[PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[shao.mingyin]

From: Shao Mingyin <shao.mingyin@zte.com.cn>

For zone append, the block layer copies req->__sector back into the
bio in blk_zone_append_update_request_bio(), so the driver must update
req->__sector with the sector reported by the device.
virtblk_request_done() does that, but the batched completion path used
by virtblk_poll() only unmaps and cleans up the requests before calling
blk_mq_end_request_batch().

A zone append completed through polling therefore keeps the sector that
was assigned at submission time, and the file system or application sees
the wrong LBA for the written data.

Move the sector update into a helper and call it from both completion
paths.

Fixes: 95bfec41bd3d ("virtio-blk: add support for zoned block devices")
Signed-off-by: Shao Mingyin <shao.mingyin@zte.com.cn>
---
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index c8bfbf8435ac..cc29b8238a4d 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -331,18 +331,25 @@ static inline u8 virtblk_vbr_status(struct virtblk_req *vbr)
 	return *((u8 *)&vbr->in_hdr + vbr->in_hdr_len - 1);
 }

+static inline void virtblk_update_zone_append_sector(struct request *req,
+						     struct virtblk_req *vbr)
+{
+	struct virtio_blk *vblk = req->mq_hctx->queue->queuedata;
+
+	if (req_op(req) == REQ_OP_ZONE_APPEND)
+		req->__sector = virtio64_to_cpu(vblk->vdev,
+						vbr->in_hdr.zone_append.sector);
+}
+
 static inline void virtblk_request_done(struct request *req)
 {
 	struct virtblk_req *vbr = blk_mq_rq_to_pdu(req);
 	blk_status_t status = virtblk_result(virtblk_vbr_status(vbr));
-	struct virtio_blk *vblk = req->mq_hctx->queue->queuedata;

 	virtblk_unmap_data(req, vbr);
 	virtblk_cleanup_cmd(req);

-	if (req_op(req) == REQ_OP_ZONE_APPEND)
-		req->__sector = virtio64_to_cpu(vblk->vdev,
-						vbr->in_hdr.zone_append.sector);
+	virtblk_update_zone_append_sector(req, vbr);

 	blk_mq_end_request(req, status);
 }
@@ -1196,8 +1203,11 @@ static void virtblk_complete_batch(struct io_comp_batch *iob)
 	struct request *req;

 	rq_list_for_each(&iob->req_list, req) {
-		virtblk_unmap_data(req, blk_mq_rq_to_pdu(req));
+		struct virtblk_req *vbr = blk_mq_rq_to_pdu(req);
+
+		virtblk_unmap_data(req, vbr);
 		virtblk_cleanup_cmd(req);
+		virtblk_update_zone_append_sector(req, vbr);
 	}
 	blk_mq_end_request_batch(iob);
 }

Re: [PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[Christoph Hellwig]

On Mon, Oct 26, 2026 at 07:55:19PM +0800, shao.mingyin@zte.com.cn wrote:
> From: Shao Mingyin <shao.mingyin@zte.com.cn>
> 
> For zone append, the block layer copies req->__sector back into the
> bio in blk_zone_append_update_request_bio(), so the driver must update
> req->__sector with the sector reported by the device.
> virtblk_request_done() does that, but the batched completion path used
> by virtblk_poll() only unmaps and cleans up the requests before calling
> blk_mq_end_request_batch().
> 
> A zone append completed through polling therefore keeps the sector that
> was assigned at submission time, and the file system or application sees
> the wrong LBA for the written data.
> 
> Move the sector update into a helper and call it from both completion
> paths.

Looks good:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Can you add something to blktests to exercises this?

Re: [PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[Stefan Hajnoczi]

On Mon, Oct 26, 2026 at 07:55:19PM +0800, shao.mingyin@zte.com.cn wrote:
> From: Shao Mingyin <shao.mingyin@zte.com.cn>
> 
> For zone append, the block layer copies req->__sector back into the
> bio in blk_zone_append_update_request_bio(), so the driver must update
> req->__sector with the sector reported by the device.
> virtblk_request_done() does that, but the batched completion path used
> by virtblk_poll() only unmaps and cleans up the requests before calling
> blk_mq_end_request_batch().
> 
> A zone append completed through polling therefore keeps the sector that
> was assigned at submission time, and the file system or application sees
> the wrong LBA for the written data.
> 
> Move the sector update into a helper and call it from both completion
> paths.
> 
> Fixes: 95bfec41bd3d ("virtio-blk: add support for zoned block devices")
> Signed-off-by: Shao Mingyin <shao.mingyin@zte.com.cn>

Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>

Re: [PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[Michael S. Tsirkin]

On Mon, Oct 26, 2026 at 07:55:19PM +0800, shao.mingyin@zte.com.cn wrote:
> From: Shao Mingyin <shao.mingyin@zte.com.cn>
> 
> For zone append, the block layer copies req->__sector back into the
> bio in blk_zone_append_update_request_bio(), so the driver must update
> req->__sector with the sector reported by the device.
> virtblk_request_done() does that, but the batched completion path used
> by virtblk_poll() only unmaps and cleans up the requests before calling
> blk_mq_end_request_batch().
> 
> A zone append completed through polling therefore keeps the sector that
> was assigned at submission time, and the file system or application sees
> the wrong LBA for the written data.
> 
> Move the sector update into a helper and call it from both completion
> paths.
> 
> Fixes: 95bfec41bd3d ("virtio-blk: add support for zoned block devices")
> Signed-off-by: Shao Mingyin <shao.mingyin@zte.com.cn>

Acked-by: Michael S. Tsirkin <mst@redhat.com>

> ---
> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> index c8bfbf8435ac..cc29b8238a4d 100644
> --- a/drivers/block/virtio_blk.c
> +++ b/drivers/block/virtio_blk.c
> @@ -331,18 +331,25 @@ static inline u8 virtblk_vbr_status(struct virtblk_req *vbr)
>  	return *((u8 *)&vbr->in_hdr + vbr->in_hdr_len - 1);
>  }
> 
> +static inline void virtblk_update_zone_append_sector(struct request *req,
> +						     struct virtblk_req *vbr)
> +{
> +	struct virtio_blk *vblk = req->mq_hctx->queue->queuedata;
> +
> +	if (req_op(req) == REQ_OP_ZONE_APPEND)
> +		req->__sector = virtio64_to_cpu(vblk->vdev,
> +						vbr->in_hdr.zone_append.sector);
> +}
> +
>  static inline void virtblk_request_done(struct request *req)
>  {
>  	struct virtblk_req *vbr = blk_mq_rq_to_pdu(req);
>  	blk_status_t status = virtblk_result(virtblk_vbr_status(vbr));
> -	struct virtio_blk *vblk = req->mq_hctx->queue->queuedata;
> 
>  	virtblk_unmap_data(req, vbr);
>  	virtblk_cleanup_cmd(req);
> 
> -	if (req_op(req) == REQ_OP_ZONE_APPEND)
> -		req->__sector = virtio64_to_cpu(vblk->vdev,
> -						vbr->in_hdr.zone_append.sector);
> +	virtblk_update_zone_append_sector(req, vbr);
> 
>  	blk_mq_end_request(req, status);
>  }
> @@ -1196,8 +1203,11 @@ static void virtblk_complete_batch(struct io_comp_batch *iob)
>  	struct request *req;
> 
>  	rq_list_for_each(&iob->req_list, req) {
> -		virtblk_unmap_data(req, blk_mq_rq_to_pdu(req));
> +		struct virtblk_req *vbr = blk_mq_rq_to_pdu(req);
> +
> +		virtblk_unmap_data(req, vbr);
>  		virtblk_cleanup_cmd(req);
> +		virtblk_update_zone_append_sector(req, vbr);
>  	}
>  	blk_mq_end_request_batch(iob);
>  }

Re: [PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[Christoph Hellwig]

On Wed, Oct 28, 2026 at 10:27:08AM +0800, shao.mingyin@zte.com.cn wrote:
> I looked into adding blktests coverage, and there is a structural
> problem. blktests cannot issue REQ_OP_ZONE_APPEND from userspace: fio's
> --zone_append goes through the io_uring_cmd NVMe passthrough path, so it
> never reaches the driver's zone-append completion, and zbdioctl has no
> append operation. blktests has no zone-append test today for the same
> reason.

Right.

> That does not fit blktests, which is a block-layer test suite driven
> from userspace. Two options as I see it:
> 
>   1. an XFS reproducer (e.g. xfstests) - the one that actually
>      exercises the append completion path, or
>   2. a blktests zbd test that covers a zoned virtio-blk with poll
>      queues, which would not reach the append part itself.
> 
> I am happy to write either, or to leave it if you think the coverage
> is not worth the churn. Which do you prefer?

We do have some blktests test that use a file system in the zbd
category (zbd/009, zbd/010, zbd/014) so I think we could do this in
blktests.  I.e. do a blktest that tests zoned polled I/O if we're
on a zoned device that support polling.

Re: [PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[Christoph Hellwig]

On Tue, Nov 03, 2026 at 09:02:23AM +0100, Christoph Hellwig wrote:
> On Wed, Oct 28, 2026 at 10:27:08AM +0800, shao.mingyin@zte.com.cn wrote:
> > I looked into adding blktests coverage, and there is a structural
> > problem. blktests cannot issue REQ_OP_ZONE_APPEND from userspace: fio's
> > --zone_append goes through the io_uring_cmd NVMe passthrough path, so it
> > never reaches the driver's zone-append completion, and zbdioctl has no
> > append operation. blktests has no zone-append test today for the same
> > reason.
> 
> Right.
> 
> > That does not fit blktests, which is a block-layer test suite driven
> > from userspace. Two options as I see it:
> > 
> >   1. an XFS reproducer (e.g. xfstests) - the one that actually
> >      exercises the append completion path, or
> >   2. a blktests zbd test that covers a zoned virtio-blk with poll
> >      queues, which would not reach the append part itself.
> > 
> > I am happy to write either, or to leave it if you think the coverage
> > is not worth the churn. Which do you prefer?
> 
> We do have some blktests test that use a file system in the zbd
> category (zbd/009, zbd/010, zbd/014) so I think we could do this in
> blktests.  I.e. do a blktest that tests zoned polled I/O if we're
> on a zoned device that support polling.

OTOH xfstests might actually still fit better for this.  I guess
eiter way will work..

Re: [PATCH] virtio-blk: Fix zone append sector reporting in batched completion

[Jens Axboe]

On Mon, 26 Oct 2026 19:55:19 +0800, shao.mingyin@zte.com.cn wrote:
> For zone append, the block layer copies req->__sector back into the
> bio in blk_zone_append_update_request_bio(), so the driver must update
> req->__sector with the sector reported by the device.
> virtblk_request_done() does that, but the batched completion path used
> by virtblk_poll() only unmaps and cleans up the requests before calling
> blk_mq_end_request_batch().
> 
> [...]

Applied, thanks!

[1/1] virtio-blk: Fix zone append sector reporting in batched completion
      commit: 08a6572441a4e646c96c2b8d8ead9eea0e458332

Best regards,
-- 
Jens Axboe