Re: [PATCH 2/5] iomap: use accelerated zeroing on a block device to zero a file range

[Dave Chinner <david@fromorbit.com>]

On Fri, Sep 17, 2021 at 06:30:55PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
> 
> Create a function that ensures that the storage backing part of a file
> contains zeroes and will not trip over old media errors if the contents
> are re-read.
> 
> Signed-off-by: Darrick J. Wong <djwong@kernel.org>
> ---
>  fs/iomap/direct-io.c  |   75 +++++++++++++++++++++++++++++++++++++++++++++++++
>  include/linux/iomap.h |    3 ++
>  2 files changed, 78 insertions(+)
> 
> 
> diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
> index 4ecd255e0511..48826a49f976 100644
> --- a/fs/iomap/direct-io.c
> +++ b/fs/iomap/direct-io.c
> @@ -652,3 +652,78 @@ iomap_dio_rw(struct kiocb *iocb, struct iov_iter *iter,
>  	return iomap_dio_complete(dio);
>  }
>  EXPORT_SYMBOL_GPL(iomap_dio_rw);
> +
> +static loff_t
> +iomap_zeroinit_iter(struct iomap_iter *iter)
> +{
> +	struct iomap *iomap = &iter->iomap;
> +	const struct iomap *srcmap = iomap_iter_srcmap(iter);
> +	const u64 start = iomap->addr + iter->pos - iomap->offset;
> +	const u64 nr_bytes = iomap_length(iter);
> +	sector_t sector = start >> SECTOR_SHIFT;
> +	sector_t nr_sectors = nr_bytes >> SECTOR_SHIFT;
> +	int ret;
> +
> +	if (!iomap->bdev)
> +		return -ECANCELED;
> +
> +	/* The physical extent must be sector-aligned for block layer APIs. */
> +	if ((start | nr_bytes) & (SECTOR_SIZE - 1))
> +		return -EINVAL;
> +
> +	/* Must be able to zero storage directly without fs intervention. */
> +	if (iomap->flags & IOMAP_F_SHARED)
> +		return -ECANCELED;
> +	if (srcmap != iomap)
> +		return -ECANCELED;
> +
> +	switch (iomap->type) {
> +	case IOMAP_MAPPED:
> +		ret = blkdev_issue_zeroout(iomap->bdev, sector, nr_sectors,
> +				GFP_KERNEL, 0);

Pretty sure this needs to use BLKDEV_ZERO_NOUNMAP.

The whole point of this is having zeroed space allocated ready for
write on return, so having the hardware optimise away the physical
storage zeroing by punching a hole in it's backing store and then
potentially getting ENOSPC on the next write to this range would be
.... suboptimal.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com