On Mon, 2014-11-24 at 11:53 -0800, David Daney wrote: > On 11/24/2014 11:46 AM, Aaro Koskinen wrote: > > Hi, > > > > On Mon, Nov 24, 2014 at 09:13:01PM +0200, Aaro Koskinen wrote: > >> On Mon, Nov 24, 2014 at 09:25:26AM -0800, David Daney wrote: > >>> On 11/24/2014 05:51 AM, James Cowgill wrote: > >>>> From: Markos Chandras <markos.chandras@xxxxxxxxxx> > >>>> > >>>> Add support for the UBNT E200 board (EdgeRouter/EdgeRouter Pro 8 port). > >>>> > >>>> Signed-off-by: Markos Chandras <markos.chandras@xxxxxxxxxx> > >>>> Signed-off-by: James Cowgill <James.Cowgill@xxxxxxxxxx> > >>> > >>> NACK. > >>> > >>> As far as I know, these boards have a boot loader that supplies a correct > >>> device tree, there should be no need to hack up the kernel like this. > >>> > >>> As far as I know, Andreas is running a kernel.org kernel on these boards > >>> without anything like this. > >> > >> It gets called from Octeon Ethernet driver through cvmx_helper_link_get() > >> frequently so the console gets spammed about unknown board, and probably > >> also the link status is bogus as a result. > > > > Just tested with 3.18-rc6 and this behaviour has been apparently > > fixed somehow. Cool. > > > > It is magic! > > If the kernel is built with the proper ethernet phy drivers *and* the > device tree contains proper phy topology information, then the phy > drivers handle link monitoring and none of the cvmx-* crap is used. > > So for all boards with bootloaders that supply a device tree, there > should never be any reason to patch in the hacky board identifiers to > the kernel sources. Yep you're right - everything does seem to work. I think I must have messed up different kernel versions or something when I tested this before. Sorry for the spam, you can ignore this patch. Thanks, James
