Re: ethernet<n> dt aliases implications in U-Boot and Linux

[Rob Herring <robh@kernel.org>]

On Tue, Aug 9, 2022 at 4:39 PM Tim Harvey <tharvey@gateworks.com> wrote:
>
> On Tue, Aug 9, 2022 at 2:31 PM Pali Rohár <pali@kernel.org> wrote:
> >
> > On Tuesday 09 August 2022 16:48:23 Sean Anderson wrote:
> > > On 8/8/22 5:45 PM, Michal Suchánek wrote:
> > > > On Mon, Aug 08, 2022 at 02:38:35PM -0700, Stephen Hemminger wrote:
> > > >> On Mon, 8 Aug 2022 23:09:45 +0200
> > > >> Michal Suchánek <msuchanek@suse.de> wrote:
> > > >>
> > > >> > On Mon, Aug 08, 2022 at 03:57:55PM -0400, Sean Anderson wrote:
> > > >> > > Hi Tim,
> > > >> > >
> > > >> > > On 8/8/22 3:18 PM, Tim Harvey wrote:
> > > >> > > > Greetings,
> > > >> > > >
> > > >> > > > I'm trying to understand if there is any implication of 'ethernet<n>'
> > > >> > > > aliases in Linux such as:
> > > >> > > >         aliases {
> > > >> > > >                 ethernet0 = &eqos;
> > > >> > > >                 ethernet1 = &fec;
> > > >> > > >                 ethernet2 = &lan1;
> > > >> > > >                 ethernet3 = &lan2;
> > > >> > > >                 ethernet4 = &lan3;
> > > >> > > >                 ethernet5 = &lan4;
> > > >> > > >                 ethernet6 = &lan5;
> > > >> > > >         };
> > > >> > > >
> > > >> > > > I know U-Boot boards that use device-tree will use these aliases to
> > > >> > > > name the devices in U-Boot such that the device with alias 'ethernet0'
> > > >> > > > becomes eth0 and alias 'ethernet1' becomes eth1 but for Linux it
> > > >> > > > appears that the naming of network devices that are embedded (ie SoC)
> > > >> > > > vs enumerated (ie pci/usb) are always based on device registration
> > > >> > > > order which for static drivers depends on Makefile linking order and
> > > >> > > > has nothing to do with device-tree.
> > > >> > > >
> > > >> > > > Is there currently any way to control network device naming in Linux
> > > >> > > > other than udev?

Ah, the topic that will never die.

> > > >> > >
> > > >> > > You can also use systemd-networkd et al. (but that is the same kind of mechanism)
> > > >> > >
> > > >> > > > Does Linux use the ethernet<n> aliases for anything at all?
> > > >> > >
> > > >> > > No :l
> > > >> >
> > > >> > Maybe it's a great opportunity for porting biosdevname to DT based
> > > >> > platforms ;-)
> > > >>
> > > >> Sorry, biosdevname was wrong way to do things.
> > > >> Did you look at the internals, it was dumpster diving as root into BIOS.
> > > >
> > > > When it's BIOS what defines the names then you have to read them from
> > > > the BIOS. Recently it was updated to use some sysfs file or whatver.
> > > > It's not like you would use any of that code with DT, anyway.
> > > >
> > > >> Systemd-networkd does things in much more supportable manner using existing
> > > >> sysfs API's.
> > > >
> > > > Which is a dumpster of systemd code, no thanks.
> > > >
> > > > I want my device naming independent of the init system, especially if
> > > > it's systemd.
> > >
> > > Well, there's always nameif...
> > >
> > > That said, I have made [1] for people using systemd-networkd.
> > >
> > > --Sean
> > >
> > > [1] https://github.com/systemd/systemd/pull/24265
> >
> > Hello!
> >
> > In some cases "label" DT property can be used also as interface name.
> > For example this property is already used by DSA kernel driver.
> >
> > I created very simple script which renames all interfaces in system to
> > their "label" DT property (if there is any defined).
> >
> > #!/bin/sh
> > for iface in `ls /sys/class/net/`; do
> >         for of_node in of_node device/of_node; do
> >                 if test -e /sys/class/net/$iface/$of_node/; then
> >                         label=`cat /sys/class/net/$iface/$of_node/label 2>/dev/null`
> >                         if test -n "$label" && test "$label" != "$iface"; then
> >                                 echo "Renaming net interface $iface to $label..."
> >                                 up=$((`cat /sys/class/net/$iface/flags 2>/dev/null || echo 1` & 0x1))
> >                                 if test "$up" != "0"; then
> >                                         ip link set dev $iface down
> >                                 fi
> >                                 ip link set dev $iface name "$label" && iface=$label
> >                                 if test "$up" != "0"; then
> >                                         ip link set dev $iface up
> >                                 fi
> >                         fi
> >                         break
> >                 fi
> >         done
> > done
> >
> > Maybe it would be better first to use "label" and then use ethernet alias?
>
> I've been wondering the same as well which made me wonder what the
> history of the 'aliases' node is and why its not used in most cases in
> Linux. I know for the SOC's I work with we've always defined aliases
> for ethernet<n>, gpio<n>, serial<n>, spi<n>, i2c<n>, mmc<n> etc. Where
> did this practice come from and why are we putting that in Linux dts
> files it if it's not used by Linux?

It's not 'Linux dts files', but dts files that happen to be hosted in
the Linux kernel. Many are sync'd to u-boot and other places.

While ethernetN is not used by the kernel, the rest are I think.
Personally, I'd do away with gpio, i2c and spi at least, but that ship
has sailed. Then we have some non-standard ones that crept in from
downstream I think (i.MX is the worst).

'label' is supposed to reflect the sticker on a user accessible
port/connector on device. As long as nothing is looking for particular
values of label, it's okay by me to use.

Rob