Re: [PATCH v2 4/5] soc: qcom: Add LLCC support for multi channel DDR

From: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
To: Komal Bajaj <quic_kbajaj@quicinc.com>
Cc: Rob Herring <robh+dt@kernel.org>,
	Bjorn Andersson <andersson@kernel.org>,
	Konrad Dybcio <konrad.dybcio@linaro.org>,
	Abel Vesa <abel.vesa@linaro.org>,
	Rishabh Bhatnagar <rishabhb@codeaurora.org>,
	Prakash Ranjan <saiprakash.ranjan@codeaurora.org>,
	Krzysztof Kozlowski <krzysztof.kozlowski+dt@linaro.org>,
	Andy Gross <agross@kernel.org>,
	devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-arm-msm@vger.kernel.org
Subject: Re: [PATCH v2 4/5] soc: qcom: Add LLCC support for multi channel DDR
Date: Wed, 15 Mar 2023 19:38:07 +0530	[thread overview]
Message-ID: <20230315140807.GB98488@thinkpad> (raw)
In-Reply-To: <20230313124040.9463-5-quic_kbajaj@quicinc.com>

On Mon, Mar 13, 2023 at 06:10:39PM +0530, Komal Bajaj wrote:
> Add LLCC support for multi channel DDR configurations
> based off of a feature register.
> 

Please elaborate more in the commit message on why this patch is needed and how
it is implemented.

> Signed-off-by: Komal Bajaj <quic_kbajaj@quicinc.com>
> ---
>  drivers/soc/qcom/llcc-qcom.c       | 56 ++++++++++++++++++++++++++++--
>  include/linux/soc/qcom/llcc-qcom.h |  2 ++
>  2 files changed, 55 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/soc/qcom/llcc-qcom.c b/drivers/soc/qcom/llcc-qcom.c
> index 00699a0c047e..f4d3e266c629 100644
> --- a/drivers/soc/qcom/llcc-qcom.c
> +++ b/drivers/soc/qcom/llcc-qcom.c
> @@ -17,6 +17,7 @@
>  #include <linux/regmap.h>
>  #include <linux/sizes.h>
>  #include <linux/slab.h>
> +#include <linux/firmware/qcom/qcom_scm.h>

Sort the includes alphabetically

>  #include <linux/soc/qcom/llcc-qcom.h>
>  
>  #define ACTIVATE                      BIT(0)
> @@ -924,6 +925,40 @@ static int qcom_llcc_cfg_program(struct platform_device *pdev,
>  	return ret;
>  }
>  
> +static int qcom_llcc_get_cfg_index(struct platform_device *pdev, u32 *cfg_index)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct resource *ch_res = NULL;

No need to initialize the pointer

> +

No need of a newline

> +	u32 ch_reg_sz;
> +	u32 ch_reg_off;
> +	u32 val;
> +	int ret = 0;
> +
> +	ch_res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "multi_channel_register");
> +	if (ch_res) {
> +		if (of_property_read_u32(dev->of_node, "multi-ch-bit-off", &ch_reg_off)) {
> +			dev_err(&pdev->dev,
> +				"Couldn't get offset for multi channel feature register\n");
> +			return -ENODEV;
> +		}
> +		if (of_property_read_u32_index(dev->of_node, "multi-ch-bit-off", 1, &ch_reg_sz)) {
> +			dev_err(&pdev->dev,
> +				"Couldn't get size of multi channel feature register\n");
> +			return -ENODEV;
> +		}
> +
> +		if (qcom_scm_io_readl(ch_res->start, &val)) {

You didn't mention this SCM call in the commit message. Also, for SCM, you need
to select the driver in Kconfig. 

> +			dev_err(&pdev->dev, "Couldn't access multi channel feature register\n");
> +			ret = -EINVAL;

Catch the actual error no from qcom_scm_io_readl().

So in the case of failure, you still want to calculate cfg_index?

> +		}
> +		*cfg_index = (val >> ch_reg_off) & ((1 << ch_reg_sz) - 1);
> +	} else

Use braces for else condition

> +		*cfg_index = 0;
> +
> +	return ret;
> +}
> +
>  static int qcom_llcc_remove(struct platform_device *pdev)
>  {
>  	/* Set the global pointer to a error code to avoid referencing it */
> @@ -956,10 +991,13 @@ static int qcom_llcc_probe(struct platform_device *pdev)
>  	struct device *dev = &pdev->dev;
>  	int ret, i;
>  	struct platform_device *llcc_edac;
> -	const struct qcom_llcc_config *cfg;
> +	const struct qcom_llcc_config *cfg, *entry;
>  	const struct llcc_slice_config *llcc_cfg;
> +

No need of newline

>  	u32 sz;
> +	u32 cfg_index;
>  	u32 version;
> +	u32 no_of_entries = 0;

num_entries?

>  
>  	drv_data = devm_kzalloc(dev, sizeof(*drv_data), GFP_KERNEL);
>  	if (!drv_data) {
> @@ -999,8 +1037,20 @@ static int qcom_llcc_probe(struct platform_device *pdev)
>  	num_banks >>= LLCC_LB_CNT_SHIFT;
>  	drv_data->num_banks = num_banks;
>  
> -	llcc_cfg = cfg[0].sct_data;
> -	sz = cfg[0].size;
> +	ret = qcom_llcc_get_cfg_index(pdev, &cfg_index);
> +	if (ret)
> +		goto err;
> +
> +	for (entry = cfg; entry->sct_data; entry++, no_of_entries++)
> +		;

Wrap ; in the above line itself

> +	if (cfg_index >= no_of_entries) {
> +		ret = -EINVAL;
> +		goto err;
> +	}
> +
> +	drv_data->cfg_index = cfg_index;

Where is this cached value used?

Thanks,
Mani

> +	llcc_cfg = cfg[cfg_index].sct_data;
> +	sz = cfg[cfg_index].size;
>  
>  	for (i = 0; i < sz; i++)
>  		if (llcc_cfg[i].slice_id > drv_data->max_slices)
> diff --git a/include/linux/soc/qcom/llcc-qcom.h b/include/linux/soc/qcom/llcc-qcom.h
> index ad1fd718169d..225891a02f5d 100644
> --- a/include/linux/soc/qcom/llcc-qcom.h
> +++ b/include/linux/soc/qcom/llcc-qcom.h
> @@ -125,6 +125,7 @@ struct llcc_edac_reg_offset {
>   * @cfg: pointer to the data structure for slice configuration
>   * @edac_reg_offset: Offset of the LLCC EDAC registers
>   * @lock: mutex associated with each slice
> + * @cfg_index: index of config table if multiple configs present for a target
>   * @cfg_size: size of the config data table
>   * @max_slices: max slices as read from device tree
>   * @num_banks: Number of llcc banks
> @@ -139,6 +140,7 @@ struct llcc_drv_data {
>  	const struct llcc_slice_config *cfg;
>  	const struct llcc_edac_reg_offset *edac_reg_offset;
>  	struct mutex lock;
> +	u32 cfg_index;
>  	u32 cfg_size;
>  	u32 max_slices;
>  	u32 num_banks;
> -- 
> 2.39.1
> 

-- 
மணிவண்ணன் சதாசிவம்