Re: [PATCH] f2fs: skip direct I/O iostat work when disabled
From: Wenjie Qi
Date: Mon Jun 15 2026 - 23:09:27 EST
Yes, f2fs_update_iostat() already checks sbi->iostat_enable.
I removed the extra guards around the direct read/write byte accounting in v2.
https://lore.kernel.org/linux-f2fs-devel/20260616030655.111933-1-qiwenjie@xxxxxxxxxx/T/#u
On Wed, Jun 10, 2026 at 8:33 AM Jaegeuk Kim <jaegeuk@xxxxxxxxxx> wrote:
>
> On 05/27, Wenjie Qi wrote:
> > F2FS iostat is optional and is disabled by default, but the direct
> > I/O submit path still allocates and binds a bio_iostat_ctx, updates
> > the submit timestamp, and replaces bi_end_io for every DIO bio even
> > when sbi->iostat_enable is false.
> >
> > The end_io path also calls f2fs_update_iostat(), which returns
> > immediately when iostat is disabled. This adds avoidable per-bio
> > overhead to the default direct I/O hot path.
> >
> > Skip DIO iostat context setup and the direct read/write byte updates
> > when iostat is disabled. If iostat is enabled through sysfs before
> > submission, the existing context allocation and latency accounting path
> > is still used.
> >
> > QEMU benchmark on a 1GiB F2FS virtio-blk image, with iostat_enable=0,
> > 4KiB O_DIRECT I/O over a 64MiB file, 50000 iterations per run:
> >
> > baseline patched
> > direct_read median 65264.50 ns 55470.95 ns
> > direct_read recheck 65553.75 ns 55470.95 ns
> > direct_write median 68054.62 ns 56309.44 ns
> > direct_write recheck 66873.51 ns 56309.44 ns
> >
> > Signed-off-by: Wenjie Qi <qiwenjie@xxxxxxxxxx>
> > ---
> > fs/f2fs/file.c | 9 +++++++--
> > fs/f2fs/iostat.h | 6 ++++++
> > 2 files changed, 13 insertions(+), 2 deletions(-)
> >
> > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> > index 6edf0105dbc8..3ad8bd660b33 100644
> > --- a/fs/f2fs/file.c
> > +++ b/fs/f2fs/file.c
> > @@ -4799,6 +4799,9 @@ static void f2fs_dio_iostat_start(struct f2fs_sb_info *sbi, struct bio *bio)
> > {
> > void *bi_private = bio->bi_private;
> >
> > + if (!f2fs_iostat_enabled(sbi))
> > + return;
> > +
> > iostat_alloc_and_bind_ctx(sbi, bio, bi_private);
> > iostat_update_submit_ctx(bio, DATA);
> > bio->bi_end_io = f2fs_dio_end_bio;
> > @@ -4816,7 +4819,8 @@ static int f2fs_dio_read_end_io(struct kiocb *iocb, ssize_t size, int error,
> > dec_page_count(sbi, F2FS_DIO_READ);
> > if (error)
> > return error;
> > - f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
> > + if (f2fs_iostat_enabled(sbi))
> > + f2fs_update_iostat(sbi, NULL, APP_DIRECT_READ_IO, size);
>
> f2fs_update_iostat() checks sbi->iostat_enable?
>
> > return 0;
> > }
> >
> > @@ -5097,7 +5101,8 @@ static int f2fs_dio_write_end_io(struct kiocb *iocb, ssize_t size, int error,
> > if (error)
> > return error;
> > f2fs_update_time(sbi, REQ_TIME);
> > - f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
> > + if (f2fs_iostat_enabled(sbi))
> > + f2fs_update_iostat(sbi, NULL, APP_DIRECT_IO, size);
> > return 0;
> > }
> >
> > diff --git a/fs/f2fs/iostat.h b/fs/f2fs/iostat.h
> > index 2025225b5bed..d3ef787575be 100644
> > --- a/fs/f2fs/iostat.h
> > +++ b/fs/f2fs/iostat.h
> > @@ -44,6 +44,11 @@ struct bio_iostat_ctx {
> > struct bio_post_read_ctx *post_read_ctx;
> > };
> >
> > +static inline bool f2fs_iostat_enabled(struct f2fs_sb_info *sbi)
> > +{
> > + return sbi->iostat_enable;
> > +}
> > +
> > static inline void iostat_update_submit_ctx(struct bio *bio,
> > enum page_type type)
> > {
> > @@ -72,6 +77,7 @@ static inline void f2fs_update_iostat(struct f2fs_sb_info *sbi, struct inode *in
> > enum iostat_type type, unsigned long long io_bytes) {}
> > static inline void f2fs_update_read_folio_count(struct f2fs_sb_info *sbi,
> > struct folio *folio) {}
> > +static inline bool f2fs_iostat_enabled(struct f2fs_sb_info *sbi) { return false; }
> > static inline void iostat_update_and_unbind_ctx(struct bio *bio) {}
> > static inline void iostat_alloc_and_bind_ctx(struct f2fs_sb_info *sbi,
> > struct bio *bio, struct bio_post_read_ctx *ctx) {}
> > --
> > 2.43.0
> >