Re: [PATCH v3 2/2] vdpa_sim_net: Add the support of set mac address

[Jason Wang <jasowang@xxxxxxxxxx>]

On Mon, Jul 8, 2024 at 3:19 PM Cindy Lu <lulu@xxxxxxxxxx> wrote:
>
> On Mon, 8 Jul 2024 at 15:06, Jason Wang <jasowang@xxxxxxxxxx> wrote:
> >
> > On Mon, Jul 8, 2024 at 2:48 PM Cindy Lu <lulu@xxxxxxxxxx> wrote:
> > >
> > > Add the function to support setting the MAC address.
> > > For vdpa_sim_net, the driver will write the MAC address
> > > to the config space, and other devices can implement
> > > their own functions to support this.
> > >
> > > Signed-off-by: Cindy Lu <lulu@xxxxxxxxxx>
> > > ---
> > >  drivers/vdpa/vdpa_sim/vdpa_sim_net.c | 19 ++++++++++++++++++-
> > >  1 file changed, 18 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/vdpa/vdpa_sim/vdpa_sim_net.c b/drivers/vdpa/vdpa_sim/vdpa_sim_net.c
> > > index cfe962911804..a472c3c43bfd 100644
> > > --- a/drivers/vdpa/vdpa_sim/vdpa_sim_net.c
> > > +++ b/drivers/vdpa/vdpa_sim/vdpa_sim_net.c
> > > @@ -414,6 +414,22 @@ static void vdpasim_net_get_config(struct vdpasim *vdpasim, void *config)
> > >         net_config->status = cpu_to_vdpasim16(vdpasim, VIRTIO_NET_S_LINK_UP);
> > >  }
> > >
> > > +static int vdpasim_net_set_attr(struct vdpa_mgmt_dev *mdev,
> > > +                               struct vdpa_device *dev,
> > > +                               const struct vdpa_dev_set_config *config)
> > > +{
> > > +       struct vdpasim *vdpasim = container_of(dev, struct vdpasim, vdpa);
> > > +
> > > +       struct virtio_net_config *vio_config = vdpasim->config;
> > > +       if (config->mask & (1 << VDPA_ATTR_DEV_NET_CFG_MACADDR)) {
> > > +               if (!is_zero_ether_addr(config->net.mac)) {
> > > +                       memcpy(vio_config->mac, config->net.mac, ETH_ALEN);
> > > +                       return 0;
> > > +               }
> > > +       }
> > > +       return -EINVAL;
> >
> > I think in the previous version, we agreed to have a lock to
> > synchronize the writing here?
> >
> > Thanks
> >
> Hi Jason
> I have moved the down_write(&vdev->cf_lock) and
> up_write(&vdev->cf_lock) to the function vdpa_dev_net_device_attr_set
> in vdpa/vdpa.c. Then the device itself doesn't need to call it again.
> Do you think this is ok?

I meant we have another path to modify the mac:

static virtio_net_ctrl_ack vdpasim_handle_ctrl_mac(struct vdpasim *vdpasim,
                                                   u8 cmd)
{
        struct virtio_net_config *vio_config = vdpasim->config;
        struct vdpasim_virtqueue *cvq = &vdpasim->vqs[2];
        virtio_net_ctrl_ack status = VIRTIO_NET_ERR;
        size_t read;

        switch (cmd) {
case VIRTIO_NET_CTRL_MAC_ADDR_SET:
                read = vringh_iov_pull_iotlb(&cvq->vring, &cvq->in_iov,
                                             vio_config->mac, ETH_ALEN);
                if (read == ETH_ALEN)
            status = VIRTIO_NET_OK;
        break;
        default:
                break;
        }

        return status;
}

We need to serialize between them.

Thanks

> Thanks
> Cindy
> > > +}
> > > +
> > >  static void vdpasim_net_setup_config(struct vdpasim *vdpasim,
> > >                                      const struct vdpa_dev_set_config *config)
> > >  {
> > > @@ -510,7 +526,8 @@ static void vdpasim_net_dev_del(struct vdpa_mgmt_dev *mdev,
> > >
> > >  static const struct vdpa_mgmtdev_ops vdpasim_net_mgmtdev_ops = {
> > >         .dev_add = vdpasim_net_dev_add,
> > > -       .dev_del = vdpasim_net_dev_del
> > > +       .dev_del = vdpasim_net_dev_del,
> > > +       .dev_set_attr = vdpasim_net_set_attr
> > >  };
> > >
> > >  static struct virtio_device_id id_table[] = {
> > > --
> > > 2.45.0
> > >
> >
>