> On Fri, May 24, 2019 at 12:33 PM <[email protected]> wrote:
> >
> > From: Ira Weiny <[email protected]>
> >
> > RFC I have no idea if this is correct or not. But looking at
> > release_pages() I see a call to both __ClearPageActive() and
> > __ClearPageWaiters() while in __page_cache_release() I do not.
> >
> > Is this a bug which needs to be fixed? Did I miss clearing active
> > somewhere else in the call chain of put_page?
> >
> > This was found via code inspection while determining if
> > release_pages() and the new put_user_pages() could be interchangeable.
> >
> > Signed-off-by: Ira Weiny <[email protected]>
> > ---
> > mm/swap.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/mm/swap.c b/mm/swap.c
> > index 3a75722e68a9..9d0432baddb0 100644
> > --- a/mm/swap.c
> > +++ b/mm/swap.c
> > @@ -69,6 +69,7 @@ static void __page_cache_release(struct page *page)
> > del_page_from_lru_list(page, lruvec,
> > page_off_lru(page));
>
> see page_off_lru(page) above which clear active bit.

Thanks, Sorry for the noise,
Ira


>
> > spin_unlock_irqrestore(&pgdat->lru_lock, flags);
> > }
> > + __ClearPageActive(page);
> > __ClearPageWaiters(page);
> > mem_cgroup_uncharge(page);
> > }
> > --
> > 2.20.1
> >