Re: [f2fs-dev] [PATCH] f2fs: fix deadlock between quota writes and checkpoint

From: Chao Yu <yuchao0@huawei.com>
To: Jaegeuk Kim <jaegeuk@kernel.org>, <linux-kernel@vger.kernel.org>,
	<linux-f2fs-devel@lists.sourceforge.net>,
	<kernel-team@android.com>
Cc: Daeho Jeong <daehojeong@google.com>
Subject: Re: [f2fs-dev] [PATCH] f2fs: fix deadlock between quota writes and checkpoint
Date: Tue, 4 Aug 2020 11:01:10 +0800	[thread overview]
Message-ID: <670f35e1-872b-6602-320c-dd73bcb62510@huawei.com> (raw)
In-Reply-To: <20200729070244.584518-1-jaegeuk@kernel.org>

On 2020/7/29 15:02, Jaegeuk Kim wrote:
> f2fs_write_data_pages(quota_mapping)
>   __f2fs_write_data_pages             f2fs_write_checkpoint
>    * blk_start_plug(&plug);
>    * add bio in write_io[DATA]
>                                        - block_operations
>                                        - skip syncing quota by
>                                                  >DEFAULT_RETRY_QUOTA_FLUSH_COUNT
>                                        - down_write(&sbi->node_write);

f2fs_flush_merged_writes() will be called after block_operations(), why this doesn't
help?

>    - f2fs_write_single_data_page
>      - f2fs_do_write_data_page
>        - f2fs_outplace_write_data
>          - do_write_page
>             - f2fs_allocate_data_block
>              - down_write(node_write)
>                                        - f2fs_wait_on_all_pages(F2FS_WB_CP_DATA);
> 
> Signed-off-by: Daeho Jeong <daehojeong@google.com>
> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
> ---
>   fs/f2fs/checkpoint.c | 2 ++
>   1 file changed, 2 insertions(+)
> 
> diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
> index 8c782d3f324f0..99c8061da55b9 100644
> --- a/fs/f2fs/checkpoint.c
> +++ b/fs/f2fs/checkpoint.c
> @@ -1269,6 +1269,8 @@ void f2fs_wait_on_all_pages(struct f2fs_sb_info *sbi, int type)
>   		if (type == F2FS_DIRTY_META)
>   			f2fs_sync_meta_pages(sbi, META, LONG_MAX,
>   							FS_CP_META_IO);
> +		else if (type == F2FS_WB_CP_DATA)
> +			f2fs_submit_merged_write(sbi, DATA);
>   		io_schedule_timeout(DEFAULT_IO_TIMEOUT);
>   	}
>   	finish_wait(&sbi->cp_wait, &wait);
> 