Re: [PATCH wpan-next 17/20] net: ieee802154: Handle limited devices with only datagram support

[Alexander Aring <aahringo@redhat.com>]

Hi,

On Fri, Jul 1, 2022 at 10:37 AM Miquel Raynal <miquel.raynal@bootlin.com> wrote:
>
> Some devices, like HardMAC ones can be a bit limited in the way they
> handle mac commands. In particular, they might just not support it at
> all and instead only be able to transmit and receive regular data
> packets. In this case, they cannot be used for any of the internal
> management commands that we have introduced so far and must be flagged
> accordingly.
>
> Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
> ---
>  include/net/cfg802154.h   | 3 +++
>  net/ieee802154/nl802154.c | 6 ++++++
>  2 files changed, 9 insertions(+)
>
> diff --git a/include/net/cfg802154.h b/include/net/cfg802154.h
> index d6ff60d900a9..20ac4df9dc7b 100644
> --- a/include/net/cfg802154.h
> +++ b/include/net/cfg802154.h
> @@ -178,12 +178,15 @@ wpan_phy_cca_cmp(const struct wpan_phy_cca *a, const struct wpan_phy_cca *b)
>   *     setting.
>   * @WPAN_PHY_FLAG_STATE_QUEUE_STOPPED: Indicates that the transmit queue was
>   *     temporarily stopped.
> + * @WPAN_PHY_FLAG_DATAGRAMS_ONLY: Indicates that transceiver is only able to
> + *     send/receive datagrams.
>   */
>  enum wpan_phy_flags {
>         WPAN_PHY_FLAG_TXPOWER           = BIT(1),
>         WPAN_PHY_FLAG_CCA_ED_LEVEL      = BIT(2),
>         WPAN_PHY_FLAG_CCA_MODE          = BIT(3),
>         WPAN_PHY_FLAG_STATE_QUEUE_STOPPED = BIT(4),
> +       WPAN_PHY_FLAG_DATAGRAMS_ONLY    = BIT(5),
>  };
>
>  struct wpan_phy {
> diff --git a/net/ieee802154/nl802154.c b/net/ieee802154/nl802154.c
> index 00b03c33e826..b31a0bd36b08 100644
> --- a/net/ieee802154/nl802154.c
> +++ b/net/ieee802154/nl802154.c
> @@ -1404,6 +1404,9 @@ static int nl802154_trigger_scan(struct sk_buff *skb, struct genl_info *info)
>         if (wpan_dev->iftype == NL802154_IFTYPE_MONITOR)
>                 return -EPERM;
>
> +       if (wpan_phy->flags & WPAN_PHY_FLAG_DATAGRAMS_ONLY)
> +               return -EOPNOTSUPP;
> +

for doing a scan it's also required to turn the transceiver into
promiscuous mode, right?

There is currently a flag if a driver supports promiscuous mode or
not... I am not sure if all drivers have support for it. For me it
looks like a mandatory feature but I am not sure if every driver
supports it.
We have a similar situation with acknowledge retransmit handling...
some transceivers can't handle it and for normal dataframes we have a
default behaviour that we don't set it. However sometimes it's
required by the spec, then we can't do anything here.

I think we should check on it but we should plan to drop this flag if
promiscuous mode is supported or not. I also think that the
promiscuous driver_ops should be removed and moved as a parameter for
start() driver_ops to declare which "receive mode" should be
enabled... but we can do that in due course.

- Alex