[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210311231730.23wcckyzihmp6elk@skbuf>
Date: Fri, 12 Mar 2021 01:17:30 +0200
From: Vladimir Oltean <olteanv@...il.com>
To: Andrew Lunn <andrew@...n.ch>
Cc: Kurt Kanzenbach <kurt@...-computers.de>,
Vivien Didelot <vivien.didelot@...il.com>,
Florian Fainelli <f.fainelli@...il.com>,
"David S. Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>, netdev@...r.kernel.org
Subject: Re: [PATCH net-next 2/6] net: dsa: hellcreek: Report META data usage
On Thu, Mar 11, 2021 at 11:52:41PM +0100, Andrew Lunn wrote:
> On Thu, Mar 11, 2021 at 06:53:40PM +0100, Kurt Kanzenbach wrote:
> > Report the META data descriptor usage via devlink.
>
> Jakubs question is also relevant here. Please could you give a bit
> more background about what the meta data is?
Not having seen any documentation for this device, my guess is that
metadata descriptors are frame references, and the RAM page count is for
packet memory buffers. Nonetheless, I would still like to hear it from
Kurt. There is still a lot unknown even if I am correct. For example, if
the frame references or buffers can be partitioned, or if watermarks for
things like congestion/flow control can be set, then maybe devlink-sb is
a better choice (as that has an occupancy facility as well)?
Fully understand that it is not as trivial as exposing a devlink
resource, but on Ocelot/Felix I quite appreciate having the feature
(drivers/net/ethernet/mscc/ocelot_devlink.c). And knowing that this is a
TSN switch, I expect that sooner or later, the need to have control over
resource partitioning per traffic class will arise anyway.
Powered by blists - more mailing lists