On Mon, Dec 5, 2022 at 11:02 AM Vladimir Oltean <vladimir.oltean@xxxxxxx> wrote: > > On Mon, Dec 05, 2022 at 10:01:35AM -0800, Tim Harvey wrote: > > Vladimir, > > > > Yes, this patch resolves the issue. Enabling CONFIG_MARVELL_PHY > > without this patch still shows the issue. > > Yeah, that makes sense considering what the issue is. > > > Thanks - please cc me on that and I'll respond (unless you want me to send it). > > > > I'll submit a v2 of my dt patch with fec phy-mode = 'rgmii' after this > > patch lands. > > The question is how far to backport this change. > > I see that the GW5904 board was introduced in 2017. Which kernels do you > need to support the complete/unambiguous dt-binding? before adding the internal PHY nodes it works fine so no backporting needed. I just don't want to add the complete/unambiguous dt-bindings until your patch lands as at that point it would otherwise be broken. Best Regards, Tim > > I'm tempted to submit the patch to the "net" tree (making it a candidate > for stable releases) but without a Fixes: tag, making it go as far as it > will. That would be down to kernel 5.18, where the mv88e6xxx_get_caps() > function was introduced. Further down, and we'd need different patches > for older stable trees.
