[Devel] [PATCH VZ10 v3 5/5] drivers/vhost/blk: rework queue/backend setup

[Konstantin Khorenko]

On 8/25/26 14:51, Andrey Zhadchenko wrote:
> vhost_blk_setup() is pretty bad: silently refusing changed vq->num
> if requests are already allocated, fetching user input second time
> (double-fetch vulnerability).
> To handle this, tie request allocation to backend existence. After
> all, if there is no backend, there is no point in having requests.
> Also expand it to get rid of boilerplate drop_backend, flush,
> fput sequence in a few places.
> 
> https://virtuozzo.atlassian.net/browse/VSTOR-138640
> Signed-off-by: Andrey Zhadchenko <andrey.zhadchenko at virtuozzo.com>
> ---
...

> +static int vhost_blk_setup_vqs(struct vhost_blk *blk)
> +{
> +	struct vhost_virtqueue *vq;
> +	int i;
> +
> +	for (i = 0; i < VHOST_BLK_VQ_MAX; i++) {
> +		vq = &blk->vqs[i].vq;
> +
> +		if (!vhost_vq_is_setup(vq))
> +			continue;
> +
> +		blk->vqs[i].req = kvmalloc_array(vq->num, sizeof(struct vhost_blk_req),
> +						 GFP_KERNEL);
> +		if (!blk->vqs[i].req)
> +			return -ENOMEM;
> +
> +		mutex_lock(&vq->mutex);
> +		vhost_vq_set_backend(vq, blk->backend);
> +		if (vhost_vq_init_access(vq)) {
> +			mutex_unlock(&vq->mutex);
> +			return -EFAULT;
> +		}
> +		mutex_unlock(&vq->mutex);

Well, i agree that currently vhost_vq_init_access() can return -EFAULT only as an error,
but may be it's still worth to write a more generic return ret code?

        int ret, i;

        for (i = 0; i < VHOST_BLK_VQ_MAX; i++) {
...
                mutex_lock(&vq->mutex);
                vhost_vq_set_backend(vq, blk->backend);
                ret = vhost_vq_init_access(vq);
                mutex_unlock(&vq->mutex);
                if (ret)
                        return ret;



> +	}
> +
> +	return 0;
>  }
>...