Re: [PATCH v9 02/32] virtio: struct virtio_config_ops add callbacks for queue_reset

From: Jason Wang <jasowang@redhat.com>
To: Xuan Zhuo <xuanzhuo@linux.alibaba.com>,
	virtualization@lists.linux-foundation.org
Cc: Jeff Dike <jdike@addtoit.com>,
	Richard Weinberger <richard@nod.at>,
	Anton Ivanov <anton.ivanov@cambridgegreys.com>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	"David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	Hans de Goede <hdegoede@redhat.com>,
	Mark Gross <markgross@kernel.org>,
	Vadim Pasternak <vadimp@nvidia.com>,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	Mathieu Poirier <mathieu.poirier@linaro.org>,
	Cornelia Huck <cohuck@redhat.com>,
	Halil Pasic <pasic@linux.ibm.com>,
	Heiko Carstens <hca@linux.ibm.com>,
	Vasily Gorbik <gor@linux.ibm.com>,
	Christian Borntraeger <borntraeger@linux.ibm.com>,
	Alexander Gordeev <agordeev@linux.ibm.com>,
	Sven Schnelle <svens@linux.ibm.com>,
	Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Jesper Dangaard Brouer <hawk@kernel.org>,
	John Fastabend <john.fastabend@gmail.com>,
	Johannes Berg <johannes.berg@intel.com>,
	Vincent Whitchurch <vincent.whitchurch@axis.com>,
	linux-um@lists.infradead.org, netdev@vger.kernel.org,
	platform-driver-x86@vger.kernel.org,
	linux-remoteproc@vger.kernel.org, linux-s390@vger.kernel.org,
	kvm@vger.kernel.org, bpf@vger.kernel.org
Subject: Re: [PATCH v9 02/32] virtio: struct virtio_config_ops add callbacks for queue_reset
Date: Tue, 12 Apr 2022 10:47:41 +0800	[thread overview]
Message-ID: <9868de23-c171-2492-a43f-78f51df84640@redhat.com> (raw)
In-Reply-To: <20220406034346.74409-3-xuanzhuo@linux.alibaba.com>

在 2022/4/6 上午11:43, Xuan Zhuo 写道:
> Performing reset on a queue is divided into four steps:
>
>   1. transport: notify the device to reset the queue
>   2. vring:     recycle the buffer submitted
>   3. vring:     reset/resize the vring (may re-alloc)
>   4. transport: mmap vring to device, and enable the queue

Nit: it looks to me we'd better say it's an example (since step 3 or 
even 2 is not a must).

>
> In order to support queue reset, add two callbacks(reset_vq,
> enable_reset_vq) in struct virtio_config_ops to implement steps 1 and 4.
>
> Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> ---
>   include/linux/virtio_config.h | 12 ++++++++++++
>   1 file changed, 12 insertions(+)
>
> diff --git a/include/linux/virtio_config.h b/include/linux/virtio_config.h
> index 4d107ad31149..d4adcd0e1c57 100644
> --- a/include/linux/virtio_config.h
> +++ b/include/linux/virtio_config.h
> @@ -74,6 +74,16 @@ struct virtio_shm_region {
>    * @set_vq_affinity: set the affinity for a virtqueue (optional).
>    * @get_vq_affinity: get the affinity for a virtqueue (optional).
>    * @get_shm_region: get a shared memory region based on the index.
> + * @reset_vq: reset a queue individually (optional).
> + *	vq: the virtqueue
> + *	Returns 0 on success or error status
> + *	reset_vq will guarantee that the callbacks are disabled and synchronized.
> + *	Except for the callback, the caller should guarantee that the vring is

I wonder what's the implications for virtio hardening[1]. In that 
series, we agree to have a synchronize_vqs() config ops to make sure 
callbacks are synchronized.

It uses a global flag and a device wise synchronization mechanism. It 
looks to me we need to switch to

1) per virtqueue flag
2) per virtqueue synchronization

Thanks

> + *	not accessed by any functions of virtqueue.
> + * @enable_reset_vq: enable a reset queue
> + *	vq: the virtqueue
> + *	Returns 0 on success or error status
> + *	If reset_vq is set, then enable_reset_vq must also be set.
>    */
>   typedef void vq_callback_t(struct virtqueue *);
>   struct virtio_config_ops {
> @@ -100,6 +110,8 @@ struct virtio_config_ops {
>   			int index);
>   	bool (*get_shm_region)(struct virtio_device *vdev,
>   			       struct virtio_shm_region *region, u8 id);
> +	int (*reset_vq)(struct virtqueue *vq);
> +	int (*enable_reset_vq)(struct virtqueue *vq);
>   };
>   
>   /* If driver didn't advertise the feature, it will never appear. */