Linux filesystem development
 help / color / mirror / Atom feed
From: "Darrick J. Wong" <djwong@kernel.org>
To: Al Viro <viro@zeniv.linux.org.uk>
Cc: linux-fsdevel@vger.kernel.org, Christoph Hellwig <hch@lst.de>,
	Jens Axboe <axboe@kernel.dk>,
	Matthew Wilcox <willy@infradead.org>
Subject: Re: [PATCH 02/10] teach iomap_dio_rw() to suppress dsync
Date: Wed, 8 Jun 2022 08:17:49 -0700	[thread overview]
Message-ID: <YqC9nZwgOHk2x0pg@magnolia> (raw)
In-Reply-To: <20220607233143.1168114-2-viro@zeniv.linux.org.uk>

On Tue, Jun 07, 2022 at 11:31:35PM +0000, Al Viro wrote:
> New flag, equivalent to removal of IOCB_DSYNC from iocb flags.
> This mimics what btrfs is doing (and that's what btrfs will
> switch to).  However, I'm not at all sure that we want to
> suppress REQ_FUA for those - all btrfs hack really cares about
> is suppression of generic_write_sync().  For now let's keep
> the existing behaviour, but I really want to hear more detailed
> arguments pro or contra.
> 
> [folded brain fix from willy]
> 
> Suggested-by: Christoph Hellwig <hch@lst.de>
> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>

Looks ok,
Reviewed-by: Darrick J. Wong <djwong@kernel.org>

--D

> ---
>  fs/iomap/direct-io.c  | 20 +++++++++++---------
>  include/linux/iomap.h |  6 ++++++
>  2 files changed, 17 insertions(+), 9 deletions(-)
> 
> diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
> index 370c3241618a..c10c69e2de24 100644
> --- a/fs/iomap/direct-io.c
> +++ b/fs/iomap/direct-io.c
> @@ -548,17 +548,19 @@ __iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
>  		}
>  
>  		/* for data sync or sync, we need sync completion processing */
> -		if (iocb->ki_flags & IOCB_DSYNC)
> +		if (iocb->ki_flags & IOCB_DSYNC &&
> +		    !(dio_flags & IOMAP_DIO_NOSYNC)) {
>  			dio->flags |= IOMAP_DIO_NEED_SYNC;
>  
> -		/*
> -		 * For datasync only writes, we optimistically try using FUA for
> -		 * this IO.  Any non-FUA write that occurs will clear this flag,
> -		 * hence we know before completion whether a cache flush is
> -		 * necessary.
> -		 */
> -		if ((iocb->ki_flags & (IOCB_DSYNC | IOCB_SYNC)) == IOCB_DSYNC)
> -			dio->flags |= IOMAP_DIO_WRITE_FUA;
> +		       /*
> +			* For datasync only writes, we optimistically try
> +			* using FUA for this IO.  Any non-FUA write that
> +			* occurs will clear this flag, hence we know before
> +			* completion whether a cache flush is necessary.
> +			*/
> +			if (!(iocb->ki_flags & IOCB_SYNC))
> +				dio->flags |= IOMAP_DIO_WRITE_FUA;
> +		}
>  	}
>  
>  	if (dio_flags & IOMAP_DIO_OVERWRITE_ONLY) {
> diff --git a/include/linux/iomap.h b/include/linux/iomap.h
> index e552097c67e0..c8622d8f064e 100644
> --- a/include/linux/iomap.h
> +++ b/include/linux/iomap.h
> @@ -353,6 +353,12 @@ struct iomap_dio_ops {
>   */
>  #define IOMAP_DIO_PARTIAL		(1 << 2)
>  
> +/*
> + * The caller will sync the write if needed; do not sync it within
> + * iomap_dio_rw.  Overrides IOMAP_DIO_FORCE_WAIT.
> + */
> +#define IOMAP_DIO_NOSYNC		(1 << 3)
> +
>  ssize_t iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
>  		const struct iomap_ops *ops, const struct iomap_dio_ops *dops,
>  		unsigned int dio_flags, void *private, size_t done_before);
> -- 
> 2.30.2
> 

  parent reply	other threads:[~2022-06-08 15:21 UTC|newest]

Thread overview: 93+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-06-07  4:08 [RFC][PATCHES] iov_iter stuff Al Viro
2022-06-07  4:09 ` [PATCH 1/9] No need of likely/unlikely on calls of check_copy_size() Al Viro
2022-06-07  4:41   ` Christoph Hellwig
2022-06-07 11:49   ` Christian Brauner
2022-06-07  4:09 ` [PATCH 2/9] btrfs_direct_write(): cleaner way to handle generic_write_sync() suppression Al Viro
2022-06-07  4:42   ` Christoph Hellwig
2022-06-07 16:06     ` Al Viro
2022-06-07 23:27       ` Al Viro
2022-06-07 23:31         ` [PATCH 01/10] No need of likely/unlikely on calls of check_copy_size() Al Viro
2022-06-07 23:31           ` [PATCH 02/10] teach iomap_dio_rw() to suppress dsync Al Viro
2022-06-08  6:18             ` Christoph Hellwig
2022-06-08 15:17             ` Darrick J. Wong [this message]
2022-06-10 11:38             ` Christian Brauner
2022-06-07 23:31           ` [PATCH 03/10] btrfs: use IOMAP_DIO_NOSYNC Al Viro
2022-06-08  6:18             ` Christoph Hellwig
2022-06-10 11:09             ` Christian Brauner
2022-06-07 23:31           ` [PATCH 04/10] struct file: use anonymous union member for rcuhead and llist Al Viro
2022-06-07 23:31           ` [PATCH 05/10] iocb: delay evaluation of IS_SYNC(...) until we want to check IOCB_DSYNC Al Viro
2022-06-10 11:41             ` Christian Brauner
2022-06-07 23:31           ` [PATCH 06/10] keep iocb_flags() result cached in struct file Al Viro
2022-06-09  0:35             ` Dave Chinner
2022-06-10 11:43             ` Christian Brauner
2022-06-07 23:31           ` [PATCH 07/10] copy_page_{to,from}_iter(): switch iovec variants to generic Al Viro
2022-06-07 23:31           ` [PATCH 08/10] new iov_iter flavour - ITER_UBUF Al Viro
2022-06-07 23:31           ` [PATCH 09/10] switch new_sync_{read,write}() to ITER_UBUF Al Viro
2022-06-10 11:11             ` Christian Brauner
2022-06-07 23:31           ` [PATCH 10/10] iov_iter_bvec_advance(): don't bother with bvec_iter Al Viro
2022-06-08  6:16       ` [PATCH 2/9] btrfs_direct_write(): cleaner way to handle generic_write_sync() suppression Christoph Hellwig
2022-06-07 14:49   ` Matthew Wilcox
2022-06-07 20:17     ` Al Viro
2022-06-07  4:10 ` [PATCH 3/9] struct file: use anonymous union member for rcuhead and llist Al Viro
2022-06-07 10:18   ` Jan Kara
2022-06-07 11:46   ` Christian Brauner
2022-06-07  4:11 ` [PATCH 4/9] iocb: delay evaluation of IS_SYNC(...) until we want to check IOCB_DSYNC Al Viro
2022-06-07 10:34   ` Jan Kara
2022-06-07 15:34     ` Al Viro
2022-06-07  4:11 ` [PATCH 5/9] keep iocb_flags() result cached in struct file Al Viro
2022-06-07  4:12 ` [PATCH 6/9] copy_page_{to,from}_iter(): switch iovec variants to generic Al Viro
2022-06-07  4:12 ` [PATCH 7/9] new iov_iter flavour - ITER_UBUF Al Viro
2022-06-07  4:13 ` [PATCH 8/9] switch new_sync_{read,write}() to ITER_UBUF Al Viro
2022-06-07  4:13 ` [PATCH 9/9] iov_iter_bvec_advance(): don't bother with bvec_iter Al Viro
2022-06-08 19:28 ` [RFC][PATCHES] iov_iter stuff Sedat Dilek
2022-06-08 20:39   ` Al Viro
2022-06-09 19:10     ` Sedat Dilek
2022-06-09 19:22       ` Matthew Wilcox
2022-06-09 19:58         ` Matthew Wilcox
2022-06-09 19:45       ` Al Viro
2022-06-17 22:30 ` Jens Axboe
2022-06-17 22:48   ` Al Viro
2022-06-18  5:27     ` Al Viro
2022-06-18  5:35       ` [PATCH 01/31] splice: stop abusing iov_iter_advance() to flush a pipe Al Viro
2022-06-18  5:35         ` [PATCH 02/31] ITER_PIPE: helper for getting pipe buffer by index Al Viro
2022-06-18  5:35         ` [PATCH 03/31] ITER_PIPE: helpers for adding pipe buffers Al Viro
2022-06-18  5:35         ` [PATCH 04/31] ITER_PIPE: allocate buffers as we go in copy-to-pipe primitives Al Viro
2022-06-19  1:34           ` Al Viro
2022-06-18  5:35         ` [PATCH 05/31] ITER_PIPE: fold push_pipe() into __pipe_get_pages() Al Viro
2022-06-18  5:35         ` [PATCH 06/31] ITER_PIPE: lose iter_head argument of __pipe_get_pages() Al Viro
2022-06-18  5:35         ` [PATCH 07/31] ITER_PIPE: clean pipe_advance() up Al Viro
2022-06-18  5:35         ` [PATCH 08/31] ITER_PIPE: clean iov_iter_revert() Al Viro
2022-06-18  5:35         ` [PATCH 09/31] ITER_PIPE: cache the type of last buffer Al Viro
2022-06-18  5:35         ` [PATCH 10/10] iov_iter_bvec_advance(): don't bother with bvec_iter Al Viro
2022-06-18  5:35         ` [PATCH 10/31] ITER_PIPE: fold data_start() and pipe_space_for_user() together Al Viro
2022-06-19  2:25           ` Al Viro
2022-06-18  5:35         ` [PATCH 11/31] iov_iter_get_pages{,_alloc}(): cap the maxsize with LONG_MAX Al Viro
2022-06-18  5:35         ` [PATCH 12/31] iov_iter_get_pages_alloc(): lift freeing pages array on failure exits into wrapper Al Viro
2022-06-18  5:35         ` [PATCH 13/31] iov_iter_get_pages(): sanity-check arguments Al Viro
2022-06-19  3:07           ` Al Viro
2022-06-18  5:35         ` [PATCH 14/31] unify pipe_get_pages() and pipe_get_pages_alloc() Al Viro
2022-06-18  5:35         ` [PATCH 15/31] unify xarray_get_pages() and xarray_get_pages_alloc() Al Viro
2022-06-18  5:35         ` [PATCH 16/31] unify the rest of iov_iter_get_pages()/iov_iter_get_pages_alloc() guts Al Viro
2022-06-19  3:56           ` Al Viro
2022-06-18  5:35         ` [PATCH 17/31] ITER_XARRAY: don't open-code DIV_ROUND_UP() Al Viro
2022-06-18  5:35         ` [PATCH 18/31] iov_iter: lift dealing with maxpages out of first_{iovec,bvec}_segment() Al Viro
2022-06-18  5:35         ` [PATCH 19/31] iov_iter: massage calling conventions for first_{iovec,bvec}_segment() Al Viro
2022-06-18 11:13           ` Al Viro
2022-06-18 11:18             ` Al Viro
2022-06-18  5:35         ` [PATCH 20/31] found_iovec_segment(): just return address Al Viro
2022-06-18  5:35         ` [PATCH 21/31] fold __pipe_get_pages() into pipe_get_pages() Al Viro
2022-06-18  5:35         ` [PATCH 22/31] iov_iter: saner helper for page array allocation Al Viro
2022-06-18 11:14           ` Al Viro
2022-06-18  5:35         ` [PATCH 23/31] iov_iter: advancing variants of iov_iter_get_pages{,_alloc}() Al Viro
2022-06-18  5:35         ` [PATCH 24/31] block: convert to " Al Viro
2022-06-18  5:35         ` [PATCH 25/31] iter_to_pipe(): switch to advancing variant of iov_iter_get_pages() Al Viro
2022-06-18  5:35         ` [PATCH 26/31] af_alg_make_sg(): " Al Viro
2022-06-18  5:35         ` [PATCH 27/31] 9p: convert to advancing variant of iov_iter_get_pages_alloc() Al Viro
2022-06-18  5:35         ` [PATCH 28/31] ceph: switch the last caller " Al Viro
2022-06-18  5:35         ` [PATCH 29/31] get rid of non-advancing variants Al Viro
2022-06-18 11:14           ` Al Viro
2022-06-18  5:35         ` [PATCH 30/31] pipe_get_pages(): switch to append_pipe() Al Viro
2022-06-19  4:01           ` Al Viro
2022-06-19  4:09             ` Al Viro
2022-06-18  5:35         ` [PATCH 31/31] expand those iov_iter_advance() Al Viro
2022-06-18 11:14           ` Al Viro

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=YqC9nZwgOHk2x0pg@magnolia \
    --to=djwong@kernel.org \
    --cc=axboe@kernel.dk \
    --cc=hch@lst.de \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=viro@zeniv.linux.org.uk \
    --cc=willy@infradead.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox