[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <YuXZSTw4reBDtLgk@google.com>
Date: Sat, 30 Jul 2022 18:22:17 -0700
From: Jaegeuk Kim <jaegeuk@...nel.org>
To: Eunhee Rho <eunhee83.rho@...sung.com>
Cc: "chao@...nel.org" <chao@...nel.org>,
"linux-f2fs-devel@...ts.sourceforge.net"
<linux-f2fs-devel@...ts.sourceforge.net>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v0] f2fs: allow direct read for zoned device
On 07/29, Eunhee Rho wrote:
> For zoned devices, f2fs forbids direct IO and forces buffered IO
> to serialize write IOs. However, the constraint does not apply to
> read IOs.
>
> Signed-off-by: Eunhee Rho <eunhee83.rho@...sung.com>
> ---
> fs/f2fs/f2fs.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index d0f428aef34b..f69731f17402 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -4471,7 +4471,7 @@ static inline bool f2fs_force_buffered_io(struct inode *inode,
> * for blkzoned device, fallback direct IO to buffered IO, so
> * all IOs can be serialized by log-structured write.
> */
> - if (f2fs_sb_has_blkzoned(sbi))
> + if (f2fs_sb_has_blkzoned(sbi) && (rw == WRITE))
> return true;
Actually we don't need to check the device type, but rely on the
filesystem allocation policy like the below lfs mode. Can we remove this
line simply?
> if (f2fs_lfs_mode(sbi) && (rw == WRITE)) {
> if (block_unaligned_IO(inode, iocb, iter))
> --
> 2.25.1
Powered by blists - more mailing lists