Re: [EXT] Re: [PATCH] net: fec: Keep device numbering consistent with datasheet

[Stefan Riedmüller]

Hi Andy, David and Andrew,

first of all, thanks for your review. I really appreciate it!

On 24.09.20 08:36, Andy Duan wrote:
> From: David Miller <davem@xxxxxxxxxxxxx> Sent: Thursday, September 24, 2020 4:32 AM
> > From: Stefan Riedmueller <s.riedmueller@xxxxxxxxx>
> > Date: Wed, 23 Sep 2020 16:25:28 +0200
> >
> > > From: Christian Hemp <c.hemp@xxxxxxxxx>
> > >
> > > Make use of device tree alias for device enumeration to keep the
> > > device order consistent with the naming in the datasheet.
> > >
> > > Otherwise for the i.MX 6UL/ULL the ENET1 interface is enumerated as
> > > eth1 and ENET2 as eth0.
> > >
> > > Signed-off-by: Christian Hemp <c.hemp@xxxxxxxxx>
> > > Signed-off-by: Stefan Riedmueller <s.riedmueller@xxxxxxxxx>
> >
> > Device naming and ordering for networking devices was never, ever,
> > guaranteed.
> >
> > Use udev or similar.
> >
> > > @@ -3691,6 +3692,10 @@ fec_probe(struct platform_device *pdev)
> > >
> > >        ndev->max_mtu = PKT_MAXBUF_SIZE - ETH_HLEN - ETH_FCS_LEN;
> > >
> > > +     eth_id = of_alias_get_id(pdev->dev.of_node, "ethernet");
> > > +     if (eth_id >= 0)
> > > +             sprintf(ndev->name, "eth%d", eth_id);
> >
> > You can't ever just write into ndev->name, what if another networking device is
> > already using that name?
> >
> > This change is incorrect on many levels.
>
> David is correct.
>
> For example, imx8DXL has ethernet0 is EQOS TSN, ethernet1 is FEC.
> EQOS TSN is andother driver and is registered early, the dev->name is eth0.
> So the patch will bring conflict in such case.

I was not aware of that conflict, but now that you mention it it makes total 
sense.

I wanted to make life a little easier for myself but underestimated the 
global context. I will try to find a solution with udev or something similar.

So please drop this patch and sorry for the noise.

Stefan

> Andy