[PATCH] f2fs: move dir data flush to write checkpoint process

[PATCH] f2fs: move dir data flush to write checkpoint process

[Yunlei He]

This patch move dir data flush to write checkpoint process, by
doing this, it may reduce some time for dir fsync.

pre:
	-f2fs_do_sync_file enter
		-file_write_and_wait_range  <- flush & wait
		-write_checkpoint
			-do_checkpoint	    <- wait all
	-f2fs_do_sync_file exit

now:
	-f2fs_do_sync_file enter
		-write_checkpoint
			-block_operations   <- flush dir & no wait
			-do_checkpoint	    <- wait all
	-f2fs_do_sync_file exit

Signed-off-by: Yunlei He <heyunlei@huawei.com>
---
 fs/f2fs/file.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
index 88b1246..9eaf07f 100644
--- a/fs/f2fs/file.c
+++ b/fs/f2fs/file.c
@@ -216,6 +216,9 @@ static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
 
 	trace_f2fs_sync_file_enter(inode);
 
+	if (S_ISDIR(inode->i_mode))
+		goto go_write;
+
 	/* if fdatasync is triggered, let's do in-place-update */
 	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
 		set_inode_flag(inode, FI_NEED_IPU);
-- 
1.9.1

Re: [PATCH] f2fs: move dir data flush to write checkpoint process

[Chao Yu]

On 2018/11/6 10:25, Yunlei He wrote:
> This patch move dir data flush to write checkpoint process, by
> doing this, it may reduce some time for dir fsync.
> 
> pre:
> 	-f2fs_do_sync_file enter
> 		-file_write_and_wait_range  <- flush & wait
> 		-write_checkpoint
> 			-do_checkpoint	    <- wait all
> 	-f2fs_do_sync_file exit
> 
> now:
> 	-f2fs_do_sync_file enter
> 		-write_checkpoint
> 			-block_operations   <- flush dir & no wait
> 			-do_checkpoint	    <- wait all
> 	-f2fs_do_sync_file exit
> 
> Signed-off-by: Yunlei He <heyunlei@huawei.com>
> ---
>  fs/f2fs/file.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> index 88b1246..9eaf07f 100644
> --- a/fs/f2fs/file.c
> +++ b/fs/f2fs/file.c
> @@ -216,6 +216,9 @@ static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
>  
>  	trace_f2fs_sync_file_enter(inode);
>  
> +	if (S_ISDIR(inode->i_mode))
> +		goto go_write;

Not sure, will we skip calling f2fs_write_inode due to this check? so we
will miss some metadata update cached in inode cache?

Thanks,

> +
>  	/* if fdatasync is triggered, let's do in-place-update */
>  	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
>  		set_inode_flag(inode, FI_NEED_IPU);
> 

Re: [PATCH] f2fs: move dir data flush to write checkpoint process

[heyunlei]

>-----Original Message-----
>From: Yuchao (T)
>Sent: Tuesday, November 06, 2018 5:20 PM
>To: heyunlei; jaegeuk@kernel.org; chao@kernel.org
>Cc: linux-f2fs-devel@lists.sourceforge.net; miaoxie (A)
>Subject: Re: [f2fs-dev][PATCH] f2fs: move dir data flush to write checkpoint process
>
>
>
>On 2018/11/6 10:25, Yunlei He wrote:
>> This patch move dir data flush to write checkpoint process, by
>> doing this, it may reduce some time for dir fsync.
>>
>> pre:
>> 	-f2fs_do_sync_file enter
>> 		-file_write_and_wait_range  <- flush & wait
>> 		-write_checkpoint
>> 			-do_checkpoint	    <- wait all
>> 	-f2fs_do_sync_file exit
>>
>> now:
>> 	-f2fs_do_sync_file enter
>> 		-write_checkpoint
>> 			-block_operations   <- flush dir & no wait
>> 			-do_checkpoint	    <- wait all
>> 	-f2fs_do_sync_file exit
>>
>> Signed-off-by: Yunlei He <heyunlei@huawei.com>
>> ---
>>  fs/f2fs/file.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
>> index 88b1246..9eaf07f 100644
>> --- a/fs/f2fs/file.c
>> +++ b/fs/f2fs/file.c
>> @@ -216,6 +216,9 @@ static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end,
>>
>>  	trace_f2fs_sync_file_enter(inode);
>>
>> +	if (S_ISDIR(inode->i_mode))
>> +		goto go_write;
>
>Not sure, will we skip calling f2fs_write_inode due to this check? so we
>will miss some metadata update cached in inode cache?

Here，new checkpoint will update all dirty inodes metadata from inode cache to inode page？
Or maybe we just skip file_write_and_wait_range action for I found it will always write a checkpoint
even the dir is clean.

And then, I test the original code：

Mkdir  A
Sync A
While true
	Fsync A

Everytime fsync the clean directory A，f2fs will write a new checkpoint also. 

Thanks.
>
>Thanks,
>
>> +
>>  	/* if fdatasync is triggered, let's do in-place-update */
>>  	if (datasync || get_dirty_pages(inode) <= SM_I(sbi)->min_fsync_blocks)
>>  		set_inode_flag(inode, FI_NEED_IPU);
>>


_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel