[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20171106220230.2081aceb@xps13>
Date: Mon, 6 Nov 2017 22:02:30 +0100
From: Miquel RAYNAL <miquel.raynal@...e-electrons.com>
To: Andrew Lunn <andrew@...n.ch>
Cc: "David S . Miller" <davem@...emloft.net>,
Thomas Petazzoni <thomas.petazzoni@...e-electrons.com>,
Antoine Tenart <antoine.tenart@...e-electrons.com>,
Gregory Clement <gregory.clement@...e-electrons.com>,
Nadav Haklai <nadavh@...vell.com>, netdev@...r.kernel.org,
Stefan Chulski <stefanc@...vell.com>,
Florian Fainelli <f.fainelli@...il.com>
Subject: Re: [PATCH net-next v2] net: mvpp2: add ethtool GOP statistics
Hi Andrew,
> > Here I do have a problem: I choose the IDA solution because it was
> > quite straightforward but I agree it would be better to use an
> > unique name. Unfortunately, on the Armada-8040-DB that instantiate
> > this driver twice, the node name is not unique. There are two CP
> > (master and slave) and both names are "ethernet@0". Otherwise, if I
> > use the full path, I get something like
> > "/cp110-master/config-space@...00000/ethernet@0" and
> > "/cp110-slave/config-space@...00000/ethernet@0" but the problem is
> > that workqueue names are truncated to 24 characters and only 15
> > appears in ps, so it would not solve the issue and choosing the
> > "parent parent node name" would work here but does not scale very
> > well. Do you have any idea to get this right?
>
> Hi Miquel
>
> You could move the starting of the thread into mvpp2_port_probe(). If
> you do it after register_netdev(dev), you can use netdev_name(dev).
>
> Andrew
There is one workqueue per instance of the driver, not per port.
Hence, I choose to use the netdev_name() (short) with a '+' after it
if there are other ports involved, ie. "stats-wq-eth0+" and
"stats-wq-eth2+".
Thanks for your help,
Miquèl
Powered by blists - more mailing lists