Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 7 Feb 2022 11:16:03 -0800
From:   Jaegeuk Kim <jaegeuk@kernel.org>
To:     Chao Yu <chao@kernel.org>
Cc:     linux-f2fs-devel@lists.sourceforge.net,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] f2fs: skip f2fs_preallocate_blocks() for overwrite
 case
Message-ID: <YgFv85dH0dWFTZBx@google.com>
References: <20220204091005.49407-1-chao@kernel.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220204091005.49407-1-chao@kernel.org>
Precedence: bulk

On 02/04, Chao Yu wrote:
> There is potential hangtask happened during swapfile's writeback:
> 
> - loop_kthread_worker_fn		- do_checkpoint
>   - kthread_worker_fn
>    - loop_queue_work
>     - lo_rw_aio
>      - f2fs_file_write_iter
>       - f2fs_preallocate_blocks
>        - f2fs_map_blocks
> 					 - down_write
>         - down_read
>          - rwsem_down_read_slowpath
>           - schedule
> 
> One cause is f2fs_preallocate_blocks() will always be called no matter
> the physical block addresses are allocated or not.
> 
> This patch tries to check whether block addresses are all allocated with
> i_size and i_blocks of inode, it's rough because blocks can be allocated
> beyond i_size, however, we can afford skipping block preallocation in this
> condition since it's not necessary to do preallocation all the time.
> 
> Signed-off-by: Chao Yu <chao@kernel.org>
> ---
> v2:
> - check overwrite case with i_size and i_blocks roughly.
>  fs/f2fs/file.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> index cfdc41f87f5d..09565d10611d 100644
> --- a/fs/f2fs/file.c
> +++ b/fs/f2fs/file.c
> @@ -4390,6 +4390,16 @@ static int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *iter,
>  	int flag;
>  	int ret;
>  
> +	/*
> +	 * It tries to check whether block addresses are all allocated,
> +	 * it's rough because blocks can be allocated beyond i_size,
> +	 * however, we can afford skipping block preallocation since
> +	 * it's not necessary all the time.
> +	 */
> +	if (F2FS_BLK_ALIGN(i_size_read(inode)) ==
> +			SECTOR_TO_BLOCK(inode->i_blocks))

Do we count i_blocks only for data?

> +		return 0;
> +
>  	/* If it will be an out-of-place direct write, don't bother. */
>  	if (dio && f2fs_lfs_mode(sbi))
>  		return 0;
> -- 
> 2.32.0