Re: [PATCH v5 17/22] virtio_net: support rx/tx queue reset

["Michael S. Tsirkin" <mst@redhat.com>]

On Wed, Feb 16, 2022 at 03:56:13PM +0800, Xuan Zhuo wrote:
> On Wed, 16 Feb 2022 12:14:11 +0800, Jason Wang <jasowang@redhat.com> wrote:
> > On Mon, Feb 14, 2022 at 4:14 PM Xuan Zhuo <xuanzhuo@linux.alibaba.com> wrote:
> > >
> > > This patch implements the reset function of the rx, tx queues.
> > >
> > > Based on this function, it is possible to modify the ring num of the
> > > queue. And quickly recycle the buffer in the queue.
> > >
> > > In the process of the queue disable, in theory, as long as virtio
> > > supports queue reset, there will be no exceptions.
> > >
> > > However, in the process of the queue enable, there may be exceptions due to
> > > memory allocation.  In this case, vq is not available, but we still have
> > > to execute napi_enable(). Because napi_disable is similar to a lock,
> > > napi_enable must be called after calling napi_disable.
> > >
> > > Signed-off-by: Xuan Zhuo <xuanzhuo@linux.alibaba.com>
> > > ---
> > >  drivers/net/virtio_net.c | 123 +++++++++++++++++++++++++++++++++++++++
> > >  1 file changed, 123 insertions(+)
> > >
> > > diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> > > index 9a1445236e23..a4ffd7cdf623 100644
> > > --- a/drivers/net/virtio_net.c
> > > +++ b/drivers/net/virtio_net.c
> > > @@ -251,6 +251,11 @@ struct padded_vnet_hdr {
> > >         char padding[4];
> > >  };
> > >
> > > +static void virtnet_sq_free_unused_bufs(struct virtnet_info *vi,
> > > +                                       struct send_queue *sq);
> > > +static void virtnet_rq_free_unused_bufs(struct virtnet_info *vi,
> > > +                                       struct receive_queue *rq);
> > > +
> > >  static bool is_xdp_frame(void *ptr)
> > >  {
> > >         return (unsigned long)ptr & VIRTIO_XDP_FLAG;
> > > @@ -1369,6 +1374,9 @@ static void virtnet_napi_enable(struct virtqueue *vq, struct napi_struct *napi)
> > >  {
> > >         napi_enable(napi);
> > >
> > > +       if (vq->reset)
> > > +               return;
> > > +
> > >         /* If all buffers were filled by other side before we napi_enabled, we
> > >          * won't get another interrupt, so process any outstanding packets now.
> > >          * Call local_bh_enable after to trigger softIRQ processing.
> > > @@ -1413,6 +1421,10 @@ static void refill_work(struct work_struct *work)
> > >                 struct receive_queue *rq = &vi->rq[i];
> > >
> > >                 napi_disable(&rq->napi);
> > > +               if (rq->vq->reset) {
> > > +                       virtnet_napi_enable(rq->vq, &rq->napi);
> > > +                       continue;
> > > +               }
> > >                 still_empty = !try_fill_recv(vi, rq, GFP_KERNEL);
> > >                 virtnet_napi_enable(rq->vq, &rq->napi);
> > >
> > > @@ -1523,6 +1535,9 @@ static void virtnet_poll_cleantx(struct receive_queue *rq)
> > >         if (!sq->napi.weight || is_xdp_raw_buffer_queue(vi, index))
> > >                 return;
> > >
> > > +       if (sq->vq->reset)
> > > +               return;
> > > +
> > >         if (__netif_tx_trylock(txq)) {
> > >                 do {
> > >                         virtqueue_disable_cb(sq->vq);
> > > @@ -1769,6 +1784,114 @@ static netdev_tx_t start_xmit(struct sk_buff *skb, struct net_device *dev)
> > >         return NETDEV_TX_OK;
> > >  }
> > >
> > > +static int virtnet_rx_vq_disable(struct virtnet_info *vi,
> > > +                                struct receive_queue *rq)
> > > +{
> > > +       int err;
> > > +
> > > +       napi_disable(&rq->napi);
> > > +
> > > +       err = virtio_reset_vq(rq->vq);
> > > +       if (err)
> > > +               goto err;
> > > +
> > > +       virtnet_rq_free_unused_bufs(vi, rq);
> > > +
> > > +       vring_release_virtqueue(rq->vq);
> > > +
> > > +       return 0;
> > > +
> > > +err:
> > > +       virtnet_napi_enable(rq->vq, &rq->napi);
> > > +       return err;
> > > +}
> > > +
> > > +static int virtnet_tx_vq_disable(struct virtnet_info *vi,
> > > +                                struct send_queue *sq)
> > > +{
> > > +       struct netdev_queue *txq;
> > > +       int err, qindex;
> > > +
> > > +       qindex = sq - vi->sq;
> > > +
> > > +       txq = netdev_get_tx_queue(vi->dev, qindex);
> > > +       __netif_tx_lock_bh(txq);
> > > +
> > > +       netif_stop_subqueue(vi->dev, qindex);
> > > +       virtnet_napi_tx_disable(&sq->napi);
> > > +
> > > +       err = virtio_reset_vq(sq->vq);
> > > +       if (err) {
> > > +               virtnet_napi_tx_enable(vi, sq->vq, &sq->napi);
> > > +               netif_start_subqueue(vi->dev, qindex);
> > > +
> > > +               __netif_tx_unlock_bh(txq);
> > > +               return err;
> > > +       }
> > > +       __netif_tx_unlock_bh(txq);
> > > +
> > > +       virtnet_sq_free_unused_bufs(vi, sq);
> > > +
> > > +       vring_release_virtqueue(sq->vq);
> > > +
> > > +       return 0;
> > > +}
> > > +
> > > +static int virtnet_tx_vq_enable(struct virtnet_info *vi, struct send_queue *sq)
> > > +{
> > > +       int err;
> > > +
> > > +       err = virtio_enable_resetq(sq->vq);
> > > +       if (!err)
> > > +               netif_start_subqueue(vi->dev, sq - vi->sq);
> > > +
> > > +       virtnet_napi_tx_enable(vi, sq->vq, &sq->napi);
> > > +
> > > +       return err;
> > > +}
> > > +
> > > +static int virtnet_rx_vq_enable(struct virtnet_info *vi,
> > > +                               struct receive_queue *rq)
> > > +{
> > > +       int err;
> >
> > So the API should be design in a consistent way.
> >
> > In rx_vq_disable() we do:
> >
> > reset()
> > detach_unused_bufs()
> > vring_release_virtqueue()
> >
> > here it's better to exactly the reverse
> >
> > vring_attach_virtqueue() // this is the helper I guess in patch 5,
> > reverse of the vring_release_virtqueue()
> > try_refill_recv() // reverse of the detach_unused_bufs()
> > enable_reset() // reverse of the reset
> 
> Such an api is ok
> 
> 1. reset()
> 2. detach_unused_bufs()
> 3. vring_release_virtqueue()
>    ---------------
> 4. vring_attach_virtqueue()
> 5. try_refill_recv()
> 6. enable_reset()
> 
> 
> But if, we just want to recycle the buffer without modifying the ring num. As
> you mentioned before, in the case where the ring num is not modified, we don't
> have to reallocate, but can use the original vring.
> 
> 1. reset()
> 2. detach_unused_bufs()
>    ---------------
> 3. vring_reset_virtqueue() // just reset, no reallocate
> 4. try_refill_recv()
> 5. enable_reset()
> 
> Thanks.

Further, can we queue the buffers instead of detach_unused_bufs
and just requeue them instead of try_refill_recv?

> >
> > So did for the tx (no need for refill in that case).
> >
> > > +
> > > +       err = virtio_enable_resetq(rq->vq);
> > > +
> > > +       virtnet_napi_enable(rq->vq, &rq->napi);
> > > +
> > > +       return err;
> > > +}
> > > +
> > > +static int virtnet_rx_vq_reset(struct virtnet_info *vi, int i)
> > > +{
> > > +       int err;
> > > +
> > > +       err = virtnet_rx_vq_disable(vi, vi->rq + i);
> > > +       if (err)
> > > +               return err;
> > > +
> > > +       err = virtnet_rx_vq_enable(vi, vi->rq + i);
> > > +       if (err)
> > > +               netdev_err(vi->dev,
> > > +                          "enable rx reset vq fail: rx queue index: %d err: %d\n", i, err);
> > > +       return err;
> > > +}
> > > +
> > > +static int virtnet_tx_vq_reset(struct virtnet_info *vi, int i)
> > > +{
> > > +       int err;
> > > +
> > > +       err = virtnet_tx_vq_disable(vi, vi->sq + i);
> > > +       if (err)
> > > +               return err;
> > > +
> > > +       err = virtnet_tx_vq_enable(vi, vi->sq + i);
> > > +       if (err)
> > > +               netdev_err(vi->dev,
> > > +                          "enable tx reset vq fail: tx queue index: %d err: %d\n", i, err);
> > > +       return err;
> > > +}
> > > +
> > >  /*
> > >   * Send command via the control virtqueue and check status.  Commands
> > >   * supported by the hypervisor, as indicated by feature bits, should
> > > --
> > > 2.31.0
> > >
> >