Re: [PATCH net-next v2 0/4] net: mdio: Clause 45 support for mdio-bcm-unimac
From: Kordian Chmielewski
Date: Mon Nov 16 2026 - 17:38:22 EST
On Mon, 16 Nov 2026 13:34:45 -0800, Florian Fainelli wrote:
> This patch series adds support for Clause 45 MDIO operations within the
> mdio-bcm-unimac driver.
Tested v2 on the same TP-Link Archer AX6000 (BCM4908) with OpenWrt and
Linux 6.18.55, again in place of our own C45 support. Patch 1 needed the
context of one hunk adjusted to apply there, because OpenWrt carries a
hack patch that adds a function right after get_phy_c45_devs_in_pkg();
the added and removed lines are unchanged.
The BCM54991E at address 0x1d, which our out-of-tree PHY driver switches
to C45 in probe, still takes its firmware over C45 and links at 1000
Mbit/s, the link comes back after ethtool -r, and 8400 C45 reads through
SIOCGMIIREG gave no errors.
This bus is registered from DT, so the C45 scan does not run on it. For
patch 1, a test module registered a bus without DT whose reads go to the
UniMAC bus and whose writes fail, with read_c45 set and phy_mask leaving
only three empty addresses. On this UniMAC, a C22 or C45 read from an
empty address fails with -EIO. With v1, get_phy_device() for C45
returned -EIO there and mdiobus_register() failed with -EIO. With v2 it
returns -ENODEV, and the bus registers with no PHY found, as it does
with C22 alone.
The new check in unimac_mdio_reset() still never matches here, as no
node on this bus has the ethernet-phy-ieee802.3-c45 compatible; the
PHY's node has only its ID compatible.
Tested-by: Kordian Chmielewski <kudzo51@xxxxxxxxx>