Re: [PATCH net-next v3 2/2] net: stmmac: dwmac-rk: Add RK3572 support
From: netdev-bot+sashiko
Date: Wed Nov 18 2026 - 05:41:50 EST
Thank you for your contribution! Sashiko AI review found 1 potential
issue(s) to consider.
Critical: 0 · High: 0 · Medium: 1 · Low: 0
If you believe none of the issues are legitimate, please reply to each
issue of Medium severity or higher and explain why the AI is wrong.
If any of them are legitimate and you plan to respin, please let patchwork
know by including "pw-bot: cr" as a separate line at the end of your reply
(one such reply per series is enough).
- [Medium] RK3572 RGMII delay lines get fixed values that the DT is not
allowed to describe.
--
net: stmmac: dwmac-rk: Add RK3572 support
This patch adds RK3572 support to the Rockchip dwmac glue driver. It
covers the register layout for the two GMAC controllers in the PHP and
NVM0 GRFs, plus a set_to_rgmii callback that programs the RGMII delay
lines through the IOC.
> diff --git a/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c b/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
> index a37df5767d84e..c9d8d8bf51f7b 100644
> --- a/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
> +++ b/drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c
[ ... ]
> +static void rk3572_set_to_rgmii(struct rk_priv_data *bsp_priv,
> + int tx_delay, int rx_delay)
> +{
> + u32 val = GRF_FIELD(14, 8, rx_delay) | GRF_FIELD(6, 0, tx_delay) |
> + GRF_FIELD(15, 15, !!rx_delay) | GRF_FIELD(7, 7, !!tx_delay);
> +
> + if (bsp_priv->id == 1) {
> + regmap_write(bsp_priv->php_grf, RK3572_VCCIO0_3_VO_IOC_CON1, val);
> + regmap_write(bsp_priv->php_grf, RK3572_VCCIO5_6_IOC_MISC0, val);
> + } else {
> + regmap_write(bsp_priv->php_grf, RK3572_VCCIO1_2_4_VI_IOC_CON1, val);
> + }
> +}
[Severity: Medium]
Where are the tx_delay and rx_delay values for RK3572 meant to come from?
The companion binding change in rockchip-dwmac.yaml forbids the legacy
properties for this compatible:
- if:
properties:
compatible:
contains:
const: rockchip,rk3572-gmac
then:
properties:
rx_delay: false
tx_delay: false
That means a schema-valid RK3572 node never has them. rk_gmac_setup() then
falls back to the old defaults and logs an error on every probe:
drivers/net/ethernet/stmicro/stmmac/dwmac-rk.c:rk_gmac_setup() {
...
ret = of_property_read_u32(dev->of_node, "tx_delay", &value);
if (ret) {
bsp_priv->tx_delay = 0x30;
dev_err(dev, "Can not read property: tx_delay.");
...
ret = of_property_read_u32(dev->of_node, "rx_delay", &value);
if (ret) {
bsp_priv->rx_delay = 0x10;
dev_err(dev, "Can not read property: rx_delay.");
...
}
rk_gmac_powerup() passes these defaults on for "rgmii" (both delays),
"rgmii-rxid" (tx_delay) and "rgmii-txid" (rx_delay). Since bits 15 and 7
are set from !!rx_delay and !!tx_delay, rk3572_set_to_rgmii() switches on
the IOC delay lines with 0x30/0x10 taps for every RGMII mode except
"rgmii-id".
The 0x30/0x10 defaults come from older Rockchip SoCs. Do they match the
RK3572 tap size?
A board can't override them without failing schema validation. The driver
also doesn't read rx-internal-delay-ps or tx-internal-delay-ps.
Consider a board that uses "rgmii" with PCB delays, or a PHY that adds its
own delay. Could this leave the RX/TX sampling skewed, with no supported
way to fix it in DT?
Would it make sense to either zero the MAC-side delays for RK3572 when the
legacy properties are absent, or take them from the standard
*-internal-delay-ps properties? If not, should the binding keep allowing
tx_delay/rx_delay for now?
> +
> +static const struct rk_gmac_ops rk3572_ops = {
> + .init = rk3572_init,
> + .set_to_rgmii = rk3572_set_to_rgmii,
[ ... ]
--
Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261116-add-board-ok3572-c-net-v3-0-a957fdcd9c7a%40qq.com