Hi,
about a year ago we resorted to using port pin MDIO in this situation
for the FCCs because the fcc_enet driver and the gianfar driver live
in separate worlds, so to speak.
But in the meantime, Andy Fleming was very active regarding the PHY
abstraction layer. I still haven't seen patches that migrate fcc_enet
to using this layer, but the gianfar driver seems to use this code now.
So I'd say quick solution: use TSEC1 MDIO for TSECs and cpm port pins
for FCC, clean solution: change fcc_enet to use the PHY abstraction
layer.
--=20
Stefan Nickl
Kontron Modular Computers
-----Original Message-----
From: linuxppc-embedded-bounces@ozlabs.org=20
[mailto:linuxppc-embedded-bounces@ozlabs.org] On Behalf Of=20
Gerhard Jaeger
Sent: Tuesday, October 11, 2005 12:14 PM
To: linuxppc-embedded@ozlabs.org
Subject: Which PHY connection to use on MPC8541...
=20
Hi list,
=20
we're currently designing a custon MPC8541 based board having=20
three ethernet connections.
For some reasons, we'd like to use the FCC1 and 2 + TSEC1.=20
From what I see from the docs, it should be no problem to=20
connect i.e. a quad phy (for FCC1/2) and a gigabit phy (for=20
TSEC1) to the EC_MDIO of the TSEC as long as they have=20
different addresses.=20
I don't want to use the GPIO stuff for MDIO.
=20
The question is, what does this mean to the drivers? For the=20
gianfar driver this should be okay and from what I see, the=20
upcoming fs_enet driver supports also a phy connected either=20
to the GPIOs (bitbanging) or the TSEC MDIO. Is this right, or=20
am I missing something essential?=20
Would it be better to use the GPIO-MDIO for the FCC phy and=20
the TSEC-MDIO for the TSEC phy?
=20
TIA
Gerhard
=20
_______________________________________________
Linuxppc-embedded mailing list
Linuxppc-embedded@ozlabs.org
https://ozlabs.org/mailman/listinfo/linuxppc-embedded
=20
On Oct 11, 2005, at 05:57, Stefan Nickl wrote:
Hi,
about a year ago we resorted to using port pin MDIO in this situation
for the FCCs because the fcc_enet driver and the gianfar driver live
in separate worlds, so to speak.
But in the meantime, Andy Fleming was very active regarding the PHY
abstraction layer. I still haven't seen patches that migrate fcc_enet
to using this layer, but the gianfar driver seems to use this code
now.
The work to migrate the new fs_enet driver to use the TSEC MDIO is in
progress. I have a few other things I'm working on right now, and
I've been having some hardware issues, but I've already done most of
the work to get the FCC to use the TSEC's MDIO bus.
So I'd say quick solution: use TSEC1 MDIO for TSECs and cpm port pins
for FCC, clean solution: change fcc_enet to use the PHY abstraction
layer.
I'd say it depends on how long until your hardware is ready. You
might want to wire up the cpm IO pins, just in case. I have yet to
actually get the fs_enet driver to send packets on the 8560, so
there's no telling how long it will be before that driver works on 8541.
Andy
On Tuesday 11 October 2005 20:05, Andy Fleming wrote:
On Oct 11, 2005, at 05:57, Stefan Nickl wrote:
quoted
Hi,
about a year ago we resorted to using port pin MDIO in this situation
for the FCCs because the fcc_enet driver and the gianfar driver live
in separate worlds, so to speak.
But in the meantime, Andy Fleming was very active regarding the PHY
abstraction layer. I still haven't seen patches that migrate fcc_enet
to using this layer, but the gianfar driver seems to use this code
now.
The work to migrate the new fs_enet driver to use the TSEC MDIO is in
progress. I have a few other things I'm working on right now, and
I've been having some hardware issues, but I've already done most of
the work to get the FCC to use the TSEC's MDIO bus.
quoted
So I'd say quick solution: use TSEC1 MDIO for TSECs and cpm port pins
for FCC, clean solution: change fcc_enet to use the PHY abstraction
layer.
I'd say it depends on how long until your hardware is ready. You
might want to wire up the cpm IO pins, just in case. I have yet to
actually get the fs_enet driver to send packets on the 8560, so
there's no telling how long it will be before that driver works on 8541.
Andy
Thank you guys for the quick answers. We already thought of additionally
wireing up the cpm IO pins to have some kind of backup ;) We expect the
hardware to be available in february/march next year so I guess there's
pretty much time.
Gerhard