[v2,1/3] net: add public qemu_net_poll() function

From: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>

The NetClientInfo .poll() callback is being called directly by
hw/vhost_net.c.  Create a public net.c function so callers do not
depend on internals.

This change is useful because later patches change net internals.  Those
changes shouldn't affect .poll() callers.

Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
---
 hw/vhost_net.c | 6 +++---
 net.c          | 7 +++++++
 net.h          | 2 ++
 3 files changed, 12 insertions(+), 3 deletions(-)

Stefan Hajnoczi <stefanha@redhat.com> writes:

> From: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
>
> The NetClientInfo .poll() callback is being called directly by
> hw/vhost_net.c.  Create a public net.c function so callers do not
> depend on internals.
>
> This change is useful because later patches change net internals.  Those
> changes shouldn't affect .poll() callers.
>
> Signed-off-by: Stefan Hajnoczi <stefanha@linux.vnet.ibm.com>
> ---
>  hw/vhost_net.c | 6 +++---
>  net.c          | 7 +++++++
>  net.h          | 2 ++
>  3 files changed, 12 insertions(+), 3 deletions(-)
>
> diff --git a/hw/vhost_net.c b/hw/vhost_net.c
> index df2c4a3..6737600 100644
> --- a/hw/vhost_net.c
> +++ b/hw/vhost_net.c
> @@ -160,7 +160,7 @@ int vhost_net_start(struct vhost_net *net,
>          goto fail_start;
>      }
>  
> -    net->nc->info->poll(net->nc, false);
> +    qemu_net_poll(net->nc, false);

Unconditional call of info->poll() is safe, because we know the backend
is tap (see vhost_net_get_fd()), and tap implements poll.

>      qemu_set_fd_handler(net->backend, NULL, NULL, NULL);
>      file.fd = net->backend;
>      for (file.index = 0; file.index < net->dev.nvqs; ++file.index) {
> @@ -177,7 +177,7 @@ fail:
>          int r = ioctl(net->dev.control, VHOST_NET_SET_BACKEND, &file);
>          assert(r >= 0);
>      }
> -    net->nc->info->poll(net->nc, true);
> +    qemu_net_poll(net->nc, true);
>      vhost_dev_stop(&net->dev, dev);
>      if (net->dev.acked_features & (1 << VIRTIO_NET_F_MRG_RXBUF)) {
>          tap_set_vnet_hdr_len(net->nc, sizeof(struct virtio_net_hdr));
> @@ -197,7 +197,7 @@ void vhost_net_stop(struct vhost_net *net,
>          int r = ioctl(net->dev.control, VHOST_NET_SET_BACKEND, &file);
>          assert(r >= 0);
>      }
> -    net->nc->info->poll(net->nc, true);
> +    qemu_net_poll(net->nc, true);
>      vhost_dev_stop(&net->dev, dev);
>      if (net->dev.acked_features & (1 << VIRTIO_NET_F_MRG_RXBUF)) {
>          tap_set_vnet_hdr_len(net->nc, sizeof(struct virtio_net_hdr));
> diff --git a/net.c b/net.c
> index ae4bc0d..eb4f4bf 100644
> --- a/net.c
> +++ b/net.c
> @@ -466,6 +466,13 @@ qemu_sendv_packet(NetClientState *nc, const struct iovec *iov, int iovcnt)
>      return qemu_sendv_packet_async(nc, iov, iovcnt, NULL);
>  }
>  
> +void qemu_net_poll(NetClientState *nc, bool enable)
> +{
> +    if (nc->info->poll) {
> +        nc->info->poll(nc, enable);
> +    }
> +}
> +

Here, we don't have such context, so we check, defensively.

poll()'s semantics aren't documented, so I can't say whether "do
nothing" is appropriate for backends that don't implement it.

>  NetClientState *qemu_find_netdev(const char *id)
>  {
>      NetClientState *nc;
> diff --git a/net.h b/net.h
> index 04fda1d..763f926 100644
> --- a/net.h
> +++ b/net.h
> @@ -111,6 +111,8 @@ ssize_t qemu_deliver_packet_iov(NetClientState *sender,
>                              int iovcnt,
>                              void *opaque);
>  
> +void qemu_net_poll(NetClientState *nc, bool enable);
> +
>  void print_net_client(Monitor *mon, NetClientState *nc);
>  void do_info_network(Monitor *mon);