Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Tue, 10 Sep 2019 12:23:04 -0700
From:   Minchan Kim <minchan@kernel.org>
To:     sunqiuyang <sunqiuyang@huawei.com>
Cc:     linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: [PATCH 1/1] mm/migrate: fix list corruption in migration of
 non-LRU movable pages
Message-ID: <20190910192304.GA220078@google.com>
References: <20190903082746.20736-1-sunqiuyang@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20190903082746.20736-1-sunqiuyang@huawei.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Sep 03, 2019 at 04:27:46PM +0800, sunqiuyang wrote:
> From: Qiuyang Sun <sunqiuyang@huawei.com>
> 
> Currently, after a page is migrated, it
> 1) has its PG_isolated flag cleared in move_to_new_page(), and
> 2) is deleted from its LRU list (cc->migratepages) in unmap_and_move().
> However, between steps 1) and 2), the page could be isolated by another
> thread in isolate_movable_page(), and added to another LRU list, leading
> to list_del corruption later.

Once non-LRU page is migrated out successfully, driver should clear
the movable flag in the page. Look at reset_page in zs_page_migrate.
So, other thread couldn't isolate the page during the window.

If I miss something, let me know it.
Thanks.

> 
> This patch fixes the bug by moving list_del into the critical section
> protected by lock_page(), so that a page will not be isolated again before
> it has been deleted from its LRU list.
> 
> Signed-off-by: Qiuyang Sun <sunqiuyang@huawei.com>
> ---
>  mm/migrate.c | 11 +++--------
>  1 file changed, 3 insertions(+), 8 deletions(-)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index a42858d..c58a606 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1124,6 +1124,8 @@ static int __unmap_and_move(struct page *page, struct page *newpage,
>  	/* Drop an anon_vma reference if we took one */
>  	if (anon_vma)
>  		put_anon_vma(anon_vma);
> +	if (rc != -EAGAIN)
> +		list_del(&page->lru);
>  	unlock_page(page);
>  out:
>  	/*
> @@ -1190,6 +1192,7 @@ static ICE_noinline int unmap_and_move(new_page_t get_new_page,
>  			put_new_page(newpage, private);
>  		else
>  			put_page(newpage);
> +		list_del(&page->lru);
>  		goto out;
>  	}
>  
> @@ -1200,14 +1203,6 @@ static ICE_noinline int unmap_and_move(new_page_t get_new_page,
>  out:
>  	if (rc != -EAGAIN) {
>  		/*
> -		 * A page that has been migrated has all references
> -		 * removed and will be freed. A page that has not been
> -		 * migrated will have kepts its references and be
> -		 * restored.
> -		 */
> -		list_del(&page->lru);
> -
> -		/*
>  		 * Compaction can migrate also non-LRU pages which are
>  		 * not accounted to NR_ISOLATED_*. They can be recognized
>  		 * as __PageMovable
> -- 
> 1.8.3.1
> 
>