[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190528100211.GX18865@dhcp-12-139.nay.redhat.com>
Date: Tue, 28 May 2019 18:02:11 +0800
From: Hangbin Liu <liuhangbin@...il.com>
To: Jiri Pirko <jiri@...nulli.us>
Cc: netdev@...r.kernel.org, davem@...emloft.net
Subject: Re: [PATCH net-next] team: add ethtool get_link_ksettings
On Tue, May 28, 2019 at 11:08:23AM +0200, Jiri Pirko wrote:
> >+static int team_ethtool_get_link_ksettings(struct net_device *dev,
> >+ struct ethtool_link_ksettings *cmd)
> >+{
> >+ struct team *team= netdev_priv(dev);
> >+ unsigned long speed = 0;
> >+ struct team_port *port;
> >+
> >+ cmd->base.duplex = DUPLEX_UNKNOWN;
> >+ cmd->base.port = PORT_OTHER;
> >+
> >+ list_for_each_entry(port, &team->port_list, list) {
> >+ if (team_port_txable(port)) {
> >+ if (port->state.speed != SPEED_UNKNOWN)
> >+ speed += port->state.speed;
> >+ if (cmd->base.duplex == DUPLEX_UNKNOWN &&
> >+ port->state.duplex != DUPLEX_UNKNOWN)
> >+ cmd->base.duplex = port->state.duplex;
>
> What is exactly the point of this patch? Why do you need such
> information. This is hw-related info. If you simply sum-up all txable
> ports, the value is always highly misleading.
>
> For example for hash-based port selection with 2 100Mbit ports,
> you will get 200Mbit, but it is not true. It is up to the traffic and
> hash function what is the actual TX speed you can get.
> On the RX side, this is even more misleading as the actual speed depends
> on the other side of the wire.
The number is the maximum speed in theory. I added it because someone
said bond interface could show total speed while team could not...
The usage is customer could get team link-speed and throughput via SNMP.
Thanks
Hangbin
>
>
> >+ }
> >+ }
> >+ cmd->base.speed = speed ? : SPEED_UNKNOWN;
> >+
> >+ return 0;
> >+}
> >+
> > static const struct ethtool_ops team_ethtool_ops = {
> > .get_drvinfo = team_ethtool_get_drvinfo,
> > .get_link = ethtool_op_get_link,
> >+ .get_link_ksettings = team_ethtool_get_link_ksettings,
> > };
> >
> > /***********************
> >--
> >2.19.2
> >
Powered by blists - more mailing lists