[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20190730121950.pyxtu5k4qphmshy4@lx-anielsen.microsemi.net>
Date: Tue, 30 Jul 2019 14:19:51 +0200
From: "Allan W. Nielsen" <allan.nielsen@...rochip.com>
To: Ido Schimmel <idosch@...sch.org>
CC: Nikolay Aleksandrov <nikolay@...ulusnetworks.com>,
Horatiu Vultur <horatiu.vultur@...rochip.com>,
<roopa@...ulusnetworks.com>, <davem@...emloft.net>,
<bridge@...ts.linux-foundation.org>, <netdev@...r.kernel.org>,
<linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] net: bridge: Allow bridge to joing multicast groups
The 07/30/2019 13:04, Ido Schimmel wrote:
> On Tue, Jul 30, 2019 at 10:30:28AM +0200, Allan W. Nielsen wrote:
> > The 07/30/2019 10:06, Ido Schimmel wrote:
> > > As a bonus, existing drivers could benefit from it, as MDB entries are already
> > > notified by MAC.
> > Not sure I follow. When FDB entries are added, it also generates notification
> > events.
>
> I meant the switchdev notification sent to drivers:
>
> /* SWITCHDEV_OBJ_ID_PORT_MDB */
> struct switchdev_obj_port_mdb {
> struct switchdev_obj obj;
> unsigned char addr[ETH_ALEN];
> u16 vid;
> };
>
> By extending MDB entries to also be keyed by MAC you basically get a lot
> of things for free without duplicating the same code for multicast FDBs.
Agree, this should be the same.
> AFAICS, then only change in the fast path is in br_mdb_get() where you
> need to use DMAC as key in case Ethertype is not IPv4/IPv6.
That would be nice.
--
/Allan
Powered by blists - more mailing lists