linux-kernel - Re: [PATCH] virtio_ring: Use workqueue to execute virtqueue callback

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20200115084601-mutt-send-email-mst@kernel.org>
Date:   Wed, 15 Jan 2020 08:49:16 -0500
From:   "Michael S. Tsirkin" <mst@...hat.com>
To:     Clement Leger <cleger@...ray.eu>
Cc:     Jason Wang <jasowang@...hat.com>,
        virtualization@...ts.linux-foundation.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] virtio_ring: Use workqueue to execute virtqueue callback

On Wed, Jan 15, 2020 at 01:05:35PM +0100, Clement Leger wrote:
> Currently, in vring_interrupt, the vq callbacks are called directly.
> However, these callbacks are not meant to be executed in IRQ context.
> They do not return any irq_return_t value and some of them can do
> locking (rpmsg_recv_done -> rpmsg_recv_single -> mutex_lock).

That's a bug in rpmsg. Pls fix there.

> When compiled with DEBUG_ATOMIC_SLEEP, the kernel will spit out warnings
> when such case shappen.
> 
> In order to allow calling these callbacks safely (without sleeping in
> IRQ context), execute them in a workqueue if needed.
> 
> Signed-off-by: Clement Leger <cleger@...ray.eu>

If applied this would slow data path handling of VQ events
significantly. Teaching callbacks to return irqreturn_t
might be a good idea, though it's not an insignificant
amount of work.


> ---
>  drivers/virtio/virtio_ring.c | 20 ++++++++++++++++++--
>  1 file changed, 18 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> index 867c7ebd3f10..0e4d0e5ca227 100644
> --- a/drivers/virtio/virtio_ring.c
> +++ b/drivers/virtio/virtio_ring.c
> @@ -183,6 +183,9 @@ struct vring_virtqueue {
>  	/* DMA, allocation, and size information */
>  	bool we_own_ring;
>  
> +	/* Work struct for vq callback handling */
> +	struct work_struct work;
> +
>  #ifdef DEBUG
>  	/* They're supposed to lock for us. */
>  	unsigned int in_use;
> @@ -2029,6 +2032,16 @@ static inline bool more_used(const struct vring_virtqueue *vq)
>  	return vq->packed_ring ? more_used_packed(vq) : more_used_split(vq);
>  }
>  
> +static void vring_work_handler(struct work_struct *work_struct)
> +{
> +	struct vring_virtqueue *vq = container_of(work_struct,
> +						  struct vring_virtqueue,
> +						  work);
> +	pr_debug("virtqueue callback for %p (%p)\n", vq, vq->vq.callback);
> +
> +	vq->vq.callback(&vq->vq);
> +}
> +
>  irqreturn_t vring_interrupt(int irq, void *_vq)
>  {
>  	struct vring_virtqueue *vq = to_vvq(_vq);
> @@ -2041,9 +2054,8 @@ irqreturn_t vring_interrupt(int irq, void *_vq)
>  	if (unlikely(vq->broken))
>  		return IRQ_HANDLED;
>  
> -	pr_debug("virtqueue callback for %p (%p)\n", vq, vq->vq.callback);
>  	if (vq->vq.callback)
> -		vq->vq.callback(&vq->vq);
> +		schedule_work(&vq->work);
>  
>  	return IRQ_HANDLED;
>  }
> @@ -2110,6 +2122,8 @@ struct virtqueue *__vring_new_virtqueue(unsigned int index,
>  					vq->split.avail_flags_shadow);
>  	}
>  
> +	INIT_WORK(&vq->work, vring_work_handler);
> +
>  	vq->split.desc_state = kmalloc_array(vring.num,
>  			sizeof(struct vring_desc_state_split), GFP_KERNEL);
>  	if (!vq->split.desc_state) {
> @@ -2179,6 +2193,8 @@ void vring_del_virtqueue(struct virtqueue *_vq)
>  {
>  	struct vring_virtqueue *vq = to_vvq(_vq);
>  
> +	cancel_work_sync(&vq->work);
> +
>  	if (vq->we_own_ring) {
>  		if (vq->packed_ring) {
>  			vring_free_queue(vq->vq.vdev,
> -- 
> 2.15.0.276.g89ea799