Re: [PATCH RFC 4/7] virtio_blk: use dummy virtqueue_notify() to detect host kick error

[Rusty Russell <rusty@rustcorp.com.au>]

Heinz Graalfs <graalfs@linux.vnet.ibm.com> writes:
> Deleting the disk and partitions in virtblk_remove() via del_gendisk() causes
> never ending waits when trying to synch dirty inode pages.
>
> A dummy virtqueue_notify() in virtblk_remove() is used to detect a host
> notification error, latter occurs when block device was hot-unplugged.
> When the dummy host kick failed blk_cleanup_queue() should be invoked
> prior to del_gendisk().
>
> Signed-off-by: Heinz Graalfs <graalfs@linux.vnet.ibm.com>
> ---
>  drivers/block/virtio_blk.c | 10 ++++++++--
>  1 file changed, 8 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
> index 6472395..98f081a 100644
> --- a/drivers/block/virtio_blk.c
> +++ b/drivers/block/virtio_blk.c
> @@ -880,8 +880,14 @@ static void virtblk_remove(struct virtio_device *vdev)
>  	vblk->config_enable = false;
>  	mutex_unlock(&vblk->config_lock);
>  
> -	del_gendisk(vblk->disk);
> -	blk_cleanup_queue(vblk->disk->queue);
> +	virtqueue_notify(vblk->vq);
> +	if (virtqueue_is_broken(vblk->vq)) {
> +		blk_cleanup_queue(vblk->disk->queue);
> +		del_gendisk(vblk->disk);
> +	} else {
> +		del_gendisk(vblk->disk);
> +		blk_cleanup_queue(vblk->disk->queue);
> +	}

This seems horribly wrong.  Firstly, it's ugly to have a gratuitous
kick.  Secondly, it's racy: what if there's a hot unplug (or other
failure) after your virtqueue_is_broken() test?  We should be doing I/O
failures which should be handled by del_gendisk() correctly.

Cheers,
Rusty.