[PATCH 1/8] net: phy: Introduce Qualcomm IPQ5018 internal PHY driver

Andrew Lunn andrew at lunn.ch
Sun Jan 21 08:19:14 PST 2024


On Sun, Jan 21, 2024 at 08:42:30PM +0800, Ziyang Huang wrote:
> Signed-off-by: Ziyang Huang <hzyitc at outlook.com>

You need to say something in the commit message. One obvious thing is
to justify not using the at803x driver, since 

> +#define IPQ5018_PHY_ID			0x004dd0c0

is in the Atheros OUI range.

> +static int ipq5018_probe(struct phy_device *phydev)
> +{
> +	struct ipq5018_phy *priv;
> +	struct device *dev = &phydev->mdio.dev;
> +	char name[64];
> +	int ret;

I guess you are new to mainline network. Please read:

https://www.kernel.org/doc/html/latest/process/maintainer-netdev.html

Section 1.6.4.

> +
> +	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> +	if (!priv)
> +		return dev_err_probe(dev, -ENOMEM,
> +				     "failed to allocate priv\n");

Please read the documentation of dev_err_probe() and this fix the
obvious problem with this.

> +	snprintf(name, sizeof(name), "%s#rx", dev_name(dev));
> +	priv->clk_rx = clk_hw_register_fixed_rate(dev, name, NULL, 0,
> +						  TX_RX_CLK_RATE);
> +	if (IS_ERR_OR_NULL(priv->clk_rx))
> +		return dev_err_probe(dev, PTR_ERR(priv->clk_rx),
> +				     "failed to register rx clock\n");
> +
> +	snprintf(name, sizeof(name), "%s#tx", dev_name(dev));
> +	priv->clk_tx = clk_hw_register_fixed_rate(dev, name, NULL, 0,
> +						  TX_RX_CLK_RATE);
> +	if (IS_ERR_OR_NULL(priv->clk_tx))
> +		return dev_err_probe(dev, PTR_ERR(priv->clk_tx),
> +				     "failed to register tx clock\n");
> +
> +	priv->clk_data = devm_kzalloc(dev,
> +				      struct_size(priv->clk_data, hws, 2),
> +				      GFP_KERNEL);
> +	if (!priv->clk_data)
> +		return dev_err_probe(dev, -ENOMEM,
> +				     "failed to allocate clk_data\n");
> +
> +	priv->clk_data->num = 2;
> +	priv->clk_data->hws[0] = priv->clk_rx;
> +	priv->clk_data->hws[1] = priv->clk_tx;
> +	ret = of_clk_add_hw_provider(dev->of_node, of_clk_hw_onecell_get,
> +				     priv->clk_data);
> +	if (ret)
> +		return dev_err_probe(dev, ret,
> +				     "fail to register clock provider\n");

This needs an explanation. Why register two fixed clocks, which you
never use? Why not put these two clocks in DT?

> +
> +	return 0;
> +}
> +
> +static int ipq5018_soft_reset(struct phy_device *phydev)
> +{
> +	int ret;
> +
> +	ret = phy_modify(phydev, IPQ5018_PHY_FIFO_CONTROL,
> +			 IPQ5018_PHY_FIFO_RESET, 0);
> +	if (ret < 0)
> +		return ret;
> +
> +	msleep(50);
> +
> +	ret = phy_modify(phydev, IPQ5018_PHY_FIFO_CONTROL,
> +			 IPQ5018_PHY_FIFO_RESET, IPQ5018_PHY_FIFO_RESET);
> +	if (ret < 0)
> +		return ret;
> +
> +	return 0;
> +}

This needs an explanation. It is also somewhat like
qca808x_link_change_notify(). Is it really sufficient to only do this
reset FIFO during a soft reset, or is it needed when ever the link
changes?

You also appear to be missing device tree bindings.

    Andrew

---
pw-bot: cr




More information about the Linux-mediatek mailing list