On 08/08/2022 21.07, Tom Rini wrote:
On Thu, May 12, 2022 at 09:33:09AM +0200, Rasmus Villemoes wrote:
Similar to what was done for the FEC driver in commit 89b5bd54c1a4 (net: fec: Allow the PHY node to be retrieved), make sure the PHY is associated with the right device tree node, so that phy specific DT properties is accessible by the phy driver.
This is required on a custom iMX8MP board with a ti,dp83867 phy sitting in front of the eqos interface.
Signed-off-by: Rasmus Villemoes rasmus.villemoes@prevas.dk Reviewed-by: Ramon Fried rfried.dev@gmail.com
I'm deferring 2/3 and now for 3/3, given a6acf95508e2 ("net: eqos: add function to get phy node and address") (which yes, sigh, came after your series but has been applied already, sorry) I don't see the obvious way to migrate the changes in without testing. If needed still, can you please rebase and resend? Thanks.
Well, I have/had other changes on top that I didn't want to send before this series went in, and those were a bit annoying to rebase. But a6acf95508e2 does seem to achieve what 2/3+3/3 did, so those can be dropped. Thanks for applying the other patches.
<moderate grumpiness>
However, we now have one more copy of more or less the exact same code, so there's certainly an opportunity for somebody to do something like 2/3 and make both fec_mxc.c, dwc_eth_qos.c and probably others use such a helper.
And on the subject of duplicate functionality with perhaps ever so slightly different semantics, it's impressive that we have at least
reset-delay-us/reset-post-delay-us reset-assert-us/reset-deassert-us phy-reset-duration/phy-reset-post-delay
making it hard to figure out what setting is honored in which contexts.
Oh well.
</moderate grumpiness>
Rasmus