[PATCH v8 0/4] Add ethernet support for the Amlogic T7
From: Lucas Tanure
Date: Sat Nov 21 2026 - 16:31:10 EST
Patches 1 to 4 of v7, the bindings and the driver, are in net-next:
2297f88ce0b8 dt-bindings: net: amlogic,meson-dwmac: list the compatible combinations
18029931478b dt-bindings: net: add amlogic,t7-dwmac
c4c851a3325e net: stmmac: dwmac-meson8b: add support for the Amlogic T7
4fdc7da7e13d net: stmmac: dwmac-meson8b: follow the generic RGMII delay rules on the T7
This version carries the rest: a new binding patch for the MDIO
multiplexer, and the three device tree patches. Patch 1 is for netdev,
patches 2 to 4 for the Amlogic tree; the device tree patches build
without the driver part.
The T7 has the same ethernet glue as the G12A. The glue registers sit
in a block gated by sys_ethphy, so the controller node lists that
clock as "pclk", and the pipeline stage between the controller and the
DRAM has a gate of its own, listed as well. The MDIO multiplexer gets a
T7 compatible with the G12A one as fallback. The VIM4 is plain
rgmii-id, with the PHY at the address its pins strap.
Tested on v7.4-rc2 on the Khadas VIM4 with the PHY at address 1:
941/944Mbit/s with iperf3, no error. A scan of the bus shows the
PHY at 0 and at 1 and nothing else.
Overlaps with Xianwei Zhao's "[PATCH v2 0/3] t7: add some nodes for
Amlogc t7."; I will drop or rebase whatever overlaps once we agree.
Changes since v7:
- patches 1 to 4 of v7 are in net-next and are not resent; the
numbers below are the new ones
- new patch 1: the MDIO multiplexer binding gets "amlogic,t7-mdio-mux"
with the G12A compatible as fallback, and patch 3 uses it in the
multiplexer node (Martin)
- patch 4: the VIM4 PHY node uses address 1, where the PHY is strapped;
address 0 only worked through the RTL8211F broadcast address (Martin)
- patch 2: Martin's Reviewed-by
Changes since v6:
- no Assisted-by trailer (Krzysztof)
- patch 2: the T7 gets a binding file of its own (Rob, Sashiko); the
pclk description names the glue registers (Krzysztof);
rx-internal-delay-ps as constraints, default 0, maximum 3000,
multipleOf 200 (Rob); tx-internal-delay-ps stays false, the
meta-schema refuses "not" on a top-level property, so the reason is
a comment above it (Rob); "snps,dwmac" dropped from the compatible
list (Rob)
- patch 6: "snps,dwmac" dropped from the node (Rob)
- patch 4: Andrew's Reviewed-by
Changes since v5:
- patch 2: the T7 block follows the Meson8b..G12A block instead of
splitting it in two; default: 0 for rx-internal-delay-ps on the T7
(Sashiko). The explicit false lines stay, as asked in review.
Changes since v4:
- patch 4: the old delay code is left untouched; the T7 only gets a
callback of its own, and the probe parsing it shares validates the
RX value (Andrew); no NULL check on the callback (Martin)
- 0 accepted for rx-internal-delay-ps, the T7 joins the G12A list
(Andrew)
- comment on the deprecated vendor delay properties (Andrew)
- the MDIO mux and the on-chip PHY bus start disabled in the dtsi; the
VIM4 enables the mux (Sashiko)
Changes since v3:
- phy-mode untouched, the PHY adds the 2ns, rx-internal-delay-ps is
tuning only, tx-internal-delay-ps not valid for the T7 (Andrew)
- 0 dropped from rx-internal-delay-ps (Martin); vendor delay
properties explicitly false for the T7 (Andrew)
- init_delays callback; patch 4 renamed (Martin, Andrew)
- VIM4 without tx-internal-delay-ps, the PHY TX delay works (Martin)
- Meson8m2 moved to the dwmac-3.70a list; its boards warn in
dtbs_check until Martin's dtsi patch lands (Martin, Conor)
- the delay table in ethernet-controller.yaml dropped
- Martin's Reviewed-by on patch 3
Changes since v2:
- T7-only delay path (Andrew), PHY mode from
phy_fix_phy_mode_for_mac_delays() (Maxime)
- tx-internal-delay-ps checked against the hardware values (Sashiko)
- vendor delay properties not valid for the T7 (Neil, Andrew)
- no max-speed on either PHY (Andrew); VIM4 PHY interrupt on GPIOH_6
(Sashiko)
- the T7 node owns sys_ethphy as "pclk"
- no longer RFC; Conor's Ack dropped from patch 2 and Neil's R-b from
patch 6
Changes since v1:
- the driver honours the generic delay properties; VIM4 moved to
rgmii-id with tx-internal-delay-ps = <2000>, Neil's R-b dropped
Lucas Tanure (4):
dt-bindings: net: amlogic,g12a-mdio-mux: add the T7 compatible
arm64: dts: amlogic: t7: add the ethernet pinctrl nodes
arm64: dts: amlogic: t7: add the ethernet controller
arm64: dts: amlogic: t7: khadas-vim4: enable the ethernet port
.../bindings/net/amlogic,g12a-mdio-mux.yaml | 6 +-
.../amlogic/amlogic-t7-a311d2-khadas-vim4.dts | 23 +++++
arch/arm64/boot/dts/amlogic/amlogic-t7.dtsi | 88 +++++++++++++++++++
3 files changed, 116 insertions(+), 1 deletion(-)
base-commit: 7e1392a42ea288d70a266f6553fe06f57f8647ea
--
2.56.1