[PATCH] exfat: fdatasync flag should be same like generic_write_sync()

[PATCH] exfat: fdatasync flag should be same like generic_write_sync()

[Cixi Geng]

From: Zhengxu Zhang <zhengxu.zhang@unisoc.com>

Test: androbench by default setting, use 64GB sdcard.
 the random write speed:
	without this patch 3.5MB/s
	with this patch 7MB/s

After patch "11a347fb6cef", the random write speed decreased significantly.
the .write_iter() interface had been modified, and check the differences with
generic_file_write_iter(), when calling generic_rite_sync() and
exfat_file_write_iter() to call vfs_fsync_range(), the fdatasync flag is wrong,
and make not use the fdatasync mode, and make random write speed decreased.

so make the fdatasync flag like generic_write_sync().

Fixes: 11a347fb6cef ("exfat: change to get file size from DataLength")

Signed-off-by: Zhengxu Zhang <zhengxu.zhang@unisoc.com>
---
 fs/exfat/file.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/fs/exfat/file.c b/fs/exfat/file.c
index 841a5b18e3df..01d983fa7f55 100644
--- a/fs/exfat/file.c
+++ b/fs/exfat/file.c
@@ -625,7 +625,7 @@ static ssize_t exfat_file_write_iter(struct kiocb *iocb, struct iov_iter *iter)
 
 	if (iocb_is_dsync(iocb) && iocb->ki_pos > pos) {
 		ssize_t err = vfs_fsync_range(file, pos, iocb->ki_pos - 1,
-				iocb->ki_flags & IOCB_SYNC);
+				(iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
 		if (err < 0)
 			return err;
 	}
-- 
2.25.1

Re: [PATCH] exfat: fdatasync flag should be same like generic_write_sync()

[Yuezhang.Mo]

> generic_file_write_iter(), when calling generic_rite_sync() and

s/_rite/_write

> --- a/fs/exfat/file.c
> +++ b/fs/exfat/file.c
> @@ -625,7 +625,7 @@ static ssize_t exfat_file_write_iter(struct kiocb *iocb, struct iov_iter *iter)
> 
>        if (iocb_is_dsync(iocb) && iocb->ki_pos > pos) {
>                 ssize_t err = vfs_fsync_range(file, pos, iocb->ki_pos - 1,
> -                               iocb->ki_flags & IOCB_SYNC);
> +                               (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);

How about calling generic_write_sync() instead of vfs_fsync_range(), like in 
generic_file_write_iter()?

答复: [PATCH] exfat: fdatasync flag should be same like generic_write_sync()

[张政旭 (Zhengxu Zhang)]

> -----邮件原件-----
> 发件人: Yuezhang.Mo@sony.com <Yuezhang.Mo@sony.com>
> 发送时间: 2025年6月13日 18:14
> 收件人: Cixi Geng <cixi.geng@linux.dev>; linkinjeon@kernel.org;
> sj1557.seo@samsung.com
> 抄送: linux-fsdevel@vger.kernel.org; linux-kernel@vger.kernel.org; 张政旭
> (Zhengxu Zhang) <Zhengxu.Zhang@unisoc.com>
> 主题: Re: [PATCH] exfat: fdatasync flag should be same like generic_write_sync()
> 
> 
> 
> > generic_file_write_iter(), when calling generic_rite_sync() and
> 
> s/_rite/_write
>
I will fix this by next patch.
> > --- a/fs/exfat/file.c
> > +++ b/fs/exfat/file.c
> > @@ -625,7 +625,7 @@ static ssize_t exfat_file_write_iter(struct kiocb *iocb,
> struct iov_iter *iter)
> >
> >        if (iocb_is_dsync(iocb) && iocb->ki_pos > pos) {
> >                 ssize_t err = vfs_fsync_range(file, pos, iocb->ki_pos - 1,
> > -                               iocb->ki_flags & IOCB_SYNC);
> > +                               (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
> 
> How about calling generic_write_sync() instead of vfs_fsync_range(), like in
> generic_file_write_iter()?
The second arg of vfs_fsync_range "pos" maybe changed by valid_size (if pos > valid_size). 
It can not replace by iocb->ki_pos - ret (ret by __generic_file_write_iter).
So current way maybe better.

Re: [PATCH] exfat: fdatasync flag should be same like generic_write_sync()

[Yuezhang.Mo]

> > > --- a/fs/exfat/file.c
> > > +++ b/fs/exfat/file.c
> > > @@ -625,7 +625,7 @@ static ssize_t exfat_file_write_iter(struct kiocb *iocb,
> > > struct iov_iter *iter)
> > >
> > >        if (iocb_is_dsync(iocb) && iocb->ki_pos > pos) {
> > >                 ssize_t err = vfs_fsync_range(file, pos, iocb->ki_pos - 1,
> > > -                               iocb->ki_flags & IOCB_SYNC);
> > > +                               (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
> > 
> > How about calling generic_write_sync() instead of vfs_fsync_range(), like in
> > generic_file_write_iter()?
> The second arg of vfs_fsync_range "pos" maybe changed by valid_size (if pos > valid_size). 
> It can not replace by iocb->ki_pos - ret (ret by __generic_file_write_iter).
> So current way maybe better.

Here we synchronize the areas written by exfat_extend_valid_size() and
__generic_file_write_iter() if valid_size < pos.

The lengths of these two write areas are 'pos-valid_size' and 'ret'.
We can use generic_write_sync() and pass it the sum of these two lengths.

Of course, regardless of whether valid_size < pos, exfat_file_write_iter() only 
needs to return the length written by __generic_file_write_iter().

答复: [PATCH] exfat: fdatasync flag should be same like generic_write_sync()

[张政旭 (Zhengxu Zhang)]

> -----邮件原件-----
> 发件人: Yuezhang.Mo@sony.com <Yuezhang.Mo@sony.com>
> 发送时间: 2025年6月17日 11:32
> 收件人: 张政旭 (Zhengxu Zhang) <Zhengxu.Zhang@unisoc.com>; Cixi Geng
> <cixi.geng@linux.dev>; linkinjeon@kernel.org; sj1557.seo@samsung.com
> 抄送: linux-fsdevel@vger.kernel.org; linux-kernel@vger.kernel.org; 王皓 (Hao_hao
> Wang) <Hao_hao.Wang@unisoc.com>
> 主题: Re: [PATCH] exfat: fdatasync flag should be same like generic_write_sync()
> 
> 
> 
> > > > --- a/fs/exfat/file.c
> > > > +++ b/fs/exfat/file.c
> > > > @@ -625,7 +625,7 @@ static ssize_t exfat_file_write_iter(struct kiocb *iocb,
> > > > struct iov_iter *iter)
> > > >
> > > >        if (iocb_is_dsync(iocb) && iocb->ki_pos > pos) {
> > > >                 ssize_t err = vfs_fsync_range(file, pos, iocb->ki_pos - 1,
> > > > -                               iocb->ki_flags & IOCB_SYNC);
> > > > +                               (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
> > >
> > > How about calling generic_write_sync() instead of vfs_fsync_range(), like in
> > > generic_file_write_iter()?
> > The second arg of vfs_fsync_range "pos" maybe changed by valid_size (if pos >
> valid_size).
> > It can not replace by iocb->ki_pos - ret (ret by __generic_file_write_iter).
> > So current way maybe better.
> 
> Here we synchronize the areas written by exfat_extend_valid_size() and
> __generic_file_write_iter() if valid_size < pos.
> 
> The lengths of these two write areas are 'pos-valid_size' and 'ret'.
> We can use generic_write_sync() and pass it the sum of these two lengths.
> 
> Of course, regardless of whether valid_size < pos, exfat_file_write_iter() only
> needs to return the length written by __generic_file_write_iter().

I try the sum of 'pos-valid_size' and 'ret',like this:
        if (iocb->ki_pos > pos) {
                ssize_t err = generic_write_sync(iocb, pos + ret - valid_size);
                if (err < 0)
                        return err;
        }
The test crashed, that maybe io error. 
So I try a simple way that use iocb->ki_pos - pos. like this:
        if (iocb->ki_pos > pos) {
                ssize_t err = generic_write_sync(iocb, iocb->ki_pos - pos);
                if (err < 0)
                        return err;
        }
The test pass. pls check again.

Re: [PATCH] exfat: fdatasync flag should be same like generic_write_sync()

[Yuezhang.Mo]

> > > > > --- a/fs/exfat/file.c
> > > > > +++ b/fs/exfat/file.c
> > > > > @@ -625,7 +625,7 @@ static ssize_t exfat_file_write_iter(struct kiocb *iocb,
> > > > > struct iov_iter *iter)
> > > > >
> > > > >        if (iocb_is_dsync(iocb) && iocb->ki_pos > pos) {
> > > > >                 ssize_t err = vfs_fsync_range(file, pos, iocb->ki_pos - 1,
> > > > > -                               iocb->ki_flags & IOCB_SYNC);
> > > > > +                               (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
> > > >
> > > > How about calling generic_write_sync() instead of vfs_fsync_range(), like in
> > > > generic_file_write_iter()?
> > > The second arg of vfs_fsync_range "pos" maybe changed by valid_size (if pos >
> > valid_size).
> > > It can not replace by iocb->ki_pos - ret (ret by __generic_file_write_iter).
> > > So current way maybe better.
> >
> > Here we synchronize the areas written by exfat_extend_valid_size() and
> > __generic_file_write_iter() if valid_size < pos.
> >
> > The lengths of these two write areas are 'pos-valid_size' and 'ret'.
> > We can use generic_write_sync() and pass it the sum of these two lengths.
> > 
> > Of course, regardless of whether valid_size < pos, exfat_file_write_iter() only
> > needs to return the length written by __generic_file_write_iter().
> 
> I try the sum of 'pos-valid_size' and 'ret',like this:
>         if (iocb->ki_pos > pos) {
>                 ssize_t err = generic_write_sync(iocb, pos + ret - valid_size);
>                 if (err < 0)
>                         return err;
>         }
> The test crashed, that maybe io error.

I think the crash happens when pos < valid_size, because exfat_extend_valid_size()
does not write data in this case.

> So I try a simple way that use iocb->ki_pos - pos. like this:
>         if (iocb->ki_pos > pos) {
>                 ssize_t err = generic_write_sync(iocb, iocb->ki_pos - pos);
>                 if (err < 0)
>                         return err;
>         }
> The test pass. pls check again.

'pos' is set to the write position of exfat_extend_valid_size() by:

if (pos > valid_size)
          pos = valid_size;

'iocb->ki_pos - pos' is the total write length, this way is fine.