Le 28/03/2015 00:02, Boris Brezillon a écrit :

Hi Nicolas,

On Fri, 27 Mar 2015 16:34:11 +0100
Nicolas Ferre [off-list ref] wrote:

quoted

Capabilities configuration by macb_configure_caps() was moved far too late by
421d9df0628b (net/macb: merge at91_ether driver into macb driver) which would
lead to badly configured hardware.

Indeed, the macb_configure_caps function is called a bit too late,
but ...

quoted

So, move this function to early probe and modify its prototype to re-gain its
original behavior.
DT data retrieval is also moved to simplify the probe code flow.

... I'm not happy with these changes.
I tried to keep  specific init steps of macb and at91_ether separated
and you're moving macb_configure_caps call (not required on at91_ether
HW) into macb_probe (the common probe part).

Well, this function is about configuring the capabilities of the
hardware both from the configuration registers and the device tree
entries (this last source applies to all flavors of hardware).

I only see advantages to set these flags early (Cf. above).

How about moving macb_configure_caps a bit earlier in the macb_init
function [1] ?

No, it won't be sufficient. The very first function needing the
capabilities set is macb_get_hwaddr() which is pretty early in macb_probe().


Bye,

Best Regards,

Boris

[1]http://code.bulix.org/8gyi6b-88141

-- 
Nicolas Ferre