On Thu, Sep 28, 2017 at 04:09:36PM +0200, Mark Kettenis wrote: > Jonathan Gray schreef op 2017-09-28 09:51: > > On Thu, Sep 28, 2017 at 09:34:49AM +0200, Mark Kettenis wrote: > > > Jonathan Gray schreef op 2017-09-28 05:02: > > > > On Wed, Sep 27, 2017 at 10:56:44PM +0200, Patrick Wildt wrote: > > > > > So, first of all copying the dtb entries is not a good idea. The > > > > > reason > > > > > is that the phandles are gonna be all wrong and overriden, because > > > > > those > > > > > are _generated_ on compile time. As you can see, the ethernet > > > > > controller references phy handle 0x7, but the phy has a phandle of > > > > > 0x47. > > > > > Something is wrong there. > > > > > > > > > > You should try to apply this "revert" that was committed in linux. > > > > > Hope > > > > > they'll soon make up their minds and commit a newer version. > > > > > > > > The device tree that is included with > > > > /usr/local/share/u-boot/orangepi_one/u-boot-sunxi-with-spl.bin > > > > should still have emac nodes. > > > > > > But they use a different binding from the Linux kernel, which are > > > the ones > > > Patrick > > > used for the implementation. And some of the regulator stuff is > > > missing > > > so things don't actually work even if we would change the driver to > > > use the > > > bindings currently used by U-Boot. It's a mess. > > > > > > So using the dtb from Linux with the revert applied is the way to go > > > for > > > now. > > > > > > > So we patch the dtb port to revert the revert commits for now? > > Doesn't appear to have been added back in 3.14 rcs yet. > > We could do that.
The port now builds 4.15-rc1 which has EMAC for A64/H5/H3 added back. The mdio child nodes change but it looks like it should already be compatible with what is in the tree as that is not used.
