linux-xfs.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Nikolay Borisov <nborisov@suse.com>
To: Dave Chinner <david@fromorbit.com>, linux-xfs@vger.kernel.org
Cc: linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH 06/16] iomap: support block size > page size for direct IO
Date: Thu, 8 Nov 2018 13:28:46 +0200	[thread overview]
Message-ID: <ef5a8148-fff2-d323-8b69-01a2062dde9d@suse.com> (raw)
In-Reply-To: <20181107063127.3902-7-david@fromorbit.com>



On 7.11.18 г. 8:31 ч., Dave Chinner wrote:
> From: Dave Chinner <dchinner@redhat.com>
> 
> iomap_dio_rw() has all the infrastructure in place to support block
> size > page size filesystems because it is essentially just
> sub-block DIO. It needs help, however, with the sub-block zeroing
> code (needs multi-page IOs) page cache invalidation over the block
> being written.
> 
> Signed-off-by: Dave Chinner <dchinner@redhat.com>
> ---
>  fs/iomap.c | 65 ++++++++++++++++++++++++++++++++++++++++--------------
>  1 file changed, 49 insertions(+), 16 deletions(-)
> 
> diff --git a/fs/iomap.c b/fs/iomap.c
> index 16d16596b00f..8878b1f1f9c7 100644
> --- a/fs/iomap.c
> +++ b/fs/iomap.c
> @@ -1548,21 +1548,34 @@ static void iomap_dio_bio_end_io(struct bio *bio)
>  	}
>  }
>  
> +/*
> + * With zeroing for block size larger than page size, the zeroing length can
> + * span multiple pages.
> + */
> +#define howmany(x, y) (((x)+((y)-1))/(y))

nit: This could be replaced by DIV_ROUND_UP

>  static blk_qc_t
>  iomap_dio_zero(struct iomap_dio *dio, struct iomap *iomap, loff_t pos,
>  		unsigned len)
>  {
>  	struct page *page = ZERO_PAGE(0);
>  	struct bio *bio;
> +	int npages = howmany(len, PAGE_SIZE);
> +
> +	WARN_ON_ONCE(npages > 16);
>  
> -	bio = bio_alloc(GFP_KERNEL, 1);
> +	bio = bio_alloc(GFP_KERNEL, npages);
>  	bio_set_dev(bio, iomap->bdev);
>  	bio->bi_iter.bi_sector = iomap_sector(iomap, pos);
>  	bio->bi_private = dio;
>  	bio->bi_end_io = iomap_dio_bio_end_io;
>  
> -	get_page(page);
> -	__bio_add_page(bio, page, len, 0);
> +	while (npages-- > 0) {
> +		unsigned plen = min_t(unsigned, PAGE_SIZE, len);
> +		get_page(page);
> +		__bio_add_page(bio, page, plen, 0);
> +		len -= plen;
> +	}
> +	WARN_ON(len != 0);
>  	bio_set_op_attrs(bio, REQ_OP_WRITE, REQ_SYNC | REQ_IDLE);
>  
>  	atomic_inc(&dio->ref);
> @@ -1752,6 +1765,38 @@ iomap_dio_actor(struct inode *inode, loff_t pos, loff_t length,
>  	}
>  }
>  
> +/*
> + * This is lifted almost straight from xfs_flush_unmap_range(). Need a generic
> + * version of the block size rounding for these purposes.
> + */
> +static int
> +iomap_flush_unmap_range(struct file *f, loff_t offset, loff_t len)
> +{
> +	struct inode *inode = file_inode(f);
> +	loff_t rounding, start, end;
> +	int ret;
> +
> +	rounding = max_t(loff_t, i_blocksize(inode), PAGE_SIZE);
> +	start = round_down(offset, rounding);
> +	end = round_up(offset + len, rounding) - 1;
> +
> +	ret = filemap_write_and_wait_range(inode->i_mapping, start, end);
> +	if (ret)
> +		return ret;
> +
> +	/*
> +	 * Try to invalidate cache pages for the range we're direct
> +	 * writing.  If this invalidation fails, tough, the write will
> +	 * still work, but racing two incompatible write paths is a
> +	 * pretty crazy thing to do, so we don't support it 100%.
> +	 */
> +	ret = invalidate_inode_pages2_range(inode->i_mapping,
> +			start >> PAGE_SHIFT, end >> PAGE_SHIFT);
> +	if (ret)
> +		dio_warn_stale_pagecache(f);
> +	return 0;
> +}
> +
>  /*
>   * iomap_dio_rw() always completes O_[D]SYNC writes regardless of whether the IO
>   * is being issued as AIO or not.  This allows us to optimise pure data writes
> @@ -1829,22 +1874,10 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
>  		flags |= IOMAP_NOWAIT;
>  	}
>  
> -	ret = filemap_write_and_wait_range(mapping, start, end);
> +	ret = iomap_flush_unmap_range(iocb->ki_filp, start, end);
>  	if (ret)
>  		goto out_free_dio;
>  
> -	/*
> -	 * Try to invalidate cache pages for the range we're direct
> -	 * writing.  If this invalidation fails, tough, the write will
> -	 * still work, but racing two incompatible write paths is a
> -	 * pretty crazy thing to do, so we don't support it 100%.
> -	 */
> -	ret = invalidate_inode_pages2_range(mapping,
> -			start >> PAGE_SHIFT, end >> PAGE_SHIFT);
> -	if (ret)
> -		dio_warn_stale_pagecache(iocb->ki_filp);
> -	ret = 0;
> -
>  	if (iov_iter_rw(iter) == WRITE && !dio->wait_for_completion &&
>  	    !inode->i_sb->s_dio_done_wq) {
>  		ret = sb_init_dio_done_wq(inode->i_sb);
> 

  reply	other threads:[~2018-11-08 21:03 UTC|newest]

Thread overview: 44+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-11-07  6:31 [RFC PATCH 00/16] xfs: Block size > PAGE_SIZE support Dave Chinner
2018-11-07  6:31 ` [PATCH 01/16] xfs: drop ->writepage completely Dave Chinner
2018-11-09 15:12   ` Christoph Hellwig
2018-11-12 21:08     ` Dave Chinner
2021-02-02 20:51       ` Darrick J. Wong
2018-11-07  6:31 ` [PATCH 02/16] xfs: move writepage context warnings to writepages Dave Chinner
2018-11-07  6:31 ` [PATCH 03/16] xfs: finobt AG reserves don't consider last AG can be a runt Dave Chinner
2018-11-07 16:55   ` Darrick J. Wong
2018-11-09  0:21     ` Dave Chinner
2018-11-07  6:31 ` [PATCH 04/16] xfs: extent shifting doesn't fully invalidate page cache Dave Chinner
2018-11-07  6:31 ` [PATCH 05/16] iomap: sub-block dio needs to zeroout beyond EOF Dave Chinner
2018-11-09 15:15   ` Christoph Hellwig
2018-11-07  6:31 ` [PATCH 06/16] iomap: support block size > page size for direct IO Dave Chinner
2018-11-08 11:28   ` Nikolay Borisov [this message]
2018-11-09 15:18   ` Christoph Hellwig
2018-11-11  1:12     ` Dave Chinner
2018-11-07  6:31 ` [PATCH 07/16] iomap: prepare buffered IO paths for block size > page size Dave Chinner
2018-11-09 15:19   ` Christoph Hellwig
2018-11-11  1:15     ` Dave Chinner
2018-11-07  6:31 ` [PATCH 08/16] iomap: mode iomap_zero_range and friends Dave Chinner
2018-11-09 15:19   ` Christoph Hellwig
2018-11-07  6:31 ` [PATCH 09/16] iomap: introduce zero-around functionality Dave Chinner
2018-11-07  6:31 ` [PATCH 10/16] iomap: enable zero-around for iomap_zero_range() Dave Chinner
2018-11-07  6:31 ` [PATCH 11/16] iomap: Don't mark partial pages zeroing uptodate for zero-around Dave Chinner
2018-11-07  6:31 ` [PATCH 12/16] iomap: zero-around in iomap_page_mkwrite Dave Chinner
2018-11-07  6:31 ` [PATCH 13/16] xfs: add zero-around controls to iomap Dave Chinner
2018-11-07  6:31 ` [PATCH 14/16] xfs: align writepages to large block sizes Dave Chinner
2018-11-09 15:22   ` Christoph Hellwig
2018-11-11  1:20     ` Dave Chinner
2018-11-11 16:32       ` Christoph Hellwig
2018-11-14 14:19   ` Brian Foster
2018-11-14 21:18     ` Dave Chinner
2018-11-15 12:55       ` Brian Foster
2018-11-16  6:19         ` Dave Chinner
2018-11-16 13:29           ` Brian Foster
2018-11-19  1:14             ` Dave Chinner
2018-11-07  6:31 ` [PATCH 15/16] xfs: expose block size in stat Dave Chinner
2018-11-07  6:31 ` [PATCH 16/16] xfs: enable block size larger than page size support Dave Chinner
2018-11-07 17:14 ` [RFC PATCH 00/16] xfs: Block size > PAGE_SIZE support Darrick J. Wong
2018-11-07 22:04   ` Dave Chinner
2018-11-08  1:38     ` Darrick J. Wong
2018-11-08  9:04       ` Dave Chinner
2018-11-08 22:17         ` Darrick J. Wong
2018-11-08 22:22           ` Dave Chinner

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=ef5a8148-fff2-d323-8b69-01a2062dde9d@suse.com \
    --to=nborisov@suse.com \
    --cc=david@fromorbit.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-xfs@vger.kernel.org \
    --subject='Re: [PATCH 06/16] iomap: support block size > page size for direct IO' \
    /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

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).