Re: [PATCH v3 10/10] zonefs: use REQ_OP_ZONE_APPEND for sync DIO

[Christoph Hellwig <hch@infradead.org>]

On Sat, Mar 28, 2020 at 01:50:12AM +0900, Johannes Thumshirn wrote:
> Synchronous direct I/O to a sequential write only zone can be issued using
> the new REQ_OP_ZONE_APPEND request operation. As dispatching multiple
> BIOs can potentially result in reordering, we cannot support asynchronous
> IO via this interface.

We trivially can if the write size is smaller than the supported zone
append size.  We could slightly less trivially by chaining a new
submission after the first bio completes.

> +static void zonefs_zone_append_bio_endio(struct bio *bio)
> +{
> +	struct task_struct *waiter = bio->bi_private;
> +
> +	WRITE_ONCE(bio->bi_private, NULL);
> +	blk_wake_io_task(waiter);
> +
> +	bio_release_pages(bio, false);
> +	bio_put(bio);
> +}
> +
> +static ssize_t zonefs_file_dio_append(struct kiocb *iocb, struct iov_iter *from)
> +{
> +	struct inode *inode = file_inode(iocb->ki_filp);
> +	struct zonefs_inode_info *zi = ZONEFS_I(inode);
> +	struct block_device *bdev = inode->i_sb->s_bdev;
> +	ssize_t ret = 0;
> +	ssize_t size;
> +	struct bio *bio;
> +	unsigned max;
> +	int nr_pages;
> +	blk_qc_t qc;
> +
> +	nr_pages = iov_iter_npages(from, BIO_MAX_PAGES);
> +	if (!nr_pages)
> +		return 0;
> +
> +	max = queue_max_zone_append_sectors(bdev_get_queue(bdev)) << 9;
> +	max = ALIGN_DOWN(max, inode->i_sb->s_blocksize);
> +	iov_iter_truncate(from, max);
> +
> +	bio = bio_alloc_bioset(GFP_NOFS, nr_pages, &fs_bio_set);
> +	if (!bio)
> +		return -ENOMEM;
> +
> +	bio_set_dev(bio, bdev);
> +	bio->bi_iter.bi_sector = zi->i_zsector;
> +	bio->bi_write_hint = iocb->ki_hint;
> +	bio->bi_private = current;
> +	bio->bi_end_io = zonefs_zone_append_bio_endio;
> +	bio->bi_ioprio = iocb->ki_ioprio;
> +	bio->bi_opf = REQ_OP_ZONE_APPEND | REQ_SYNC | REQ_IDLE;
> +	if (iocb->ki_flags & IOCB_DSYNC)
> +		bio->bi_opf |= REQ_FUA;
> +
> +	ret = bio_iov_iter_get_pages(bio, from);
> +	if (unlikely(ret)) {
> +		bio->bi_status = BLK_STS_IOERR;
> +		bio_endio(bio);
> +		return ret;
> +	}
> +	size = bio->bi_iter.bi_size;
> +	task_io_account_write(ret);
> +
> +	if (iocb->ki_flags & IOCB_HIPRI)
> +		bio_set_polled(bio, iocb);
> +
> +	bio_get(bio);
> +	qc = submit_bio(bio);
> +	for (;;) {
> +		set_current_state(TASK_UNINTERRUPTIBLE);
> +		if (!READ_ONCE(bio->bi_private))
> +			break;
> +		if (!(iocb->ki_flags & IOCB_HIPRI) ||
> +		    !blk_poll(bdev_get_queue(bdev), qc, true))
> +			io_schedule();
> +	}
> +	__set_current_state(TASK_RUNNING);
> +
> +	if (unlikely(bio->bi_status))
> +		ret = blk_status_to_errno(bio->bi_status);
> +
> +	bio_put(bio);
> +
> +	zonefs_file_write_dio_end_io(iocb, size, ret, 0);
> +	if (ret >= 0) {
> +		iocb->ki_pos += size;
> +		return size;
> +	}
> +
> +	return ret;

This looks like no one waits for I/O completion?   Also it looks
like it silently causes a short write, which probably needs to be
documented..