On 07/30/2013 05:22:11 AM, Chunhe Lan wrote:

=20
On 07/30/2013 09:09 AM, Scott Wood wrote:

quoted

On Fri, Jun 14, 2013 at 04:26:20PM +0800, Chunhe Lan wrote:

quoted

Ethernet:
    eTSEC1: Connected to Atheros AR8035 GETH PHY
    eTSEC2: Connected to Atheros AR8035 GETH PHY

Where are the PHYs in the device tree?

    "Atheros AR8035 GETH PHY" driver is module_init driver. It uses =20
the two structs of "static struct phy_driver at8035_driver" and
    "static struct mdio_device_id __maybe_unused atheros_tbl" to =20
register at8035_driver.
=20
    So do not need to add PHYs in the device tree.

Huh?  How does registering a driver eliminate the need to describe the =20
devices in the device tree?  If you're trying to say that the device =20
can be probed (like a PCI device), how do you determine which PHY goes =20
to which MAC?  I suspect the actual answer is "this chip has datapath =20
ethernet, and datapath stuff is not upstream (still!)".  That's no =20
excuse for not describing it in the device tree, though.  The device =20
tree describes the hardware, not what Linux has drivers for.

FWIW, I don't see the string "at8035_driver" anywhere in the kernel =20
(except in the SDK, which doesn't count here).  Maybe you meant =20
at803x_driver?

-Scott=