Re: [PATCH v6] phy: qcom-qmp-ufs: provide symbol clocks

[Johan Hovold <johan@kernel.org>]

On Fri, Oct 28, 2022 at 07:48:47PM +0300, Dmitry Baryshkov wrote:
> Register three UFS symbol clocks (ufs_rx_symbol_0_clk_src,
> ufs_rx_symbol_1_clk_src ufs_tx_symbol_0_clk_src). Register OF clock
> provider to let other devices link these clocks through the DT.
> 
> Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
> ---
> 
> Changes since v5:
> - Rebased on top of phy/next
> 
> Changes since v4:
> - Rebased, dropping merged clk patches
> - Fixed whitespace errors
> - Added linebreaks to fit into 100 chars limit
> 
> ---
>  drivers/phy/qualcomm/phy-qcom-qmp-ufs.c | 67 +++++++++++++++++++++++++
>  1 file changed, 67 insertions(+)
> 
> diff --git a/drivers/phy/qualcomm/phy-qcom-qmp-ufs.c b/drivers/phy/qualcomm/phy-qcom-qmp-ufs.c
> index 189103d1bd18..c1c9c6e7949d 100644
> --- a/drivers/phy/qualcomm/phy-qcom-qmp-ufs.c
> +++ b/drivers/phy/qualcomm/phy-qcom-qmp-ufs.c
> @@ -1023,6 +1023,66 @@ static int qmp_ufs_clk_init(struct qmp_ufs *qmp)
>  	return devm_clk_bulk_get(dev, num, qmp->clks);
>  }
>  
> +static void phy_clk_release_provider(void *res)
> +{
> +	of_clk_del_provider(res);
> +}
> +
> +#define UFS_SYMBOL_CLOCKS 3
> +
> +static int phy_symbols_clk_register(struct qmp_ufs *qmp, struct device_node *np)
> +{
> +	struct clk_hw_onecell_data *clk_data;
> +	struct clk_hw *hw;
> +	char name[64];
> +	int ret;
> +
> +	clk_data = devm_kzalloc(qmp->dev,
> +				struct_size(clk_data, hws, UFS_SYMBOL_CLOCKS),
> +				GFP_KERNEL);
> +	if (!clk_data)
> +		return -ENOMEM;
> +
> +	clk_data->num = UFS_SYMBOL_CLOCKS;
> +
> +	snprintf(name, sizeof(name), "%s::rx_symbol_0", dev_name(qmp->dev));
> +	hw = devm_clk_hw_register_fixed_rate(qmp->dev, name, NULL, 0, 0);
> +	if (IS_ERR(hw))
> +		return PTR_ERR(hw);
> +
> +	clk_data->hws[0] = hw;
> +
> +	snprintf(name, sizeof(name), "%s::rx_symbol_1", dev_name(qmp->dev));
> +	hw = devm_clk_hw_register_fixed_rate(qmp->dev, name, NULL, 0, 0);
> +	if (IS_ERR(hw))
> +		return PTR_ERR(hw);
> +
> +	clk_data->hws[1] = hw;
> +
> +	snprintf(name, sizeof(name), "%s::tx_symbol_0", dev_name(qmp->dev));
> +	hw = devm_clk_hw_register_fixed_rate(qmp->dev, name, NULL, 0, 0);
> +	if (IS_ERR(hw))
> +		return PTR_ERR(hw);
> +
> +	clk_data->hws[2] = hw;
> +
> +	ret = of_clk_add_hw_provider(np, of_clk_hw_onecell_get, clk_data);
> +	if (ret)
> +		return ret;
> +
> +	/*
> +	 * Roll a devm action because the clock provider is the child node, but
> +	 * the child node is not actually a device.
> +	 */
> +	return devm_add_action_or_reset(qmp->dev, phy_clk_release_provider, np);
> +}
> +
> +static const struct phy_ops qcom_qmp_ufs_ops = {
> +	.power_on	= qmp_ufs_enable,
> +	.power_off	= qmp_ufs_disable,
> +	.owner		= THIS_MODULE,
> +};

This phy_ops addition looks an artefact of the rebase and should not be
here.

>  static int qmp_ufs_parse_dt_legacy(struct qmp_ufs *qmp, struct device_node *np)
>  {
>  	struct platform_device *pdev = to_platform_device(qmp->dev);
> @@ -1135,6 +1195,13 @@ static int qmp_ufs_probe(struct platform_device *pdev)
>  	if (ret)
>  		goto err_node_put;
>  
> +	ret = phy_symbols_clk_register(qmp, np);

Looks like this should go in probe() instead, or was there a reason for
not registering these clocks when using the new bindings?

And don't they need to be described in both the old and new bindings
first either way?

> +	if (ret) {
> +		dev_err(dev, "failed to create symbol clocks, %d\n",
> +			ret);

Please use the "...: %d\n" form for consistency.

But you can probably just drop this error message instead.

> +		goto err_node_put;
> +	}
> +
>  	qmp->phy = devm_phy_create(dev, np, &qcom_qmp_ufs_phy_ops);
>  	if (IS_ERR(qmp->phy)) {
>  		ret = PTR_ERR(qmp->phy);

Johan

-- 
linux-phy mailing list
linux-phy@lists.infradead.org
https://lists.infradead.org/mailman/listinfo/linux-phy