On Samstag, 25. Februar 2012, Geert Uytterhoeven wrote:

2012/2/24 Michał Mirosław [off-list ref]:

quoted

2012/2/24 Danny Kukawka [off-list ref]:

quoted

Second Part of series patches to unifiy the return value of
.ndo_set_mac_address if the given address isn't valid.

These changes check if a given (MAC) address is valid in
.ndo_set_mac_address, if invalid return -EADDRNOTAVAIL
as eth_mac_addr() already does if is_valid_ether_addr() fails.

Why not just fix dev_set_mac_address() and make do_setlink() use that?

BTW, it's also called from dev_set_mac_address().

quoted

Checks are specific to address family, not device model I assume.

Indeed, why can't this be done in one single place, instead of sprinkling
these checks over all drivers, missing all out-of-tree (note: I don't care)
and all soon-to-be-submitted drivers?

Since the .ndo_set_mac_address functions are used by some drivers internally 
too, you may get some new checks on other places. But I'll take a look at it.

Danny