Re: [PATCH 11/12] iomap: add read_folio_range() handler for buffered writes

[Brian Foster <bfoster@redhat.com>]

On Fri, Jun 27, 2025 at 09:02:44AM +0200, Christoph Hellwig wrote:
> From: Joanne Koong <joannelkoong@gmail.com>
> 
> Add a read_folio_range() handler for buffered writes that filesystems
> may pass in if they wish to provide a custom handler for synchronously
> reading in the contents of a folio.
> 
> Signed-off-by: Joanne Koong <joannelkoong@gmail.com>
> [hch: renamed to read_folio_range, pass less arguments]
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  .../filesystems/iomap/operations.rst          |  6 +++++
>  fs/iomap/buffered-io.c                        | 25 +++++++++++--------
>  include/linux/iomap.h                         | 10 ++++++++
>  3 files changed, 31 insertions(+), 10 deletions(-)
> 
> diff --git a/Documentation/filesystems/iomap/operations.rst b/Documentation/filesystems/iomap/operations.rst
> index 167d3ca7819c..04432f40e7a2 100644
> --- a/Documentation/filesystems/iomap/operations.rst
> +++ b/Documentation/filesystems/iomap/operations.rst
> @@ -68,6 +68,8 @@ The following address space operations can be wrapped easily:
>       void (*put_folio)(struct inode *inode, loff_t pos, unsigned copied,
>                         struct folio *folio);
>       bool (*iomap_valid)(struct inode *inode, const struct iomap *iomap);
> +     int (*read_folio_range)(const struct iomap_iter *iter,
> +     			struct folio *folio, loff_t pos, size_t len);

Whitespace ^

>   };
>  
>  iomap calls these functions:
...
> diff --git a/fs/iomap/buffered-io.c b/fs/iomap/buffered-io.c
> index a77686977a2e..1a9ade77aeeb 100644
> --- a/fs/iomap/buffered-io.c
> +++ b/fs/iomap/buffered-io.c
> @@ -667,22 +667,23 @@ iomap_write_failed(struct inode *inode, loff_t pos, unsigned len)
>  					 pos + len - 1);
>  }
>  
> -static int iomap_read_folio_sync(loff_t block_start, struct folio *folio,
> -		size_t poff, size_t plen, const struct iomap *iomap)
> +static int iomap_read_folio_range(const struct iomap_iter *iter,
> +		struct folio *folio, loff_t pos, size_t len)
>  {
> +	const struct iomap *srcmap = iomap_iter_srcmap(iter);
>  	struct bio_vec bvec;
>  	struct bio bio;
>  
> -	bio_init(&bio, iomap->bdev, &bvec, 1, REQ_OP_READ);
> -	bio.bi_iter.bi_sector = iomap_sector(iomap, block_start);
> -	bio_add_folio_nofail(&bio, folio, plen, poff);
> +	bio_init(&bio, srcmap->bdev, &bvec, 1, REQ_OP_READ);
> +	bio.bi_iter.bi_sector = iomap_sector(srcmap, pos);
> +	bio_add_folio_nofail(&bio, folio, len, offset_in_folio(folio, pos));
>  	return submit_bio_wait(&bio);
>  }

Hmm, so this kind of makes my brain hurt... pos here is now the old
block_start and len is the old plen. We used to pass poff to the
add_folio_nofail() call, and now that is dropped and instead we use
offset_in_folio(..., pos). The old poff is an output of the previous
iomap_adjust_read_range() call, which is initially set to
offset_in_folio(folio, *pos), of which *pos is block_start and is bumped
in that function in the same places that poff is. Therefore old poff and
new offset_in_folio(folio, pos) are logically equivalent. Am I following
that correctly?

Brian

>  
> -static int __iomap_write_begin(const struct iomap_iter *iter, size_t len,
> +static int __iomap_write_begin(const struct iomap_iter *iter,
> +		const struct iomap_write_ops *write_ops, size_t len,
>  		struct folio *folio)
>  {
> -	const struct iomap *srcmap = iomap_iter_srcmap(iter);
>  	struct iomap_folio_state *ifs;
>  	loff_t pos = iter->pos;
>  	loff_t block_size = i_blocksize(iter->inode);
> @@ -731,8 +732,12 @@ static int __iomap_write_begin(const struct iomap_iter *iter, size_t len,
>  			if (iter->flags & IOMAP_NOWAIT)
>  				return -EAGAIN;
>  
> -			status = iomap_read_folio_sync(block_start, folio,
> -					poff, plen, srcmap);
> +			if (write_ops && write_ops->read_folio_range)
> +				status = write_ops->read_folio_range(iter,
> +						folio, block_start, plen);
> +			else
> +				status = iomap_read_folio_range(iter,
> +						folio, block_start, plen);
>  			if (status)
>  				return status;
>  		}
> @@ -848,7 +853,7 @@ static int iomap_write_begin(struct iomap_iter *iter,
>  	else if (srcmap->flags & IOMAP_F_BUFFER_HEAD)
>  		status = __block_write_begin_int(folio, pos, len, NULL, srcmap);
>  	else
> -		status = __iomap_write_begin(iter, len, folio);
> +		status = __iomap_write_begin(iter, write_ops, len, folio);
>  
>  	if (unlikely(status))
>  		goto out_unlock;
> diff --git a/include/linux/iomap.h b/include/linux/iomap.h
> index 482787013ff7..b3588dd43105 100644
> --- a/include/linux/iomap.h
> +++ b/include/linux/iomap.h
> @@ -166,6 +166,16 @@ struct iomap_write_ops {
>  	 * locked by the iomap code.
>  	 */
>  	bool (*iomap_valid)(struct inode *inode, const struct iomap *iomap);
> +
> +	/*
> +	 * Optional if the filesystem wishes to provide a custom handler for
> +	 * reading in the contents of a folio, otherwise iomap will default to
> +	 * submitting a bio read request.
> +	 *
> +	 * The read must be done synchronously.
> +	 */
> +	int (*read_folio_range)(const struct iomap_iter *iter,
> +			struct folio *folio, loff_t pos, size_t len);
>  };
>  
>  /*
> -- 
> 2.47.2
> 
>